ip_fw2.c revision 171173
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 171173 2007-07-03 12:46:08Z mlaier $ 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) do { \ 151 rw_assert(&(_chain)->rwmtx, RA_WLOCKED); \ 152 NET_ASSERT_GIANT(); \ 153} while (0) 154 155#define IPFW_RLOCK(p) rw_rlock(&(p)->rwmtx) 156#define IPFW_RUNLOCK(p) rw_runlock(&(p)->rwmtx) 157#define IPFW_WLOCK(p) rw_wlock(&(p)->rwmtx) 158#define IPFW_WUNLOCK(p) rw_wunlock(&(p)->rwmtx) 159 160/* 161 * list of rules for layer 3 162 */ 163static struct ip_fw_chain layer3_chain; 164 165MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 166MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 167 168struct table_entry { 169 struct radix_node rn[2]; 170 struct sockaddr_in addr, mask; 171 u_int32_t value; 172}; 173 174static int fw_debug = 1; 175static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 176 177extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS); 178 179#ifdef SYSCTL_NODE 180SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 181SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable, 182 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &fw_enable, 0, 183 ipfw_chg_hook, "I", "Enable ipfw"); 184SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 185 &autoinc_step, 0, "Rule number autincrement step"); 186SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 187 CTLFLAG_RW | CTLFLAG_SECURE3, 188 &fw_one_pass, 0, 189 "Only do a single pass through ipfw when using dummynet(4)"); 190SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 191 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 192SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 193 CTLFLAG_RW | CTLFLAG_SECURE3, 194 &fw_verbose, 0, "Log matches to ipfw rules"); 195SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 196 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 197 198/* 199 * Description of dynamic rules. 200 * 201 * Dynamic rules are stored in lists accessed through a hash table 202 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 203 * be modified through the sysctl variable dyn_buckets which is 204 * updated when the table becomes empty. 205 * 206 * XXX currently there is only one list, ipfw_dyn. 207 * 208 * When a packet is received, its address fields are first masked 209 * with the mask defined for the rule, then hashed, then matched 210 * against the entries in the corresponding list. 211 * Dynamic rules can be used for different purposes: 212 * + stateful rules; 213 * + enforcing limits on the number of sessions; 214 * + in-kernel NAT (not implemented yet) 215 * 216 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 217 * measured in seconds and depending on the flags. 218 * 219 * The total number of dynamic rules is stored in dyn_count. 220 * The max number of dynamic rules is dyn_max. When we reach 221 * the maximum number of rules we do not create anymore. This is 222 * done to avoid consuming too much memory, but also too much 223 * time when searching on each packet (ideally, we should try instead 224 * to put a limit on the length of the list on each bucket...). 225 * 226 * Each dynamic rule holds a pointer to the parent ipfw rule so 227 * we know what action to perform. Dynamic rules are removed when 228 * the parent rule is deleted. XXX we should make them survive. 229 * 230 * There are some limitations with dynamic rules -- we do not 231 * obey the 'randomized match', and we do not do multiple 232 * passes through the firewall. XXX check the latter!!! 233 */ 234static ipfw_dyn_rule **ipfw_dyn_v = NULL; 235static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 236static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 237 238static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 239#define IPFW_DYN_LOCK_INIT() \ 240 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 241#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 242#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 243#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 244#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 245 246/* 247 * Timeouts for various events in handing dynamic rules. 248 */ 249static u_int32_t dyn_ack_lifetime = 300; 250static u_int32_t dyn_syn_lifetime = 20; 251static u_int32_t dyn_fin_lifetime = 1; 252static u_int32_t dyn_rst_lifetime = 1; 253static u_int32_t dyn_udp_lifetime = 10; 254static u_int32_t dyn_short_lifetime = 5; 255 256/* 257 * Keepalives are sent if dyn_keepalive is set. They are sent every 258 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 259 * seconds of lifetime of a rule. 260 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 261 * than dyn_keepalive_period. 262 */ 263 264static u_int32_t dyn_keepalive_interval = 20; 265static u_int32_t dyn_keepalive_period = 5; 266static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 267 268static u_int32_t static_count; /* # of static rules */ 269static u_int32_t static_len; /* size in bytes of static rules */ 270static u_int32_t dyn_count; /* # of dynamic rules */ 271static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 272 273SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 274 &dyn_buckets, 0, "Number of dyn. buckets"); 275SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 276 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 277SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 278 &dyn_count, 0, "Number of dyn. rules"); 279SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 280 &dyn_max, 0, "Max number of dyn. rules"); 281SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 282 &static_count, 0, "Number of static rules"); 283SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 284 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 285SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 286 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 287SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 288 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 289SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 290 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 291SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 292 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 293SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 294 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 295SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 296 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 297 298#ifdef INET6 299/* 300 * IPv6 specific variables 301 */ 302SYSCTL_DECL(_net_inet6_ip6); 303 304static struct sysctl_ctx_list ip6_fw_sysctl_ctx; 305static struct sysctl_oid *ip6_fw_sysctl_tree; 306#endif /* INET6 */ 307#endif /* SYSCTL_NODE */ 308 309#ifdef IPFIREWALL_NAT 310MODULE_DEPEND(ipfw, libalias, 1, 1, 1); 311#endif 312static int fw_deny_unknown_exthdrs = 1; 313 314 315/* 316 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 317 * Other macros just cast void * into the appropriate type 318 */ 319#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 320#define TCP(p) ((struct tcphdr *)(p)) 321#define SCTP(p) ((struct sctphdr *)(p)) 322#define UDP(p) ((struct udphdr *)(p)) 323#define ICMP(p) ((struct icmphdr *)(p)) 324#define ICMP6(p) ((struct icmp6_hdr *)(p)) 325 326static __inline int 327icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 328{ 329 int type = icmp->icmp_type; 330 331 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 332} 333 334#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 335 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 336 337static int 338is_icmp_query(struct icmphdr *icmp) 339{ 340 int type = icmp->icmp_type; 341 342 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 343} 344#undef TT 345 346/* 347 * The following checks use two arrays of 8 or 16 bits to store the 348 * bits that we want set or clear, respectively. They are in the 349 * low and high half of cmd->arg1 or cmd->d[0]. 350 * 351 * We scan options and store the bits we find set. We succeed if 352 * 353 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 354 * 355 * The code is sometimes optimized not to store additional variables. 356 */ 357 358static int 359flags_match(ipfw_insn *cmd, u_int8_t bits) 360{ 361 u_char want_clear; 362 bits = ~bits; 363 364 if ( ((cmd->arg1 & 0xff) & bits) != 0) 365 return 0; /* some bits we want set were clear */ 366 want_clear = (cmd->arg1 >> 8) & 0xff; 367 if ( (want_clear & bits) != want_clear) 368 return 0; /* some bits we want clear were set */ 369 return 1; 370} 371 372static int 373ipopts_match(struct ip *ip, ipfw_insn *cmd) 374{ 375 int optlen, bits = 0; 376 u_char *cp = (u_char *)(ip + 1); 377 int x = (ip->ip_hl << 2) - sizeof (struct ip); 378 379 for (; x > 0; x -= optlen, cp += optlen) { 380 int opt = cp[IPOPT_OPTVAL]; 381 382 if (opt == IPOPT_EOL) 383 break; 384 if (opt == IPOPT_NOP) 385 optlen = 1; 386 else { 387 optlen = cp[IPOPT_OLEN]; 388 if (optlen <= 0 || optlen > x) 389 return 0; /* invalid or truncated */ 390 } 391 switch (opt) { 392 393 default: 394 break; 395 396 case IPOPT_LSRR: 397 bits |= IP_FW_IPOPT_LSRR; 398 break; 399 400 case IPOPT_SSRR: 401 bits |= IP_FW_IPOPT_SSRR; 402 break; 403 404 case IPOPT_RR: 405 bits |= IP_FW_IPOPT_RR; 406 break; 407 408 case IPOPT_TS: 409 bits |= IP_FW_IPOPT_TS; 410 break; 411 } 412 } 413 return (flags_match(cmd, bits)); 414} 415 416static int 417tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 418{ 419 int optlen, bits = 0; 420 u_char *cp = (u_char *)(tcp + 1); 421 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 422 423 for (; x > 0; x -= optlen, cp += optlen) { 424 int opt = cp[0]; 425 if (opt == TCPOPT_EOL) 426 break; 427 if (opt == TCPOPT_NOP) 428 optlen = 1; 429 else { 430 optlen = cp[1]; 431 if (optlen <= 0) 432 break; 433 } 434 435 switch (opt) { 436 437 default: 438 break; 439 440 case TCPOPT_MAXSEG: 441 bits |= IP_FW_TCPOPT_MSS; 442 break; 443 444 case TCPOPT_WINDOW: 445 bits |= IP_FW_TCPOPT_WINDOW; 446 break; 447 448 case TCPOPT_SACK_PERMITTED: 449 case TCPOPT_SACK: 450 bits |= IP_FW_TCPOPT_SACK; 451 break; 452 453 case TCPOPT_TIMESTAMP: 454 bits |= IP_FW_TCPOPT_TS; 455 break; 456 457 } 458 } 459 return (flags_match(cmd, bits)); 460} 461 462static int 463iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 464{ 465 if (ifp == NULL) /* no iface with this packet, match fails */ 466 return 0; 467 /* Check by name or by IP address */ 468 if (cmd->name[0] != '\0') { /* match by name */ 469 /* Check name */ 470 if (cmd->p.glob) { 471 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 472 return(1); 473 } else { 474 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 475 return(1); 476 } 477 } else { 478 struct ifaddr *ia; 479 480 /* XXX lock? */ 481 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 482 if (ia->ifa_addr->sa_family != AF_INET) 483 continue; 484 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 485 (ia->ifa_addr))->sin_addr.s_addr) 486 return(1); /* match */ 487 } 488 } 489 return(0); /* no match, fail ... */ 490} 491 492/* 493 * The verify_path function checks if a route to the src exists and 494 * if it is reachable via ifp (when provided). 495 * 496 * The 'verrevpath' option checks that the interface that an IP packet 497 * arrives on is the same interface that traffic destined for the 498 * packet's source address would be routed out of. The 'versrcreach' 499 * option just checks that the source address is reachable via any route 500 * (except default) in the routing table. These two are a measure to block 501 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 502 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 503 * is purposely reminiscent of the Cisco IOS command, 504 * 505 * ip verify unicast reverse-path 506 * ip verify unicast source reachable-via any 507 * 508 * which implements the same functionality. But note that syntax is 509 * misleading. The check may be performed on all IP packets whether unicast, 510 * multicast, or broadcast. 511 */ 512static int 513verify_path(struct in_addr src, struct ifnet *ifp) 514{ 515 struct route ro; 516 struct sockaddr_in *dst; 517 518 bzero(&ro, sizeof(ro)); 519 520 dst = (struct sockaddr_in *)&(ro.ro_dst); 521 dst->sin_family = AF_INET; 522 dst->sin_len = sizeof(*dst); 523 dst->sin_addr = src; 524 rtalloc_ign(&ro, RTF_CLONING); 525 526 if (ro.ro_rt == NULL) 527 return 0; 528 529 /* 530 * If ifp is provided, check for equality with rtentry. 531 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 532 * in order to pass packets injected back by if_simloop(): 533 * if useloopback == 1 routing entry (via lo0) for our own address 534 * may exist, so we need to handle routing assymetry. 535 */ 536 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 537 RTFREE(ro.ro_rt); 538 return 0; 539 } 540 541 /* if no ifp provided, check if rtentry is not default route */ 542 if (ifp == NULL && 543 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 544 RTFREE(ro.ro_rt); 545 return 0; 546 } 547 548 /* or if this is a blackhole/reject route */ 549 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 550 RTFREE(ro.ro_rt); 551 return 0; 552 } 553 554 /* found valid route */ 555 RTFREE(ro.ro_rt); 556 return 1; 557} 558 559#ifdef INET6 560/* 561 * ipv6 specific rules here... 562 */ 563static __inline int 564icmp6type_match (int type, ipfw_insn_u32 *cmd) 565{ 566 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 567} 568 569static int 570flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 571{ 572 int i; 573 for (i=0; i <= cmd->o.arg1; ++i ) 574 if (curr_flow == cmd->d[i] ) 575 return 1; 576 return 0; 577} 578 579/* support for IP6_*_ME opcodes */ 580static int 581search_ip6_addr_net (struct in6_addr * ip6_addr) 582{ 583 struct ifnet *mdc; 584 struct ifaddr *mdc2; 585 struct in6_ifaddr *fdm; 586 struct in6_addr copia; 587 588 TAILQ_FOREACH(mdc, &ifnet, if_link) 589 TAILQ_FOREACH(mdc2, &mdc->if_addrlist, ifa_list) { 590 if (mdc2->ifa_addr->sa_family == AF_INET6) { 591 fdm = (struct in6_ifaddr *)mdc2; 592 copia = fdm->ia_addr.sin6_addr; 593 /* need for leaving scope_id in the sock_addr */ 594 in6_clearscope(&copia); 595 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) 596 return 1; 597 } 598 } 599 return 0; 600} 601 602static int 603verify_path6(struct in6_addr *src, struct ifnet *ifp) 604{ 605 struct route_in6 ro; 606 struct sockaddr_in6 *dst; 607 608 bzero(&ro, sizeof(ro)); 609 610 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 611 dst->sin6_family = AF_INET6; 612 dst->sin6_len = sizeof(*dst); 613 dst->sin6_addr = *src; 614 rtalloc_ign((struct route *)&ro, RTF_CLONING); 615 616 if (ro.ro_rt == NULL) 617 return 0; 618 619 /* 620 * if ifp is provided, check for equality with rtentry 621 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 622 * to support the case of sending packets to an address of our own. 623 * (where the former interface is the first argument of if_simloop() 624 * (=ifp), the latter is lo0) 625 */ 626 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 627 RTFREE(ro.ro_rt); 628 return 0; 629 } 630 631 /* if no ifp provided, check if rtentry is not default route */ 632 if (ifp == NULL && 633 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 634 RTFREE(ro.ro_rt); 635 return 0; 636 } 637 638 /* or if this is a blackhole/reject route */ 639 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 640 RTFREE(ro.ro_rt); 641 return 0; 642 } 643 644 /* found valid route */ 645 RTFREE(ro.ro_rt); 646 return 1; 647 648} 649static __inline int 650hash_packet6(struct ipfw_flow_id *id) 651{ 652 u_int32_t i; 653 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 654 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 655 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 656 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 657 (id->dst_port) ^ (id->src_port); 658 return i; 659} 660 661static int 662is_icmp6_query(int icmp6_type) 663{ 664 if ((icmp6_type <= ICMP6_MAXTYPE) && 665 (icmp6_type == ICMP6_ECHO_REQUEST || 666 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 667 icmp6_type == ICMP6_WRUREQUEST || 668 icmp6_type == ICMP6_FQDN_QUERY || 669 icmp6_type == ICMP6_NI_QUERY)) 670 return (1); 671 672 return (0); 673} 674 675static void 676send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 677{ 678 struct mbuf *m; 679 680 m = args->m; 681 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 682 struct tcphdr *tcp; 683 tcp_seq ack, seq; 684 int flags; 685 struct { 686 struct ip6_hdr ip6; 687 struct tcphdr th; 688 } ti; 689 tcp = (struct tcphdr *)((char *)ip6 + hlen); 690 691 if ((tcp->th_flags & TH_RST) != 0) { 692 m_freem(m); 693 args->m = NULL; 694 return; 695 } 696 697 ti.ip6 = *ip6; 698 ti.th = *tcp; 699 ti.th.th_seq = ntohl(ti.th.th_seq); 700 ti.th.th_ack = ntohl(ti.th.th_ack); 701 ti.ip6.ip6_nxt = IPPROTO_TCP; 702 703 if (ti.th.th_flags & TH_ACK) { 704 ack = 0; 705 seq = ti.th.th_ack; 706 flags = TH_RST; 707 } else { 708 ack = ti.th.th_seq; 709 if ((m->m_flags & M_PKTHDR) != 0) { 710 /* 711 * total new data to ACK is: 712 * total packet length, 713 * minus the header length, 714 * minus the tcp header length. 