ip_fw2.c revision 159398
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 159398 2006-06-08 11:27:45Z oleg $ 26 */ 27 28#define DEB(x) 29#define DDB(x) x 30 31/* 32 * Implement IP packet firewall (new version) 33 */ 34 35#if !defined(KLD_MODULE) 36#include "opt_ipfw.h" 37#include "opt_ipdn.h" 38#include "opt_inet.h" 39#ifndef INET 40#error IPFIREWALL requires INET. 41#endif /* INET */ 42#endif 43#include "opt_inet6.h" 44#include "opt_ipsec.h" 45 46#include <sys/param.h> 47#include <sys/systm.h> 48#include <sys/condvar.h> 49#include <sys/malloc.h> 50#include <sys/mbuf.h> 51#include <sys/kernel.h> 52#include <sys/lock.h> 53#include <sys/jail.h> 54#include <sys/module.h> 55#include <sys/proc.h> 56#include <sys/rwlock.h> 57#include <sys/socket.h> 58#include <sys/socketvar.h> 59#include <sys/sysctl.h> 60#include <sys/syslog.h> 61#include <sys/ucred.h> 62#include <net/if.h> 63#include <net/radix.h> 64#include <net/route.h> 65#include <netinet/in.h> 66#include <netinet/in_systm.h> 67#include <netinet/in_var.h> 68#include <netinet/in_pcb.h> 69#include <netinet/ip.h> 70#include <netinet/ip_var.h> 71#include <netinet/ip_icmp.h> 72#include <netinet/ip_fw.h> 73#include <netinet/ip_divert.h> 74#include <netinet/ip_dummynet.h> 75#include <netinet/tcp.h> 76#include <netinet/tcp_timer.h> 77#include <netinet/tcp_var.h> 78#include <netinet/tcpip.h> 79#include <netinet/udp.h> 80#include <netinet/udp_var.h> 81 82#include <netgraph/ng_ipfw.h> 83 84#include <altq/if_altq.h> 85 86#ifdef IPSEC 87#include <netinet6/ipsec.h> 88#endif 89 90#include <netinet/ip6.h> 91#include <netinet/icmp6.h> 92#ifdef INET6 93#include <netinet6/scope6_var.h> 94#endif 95 96#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 97 98#include <machine/in_cksum.h> /* XXX for in_cksum */ 99 100/* 101 * set_disable contains one bit per set value (0..31). 102 * If the bit is set, all rules with the corresponding set 103 * are disabled. Set RESVD_SET(31) is reserved for the default rule 104 * and rules that are not deleted by the flush command, 105 * and CANNOT be disabled. 106 * Rules in set RESVD_SET can only be deleted explicitly. 107 */ 108static u_int32_t set_disable; 109 110static int fw_verbose; 111static int verbose_limit; 112 113static struct callout ipfw_timeout; 114static uma_zone_t ipfw_dyn_rule_zone; 115#define IPFW_DEFAULT_RULE 65535 116 117/* 118 * Data structure to cache our ucred related 119 * information. This structure only gets used if 120 * the user specified UID/GID based constraints in 121 * a firewall rule. 122 */ 123struct ip_fw_ugid { 124 gid_t fw_groups[NGROUPS]; 125 int fw_ngroups; 126 uid_t fw_uid; 127 int fw_prid; 128}; 129 130#define IPFW_TABLES_MAX 128 131struct ip_fw_chain { 132 struct ip_fw *rules; /* list of rules */ 133 struct ip_fw *reap; /* list of rules to reap */ 134 struct radix_node_head *tables[IPFW_TABLES_MAX]; 135 struct rwlock rwmtx; 136}; 137#define IPFW_LOCK_INIT(_chain) \ 138 rw_init(&(_chain)->rwmtx, "IPFW static rules") 139#define IPFW_LOCK_DESTROY(_chain) rw_destroy(&(_chain)->rwmtx) 140#define IPFW_WLOCK_ASSERT(_chain) do { \ 141 rw_assert(&(_chain)->rwmtx, RA_WLOCKED); \ 142 NET_ASSERT_GIANT(); \ 143} while (0) 144 145#define IPFW_RLOCK(p) rw_rlock(&(p)->rwmtx) 146#define IPFW_RUNLOCK(p) rw_runlock(&(p)->rwmtx) 147#define IPFW_WLOCK(p) rw_wlock(&(p)->rwmtx) 148#define IPFW_WUNLOCK(p) rw_wunlock(&(p)->rwmtx) 149 150/* 151 * list of rules for layer 3 152 */ 153static struct ip_fw_chain layer3_chain; 154 155MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 156MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 157 158struct table_entry { 159 struct radix_node rn[2]; 160 struct sockaddr_in addr, mask; 161 u_int32_t value; 162}; 163 164static int fw_debug = 1; 165static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 166 167extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS); 168 169#ifdef SYSCTL_NODE 170SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 171SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable, 172 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &fw_enable, 0, 173 ipfw_chg_hook, "I", "Enable ipfw"); 174SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 175 &autoinc_step, 0, "Rule number autincrement step"); 176SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 177 CTLFLAG_RW | CTLFLAG_SECURE3, 178 &fw_one_pass, 0, 179 "Only do a single pass through ipfw when using dummynet(4)"); 180SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 181 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 182SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 183 CTLFLAG_RW | CTLFLAG_SECURE3, 184 &fw_verbose, 0, "Log matches to ipfw rules"); 185SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 186 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 187 188/* 189 * Description of dynamic rules. 190 * 191 * Dynamic rules are stored in lists accessed through a hash table 192 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 193 * be modified through the sysctl variable dyn_buckets which is 194 * updated when the table becomes empty. 195 * 196 * XXX currently there is only one list, ipfw_dyn. 197 * 198 * When a packet is received, its address fields are first masked 199 * with the mask defined for the rule, then hashed, then matched 200 * against the entries in the corresponding list. 201 * Dynamic rules can be used for different purposes: 202 * + stateful rules; 203 * + enforcing limits on the number of sessions; 204 * + in-kernel NAT (not implemented yet) 205 * 206 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 207 * measured in seconds and depending on the flags. 208 * 209 * The total number of dynamic rules is stored in dyn_count. 210 * The max number of dynamic rules is dyn_max. When we reach 211 * the maximum number of rules we do not create anymore. This is 212 * done to avoid consuming too much memory, but also too much 213 * time when searching on each packet (ideally, we should try instead 214 * to put a limit on the length of the list on each bucket...). 215 * 216 * Each dynamic rule holds a pointer to the parent ipfw rule so 217 * we know what action to perform. Dynamic rules are removed when 218 * the parent rule is deleted. XXX we should make them survive. 219 * 220 * There are some limitations with dynamic rules -- we do not 221 * obey the 'randomized match', and we do not do multiple 222 * passes through the firewall. XXX check the latter!!! 223 */ 224static ipfw_dyn_rule **ipfw_dyn_v = NULL; 225static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 226static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 227 228static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 229#define IPFW_DYN_LOCK_INIT() \ 230 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 231#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 232#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 233#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 234#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 235 236/* 237 * Timeouts for various events in handing dynamic rules. 238 */ 239static u_int32_t dyn_ack_lifetime = 300; 240static u_int32_t dyn_syn_lifetime = 20; 241static u_int32_t dyn_fin_lifetime = 1; 242static u_int32_t dyn_rst_lifetime = 1; 243static u_int32_t dyn_udp_lifetime = 10; 244static u_int32_t dyn_short_lifetime = 5; 245 246/* 247 * Keepalives are sent if dyn_keepalive is set. They are sent every 248 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 249 * seconds of lifetime of a rule. 250 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 251 * than dyn_keepalive_period. 252 */ 253 254static u_int32_t dyn_keepalive_interval = 20; 255static u_int32_t dyn_keepalive_period = 5; 256static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 257 258static u_int32_t static_count; /* # of static rules */ 259static u_int32_t static_len; /* size in bytes of static rules */ 260static u_int32_t dyn_count; /* # of dynamic rules */ 261static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 262 263SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 264 &dyn_buckets, 0, "Number of dyn. buckets"); 265SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 266 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 267SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 268 &dyn_count, 0, "Number of dyn. rules"); 269SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 270 &dyn_max, 0, "Max number of dyn. rules"); 271SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 272 &static_count, 0, "Number of static rules"); 273SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 274 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 275SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 276 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 277SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 278 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 279SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 280 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 281SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 282 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 283SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 284 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 285SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 286 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 287 288#ifdef INET6 289/* 290 * IPv6 specific variables 291 */ 292SYSCTL_DECL(_net_inet6_ip6); 293 294static struct sysctl_ctx_list ip6_fw_sysctl_ctx; 295static struct sysctl_oid *ip6_fw_sysctl_tree; 296#endif /* INET6 */ 297#endif /* SYSCTL_NODE */ 298 299static int fw_deny_unknown_exthdrs = 1; 300 301 302/* 303 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 304 * Other macros just cast void * into the appropriate type 305 */ 306#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 307#define TCP(p) ((struct tcphdr *)(p)) 308#define UDP(p) ((struct udphdr *)(p)) 309#define ICMP(p) ((struct icmphdr *)(p)) 310#define ICMP6(p) ((struct icmp6_hdr *)(p)) 311 312static __inline int 313icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 314{ 315 int type = icmp->icmp_type; 316 317 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 318} 319 320#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 321 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 322 323static int 324is_icmp_query(struct icmphdr *icmp) 325{ 326 int type = icmp->icmp_type; 327 328 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 329} 330#undef TT 331 332/* 333 * The following checks use two arrays of 8 or 16 bits to store the 334 * bits that we want set or clear, respectively. They are in the 335 * low and high half of cmd->arg1 or cmd->d[0]. 336 * 337 * We scan options and store the bits we find set. We succeed if 338 * 339 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 340 * 341 * The code is sometimes optimized not to store additional variables. 342 */ 343 344static int 345flags_match(ipfw_insn *cmd, u_int8_t bits) 346{ 347 u_char want_clear; 348 bits = ~bits; 349 350 if ( ((cmd->arg1 & 0xff) & bits) != 0) 351 return 0; /* some bits we want set were clear */ 352 want_clear = (cmd->arg1 >> 8) & 0xff; 353 if ( (want_clear & bits) != want_clear) 354 return 0; /* some bits we want clear were set */ 355 return 1; 356} 357 358static int 359ipopts_match(struct ip *ip, ipfw_insn *cmd) 360{ 361 int optlen, bits = 0; 362 u_char *cp = (u_char *)(ip + 1); 363 int x = (ip->ip_hl << 2) - sizeof (struct ip); 364 365 for (; x > 0; x -= optlen, cp += optlen) { 366 int opt = cp[IPOPT_OPTVAL]; 367 368 if (opt == IPOPT_EOL) 369 break; 370 if (opt == IPOPT_NOP) 371 optlen = 1; 372 else { 373 optlen = cp[IPOPT_OLEN]; 374 if (optlen <= 0 || optlen > x) 375 return 0; /* invalid or truncated */ 376 } 377 switch (opt) { 378 379 default: 380 break; 381 382 case IPOPT_LSRR: 383 bits |= IP_FW_IPOPT_LSRR; 384 break; 385 386 case IPOPT_SSRR: 387 bits |= IP_FW_IPOPT_SSRR; 388 break; 389 390 case IPOPT_RR: 391 bits |= IP_FW_IPOPT_RR; 392 break; 393 394 case IPOPT_TS: 395 bits |= IP_FW_IPOPT_TS; 396 break; 397 } 398 } 399 return (flags_match(cmd, bits)); 400} 401 402static int 403tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 404{ 405 int optlen, bits = 0; 406 u_char *cp = (u_char *)(tcp + 1); 407 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 408 409 for (; x > 0; x -= optlen, cp += optlen) { 410 int opt = cp[0]; 411 if (opt == TCPOPT_EOL) 412 break; 413 if (opt == TCPOPT_NOP) 414 optlen = 1; 415 else { 416 optlen = cp[1]; 417 if (optlen <= 0) 418 break; 419 } 420 421 switch (opt) { 422 423 default: 424 break; 425 426 case TCPOPT_MAXSEG: 427 bits |= IP_FW_TCPOPT_MSS; 428 break; 429 430 case TCPOPT_WINDOW: 431 bits |= IP_FW_TCPOPT_WINDOW; 432 break; 433 434 case TCPOPT_SACK_PERMITTED: 435 case TCPOPT_SACK: 436 bits |= IP_FW_TCPOPT_SACK; 437 break; 438 439 case TCPOPT_TIMESTAMP: 440 bits |= IP_FW_TCPOPT_TS; 441 break; 442 443 } 444 } 445 return (flags_match(cmd, bits)); 446} 447 448static int 449iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 450{ 451 if (ifp == NULL) /* no iface with this packet, match fails */ 452 return 0; 453 /* Check by name or by IP address */ 454 if (cmd->name[0] != '\0') { /* match by name */ 455 /* Check name */ 456 if (cmd->p.glob) { 457 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 458 return(1); 459 } else { 460 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 461 return(1); 462 } 463 } else { 464 struct ifaddr *ia; 465 466 /* XXX lock? */ 467 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 468 if (ia->ifa_addr == NULL) 469 continue; 470 if (ia->ifa_addr->sa_family != AF_INET) 471 continue; 472 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 473 (ia->ifa_addr))->sin_addr.s_addr) 474 return(1); /* match */ 475 } 476 } 477 return(0); /* no match, fail ... */ 478} 479 480/* 481 * The verify_path function checks if a route to the src exists and 482 * if it is reachable via ifp (when provided). 