715 */ 716 ack += m->m_pkthdr.len - hlen 717 - (ti.th.th_off << 2); 718 } else if (ip6->ip6_plen) { 719 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) - 720 hlen - (ti.th.th_off << 2); 721 } else { 722 m_freem(m); 723 return; 724 } 725 if (tcp->th_flags & TH_SYN) 726 ack++; 727 seq = 0; 728 flags = TH_RST|TH_ACK; 729 } 730 bcopy(&ti, ip6, sizeof(ti)); 731 /* 732 * m is only used to recycle the mbuf 733 * The data in it is never read so we don't need 734 * to correct the offsets or anything 735 */ 736 tcp_respond(NULL, ip6, tcp, m, ack, seq, flags); 737 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 738#if 0 739 /* 740 * Unlike above, the mbufs need to line up with the ip6 hdr, 741 * as the contents are read. We need to m_adj() the 742 * needed amount. 743 * The mbuf will however be thrown away so we can adjust it. 744 * Remember we did an m_pullup on it already so we 745 * can make some assumptions about contiguousness. 746 */ 747 if (args->L3offset) 748 m_adj(m, args->L3offset); 749#endif 750 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 751 } else 752 m_freem(m); 753 754 args->m = NULL; 755} 756 757#endif /* INET6 */ 758 759static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 760 761#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 762#define SNP(buf) buf, sizeof(buf) 763 764/* 765 * We enter here when we have a rule with O_LOG. 766 * XXX this function alone takes about 2Kbytes of code! 767 */ 768static void 769ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 770 struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg, 771 struct ip *ip) 772{ 773 struct ether_header *eh = args->eh; 774 char *action; 775 int limit_reached = 0; 776 char action2[40], proto[128], fragment[32]; 777 778 fragment[0] = '\0'; 779 proto[0] = '\0'; 780 781 if (f == NULL) { /* bogus pkt */ 782 if (verbose_limit != 0 && norule_counter >= verbose_limit) 783 return; 784 norule_counter++; 785 if (norule_counter == verbose_limit) 786 limit_reached = verbose_limit; 787 action = "Refuse"; 788 } else { /* O_LOG is the first action, find the real one */ 789 ipfw_insn *cmd = ACTION_PTR(f); 790 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 791 792 if (l->max_log != 0 && l->log_left == 0) 793 return; 794 l->log_left--; 795 if (l->log_left == 0) 796 limit_reached = l->max_log; 797 cmd += F_LEN(cmd); /* point to first action */ 798 if (cmd->opcode == O_ALTQ) { 799 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 800 801 snprintf(SNPARGS(action2, 0), "Altq %d", 802 altq->qid); 803 cmd += F_LEN(cmd); 804 } 805 if (cmd->opcode == O_PROB) 806 cmd += F_LEN(cmd); 807 808 if (cmd->opcode == O_TAG) 809 cmd += F_LEN(cmd); 810 811 action = action2; 812 switch (cmd->opcode) { 813 case O_DENY: 814 action = "Deny"; 815 break; 816 817 case O_REJECT: 818 if (cmd->arg1==ICMP_REJECT_RST) 819 action = "Reset"; 820 else if (cmd->arg1==ICMP_UNREACH_HOST) 821 action = "Reject"; 822 else 823 snprintf(SNPARGS(action2, 0), "Unreach %d", 824 cmd->arg1); 825 break; 826 827 case O_UNREACH6: 828 if (cmd->arg1==ICMP6_UNREACH_RST) 829 action = "Reset"; 830 else 831 snprintf(SNPARGS(action2, 0), "Unreach %d", 832 cmd->arg1); 833 break; 834 835 case O_ACCEPT: 836 action = "Accept"; 837 break; 838 case O_COUNT: 839 action = "Count"; 840 break; 841 case O_DIVERT: 842 snprintf(SNPARGS(action2, 0), "Divert %d", 843 cmd->arg1); 844 break; 845 case O_TEE: 846 snprintf(SNPARGS(action2, 0), "Tee %d", 847 cmd->arg1); 848 break; 849 case O_SKIPTO: 850 snprintf(SNPARGS(action2, 0), "SkipTo %d", 851 cmd->arg1); 852 break; 853 case O_PIPE: 854 snprintf(SNPARGS(action2, 0), "Pipe %d", 855 cmd->arg1); 856 break; 857 case O_QUEUE: 858 snprintf(SNPARGS(action2, 0), "Queue %d", 859 cmd->arg1); 860 break; 861 case O_FORWARD_IP: { 862 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 863 int len; 864 struct in_addr dummyaddr; 865 if (sa->sa.sin_addr.s_addr == INADDR_ANY) 866 dummyaddr.s_addr = htonl(tablearg); 867 else 868 dummyaddr.s_addr = sa->sa.sin_addr.s_addr; 869 870 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 871 inet_ntoa(dummyaddr)); 872 873 if (sa->sa.sin_port) 874 snprintf(SNPARGS(action2, len), ":%d", 875 sa->sa.sin_port); 876 } 877 break; 878 case O_NETGRAPH: 879 snprintf(SNPARGS(action2, 0), "Netgraph %d", 880 cmd->arg1); 881 break; 882 case O_NGTEE: 883 snprintf(SNPARGS(action2, 0), "Ngtee %d", 884 cmd->arg1); 885 break; 886 case O_NAT: 887 action = "Nat"; 888 break; 889 default: 890 action = "UNKNOWN"; 891 break; 892 } 893 } 894 895 if (hlen == 0) { /* non-ip */ 896 snprintf(SNPARGS(proto, 0), "MAC"); 897 898 } else { 899 int len; 900 char src[48], dst[48]; 901 struct icmphdr *icmp; 902 struct tcphdr *tcp; 903 struct udphdr *udp; 904#ifdef INET6 905 struct ip6_hdr *ip6 = NULL; 906 struct icmp6_hdr *icmp6; 907#endif 908 src[0] = '\0'; 909 dst[0] = '\0'; 910#ifdef INET6 911 if (IS_IP6_FLOW_ID(&(args->f_id))) { 912 char ip6buf[INET6_ADDRSTRLEN]; 913 snprintf(src, sizeof(src), "[%s]", 914 ip6_sprintf(ip6buf, &args->f_id.src_ip6)); 915 snprintf(dst, sizeof(dst), "[%s]", 916 ip6_sprintf(ip6buf, &args->f_id.dst_ip6)); 917 918 ip6 = (struct ip6_hdr *)ip; 919 tcp = (struct tcphdr *)(((char *)ip) + hlen); 920 udp = (struct udphdr *)(((char *)ip) + hlen); 921 } else 922#endif 923 { 924 tcp = L3HDR(struct tcphdr, ip); 925 udp = L3HDR(struct udphdr, ip); 926 927 inet_ntoa_r(ip->ip_src, src); 928 inet_ntoa_r(ip->ip_dst, dst); 929 } 930 931 switch (args->f_id.proto) { 932 case IPPROTO_TCP: 933 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 934 if (offset == 0) 935 snprintf(SNPARGS(proto, len), ":%d %s:%d", 936 ntohs(tcp->th_sport), 937 dst, 938 ntohs(tcp->th_dport)); 939 else 940 snprintf(SNPARGS(proto, len), " %s", dst); 941 break; 942 943 case IPPROTO_UDP: 944 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 945 if (offset == 0) 946 snprintf(SNPARGS(proto, len), ":%d %s:%d", 947 ntohs(udp->uh_sport), 948 dst, 949 ntohs(udp->uh_dport)); 950 else 951 snprintf(SNPARGS(proto, len), " %s", dst); 952 break; 953 954 case IPPROTO_ICMP: 955 icmp = L3HDR(struct icmphdr, ip); 956 if (offset == 0) 957 len = snprintf(SNPARGS(proto, 0), 958 "ICMP:%u.%u ", 959 icmp->icmp_type, icmp->icmp_code); 960 else 961 len = snprintf(SNPARGS(proto, 0), "ICMP "); 962 len += snprintf(SNPARGS(proto, len), "%s", src); 963 snprintf(SNPARGS(proto, len), " %s", dst); 964 break; 965#ifdef INET6 966 case IPPROTO_ICMPV6: 967 icmp6 = (struct icmp6_hdr *)(((char *)ip) + hlen); 968 if (offset == 0) 969 len = snprintf(SNPARGS(proto, 0), 970 "ICMPv6:%u.%u ", 971 icmp6->icmp6_type, icmp6->icmp6_code); 972 else 973 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 974 len += snprintf(SNPARGS(proto, len), "%s", src); 975 snprintf(SNPARGS(proto, len), " %s", dst); 976 break; 977#endif 978 default: 979 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 980 args->f_id.proto, src); 981 snprintf(SNPARGS(proto, len), " %s", dst); 982 break; 983 } 984 985#ifdef INET6 986 if (IS_IP6_FLOW_ID(&(args->f_id))) { 987 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 988 snprintf(SNPARGS(fragment, 0), 989 " (frag %08x:%d@%d%s)", 990 args->f_id.frag_id6, 991 ntohs(ip6->ip6_plen) - hlen, 992 ntohs(offset & IP6F_OFF_MASK) << 3, 993 (offset & IP6F_MORE_FRAG) ? "+" : ""); 994 } else 995#endif 996 { 997 int ip_off, ip_len; 998 if (eh != NULL) { /* layer 2 packets are as on the wire */ 999 ip_off = ntohs(ip->ip_off); 1000 ip_len = ntohs(ip->ip_len); 1001 } else { 1002 ip_off = ip->ip_off; 1003 ip_len = ip->ip_len; 1004 } 1005 if (ip_off & (IP_MF | IP_OFFMASK)) 1006 snprintf(SNPARGS(fragment, 0), 1007 " (frag %d:%d@%d%s)", 1008 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 1009 offset << 3, 1010 (ip_off & IP_MF) ? "+" : ""); 1011 } 1012 } 1013 if (oif || m->m_pkthdr.rcvif) 1014 log(LOG_SECURITY | LOG_INFO, 1015 "ipfw: %d %s %s %s via %s%s\n", 1016 f ? f->rulenum : -1, 1017 action, proto, oif ? "out" : "in", 1018 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 1019 fragment); 1020 else 1021 log(LOG_SECURITY | LOG_INFO, 1022 "ipfw: %d %s %s [no if info]%s\n", 1023 f ? f->rulenum : -1, 1024 action, proto, fragment); 1025 if (limit_reached) 1026 log(LOG_SECURITY | LOG_NOTICE, 1027 "ipfw: limit %d reached on entry %d\n", 1028 limit_reached, f ? f->rulenum : -1); 1029} 1030 1031/* 1032 * IMPORTANT: the hash function for dynamic rules must be commutative 1033 * in source and destination (ip,port), because rules are bidirectional 1034 * and we want to find both in the same bucket. 1035 */ 1036static __inline int 1037hash_packet(struct ipfw_flow_id *id) 1038{ 1039 u_int32_t i; 1040 1041#ifdef INET6 1042 if (IS_IP6_FLOW_ID(id)) 1043 i = hash_packet6(id); 1044 else 1045#endif /* INET6 */ 1046 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1047 i &= (curr_dyn_buckets - 1); 1048 return i; 1049} 1050 1051/** 1052 * unlink a dynamic rule from a chain. prev is a pointer to 1053 * the previous one, q is a pointer to the rule to delete, 1054 * head is a pointer to the head of the queue. 1055 * Modifies q and potentially also head. 1056 */ 1057#define UNLINK_DYN_RULE(prev, head, q) { \ 1058 ipfw_dyn_rule *old_q = q; \ 1059 \ 1060 /* remove a refcount to the parent */ \ 1061 if (q->dyn_type == O_LIMIT) \ 1062 q->parent->count--; \ 1063 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1064 (q->id.src_ip), (q->id.src_port), \ 1065 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 1066 if (prev != NULL) \ 1067 prev->next = q = q->next; \ 1068 else \ 1069 head = q = q->next; \ 1070 dyn_count--; \ 1071 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1072 1073#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1074 1075/** 1076 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1077 * 1078 * If keep_me == NULL, rules are deleted even if not expired, 1079 * otherwise only expired rules are removed. 1080 * 1081 * The value of the second parameter is also used to point to identify 1082 * a rule we absolutely do not want to remove (e.g. because we are 1083 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1084 * rules). The pointer is only used for comparison, so any non-null 1085 * value will do. 1086 */ 1087static void 1088remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1089{ 1090 static u_int32_t last_remove = 0; 1091 1092#define FORCE (keep_me == NULL) 1093 1094 ipfw_dyn_rule *prev, *q; 1095 int i, pass = 0, max_pass = 0; 1096 1097 IPFW_DYN_LOCK_ASSERT(); 1098 1099 if (ipfw_dyn_v == NULL || dyn_count == 0) 1100 return; 1101 /* do not expire more than once per second, it is useless */ 1102 if (!FORCE && last_remove == time_uptime) 1103 return; 1104 last_remove = time_uptime; 1105 1106 /* 1107 * because O_LIMIT refer to parent rules, during the first pass only 1108 * remove child and mark any pending LIMIT_PARENT, and remove 1109 * them in a second pass. 1110 */ 1111next_pass: 1112 for (i = 0 ; i < curr_dyn_buckets ; i++) { 1113 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 1114 /* 1115 * Logic can become complex here, so we split tests. 1116 */ 1117 if (q == keep_me) 1118 goto next; 1119 if (rule != NULL && rule != q->rule) 1120 goto next; /* not the one we are looking for */ 1121 if (q->dyn_type == O_LIMIT_PARENT) { 1122 /* 1123 * handle parent in the second pass, 1124 * record we need one. 1125 */ 1126 max_pass = 1; 1127 if (pass == 0) 1128 goto next; 1129 if (FORCE && q->count != 0 ) { 1130 /* XXX should not happen! */ 1131 printf("ipfw: OUCH! cannot remove rule," 1132 " count %d\n", q->count); 1133 } 1134 } else { 1135 if (!FORCE && 1136 !TIME_LEQ( q->expire, time_uptime )) 1137 goto next; 1138 } 1139 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1140 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1141 continue; 1142 } 1143next: 1144 prev=q; 1145 q=q->next; 1146 } 1147 } 1148 if (pass++ < max_pass) 1149 goto next_pass; 1150} 1151 1152 1153/** 1154 * lookup a dynamic rule. 1155 */ 1156static ipfw_dyn_rule * 1157lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1158 struct tcphdr *tcp) 1159{ 1160 /* 1161 * stateful ipfw extensions. 1162 * Lookup into dynamic session queue 1163 */ 1164#define MATCH_REVERSE 0 1165#define MATCH_FORWARD 1 1166#define MATCH_NONE 2 1167#define MATCH_UNKNOWN 3 1168 int i, dir = MATCH_NONE; 1169 ipfw_dyn_rule *prev, *q=NULL; 1170 1171 IPFW_DYN_LOCK_ASSERT(); 1172 1173 if (ipfw_dyn_v == NULL) 1174 goto done; /* not found */ 1175 i = hash_packet( pkt ); 1176 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 1177 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1178 goto next; 1179 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1180 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1181 continue; 1182 } 1183 if (pkt->proto == q->id.proto && 1184 q->dyn_type != O_LIMIT_PARENT) { 1185 if (IS_IP6_FLOW_ID(pkt)) { 1186 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1187 &(q->id.src_ip6)) && 1188 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1189 &(q->id.dst_ip6)) && 1190 pkt->src_port == q->id.src_port && 1191 pkt->dst_port == q->id.dst_port ) { 1192 dir = MATCH_FORWARD; 1193 break; 1194 } 1195 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1196 &(q->id.dst_ip6)) && 1197 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1198 &(q->id.src_ip6)) && 1199 pkt->src_port == q->id.dst_port && 1200 pkt->dst_port == q->id.src_port ) { 1201 dir = MATCH_REVERSE; 1202 break; 1203 } 1204 } else { 1205 if (pkt->src_ip == q->id.src_ip && 1206 pkt->dst_ip == q->id.dst_ip && 1207 pkt->src_port == q->id.src_port && 1208 pkt->dst_port == q->id.dst_port ) { 1209 dir = MATCH_FORWARD; 1210 break; 1211 } 1212 if (pkt->src_ip == q->id.dst_ip && 1213 pkt->dst_ip == q->id.src_ip && 1214 pkt->src_port == q->id.dst_port && 1215 pkt->dst_port == q->id.src_port ) { 1216 dir = MATCH_REVERSE; 1217 break; 1218 } 1219 } 1220 } 1221next: 1222 prev = q; 1223 q = q->next; 1224 } 1225 if (q == NULL) 1226 goto done; /* q = NULL, not found */ 1227 1228 if ( prev != NULL) { /* found and not in front */ 1229 prev->next = q->next; 1230 q->next = ipfw_dyn_v[i]; 1231 ipfw_dyn_v[i] = q; 1232 } 1233 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1234 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1235 1236#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1237#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1238 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1239 switch (q->state) { 1240 case TH_SYN: /* opening */ 1241 q->expire = time_uptime + dyn_syn_lifetime; 1242 break; 1243 1244 case BOTH_SYN: /* move to established */ 1245 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1246 case BOTH_SYN | (TH_FIN << 8) : 1247 if (tcp) { 1248#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1249 u_int32_t ack = ntohl(tcp->th_ack); 1250 if (dir == MATCH_FORWARD) { 1251 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1252 q->ack_fwd = ack; 1253 else { /* ignore out-of-sequence */ 1254 break; 1255 } 1256 } else { 1257 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1258 q->ack_rev = ack; 1259 else { /* ignore out-of-sequence */ 1260 break; 1261 } 1262 } 1263 } 1264 q->expire = time_uptime + dyn_ack_lifetime; 1265 break; 1266 1267 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1268 if (dyn_fin_lifetime >= dyn_keepalive_period) 1269 dyn_fin_lifetime = dyn_keepalive_period - 1; 1270 q->expire = time_uptime + dyn_fin_lifetime; 1271 break; 1272 1273 default: 1274#if 0 1275 /* 1276 * reset or some invalid combination, but can also 1277 * occur if we use keep-state the wrong way. 1278 */ 1279 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1280 printf("invalid state: 0x%x\n", q->state); 1281#endif 1282 if (dyn_rst_lifetime >= dyn_keepalive_period) 1283 dyn_rst_lifetime = dyn_keepalive_period - 1; 1284 q->expire = time_uptime + dyn_rst_lifetime; 1285 break; 1286 } 1287 } else if (pkt->proto == IPPROTO_UDP) { 1288 q->expire = time_uptime + dyn_udp_lifetime; 1289 } else { 1290 /* other protocols */ 1291 q->expire = time_uptime + dyn_short_lifetime; 1292 } 1293done: 1294 if (match_direction) 1295 *match_direction = dir; 1296 return q; 1297} 1298 1299static ipfw_dyn_rule * 1300lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1301 struct tcphdr *tcp) 1302{ 1303 ipfw_dyn_rule *q; 1304 1305 IPFW_DYN_LOCK(); 1306 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1307 if (q == NULL) 1308 IPFW_DYN_UNLOCK(); 1309 /* NB: return table locked when q is not NULL */ 1310 return q; 1311} 1312 1313static void 1314realloc_dynamic_table(void) 1315{ 1316 IPFW_DYN_LOCK_ASSERT(); 1317 1318 /* 1319 * Try reallocation, make sure we have a power of 2 and do 1320 * not allow more than 64k entries. In case of overflow, 1321 * default to 1024. 1322 */ 1323 1324 if (dyn_buckets > 65536) 1325 dyn_buckets = 1024; 1326 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1327 dyn_buckets = curr_dyn_buckets; /* reset */ 1328 return; 1329 } 1330 curr_dyn_buckets = dyn_buckets; 1331 if (ipfw_dyn_v != NULL) 1332 free(ipfw_dyn_v, M_IPFW); 1333 for (;;) { 1334 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1335 M_IPFW, M_NOWAIT | M_ZERO); 1336 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 1337 break; 1338 curr_dyn_buckets /= 2; 1339 } 1340} 1341 1342/** 1343 * Install state of type 'type' for a dynamic session. 1344 * The hash table contains two type of rules: 1345 * - regular rules (O_KEEP_STATE) 1346 * - rules for sessions with limited number of sess per user 1347 * (O_LIMIT). When they are created, the parent is 1348 * increased by 1, and decreased on delete. In this case, 1349 * the third parameter is the parent rule and not the chain. 1350 * - "parent" rules for the above (O_LIMIT_PARENT). 1351 */ 1352static ipfw_dyn_rule * 1353add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1354{ 1355 ipfw_dyn_rule *r; 1356 int i; 1357 1358 IPFW_DYN_LOCK_ASSERT(); 1359 1360 if (ipfw_dyn_v == NULL || 1361 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1362 realloc_dynamic_table(); 1363 if (ipfw_dyn_v == NULL) 1364 return NULL; /* failed ! */ 1365 } 1366 i = hash_packet(id); 1367 1368 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1369 if (r == NULL) { 1370 printf ("ipfw: sorry cannot allocate state\n"); 1371 return NULL; 1372 } 1373 1374 /* increase refcount on parent, and set pointer */ 1375 if (dyn_type == O_LIMIT) { 1376 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1377 if ( parent->dyn_type != O_LIMIT_PARENT) 1378 panic("invalid parent"); 1379 parent->count++; 1380 r->parent = parent; 1381 rule = parent->rule; 1382 } 1383 1384 r->id = *id; 1385 r->expire = time_uptime + dyn_syn_lifetime; 1386 r->rule = rule; 1387 r->dyn_type = dyn_type; 1388 r->pcnt = r->bcnt = 0; 1389 r->count = 0; 1390 1391 r->bucket = i; 1392 r->next = ipfw_dyn_v[i]; 1393 ipfw_dyn_v[i] = r; 1394 dyn_count++; 1395 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1396 dyn_type, 1397 (r->id.src_ip), (r->id.src_port), 1398 (r->id.dst_ip), (r->id.dst_port), 1399 dyn_count ); ) 1400 return r; 1401} 1402 1403/** 1404 * lookup dynamic parent rule using pkt and rule as search keys. 1405 * If the lookup fails, then install one. 1406 */ 1407static ipfw_dyn_rule * 1408lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1409{ 1410 ipfw_dyn_rule *q; 1411 int i; 1412 1413 IPFW_DYN_LOCK_ASSERT(); 1414 1415 if (ipfw_dyn_v) { 1416 int is_v6 = IS_IP6_FLOW_ID(pkt); 1417 i = hash_packet( pkt ); 1418 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1419 if (q->dyn_type == O_LIMIT_PARENT && 1420 rule== q->rule && 1421 pkt->proto == q->id.proto && 1422 pkt->src_port == q->id.src_port && 1423 pkt->dst_port == q->id.dst_port && 1424 ( 1425 (is_v6 && 1426 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1427 &(q->id.src_ip6)) && 1428 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1429 &(q->id.dst_ip6))) || 1430 (!is_v6 && 1431 pkt->src_ip == q->id.src_ip && 1432 pkt->dst_ip == q->id.dst_ip) 1433 ) 1434 ) { 1435 q->expire = time_uptime + dyn_short_lifetime; 1436 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1437 return q; 1438 } 1439 } 1440 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1441} 1442 1443/** 1444 * Install dynamic state for rule type cmd->o.opcode 1445 * 1446 * Returns 1 (failure) if state is not installed because of errors or because 1447 * session limitations are enforced. 1448 */ 1449static int 1450install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1451 struct ip_fw_args *args, uint32_t tablearg) 1452{ 1453 static int last_log; 1454 ipfw_dyn_rule *q; 1455 struct in_addr da; 1456 char src[48], dst[48]; 1457 1458 src[0] = '\0'; 1459 dst[0] = '\0'; 1460 1461 DEB( 1462 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1463 __func__, cmd->o.opcode, 1464 (args->f_id.src_ip), (args->f_id.src_port), 1465 (args->f_id.dst_ip), (args->f_id.dst_port)); 1466 ) 1467 1468 IPFW_DYN_LOCK(); 1469 1470 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1471 1472 if (q != NULL) { /* should never occur */ 1473 if (last_log != time_uptime) { 1474 last_log = time_uptime; 1475 printf("ipfw: %s: entry already present, done\n", 1476 __func__); 1477 } 1478 IPFW_DYN_UNLOCK(); 1479 return (0); 1480 } 1481 1482 if (dyn_count >= dyn_max) 1483 /* Run out of slots, try to remove any expired rule. */ 1484 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1485 1486 if (dyn_count >= dyn_max) { 1487 if (last_log != time_uptime) { 1488 last_log = time_uptime; 1489 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1490 } 1491 IPFW_DYN_UNLOCK(); 1492 return (1); /* cannot install, notify caller */ 1493 } 1494 1495 switch (cmd->o.opcode) { 1496 case O_KEEP_STATE: /* bidir rule */ 1497 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1498 break; 1499 1500 case O_LIMIT: { /* limit number of sessions */ 1501 struct ipfw_flow_id id; 1502 ipfw_dyn_rule *parent; 1503 uint32_t conn_limit; 1504 uint16_t limit_mask = cmd->limit_mask; 1505 1506 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1507 tablearg : cmd->conn_limit; 1508 1509 DEB( 1510 if (cmd->conn_limit == IP_FW_TABLEARG) 1511 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1512 "(tablearg)\n", __func__, conn_limit); 1513 else 1514 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1515 __func__, conn_limit); 1516 ) 1517 1518 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1519 id.proto = args->f_id.proto; 1520 id.addr_type = args->f_id.addr_type; 1521 1522 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1523 if (limit_mask & DYN_SRC_ADDR) 1524 id.src_ip6 = args->f_id.src_ip6; 1525 if (limit_mask & DYN_DST_ADDR) 1526 id.dst_ip6 = args->f_id.dst_ip6; 1527 } else { 1528 if (limit_mask & DYN_SRC_ADDR) 1529 id.src_ip = args->f_id.src_ip; 1530 if (limit_mask & DYN_DST_ADDR) 1531 id.dst_ip = args->f_id.dst_ip; 1532 } 1533 if (limit_mask & DYN_SRC_PORT) 1534 id.src_port = args->f_id.src_port; 1535 if (limit_mask & DYN_DST_PORT) 1536 id.dst_port = args->f_id.dst_port; 1537 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1538 printf("ipfw: %s: add parent failed\n", __func__); 1539 IPFW_DYN_UNLOCK(); 1540 return (1); 1541 } 1542 1543 if (parent->count >= conn_limit) { 1544 /* See if we can remove some expired rule. */ 1545 remove_dyn_rule(rule, parent); 1546 if (parent->count >= conn_limit) { 1547 if (fw_verbose && last_log != time_uptime) { 1548 last_log = time_uptime; 1549#ifdef INET6 1550 /* 1551 * XXX IPv6 flows are not 1552 * supported yet. 1553 */ 1554 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1555 char ip6buf[INET6_ADDRSTRLEN]; 1556 snprintf(src, sizeof(src), 1557 "[%s]", ip6_sprintf(ip6buf, 1558 &args->f_id.src_ip6)); 1559 snprintf(dst, sizeof(dst), 1560 "[%s]", ip6_sprintf(ip6buf, 1561 &args->f_id.dst_ip6)); 1562 } else 1563#endif 1564 { 1565 da.s_addr = 1566 htonl(args->f_id.src_ip); 1567 inet_ntoa_r(da, src); 1568 da.s_addr = 1569 htonl(args->f_id.dst_ip); 1570 inet_ntoa_r(da, dst); 1571 } 1572 log(LOG_SECURITY | LOG_DEBUG, 1573 "%s %s:%u -> %s:%u, %s\n", 1574 "drop session", 1575 src, (args->f_id.src_port), 1576 dst, (args->f_id.dst_port), 1577 "too many entries"); 1578 } 1579 IPFW_DYN_UNLOCK(); 1580 return (1); 1581 } 1582 } 1583 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1584 break; 1585 } 1586 default: 1587 printf("ipfw: %s: unknown dynamic rule type %u\n", 1588 __func__, cmd->o.opcode); 1589 IPFW_DYN_UNLOCK(); 1590 return (1); 1591 } 1592 1593 /* XXX just set lifetime */ 1594 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1595 1596 IPFW_DYN_UNLOCK(); 1597 return (0); 1598} 1599 1600/* 1601 * Generate a TCP packet, containing either a RST or a keepalive. 1602 * When flags & TH_RST, we are sending a RST packet, because of a 1603 * "reset" action matched the packet. 1604 * Otherwise we are sending a keepalive, and flags & TH_ 1605 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 1606 * so that MAC can label the reply appropriately. 1607 */ 1608static struct mbuf * 1609send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 1610 u_int32_t ack, int flags) 1611{ 1612 struct mbuf *m; 1613 struct ip *ip; 1614 struct tcphdr *tcp; 1615 1616 MGETHDR(m, M_DONTWAIT, MT_DATA); 1617 if (m == 0) 1618 return (NULL); 1619 m->m_pkthdr.rcvif = (struct ifnet *)0; 1620 1621#ifdef MAC 1622 if (replyto != NULL) 1623 mac_create_mbuf_netlayer(replyto, m); 1624 else 1625 mac_create_mbuf_from_firewall(m); 1626#else 1627 (void)replyto; /* don't warn about unused arg */ 1628#endif 1629 1630 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1631 m->m_data += max_linkhdr; 1632 1633 ip = mtod(m, struct ip *); 1634 bzero(ip, m->m_len); 1635 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1636 ip->ip_p = IPPROTO_TCP; 1637 tcp->th_off = 5; 1638 /* 1639 * Assume we are sending a RST (or a keepalive in the reverse 1640 * direction), swap src and destination addresses and ports. 1641 */ 1642 ip->ip_src.s_addr = htonl(id->dst_ip); 1643 ip->ip_dst.s_addr = htonl(id->src_ip); 1644 tcp->th_sport = htons(id->dst_port); 1645 tcp->th_dport = htons(id->src_port); 1646 if (flags & TH_RST) { /* we are sending a RST */ 1647 if (flags & TH_ACK) { 1648 tcp->th_seq = htonl(ack); 1649 tcp->th_ack = htonl(0); 1650 tcp->th_flags = TH_RST; 1651 } else { 1652 if (flags & TH_SYN) 1653 seq++; 1654 tcp->th_seq = htonl(0); 1655 tcp->th_ack = htonl(seq); 1656 tcp->th_flags = TH_RST | TH_ACK; 1657 } 1658 } else { 1659 /* 1660 * We are sending a keepalive. flags & TH_SYN determines 1661 * the direction, forward if set, reverse if clear. 1662 * NOTE: seq and ack are always assumed to be correct 1663 * as set by the caller. This may be confusing... 1664 */ 1665 if (flags & TH_SYN) { 1666 /* 1667 * we have to rewrite the correct addresses! 1668 */ 1669 ip->ip_dst.s_addr = htonl(id->dst_ip); 1670 ip->ip_src.s_addr = htonl(id->src_ip); 1671 tcp->th_dport = htons(id->dst_port); 1672 tcp->th_sport = htons(id->src_port); 1673 } 1674 tcp->th_seq = htonl(seq); 1675 tcp->th_ack = htonl(ack); 1676 tcp->th_flags = TH_ACK; 1677 } 1678 /* 1679 * set ip_len to the payload size so we can compute 1680 * the tcp checksum on the pseudoheader 1681 * XXX check this, could save a couple of words ? 1682 */ 1683 ip->ip_len = htons(sizeof(struct tcphdr)); 1684 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1685 /* 1686 * now fill fields left out earlier 1687 */ 1688 ip->ip_ttl = ip_defttl; 1689 ip->ip_len = m->m_pkthdr.len; 1690 m->m_flags |= M_SKIP_FIREWALL; 1691 return (m); 1692} 1693 1694/* 1695 * sends a reject message, consuming the mbuf passed as an argument. 1696 */ 1697static void 1698send_reject(struct ip_fw_args *args, int code, int ip_len, struct ip *ip) 1699{ 1700 1701#if 0 1702 /* XXX When ip is not guaranteed to be at mtod() we will 1703 * need to account for this */ 1704 * The mbuf will however be thrown away so we can adjust it. 1705 * Remember we did an m_pullup on it already so we 1706 * can make some assumptions about contiguousness. 1707 */ 1708 if (args->L3offset) 1709 m_adj(m, args->L3offset); 1710#endif 1711 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1712 /* We need the IP header in host order for icmp_error(). */ 1713 if (args->eh != NULL) { 1714 ip->ip_len = ntohs(ip->ip_len); 1715 ip->ip_off = ntohs(ip->ip_off); 1716 } 1717 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1718 } else if (args->f_id.proto == IPPROTO_TCP) { 1719 struct tcphdr *const tcp = 1720 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1721 if ( (tcp->th_flags & TH_RST) == 0) { 1722 struct mbuf *m; 1723 m = send_pkt(args->m, &(args->f_id), 1724 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1725 tcp->th_flags | TH_RST); 1726 if (m != NULL) 1727 ip_output(m, NULL, NULL, 0, NULL, NULL); 1728 } 1729 m_freem(args->m); 1730 } else 1731 m_freem(args->m); 1732 args->m = NULL; 1733} 1734 1735/** 1736 * 1737 * Given an ip_fw *, lookup_next_rule will return a pointer 1738 * to the next rule, which can be either the jump 1739 * target (for skipto instructions) or the next one in the list (in 1740 * all other cases including a missing jump target). 1741 * The result is also written in the "next_rule" field of the rule. 1742 * Backward jumps are not allowed, so start looking from the next 1743 * rule... 1744 * 1745 * This never returns NULL -- in case we do not have an exact match, 1746 * the next rule is returned. When the ruleset is changed, 1747 * pointers are flushed so we are always correct. 1748 */ 1749 1750static struct ip_fw * 1751lookup_next_rule(struct ip_fw *me) 1752{ 1753 struct ip_fw *rule = NULL; 1754 ipfw_insn *cmd; 1755 1756 /* look for action, in case it is a skipto */ 1757 cmd = ACTION_PTR(me); 1758 if (cmd->opcode == O_LOG) 1759 cmd += F_LEN(cmd); 1760 if (cmd->opcode == O_ALTQ) 1761 cmd += F_LEN(cmd); 1762 if (cmd->opcode == O_TAG) 1763 cmd += F_LEN(cmd); 1764 if ( cmd->opcode == O_SKIPTO ) 1765 for (rule = me->next; rule ; rule = rule->next) 1766 if (rule->rulenum >= cmd->arg1) 1767 break; 1768 if (rule == NULL) /* failure or not a skipto */ 1769 rule = me->next; 1770 me->next_rule = rule; 1771 return rule; 1772} 1773 1774static int 1775add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1776 uint8_t mlen, uint32_t value) 1777{ 1778 struct radix_node_head *rnh; 1779 struct table_entry *ent; 1780 1781 if (tbl >= IPFW_TABLES_MAX) 1782 return (EINVAL); 1783 rnh = ch->tables[tbl]; 1784 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1785 if (ent == NULL) 1786 return (ENOMEM); 1787 ent->value = value; 1788 ent->addr.sin_len = ent->mask.sin_len = 8; 1789 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1790 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1791 IPFW_WLOCK(&layer3_chain); 1792 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1793 NULL) { 1794 IPFW_WUNLOCK(&layer3_chain); 1795 free(ent, M_IPFW_TBL); 1796 return (EEXIST); 1797 } 1798 IPFW_WUNLOCK(&layer3_chain); 1799 return (0); 1800} 1801 1802static int 1803del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1804 uint8_t mlen) 1805{ 1806 struct radix_node_head *rnh; 1807 struct table_entry *ent; 1808 struct sockaddr_in sa, mask; 1809 1810 if (tbl >= IPFW_TABLES_MAX) 1811 return (EINVAL); 1812 rnh = ch->tables[tbl]; 1813 sa.sin_len = mask.sin_len = 8; 1814 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1815 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1816 IPFW_WLOCK(ch); 1817 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1818 if (ent == NULL) { 1819 IPFW_WUNLOCK(ch); 1820 return (ESRCH); 1821 } 1822 IPFW_WUNLOCK(ch); 1823 free(ent, M_IPFW_TBL); 1824 return (0); 1825} 1826 1827static int 1828flush_table_entry(struct radix_node *rn, void *arg) 1829{ 1830 struct radix_node_head * const rnh = arg; 1831 struct table_entry *ent; 1832 1833 ent = (struct table_entry *) 1834 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1835 if (ent != NULL) 1836 free(ent, M_IPFW_TBL); 1837 return (0); 1838} 1839 1840static int 1841flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1842{ 1843 struct radix_node_head *rnh; 1844 1845 IPFW_WLOCK_ASSERT(ch); 1846 1847 if (tbl >= IPFW_TABLES_MAX) 1848 return (EINVAL); 1849 rnh = ch->tables[tbl]; 1850 KASSERT(rnh != NULL, ("NULL IPFW table")); 1851 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1852 return (0); 1853} 1854 1855static void 1856flush_tables(struct ip_fw_chain *ch) 1857{ 1858 uint16_t tbl; 1859 1860 IPFW_WLOCK_ASSERT(ch); 1861 1862 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1863 flush_table(ch, tbl); 1864} 1865 1866static int 1867init_tables(struct ip_fw_chain *ch) 1868{ 1869 int i; 1870 uint16_t j; 1871 1872 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1873 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1874 for (j = 0; j < i; j++) { 1875 (void) flush_table(ch, j); 1876 } 1877 return (ENOMEM); 1878 } 1879 } 1880 return (0); 1881} 1882 1883static int 1884lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1885 uint32_t *val) 1886{ 1887 struct radix_node_head *rnh; 1888 struct table_entry *ent; 1889 struct sockaddr_in sa; 1890 1891 if (tbl >= IPFW_TABLES_MAX) 1892 return (0); 1893 rnh = ch->tables[tbl]; 1894 sa.sin_len = 8; 1895 sa.sin_addr.s_addr = addr; 1896 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1897 if (ent != NULL) { 1898 *val = ent->value; 1899 return (1); 1900 } 1901 return (0); 1902} 1903 1904static int 1905count_table_entry(struct radix_node *rn, void *arg) 1906{ 1907 u_int32_t * const cnt = arg; 1908 1909 (*cnt)++; 1910 return (0); 1911} 1912 1913static int 1914count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1915{ 1916 struct radix_node_head *rnh; 1917 1918 if (tbl >= IPFW_TABLES_MAX) 1919 return (EINVAL); 1920 rnh = ch->tables[tbl]; 1921 *cnt = 0; 1922 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1923 return (0); 1924} 1925 1926static int 1927dump_table_entry(struct radix_node *rn, void *arg) 1928{ 1929 struct table_entry * const n = (struct table_entry *)rn; 1930 ipfw_table * const tbl = arg; 1931 ipfw_table_entry *ent; 1932 1933 if (tbl->cnt == tbl->size) 1934 return (1); 1935 ent = &tbl->ent[tbl->cnt]; 1936 ent->tbl = tbl->tbl; 1937 if (in_nullhost(n->mask.sin_addr)) 1938 ent->masklen = 0; 1939 else 1940 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1941 ent->addr = n->addr.sin_addr.s_addr; 1942 ent->value = n->value; 1943 tbl->cnt++; 1944 return (0); 1945} 1946 1947static int 1948dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1949{ 1950 struct radix_node_head *rnh; 1951 1952 if (tbl->tbl >= IPFW_TABLES_MAX) 1953 return (EINVAL); 1954 rnh = ch->tables[tbl->tbl]; 1955 tbl->cnt = 0; 1956 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1957 return (0); 1958} 1959 1960static void 1961fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1962{ 1963 struct ucred *cr; 1964 1965 if (inp->inp_socket != NULL) { 1966 cr = inp->inp_socket->so_cred; 1967 ugp->fw_prid = jailed(cr) ? 