483 * 484 * The 'verrevpath' option checks that the interface that an IP packet 485 * arrives on is the same interface that traffic destined for the 486 * packet's source address would be routed out of. The 'versrcreach' 487 * option just checks that the source address is reachable via any route 488 * (except default) in the routing table. These two are a measure to block 489 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 490 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 491 * is purposely reminiscent of the Cisco IOS command, 492 * 493 * ip verify unicast reverse-path 494 * ip verify unicast source reachable-via any 495 * 496 * which implements the same functionality. But note that syntax is 497 * misleading. The check may be performed on all IP packets whether unicast, 498 * multicast, or broadcast. 499 */ 500static int 501verify_path(struct in_addr src, struct ifnet *ifp) 502{ 503 struct route ro; 504 struct sockaddr_in *dst; 505 506 bzero(&ro, sizeof(ro)); 507 508 dst = (struct sockaddr_in *)&(ro.ro_dst); 509 dst->sin_family = AF_INET; 510 dst->sin_len = sizeof(*dst); 511 dst->sin_addr = src; 512 rtalloc_ign(&ro, RTF_CLONING); 513 514 if (ro.ro_rt == NULL) 515 return 0; 516 517 /* 518 * If ifp is provided, check for equality with rtentry. 519 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 520 * in order to pass packets injected back by if_simloop(): 521 * if useloopback == 1 routing entry (via lo0) for our own address 522 * may exist, so we need to handle routing assymetry. 523 */ 524 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 525 RTFREE(ro.ro_rt); 526 return 0; 527 } 528 529 /* if no ifp provided, check if rtentry is not default route */ 530 if (ifp == NULL && 531 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 532 RTFREE(ro.ro_rt); 533 return 0; 534 } 535 536 /* or if this is a blackhole/reject route */ 537 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 538 RTFREE(ro.ro_rt); 539 return 0; 540 } 541 542 /* found valid route */ 543 RTFREE(ro.ro_rt); 544 return 1; 545} 546 547#ifdef INET6 548/* 549 * ipv6 specific rules here... 550 */ 551static __inline int 552icmp6type_match (int type, ipfw_insn_u32 *cmd) 553{ 554 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 555} 556 557static int 558flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 559{ 560 int i; 561 for (i=0; i <= cmd->o.arg1; ++i ) 562 if (curr_flow == cmd->d[i] ) 563 return 1; 564 return 0; 565} 566 567/* support for IP6_*_ME opcodes */ 568static int 569search_ip6_addr_net (struct in6_addr * ip6_addr) 570{ 571 struct ifnet *mdc; 572 struct ifaddr *mdc2; 573 struct in6_ifaddr *fdm; 574 struct in6_addr copia; 575 576 TAILQ_FOREACH(mdc, &ifnet, if_link) 577 for (mdc2 = mdc->if_addrlist.tqh_first; mdc2; 578 mdc2 = mdc2->ifa_list.tqe_next) { 579 if (!mdc2->ifa_addr) 580 continue; 581 if (mdc2->ifa_addr->sa_family == AF_INET6) { 582 fdm = (struct in6_ifaddr *)mdc2; 583 copia = fdm->ia_addr.sin6_addr; 584 /* need for leaving scope_id in the sock_addr */ 585 in6_clearscope(&copia); 586 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) 587 return 1; 588 } 589 } 590 return 0; 591} 592 593static int 594verify_path6(struct in6_addr *src, struct ifnet *ifp) 595{ 596 struct route_in6 ro; 597 struct sockaddr_in6 *dst; 598 599 bzero(&ro, sizeof(ro)); 600 601 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 602 dst->sin6_family = AF_INET6; 603 dst->sin6_len = sizeof(*dst); 604 dst->sin6_addr = *src; 605 rtalloc_ign((struct route *)&ro, RTF_CLONING); 606 607 if (ro.ro_rt == NULL) 608 return 0; 609 610 /* 611 * if ifp is provided, check for equality with rtentry 612 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 613 * to support the case of sending packets to an address of our own. 614 * (where the former interface is the first argument of if_simloop() 615 * (=ifp), the latter is lo0) 616 */ 617 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 618 RTFREE(ro.ro_rt); 619 return 0; 620 } 621 622 /* if no ifp provided, check if rtentry is not default route */ 623 if (ifp == NULL && 624 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 625 RTFREE(ro.ro_rt); 626 return 0; 627 } 628 629 /* or if this is a blackhole/reject route */ 630 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 631 RTFREE(ro.ro_rt); 632 return 0; 633 } 634 635 /* found valid route */ 636 RTFREE(ro.ro_rt); 637 return 1; 638 639} 640static __inline int 641hash_packet6(struct ipfw_flow_id *id) 642{ 643 u_int32_t i; 644 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 645 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 646 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 647 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 648 (id->dst_port) ^ (id->src_port); 649 return i; 650} 651 652static int 653is_icmp6_query(int icmp6_type) 654{ 655 if ((icmp6_type <= ICMP6_MAXTYPE) && 656 (icmp6_type == ICMP6_ECHO_REQUEST || 657 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 658 icmp6_type == ICMP6_WRUREQUEST || 659 icmp6_type == ICMP6_FQDN_QUERY || 660 icmp6_type == ICMP6_NI_QUERY)) 661 return (1); 662 663 return (0); 664} 665 666static void 667send_reject6(struct ip_fw_args *args, int code, u_short offset, u_int hlen) 668{ 669 if (code == ICMP6_UNREACH_RST && offset == 0 && 670 args->f_id.proto == IPPROTO_TCP) { 671 struct ip6_hdr *ip6; 672 struct tcphdr *tcp; 673 tcp_seq ack, seq; 674 int flags; 675 struct { 676 struct ip6_hdr ip6; 677 struct tcphdr th; 678 } ti; 679 680 if (args->m->m_len < (hlen+sizeof(struct tcphdr))) { 681 args->m = m_pullup(args->m, hlen+sizeof(struct tcphdr)); 682 if (args->m == NULL) 683 return; 684 } 685 686 ip6 = mtod(args->m, struct ip6_hdr *); 687 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 688 689 if ((tcp->th_flags & TH_RST) != 0) { 690 m_freem(args->m); 691 return; 692 } 693 694 ti.ip6 = *ip6; 695 ti.th = *tcp; 696 ti.th.th_seq = ntohl(ti.th.th_seq); 697 ti.th.th_ack = ntohl(ti.th.th_ack); 698 ti.ip6.ip6_nxt = IPPROTO_TCP; 699 700 if (ti.th.th_flags & TH_ACK) { 701 ack = 0; 702 seq = ti.th.th_ack; 703 flags = TH_RST; 704 } else { 705 ack = ti.th.th_seq; 706 if (((args->m)->m_flags & M_PKTHDR) != 0) { 707 ack += (args->m)->m_pkthdr.len - hlen 708 - (ti.th.th_off << 2); 709 } else if (ip6->ip6_plen) { 710 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) 711 - hlen - (ti.th.th_off << 2); 712 } else { 713 m_freem(args->m); 714 return; 715 } 716 if (tcp->th_flags & TH_SYN) 717 ack++; 718 seq = 0; 719 flags = TH_RST|TH_ACK; 720 } 721 bcopy(&ti, ip6, sizeof(ti)); 722 tcp_respond(NULL, ip6, (struct tcphdr *)(ip6 + 1), 723 args->m, ack, seq, flags); 724 725 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 726 icmp6_error(args->m, ICMP6_DST_UNREACH, code, 0); 727 728 } else 729 m_freem(args->m); 730 731 args->m = NULL; 732} 733 734#endif /* INET6 */ 735 736static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 737 738#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 739#define SNP(buf) buf, sizeof(buf) 740 741/* 742 * We enter here when we have a rule with O_LOG. 743 * XXX this function alone takes about 2Kbytes of code! 744 */ 745static void 746ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 747 struct mbuf *m, struct ifnet *oif, u_short offset) 748{ 749 struct ether_header *eh = args->eh; 750 char *action; 751 int limit_reached = 0; 752 char action2[40], proto[128], fragment[32]; 753 754 fragment[0] = '\0'; 755 proto[0] = '\0'; 756 757 if (f == NULL) { /* bogus pkt */ 758 if (verbose_limit != 0 && norule_counter >= verbose_limit) 759 return; 760 norule_counter++; 761 if (norule_counter == verbose_limit) 762 limit_reached = verbose_limit; 763 action = "Refuse"; 764 } else { /* O_LOG is the first action, find the real one */ 765 ipfw_insn *cmd = ACTION_PTR(f); 766 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 767 768 if (l->max_log != 0 && l->log_left == 0) 769 return; 770 l->log_left--; 771 if (l->log_left == 0) 772 limit_reached = l->max_log; 773 cmd += F_LEN(cmd); /* point to first action */ 774 if (cmd->opcode == O_ALTQ) { 775 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 776 777 snprintf(SNPARGS(action2, 0), "Altq %d", 778 altq->qid); 779 cmd += F_LEN(cmd); 780 } 781 if (cmd->opcode == O_PROB) 782 cmd += F_LEN(cmd); 783 784 if (cmd->opcode == O_TAG) 785 cmd += F_LEN(cmd); 786 787 action = action2; 788 switch (cmd->opcode) { 789 case O_DENY: 790 action = "Deny"; 791 break; 792 793 case O_REJECT: 794 if (cmd->arg1==ICMP_REJECT_RST) 795 action = "Reset"; 796 else if (cmd->arg1==ICMP_UNREACH_HOST) 797 action = "Reject"; 798 else 799 snprintf(SNPARGS(action2, 0), "Unreach %d", 800 cmd->arg1); 801 break; 802 803 case O_UNREACH6: 804 if (cmd->arg1==ICMP6_UNREACH_RST) 805 action = "Reset"; 806 else 807 snprintf(SNPARGS(action2, 0), "Unreach %d", 808 cmd->arg1); 809 break; 810 811 case O_ACCEPT: 812 action = "Accept"; 813 break; 814 case O_COUNT: 815 action = "Count"; 816 break; 817 case O_DIVERT: 818 snprintf(SNPARGS(action2, 0), "Divert %d", 819 cmd->arg1); 820 break; 821 case O_TEE: 822 snprintf(SNPARGS(action2, 0), "Tee %d", 823 cmd->arg1); 824 break; 825 case O_SKIPTO: 826 snprintf(SNPARGS(action2, 0), "SkipTo %d", 827 cmd->arg1); 828 break; 829 case O_PIPE: 830 snprintf(SNPARGS(action2, 0), "Pipe %d", 831 cmd->arg1); 832 break; 833 case O_QUEUE: 834 snprintf(SNPARGS(action2, 0), "Queue %d", 835 cmd->arg1); 836 break; 837 case O_FORWARD_IP: { 838 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 839 int len; 840 841 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 842 inet_ntoa(sa->sa.sin_addr)); 843 if (sa->sa.sin_port) 844 snprintf(SNPARGS(action2, len), ":%d", 845 sa->sa.sin_port); 846 } 847 break; 848 case O_NETGRAPH: 849 snprintf(SNPARGS(action2, 0), "Netgraph %d", 850 cmd->arg1); 851 break; 852 case O_NGTEE: 853 snprintf(SNPARGS(action2, 0), "Ngtee %d", 854 cmd->arg1); 855 break; 856 default: 857 action = "UNKNOWN"; 858 break; 859 } 860 } 861 862 if (hlen == 0) { /* non-ip */ 863 snprintf(SNPARGS(proto, 0), "MAC"); 864 865 } else { 866 int len; 867 char src[48], dst[48]; 868 struct icmphdr *icmp; 869 struct tcphdr *tcp; 870 struct udphdr *udp; 871 /* Initialize to make compiler happy. */ 872 struct ip *ip = NULL; 873#ifdef INET6 874 struct ip6_hdr *ip6 = NULL; 875 struct icmp6_hdr *icmp6; 876#endif 877 src[0] = '\0'; 878 dst[0] = '\0'; 879#ifdef INET6 880 if (args->f_id.addr_type == 6) { 881 snprintf(src, sizeof(src), "[%s]", 882 ip6_sprintf(&args->f_id.src_ip6)); 883 snprintf(dst, sizeof(dst), "[%s]", 884 ip6_sprintf(&args->f_id.dst_ip6)); 885 886 ip6 = (struct ip6_hdr *)mtod(m, struct ip6_hdr *); 887 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 888 udp = (struct udphdr *)(mtod(args->m, char *) + hlen); 889 } else 890#endif 891 { 892 ip = mtod(m, struct ip *); 893 tcp = L3HDR(struct tcphdr, ip); 894 udp = L3HDR(struct udphdr, ip); 895 896 inet_ntoa_r(ip->ip_src, src); 897 inet_ntoa_r(ip->ip_dst, dst); 898 } 899 900 switch (args->f_id.proto) { 901 case IPPROTO_TCP: 902 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 903 if (offset == 0) 904 snprintf(SNPARGS(proto, len), ":%d %s:%d", 905 ntohs(tcp->th_sport), 906 dst, 907 ntohs(tcp->th_dport)); 908 else 909 snprintf(SNPARGS(proto, len), " %s", dst); 910 break; 911 912 case IPPROTO_UDP: 913 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 914 if (offset == 0) 915 snprintf(SNPARGS(proto, len), ":%d %s:%d", 916 ntohs(udp->uh_sport), 917 dst, 918 ntohs(udp->uh_dport)); 919 else 920 snprintf(SNPARGS(proto, len), " %s", dst); 921 break; 922 923 case IPPROTO_ICMP: 924 icmp = L3HDR(struct icmphdr, ip); 925 if (offset == 0) 926 len = snprintf(SNPARGS(proto, 0), 927 "ICMP:%u.%u ", 928 icmp->icmp_type, icmp->icmp_code); 929 else 930 len = snprintf(SNPARGS(proto, 0), "ICMP "); 931 len += snprintf(SNPARGS(proto, len), "%s", src); 932 snprintf(SNPARGS(proto, len), " %s", dst); 933 break; 934#ifdef INET6 935 case IPPROTO_ICMPV6: 936 icmp6 = (struct icmp6_hdr *)(mtod(args->m, char *) + hlen); 937 if (offset == 0) 938 len = snprintf(SNPARGS(proto, 0), 939 "ICMPv6:%u.%u ", 940 icmp6->icmp6_type, icmp6->icmp6_code); 941 else 942 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 943 len += snprintf(SNPARGS(proto, len), "%s", src); 944 snprintf(SNPARGS(proto, len), " %s", dst); 945 break; 946#endif 947 default: 948 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 949 args->f_id.proto, src); 950 snprintf(SNPARGS(proto, len), " %s", dst); 951 break; 952 } 953 954#ifdef INET6 955 if (args->f_id.addr_type == 6) { 956 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 957 snprintf(SNPARGS(fragment, 0), 958 " (frag %08x:%d@%d%s)", 959 args->f_id.frag_id6, 960 ntohs(ip6->ip6_plen) - hlen, 961 ntohs(offset & IP6F_OFF_MASK) << 3, 962 (offset & IP6F_MORE_FRAG) ? "+" : ""); 963 } else 964#endif 965 { 966 int ip_off, ip_len; 967 if (eh != NULL) { /* layer 2 packets are as on the wire */ 968 ip_off = ntohs(ip->ip_off); 969 ip_len = ntohs(ip->ip_len); 970 } else { 971 ip_off = ip->ip_off; 972 ip_len = ip->ip_len; 973 } 974 if (ip_off & (IP_MF | IP_OFFMASK)) 975 snprintf(SNPARGS(fragment, 0), 976 " (frag %d:%d@%d%s)", 977 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 978 offset << 3, 979 (ip_off & IP_MF) ? "+" : ""); 980 } 981 } 982 if (oif || m->m_pkthdr.rcvif) 983 log(LOG_SECURITY | LOG_INFO, 984 "ipfw: %d %s %s %s via %s%s\n", 985 f ? f->rulenum : -1, 986 action, proto, oif ? "out" : "in", 987 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 988 fragment); 989 else 990 log(LOG_SECURITY | LOG_INFO, 991 "ipfw: %d %s %s [no if info]%s\n", 992 f ? f->rulenum : -1, 993 action, proto, fragment); 994 if (limit_reached) 995 log(LOG_SECURITY | LOG_NOTICE, 996 "ipfw: limit %d reached on entry %d\n", 997 limit_reached, f ? f->rulenum : -1); 998} 999 1000/* 1001 * IMPORTANT: the hash function for dynamic rules must be commutative 1002 * in source and destination (ip,port), because rules are bidirectional 1003 * and we want to find both in the same bucket. 