1968 cr->cr_prison->pr_id : -1; 1969 ugp->fw_uid = cr->cr_uid; 1970 ugp->fw_ngroups = cr->cr_ngroups; 1971 bcopy(cr->cr_groups, ugp->fw_groups, 1972 sizeof(ugp->fw_groups)); 1973 } 1974} 1975 1976static int 1977check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif, 1978 struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip, 1979 u_int16_t src_port, struct ip_fw_ugid *ugp, int *lookup, 1980 struct inpcb *inp) 1981{ 1982 struct inpcbinfo *pi; 1983 int wildcard; 1984 struct inpcb *pcb; 1985 int match; 1986 gid_t *gp; 1987 1988 /* 1989 * Check to see if the UDP or TCP stack supplied us with 1990 * the PCB. If so, rather then holding a lock and looking 1991 * up the PCB, we can use the one that was supplied. 1992 */ 1993 if (inp && *lookup == 0) { 1994 INP_LOCK_ASSERT(inp); 1995 if (inp->inp_socket != NULL) { 1996 fill_ugid_cache(inp, ugp); 1997 *lookup = 1; 1998 } 1999 } 2000 /* 2001 * If we have already been here and the packet has no 2002 * PCB entry associated with it, then we can safely 2003 * assume that this is a no match. 2004 */ 2005 if (*lookup == -1) 2006 return (0); 2007 if (proto == IPPROTO_TCP) { 2008 wildcard = 0; 2009 pi = &tcbinfo; 2010 } else if (proto == IPPROTO_UDP) { 2011 wildcard = INPLOOKUP_WILDCARD; 2012 pi = &udbinfo; 2013 } else 2014 return 0; 2015 match = 0; 2016 if (*lookup == 0) { 2017 INP_INFO_RLOCK(pi); 2018 pcb = (oif) ? 2019 in_pcblookup_hash(pi, 2020 dst_ip, htons(dst_port), 2021 src_ip, htons(src_port), 2022 wildcard, oif) : 2023 in_pcblookup_hash(pi, 2024 src_ip, htons(src_port), 2025 dst_ip, htons(dst_port), 2026 wildcard, NULL); 2027 if (pcb != NULL) { 2028 INP_LOCK(pcb); 2029 if (pcb->inp_socket != NULL) { 2030 fill_ugid_cache(pcb, ugp); 2031 *lookup = 1; 2032 } 2033 INP_UNLOCK(pcb); 2034 } 2035 INP_INFO_RUNLOCK(pi); 2036 if (*lookup == 0) { 2037 /* 2038 * If the lookup did not yield any results, there 2039 * is no sense in coming back and trying again. So 2040 * we can set lookup to -1 and ensure that we wont 2041 * bother the pcb system again. 2042 */ 2043 *lookup = -1; 2044 return (0); 2045 } 2046 } 2047 if (insn->o.opcode == O_UID) 2048 match = (ugp->fw_uid == (uid_t)insn->d[0]); 2049 else if (insn->o.opcode == O_GID) { 2050 for (gp = ugp->fw_groups; 2051 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 2052 if (*gp == (gid_t)insn->d[0]) { 2053 match = 1; 2054 break; 2055 } 2056 } else if (insn->o.opcode == O_JAIL) 2057 match = (ugp->fw_prid == (int)insn->d[0]); 2058 return match; 2059} 2060 2061#ifdef IPFIREWALL_NAT 2062static eventhandler_tag ifaddr_event_tag; 2063 2064static void 2065ifaddr_change(void *arg __unused, struct ifnet *ifp) 2066{ 2067 struct cfg_nat *ptr; 2068 struct ifaddr *ifa; 2069 2070 IPFW_WLOCK(&layer3_chain); 2071 /* Check every nat entry... */ 2072 LIST_FOREACH(ptr, &layer3_chain.nat, _next) { 2073 /* ...using nic 'ifp->if_xname' as dynamic alias address. */ 2074 if (strncmp(ptr->if_name, ifp->if_xname, IF_NAMESIZE) == 0) { 2075 mtx_lock(&ifp->if_addr_mtx); 2076 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) { 2077 if (ifa->ifa_addr == NULL) 2078 continue; 2079 if (ifa->ifa_addr->sa_family != AF_INET) 2080 continue; 2081 ptr->ip = ((struct sockaddr_in *) 2082 (ifa->ifa_addr))->sin_addr; 2083 LibAliasSetAddress(ptr->lib, ptr->ip); 2084 } 2085 mtx_unlock(&ifp->if_addr_mtx); 2086 } 2087 } 2088 IPFW_WUNLOCK(&layer3_chain); 2089} 2090 2091static void 2092flush_nat_ptrs(const int i) 2093{ 2094 struct ip_fw *rule; 2095 2096 IPFW_WLOCK_ASSERT(&layer3_chain); 2097 for (rule = layer3_chain.rules; rule; rule = rule->next) { 2098 ipfw_insn_nat *cmd = (ipfw_insn_nat *)ACTION_PTR(rule); 2099 if (cmd->o.opcode != O_NAT) 2100 continue; 2101 if (cmd->nat != NULL && cmd->nat->id == i) 2102 cmd->nat = NULL; 2103 } 2104} 2105 2106static struct cfg_nat * 2107lookup_nat(const int i) 2108{ 2109 struct cfg_nat *ptr; 2110 2111 LIST_FOREACH(ptr, &layer3_chain.nat, _next) 2112 if (ptr->id == i) 2113 return(ptr); 2114 return (NULL); 2115} 2116 2117#define HOOK_NAT(b, p) do { \ 2118 IPFW_WLOCK_ASSERT(&layer3_chain); \ 2119 LIST_INSERT_HEAD(b, p, _next); \ 2120} while (0) 2121 2122#define UNHOOK_NAT(p) do { \ 2123 IPFW_WLOCK_ASSERT(&layer3_chain); \ 2124 LIST_REMOVE(p, _next); \ 2125} while (0) 2126 2127#define HOOK_REDIR(b, p) do { \ 2128 LIST_INSERT_HEAD(b, p, _next); \ 2129} while (0) 2130 2131#define HOOK_SPOOL(b, p) do { \ 2132 LIST_INSERT_HEAD(b, p, _next); \ 2133} while (0) 2134 2135static void 2136del_redir_spool_cfg(struct cfg_nat *n, struct redir_chain *head) 2137{ 2138 struct cfg_redir *r, *tmp_r; 2139 struct cfg_spool *s, *tmp_s; 2140 int i, num; 2141 2142 LIST_FOREACH_SAFE(r, head, _next, tmp_r) { 2143 num = 1; /* Number of alias_link to delete. */ 2144 switch (r->mode) { 2145 case REDIR_PORT: 2146 num = r->pport_cnt; 2147 /* FALLTHROUGH */ 2148 case REDIR_ADDR: 2149 case REDIR_PROTO: 2150 /* Delete all libalias redirect entry. */ 2151 for (i = 0; i < num; i++) 2152 LibAliasRedirectDelete(n->lib, r->alink[i]); 2153 /* Del spool cfg if any. */ 2154 LIST_FOREACH_SAFE(s, &r->spool_chain, _next, tmp_s) { 2155 LIST_REMOVE(s, _next); 2156 free(s, M_IPFW); 2157 } 2158 free(r->alink, M_IPFW); 2159 LIST_REMOVE(r, _next); 2160 free(r, M_IPFW); 2161 break; 2162 default: 2163 printf("unknown redirect mode: %u\n", r->mode); 2164 /* XXX - panic?!?!? */ 2165 break; 2166 } 2167 } 2168} 2169 2170static int 2171add_redir_spool_cfg(char *buf, struct cfg_nat *ptr) 2172{ 2173 struct cfg_redir *r, *ser_r; 2174 struct cfg_spool *s, *ser_s; 2175 int cnt, off, i; 2176 char *panic_err; 2177 2178 for (cnt = 0, off = 0; cnt < ptr->redir_cnt; cnt++) { 2179 ser_r = (struct cfg_redir *)&buf[off]; 2180 r = malloc(SOF_REDIR, M_IPFW, M_WAITOK | M_ZERO); 2181 memcpy(r, ser_r, SOF_REDIR); 2182 LIST_INIT(&r->spool_chain); 2183 off += SOF_REDIR; 2184 r->alink = malloc(sizeof(struct alias_link *) * r->pport_cnt, 2185 M_IPFW, M_WAITOK | M_ZERO); 2186 switch (r->mode) { 2187 case REDIR_ADDR: 2188 r->alink[0] = LibAliasRedirectAddr(ptr->lib, r->laddr, 2189 r->paddr); 2190 break; 2191 case REDIR_PORT: 2192 for (i = 0 ; i < r->pport_cnt; i++) { 2193 /* If remotePort is all ports, set it to 0. */ 2194 u_short remotePortCopy = r->rport + i; 2195 if (r->rport_cnt == 1 && r->rport == 0) 2196 remotePortCopy = 0; 2197 r->alink[i] = LibAliasRedirectPort(ptr->lib, 2198 r->laddr, htons(r->lport + i), r->raddr, 2199 htons(remotePortCopy), r->paddr, 2200 htons(r->pport + i), r->proto); 2201 if (r->alink[i] == NULL) { 2202 r->alink[0] = NULL; 2203 break; 2204 } 2205 } 2206 break; 2207 case REDIR_PROTO: 2208 r->alink[0] = LibAliasRedirectProto(ptr->lib ,r->laddr, 2209 r->raddr, r->paddr, r->proto); 2210 break; 2211 default: 2212 printf("unknown redirect mode: %u\n", r->mode); 2213 break; 2214 } 2215 if (r->alink[0] == NULL) { 2216 panic_err = "LibAliasRedirect* returned NULL"; 2217 goto bad; 2218 } else /* LSNAT handling. */ 2219 for (i = 0; i < r->spool_cnt; i++) { 2220 ser_s = (struct cfg_spool *)&buf[off]; 2221 s = malloc(SOF_REDIR, M_IPFW, 2222 M_WAITOK | M_ZERO); 2223 memcpy(s, ser_s, SOF_SPOOL); 2224 LibAliasAddServer(ptr->lib, r->alink[0], 2225 s->addr, htons(s->port)); 2226 off += SOF_SPOOL; 2227 /* Hook spool entry. */ 2228 HOOK_SPOOL(&r->spool_chain, s); 2229 } 2230 /* And finally hook this redir entry. */ 2231 HOOK_REDIR(&ptr->redir_chain, r); 2232 } 2233 return (1); 2234bad: 2235 /* something really bad happened: panic! */ 2236 panic("%s\n", panic_err); 2237} 2238#endif 2239 2240/* 2241 * The main check routine for the firewall. 2242 * 2243 * All arguments are in args so we can modify them and return them 2244 * back to the caller. 2245 * 2246 * Parameters: 2247 * 2248 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 2249 * Starts with the IP header. 2250 * args->eh (in) Mac header if present, or NULL for layer3 packet. 2251 * args->L3offset Number of bytes bypassed if we came from L2. 2252 * e.g. often sizeof(eh) ** NOTYET ** 2253 * args->oif Outgoing interface, or NULL if packet is incoming. 2254 * The incoming interface is in the mbuf. (in) 2255 * args->divert_rule (in/out) 2256 * Skip up to the first rule past this rule number; 2257 * upon return, non-zero port number for divert or tee. 2258 * 2259 * args->rule Pointer to the last matching rule (in/out) 2260 * args->next_hop Socket we are forwarding to (out). 2261 * args->f_id Addresses grabbed from the packet (out) 2262 * args->cookie a cookie depending on rule action 2263 * 2264 * Return value: 2265 * 2266 * IP_FW_PASS the packet must be accepted 2267 * IP_FW_DENY the packet must be dropped 2268 * IP_FW_DIVERT divert packet, port in m_tag 2269 * IP_FW_TEE tee packet, port in m_tag 2270 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2271 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2272 * 2273 */ 2274int 2275ipfw_chk(struct ip_fw_args *args) 2276{ 2277 /* 2278 * Local variables holding state during the processing of a packet: 2279 * 2280 * IMPORTANT NOTE: to speed up the processing of rules, there 2281 * are some assumption on the values of the variables, which 2282 * are documented here. Should you change them, please check 2283 * the implementation of the various instructions to make sure 2284 * that they still work. 2285 * 2286 * args->eh The MAC header. It is non-null for a layer2 2287 * packet, it is NULL for a layer-3 packet. 2288 * **notyet** 2289 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. 2290 * 2291 * m | args->m Pointer to the mbuf, as received from the caller. 2292 * It may change if ipfw_chk() does an m_pullup, or if it 2293 * consumes the packet because it calls send_reject(). 2294 * XXX This has to change, so that ipfw_chk() never modifies 2295 * or consumes the buffer. 2296 * ip is the beginning of the ip(4 or 6) header. 2297 * Calculated by adding the L3offset to the start of data. 2298 * (Until we start using L3offset, the packet is 2299 * supposed to start with the ip header). 2300 */ 2301 struct mbuf *m = args->m; 2302 struct ip *ip = mtod(m, struct ip *); 2303 2304 /* 2305 * For rules which contain uid/gid or jail constraints, cache 2306 * a copy of the users credentials after the pcb lookup has been 2307 * executed. This will speed up the processing of rules with 2308 * these types of constraints, as well as decrease contention 2309 * on pcb related locks. 2310 */ 2311 struct ip_fw_ugid fw_ugid_cache; 2312 int ugid_lookup = 0; 2313 2314 /* 2315 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2316 * associated with a packet input on a divert socket. This 2317 * will allow to distinguish traffic and its direction when 2318 * it originates from a divert socket. 2319 */ 2320 u_int divinput_flags = 0; 2321 2322 /* 2323 * oif | args->oif If NULL, ipfw_chk has been called on the 2324 * inbound path (ether_input, ip_input). 2325 * If non-NULL, ipfw_chk has been called on the outbound path 2326 * (ether_output, ip_output). 2327 */ 2328 struct ifnet *oif = args->oif; 2329 2330 struct ip_fw *f = NULL; /* matching rule */ 2331 int retval = 0; 2332 2333 /* 2334 * hlen The length of the IP header. 2335 */ 2336 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2337 2338 /* 2339 * offset The offset of a fragment. offset != 0 means that 2340 * we have a fragment at this offset of an IPv4 packet. 2341 * offset == 0 means that (if this is an IPv4 packet) 2342 * this is the first or only fragment. 2343 * For IPv6 offset == 0 means there is no Fragment Header. 2344 * If offset != 0 for IPv6 always use correct mask to 2345 * get the correct offset because we add IP6F_MORE_FRAG 2346 * to be able to dectect the first fragment which would 2347 * otherwise have offset = 0. 2348 */ 2349 u_short offset = 0; 2350 2351 /* 2352 * Local copies of addresses. They are only valid if we have 2353 * an IP packet. 2354 * 2355 * proto The protocol. Set to 0 for non-ip packets, 2356 * or to the protocol read from the packet otherwise. 2357 * proto != 0 means that we have an IPv4 packet. 2358 * 2359 * src_port, dst_port port numbers, in HOST format. Only 2360 * valid for TCP and UDP packets. 2361 * 2362 * src_ip, dst_ip ip addresses, in NETWORK format. 2363 * Only valid for IPv4 packets. 2364 */ 2365 u_int8_t proto; 2366 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2367 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2368 u_int16_t ip_len=0; 2369 int pktlen; 2370 u_int16_t etype = 0; /* Host order stored ether type */ 2371 2372 /* 2373 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2374 * MATCH_NONE when checked and not matched (q = NULL), 2375 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2376 */ 2377 int dyn_dir = MATCH_UNKNOWN; 2378 ipfw_dyn_rule *q = NULL; 2379 struct ip_fw_chain *chain = &layer3_chain; 2380 struct m_tag *mtag; 2381 2382 /* 2383 * We store in ulp a pointer to the upper layer protocol header. 2384 * In the ipv4 case this is easy to determine from the header, 2385 * but for ipv6 we might have some additional headers in the middle. 2386 * ulp is NULL if not found. 2387 */ 2388 void *ulp = NULL; /* upper layer protocol pointer. */ 2389 /* XXX ipv6 variables */ 2390 int is_ipv6 = 0; 2391 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2392 /* end of ipv6 variables */ 2393 int is_ipv4 = 0; 2394 2395 if (m->m_flags & M_SKIP_FIREWALL) 2396 return (IP_FW_PASS); /* accept */ 2397 2398 pktlen = m->m_pkthdr.len; 2399 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2400 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2401 2402/* 2403 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2404 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2405 * pointer might become stale after other pullups (but we never use it 2406 * this way). 2407 */ 2408#define PULLUP_TO(len, p, T) \ 2409do { \ 2410 int x = (len) + sizeof(T); \ 2411 if ((m)->m_len < x) { \ 2412 args->m = m = m_pullup(m, x); \ 2413 if (m == NULL) \ 2414 goto pullup_failed; \ 2415 } \ 2416 p = (mtod(m, char *) + (len)); \ 2417} while (0) 2418 2419 /* 2420 * if we have an ether header, 2421 */ 2422 if (args->eh) 2423 etype = ntohs(args->eh->ether_type); 2424 2425 /* Identify IP packets and fill up variables. */ 2426 if (pktlen >= sizeof(struct ip6_hdr) && 2427 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { 2428 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 2429 is_ipv6 = 1; 2430 args->f_id.addr_type = 6; 2431 hlen = sizeof(struct ip6_hdr); 2432 proto = ip6->ip6_nxt; 2433 2434 /* Search extension headers to find upper layer protocols */ 2435 while (ulp == NULL) { 2436 switch (proto) { 2437 case IPPROTO_ICMPV6: 2438 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2439 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2440 break; 2441 2442 case IPPROTO_TCP: 2443 PULLUP_TO(hlen, ulp, struct tcphdr); 2444 dst_port = TCP(ulp)->th_dport; 2445 src_port = TCP(ulp)->th_sport; 2446 args->f_id.flags = TCP(ulp)->th_flags; 2447 break; 2448 2449 case IPPROTO_SCTP: 2450 PULLUP_TO(hlen, ulp, struct sctphdr); 2451 src_port = SCTP(ulp)->src_port; 2452 dst_port = SCTP(ulp)->dest_port; 2453 break; 2454 2455 case IPPROTO_UDP: 2456 PULLUP_TO(hlen, ulp, struct udphdr); 2457 dst_port = UDP(ulp)->uh_dport; 2458 src_port = UDP(ulp)->uh_sport; 2459 break; 2460 2461 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2462 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2463 ext_hd |= EXT_HOPOPTS; 2464 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2465 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2466 ulp = NULL; 2467 break; 2468 2469 case IPPROTO_ROUTING: /* RFC 2460 */ 2470 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2471 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2472 case 0: 2473 ext_hd |= EXT_RTHDR0; 2474 break; 2475 case 2: 2476 ext_hd |= EXT_RTHDR2; 2477 break; 2478 default: 2479 printf("IPFW2: IPV6 - Unknown Routing " 2480 "Header type(%d)\n", 2481 ((struct ip6_rthdr *)ulp)->ip6r_type); 2482 if (fw_deny_unknown_exthdrs) 2483 return (IP_FW_DENY); 2484 break; 2485 } 2486 ext_hd |= EXT_ROUTING; 2487 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2488 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2489 ulp = NULL; 2490 break; 2491 2492 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2493 PULLUP_TO(hlen, ulp, struct ip6_frag); 2494 ext_hd |= EXT_FRAGMENT; 2495 hlen += sizeof (struct ip6_frag); 2496 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2497 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2498 IP6F_OFF_MASK; 2499 /* Add IP6F_MORE_FRAG for offset of first 2500 * fragment to be != 0. */ 2501 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2502 IP6F_MORE_FRAG; 2503 if (offset == 0) { 2504 printf("IPFW2: IPV6 - Invalid Fragment " 2505 "Header\n"); 2506 if (fw_deny_unknown_exthdrs) 2507 return (IP_FW_DENY); 2508 break; 2509 } 2510 args->f_id.frag_id6 = 2511 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2512 ulp = NULL; 2513 break; 2514 2515 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2516 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2517 ext_hd |= EXT_DSTOPTS; 2518 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2519 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2520 ulp = NULL; 2521 break; 2522 2523 case IPPROTO_AH: /* RFC 2402 */ 2524 PULLUP_TO(hlen, ulp, struct ip6_ext); 2525 ext_hd |= EXT_AH; 2526 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2527 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2528 ulp = NULL; 2529 break; 2530 2531 case IPPROTO_ESP: /* RFC 2406 */ 2532 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2533 /* Anything past Seq# is variable length and 2534 * data past this ext. header is encrypted. */ 2535 ext_hd |= EXT_ESP; 2536 break; 2537 2538 case IPPROTO_NONE: /* RFC 2460 */ 2539 PULLUP_TO(hlen, ulp, struct ip6_ext); 2540 /* Packet ends here. if ip6e_len!=0 octets 2541 * must be ignored. */ 2542 break; 2543 2544 case IPPROTO_OSPFIGP: 2545 /* XXX OSPF header check? */ 2546 PULLUP_TO(hlen, ulp, struct ip6_ext); 2547 break; 2548 2549 case IPPROTO_PIM: 2550 /* XXX PIM header check? */ 2551 PULLUP_TO(hlen, ulp, struct pim); 2552 break; 2553 2554 case IPPROTO_CARP: 2555 PULLUP_TO(hlen, ulp, struct carp_header); 2556 if (((struct carp_header *)ulp)->carp_version != 2557 CARP_VERSION) 2558 return (IP_FW_DENY); 2559 if (((struct carp_header *)ulp)->carp_type != 2560 CARP_ADVERTISEMENT) 2561 return (IP_FW_DENY); 2562 break; 2563 2564 case IPPROTO_IPV6: /* RFC 2893 */ 2565 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2566 break; 2567 2568 case IPPROTO_IPV4: /* RFC 2893 */ 2569 PULLUP_TO(hlen, ulp, struct ip); 2570 break; 2571 2572 default: 2573 printf("IPFW2: IPV6 - Unknown Extension " 2574 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2575 if (fw_deny_unknown_exthdrs) 2576 return (IP_FW_DENY); 2577 PULLUP_TO(hlen, ulp, struct ip6_ext); 2578 break; 2579 } /*switch */ 2580 } 2581 ip = mtod(m, struct ip *); 2582 ip6 = (struct ip6_hdr *)ip; 2583 args->f_id.src_ip6 = ip6->ip6_src; 2584 args->f_id.dst_ip6 = ip6->ip6_dst; 2585 args->f_id.src_ip = 0; 2586 args->f_id.dst_ip = 0; 2587 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 2588 } else if (pktlen >= sizeof(struct ip) && 2589 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { 2590 is_ipv4 = 1; 2591 hlen = ip->ip_hl << 2; 2592 args->f_id.addr_type = 4; 2593 2594 /* 2595 * Collect parameters into local variables for faster matching. 2596 */ 2597 proto = ip->ip_p; 2598 src_ip = ip->ip_src; 2599 dst_ip = ip->ip_dst; 2600 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2601 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2602 ip_len = ntohs(ip->ip_len); 2603 } else { 2604 offset = ip->ip_off & IP_OFFMASK; 2605 ip_len = ip->ip_len; 2606 } 2607 pktlen = ip_len < pktlen ? ip_len : pktlen; 2608 2609 if (offset == 0) { 2610 switch (proto) { 2611 case IPPROTO_TCP: 2612 PULLUP_TO(hlen, ulp, struct tcphdr); 2613 dst_port = TCP(ulp)->th_dport; 2614 src_port = TCP(ulp)->th_sport; 2615 args->f_id.flags = TCP(ulp)->th_flags; 2616 break; 2617 2618 case IPPROTO_UDP: 2619 PULLUP_TO(hlen, ulp, struct udphdr); 2620 dst_port = UDP(ulp)->uh_dport; 2621 src_port = UDP(ulp)->uh_sport; 2622 break; 2623 2624 case IPPROTO_ICMP: 2625 PULLUP_TO(hlen, ulp, struct icmphdr); 2626 args->f_id.flags = ICMP(ulp)->icmp_type; 2627 break; 2628 2629 default: 2630 break; 2631 } 2632 } 2633 2634 ip = mtod(m, struct ip *); 2635 args->f_id.src_ip = ntohl(src_ip.s_addr); 2636 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2637 } 2638#undef PULLUP_TO 2639 if (proto) { /* we may have port numbers, store them */ 2640 args->f_id.proto = proto; 2641 args->f_id.src_port = src_port = ntohs(src_port); 2642 args->f_id.dst_port = dst_port = ntohs(dst_port); 2643 } 2644 2645 IPFW_RLOCK(chain); 2646 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2647 if (args->rule) { 2648 /* 2649 * Packet has already been tagged. Look for the next rule 2650 * to restart processing. 2651 * 2652 * If fw_one_pass != 0 then just accept it. 2653 * XXX should not happen here, but optimized out in 2654 * the caller. 2655 */ 2656 if (fw_one_pass) { 2657 IPFW_RUNLOCK(chain); 2658 return (IP_FW_PASS); 2659 } 2660 2661 f = args->rule->next_rule; 2662 if (f == NULL) 2663 f = lookup_next_rule(args->rule); 2664 } else { 2665 /* 2666 * Find the starting rule. It can be either the first 2667 * one, or the one after divert_rule if asked so. 2668 */ 2669 int skipto = mtag ? divert_cookie(mtag) : 0; 2670 2671 f = chain->rules; 2672 if (args->eh == NULL && skipto != 0) { 2673 if (skipto >= IPFW_DEFAULT_RULE) { 2674 IPFW_RUNLOCK(chain); 2675 return (IP_FW_DENY); /* invalid */ 2676 } 2677 while (f && f->rulenum <= skipto) 2678 f = f->next; 2679 if (f == NULL) { /* drop packet */ 2680 IPFW_RUNLOCK(chain); 2681 return (IP_FW_DENY); 2682 } 2683 } 2684 } 2685 /* reset divert rule to avoid confusion later */ 2686 if (mtag) { 2687 divinput_flags = divert_info(mtag) & 2688 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2689 m_tag_delete(m, mtag); 2690 } 2691 2692 /* 2693 * Now scan the rules, and parse microinstructions for each rule. 2694 */ 2695 for (; f; f = f->next) { 2696 ipfw_insn *cmd; 2697 uint32_t tablearg = 0; 2698 int l, cmdlen, skip_or; /* skip rest of OR block */ 2699 2700again: 2701 if (set_disable & (1 << f->set) ) 2702 continue; 2703 2704 skip_or = 0; 2705 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2706 l -= cmdlen, cmd += cmdlen) { 2707 int match; 2708 2709 /* 2710 * check_body is a jump target used when we find a 2711 * CHECK_STATE, and need to jump to the body of 2712 * the target rule. 2713 */ 2714 2715check_body: 2716 cmdlen = F_LEN(cmd); 2717 /* 2718 * An OR block (insn_1 || .. || insn_n) has the 2719 * F_OR bit set in all but the last instruction. 2720 * The first match will set "skip_or", and cause 2721 * the following instructions to be skipped until 2722 * past the one with the F_OR bit clear. 2723 */ 2724 if (skip_or) { /* skip this instruction */ 2725 if ((cmd->len & F_OR) == 0) 2726 skip_or = 0; /* next one is good */ 2727 continue; 2728 } 2729 match = 0; /* set to 1 if we succeed */ 2730 2731 switch (cmd->opcode) { 2732 /* 2733 * The first set of opcodes compares the packet's 2734 * fields with some pattern, setting 'match' if a 2735 * match is found. At the end of the loop there is 2736 * logic to deal with F_NOT and F_OR flags associated 2737 * with the opcode. 2738 */ 2739 case O_NOP: 2740 match = 1; 2741 break; 2742 2743 case O_FORWARD_MAC: 2744 printf("ipfw: opcode %d unimplemented\n", 2745 cmd->opcode); 2746 break; 2747 2748 case O_GID: 2749 case O_UID: 2750 case O_JAIL: 2751 /* 2752 * We only check offset == 0 && proto != 0, 2753 * as this ensures that we have a 2754 * packet with the ports info. 2755 */ 2756 if (offset!=0) 2757 break; 2758 if (is_ipv6) /* XXX to be fixed later */ 2759 break; 2760 if (proto == IPPROTO_TCP || 2761 proto == IPPROTO_UDP) 2762 match = check_uidgid( 2763 (ipfw_insn_u32 *)cmd, 2764 proto, oif, 2765 dst_ip, dst_port, 2766 src_ip, src_port, &fw_ugid_cache, 2767 &ugid_lookup, args->inp); 2768 break; 2769 2770 case O_RECV: 2771 match = iface_match(m->m_pkthdr.rcvif, 2772 (ipfw_insn_if *)cmd); 2773 break; 2774 2775 case O_XMIT: 2776 match = iface_match(oif, (ipfw_insn_if *)cmd); 2777 break; 2778 2779 case O_VIA: 2780 match = iface_match(oif ? oif : 2781 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2782 break; 2783 2784 case O_MACADDR2: 2785 if (args->eh != NULL) { /* have MAC header */ 2786 u_int32_t *want = (u_int32_t *) 2787 ((ipfw_insn_mac *)cmd)->addr; 2788 u_int32_t *mask = (u_int32_t *) 2789 ((ipfw_insn_mac *)cmd)->mask; 2790 u_int32_t *hdr = (u_int32_t *)args->eh; 2791 2792 match = 2793 ( want[0] == (hdr[0] & mask[0]) && 2794 want[1] == (hdr[1] & mask[1]) && 2795 want[2] == (hdr[2] & mask[2]) ); 2796 } 2797 break; 2798 2799 case O_MAC_TYPE: 2800 if (args->eh != NULL) { 2801 u_int16_t *p = 2802 ((ipfw_insn_u16 *)cmd)->ports; 2803 int i; 2804 2805 for (i = cmdlen - 1; !match && i>0; 2806 i--, p += 2) 2807 match = (etype >= p[0] && 2808 etype <= p[1]); 2809 } 2810 break; 2811 2812 case O_FRAG: 2813 match = (offset != 0); 2814 break; 2815 2816 case O_IN: /* "out" is "not in" */ 2817 match = (oif == NULL); 2818 break; 2819 2820 case O_LAYER2: 2821 match = (args->eh != NULL); 2822 break; 2823 2824 case O_DIVERTED: 2825 match = (cmd->arg1 & 1 && divinput_flags & 2826 IP_FW_DIVERT_LOOPBACK_FLAG) || 2827 (cmd->arg1 & 2 && divinput_flags & 2828 IP_FW_DIVERT_OUTPUT_FLAG); 2829 break; 2830 2831 case O_PROTO: 2832 /* 2833 * We do not allow an arg of 0 so the 2834 * check of "proto" only suffices. 2835 */ 2836 match = (proto == cmd->arg1); 2837 break; 2838 2839 case O_IP_SRC: 2840 match = is_ipv4 && 2841 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2842 src_ip.s_addr); 2843 break; 2844 2845 case O_IP_SRC_LOOKUP: 2846 case O_IP_DST_LOOKUP: 2847 if (is_ipv4) { 2848 uint32_t a = 2849 (cmd->opcode == O_IP_DST_LOOKUP) ? 2850 dst_ip.s_addr : src_ip.s_addr; 2851 uint32_t v; 2852 2853 match = lookup_table(chain, cmd->arg1, a, 2854 &v); 2855 if (!match) 2856 break; 2857 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2858 match = 2859 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2860 else 2861 tablearg = v; 2862 } 2863 break; 2864 2865 case O_IP_SRC_MASK: 2866 case O_IP_DST_MASK: 2867 if (is_ipv4) { 2868 uint32_t a = 2869 (cmd->opcode == O_IP_DST_MASK) ? 2870 dst_ip.s_addr : src_ip.s_addr; 2871 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2872 int i = cmdlen-1; 2873 2874 for (; !match && i>0; i-= 2, p+= 2) 2875 match = (p[0] == (a & p[1])); 2876 } 2877 break; 2878 2879 case O_IP_SRC_ME: 2880 if (is_ipv4) { 2881 struct ifnet *tif; 2882 2883 INADDR_TO_IFP(src_ip, tif); 2884 match = (tif != NULL); 2885 } 2886 break; 2887 2888 case O_IP_DST_SET: 2889 case O_IP_SRC_SET: 2890 if (is_ipv4) { 2891 u_int32_t *d = (u_int32_t *)(cmd+1); 2892 u_int32_t addr = 2893 cmd->opcode == O_IP_DST_SET ? 2894 args->f_id.dst_ip : 2895 args->f_id.src_ip; 2896 2897 if (addr < d[0]) 2898 break; 2899 addr -= d[0]; /* subtract base */ 2900 match = (addr < cmd->arg1) && 2901 ( d[ 1 + (addr>>5)] & 2902 (1<<(addr & 0x1f)) ); 2903 } 2904 break; 2905 2906 case O_IP_DST: 2907 match = is_ipv4 && 2908 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2909 dst_ip.s_addr); 2910 break; 2911 2912 case O_IP_DST_ME: 2913 if (is_ipv4) { 2914 struct ifnet *tif; 2915 2916 INADDR_TO_IFP(dst_ip, tif); 2917 match = (tif != NULL); 2918 } 2919 break; 2920 2921 case O_IP_SRCPORT: 2922 case O_IP_DSTPORT: 2923 /* 2924 * offset == 0 && proto != 0 is enough 2925 * to guarantee that we have a 2926 * packet with port info. 2927 */ 2928 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2929 && offset == 0) { 2930 u_int16_t x = 2931 (cmd->opcode == O_IP_SRCPORT) ? 2932 src_port : dst_port ; 2933 u_int16_t *p = 2934 ((ipfw_insn_u16 *)cmd)->ports; 2935 int i; 2936 2937 for (i = cmdlen - 1; !match && i>0; 2938 i--, p += 2) 2939 match = (x>=p[0] && x<=p[1]); 2940 } 2941 break; 2942 2943 case O_ICMPTYPE: 2944 match = (offset == 0 && proto==IPPROTO_ICMP && 2945 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2946 break; 2947 2948#ifdef INET6 2949 case O_ICMP6TYPE: 2950 match = is_ipv6 && offset == 0 && 2951 proto==IPPROTO_ICMPV6 && 2952 icmp6type_match( 2953 ICMP6(ulp)->icmp6_type, 2954 (ipfw_insn_u32 *)cmd); 2955 break; 2956#endif /* INET6 */ 2957 2958 case O_IPOPT: 2959 match = (is_ipv4 && 2960 ipopts_match(ip, cmd) ); 2961 break; 2962 2963 case O_IPVER: 2964 match = (is_ipv4 && 2965 cmd->arg1 == ip->ip_v); 2966 break; 2967 2968 case O_IPID: 2969 case O_IPLEN: 2970 case O_IPTTL: 2971 if (is_ipv4) { /* only for IP packets */ 2972 uint16_t x; 2973 uint16_t *p; 2974 int i; 2975 2976 if (cmd->opcode == O_IPLEN) 2977 x = ip_len; 2978 else if (cmd->opcode == O_IPTTL) 2979 x = ip->ip_ttl; 2980 else /* must be IPID */ 2981 x = ntohs(ip->ip_id); 2982 if (cmdlen == 1) { 2983 match = (cmd->arg1 == x); 2984 break; 2985 } 2986 /* otherwise we have ranges */ 2987 p = ((ipfw_insn_u16 *)cmd)->ports; 2988 i = cmdlen - 1; 2989 for (; !match && i>0; i--, p += 2) 2990 match = (x >= p[0] && x <= p[1]); 2991 } 2992 break; 2993 2994 case O_IPPRECEDENCE: 2995 match = (is_ipv4 && 2996 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2997 break; 2998 2999 case O_IPTOS: 3000 match = (is_ipv4 && 3001 flags_match(cmd, ip->ip_tos)); 3002 break; 3003 3004 case O_TCPDATALEN: 3005 if (proto == IPPROTO_TCP && offset == 0) { 3006 struct tcphdr *tcp; 3007 uint16_t x; 3008 uint16_t *p; 3009 int i; 3010 3011 tcp = TCP(ulp); 3012 x = ip_len - 3013 ((ip->ip_hl + tcp->th_off) << 2); 3014 if (cmdlen == 1) { 3015 match = (cmd->arg1 == x); 3016 break; 3017 } 3018 /* otherwise we have ranges */ 3019 p = ((ipfw_insn_u16 *)cmd)->ports; 3020 i = cmdlen - 1; 3021 for (; !match && i>0; i--, p += 2) 3022 match = (x >= p[0] && x <= p[1]); 3023 } 3024 break; 3025 3026 case O_TCPFLAGS: 3027 match = (proto == IPPROTO_TCP && offset == 0 && 3028 flags_match(cmd, TCP(ulp)->th_flags)); 3029 break; 3030 3031 case O_TCPOPTS: 3032 match = (proto == IPPROTO_TCP && offset == 0 && 3033 tcpopts_match(TCP(ulp), cmd)); 3034 break; 3035 3036 case O_TCPSEQ: 3037 match = (proto == IPPROTO_TCP && offset == 0 && 3038 ((ipfw_insn_u32 *)cmd)->d[0] == 3039 TCP(ulp)->th_seq); 3040 break; 3041 3042 case O_TCPACK: 3043 match = (proto == IPPROTO_TCP && offset == 0 && 3044 ((ipfw_insn_u32 *)cmd)->d[0] == 3045 TCP(ulp)->th_ack); 3046 break; 3047 3048 case O_TCPWIN: 3049 match = (proto == IPPROTO_TCP && offset == 0 && 3050 cmd->arg1 == TCP(ulp)->th_win); 3051 break; 3052 3053 case O_ESTAB: 3054 /* reject packets which have SYN only */ 3055 /* XXX should i also check for TH_ACK ? */ 3056 match = (proto == IPPROTO_TCP && offset == 0 && 3057 (TCP(ulp)->th_flags & 3058 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 3059 break; 3060 3061 case O_ALTQ: { 3062 struct pf_mtag *at; 3063 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 3064 3065 match = 1; 3066 at = pf_find_mtag(m); 3067 if (at != NULL && at->qid != 0) 3068 break; 3069 at = pf_get_mtag(m); 3070 if (at == NULL) { 3071 /* 3072 * Let the packet fall back to the 3073 * default ALTQ. 3074 */ 3075 break; 3076 } 3077 at->qid = altq->qid; 3078 if (is_ipv4) 3079 at->af = AF_INET; 3080 else 3081 at->af = AF_LINK; 3082 at->hdr = ip; 3083 m_tag_prepend(m, mtag); 3084 break; 3085 } 3086 3087 case O_LOG: 3088 if (fw_verbose) 3089 ipfw_log(f, hlen, args, m, 3090 oif, offset, tablearg, ip); 3091 match = 1; 3092 break; 3093 3094 case O_PROB: 3095 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 3096 break; 3097 3098 case O_VERREVPATH: 3099 /* Outgoing packets automatically pass/match */ 3100 match = ((oif != NULL) || 3101 (m->m_pkthdr.rcvif == NULL) || 3102 ( 3103#ifdef INET6 3104 is_ipv6 ? 3105 verify_path6(&(args->f_id.src_ip6), 3106 m->m_pkthdr.rcvif) : 3107#endif 3108 verify_path(src_ip, m->m_pkthdr.rcvif))); 3109 break; 3110 3111 case O_VERSRCREACH: 3112 /* Outgoing packets automatically pass/match */ 3113 match = (hlen > 0 && ((oif != NULL) || 3114#ifdef INET6 3115 is_ipv6 ? 3116 verify_path6(&(args->f_id.src_ip6), 3117 NULL) : 3118#endif 3119 verify_path(src_ip, NULL))); 3120 break; 3121 3122 case O_ANTISPOOF: 3123 /* Outgoing packets automatically pass/match */ 3124 if (oif == NULL && hlen > 0 && 3125 ( (is_ipv4 && in_localaddr(src_ip)) 3126#ifdef INET6 3127 || (is_ipv6 && 3128 in6_localaddr(&(args->f_id.src_ip6))) 3129#endif 3130 )) 3131 match = 3132#ifdef INET6 3133 is_ipv6 ? verify_path6( 3134 &(args->f_id.src_ip6), 3135 m->m_pkthdr.rcvif) : 3136#endif 3137 verify_path(src_ip, 3138 m->m_pkthdr.rcvif); 3139 else 3140 match = 1; 3141 break; 3142 3143 case O_IPSEC: 3144#ifdef IPSEC 3145 match = (m_tag_find(m, 3146 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 3147#endif 3148 /* otherwise no match */ 3149 break; 3150 3151#ifdef INET6 3152 case O_IP6_SRC: 3153 match = is_ipv6 && 3154 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 3155 &((ipfw_insn_ip6 *)cmd)->addr6); 3156 break; 3157 3158 case O_IP6_DST: 3159 match = is_ipv6 && 3160 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 3161 &((ipfw_insn_ip6 *)cmd)->addr6); 3162 break; 3163 case O_IP6_SRC_MASK: 3164 case O_IP6_DST_MASK: 3165 if (is_ipv6) { 3166 int i = cmdlen - 1; 3167 struct in6_addr p; 3168 struct in6_addr *d = 3169 &((ipfw_insn_ip6 *)cmd)->addr6; 3170 3171 for (; !match && i > 0; d += 2, 3172 i -= F_INSN_SIZE(struct in6_addr) 3173 * 2) { 3174 p = (cmd->opcode == 3175 O_IP6_SRC_MASK) ? 