1004 */ 1005static __inline int 1006hash_packet(struct ipfw_flow_id *id) 1007{ 1008 u_int32_t i; 1009 1010#ifdef INET6 1011 if (IS_IP6_FLOW_ID(id)) 1012 i = hash_packet6(id); 1013 else 1014#endif /* INET6 */ 1015 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1016 i &= (curr_dyn_buckets - 1); 1017 return i; 1018} 1019 1020/** 1021 * unlink a dynamic rule from a chain. prev is a pointer to 1022 * the previous one, q is a pointer to the rule to delete, 1023 * head is a pointer to the head of the queue. 1024 * Modifies q and potentially also head. 1025 */ 1026#define UNLINK_DYN_RULE(prev, head, q) { \ 1027 ipfw_dyn_rule *old_q = q; \ 1028 \ 1029 /* remove a refcount to the parent */ \ 1030 if (q->dyn_type == O_LIMIT) \ 1031 q->parent->count--; \ 1032 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1033 (q->id.src_ip), (q->id.src_port), \ 1034 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 1035 if (prev != NULL) \ 1036 prev->next = q = q->next; \ 1037 else \ 1038 head = q = q->next; \ 1039 dyn_count--; \ 1040 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1041 1042#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1043 1044/** 1045 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1046 * 1047 * If keep_me == NULL, rules are deleted even if not expired, 1048 * otherwise only expired rules are removed. 1049 * 1050 * The value of the second parameter is also used to point to identify 1051 * a rule we absolutely do not want to remove (e.g. because we are 1052 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1053 * rules). The pointer is only used for comparison, so any non-null 1054 * value will do. 1055 */ 1056static void 1057remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1058{ 1059 static u_int32_t last_remove = 0; 1060 1061#define FORCE (keep_me == NULL) 1062 1063 ipfw_dyn_rule *prev, *q; 1064 int i, pass = 0, max_pass = 0; 1065 1066 IPFW_DYN_LOCK_ASSERT(); 1067 1068 if (ipfw_dyn_v == NULL || dyn_count == 0) 1069 return; 1070 /* do not expire more than once per second, it is useless */ 1071 if (!FORCE && last_remove == time_uptime) 1072 return; 1073 last_remove = time_uptime; 1074 1075 /* 1076 * because O_LIMIT refer to parent rules, during the first pass only 1077 * remove child and mark any pending LIMIT_PARENT, and remove 1078 * them in a second pass. 1079 */ 1080next_pass: 1081 for (i = 0 ; i < curr_dyn_buckets ; i++) { 1082 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 1083 /* 1084 * Logic can become complex here, so we split tests. 1085 */ 1086 if (q == keep_me) 1087 goto next; 1088 if (rule != NULL && rule != q->rule) 1089 goto next; /* not the one we are looking for */ 1090 if (q->dyn_type == O_LIMIT_PARENT) { 1091 /* 1092 * handle parent in the second pass, 1093 * record we need one. 1094 */ 1095 max_pass = 1; 1096 if (pass == 0) 1097 goto next; 1098 if (FORCE && q->count != 0 ) { 1099 /* XXX should not happen! */ 1100 printf("ipfw: OUCH! cannot remove rule," 1101 " count %d\n", q->count); 1102 } 1103 } else { 1104 if (!FORCE && 1105 !TIME_LEQ( q->expire, time_uptime )) 1106 goto next; 1107 } 1108 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1109 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1110 continue; 1111 } 1112next: 1113 prev=q; 1114 q=q->next; 1115 } 1116 } 1117 if (pass++ < max_pass) 1118 goto next_pass; 1119} 1120 1121 1122/** 1123 * lookup a dynamic rule. 1124 */ 1125static ipfw_dyn_rule * 1126lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1127 struct tcphdr *tcp) 1128{ 1129 /* 1130 * stateful ipfw extensions. 1131 * Lookup into dynamic session queue 1132 */ 1133#define MATCH_REVERSE 0 1134#define MATCH_FORWARD 1 1135#define MATCH_NONE 2 1136#define MATCH_UNKNOWN 3 1137 int i, dir = MATCH_NONE; 1138 ipfw_dyn_rule *prev, *q=NULL; 1139 1140 IPFW_DYN_LOCK_ASSERT(); 1141 1142 if (ipfw_dyn_v == NULL) 1143 goto done; /* not found */ 1144 i = hash_packet( pkt ); 1145 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 1146 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1147 goto next; 1148 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1149 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1150 continue; 1151 } 1152 if (pkt->proto == q->id.proto && 1153 q->dyn_type != O_LIMIT_PARENT) { 1154 if (IS_IP6_FLOW_ID(pkt)) { 1155 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1156 &(q->id.src_ip6)) && 1157 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1158 &(q->id.dst_ip6)) && 1159 pkt->src_port == q->id.src_port && 1160 pkt->dst_port == q->id.dst_port ) { 1161 dir = MATCH_FORWARD; 1162 break; 1163 } 1164 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1165 &(q->id.dst_ip6)) && 1166 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1167 &(q->id.src_ip6)) && 1168 pkt->src_port == q->id.dst_port && 1169 pkt->dst_port == q->id.src_port ) { 1170 dir = MATCH_REVERSE; 1171 break; 1172 } 1173 } else { 1174 if (pkt->src_ip == q->id.src_ip && 1175 pkt->dst_ip == q->id.dst_ip && 1176 pkt->src_port == q->id.src_port && 1177 pkt->dst_port == q->id.dst_port ) { 1178 dir = MATCH_FORWARD; 1179 break; 1180 } 1181 if (pkt->src_ip == q->id.dst_ip && 1182 pkt->dst_ip == q->id.src_ip && 1183 pkt->src_port == q->id.dst_port && 1184 pkt->dst_port == q->id.src_port ) { 1185 dir = MATCH_REVERSE; 1186 break; 1187 } 1188 } 1189 } 1190next: 1191 prev = q; 1192 q = q->next; 1193 } 1194 if (q == NULL) 1195 goto done; /* q = NULL, not found */ 1196 1197 if ( prev != NULL) { /* found and not in front */ 1198 prev->next = q->next; 1199 q->next = ipfw_dyn_v[i]; 1200 ipfw_dyn_v[i] = q; 1201 } 1202 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1203 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1204 1205#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1206#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1207 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1208 switch (q->state) { 1209 case TH_SYN: /* opening */ 1210 q->expire = time_uptime + dyn_syn_lifetime; 1211 break; 1212 1213 case BOTH_SYN: /* move to established */ 1214 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1215 case BOTH_SYN | (TH_FIN << 8) : 1216 if (tcp) { 1217#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1218 u_int32_t ack = ntohl(tcp->th_ack); 1219 if (dir == MATCH_FORWARD) { 1220 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1221 q->ack_fwd = ack; 1222 else { /* ignore out-of-sequence */ 1223 break; 1224 } 1225 } else { 1226 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1227 q->ack_rev = ack; 1228 else { /* ignore out-of-sequence */ 1229 break; 1230 } 1231 } 1232 } 1233 q->expire = time_uptime + dyn_ack_lifetime; 1234 break; 1235 1236 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1237 if (dyn_fin_lifetime >= dyn_keepalive_period) 1238 dyn_fin_lifetime = dyn_keepalive_period - 1; 1239 q->expire = time_uptime + dyn_fin_lifetime; 1240 break; 1241 1242 default: 1243#if 0 1244 /* 1245 * reset or some invalid combination, but can also 1246 * occur if we use keep-state the wrong way. 1247 */ 1248 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1249 printf("invalid state: 0x%x\n", q->state); 1250#endif 1251 if (dyn_rst_lifetime >= dyn_keepalive_period) 1252 dyn_rst_lifetime = dyn_keepalive_period - 1; 1253 q->expire = time_uptime + dyn_rst_lifetime; 1254 break; 1255 } 1256 } else if (pkt->proto == IPPROTO_UDP) { 1257 q->expire = time_uptime + dyn_udp_lifetime; 1258 } else { 1259 /* other protocols */ 1260 q->expire = time_uptime + dyn_short_lifetime; 1261 } 1262done: 1263 if (match_direction) 1264 *match_direction = dir; 1265 return q; 1266} 1267 1268static ipfw_dyn_rule * 1269lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1270 struct tcphdr *tcp) 1271{ 1272 ipfw_dyn_rule *q; 1273 1274 IPFW_DYN_LOCK(); 1275 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1276 if (q == NULL) 1277 IPFW_DYN_UNLOCK(); 1278 /* NB: return table locked when q is not NULL */ 1279 return q; 1280} 1281 1282static void 1283realloc_dynamic_table(void) 1284{ 1285 IPFW_DYN_LOCK_ASSERT(); 1286 1287 /* 1288 * Try reallocation, make sure we have a power of 2 and do 1289 * not allow more than 64k entries. In case of overflow, 1290 * default to 1024. 1291 */ 1292 1293 if (dyn_buckets > 65536) 1294 dyn_buckets = 1024; 1295 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1296 dyn_buckets = curr_dyn_buckets; /* reset */ 1297 return; 1298 } 1299 curr_dyn_buckets = dyn_buckets; 1300 if (ipfw_dyn_v != NULL) 1301 free(ipfw_dyn_v, M_IPFW); 1302 for (;;) { 1303 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1304 M_IPFW, M_NOWAIT | M_ZERO); 1305 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 1306 break; 1307 curr_dyn_buckets /= 2; 1308 } 1309} 1310 1311/** 1312 * Install state of type 'type' for a dynamic session. 1313 * The hash table contains two type of rules: 1314 * - regular rules (O_KEEP_STATE) 1315 * - rules for sessions with limited number of sess per user 1316 * (O_LIMIT). When they are created, the parent is 1317 * increased by 1, and decreased on delete. In this case, 1318 * the third parameter is the parent rule and not the chain. 1319 * - "parent" rules for the above (O_LIMIT_PARENT). 1320 */ 1321static ipfw_dyn_rule * 1322add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1323{ 1324 ipfw_dyn_rule *r; 1325 int i; 1326 1327 IPFW_DYN_LOCK_ASSERT(); 1328 1329 if (ipfw_dyn_v == NULL || 1330 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1331 realloc_dynamic_table(); 1332 if (ipfw_dyn_v == NULL) 1333 return NULL; /* failed ! */ 1334 } 1335 i = hash_packet(id); 1336 1337 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1338 if (r == NULL) { 1339 printf ("ipfw: sorry cannot allocate state\n"); 1340 return NULL; 1341 } 1342 1343 /* increase refcount on parent, and set pointer */ 1344 if (dyn_type == O_LIMIT) { 1345 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1346 if ( parent->dyn_type != O_LIMIT_PARENT) 1347 panic("invalid parent"); 1348 parent->count++; 1349 r->parent = parent; 1350 rule = parent->rule; 1351 } 1352 1353 r->id = *id; 1354 r->expire = time_uptime + dyn_syn_lifetime; 1355 r->rule = rule; 1356 r->dyn_type = dyn_type; 1357 r->pcnt = r->bcnt = 0; 1358 r->count = 0; 1359 1360 r->bucket = i; 1361 r->next = ipfw_dyn_v[i]; 1362 ipfw_dyn_v[i] = r; 1363 dyn_count++; 1364 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1365 dyn_type, 1366 (r->id.src_ip), (r->id.src_port), 1367 (r->id.dst_ip), (r->id.dst_port), 1368 dyn_count ); ) 1369 return r; 1370} 1371 1372/** 1373 * lookup dynamic parent rule using pkt and rule as search keys. 1374 * If the lookup fails, then install one. 1375 */ 1376static ipfw_dyn_rule * 1377lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1378{ 1379 ipfw_dyn_rule *q; 1380 int i; 1381 1382 IPFW_DYN_LOCK_ASSERT(); 1383 1384 if (ipfw_dyn_v) { 1385 int is_v6 = IS_IP6_FLOW_ID(pkt); 1386 i = hash_packet( pkt ); 1387 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1388 if (q->dyn_type == O_LIMIT_PARENT && 1389 rule== q->rule && 1390 pkt->proto == q->id.proto && 1391 pkt->src_port == q->id.src_port && 1392 pkt->dst_port == q->id.dst_port && 1393 ( 1394 (is_v6 && 1395 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1396 &(q->id.src_ip6)) && 1397 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1398 &(q->id.dst_ip6))) || 1399 (!is_v6 && 1400 pkt->src_ip == q->id.src_ip && 1401 pkt->dst_ip == q->id.dst_ip) 1402 ) 1403 ) { 1404 q->expire = time_uptime + dyn_short_lifetime; 1405 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1406 return q; 1407 } 1408 } 1409 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1410} 1411 1412/** 1413 * Install dynamic state for rule type cmd->o.opcode 1414 * 1415 * Returns 1 (failure) if state is not installed because of errors or because 1416 * session limitations are enforced. 