3176 args->f_id.src_ip6: 3177 args->f_id.dst_ip6; 3178 APPLY_MASK(&p, &d[1]); 3179 match = 3180 IN6_ARE_ADDR_EQUAL(&d[0], 3181 &p); 3182 } 3183 } 3184 break; 3185 3186 case O_IP6_SRC_ME: 3187 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 3188 break; 3189 3190 case O_IP6_DST_ME: 3191 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 3192 break; 3193 3194 case O_FLOW6ID: 3195 match = is_ipv6 && 3196 flow6id_match(args->f_id.flow_id6, 3197 (ipfw_insn_u32 *) cmd); 3198 break; 3199 3200 case O_EXT_HDR: 3201 match = is_ipv6 && 3202 (ext_hd & ((ipfw_insn *) cmd)->arg1); 3203 break; 3204 3205 case O_IP6: 3206 match = is_ipv6; 3207 break; 3208#endif 3209 3210 case O_IP4: 3211 match = is_ipv4; 3212 break; 3213 3214 case O_TAG: { 3215 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3216 tablearg : cmd->arg1; 3217 3218 /* Packet is already tagged with this tag? */ 3219 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 3220 3221 /* We have `untag' action when F_NOT flag is 3222 * present. And we must remove this mtag from 3223 * mbuf and reset `match' to zero (`match' will 3224 * be inversed later). 3225 * Otherwise we should allocate new mtag and 3226 * push it into mbuf. 3227 */ 3228 if (cmd->len & F_NOT) { /* `untag' action */ 3229 if (mtag != NULL) 3230 m_tag_delete(m, mtag); 3231 } else if (mtag == NULL) { 3232 if ((mtag = m_tag_alloc(MTAG_IPFW, 3233 tag, 0, M_NOWAIT)) != NULL) 3234 m_tag_prepend(m, mtag); 3235 } 3236 match = (cmd->len & F_NOT) ? 0: 1; 3237 break; 3238 } 3239 3240 case O_TAGGED: { 3241 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3242 tablearg : cmd->arg1; 3243 3244 if (cmdlen == 1) { 3245 match = m_tag_locate(m, MTAG_IPFW, 3246 tag, NULL) != NULL; 3247 break; 3248 } 3249 3250 /* we have ranges */ 3251 for (mtag = m_tag_first(m); 3252 mtag != NULL && !match; 3253 mtag = m_tag_next(m, mtag)) { 3254 uint16_t *p; 3255 int i; 3256 3257 if (mtag->m_tag_cookie != MTAG_IPFW) 3258 continue; 3259 3260 p = ((ipfw_insn_u16 *)cmd)->ports; 3261 i = cmdlen - 1; 3262 for(; !match && i > 0; i--, p += 2) 3263 match = 3264 mtag->m_tag_id >= p[0] && 3265 mtag->m_tag_id <= p[1]; 3266 } 3267 break; 3268 } 3269 3270 /* 3271 * The second set of opcodes represents 'actions', 3272 * i.e. the terminal part of a rule once the packet 3273 * matches all previous patterns. 3274 * Typically there is only one action for each rule, 3275 * and the opcode is stored at the end of the rule 3276 * (but there are exceptions -- see below). 3277 * 3278 * In general, here we set retval and terminate the 3279 * outer loop (would be a 'break 3' in some language, 3280 * but we need to do a 'goto done'). 3281 * 3282 * Exceptions: 3283 * O_COUNT and O_SKIPTO actions: 3284 * instead of terminating, we jump to the next rule 3285 * ('goto next_rule', equivalent to a 'break 2'), 3286 * or to the SKIPTO target ('goto again' after 3287 * having set f, cmd and l), respectively. 3288 * 3289 * O_TAG, O_LOG and O_ALTQ action parameters: 3290 * perform some action and set match = 1; 3291 * 3292 * O_LIMIT and O_KEEP_STATE: these opcodes are 3293 * not real 'actions', and are stored right 3294 * before the 'action' part of the rule. 3295 * These opcodes try to install an entry in the 3296 * state tables; if successful, we continue with 3297 * the next opcode (match=1; break;), otherwise 3298 * the packet * must be dropped 3299 * ('goto done' after setting retval); 3300 * 3301 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3302 * cause a lookup of the state table, and a jump 3303 * to the 'action' part of the parent rule 3304 * ('goto check_body') if an entry is found, or 3305 * (CHECK_STATE only) a jump to the next rule if 3306 * the entry is not found ('goto next_rule'). 3307 * The result of the lookup is cached to make 3308 * further instances of these opcodes are 3309 * effectively NOPs. 3310 */ 3311 case O_LIMIT: 3312 case O_KEEP_STATE: 3313 if (install_state(f, 3314 (ipfw_insn_limit *)cmd, args, tablearg)) { 3315 retval = IP_FW_DENY; 3316 goto done; /* error/limit violation */ 3317 } 3318 match = 1; 3319 break; 3320 3321 case O_PROBE_STATE: 3322 case O_CHECK_STATE: 3323 /* 3324 * dynamic rules are checked at the first 3325 * keep-state or check-state occurrence, 3326 * with the result being stored in dyn_dir. 3327 * The compiler introduces a PROBE_STATE 3328 * instruction for us when we have a 3329 * KEEP_STATE (because PROBE_STATE needs 3330 * to be run first). 3331 */ 3332 if (dyn_dir == MATCH_UNKNOWN && 3333 (q = lookup_dyn_rule(&args->f_id, 3334 &dyn_dir, proto == IPPROTO_TCP ? 3335 TCP(ulp) : NULL)) 3336 != NULL) { 3337 /* 3338 * Found dynamic entry, update stats 3339 * and jump to the 'action' part of 3340 * the parent rule. 3341 */ 3342 q->pcnt++; 3343 q->bcnt += pktlen; 3344 f = q->rule; 3345 cmd = ACTION_PTR(f); 3346 l = f->cmd_len - f->act_ofs; 3347 IPFW_DYN_UNLOCK(); 3348 goto check_body; 3349 } 3350 /* 3351 * Dynamic entry not found. If CHECK_STATE, 3352 * skip to next rule, if PROBE_STATE just 3353 * ignore and continue with next opcode. 3354 */ 3355 if (cmd->opcode == O_CHECK_STATE) 3356 goto next_rule; 3357 match = 1; 3358 break; 3359 3360 case O_ACCEPT: 3361 retval = 0; /* accept */ 3362 goto done; 3363 3364 case O_PIPE: 3365 case O_QUEUE: 3366 args->rule = f; /* report matching rule */ 3367 if (cmd->arg1 == IP_FW_TABLEARG) 3368 args->cookie = tablearg; 3369 else 3370 args->cookie = cmd->arg1; 3371 retval = IP_FW_DUMMYNET; 3372 goto done; 3373 3374 case O_DIVERT: 3375 case O_TEE: { 3376 struct divert_tag *dt; 3377 3378 if (args->eh) /* not on layer 2 */ 3379 break; 3380 mtag = m_tag_get(PACKET_TAG_DIVERT, 3381 sizeof(struct divert_tag), 3382 M_NOWAIT); 3383 if (mtag == NULL) { 3384 /* XXX statistic */ 3385 /* drop packet */ 3386 IPFW_RUNLOCK(chain); 3387 return (IP_FW_DENY); 3388 } 3389 dt = (struct divert_tag *)(mtag+1); 3390 dt->cookie = f->rulenum; 3391 if (cmd->arg1 == IP_FW_TABLEARG) 3392 dt->info = tablearg; 3393 else 3394 dt->info = cmd->arg1; 3395 m_tag_prepend(m, mtag); 3396 retval = (cmd->opcode == O_DIVERT) ? 3397 IP_FW_DIVERT : IP_FW_TEE; 3398 goto done; 3399 } 3400 3401 case O_COUNT: 3402 case O_SKIPTO: 3403 f->pcnt++; /* update stats */ 3404 f->bcnt += pktlen; 3405 f->timestamp = time_uptime; 3406 if (cmd->opcode == O_COUNT) 3407 goto next_rule; 3408 /* handle skipto */ 3409 if (f->next_rule == NULL) 3410 lookup_next_rule(f); 3411 f = f->next_rule; 3412 goto again; 3413 3414 case O_REJECT: 3415 /* 3416 * Drop the packet and send a reject notice 3417 * if the packet is not ICMP (or is an ICMP 3418 * query), and it is not multicast/broadcast. 3419 */ 3420 if (hlen > 0 && is_ipv4 && offset == 0 && 3421 (proto != IPPROTO_ICMP || 3422 is_icmp_query(ICMP(ulp))) && 3423 !(m->m_flags & (M_BCAST|M_MCAST)) && 3424 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3425 send_reject(args, cmd->arg1, ip_len, ip); 3426 m = args->m; 3427 } 3428 /* FALLTHROUGH */ 3429#ifdef INET6 3430 case O_UNREACH6: 3431 if (hlen > 0 && is_ipv6 && 3432 ((offset & IP6F_OFF_MASK) == 0) && 3433 (proto != IPPROTO_ICMPV6 || 3434 (is_icmp6_query(args->f_id.flags) == 1)) && 3435 !(m->m_flags & (M_BCAST|M_MCAST)) && 3436 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3437 send_reject6( 3438 args, cmd->arg1, hlen, 3439 (struct ip6_hdr *)ip); 3440 m = args->m; 3441 } 3442 /* FALLTHROUGH */ 3443#endif 3444 case O_DENY: 3445 retval = IP_FW_DENY; 3446 goto done; 3447 3448 case O_FORWARD_IP: { 3449 struct sockaddr_in *sa; 3450 sa = &(((ipfw_insn_sa *)cmd)->sa); 3451 if (args->eh) /* not valid on layer2 pkts */ 3452 break; 3453 if (!q || dyn_dir == MATCH_FORWARD) { 3454 if (sa->sin_addr.s_addr == INADDR_ANY) { 3455 bcopy(sa, &args->hopstore, 3456 sizeof(*sa)); 3457 args->hopstore.sin_addr.s_addr = 3458 htonl(tablearg); 3459 args->next_hop = 3460 &args->hopstore; 3461 } else { 3462 args->next_hop = sa; 3463 } 3464 } 3465 retval = IP_FW_PASS; 3466 } 3467 goto done; 3468 3469 case O_NETGRAPH: 3470 case O_NGTEE: 3471 args->rule = f; /* report matching rule */ 3472 if (cmd->arg1 == IP_FW_TABLEARG) 3473 args->cookie = tablearg; 3474 else 3475 args->cookie = cmd->arg1; 3476 retval = (cmd->opcode == O_NETGRAPH) ? 3477 IP_FW_NETGRAPH : IP_FW_NGTEE; 3478 goto done; 3479 3480#ifdef IPFIREWALL_NAT 3481 case O_NAT: { 3482 struct cfg_nat *t; 3483 struct mbuf *mcl; 3484 /* XXX - libalias duct tape */ 3485 int ldt; 3486 char *c; 3487 3488 ldt = 0; 3489 args->rule = f; /* Report matching rule. */ 3490 retval = 0; 3491 t = ((ipfw_insn_nat *)cmd)->nat; 3492 if (t == NULL) { 3493 t = lookup_nat(cmd->arg1); 3494 if (t == NULL) { 3495 retval = IP_FW_DENY; 3496 goto done; 3497 } else 3498 ((ipfw_insn_nat *)cmd)->nat = 3499 t; 3500 } 3501 if ((mcl = m_megapullup(m, m->m_pkthdr.len)) == 3502 NULL) 3503 goto badnat; 3504 ip = mtod(mcl, struct ip *); 3505 if (args->eh == NULL) { 3506 ip->ip_len = htons(ip->ip_len); 3507 ip->ip_off = htons(ip->ip_off); 3508 } 3509 3510 /* 3511 * XXX - Libalias checksum offload 'duct tape': 3512 * 3513 * locally generated packets have only 3514 * pseudo-header checksum calculated 3515 * and libalias will screw it[1], so 3516 * mark them for later fix. Moreover 3517 * there are cases when libalias 3518 * modify tcp packet data[2], mark it 3519 * for later fix too. 3520 * 3521 * [1] libalias was never meant to run 3522 * in kernel, so it doesn't have any 3523 * knowledge about checksum 3524 * offloading, and it expects a packet 3525 * with a full internet 3526 * checksum. Unfortunately, packets 3527 * generated locally will have just the 3528 * pseudo header calculated, and when 3529 * libalias tries to adjust the 3530 * checksum it will actually screw it. 3531 * 3532 * [2] when libalias modify tcp's data 3533 * content, full TCP checksum has to 3534 * be recomputed: the problem is that 3535 * libalias doesn't have any idea 3536 * about checksum offloading To 3537 * workaround this, we do not do 3538 * checksumming in LibAlias, but only 3539 * mark the packets in th_x2 field. If 3540 * we receive a marked packet, we 3541 * calculate correct checksum for it 3542 * aware of offloading. Why such a 3543 * terrible hack instead of 3544 * recalculating checksum for each 3545 * packet? Because the previous 3546 * checksum was not checked! 3547 * Recalculating checksums for EVERY 3548 * packet will hide ALL transmission 3549 * errors. Yes, marked packets still 3550 * suffer from this problem. But, 3551 * sigh, natd(8) has this problem, 3552 * too. 3553 * 3554 * TODO: -make libalias mbuf aware (so 3555 * it can handle delayed checksum and tso) 3556 */ 3557 3558 if (mcl->m_pkthdr.rcvif == NULL && 3559 mcl->m_pkthdr.csum_flags & 3560 CSUM_DELAY_DATA) 3561 ldt = 1; 3562 3563 c = mtod(mcl, char *); 3564 if (oif == NULL) 3565 retval = LibAliasIn(t->lib, c, 3566 MCLBYTES); 3567 else 3568 retval = LibAliasOut(t->lib, c, 3569 MCLBYTES); 3570 if (retval != PKT_ALIAS_OK) { 3571 /* XXX - should i add some logging? */ 3572 m_free(mcl); 3573 badnat: 3574 args->m = NULL; 3575 retval = IP_FW_DENY; 3576 goto done; 3577 } 3578 mcl->m_pkthdr.len = mcl->m_len = 3579 ntohs(ip->ip_len); 3580 3581 /* 3582 * XXX - libalias checksum offload 3583 * 'duct tape' (see above) 3584 */ 3585 3586 if ((ip->ip_off & htons(IP_OFFMASK)) == 0 && 3587 ip->ip_p == IPPROTO_TCP) { 3588 struct tcphdr *th; 3589 3590 th = (struct tcphdr *)(ip + 1); 3591 if (th->th_x2) 3592 ldt = 1; 3593 } 3594 3595 if (ldt) { 3596 struct tcphdr *th; 3597 struct udphdr *uh; 3598 u_short cksum; 3599 3600 ip->ip_len = ntohs(ip->ip_len); 3601 cksum = in_pseudo( 3602 ip->ip_src.s_addr, 3603 ip->ip_dst.s_addr, 3604 htons(ip->ip_p + ip->ip_len - 3605 (ip->ip_hl << 2)) 3606 ); 3607 3608 switch (ip->ip_p) { 3609 case IPPROTO_TCP: 3610 th = (struct tcphdr *)(ip + 1); 3611 /* 3612 * Maybe it was set in 3613 * libalias... 3614 */ 3615 th->th_x2 = 0; 3616 th->th_sum = cksum; 3617 mcl->m_pkthdr.csum_data = 3618 offsetof(struct tcphdr, 3619 th_sum); 3620 break; 3621 case IPPROTO_UDP: 3622 uh = (struct udphdr *)(ip + 1); 3623 uh->uh_sum = cksum; 3624 mcl->m_pkthdr.csum_data = 3625 offsetof(struct udphdr, 3626 uh_sum); 3627 break; 3628 } 3629 /* 3630 * No hw checksum offloading: do it 3631 * by ourself. 3632 */ 3633 if ((mcl->m_pkthdr.csum_flags & 3634 CSUM_DELAY_DATA) == 0) { 3635 in_delayed_cksum(mcl); 3636 mcl->m_pkthdr.csum_flags &= 3637 ~CSUM_DELAY_DATA; 3638 } 3639 ip->ip_len = htons(ip->ip_len); 3640 } 3641 3642 if (args->eh == NULL) { 3643 ip->ip_len = ntohs(ip->ip_len); 3644 ip->ip_off = ntohs(ip->ip_off); 3645 } 3646 3647 args->m = mcl; 3648 retval = IP_FW_NAT; 3649 goto done; 3650 } 3651#endif 3652 3653 default: 3654 panic("-- unknown opcode %d\n", cmd->opcode); 3655 } /* end of switch() on opcodes */ 3656 3657 if (cmd->len & F_NOT) 3658 match = !match; 3659 3660 if (match) { 3661 if (cmd->len & F_OR) 3662 skip_or = 1; 3663 } else { 3664 if (!(cmd->len & F_OR)) /* not an OR block, */ 3665 break; /* try next rule */ 3666 } 3667 3668 } /* end of inner for, scan opcodes */ 3669 3670next_rule:; /* try next rule */ 3671 3672 } /* end of outer for, scan rules */ 3673 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3674 IPFW_RUNLOCK(chain); 3675 return (IP_FW_DENY); 3676 3677done: 3678 /* Update statistics */ 3679 f->pcnt++; 3680 f->bcnt += pktlen; 3681 f->timestamp = time_uptime; 3682 IPFW_RUNLOCK(chain); 3683 return (retval); 3684 3685pullup_failed: 3686 if (fw_verbose) 3687 printf("ipfw: pullup failed\n"); 3688 return (IP_FW_DENY); 3689} 3690 3691/* 3692 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3693 * These will be reconstructed on the fly as packets are matched. 3694 */ 3695static void 3696flush_rule_ptrs(struct ip_fw_chain *chain) 3697{ 3698 struct ip_fw *rule; 3699 3700 IPFW_WLOCK_ASSERT(chain); 3701 3702 for (rule = chain->rules; rule; rule = rule->next) 3703 rule->next_rule = NULL; 3704} 3705 3706/* 3707 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3708 * possibly create a rule number and add the rule to the list. 3709 * Update the rule_number in the input struct so the caller knows it as well. 3710 */ 3711static int 3712add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3713{ 3714 struct ip_fw *rule, *f, *prev; 3715 int l = RULESIZE(input_rule); 3716 3717 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3718 return (EINVAL); 3719 3720 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3721 if (rule == NULL) 3722 return (ENOSPC); 3723 3724 bcopy(input_rule, rule, l); 3725 3726 rule->next = NULL; 3727 rule->next_rule = NULL; 3728 3729 rule->pcnt = 0; 3730 rule->bcnt = 0; 3731 rule->timestamp = 0; 3732 3733 IPFW_WLOCK(chain); 3734 3735 if (chain->rules == NULL) { /* default rule */ 3736 chain->rules = rule; 3737 goto done; 3738 } 3739 3740 /* 3741 * If rulenum is 0, find highest numbered rule before the 3742 * default rule, and add autoinc_step 3743 */ 3744 if (autoinc_step < 1) 3745 autoinc_step = 1; 3746 else if (autoinc_step > 1000) 3747 autoinc_step = 1000; 3748 if (rule->rulenum == 0) { 3749 /* 3750 * locate the highest numbered rule before default 3751 */ 3752 for (f = chain->rules; f; f = f->next) { 3753 if (f->rulenum == IPFW_DEFAULT_RULE) 3754 break; 3755 rule->rulenum = f->rulenum; 3756 } 3757 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 3758 rule->rulenum += autoinc_step; 3759 input_rule->rulenum = rule->rulenum; 3760 } 3761 3762 /* 3763 * Now insert the new rule in the right place in the sorted list. 3764 */ 3765 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3766 if (f->rulenum > rule->rulenum) { /* found the location */ 3767 if (prev) { 3768 rule->next = f; 3769 prev->next = rule; 3770 } else { /* head insert */ 3771 rule->next = chain->rules; 3772 chain->rules = rule; 3773 } 3774 break; 3775 } 3776 } 3777 flush_rule_ptrs(chain); 3778done: 3779 static_count++; 3780 static_len += l; 3781 IPFW_WUNLOCK(chain); 3782 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3783 rule->rulenum, static_count);) 3784 return (0); 3785} 3786 3787/** 3788 * Remove a static rule (including derived * dynamic rules) 3789 * and place it on the ``reap list'' for later reclamation. 3790 * The caller is in charge of clearing rule pointers to avoid 3791 * dangling pointers. 3792 * @return a pointer to the next entry. 3793 * Arguments are not checked, so they better be correct. 