1417 */ 1418static int 1419install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1420 struct ip_fw_args *args) 1421{ 1422 static int last_log; 1423 1424 ipfw_dyn_rule *q; 1425 1426 DEB(printf("ipfw: install state type %d 0x%08x %u -> 0x%08x %u\n", 1427 cmd->o.opcode, 1428 (args->f_id.src_ip), (args->f_id.src_port), 1429 (args->f_id.dst_ip), (args->f_id.dst_port) );) 1430 1431 IPFW_DYN_LOCK(); 1432 1433 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1434 1435 if (q != NULL) { /* should never occur */ 1436 if (last_log != time_uptime) { 1437 last_log = time_uptime; 1438 printf("ipfw: install_state: entry already present, done\n"); 1439 } 1440 IPFW_DYN_UNLOCK(); 1441 return 0; 1442 } 1443 1444 if (dyn_count >= dyn_max) 1445 /* 1446 * Run out of slots, try to remove any expired rule. 1447 */ 1448 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1449 1450 if (dyn_count >= dyn_max) { 1451 if (last_log != time_uptime) { 1452 last_log = time_uptime; 1453 printf("ipfw: install_state: Too many dynamic rules\n"); 1454 } 1455 IPFW_DYN_UNLOCK(); 1456 return 1; /* cannot install, notify caller */ 1457 } 1458 1459 switch (cmd->o.opcode) { 1460 case O_KEEP_STATE: /* bidir rule */ 1461 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1462 break; 1463 1464 case O_LIMIT: /* limit number of sessions */ 1465 { 1466 u_int16_t limit_mask = cmd->limit_mask; 1467 struct ipfw_flow_id id; 1468 ipfw_dyn_rule *parent; 1469 1470 DEB(printf("ipfw: installing dyn-limit rule %d\n", 1471 cmd->conn_limit);) 1472 1473 id.dst_ip = id.src_ip = 0; 1474 id.dst_port = id.src_port = 0; 1475 id.proto = args->f_id.proto; 1476 id.addr_type = args->f_id.addr_type; 1477 1478 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1479 if (limit_mask & DYN_SRC_ADDR) 1480 id.src_ip6 = args->f_id.src_ip6; 1481 if (limit_mask & DYN_DST_ADDR) 1482 id.dst_ip6 = args->f_id.dst_ip6; 1483 } else { 1484 if (limit_mask & DYN_SRC_ADDR) 1485 id.src_ip = args->f_id.src_ip; 1486 if (limit_mask & DYN_DST_ADDR) 1487 id.dst_ip = args->f_id.dst_ip; 1488 } 1489 if (limit_mask & DYN_SRC_PORT) 1490 id.src_port = args->f_id.src_port; 1491 if (limit_mask & DYN_DST_PORT) 1492 id.dst_port = args->f_id.dst_port; 1493 parent = lookup_dyn_parent(&id, rule); 1494 if (parent == NULL) { 1495 printf("ipfw: add parent failed\n"); 1496 IPFW_DYN_UNLOCK(); 1497 return 1; 1498 } 1499 if (parent->count >= cmd->conn_limit) { 1500 /* 1501 * See if we can remove some expired rule. 1502 */ 1503 remove_dyn_rule(rule, parent); 1504 if (parent->count >= cmd->conn_limit) { 1505 if (fw_verbose && last_log != time_uptime) { 1506 last_log = time_uptime; 1507 log(LOG_SECURITY | LOG_DEBUG, 1508 "drop session, too many entries\n"); 1509 } 1510 IPFW_DYN_UNLOCK(); 1511 return 1; 1512 } 1513 } 1514 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1515 } 1516 break; 1517 default: 1518 printf("ipfw: unknown dynamic rule type %u\n", cmd->o.opcode); 1519 IPFW_DYN_UNLOCK(); 1520 return 1; 1521 } 1522 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); /* XXX just set lifetime */ 1523 IPFW_DYN_UNLOCK(); 1524 return 0; 1525} 1526 1527/* 1528 * Generate a TCP packet, containing either a RST or a keepalive. 1529 * When flags & TH_RST, we are sending a RST packet, because of a 1530 * "reset" action matched the packet. 1531 * Otherwise we are sending a keepalive, and flags & TH_ 1532 */ 1533static struct mbuf * 1534send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags) 1535{ 1536 struct mbuf *m; 1537 struct ip *ip; 1538 struct tcphdr *tcp; 1539 1540 MGETHDR(m, M_DONTWAIT, MT_DATA); 1541 if (m == 0) 1542 return (NULL); 1543 m->m_pkthdr.rcvif = (struct ifnet *)0; 1544 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1545 m->m_data += max_linkhdr; 1546 1547 ip = mtod(m, struct ip *); 1548 bzero(ip, m->m_len); 1549 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1550 ip->ip_p = IPPROTO_TCP; 1551 tcp->th_off = 5; 1552 /* 1553 * Assume we are sending a RST (or a keepalive in the reverse 1554 * direction), swap src and destination addresses and ports. 1555 */ 1556 ip->ip_src.s_addr = htonl(id->dst_ip); 1557 ip->ip_dst.s_addr = htonl(id->src_ip); 1558 tcp->th_sport = htons(id->dst_port); 1559 tcp->th_dport = htons(id->src_port); 1560 if (flags & TH_RST) { /* we are sending a RST */ 1561 if (flags & TH_ACK) { 1562 tcp->th_seq = htonl(ack); 1563 tcp->th_ack = htonl(0); 1564 tcp->th_flags = TH_RST; 1565 } else { 1566 if (flags & TH_SYN) 1567 seq++; 1568 tcp->th_seq = htonl(0); 1569 tcp->th_ack = htonl(seq); 1570 tcp->th_flags = TH_RST | TH_ACK; 1571 } 1572 } else { 1573 /* 1574 * We are sending a keepalive. flags & TH_SYN determines 1575 * the direction, forward if set, reverse if clear. 1576 * NOTE: seq and ack are always assumed to be correct 1577 * as set by the caller. This may be confusing... 1578 */ 1579 if (flags & TH_SYN) { 1580 /* 1581 * we have to rewrite the correct addresses! 1582 */ 1583 ip->ip_dst.s_addr = htonl(id->dst_ip); 1584 ip->ip_src.s_addr = htonl(id->src_ip); 1585 tcp->th_dport = htons(id->dst_port); 1586 tcp->th_sport = htons(id->src_port); 1587 } 1588 tcp->th_seq = htonl(seq); 1589 tcp->th_ack = htonl(ack); 1590 tcp->th_flags = TH_ACK; 1591 } 1592 /* 1593 * set ip_len to the payload size so we can compute 1594 * the tcp checksum on the pseudoheader 1595 * XXX check this, could save a couple of words ? 1596 */ 1597 ip->ip_len = htons(sizeof(struct tcphdr)); 1598 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1599 /* 1600 * now fill fields left out earlier 1601 */ 1602 ip->ip_ttl = ip_defttl; 1603 ip->ip_len = m->m_pkthdr.len; 1604 m->m_flags |= M_SKIP_FIREWALL; 1605 return (m); 1606} 1607 1608/* 1609 * sends a reject message, consuming the mbuf passed as an argument. 1610 */ 1611static void 1612send_reject(struct ip_fw_args *args, int code, u_short offset, int ip_len) 1613{ 1614 1615 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1616 /* We need the IP header in host order for icmp_error(). */ 1617 if (args->eh != NULL) { 1618 struct ip *ip = mtod(args->m, struct ip *); 1619 ip->ip_len = ntohs(ip->ip_len); 1620 ip->ip_off = ntohs(ip->ip_off); 1621 } 1622 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1623 } else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) { 1624 struct tcphdr *const tcp = 1625 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1626 if ( (tcp->th_flags & TH_RST) == 0) { 1627 struct mbuf *m; 1628 m = send_pkt(&(args->f_id), ntohl(tcp->th_seq), 1629 ntohl(tcp->th_ack), 1630 tcp->th_flags | TH_RST); 1631 if (m != NULL) 1632 ip_output(m, NULL, NULL, 0, NULL, NULL); 1633 } 1634 m_freem(args->m); 1635 } else 1636 m_freem(args->m); 1637 args->m = NULL; 1638} 1639 1640/** 1641 * 1642 * Given an ip_fw *, lookup_next_rule will return a pointer 1643 * to the next rule, which can be either the jump 1644 * target (for skipto instructions) or the next one in the list (in 1645 * all other cases including a missing jump target). 1646 * The result is also written in the "next_rule" field of the rule. 1647 * Backward jumps are not allowed, so start looking from the next 1648 * rule... 1649 * 1650 * This never returns NULL -- in case we do not have an exact match, 1651 * the next rule is returned. When the ruleset is changed, 1652 * pointers are flushed so we are always correct. 1653 */ 1654 1655static struct ip_fw * 1656lookup_next_rule(struct ip_fw *me) 1657{ 1658 struct ip_fw *rule = NULL; 1659 ipfw_insn *cmd; 1660 1661 /* look for action, in case it is a skipto */ 1662 cmd = ACTION_PTR(me); 1663 if (cmd->opcode == O_LOG) 1664 cmd += F_LEN(cmd); 1665 if (cmd->opcode == O_ALTQ) 1666 cmd += F_LEN(cmd); 1667 if (cmd->opcode == O_TAG) 1668 cmd += F_LEN(cmd); 1669 if ( cmd->opcode == O_SKIPTO ) 1670 for (rule = me->next; rule ; rule = rule->next) 1671 if (rule->rulenum >= cmd->arg1) 1672 break; 1673 if (rule == NULL) /* failure or not a skipto */ 1674 rule = me->next; 1675 me->next_rule = rule; 1676 return rule; 1677} 1678 1679static int 1680add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1681 uint8_t mlen, uint32_t value) 1682{ 1683 struct radix_node_head *rnh; 1684 struct table_entry *ent; 1685 1686 if (tbl >= IPFW_TABLES_MAX) 1687 return (EINVAL); 1688 rnh = ch->tables[tbl]; 1689 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1690 if (ent == NULL) 1691 return (ENOMEM); 1692 ent->value = value; 1693 ent->addr.sin_len = ent->mask.sin_len = 8; 1694 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1695 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1696 IPFW_WLOCK(&layer3_chain); 1697 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1698 NULL) { 1699 IPFW_WUNLOCK(&layer3_chain); 1700 free(ent, M_IPFW_TBL); 1701 return (EEXIST); 1702 } 1703 IPFW_WUNLOCK(&layer3_chain); 1704 return (0); 1705} 1706 1707static int 1708del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1709 uint8_t mlen) 1710{ 1711 struct radix_node_head *rnh; 1712 struct table_entry *ent; 1713 struct sockaddr_in sa, mask; 1714 1715 if (tbl >= IPFW_TABLES_MAX) 1716 return (EINVAL); 1717 rnh = ch->tables[tbl]; 1718 sa.sin_len = mask.sin_len = 8; 1719 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1720 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1721 IPFW_WLOCK(ch); 1722 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1723 if (ent == NULL) { 1724 IPFW_WUNLOCK(ch); 1725 return (ESRCH); 1726 } 1727 IPFW_WUNLOCK(ch); 1728 free(ent, M_IPFW_TBL); 1729 return (0); 1730} 1731 1732static int 1733flush_table_entry(struct radix_node *rn, void *arg) 1734{ 1735 struct radix_node_head * const rnh = arg; 1736 struct table_entry *ent; 1737 1738 ent = (struct table_entry *) 1739 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1740 if (ent != NULL) 1741 free(ent, M_IPFW_TBL); 1742 return (0); 1743} 1744 1745static int 1746flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1747{ 1748 struct radix_node_head *rnh; 1749 1750 IPFW_WLOCK_ASSERT(ch); 1751 1752 if (tbl >= IPFW_TABLES_MAX) 1753 return (EINVAL); 1754 rnh = ch->tables[tbl]; 1755 KASSERT(rnh != NULL, ("NULL IPFW table")); 1756 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1757 return (0); 1758} 1759 1760static void 1761flush_tables(struct ip_fw_chain *ch) 1762{ 1763 uint16_t tbl; 1764 1765 IPFW_WLOCK_ASSERT(ch); 1766 1767 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1768 flush_table(ch, tbl); 1769} 1770 1771static int 1772init_tables(struct ip_fw_chain *ch) 1773{ 1774 int i; 1775 uint16_t j; 1776 1777 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1778 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1779 for (j = 0; j < i; j++) { 1780 (void) flush_table(ch, j); 1781 } 1782 return (ENOMEM); 1783 } 1784 } 1785 return (0); 1786} 1787 1788static int 1789lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1790 uint32_t *val) 1791{ 1792 struct radix_node_head *rnh; 1793 struct table_entry *ent; 1794 struct sockaddr_in sa; 1795 1796 if (tbl >= IPFW_TABLES_MAX) 1797 return (0); 1798 rnh = ch->tables[tbl]; 1799 sa.sin_len = 8; 1800 sa.sin_addr.s_addr = addr; 1801 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1802 if (ent != NULL) { 1803 *val = ent->value; 1804 return (1); 1805 } 1806 return (0); 1807} 1808 1809static int 1810count_table_entry(struct radix_node *rn, void *arg) 1811{ 1812 u_int32_t * const cnt = arg; 1813 1814 (*cnt)++; 1815 return (0); 1816} 1817 1818static int 1819count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1820{ 1821 struct radix_node_head *rnh; 1822 1823 if (tbl >= IPFW_TABLES_MAX) 1824 return (EINVAL); 1825 rnh = ch->tables[tbl]; 1826 *cnt = 0; 1827 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1828 return (0); 1829} 1830 1831static int 1832dump_table_entry(struct radix_node *rn, void *arg) 1833{ 1834 struct table_entry * const n = (struct table_entry *)rn; 1835 ipfw_table * const tbl = arg; 1836 ipfw_table_entry *ent; 1837 1838 if (tbl->cnt == tbl->size) 1839 return (1); 1840 ent = &tbl->ent[tbl->cnt]; 1841 ent->tbl = tbl->tbl; 1842 if (in_nullhost(n->mask.sin_addr)) 1843 ent->masklen = 0; 1844 else 1845 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1846 ent->addr = n->addr.sin_addr.s_addr; 1847 ent->value = n->value; 1848 tbl->cnt++; 1849 return (0); 1850} 1851 1852static int 1853dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1854{ 1855 struct radix_node_head *rnh; 1856 1857 if (tbl->tbl >= IPFW_TABLES_MAX) 1858 return (EINVAL); 1859 rnh = ch->tables[tbl->tbl]; 1860 tbl->cnt = 0; 1861 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1862 return (0); 1863} 1864 1865static void 1866fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1867{ 1868 struct ucred *cr; 1869 1870 if (inp->inp_socket != NULL) { 1871 cr = inp->inp_socket->so_cred; 1872 ugp->fw_prid = jailed(cr) ? 1873 cr->cr_prison->pr_id : -1; 1874 ugp->fw_uid = cr->cr_uid; 1875 ugp->fw_ngroups = cr->cr_ngroups; 1876 bcopy(cr->cr_groups, ugp->fw_groups, 1877 sizeof(ugp->fw_groups)); 1878 } 1879} 1880 1881static int 1882check_uidgid(ipfw_insn_u32 *insn, 1883 int proto, struct ifnet *oif, 1884 struct in_addr dst_ip, u_int16_t dst_port, 1885 struct in_addr src_ip, u_int16_t src_port, 1886 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 1887{ 1888 struct inpcbinfo *pi; 1889 int wildcard; 1890 struct inpcb *pcb; 1891 int match; 1892 gid_t *gp; 1893 1894 /* 1895 * Check to see if the UDP or TCP stack supplied us with 1896 * the PCB. If so, rather then holding a lock and looking 1897 * up the PCB, we can use the one that was supplied. 1898 */ 1899 if (inp && *lookup == 0) { 1900 INP_LOCK_ASSERT(inp); 1901 if (inp->inp_socket != NULL) { 1902 fill_ugid_cache(inp, ugp); 1903 *lookup = 1; 1904 } 1905 } 1906 /* 1907 * If we have already been here and the packet has no 1908 * PCB entry associated with it, then we can safely 1909 * assume that this is a no match. 1910 */ 1911 if (*lookup == -1) 1912 return (0); 1913 if (proto == IPPROTO_TCP) { 1914 wildcard = 0; 1915 pi = &tcbinfo; 1916 } else if (proto == IPPROTO_UDP) { 1917 wildcard = 1; 1918 pi = &udbinfo; 1919 } else 1920 return 0; 1921 match = 0; 1922 if (*lookup == 0) { 1923 INP_INFO_RLOCK(pi); 1924 pcb = (oif) ? 1925 in_pcblookup_hash(pi, 1926 dst_ip, htons(dst_port), 1927 src_ip, htons(src_port), 1928 wildcard, oif) : 1929 in_pcblookup_hash(pi, 1930 src_ip, htons(src_port), 1931 dst_ip, htons(dst_port), 1932 wildcard, NULL); 1933 if (pcb != NULL) { 1934 INP_LOCK(pcb); 1935 if (pcb->inp_socket != NULL) { 1936 fill_ugid_cache(pcb, ugp); 1937 *lookup = 1; 1938 } 1939 INP_UNLOCK(pcb); 1940 } 1941 INP_INFO_RUNLOCK(pi); 1942 if (*lookup == 0) { 1943 /* 1944 * If the lookup did not yield any results, there 1945 * is no sense in coming back and trying again. So 1946 * we can set lookup to -1 and ensure that we wont 1947 * bother the pcb system again. 1948 */ 1949 *lookup = -1; 1950 return (0); 1951 } 1952 } 1953 if (insn->o.opcode == O_UID) 1954 match = (ugp->fw_uid == (uid_t)insn->d[0]); 1955 else if (insn->o.opcode == O_GID) { 1956 for (gp = ugp->fw_groups; 1957 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 1958 if (*gp == (gid_t)insn->d[0]) { 1959 match = 1; 1960 break; 1961 } 1962 } else if (insn->o.opcode == O_JAIL) 1963 match = (ugp->fw_prid == (int)insn->d[0]); 1964 return match; 1965} 1966 1967/* 1968 * The main check routine for the firewall. 1969 * 1970 * All arguments are in args so we can modify them and return them 1971 * back to the caller. 1972 * 1973 * Parameters: 1974 * 1975 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1976 * Starts with the IP header. 1977 * args->eh (in) Mac header if present, or NULL for layer3 packet. 1978 * args->oif Outgoing interface, or NULL if packet is incoming. 1979 * The incoming interface is in the mbuf. (in) 1980 * args->divert_rule (in/out) 1981 * Skip up to the first rule past this rule number; 1982 * upon return, non-zero port number for divert or tee. 1983 * 1984 * args->rule Pointer to the last matching rule (in/out) 1985 * args->next_hop Socket we are forwarding to (out). 1986 * args->f_id Addresses grabbed from the packet (out) 1987 * args->cookie a cookie depending on rule action 1988 * 1989 * Return value: 1990 * 1991 * IP_FW_PASS the packet must be accepted 1992 * IP_FW_DENY the packet must be dropped 1993 * IP_FW_DIVERT divert packet, port in m_tag 1994 * IP_FW_TEE tee packet, port in m_tag 1995 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 1996 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 1997 * 1998 */ 1999 2000int 2001ipfw_chk(struct ip_fw_args *args) 2002{ 2003 /* 2004 * Local variables hold state during the processing of a packet. 2005 * 2006 * IMPORTANT NOTE: to speed up the processing of rules, there 2007 * are some assumption on the values of the variables, which 2008 * are documented here. Should you change them, please check 2009 * the implementation of the various instructions to make sure 2010 * that they still work. 2011 * 2012 * args->eh The MAC header. It is non-null for a layer2 2013 * packet, it is NULL for a layer-3 packet. 2014 * 2015 * m | args->m Pointer to the mbuf, as received from the caller. 2016 * It may change if ipfw_chk() does an m_pullup, or if it 2017 * consumes the packet because it calls send_reject(). 2018 * XXX This has to change, so that ipfw_chk() never modifies 2019 * or consumes the buffer. 2020 * ip is simply an alias of the value of m, and it is kept 2021 * in sync with it (the packet is supposed to start with 2022 * the ip header). 2023 */ 2024 struct mbuf *m = args->m; 2025 struct ip *ip = mtod(m, struct ip *); 2026 2027 /* 2028 * For rules which contain uid/gid or jail constraints, cache 2029 * a copy of the users credentials after the pcb lookup has been 2030 * executed. This will speed up the processing of rules with 2031 * these types of constraints, as well as decrease contention 2032 * on pcb related locks. 2033 */ 2034 struct ip_fw_ugid fw_ugid_cache; 2035 int ugid_lookup = 0; 2036 2037 /* 2038 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2039 * associated with a packet input on a divert socket. This 2040 * will allow to distinguish traffic and its direction when 2041 * it originates from a divert socket. 2042 */ 2043 u_int divinput_flags = 0; 2044 2045 /* 2046 * oif | args->oif If NULL, ipfw_chk has been called on the 2047 * inbound path (ether_input, ip_input). 2048 * If non-NULL, ipfw_chk has been called on the outbound path 2049 * (ether_output, ip_output). 2050 */ 2051 struct ifnet *oif = args->oif; 2052 2053 struct ip_fw *f = NULL; /* matching rule */ 2054 int retval = 0; 2055 2056 /* 2057 * hlen The length of the IP header. 2058 */ 2059 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2060 2061 /* 2062 * offset The offset of a fragment. offset != 0 means that 2063 * we have a fragment at this offset of an IPv4 packet. 2064 * offset == 0 means that (if this is an IPv4 packet) 2065 * this is the first or only fragment. 2066 * For IPv6 offset == 0 means there is no Fragment Header. 2067 * If offset != 0 for IPv6 always use correct mask to 2068 * get the correct offset because we add IP6F_MORE_FRAG 2069 * to be able to dectect the first fragment which would 2070 * otherwise have offset = 0. 2071 */ 2072 u_short offset = 0; 2073 2074 /* 2075 * Local copies of addresses. They are only valid if we have 2076 * an IP packet. 2077 * 2078 * proto The protocol. Set to 0 for non-ip packets, 2079 * or to the protocol read from the packet otherwise. 2080 * proto != 0 means that we have an IPv4 packet. 2081 * 2082 * src_port, dst_port port numbers, in HOST format. Only 2083 * valid for TCP and UDP packets. 2084 * 2085 * src_ip, dst_ip ip addresses, in NETWORK format. 2086 * Only valid for IPv4 packets. 2087 */ 2088 u_int8_t proto; 2089 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2090 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2091 u_int16_t ip_len=0; 2092 int pktlen; 2093 2094 /* 2095 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2096 * MATCH_NONE when checked and not matched (q = NULL), 2097 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2098 */ 2099 int dyn_dir = MATCH_UNKNOWN; 2100 ipfw_dyn_rule *q = NULL; 2101 struct ip_fw_chain *chain = &layer3_chain; 2102 struct m_tag *mtag; 2103 2104 /* 2105 * We store in ulp a pointer to the upper layer protocol header. 2106 * In the ipv4 case this is easy to determine from the header, 2107 * but for ipv6 we might have some additional headers in the middle. 2108 * ulp is NULL if not found. 2109 */ 2110 void *ulp = NULL; /* upper layer protocol pointer. */ 2111 /* XXX ipv6 variables */ 2112 int is_ipv6 = 0; 2113 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2114 /* end of ipv6 variables */ 2115 int is_ipv4 = 0; 2116 2117 if (m->m_flags & M_SKIP_FIREWALL) 2118 return (IP_FW_PASS); /* accept */ 2119 2120 pktlen = m->m_pkthdr.len; 2121 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2122 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2123 2124/* 2125 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2126 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2127 * pointer might become stale after other pullups (but we never use it 2128 * this way). 2129 */ 2130#define PULLUP_TO(len, p, T) \ 2131do { \ 2132 int x = (len) + sizeof(T); \ 2133 if ((m)->m_len < x) { \ 2134 args->m = m = m_pullup(m, x); \ 2135 if (m == NULL) \ 2136 goto pullup_failed; \ 2137 } \ 2138 p = (mtod(m, char *) + (len)); \ 2139} while (0) 2140 2141 /* Identify IP packets and fill up variables. */ 2142 if (pktlen >= sizeof(struct ip6_hdr) && 2143 (args->eh == NULL || ntohs(args->eh->ether_type)==ETHERTYPE_IPV6) && 2144 mtod(m, struct ip *)->ip_v == 6) { 2145 is_ipv6 = 1; 2146 args->f_id.addr_type = 6; 2147 hlen = sizeof(struct ip6_hdr); 2148 proto = mtod(m, struct ip6_hdr *)->ip6_nxt; 2149 2150 /* Search extension headers to find upper layer protocols */ 2151 while (ulp == NULL) { 2152 switch (proto) { 2153 case IPPROTO_ICMPV6: 2154 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2155 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2156 break; 2157 2158 case IPPROTO_TCP: 2159 PULLUP_TO(hlen, ulp, struct tcphdr); 2160 dst_port = TCP(ulp)->th_dport; 2161 src_port = TCP(ulp)->th_sport; 2162 args->f_id.flags = TCP(ulp)->th_flags; 2163 break; 2164 2165 case IPPROTO_UDP: 2166 PULLUP_TO(hlen, ulp, struct udphdr); 2167 dst_port = UDP(ulp)->uh_dport; 2168 src_port = UDP(ulp)->uh_sport; 2169 break; 2170 2171 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2172 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2173 ext_hd |= EXT_HOPOPTS; 2174 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2175 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2176 ulp = NULL; 2177 break; 2178 2179 case IPPROTO_ROUTING: /* RFC 2460 */ 2180 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2181 if (((struct ip6_rthdr *)ulp)->ip6r_type != 0) { 2182 printf("IPFW2: IPV6 - Unknown Routing " 2183 "Header type(%d)\n", 2184 ((struct ip6_rthdr *)ulp)->ip6r_type); 2185 if (fw_deny_unknown_exthdrs) 2186 return (IP_FW_DENY); 2187 break; 2188 } 2189 ext_hd |= EXT_ROUTING; 2190 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2191 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2192 ulp = NULL; 2193 break; 2194 2195 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2196 PULLUP_TO(hlen, ulp, struct ip6_frag); 2197 ext_hd |= EXT_FRAGMENT; 2198 hlen += sizeof (struct ip6_frag); 2199 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2200 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2201 IP6F_OFF_MASK; 2202 /* Add IP6F_MORE_FRAG for offset of first 2203 * fragment to be != 0. */ 2204 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2205 IP6F_MORE_FRAG; 2206 if (offset == 0) { 2207 printf("IPFW2: IPV6 - Invalid Fragment " 2208 "Header\n"); 2209 if (fw_deny_unknown_exthdrs) 2210 return (IP_FW_DENY); 2211 break; 2212 } 2213 args->f_id.frag_id6 = 2214 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2215 ulp = NULL; 2216 break; 2217 2218 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2219 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2220 ext_hd |= EXT_DSTOPTS; 2221 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2222 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2223 ulp = NULL; 2224 break; 2225 2226 case IPPROTO_AH: /* RFC 2402 */ 2227 PULLUP_TO(hlen, ulp, struct ip6_ext); 2228 ext_hd |= EXT_AH; 2229 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2230 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2231 ulp = NULL; 2232 break; 2233 2234 case IPPROTO_ESP: /* RFC 2406 */ 2235 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2236 /* Anything past Seq# is variable length and 2237 * data past this ext. header is encrypted. */ 2238 ext_hd |= EXT_ESP; 2239 break; 2240 2241 case IPPROTO_NONE: /* RFC 2460 */ 2242 PULLUP_TO(hlen, ulp, struct ip6_ext); 2243 /* Packet ends here. if ip6e_len!=0 octets 2244 * must be ignored. */ 2245 break; 2246 2247 case IPPROTO_OSPFIGP: 2248 /* XXX OSPF header check? */ 2249 PULLUP_TO(hlen, ulp, struct ip6_ext); 2250 break; 2251 2252 default: 2253 printf("IPFW2: IPV6 - Unknown Extension " 2254 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2255 if (fw_deny_unknown_exthdrs) 2256 return (IP_FW_DENY); 2257 break; 2258 } /*switch */ 2259 } 2260 args->f_id.src_ip6 = mtod(m,struct ip6_hdr *)->ip6_src; 2261 args->f_id.dst_ip6 = mtod(m,struct ip6_hdr *)->ip6_dst; 2262 args->f_id.src_ip = 0; 2263 args->f_id.dst_ip = 0; 2264 args->f_id.flow_id6 = ntohl(mtod(m, struct ip6_hdr *)->ip6_flow); 2265 } else if (pktlen >= sizeof(struct ip) && 2266 (args->eh == NULL || ntohs(args->eh->ether_type) == ETHERTYPE_IP) && 2267 mtod(m, struct ip *)->ip_v == 4) { 2268 is_ipv4 = 1; 2269 ip = mtod(m, struct ip *); 2270 hlen = ip->ip_hl << 2; 2271 args->f_id.addr_type = 4; 2272 2273 /* 2274 * Collect parameters into local variables for faster matching. 2275 */ 2276 proto = ip->ip_p; 2277 src_ip = ip->ip_src; 2278 dst_ip = ip->ip_dst; 2279 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2280 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2281 ip_len = ntohs(ip->ip_len); 2282 } else { 2283 offset = ip->ip_off & IP_OFFMASK; 2284 ip_len = ip->ip_len; 2285 } 2286 pktlen = ip_len < pktlen ? ip_len : pktlen; 2287 2288 if (offset == 0) { 2289 switch (proto) { 2290 case IPPROTO_TCP: 2291 PULLUP_TO(hlen, ulp, struct tcphdr); 2292 dst_port = TCP(ulp)->th_dport; 2293 src_port = TCP(ulp)->th_sport; 2294 args->f_id.flags = TCP(ulp)->th_flags; 2295 break; 2296 2297 case IPPROTO_UDP: 2298 PULLUP_TO(hlen, ulp, struct udphdr); 2299 dst_port = UDP(ulp)->uh_dport; 2300 src_port = UDP(ulp)->uh_sport; 2301 break; 2302 2303 case IPPROTO_ICMP: 2304 PULLUP_TO(hlen, ulp, struct icmphdr); 2305 args->f_id.flags = ICMP(ulp)->icmp_type; 2306 break; 2307 2308 default: 2309 break; 2310 } 2311 } 2312 2313 args->f_id.src_ip = ntohl(src_ip.s_addr); 2314 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2315 } 2316#undef PULLUP_TO 2317 if (proto) { /* we may have port numbers, store them */ 2318 args->f_id.proto = proto; 2319 args->f_id.src_port = src_port = ntohs(src_port); 2320 args->f_id.dst_port = dst_port = ntohs(dst_port); 2321 } 2322 2323 IPFW_RLOCK(chain); 2324 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2325 if (args->rule) { 2326 /* 2327 * Packet has already been tagged. Look for the next rule 2328 * to restart processing. 2329 * 2330 * If fw_one_pass != 0 then just accept it. 2331 * XXX should not happen here, but optimized out in 2332 * the caller. 2333 */ 2334 if (fw_one_pass) { 2335 IPFW_RUNLOCK(chain); 2336 return (IP_FW_PASS); 2337 } 2338 2339 f = args->rule->next_rule; 2340 if (f == NULL) 2341 f = lookup_next_rule(args->rule); 2342 } else { 2343 /* 2344 * Find the starting rule. It can be either the first 2345 * one, or the one after divert_rule if asked so. 2346 */ 2347 int skipto = mtag ? divert_cookie(mtag) : 0; 2348 2349 f = chain->rules; 2350 if (args->eh == NULL && skipto != 0) { 2351 if (skipto >= IPFW_DEFAULT_RULE) { 2352 IPFW_RUNLOCK(chain); 2353 return (IP_FW_DENY); /* invalid */ 2354 } 2355 while (f && f->rulenum <= skipto) 2356 f = f->next; 2357 if (f == NULL) { /* drop packet */ 2358 IPFW_RUNLOCK(chain); 2359 return (IP_FW_DENY); 2360 } 2361 } 2362 } 2363 /* reset divert rule to avoid confusion later */ 2364 if (mtag) { 2365 divinput_flags = divert_info(mtag) & 2366 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2367 m_tag_delete(m, mtag); 2368 } 2369 2370 /* 2371 * Now scan the rules, and parse microinstructions for each rule. 2372 */ 2373 for (; f; f = f->next) { 2374 ipfw_insn *cmd; 2375 uint32_t tablearg = 0; 2376 int l, cmdlen, skip_or; /* skip rest of OR block */ 2377 2378again: 2379 if (set_disable & (1 << f->set) ) 2380 continue; 2381 2382 skip_or = 0; 2383 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2384 l -= cmdlen, cmd += cmdlen) { 2385 int match; 2386 2387 /* 2388 * check_body is a jump target used when we find a 2389 * CHECK_STATE, and need to jump to the body of 2390 * the target rule. 2391 */ 2392 2393check_body: 2394 cmdlen = F_LEN(cmd); 2395 /* 2396 * An OR block (insn_1 || .. || insn_n) has the 2397 * F_OR bit set in all but the last instruction. 