3794 */ 3795static struct ip_fw * 3796remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, 3797 struct ip_fw *prev) 3798{ 3799 struct ip_fw *n; 3800 int l = RULESIZE(rule); 3801 3802 IPFW_WLOCK_ASSERT(chain); 3803 3804 n = rule->next; 3805 IPFW_DYN_LOCK(); 3806 remove_dyn_rule(rule, NULL /* force removal */); 3807 IPFW_DYN_UNLOCK(); 3808 if (prev == NULL) 3809 chain->rules = n; 3810 else 3811 prev->next = n; 3812 static_count--; 3813 static_len -= l; 3814 3815 rule->next = chain->reap; 3816 chain->reap = rule; 3817 3818 return n; 3819} 3820 3821/** 3822 * Reclaim storage associated with a list of rules. This is 3823 * typically the list created using remove_rule. 3824 */ 3825static void 3826reap_rules(struct ip_fw *head) 3827{ 3828 struct ip_fw *rule; 3829 3830 while ((rule = head) != NULL) { 3831 head = head->next; 3832 if (DUMMYNET_LOADED) 3833 ip_dn_ruledel_ptr(rule); 3834 free(rule, M_IPFW); 3835 } 3836} 3837 3838/* 3839 * Remove all rules from a chain (except rules in set RESVD_SET 3840 * unless kill_default = 1). The caller is responsible for 3841 * reclaiming storage for the rules left in chain->reap. 3842 */ 3843static void 3844free_chain(struct ip_fw_chain *chain, int kill_default) 3845{ 3846 struct ip_fw *prev, *rule; 3847 3848 IPFW_WLOCK_ASSERT(chain); 3849 3850 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3851 for (prev = NULL, rule = chain->rules; rule ; ) 3852 if (kill_default || rule->set != RESVD_SET) 3853 rule = remove_rule(chain, rule, prev); 3854 else { 3855 prev = rule; 3856 rule = rule->next; 3857 } 3858} 3859 3860/** 3861 * Remove all rules with given number, and also do set manipulation. 3862 * Assumes chain != NULL && *chain != NULL. 3863 * 3864 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3865 * the next 8 bits are the new set, the top 8 bits are the command: 3866 * 3867 * 0 delete rules with given number 3868 * 1 delete rules with given set number 3869 * 2 move rules with given number to new set 3870 * 3 move rules with given set number to new set 3871 * 4 swap sets with given numbers 3872 * 5 delete rules with given number and with given set number 3873 */ 3874static int 3875del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3876{ 3877 struct ip_fw *prev = NULL, *rule; 3878 u_int16_t rulenum; /* rule or old_set */ 3879 u_int8_t cmd, new_set; 3880 3881 rulenum = arg & 0xffff; 3882 cmd = (arg >> 24) & 0xff; 3883 new_set = (arg >> 16) & 0xff; 3884 3885 if (cmd > 5 || new_set > RESVD_SET) 3886 return EINVAL; 3887 if (cmd == 0 || cmd == 2 || cmd == 5) { 3888 if (rulenum >= IPFW_DEFAULT_RULE) 3889 return EINVAL; 3890 } else { 3891 if (rulenum > RESVD_SET) /* old_set */ 3892 return EINVAL; 3893 } 3894 3895 IPFW_WLOCK(chain); 3896 rule = chain->rules; 3897 chain->reap = NULL; 3898 switch (cmd) { 3899 case 0: /* delete rules with given number */ 3900 /* 3901 * locate first rule to delete 3902 */ 3903 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3904 ; 3905 if (rule->rulenum != rulenum) { 3906 IPFW_WUNLOCK(chain); 3907 return EINVAL; 3908 } 3909 3910 /* 3911 * flush pointers outside the loop, then delete all matching 3912 * rules. prev remains the same throughout the cycle. 3913 */ 3914 flush_rule_ptrs(chain); 3915 while (rule->rulenum == rulenum) 3916 rule = remove_rule(chain, rule, prev); 3917 break; 3918 3919 case 1: /* delete all rules with given set number */ 3920 flush_rule_ptrs(chain); 3921 rule = chain->rules; 3922 while (rule->rulenum < IPFW_DEFAULT_RULE) 3923 if (rule->set == rulenum) 3924 rule = remove_rule(chain, rule, prev); 3925 else { 3926 prev = rule; 3927 rule = rule->next; 3928 } 3929 break; 3930 3931 case 2: /* move rules with given number to new set */ 3932 rule = chain->rules; 3933 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3934 if (rule->rulenum == rulenum) 3935 rule->set = new_set; 3936 break; 3937 3938 case 3: /* move rules with given set number to new set */ 3939 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3940 if (rule->set == rulenum) 3941 rule->set = new_set; 3942 break; 3943 3944 case 4: /* swap two sets */ 3945 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3946 if (rule->set == rulenum) 3947 rule->set = new_set; 3948 else if (rule->set == new_set) 3949 rule->set = rulenum; 3950 break; 3951 case 5: /* delete rules with given number and with given set number. 3952 * rulenum - given rule number; 3953 * new_set - given set number. 3954 */ 3955 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3956 ; 3957 if (rule->rulenum != rulenum) { 3958 IPFW_WUNLOCK(chain); 3959 return (EINVAL); 3960 } 3961 flush_rule_ptrs(chain); 3962 while (rule->rulenum == rulenum) { 3963 if (rule->set == new_set) 3964 rule = remove_rule(chain, rule, prev); 3965 else { 3966 prev = rule; 3967 rule = rule->next; 3968 } 3969 } 3970 } 3971 /* 3972 * Look for rules to reclaim. We grab the list before 3973 * releasing the lock then reclaim them w/o the lock to 3974 * avoid a LOR with dummynet. 3975 */ 3976 rule = chain->reap; 3977 chain->reap = NULL; 3978 IPFW_WUNLOCK(chain); 3979 if (rule) 3980 reap_rules(rule); 3981 return 0; 3982} 3983 3984/* 3985 * Clear counters for a specific rule. 3986 * The enclosing "table" is assumed locked. 3987 */ 3988static void 3989clear_counters(struct ip_fw *rule, int log_only) 3990{ 3991 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3992 3993 if (log_only == 0) { 3994 rule->bcnt = rule->pcnt = 0; 3995 rule->timestamp = 0; 3996 } 3997 if (l->o.opcode == O_LOG) 3998 l->log_left = l->max_log; 3999} 4000 4001/** 4002 * Reset some or all counters on firewall rules. 4003 * The argument `arg' is an u_int32_t. The low 16 bit are the rule number, 4004 * the next 8 bits are the set number, the top 8 bits are the command: 4005 * 0 work with rules from all set's; 4006 * 1 work with rules only from specified set. 4007 * Specified rule number is zero if we want to clear all entries. 4008 * log_only is 1 if we only want to reset logs, zero otherwise. 4009 */ 4010static int 4011zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only) 4012{ 4013 struct ip_fw *rule; 4014 char *msg; 4015 4016 uint16_t rulenum = arg & 0xffff; 4017 uint8_t set = (arg >> 16) & 0xff; 4018 uint8_t cmd = (arg >> 24) & 0xff; 4019 4020 if (cmd > 1) 4021 return (EINVAL); 4022 if (cmd == 1 && set > RESVD_SET) 4023 return (EINVAL); 4024 4025 IPFW_WLOCK(chain); 4026 if (rulenum == 0) { 4027 norule_counter = 0; 4028 for (rule = chain->rules; rule; rule = rule->next) { 4029 /* Skip rules from another set. */ 4030 if (cmd == 1 && rule->set != set) 4031 continue; 4032 clear_counters(rule, log_only); 4033 } 4034 msg = log_only ? "ipfw: All logging counts reset.\n" : 4035 "ipfw: Accounting cleared.\n"; 4036 } else { 4037 int cleared = 0; 4038 /* 4039 * We can have multiple rules with the same number, so we 4040 * need to clear them all. 4041 */ 4042 for (rule = chain->rules; rule; rule = rule->next) 4043 if (rule->rulenum == rulenum) { 4044 while (rule && rule->rulenum == rulenum) { 4045 if (cmd == 0 || rule->set == set) 4046 clear_counters(rule, log_only); 4047 rule = rule->next; 4048 } 4049 cleared = 1; 4050 break; 4051 } 4052 if (!cleared) { /* we did not find any matching rules */ 4053 IPFW_WUNLOCK(chain); 4054 return (EINVAL); 4055 } 4056 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 4057 "ipfw: Entry %d cleared.\n"; 4058 } 4059 IPFW_WUNLOCK(chain); 4060 4061 if (fw_verbose) 4062 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 4063 return (0); 4064} 4065 4066/* 4067 * Check validity of the structure before insert. 4068 * Fortunately rules are simple, so this mostly need to check rule sizes. 4069 */ 4070static int 4071check_ipfw_struct(struct ip_fw *rule, int size) 4072{ 4073 int l, cmdlen = 0; 4074 int have_action=0; 4075 ipfw_insn *cmd; 4076 4077 if (size < sizeof(*rule)) { 4078 printf("ipfw: rule too short\n"); 4079 return (EINVAL); 4080 } 4081 /* first, check for valid size */ 4082 l = RULESIZE(rule); 4083 if (l != size) { 4084 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 4085 return (EINVAL); 4086 } 4087 if (rule->act_ofs >= rule->cmd_len) { 4088 printf("ipfw: bogus action offset (%u > %u)\n", 4089 rule->act_ofs, rule->cmd_len - 1); 4090 return (EINVAL); 4091 } 4092 /* 4093 * Now go for the individual checks. Very simple ones, basically only 4094 * instruction sizes. 4095 */ 4096 for (l = rule->cmd_len, cmd = rule->cmd ; 4097 l > 0 ; l -= cmdlen, cmd += cmdlen) { 4098 cmdlen = F_LEN(cmd); 4099 if (cmdlen > l) { 4100 printf("ipfw: opcode %d size truncated\n", 4101 cmd->opcode); 4102 return EINVAL; 4103 } 4104 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 4105 switch (cmd->opcode) { 4106 case O_PROBE_STATE: 4107 case O_KEEP_STATE: 4108 case O_PROTO: 4109 case O_IP_SRC_ME: 4110 case O_IP_DST_ME: 4111 case O_LAYER2: 4112 case O_IN: 4113 case O_FRAG: 4114 case O_DIVERTED: 4115 case O_IPOPT: 4116 case O_IPTOS: 4117 case O_IPPRECEDENCE: 4118 case O_IPVER: 4119 case O_TCPWIN: 4120 case O_TCPFLAGS: 4121 case O_TCPOPTS: 4122 case O_ESTAB: 4123 case O_VERREVPATH: 4124 case O_VERSRCREACH: 4125 case O_ANTISPOOF: 4126 case O_IPSEC: 4127#ifdef INET6 4128 case O_IP6_SRC_ME: 4129 case O_IP6_DST_ME: 4130 case O_EXT_HDR: 4131 case O_IP6: 4132#endif 4133 case O_IP4: 4134 case O_TAG: 4135 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 4136 goto bad_size; 4137 break; 4138 4139 case O_UID: 4140 case O_GID: 4141 case O_JAIL: 4142 case O_IP_SRC: 4143 case O_IP_DST: 4144 case O_TCPSEQ: 4145 case O_TCPACK: 4146 case O_PROB: 4147 case O_ICMPTYPE: 4148 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 4149 goto bad_size; 4150 break; 4151 4152 case O_LIMIT: 4153 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 4154 goto bad_size; 4155 break; 4156 4157 case O_LOG: 4158 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 4159 goto bad_size; 4160 4161 ((ipfw_insn_log *)cmd)->log_left = 4162 ((ipfw_insn_log *)cmd)->max_log; 4163 4164 break; 4165 4166 case O_IP_SRC_MASK: 4167 case O_IP_DST_MASK: 4168 /* only odd command lengths */ 4169 if ( !(cmdlen & 1) || cmdlen > 31) 4170 goto bad_size; 4171 break; 4172 4173 case O_IP_SRC_SET: 4174 case O_IP_DST_SET: 4175 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 4176 printf("ipfw: invalid set size %d\n", 4177 cmd->arg1); 4178 return EINVAL; 4179 } 4180 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 4181 (cmd->arg1+31)/32 ) 4182 goto bad_size; 4183 break; 4184 4185 case O_IP_SRC_LOOKUP: 4186 case O_IP_DST_LOOKUP: 4187 if (cmd->arg1 >= IPFW_TABLES_MAX) { 4188 printf("ipfw: invalid table number %d\n", 4189 cmd->arg1); 4190 return (EINVAL); 4191 } 4192 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 4193 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 4194 goto bad_size; 4195 break; 4196 4197 case O_MACADDR2: 4198 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 4199 goto bad_size; 4200 break; 4201 4202 case O_NOP: 4203 case O_IPID: 4204 case O_IPTTL: 4205 case O_IPLEN: 4206 case O_TCPDATALEN: 4207 case O_TAGGED: 4208 if (cmdlen < 1 || cmdlen > 31) 4209 goto bad_size; 4210 break; 4211 4212 case O_MAC_TYPE: 4213 case O_IP_SRCPORT: 4214 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 4215 if (cmdlen < 2 || cmdlen > 31) 4216 goto bad_size; 4217 break; 4218 4219 case O_RECV: 4220 case O_XMIT: 4221 case O_VIA: 4222 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 4223 goto bad_size; 4224 break; 4225 4226 case O_ALTQ: 4227 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 4228 goto bad_size; 4229 break; 4230 4231 case O_PIPE: 4232 case O_QUEUE: 4233 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 4234 goto bad_size; 4235 goto check_action; 4236 4237 case O_FORWARD_IP: 4238#ifdef IPFIREWALL_FORWARD 4239 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 4240 goto bad_size; 4241 goto check_action; 4242#else 4243 return EINVAL; 4244#endif 4245 4246 case O_DIVERT: 4247 case O_TEE: 4248 if (ip_divert_ptr == NULL) 4249 return EINVAL; 4250 else 4251 goto check_size; 4252 case O_NETGRAPH: 4253 case O_NGTEE: 4254 if (!NG_IPFW_LOADED) 4255 return EINVAL; 4256 else 4257 goto check_size; 4258 case O_NAT: 4259#ifdef IPFIREWALL_NAT 4260 if (cmdlen != F_INSN_SIZE(ipfw_insn_nat)) 4261 goto bad_size; 4262 goto check_action; 4263#else 4264 return EINVAL; 4265#endif 4266 case O_FORWARD_MAC: /* XXX not implemented yet */ 4267 case O_CHECK_STATE: 4268 case O_COUNT: 4269 case O_ACCEPT: 4270 case O_DENY: 4271 case O_REJECT: 4272#ifdef INET6 4273 case O_UNREACH6: 4274#endif 4275 case O_SKIPTO: 4276check_size: 4277 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 4278 goto bad_size; 4279check_action: 4280 if (have_action) { 4281 printf("ipfw: opcode %d, multiple actions" 4282 " not allowed\n", 4283 cmd->opcode); 4284 return EINVAL; 4285 } 4286 have_action = 1; 4287 if (l != cmdlen) { 4288 printf("ipfw: opcode %d, action must be" 4289 " last opcode\n", 4290 cmd->opcode); 4291 return EINVAL; 4292 } 4293 break; 4294#ifdef INET6 4295 case O_IP6_SRC: 4296 case O_IP6_DST: 4297 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 4298 F_INSN_SIZE(ipfw_insn)) 4299 goto bad_size; 4300 break; 4301 4302 case O_FLOW6ID: 4303 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 4304 ((ipfw_insn_u32 *)cmd)->o.arg1) 4305 goto bad_size; 4306 break; 4307 4308 case O_IP6_SRC_MASK: 4309 case O_IP6_DST_MASK: 4310 if ( !(cmdlen & 1) || cmdlen > 127) 4311 goto bad_size; 4312 break; 4313 case O_ICMP6TYPE: 4314 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 4315 goto bad_size; 4316 break; 4317#endif 4318 4319 default: 4320 switch (cmd->opcode) { 4321#ifndef INET6 4322 case O_IP6_SRC_ME: 4323 case O_IP6_DST_ME: 4324 case O_EXT_HDR: 4325 case O_IP6: 4326 case O_UNREACH6: 4327 case O_IP6_SRC: 4328 case O_IP6_DST: 4329 case O_FLOW6ID: 4330 case O_IP6_SRC_MASK: 4331 case O_IP6_DST_MASK: 4332 case O_ICMP6TYPE: 4333 printf("ipfw: no IPv6 support in kernel\n"); 4334 return EPROTONOSUPPORT; 4335#endif 4336 default: 4337 printf("ipfw: opcode %d, unknown opcode\n", 4338 cmd->opcode); 4339 return EINVAL; 4340 } 4341 } 4342 } 4343 if (have_action == 0) { 4344 printf("ipfw: missing action\n"); 4345 return EINVAL; 4346 } 4347 return 0; 4348 4349bad_size: 4350 printf("ipfw: opcode %d size %d wrong\n", 4351 cmd->opcode, cmdlen); 4352 return EINVAL; 4353} 4354 4355/* 4356 * Copy the static and dynamic rules to the supplied buffer 4357 * and return the amount of space actually used. 4358 */ 4359static size_t 4360ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 4361{ 4362 char *bp = buf; 4363 char *ep = bp + space; 4364 struct ip_fw *rule; 4365 int i; 4366 time_t boot_seconds; 4367 4368 boot_seconds = boottime.tv_sec; 4369 /* XXX this can take a long time and locking will block packet flow */ 4370 IPFW_RLOCK(chain); 4371 for (rule = chain->rules; rule ; rule = rule->next) { 4372 /* 4373 * Verify the entry fits in the buffer in case the 4374 * rules changed between calculating buffer space and 4375 * now. This would be better done using a generation 4376 * number but should suffice for now. 4377 */ 4378 i = RULESIZE(rule); 4379 if (bp + i <= ep) { 4380 bcopy(rule, bp, i); 4381 /* 4382 * XXX HACK. Store the disable mask in the "next" pointer 4383 * in a wild attempt to keep the ABI the same. 4384 * Why do we do this on EVERY rule? 4385 */ 4386 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 4387 sizeof(set_disable)); 4388 if (((struct ip_fw *)bp)->timestamp) 4389 ((struct ip_fw *)bp)->timestamp += boot_seconds; 4390 bp += i; 4391 } 4392 } 4393 IPFW_RUNLOCK(chain); 4394 if (ipfw_dyn_v) { 4395 ipfw_dyn_rule *p, *last = NULL; 4396 4397 IPFW_DYN_LOCK(); 4398 for (i = 0 ; i < curr_dyn_buckets; i++) 4399 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 4400 if (bp + sizeof *p <= ep) { 4401 ipfw_dyn_rule *dst = 4402 (ipfw_dyn_rule *)bp; 4403 bcopy(p, dst, sizeof *p); 4404 bcopy(&(p->rule->rulenum), &(dst->rule), 4405 sizeof(p->rule->rulenum)); 4406 /* 4407 * store set number into high word of 4408 * dst->rule pointer. 4409 */ 4410 bcopy(&(p->rule->set), &dst->rule + 4411 sizeof(p->rule->rulenum), 4412 sizeof(p->rule->set)); 4413 /* 4414 * store a non-null value in "next". 4415 * The userland code will interpret a 4416 * NULL here as a marker 4417 * for the last dynamic rule. 4418 */ 4419 bcopy(&dst, &dst->next, sizeof(dst)); 4420 last = dst; 4421 dst->expire = 4422 TIME_LEQ(dst->expire, time_uptime) ? 4423 0 : dst->expire - time_uptime ; 4424 bp += sizeof(ipfw_dyn_rule); 4425 } 4426 } 4427 IPFW_DYN_UNLOCK(); 4428 if (last != NULL) /* mark last dynamic rule */ 4429 bzero(&last->next, sizeof(last)); 4430 } 4431 return (bp - (char *)buf); 4432} 4433 4434 4435/** 4436 * {set|get}sockopt parser. 4437 */ 4438static int 4439ipfw_ctl(struct sockopt *sopt) 4440{ 4441#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 4442 int error; 4443 size_t size; 4444 struct ip_fw *buf, *rule; 4445 u_int32_t rulenum[2]; 4446 4447 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW); 4448 if (error) 4449 return (error); 4450 4451 /* 4452 * Disallow modifications in really-really secure mode, but still allow 4453 * the logging counters to be reset. 