2398 * The first match will set "skip_or", and cause 2399 * the following instructions to be skipped until 2400 * past the one with the F_OR bit clear. 2401 */ 2402 if (skip_or) { /* skip this instruction */ 2403 if ((cmd->len & F_OR) == 0) 2404 skip_or = 0; /* next one is good */ 2405 continue; 2406 } 2407 match = 0; /* set to 1 if we succeed */ 2408 2409 switch (cmd->opcode) { 2410 /* 2411 * The first set of opcodes compares the packet's 2412 * fields with some pattern, setting 'match' if a 2413 * match is found. At the end of the loop there is 2414 * logic to deal with F_NOT and F_OR flags associated 2415 * with the opcode. 2416 */ 2417 case O_NOP: 2418 match = 1; 2419 break; 2420 2421 case O_FORWARD_MAC: 2422 printf("ipfw: opcode %d unimplemented\n", 2423 cmd->opcode); 2424 break; 2425 2426 case O_GID: 2427 case O_UID: 2428 case O_JAIL: 2429 /* 2430 * We only check offset == 0 && proto != 0, 2431 * as this ensures that we have a 2432 * packet with the ports info. 2433 */ 2434 if (offset!=0) 2435 break; 2436 if (is_ipv6) /* XXX to be fixed later */ 2437 break; 2438 if (proto == IPPROTO_TCP || 2439 proto == IPPROTO_UDP) 2440 match = check_uidgid( 2441 (ipfw_insn_u32 *)cmd, 2442 proto, oif, 2443 dst_ip, dst_port, 2444 src_ip, src_port, &fw_ugid_cache, 2445 &ugid_lookup, args->inp); 2446 break; 2447 2448 case O_RECV: 2449 match = iface_match(m->m_pkthdr.rcvif, 2450 (ipfw_insn_if *)cmd); 2451 break; 2452 2453 case O_XMIT: 2454 match = iface_match(oif, (ipfw_insn_if *)cmd); 2455 break; 2456 2457 case O_VIA: 2458 match = iface_match(oif ? oif : 2459 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2460 break; 2461 2462 case O_MACADDR2: 2463 if (args->eh != NULL) { /* have MAC header */ 2464 u_int32_t *want = (u_int32_t *) 2465 ((ipfw_insn_mac *)cmd)->addr; 2466 u_int32_t *mask = (u_int32_t *) 2467 ((ipfw_insn_mac *)cmd)->mask; 2468 u_int32_t *hdr = (u_int32_t *)args->eh; 2469 2470 match = 2471 ( want[0] == (hdr[0] & mask[0]) && 2472 want[1] == (hdr[1] & mask[1]) && 2473 want[2] == (hdr[2] & mask[2]) ); 2474 } 2475 break; 2476 2477 case O_MAC_TYPE: 2478 if (args->eh != NULL) { 2479 u_int16_t t = 2480 ntohs(args->eh->ether_type); 2481 u_int16_t *p = 2482 ((ipfw_insn_u16 *)cmd)->ports; 2483 int i; 2484 2485 for (i = cmdlen - 1; !match && i>0; 2486 i--, p += 2) 2487 match = (t>=p[0] && t<=p[1]); 2488 } 2489 break; 2490 2491 case O_FRAG: 2492 match = (offset != 0); 2493 break; 2494 2495 case O_IN: /* "out" is "not in" */ 2496 match = (oif == NULL); 2497 break; 2498 2499 case O_LAYER2: 2500 match = (args->eh != NULL); 2501 break; 2502 2503 case O_DIVERTED: 2504 match = (cmd->arg1 & 1 && divinput_flags & 2505 IP_FW_DIVERT_LOOPBACK_FLAG) || 2506 (cmd->arg1 & 2 && divinput_flags & 2507 IP_FW_DIVERT_OUTPUT_FLAG); 2508 break; 2509 2510 case O_PROTO: 2511 /* 2512 * We do not allow an arg of 0 so the 2513 * check of "proto" only suffices. 2514 */ 2515 match = (proto == cmd->arg1); 2516 break; 2517 2518 case O_IP_SRC: 2519 match = is_ipv4 && 2520 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2521 src_ip.s_addr); 2522 break; 2523 2524 case O_IP_SRC_LOOKUP: 2525 case O_IP_DST_LOOKUP: 2526 if (is_ipv4) { 2527 uint32_t a = 2528 (cmd->opcode == O_IP_DST_LOOKUP) ? 2529 dst_ip.s_addr : src_ip.s_addr; 2530 uint32_t v; 2531 2532 match = lookup_table(chain, cmd->arg1, a, 2533 &v); 2534 if (!match) 2535 break; 2536 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2537 match = 2538 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2539 else 2540 tablearg = v; 2541 } 2542 break; 2543 2544 case O_IP_SRC_MASK: 2545 case O_IP_DST_MASK: 2546 if (is_ipv4) { 2547 uint32_t a = 2548 (cmd->opcode == O_IP_DST_MASK) ? 2549 dst_ip.s_addr : src_ip.s_addr; 2550 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2551 int i = cmdlen-1; 2552 2553 for (; !match && i>0; i-= 2, p+= 2) 2554 match = (p[0] == (a & p[1])); 2555 } 2556 break; 2557 2558 case O_IP_SRC_ME: 2559 if (is_ipv4) { 2560 struct ifnet *tif; 2561 2562 INADDR_TO_IFP(src_ip, tif); 2563 match = (tif != NULL); 2564 } 2565 break; 2566 2567 case O_IP_DST_SET: 2568 case O_IP_SRC_SET: 2569 if (is_ipv4) { 2570 u_int32_t *d = (u_int32_t *)(cmd+1); 2571 u_int32_t addr = 2572 cmd->opcode == O_IP_DST_SET ? 2573 args->f_id.dst_ip : 2574 args->f_id.src_ip; 2575 2576 if (addr < d[0]) 2577 break; 2578 addr -= d[0]; /* subtract base */ 2579 match = (addr < cmd->arg1) && 2580 ( d[ 1 + (addr>>5)] & 2581 (1<<(addr & 0x1f)) ); 2582 } 2583 break; 2584 2585 case O_IP_DST: 2586 match = is_ipv4 && 2587 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2588 dst_ip.s_addr); 2589 break; 2590 2591 case O_IP_DST_ME: 2592 if (is_ipv4) { 2593 struct ifnet *tif; 2594 2595 INADDR_TO_IFP(dst_ip, tif); 2596 match = (tif != NULL); 2597 } 2598 break; 2599 2600 case O_IP_SRCPORT: 2601 case O_IP_DSTPORT: 2602 /* 2603 * offset == 0 && proto != 0 is enough 2604 * to guarantee that we have a 2605 * packet with port info. 2606 */ 2607 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2608 && offset == 0) { 2609 u_int16_t x = 2610 (cmd->opcode == O_IP_SRCPORT) ? 2611 src_port : dst_port ; 2612 u_int16_t *p = 2613 ((ipfw_insn_u16 *)cmd)->ports; 2614 int i; 2615 2616 for (i = cmdlen - 1; !match && i>0; 2617 i--, p += 2) 2618 match = (x>=p[0] && x<=p[1]); 2619 } 2620 break; 2621 2622 case O_ICMPTYPE: 2623 match = (offset == 0 && proto==IPPROTO_ICMP && 2624 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2625 break; 2626 2627#ifdef INET6 2628 case O_ICMP6TYPE: 2629 match = is_ipv6 && offset == 0 && 2630 proto==IPPROTO_ICMPV6 && 2631 icmp6type_match( 2632 ICMP6(ulp)->icmp6_type, 2633 (ipfw_insn_u32 *)cmd); 2634 break; 2635#endif /* INET6 */ 2636 2637 case O_IPOPT: 2638 match = (is_ipv4 && 2639 ipopts_match(mtod(m, struct ip *), cmd) ); 2640 break; 2641 2642 case O_IPVER: 2643 match = (is_ipv4 && 2644 cmd->arg1 == mtod(m, struct ip *)->ip_v); 2645 break; 2646 2647 case O_IPID: 2648 case O_IPLEN: 2649 case O_IPTTL: 2650 if (is_ipv4) { /* only for IP packets */ 2651 uint16_t x; 2652 uint16_t *p; 2653 int i; 2654 2655 if (cmd->opcode == O_IPLEN) 2656 x = ip_len; 2657 else if (cmd->opcode == O_IPTTL) 2658 x = mtod(m, struct ip *)->ip_ttl; 2659 else /* must be IPID */ 2660 x = ntohs(mtod(m, struct ip *)->ip_id); 2661 if (cmdlen == 1) { 2662 match = (cmd->arg1 == x); 2663 break; 2664 } 2665 /* otherwise we have ranges */ 2666 p = ((ipfw_insn_u16 *)cmd)->ports; 2667 i = cmdlen - 1; 2668 for (; !match && i>0; i--, p += 2) 2669 match = (x >= p[0] && x <= p[1]); 2670 } 2671 break; 2672 2673 case O_IPPRECEDENCE: 2674 match = (is_ipv4 && 2675 (cmd->arg1 == (mtod(m, struct ip *)->ip_tos & 0xe0)) ); 2676 break; 2677 2678 case O_IPTOS: 2679 match = (is_ipv4 && 2680 flags_match(cmd, mtod(m, struct ip *)->ip_tos)); 2681 break; 2682 2683 case O_TCPDATALEN: 2684 if (proto == IPPROTO_TCP && offset == 0) { 2685 struct tcphdr *tcp; 2686 uint16_t x; 2687 uint16_t *p; 2688 int i; 2689 2690 tcp = TCP(ulp); 2691 x = ip_len - 2692 ((ip->ip_hl + tcp->th_off) << 2); 2693 if (cmdlen == 1) { 2694 match = (cmd->arg1 == x); 2695 break; 2696 } 2697 /* otherwise we have ranges */ 2698 p = ((ipfw_insn_u16 *)cmd)->ports; 2699 i = cmdlen - 1; 2700 for (; !match && i>0; i--, p += 2) 2701 match = (x >= p[0] && x <= p[1]); 2702 } 2703 break; 2704 2705 case O_TCPFLAGS: 2706 match = (proto == IPPROTO_TCP && offset == 0 && 2707 flags_match(cmd, TCP(ulp)->th_flags)); 2708 break; 2709 2710 case O_TCPOPTS: 2711 match = (proto == IPPROTO_TCP && offset == 0 && 2712 tcpopts_match(TCP(ulp), cmd)); 2713 break; 2714 2715 case O_TCPSEQ: 2716 match = (proto == IPPROTO_TCP && offset == 0 && 2717 ((ipfw_insn_u32 *)cmd)->d[0] == 2718 TCP(ulp)->th_seq); 2719 break; 2720 2721 case O_TCPACK: 2722 match = (proto == IPPROTO_TCP && offset == 0 && 2723 ((ipfw_insn_u32 *)cmd)->d[0] == 2724 TCP(ulp)->th_ack); 2725 break; 2726 2727 case O_TCPWIN: 2728 match = (proto == IPPROTO_TCP && offset == 0 && 2729 cmd->arg1 == TCP(ulp)->th_win); 2730 break; 2731 2732 case O_ESTAB: 2733 /* reject packets which have SYN only */ 2734 /* XXX should i also check for TH_ACK ? */ 2735 match = (proto == IPPROTO_TCP && offset == 0 && 2736 (TCP(ulp)->th_flags & 2737 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2738 break; 2739 2740 case O_ALTQ: { 2741 struct altq_tag *at; 2742 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2743 2744 match = 1; 2745 mtag = m_tag_find(m, PACKET_TAG_PF_QID, NULL); 2746 if (mtag != NULL) 2747 break; 2748 mtag = m_tag_get(PACKET_TAG_PF_QID, 2749 sizeof(struct altq_tag), 2750 M_NOWAIT); 2751 if (mtag == NULL) { 2752 /* 2753 * Let the packet fall back to the 2754 * default ALTQ. 2755 */ 2756 break; 2757 } 2758 at = (struct altq_tag *)(mtag+1); 2759 at->qid = altq->qid; 2760 if (is_ipv4) 2761 at->af = AF_INET; 2762 else 2763 at->af = AF_LINK; 2764 at->hdr = ip; 2765 m_tag_prepend(m, mtag); 2766 break; 2767 } 2768 2769 case O_LOG: 2770 if (fw_verbose) 2771 ipfw_log(f, hlen, args, m, oif, offset); 2772 match = 1; 2773 break; 2774 2775 case O_PROB: 2776 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2777 break; 2778 2779 case O_VERREVPATH: 2780 /* Outgoing packets automatically pass/match */ 2781 match = ((oif != NULL) || 2782 (m->m_pkthdr.rcvif == NULL) || 2783 ( 2784#ifdef INET6 2785 is_ipv6 ? 2786 verify_path6(&(args->f_id.src_ip6), 2787 m->m_pkthdr.rcvif) : 2788#endif 2789 verify_path(src_ip, m->m_pkthdr.rcvif))); 2790 break; 2791 2792 case O_VERSRCREACH: 2793 /* Outgoing packets automatically pass/match */ 2794 match = (hlen > 0 && ((oif != NULL) || 2795#ifdef INET6 2796 is_ipv6 ? 2797 verify_path6(&(args->f_id.src_ip6), 2798 NULL) : 2799#endif 2800 verify_path(src_ip, NULL))); 2801 break; 2802 2803 case O_ANTISPOOF: 2804 /* Outgoing packets automatically pass/match */ 2805 if (oif == NULL && hlen > 0 && 2806 ( (is_ipv4 && in_localaddr(src_ip)) 2807#ifdef INET6 2808 || (is_ipv6 && 2809 in6_localaddr(&(args->f_id.src_ip6))) 2810#endif 2811 )) 2812 match = 2813#ifdef INET6 2814 is_ipv6 ? verify_path6( 2815 &(args->f_id.src_ip6), 2816 m->m_pkthdr.rcvif) : 2817#endif 2818 verify_path(src_ip, 2819 m->m_pkthdr.rcvif); 2820 else 2821 match = 1; 2822 break; 2823 2824 case O_IPSEC: 2825#ifdef FAST_IPSEC 2826 match = (m_tag_find(m, 2827 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2828#endif 2829#ifdef IPSEC 2830 match = (ipsec_getnhist(m) != 0); 2831#endif 2832 /* otherwise no match */ 2833 break; 2834 2835#ifdef INET6 2836 case O_IP6_SRC: 2837 match = is_ipv6 && 2838 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2839 &((ipfw_insn_ip6 *)cmd)->addr6); 2840 break; 2841 2842 case O_IP6_DST: 2843 match = is_ipv6 && 2844 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2845 &((ipfw_insn_ip6 *)cmd)->addr6); 2846 break; 2847 case O_IP6_SRC_MASK: 2848 if (is_ipv6) { 2849 ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd; 2850 struct in6_addr p = args->f_id.src_ip6; 2851 2852 APPLY_MASK(&p, &te->mask6); 2853 match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p); 2854 } 2855 break; 2856 2857 case O_IP6_DST_MASK: 2858 if (is_ipv6) { 2859 ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd; 2860 struct in6_addr p = args->f_id.dst_ip6; 2861 2862 APPLY_MASK(&p, &te->mask6); 2863 match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p); 2864 } 2865 break; 2866 2867 case O_IP6_SRC_ME: 2868 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 2869 break; 2870 2871 case O_IP6_DST_ME: 2872 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 2873 break; 2874 2875 case O_FLOW6ID: 2876 match = is_ipv6 && 2877 flow6id_match(args->f_id.flow_id6, 2878 (ipfw_insn_u32 *) cmd); 2879 break; 2880 2881 case O_EXT_HDR: 2882 match = is_ipv6 && 2883 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2884 break; 2885 2886 case O_IP6: 2887 match = is_ipv6; 2888 break; 2889#endif 2890 2891 case O_IP4: 2892 match = is_ipv4; 2893 break; 2894 2895 case O_TAG: 2896 /* Packet is already tagged with this tag? */ 2897 mtag = m_tag_locate(m, MTAG_IPFW, 2898 ((ipfw_insn *) cmd)->arg1, NULL); 2899 /* We have `untag' action when F_NOT flag is 2900 * present. And we must remove this mtag from 2901 * mbuf and reset `match' to zero (`match' will 2902 * be inversed later). 2903 * Otherwise we should allocate new mtag and 2904 * push it into mbuf. 2905 */ 2906 if (cmd->len & F_NOT) { /* `untag' action */ 2907 if (mtag != NULL) 2908 m_tag_delete(m, mtag); 2909 } else if (mtag == NULL) { 2910 mtag = m_tag_alloc(MTAG_IPFW, 2911 ((ipfw_insn *) cmd)->arg1, 0, M_NOWAIT); 2912 if (mtag != NULL) 2913 m_tag_prepend(m, mtag); 2914 } 2915 match = (cmd->len & F_NOT) ? 0: 1; 2916 break; 2917 2918 case O_TAGGED: 2919 if (cmdlen == 1) { 2920 match = (m_tag_locate(m, MTAG_IPFW, 2921 ((ipfw_insn *) cmd)->arg1, NULL) != NULL); 2922 break; 2923 } 2924 2925 /* we have ranges */ 2926 for (mtag = m_tag_first(m); 2927 mtag != NULL && !match; 2928 mtag = m_tag_next(m, mtag)) { 2929 uint16_t *p; 2930 int i; 2931 2932 if (mtag->m_tag_cookie != MTAG_IPFW) 2933 continue; 2934 2935 p = ((ipfw_insn_u16 *)cmd)->ports; 2936 i = cmdlen - 1; 2937 for(; !match && i > 0; i--, p += 2) 2938 match = 2939 mtag->m_tag_id >= p[0] && 2940 mtag->m_tag_id <= p[1]; 2941 } 2942 break; 2943 2944 /* 2945 * The second set of opcodes represents 'actions', 2946 * i.e. the terminal part of a rule once the packet 2947 * matches all previous patterns. 2948 * Typically there is only one action for each rule, 2949 * and the opcode is stored at the end of the rule 2950 * (but there are exceptions -- see below). 