4454 */ 4455 if (sopt->sopt_name == IP_FW_ADD || 4456 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 4457 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 4458 if (error) 4459 return (error); 4460 } 4461 4462 error = 0; 4463 4464 switch (sopt->sopt_name) { 4465 case IP_FW_GET: 4466 /* 4467 * pass up a copy of the current rules. Static rules 4468 * come first (the last of which has number IPFW_DEFAULT_RULE), 4469 * followed by a possibly empty list of dynamic rule. 4470 * The last dynamic rule has NULL in the "next" field. 4471 * 4472 * Note that the calculated size is used to bound the 4473 * amount of data returned to the user. The rule set may 4474 * change between calculating the size and returning the 4475 * data in which case we'll just return what fits. 4476 */ 4477 size = static_len; /* size of static rules */ 4478 if (ipfw_dyn_v) /* add size of dyn.rules */ 4479 size += (dyn_count * sizeof(ipfw_dyn_rule)); 4480 4481 /* 4482 * XXX todo: if the user passes a short length just to know 4483 * how much room is needed, do not bother filling up the 4484 * buffer, just jump to the sooptcopyout. 4485 */ 4486 buf = malloc(size, M_TEMP, M_WAITOK); 4487 error = sooptcopyout(sopt, buf, 4488 ipfw_getrules(&layer3_chain, buf, size)); 4489 free(buf, M_TEMP); 4490 break; 4491 4492 case IP_FW_FLUSH: 4493 /* 4494 * Normally we cannot release the lock on each iteration. 4495 * We could do it here only because we start from the head all 4496 * the times so there is no risk of missing some entries. 4497 * On the other hand, the risk is that we end up with 4498 * a very inconsistent ruleset, so better keep the lock 4499 * around the whole cycle. 4500 * 4501 * XXX this code can be improved by resetting the head of 4502 * the list to point to the default rule, and then freeing 4503 * the old list without the need for a lock. 4504 */ 4505 4506 IPFW_WLOCK(&layer3_chain); 4507 layer3_chain.reap = NULL; 4508 free_chain(&layer3_chain, 0 /* keep default rule */); 4509 rule = layer3_chain.reap; 4510 layer3_chain.reap = NULL; 4511 IPFW_WUNLOCK(&layer3_chain); 4512 if (rule != NULL) 4513 reap_rules(rule); 4514 break; 4515 4516 case IP_FW_ADD: 4517 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 4518 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 4519 sizeof(struct ip_fw) ); 4520 if (error == 0) 4521 error = check_ipfw_struct(rule, sopt->sopt_valsize); 4522 if (error == 0) { 4523 error = add_rule(&layer3_chain, rule); 4524 size = RULESIZE(rule); 4525 if (!error && sopt->sopt_dir == SOPT_GET) 4526 error = sooptcopyout(sopt, rule, size); 4527 } 4528 free(rule, M_TEMP); 4529 break; 4530 4531 case IP_FW_DEL: 4532 /* 4533 * IP_FW_DEL is used for deleting single rules or sets, 4534 * and (ab)used to atomically manipulate sets. Argument size 4535 * is used to distinguish between the two: 4536 * sizeof(u_int32_t) 4537 * delete single rule or set of rules, 4538 * or reassign rules (or sets) to a different set. 4539 * 2*sizeof(u_int32_t) 4540 * atomic disable/enable sets. 4541 * first u_int32_t contains sets to be disabled, 4542 * second u_int32_t contains sets to be enabled. 4543 */ 4544 error = sooptcopyin(sopt, rulenum, 4545 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4546 if (error) 4547 break; 4548 size = sopt->sopt_valsize; 4549 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4550 error = del_entry(&layer3_chain, rulenum[0]); 4551 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4552 set_disable = 4553 (set_disable | rulenum[0]) & ~rulenum[1] & 4554 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4555 else 4556 error = EINVAL; 4557 break; 4558 4559 case IP_FW_ZERO: 4560 case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */ 4561 rulenum[0] = 0; 4562 if (sopt->sopt_val != 0) { 4563 error = sooptcopyin(sopt, rulenum, 4564 sizeof(u_int32_t), sizeof(u_int32_t)); 4565 if (error) 4566 break; 4567 } 4568 error = zero_entry(&layer3_chain, rulenum[0], 4569 sopt->sopt_name == IP_FW_RESETLOG); 4570 break; 4571 4572 case IP_FW_TABLE_ADD: 4573 { 4574 ipfw_table_entry ent; 4575 4576 error = sooptcopyin(sopt, &ent, 4577 sizeof(ent), sizeof(ent)); 4578 if (error) 4579 break; 4580 error = add_table_entry(&layer3_chain, ent.tbl, 4581 ent.addr, ent.masklen, ent.value); 4582 } 4583 break; 4584 4585 case IP_FW_TABLE_DEL: 4586 { 4587 ipfw_table_entry ent; 4588 4589 error = sooptcopyin(sopt, &ent, 4590 sizeof(ent), sizeof(ent)); 4591 if (error) 4592 break; 4593 error = del_table_entry(&layer3_chain, ent.tbl, 4594 ent.addr, ent.masklen); 4595 } 4596 break; 4597 4598 case IP_FW_TABLE_FLUSH: 4599 { 4600 u_int16_t tbl; 4601 4602 error = sooptcopyin(sopt, &tbl, 4603 sizeof(tbl), sizeof(tbl)); 4604 if (error) 4605 break; 4606 IPFW_WLOCK(&layer3_chain); 4607 error = flush_table(&layer3_chain, tbl); 4608 IPFW_WUNLOCK(&layer3_chain); 4609 } 4610 break; 4611 4612 case IP_FW_TABLE_GETSIZE: 4613 { 4614 u_int32_t tbl, cnt; 4615 4616 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4617 sizeof(tbl)))) 4618 break; 4619 IPFW_RLOCK(&layer3_chain); 4620 error = count_table(&layer3_chain, tbl, &cnt); 4621 IPFW_RUNLOCK(&layer3_chain); 4622 if (error) 4623 break; 4624 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4625 } 4626 break; 4627 4628 case IP_FW_TABLE_LIST: 4629 { 4630 ipfw_table *tbl; 4631 4632 if (sopt->sopt_valsize < sizeof(*tbl)) { 4633 error = EINVAL; 4634 break; 4635 } 4636 size = sopt->sopt_valsize; 4637 tbl = malloc(size, M_TEMP, M_WAITOK); 4638 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4639 if (error) { 4640 free(tbl, M_TEMP); 4641 break; 4642 } 4643 tbl->size = (size - sizeof(*tbl)) / 4644 sizeof(ipfw_table_entry); 4645 IPFW_RLOCK(&layer3_chain); 4646 error = dump_table(&layer3_chain, tbl); 4647 IPFW_RUNLOCK(&layer3_chain); 4648 if (error) { 4649 free(tbl, M_TEMP); 4650 break; 4651 } 4652 error = sooptcopyout(sopt, tbl, size); 4653 free(tbl, M_TEMP); 4654 } 4655 break; 4656 4657#ifdef IPFIREWALL_NAT 4658 case IP_FW_NAT_CFG: 4659 { 4660 struct cfg_nat *ptr, *ser_n; 4661 char *buf; 4662 4663 buf = malloc(NAT_BUF_LEN, M_IPFW, M_WAITOK | M_ZERO); 4664 error = sooptcopyin(sopt, buf, NAT_BUF_LEN, 4665 sizeof(struct cfg_nat)); 4666 ser_n = (struct cfg_nat *)buf; 4667 4668 /* 4669 * Find/create nat rule. 4670 */ 4671 IPFW_WLOCK(&layer3_chain); 4672 ptr = lookup_nat(ser_n->id); 4673 if (ptr == NULL) { 4674 /* New rule: allocate and init new instance. */ 4675 ptr = malloc(sizeof(struct cfg_nat), 4676 M_IPFW, M_NOWAIT | M_ZERO); 4677 if (ptr == NULL) { 4678 IPFW_WUNLOCK(&layer3_chain); 4679 free(buf, M_IPFW); 4680 return (ENOSPC); 4681 } 4682 ptr->lib = LibAliasInit(NULL); 4683 if (ptr->lib == NULL) { 4684 IPFW_WUNLOCK(&layer3_chain); 4685 free(ptr, M_IPFW); 4686 free(buf, M_IPFW); 4687 return (EINVAL); 4688 } 4689 LIST_INIT(&ptr->redir_chain); 4690 } else { 4691 /* Entry already present: temporarly unhook it. */ 4692 UNHOOK_NAT(ptr); 4693 flush_nat_ptrs(ser_n->id); 4694 } 4695 IPFW_WUNLOCK(&layer3_chain); 4696 4697 /* 4698 * Basic nat configuration. 4699 */ 4700 ptr->id = ser_n->id; 4701 /* 4702 * XXX - what if this rule doesn't nat any ip and just 4703 * redirect? 4704 * do we set aliasaddress to 0.0.0.0? 4705 */ 4706 ptr->ip = ser_n->ip; 4707 ptr->redir_cnt = ser_n->redir_cnt; 4708 ptr->mode = ser_n->mode; 4709 LibAliasSetMode(ptr->lib, ser_n->mode, ser_n->mode); 4710 LibAliasSetAddress(ptr->lib, ptr->ip); 4711 memcpy(ptr->if_name, ser_n->if_name, IF_NAMESIZE); 4712 4713 /* 4714 * Redir and LSNAT configuration. 4715 */ 4716 /* Delete old cfgs. */ 4717 del_redir_spool_cfg(ptr, &ptr->redir_chain); 4718 /* Add new entries. */ 4719 add_redir_spool_cfg(&buf[(sizeof(struct cfg_nat))], ptr); 4720 free(buf, M_IPFW); 4721 IPFW_WLOCK(&layer3_chain); 4722 HOOK_NAT(&layer3_chain.nat, ptr); 4723 IPFW_WUNLOCK(&layer3_chain); 4724 } 4725 break; 4726 4727 case IP_FW_NAT_DEL: 4728 { 4729 struct cfg_nat *ptr; 4730 int i; 4731 4732 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 4733 IPFW_WLOCK(&layer3_chain); 4734 ptr = lookup_nat(i); 4735 if (ptr == NULL) { 4736 error = EINVAL; 4737 IPFW_WUNLOCK(&layer3_chain); 4738 break; 4739 } 4740 UNHOOK_NAT(ptr); 4741 flush_nat_ptrs(i); 4742 IPFW_WUNLOCK(&layer3_chain); 4743 del_redir_spool_cfg(ptr, &ptr->redir_chain); 4744 LibAliasUninit(ptr->lib); 4745 free(ptr, M_IPFW); 4746 } 4747 break; 4748 4749 case IP_FW_NAT_GET_CONFIG: 4750 { 4751 uint8_t *data; 4752 struct cfg_nat *n; 4753 struct cfg_redir *r; 4754 struct cfg_spool *s; 4755 int nat_cnt, off; 4756 4757 nat_cnt = 0; 4758 off = sizeof(nat_cnt); 4759 4760 data = malloc(NAT_BUF_LEN, M_IPFW, M_WAITOK | M_ZERO); 4761 IPFW_RLOCK(&layer3_chain); 4762 /* Serialize all the data. */ 4763 LIST_FOREACH(n, &layer3_chain.nat, _next) { 4764 nat_cnt++; 4765 if (off + SOF_NAT < NAT_BUF_LEN) { 4766 bcopy(n, &data[off], SOF_NAT); 4767 off += SOF_NAT; 4768 LIST_FOREACH(r, &n->redir_chain, _next) { 4769 if (off + SOF_REDIR < NAT_BUF_LEN) { 4770 bcopy(r, &data[off], 4771 SOF_REDIR); 4772 off += SOF_REDIR; 4773 LIST_FOREACH(s, &r->spool_chain, 4774 _next) { 4775 if (off + SOF_SPOOL < 4776 NAT_BUF_LEN) { 4777 bcopy(s, 4778 &data[off], 4779 SOF_SPOOL); 4780 off += 4781 SOF_SPOOL; 4782 } else 4783 goto nospace; 4784 } 4785 } else 4786 goto nospace; 4787 } 4788 } else 4789 goto nospace; 4790 } 4791 bcopy(&nat_cnt, data, sizeof(nat_cnt)); 4792 IPFW_RUNLOCK(&layer3_chain); 4793 error = sooptcopyout(sopt, data, NAT_BUF_LEN); 4794 free(data, M_IPFW); 4795 break; 4796 nospace: 4797 IPFW_RUNLOCK(&layer3_chain); 4798 printf("serialized data buffer not big enough:" 4799 "please increase NAT_BUF_LEN\n"); 4800 free(data, M_IPFW); 4801 } 4802 break; 4803 4804 case IP_FW_NAT_GET_LOG: 4805 { 4806 uint8_t *data; 4807 struct cfg_nat *ptr; 4808 int i, size, cnt, sof; 4809 4810 data = NULL; 4811 sof = LIBALIAS_BUF_SIZE; 4812 cnt = 0; 4813 4814 IPFW_RLOCK(&layer3_chain); 4815 size = i = 0; 4816 LIST_FOREACH(ptr, &layer3_chain.nat, _next) { 4817 if (ptr->lib->logDesc == NULL) 4818 continue; 4819 cnt++; 4820 size = cnt * (sof + sizeof(int)); 4821 data = realloc(data, size, M_IPFW, M_NOWAIT | M_ZERO); 4822 if (data == NULL) { 4823 IPFW_RUNLOCK(&layer3_chain); 4824 return (ENOSPC); 4825 } 4826 bcopy(&ptr->id, &data[i], sizeof(int)); 4827 i += sizeof(int); 4828 bcopy(ptr->lib->logDesc, &data[i], sof); 4829 i += sof; 4830 } 4831 IPFW_RUNLOCK(&layer3_chain); 4832 error = sooptcopyout(sopt, data, size); 4833 free(data, M_IPFW); 4834 } 4835 break; 4836#endif 4837 4838 default: 4839 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4840 error = EINVAL; 4841 } 4842 4843 return (error); 4844#undef RULE_MAXSIZE 4845} 4846 4847/** 4848 * dummynet needs a reference to the default rule, because rules can be 4849 * deleted while packets hold a reference to them. When this happens, 4850 * dummynet changes the reference to the default rule (it could well be a 4851 * NULL pointer, but this way we do not need to check for the special 4852 * case, plus here he have info on the default behaviour). 4853 */ 4854struct ip_fw *ip_fw_default_rule; 4855 4856/* 4857 * This procedure is only used to handle keepalives. It is invoked 4858 * every dyn_keepalive_period 4859 */ 4860static void 4861ipfw_tick(void * __unused unused) 4862{ 4863 struct mbuf *m0, *m, *mnext, **mtailp; 4864 int i; 4865 ipfw_dyn_rule *q; 4866 4867 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 4868 goto done; 4869 4870 /* 4871 * We make a chain of packets to go out here -- not deferring 4872 * until after we drop the IPFW dynamic rule lock would result 4873 * in a lock order reversal with the normal packet input -> ipfw 4874 * call stack. 4875 */ 4876 m0 = NULL; 4877 mtailp = &m0; 4878 IPFW_DYN_LOCK(); 4879 for (i = 0 ; i < curr_dyn_buckets ; i++) { 4880 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 4881 if (q->dyn_type == O_LIMIT_PARENT) 4882 continue; 4883 if (q->id.proto != IPPROTO_TCP) 4884 continue; 4885 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4886 continue; 4887 if (TIME_LEQ( time_uptime+dyn_keepalive_interval, 4888 q->expire)) 4889 continue; /* too early */ 4890 if (TIME_LEQ(q->expire, time_uptime)) 4891 continue; /* too late, rule expired */ 4892 4893 *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1, 4894 q->ack_fwd, TH_SYN); 4895 if (*mtailp != NULL) 4896 mtailp = &(*mtailp)->m_nextpkt; 4897 *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1, 4898 q->ack_rev, 0); 4899 if (*mtailp != NULL) 4900 mtailp = &(*mtailp)->m_nextpkt; 4901 } 4902 } 4903 IPFW_DYN_UNLOCK(); 4904 for (m = mnext = m0; m != NULL; m = mnext) { 4905 mnext = m->m_nextpkt; 4906 m->m_nextpkt = NULL; 4907 ip_output(m, NULL, NULL, 0, NULL, NULL); 4908 } 4909done: 4910 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 4911} 4912 4913int 4914ipfw_init(void) 4915{ 4916 struct ip_fw default_rule; 4917 int error; 4918 4919#ifdef INET6 4920 /* Setup IPv6 fw sysctl tree. */ 4921 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4922 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4923 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4924 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4925 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4926 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4927 &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4928 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4929 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4930 &fw_deny_unknown_exthdrs, 0, 4931 "Deny packets with unknown IPv6 Extension Headers"); 4932#endif 4933 4934 layer3_chain.rules = NULL; 4935 IPFW_LOCK_INIT(&layer3_chain); 4936 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule", 4937 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4938 UMA_ALIGN_PTR, 0); 4939 IPFW_DYN_LOCK_INIT(); 4940 callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE); 4941 4942 bzero(&default_rule, sizeof default_rule); 4943 4944 default_rule.act_ofs = 0; 4945 default_rule.rulenum = IPFW_DEFAULT_RULE; 4946 default_rule.cmd_len = 1; 4947 default_rule.set = RESVD_SET; 4948 4949 default_rule.cmd[0].len = 1; 4950 default_rule.cmd[0].opcode = 4951#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4952 1 ? O_ACCEPT : 4953#endif 4954 O_DENY; 4955 4956 error = add_rule(&layer3_chain, &default_rule); 4957 if (error != 0) { 4958 printf("ipfw2: error %u initializing default rule " 4959 "(support disabled)\n", error); 4960 IPFW_DYN_LOCK_DESTROY(); 4961 IPFW_LOCK_DESTROY(&layer3_chain); 4962 uma_zdestroy(ipfw_dyn_rule_zone); 4963 return (error); 4964 } 4965 4966 ip_fw_default_rule = layer3_chain.rules; 4967 printf("ipfw2 " 4968#ifdef INET6 4969 "(+ipv6) " 4970#endif 4971 "initialized, divert %s, " 4972 "rule-based forwarding " 4973#ifdef IPFIREWALL_FORWARD 4974 "enabled, " 4975#else 4976 "disabled, " 4977#endif 4978 "default to %s, logging ", 4979#ifdef IPDIVERT 4980 "enabled", 4981#else 4982 "loadable", 4983#endif 4984 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4985 4986#ifdef IPFIREWALL_VERBOSE 4987 fw_verbose = 1; 4988#endif 4989#ifdef IPFIREWALL_VERBOSE_LIMIT 4990 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4991#endif 4992 if (fw_verbose == 0) 4993 printf("disabled\n"); 4994 else if (verbose_limit == 0) 4995 printf("unlimited\n"); 4996 else 4997 printf("limited to %d packets/entry by default\n", 4998 verbose_limit); 4999 5000 error = init_tables(&layer3_chain); 5001 if (error) { 5002 IPFW_DYN_LOCK_DESTROY(); 5003 IPFW_LOCK_DESTROY(&layer3_chain); 5004 uma_zdestroy(ipfw_dyn_rule_zone); 5005 return (error); 5006 } 5007 ip_fw_ctl_ptr = ipfw_ctl; 5008 ip_fw_chk_ptr = ipfw_chk; 5009 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 5010#ifdef IPFIREWALL_NAT 5011 LIST_INIT(&layer3_chain.nat); 5012 ifaddr_event_tag = EVENTHANDLER_REGISTER(ifaddr_event, ifaddr_change, 5013 NULL, EVENTHANDLER_PRI_ANY); 5014#endif 5015 return (0); 5016} 5017 5018void 5019ipfw_destroy(void) 5020{ 5021 struct ip_fw *reap; 5022#ifdef IPFIREWALL_NAT 5023 struct cfg_nat *ptr, *ptr_temp; 5024#endif 5025 5026 ip_fw_chk_ptr = NULL; 5027 ip_fw_ctl_ptr = NULL; 5028 callout_drain(&ipfw_timeout); 5029 IPFW_WLOCK(&layer3_chain); 5030 flush_tables(&layer3_chain); 5031#ifdef IPFIREWALL_NAT 5032 LIST_FOREACH_SAFE(ptr, &layer3_chain.nat, _next, ptr_temp) { 5033 LIST_REMOVE(ptr, _next); 5034 del_redir_spool_cfg(ptr, &ptr->redir_chain); 5035 LibAliasUninit(ptr->lib); 5036 free(ptr, M_IPFW); 5037 } 5038 EVENTHANDLER_DEREGISTER(ifaddr_event, ifaddr_event_tag); 5039#endif 5040 layer3_chain.reap = NULL; 5041 free_chain(&layer3_chain, 1 /* kill default rule */); 5042 reap = layer3_chain.reap, layer3_chain.reap = NULL; 5043 IPFW_WUNLOCK(&layer3_chain); 5044 if (reap != NULL) 5045 reap_rules(reap); 5046 IPFW_DYN_LOCK_DESTROY(); 5047 uma_zdestroy(ipfw_dyn_rule_zone); 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