2951 * 2952 * In general, here we set retval and terminate the 2953 * outer loop (would be a 'break 3' in some language, 2954 * but we need to do a 'goto done'). 2955 * 2956 * Exceptions: 2957 * O_COUNT and O_SKIPTO actions: 2958 * instead of terminating, we jump to the next rule 2959 * ('goto next_rule', equivalent to a 'break 2'), 2960 * or to the SKIPTO target ('goto again' after 2961 * having set f, cmd and l), respectively. 2962 * 2963 * O_TAG, O_LOG and O_ALTQ action parameters: 2964 * perform some action and set match = 1; 2965 * 2966 * O_LIMIT and O_KEEP_STATE: these opcodes are 2967 * not real 'actions', and are stored right 2968 * before the 'action' part of the rule. 2969 * These opcodes try to install an entry in the 2970 * state tables; if successful, we continue with 2971 * the next opcode (match=1; break;), otherwise 2972 * the packet * must be dropped 2973 * ('goto done' after setting retval); 2974 * 2975 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2976 * cause a lookup of the state table, and a jump 2977 * to the 'action' part of the parent rule 2978 * ('goto check_body') if an entry is found, or 2979 * (CHECK_STATE only) a jump to the next rule if 2980 * the entry is not found ('goto next_rule'). 2981 * The result of the lookup is cached to make 2982 * further instances of these opcodes are 2983 * effectively NOPs. 2984 */ 2985 case O_LIMIT: 2986 case O_KEEP_STATE: 2987 if (install_state(f, 2988 (ipfw_insn_limit *)cmd, args)) { 2989 retval = IP_FW_DENY; 2990 goto done; /* error/limit violation */ 2991 } 2992 match = 1; 2993 break; 2994 2995 case O_PROBE_STATE: 2996 case O_CHECK_STATE: 2997 /* 2998 * dynamic rules are checked at the first 2999 * keep-state or check-state occurrence, 3000 * with the result being stored in dyn_dir. 3001 * The compiler introduces a PROBE_STATE 3002 * instruction for us when we have a 3003 * KEEP_STATE (because PROBE_STATE needs 3004 * to be run first). 3005 */ 3006 if (dyn_dir == MATCH_UNKNOWN && 3007 (q = lookup_dyn_rule(&args->f_id, 3008 &dyn_dir, proto == IPPROTO_TCP ? 3009 TCP(ulp) : NULL)) 3010 != NULL) { 3011 /* 3012 * Found dynamic entry, update stats 3013 * and jump to the 'action' part of 3014 * the parent rule. 3015 */ 3016 q->pcnt++; 3017 q->bcnt += pktlen; 3018 f = q->rule; 3019 cmd = ACTION_PTR(f); 3020 l = f->cmd_len - f->act_ofs; 3021 IPFW_DYN_UNLOCK(); 3022 goto check_body; 3023 } 3024 /* 3025 * Dynamic entry not found. If CHECK_STATE, 3026 * skip to next rule, if PROBE_STATE just 3027 * ignore and continue with next opcode. 3028 */ 3029 if (cmd->opcode == O_CHECK_STATE) 3030 goto next_rule; 3031 match = 1; 3032 break; 3033 3034 case O_ACCEPT: 3035 retval = 0; /* accept */ 3036 goto done; 3037 3038 case O_PIPE: 3039 case O_QUEUE: 3040 args->rule = f; /* report matching rule */ 3041 if (cmd->arg1 == IP_FW_TABLEARG) 3042 args->cookie = tablearg; 3043 else 3044 args->cookie = cmd->arg1; 3045 retval = IP_FW_DUMMYNET; 3046 goto done; 3047 3048 case O_DIVERT: 3049 case O_TEE: { 3050 struct divert_tag *dt; 3051 3052 if (args->eh) /* not on layer 2 */ 3053 break; 3054 mtag = m_tag_get(PACKET_TAG_DIVERT, 3055 sizeof(struct divert_tag), 3056 M_NOWAIT); 3057 if (mtag == NULL) { 3058 /* XXX statistic */ 3059 /* drop packet */ 3060 IPFW_RUNLOCK(chain); 3061 return (IP_FW_DENY); 3062 } 3063 dt = (struct divert_tag *)(mtag+1); 3064 dt->cookie = f->rulenum; 3065 if (cmd->arg1 == IP_FW_TABLEARG) 3066 dt->info = tablearg; 3067 else 3068 dt->info = cmd->arg1; 3069 m_tag_prepend(m, mtag); 3070 retval = (cmd->opcode == O_DIVERT) ? 3071 IP_FW_DIVERT : IP_FW_TEE; 3072 goto done; 3073 } 3074 3075 case O_COUNT: 3076 case O_SKIPTO: 3077 f->pcnt++; /* update stats */ 3078 f->bcnt += pktlen; 3079 f->timestamp = time_uptime; 3080 if (cmd->opcode == O_COUNT) 3081 goto next_rule; 3082 /* handle skipto */ 3083 if (f->next_rule == NULL) 3084 lookup_next_rule(f); 3085 f = f->next_rule; 3086 goto again; 3087 3088 case O_REJECT: 3089 /* 3090 * Drop the packet and send a reject notice 3091 * if the packet is not ICMP (or is an ICMP 3092 * query), and it is not multicast/broadcast. 3093 */ 3094 if (hlen > 0 && is_ipv4 && offset == 0 && 3095 (proto != IPPROTO_ICMP || 3096 is_icmp_query(ICMP(ulp))) && 3097 !(m->m_flags & (M_BCAST|M_MCAST)) && 3098 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3099 send_reject(args, cmd->arg1, 3100 offset,ip_len); 3101 m = args->m; 3102 } 3103 /* FALLTHROUGH */ 3104#ifdef INET6 3105 case O_UNREACH6: 3106 if (hlen > 0 && is_ipv6 && 3107 (proto != IPPROTO_ICMPV6 || 3108 (is_icmp6_query(args->f_id.flags) == 1)) && 3109 !(m->m_flags & (M_BCAST|M_MCAST)) && 3110 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3111 send_reject6(args, cmd->arg1, 3112 offset, hlen); 3113 m = args->m; 3114 } 3115 /* FALLTHROUGH */ 3116#endif 3117 case O_DENY: 3118 retval = IP_FW_DENY; 3119 goto done; 3120 3121 case O_FORWARD_IP: 3122 if (args->eh) /* not valid on layer2 pkts */ 3123 break; 3124 if (!q || dyn_dir == MATCH_FORWARD) 3125 args->next_hop = 3126 &((ipfw_insn_sa *)cmd)->sa; 3127 retval = IP_FW_PASS; 3128 goto done; 3129 3130 case O_NETGRAPH: 3131 case O_NGTEE: 3132 args->rule = f; /* report matching rule */ 3133 if (cmd->arg1 == IP_FW_TABLEARG) 3134 args->cookie = tablearg; 3135 else 3136 args->cookie = cmd->arg1; 3137 retval = (cmd->opcode == O_NETGRAPH) ? 3138 IP_FW_NETGRAPH : IP_FW_NGTEE; 3139 goto done; 3140 3141 default: 3142 panic("-- unknown opcode %d\n", cmd->opcode); 3143 } /* end of switch() on opcodes */ 3144 3145 if (cmd->len & F_NOT) 3146 match = !match; 3147 3148 if (match) { 3149 if (cmd->len & F_OR) 3150 skip_or = 1; 3151 } else { 3152 if (!(cmd->len & F_OR)) /* not an OR block, */ 3153 break; /* try next rule */ 3154 } 3155 3156 } /* end of inner for, scan opcodes */ 3157 3158next_rule:; /* try next rule */ 3159 3160 } /* end of outer for, scan rules */ 3161 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3162 IPFW_RUNLOCK(chain); 3163 return (IP_FW_DENY); 3164 3165done: 3166 /* Update statistics */ 3167 f->pcnt++; 3168 f->bcnt += pktlen; 3169 f->timestamp = time_uptime; 3170 IPFW_RUNLOCK(chain); 3171 return (retval); 3172 3173pullup_failed: 3174 if (fw_verbose) 3175 printf("ipfw: pullup failed\n"); 3176 return (IP_FW_DENY); 3177} 3178 3179/* 3180 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3181 * These will be reconstructed on the fly as packets are matched. 3182 */ 3183static void 3184flush_rule_ptrs(struct ip_fw_chain *chain) 3185{ 3186 struct ip_fw *rule; 3187 3188 IPFW_WLOCK_ASSERT(chain); 3189 3190 for (rule = chain->rules; rule; rule = rule->next) 3191 rule->next_rule = NULL; 3192} 3193 3194/* 3195 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3196 * possibly create a rule number and add the rule to the list. 3197 * Update the rule_number in the input struct so the caller knows it as well. 3198 */ 3199static int 3200add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3201{ 3202 struct ip_fw *rule, *f, *prev; 3203 int l = RULESIZE(input_rule); 3204 3205 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3206 return (EINVAL); 3207 3208 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3209 if (rule == NULL) 3210 return (ENOSPC); 3211 3212 bcopy(input_rule, rule, l); 3213 3214 rule->next = NULL; 3215 rule->next_rule = NULL; 3216 3217 rule->pcnt = 0; 3218 rule->bcnt = 0; 3219 rule->timestamp = 0; 3220 3221 IPFW_WLOCK(chain); 3222 3223 if (chain->rules == NULL) { /* default rule */ 3224 chain->rules = rule; 3225 goto done; 3226 } 3227 3228 /* 3229 * If rulenum is 0, find highest numbered rule before the 3230 * default rule, and add autoinc_step 3231 */ 3232 if (autoinc_step < 1) 3233 autoinc_step = 1; 3234 else if (autoinc_step > 1000) 3235 autoinc_step = 1000; 3236 if (rule->rulenum == 0) { 3237 /* 3238 * locate the highest numbered rule before default 3239 */ 3240 for (f = chain->rules; f; f = f->next) { 3241 if (f->rulenum == IPFW_DEFAULT_RULE) 3242 break; 3243 rule->rulenum = f->rulenum; 3244 } 3245 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 3246 rule->rulenum += autoinc_step; 3247 input_rule->rulenum = rule->rulenum; 3248 } 3249 3250 /* 3251 * Now insert the new rule in the right place in the sorted list. 3252 */ 3253 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3254 if (f->rulenum > rule->rulenum) { /* found the location */ 3255 if (prev) { 3256 rule->next = f; 3257 prev->next = rule; 3258 } else { /* head insert */ 3259 rule->next = chain->rules; 3260 chain->rules = rule; 3261 } 3262 break; 3263 } 3264 } 3265 flush_rule_ptrs(chain); 3266done: 3267 static_count++; 3268 static_len += l; 3269 IPFW_WUNLOCK(chain); 3270 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3271 rule->rulenum, static_count);) 3272 return (0); 3273} 3274 3275/** 3276 * Remove a static rule (including derived * dynamic rules) 3277 * and place it on the ``reap list'' for later reclamation. 3278 * The caller is in charge of clearing rule pointers to avoid 3279 * dangling pointers. 3280 * @return a pointer to the next entry. 3281 * Arguments are not checked, so they better be correct. 3282 */ 3283static struct ip_fw * 3284remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 3285{ 3286 struct ip_fw *n; 3287 int l = RULESIZE(rule); 3288 3289 IPFW_WLOCK_ASSERT(chain); 3290 3291 n = rule->next; 3292 IPFW_DYN_LOCK(); 3293 remove_dyn_rule(rule, NULL /* force removal */); 3294 IPFW_DYN_UNLOCK(); 3295 if (prev == NULL) 3296 chain->rules = n; 3297 else 3298 prev->next = n; 3299 static_count--; 3300 static_len -= l; 3301 3302 rule->next = chain->reap; 3303 chain->reap = rule; 3304 3305 return n; 3306} 3307 3308/** 3309 * Reclaim storage associated with a list of rules. This is 3310 * typically the list created using remove_rule. 3311 */ 3312static void 3313reap_rules(struct ip_fw *head) 3314{ 3315 struct ip_fw *rule; 3316 3317 while ((rule = head) != NULL) { 3318 head = head->next; 3319 if (DUMMYNET_LOADED) 3320 ip_dn_ruledel_ptr(rule); 3321 free(rule, M_IPFW); 3322 } 3323} 3324 3325/* 3326 * Remove all rules from a chain (except rules in set RESVD_SET 3327 * unless kill_default = 1). The caller is responsible for 3328 * reclaiming storage for the rules left in chain->reap. 3329 */ 3330static void 3331free_chain(struct ip_fw_chain *chain, int kill_default) 3332{ 3333 struct ip_fw *prev, *rule; 3334 3335 IPFW_WLOCK_ASSERT(chain); 3336 3337 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3338 for (prev = NULL, rule = chain->rules; rule ; ) 3339 if (kill_default || rule->set != RESVD_SET) 3340 rule = remove_rule(chain, rule, prev); 3341 else { 3342 prev = rule; 3343 rule = rule->next; 3344 } 3345} 3346 3347/** 3348 * Remove all rules with given number, and also do set manipulation. 3349 * Assumes chain != NULL && *chain != NULL. 3350 * 3351 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3352 * the next 8 bits are the new set, the top 8 bits are the command: 3353 * 3354 * 0 delete rules with given number 3355 * 1 delete rules with given set number 3356 * 2 move rules with given number to new set 3357 * 3 move rules with given set number to new set 3358 * 4 swap sets with given numbers 3359 */ 3360static int 3361del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3362{ 3363 struct ip_fw *prev = NULL, *rule; 3364 u_int16_t rulenum; /* rule or old_set */ 3365 u_int8_t cmd, new_set; 3366 3367 rulenum = arg & 0xffff; 3368 cmd = (arg >> 24) & 0xff; 3369 new_set = (arg >> 16) & 0xff; 3370 3371 if (cmd > 4) 3372 return EINVAL; 3373 if (new_set > RESVD_SET) 3374 return EINVAL; 3375 if (cmd == 0 || cmd == 2) { 3376 if (rulenum >= IPFW_DEFAULT_RULE) 3377 return EINVAL; 3378 } else { 3379 if (rulenum > RESVD_SET) /* old_set */ 3380 return EINVAL; 3381 } 3382 3383 IPFW_WLOCK(chain); 3384 rule = chain->rules; 3385 chain->reap = NULL; 3386 switch (cmd) { 3387 case 0: /* delete rules with given number */ 3388 /* 3389 * locate first rule to delete 3390 */ 3391 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3392 ; 3393 if (rule->rulenum != rulenum) { 3394 IPFW_WUNLOCK(chain); 3395 return EINVAL; 3396 } 3397 3398 /* 3399 * flush pointers outside the loop, then delete all matching 3400 * rules. prev remains the same throughout the cycle. 3401 */ 3402 flush_rule_ptrs(chain); 3403 while (rule->rulenum == rulenum) 3404 rule = remove_rule(chain, rule, prev); 3405 break; 3406 3407 case 1: /* delete all rules with given set number */ 3408 flush_rule_ptrs(chain); 3409 rule = chain->rules; 3410 while (rule->rulenum < IPFW_DEFAULT_RULE) 3411 if (rule->set == rulenum) 3412 rule = remove_rule(chain, rule, prev); 3413 else { 3414 prev = rule; 3415 rule = rule->next; 3416 } 3417 break; 3418 3419 case 2: /* move rules with given number to new set */ 3420 rule = chain->rules; 3421 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3422 if (rule->rulenum == rulenum) 3423 rule->set = new_set; 3424 break; 3425 3426 case 3: /* move rules with given set number to new set */ 3427 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3428 if (rule->set == rulenum) 3429 rule->set = new_set; 3430 break; 3431 3432 case 4: /* swap two sets */ 3433 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3434 if (rule->set == rulenum) 3435 rule->set = new_set; 3436 else if (rule->set == new_set) 3437 rule->set = rulenum; 3438 break; 3439 } 3440 /* 3441 * Look for rules to reclaim. We grab the list before 3442 * releasing the lock then reclaim them w/o the lock to 3443 * avoid a LOR with dummynet. 3444 */ 3445 rule = chain->reap; 3446 chain->reap = NULL; 3447 IPFW_WUNLOCK(chain); 3448 if (rule) 3449 reap_rules(rule); 3450 return 0; 3451} 3452 3453/* 3454 * Clear counters for a specific rule. 3455 * The enclosing "table" is assumed locked. 3456 */ 3457static void 3458clear_counters(struct ip_fw *rule, int log_only) 3459{ 3460 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3461 3462 if (log_only == 0) { 3463 rule->bcnt = rule->pcnt = 0; 3464 rule->timestamp = 0; 3465 } 3466 if (l->o.opcode == O_LOG) 3467 l->log_left = l->max_log; 3468} 3469 3470/** 3471 * Reset some or all counters on firewall rules. 3472 * @arg frwl is null to clear all entries, or contains a specific 3473 * rule number. 3474 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 3475 */ 3476static int 3477zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 3478{ 3479 struct ip_fw *rule; 3480 char *msg; 3481 3482 IPFW_WLOCK(chain); 3483 if (rulenum == 0) { 3484 norule_counter = 0; 3485 for (rule = chain->rules; rule; rule = rule->next) 3486 clear_counters(rule, log_only); 3487 msg = log_only ? "ipfw: All logging counts reset.\n" : 3488 "ipfw: Accounting cleared.\n"; 3489 } else { 3490 int cleared = 0; 3491 /* 3492 * We can have multiple rules with the same number, so we 3493 * need to clear them all. 3494 */ 3495 for (rule = chain->rules; rule; rule = rule->next) 3496 if (rule->rulenum == rulenum) { 3497 while (rule && rule->rulenum == rulenum) { 3498 clear_counters(rule, log_only); 3499 rule = rule->next; 3500 } 3501 cleared = 1; 3502 break; 3503 } 3504 if (!cleared) { /* we did not find any matching rules */ 3505 IPFW_WUNLOCK(chain); 3506 return (EINVAL); 3507 } 3508 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 3509 "ipfw: Entry %d cleared.\n"; 3510 } 3511 IPFW_WUNLOCK(chain); 3512 3513 if (fw_verbose) 3514 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 3515 return (0); 3516} 3517 3518/* 3519 * Check validity of the structure before insert. 3520 * Fortunately rules are simple, so this mostly need to check rule sizes. 3521 */ 3522static int 3523check_ipfw_struct(struct ip_fw *rule, int size) 3524{ 3525 int l, cmdlen = 0; 3526 int have_action=0; 3527 ipfw_insn *cmd; 3528 3529 if (size < sizeof(*rule)) { 3530 printf("ipfw: rule too short\n"); 3531 return (EINVAL); 3532 } 3533 /* first, check for valid size */ 3534 l = RULESIZE(rule); 3535 if (l != size) { 3536 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3537 return (EINVAL); 3538 } 3539 if (rule->act_ofs >= rule->cmd_len) { 3540 printf("ipfw: bogus action offset (%u > %u)\n", 3541 rule->act_ofs, rule->cmd_len - 1); 3542 return (EINVAL); 3543 } 3544 /* 3545 * Now go for the individual checks. Very simple ones, basically only 3546 * instruction sizes. 3547 */ 3548 for (l = rule->cmd_len, cmd = rule->cmd ; 3549 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3550 cmdlen = F_LEN(cmd); 3551 if (cmdlen > l) { 3552 printf("ipfw: opcode %d size truncated\n", 3553 cmd->opcode); 3554 return EINVAL; 3555 } 3556 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3557 switch (cmd->opcode) { 3558 case O_PROBE_STATE: 3559 case O_KEEP_STATE: 3560 case O_PROTO: 3561 case O_IP_SRC_ME: 3562 case O_IP_DST_ME: 3563 case O_LAYER2: 3564 case O_IN: 3565 case O_FRAG: 3566 case O_DIVERTED: 3567 case O_IPOPT: 3568 case O_IPTOS: 3569 case O_IPPRECEDENCE: 3570 case O_IPVER: 3571 case O_TCPWIN: 3572 case O_TCPFLAGS: 3573 case O_TCPOPTS: 3574 case O_ESTAB: 3575 case O_VERREVPATH: 3576 case O_VERSRCREACH: 3577 case O_ANTISPOOF: 3578 case O_IPSEC: 3579#ifdef INET6 3580 case O_IP6_SRC_ME: 3581 case O_IP6_DST_ME: 3582 case O_EXT_HDR: 3583 case O_IP6: 3584#endif 3585 case O_IP4: 3586 case O_TAG: 3587 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3588 goto bad_size; 3589 break; 3590 3591 case O_UID: 3592 case O_GID: 3593 case O_JAIL: 3594 case O_IP_SRC: 3595 case O_IP_DST: 3596 case O_TCPSEQ: 3597 case O_TCPACK: 3598 case O_PROB: 3599 case O_ICMPTYPE: 3600 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3601 goto bad_size; 3602 break; 3603 3604 case O_LIMIT: 3605 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3606 goto bad_size; 3607 break; 3608 3609 case O_LOG: 3610 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3611 goto bad_size; 3612 3613 ((ipfw_insn_log *)cmd)->log_left = 3614 ((ipfw_insn_log *)cmd)->max_log; 3615 3616 break; 3617 3618 case O_IP_SRC_MASK: 3619 case O_IP_DST_MASK: 3620 /* only odd command lengths */ 3621 if ( !(cmdlen & 1) || cmdlen > 31) 3622 goto bad_size; 3623 break; 3624 3625 case O_IP_SRC_SET: 3626 case O_IP_DST_SET: 3627 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3628 printf("ipfw: invalid set size %d\n", 3629 cmd->arg1); 3630 return EINVAL; 3631 } 3632 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3633 (cmd->arg1+31)/32 ) 3634 goto bad_size; 3635 break; 3636 3637 case O_IP_SRC_LOOKUP: 3638 case O_IP_DST_LOOKUP: 3639 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3640 printf("ipfw: invalid table number %d\n", 3641 cmd->arg1); 3642 return (EINVAL); 3643 } 3644 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3645 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3646 goto bad_size; 3647 break; 3648 3649 case O_MACADDR2: 3650 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3651 goto bad_size; 3652 break; 3653 3654 case O_NOP: 3655 case O_IPID: 3656 case O_IPTTL: 3657 case O_IPLEN: 3658 case O_TCPDATALEN: 3659 case O_TAGGED: 3660 if (cmdlen < 1 || cmdlen > 31) 3661 goto bad_size; 3662 break; 3663 3664 case O_MAC_TYPE: 3665 case O_IP_SRCPORT: 3666 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3667 if (cmdlen < 2 || cmdlen > 31) 3668 goto bad_size; 3669 break; 3670 3671 case O_RECV: 3672 case O_XMIT: 3673 case O_VIA: 3674 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3675 goto bad_size; 3676 break; 3677 3678 case O_ALTQ: 3679 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3680 goto bad_size; 3681 break; 3682 3683 case O_PIPE: 3684 case O_QUEUE: 3685 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3686 goto bad_size; 3687 goto check_action; 3688 3689 case O_FORWARD_IP: 3690#ifdef IPFIREWALL_FORWARD 3691 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3692 goto bad_size; 3693 goto check_action; 3694#else 3695 return EINVAL; 3696#endif 3697 3698 case O_DIVERT: 3699 case O_TEE: 3700 if (ip_divert_ptr == NULL) 3701 return EINVAL; 3702 else 3703 goto check_size; 3704 case O_NETGRAPH: 3705 case O_NGTEE: 3706 if (!NG_IPFW_LOADED) 3707 return EINVAL; 3708 else 3709 goto check_size; 3710 case O_FORWARD_MAC: /* XXX not implemented yet */ 3711 case O_CHECK_STATE: 3712 case O_COUNT: 3713 case O_ACCEPT: 3714 case O_DENY: 3715 case O_REJECT: 3716#ifdef INET6 3717 case O_UNREACH6: 3718#endif 3719 case O_SKIPTO: 3720check_size: 3721 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3722 goto bad_size; 3723check_action: 3724 if (have_action) { 3725 printf("ipfw: opcode %d, multiple actions" 3726 " not allowed\n", 3727 cmd->opcode); 3728 return EINVAL; 3729 } 3730 have_action = 1; 3731 if (l != cmdlen) { 3732 printf("ipfw: opcode %d, action must be" 3733 " last opcode\n", 3734 cmd->opcode); 3735 return EINVAL; 3736 } 3737 break; 3738#ifdef INET6 3739 case O_IP6_SRC: 3740 case O_IP6_DST: 3741 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 3742 F_INSN_SIZE(ipfw_insn)) 3743 goto bad_size; 3744 break; 3745 3746 case O_FLOW6ID: 3747 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3748 ((ipfw_insn_u32 *)cmd)->o.arg1) 3749 goto bad_size; 3750 break; 3751 3752 case O_IP6_SRC_MASK: 3753 case O_IP6_DST_MASK: 3754 if ( !(cmdlen & 1) || cmdlen > 127) 3755 goto bad_size; 3756 break; 3757 case O_ICMP6TYPE: 3758 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 3759 goto bad_size; 3760 break; 3761#endif 3762 3763 default: 3764 switch (cmd->opcode) { 3765#ifndef INET6 3766 case O_IP6_SRC_ME: 3767 case O_IP6_DST_ME: 3768 case O_EXT_HDR: 3769 case O_IP6: 3770 case O_UNREACH6: 3771 case O_IP6_SRC: 3772 case O_IP6_DST: 3773 case O_FLOW6ID: 3774 case O_IP6_SRC_MASK: 3775 case O_IP6_DST_MASK: 3776 case O_ICMP6TYPE: 3777 printf("ipfw: no IPv6 support in kernel\n"); 3778 return EPROTONOSUPPORT; 3779#endif 3780 default: 3781 printf("ipfw: opcode %d, unknown opcode\n", 3782 cmd->opcode); 3783 return EINVAL; 3784 } 3785 } 3786 } 3787 if (have_action == 0) { 3788 printf("ipfw: missing action\n"); 3789 return EINVAL; 3790 } 3791 return 0; 3792 3793bad_size: 3794 printf("ipfw: opcode %d size %d wrong\n", 3795 cmd->opcode, cmdlen); 3796 return EINVAL; 3797} 3798 3799/* 3800 * Copy the static and dynamic rules to the supplied buffer 3801 * and return the amount of space actually used. 3802 */ 3803static size_t 3804ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 3805{ 3806 char *bp = buf; 3807 char *ep = bp + space; 3808 struct ip_fw *rule; 3809 int i; 3810 3811 /* XXX this can take a long time and locking will block packet flow */ 3812 IPFW_RLOCK(chain); 3813 for (rule = chain->rules; rule ; rule = rule->next) { 3814 /* 3815 * Verify the entry fits in the buffer in case the 3816 * rules changed between calculating buffer space and 3817 * now. This would be better done using a generation 3818 * number but should suffice for now. 3819 */ 3820 i = RULESIZE(rule); 3821 if (bp + i <= ep) { 3822 bcopy(rule, bp, i); 3823 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 3824 sizeof(set_disable)); 3825 bp += i; 3826 } 3827 } 3828 IPFW_RUNLOCK(chain); 3829 if (ipfw_dyn_v) { 3830 ipfw_dyn_rule *p, *last = NULL; 3831 3832 IPFW_DYN_LOCK(); 3833 for (i = 0 ; i < curr_dyn_buckets; i++) 3834 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 3835 if (bp + sizeof *p <= ep) { 3836 ipfw_dyn_rule *dst = 3837 (ipfw_dyn_rule *)bp; 3838 bcopy(p, dst, sizeof *p); 3839 bcopy(&(p->rule->rulenum), &(dst->rule), 3840 sizeof(p->rule->rulenum)); 3841 /* 3842 * store a non-null value in "next". 3843 * The userland code will interpret a 3844 * NULL here as a marker 3845 * for the last dynamic rule. 3846 */ 3847 bcopy(&dst, &dst->next, sizeof(dst)); 3848 last = dst; 3849 dst->expire = 3850 TIME_LEQ(dst->expire, time_uptime) ? 3851 0 : dst->expire - time_uptime ; 3852 bp += sizeof(ipfw_dyn_rule); 3853 } 3854 } 3855 IPFW_DYN_UNLOCK(); 3856 if (last != NULL) /* mark last dynamic rule */ 3857 bzero(&last->next, sizeof(last)); 3858 } 3859 return (bp - (char *)buf); 3860} 3861 3862 3863/** 3864 * {set|get}sockopt parser. 3865 */ 3866static int 3867ipfw_ctl(struct sockopt *sopt) 3868{ 3869#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3870 int error, rule_num; 3871 size_t size; 3872 struct ip_fw *buf, *rule; 3873 u_int32_t rulenum[2]; 3874 3875 error = suser(sopt->sopt_td); 3876 if (error) 3877 return (error); 3878 3879 /* 3880 * Disallow modifications in really-really secure mode, but still allow 3881 * the logging counters to be reset. 3882 */ 3883 if (sopt->sopt_name == IP_FW_ADD || 3884 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 3885 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3886 if (error) 3887 return (error); 3888 } 3889 3890 error = 0; 3891 3892 switch (sopt->sopt_name) { 3893 case IP_FW_GET: 3894 /* 3895 * pass up a copy of the current rules. Static rules 3896 * come first (the last of which has number IPFW_DEFAULT_RULE), 3897 * followed by a possibly empty list of dynamic rule. 3898 * The last dynamic rule has NULL in the "next" field. 3899 * 3900 * Note that the calculated size is used to bound the 3901 * amount of data returned to the user. The rule set may 3902 * change between calculating the size and returning the 3903 * data in which case we'll just return what fits. 3904 */ 3905 size = static_len; /* size of static rules */ 3906 if (ipfw_dyn_v) /* add size of dyn.rules */ 3907 size += (dyn_count * sizeof(ipfw_dyn_rule)); 3908 3909 /* 3910 * XXX todo: if the user passes a short length just to know 3911 * how much room is needed, do not bother filling up the 3912 * buffer, just jump to the sooptcopyout. 3913 */ 3914 buf = malloc(size, M_TEMP, M_WAITOK); 3915 error = sooptcopyout(sopt, buf, 3916 ipfw_getrules(&layer3_chain, buf, size)); 3917 free(buf, M_TEMP); 3918 break; 3919 3920 case IP_FW_FLUSH: 3921 /* 3922 * Normally we cannot release the lock on each iteration. 3923 * We could do it here only because we start from the head all 3924 * the times so there is no risk of missing some entries. 3925 * On the other hand, the risk is that we end up with 3926 * a very inconsistent ruleset, so better keep the lock 3927 * around the whole cycle. 3928 * 3929 * XXX this code can be improved by resetting the head of 3930 * the list to point to the default rule, and then freeing 3931 * the old list without the need for a lock. 3932 */ 3933 3934 IPFW_WLOCK(&layer3_chain); 3935 layer3_chain.reap = NULL; 3936 free_chain(&layer3_chain, 0 /* keep default rule */); 3937 rule = layer3_chain.reap, layer3_chain.reap = NULL; 3938 IPFW_WUNLOCK(&layer3_chain); 3939 if (layer3_chain.reap != NULL) 3940 reap_rules(rule); 3941 break; 3942 3943 case IP_FW_ADD: 3944 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 3945 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 3946 sizeof(struct ip_fw) ); 3947 if (error == 0) 3948 error = check_ipfw_struct(rule, sopt->sopt_valsize); 3949 if (error == 0) { 3950 error = add_rule(&layer3_chain, rule); 3951 size = RULESIZE(rule); 3952 if (!error && sopt->sopt_dir == SOPT_GET) 3953 error = sooptcopyout(sopt, rule, size); 3954 } 3955 free(rule, M_TEMP); 3956 break; 3957 3958 case IP_FW_DEL: 3959 /* 3960 * IP_FW_DEL is used for deleting single rules or sets, 3961 * and (ab)used to atomically manipulate sets. Argument size 3962 * is used to distinguish between the two: 3963 * sizeof(u_int32_t) 3964 * delete single rule or set of rules, 3965 * or reassign rules (or sets) to a different set. 3966 * 2*sizeof(u_int32_t) 3967 * atomic disable/enable sets. 3968 * first u_int32_t contains sets to be disabled, 3969 * second u_int32_t contains sets to be enabled. 3970 */ 3971 error = sooptcopyin(sopt, rulenum, 3972 2*sizeof(u_int32_t), sizeof(u_int32_t)); 3973 if (error) 3974 break; 3975 size = sopt->sopt_valsize; 3976 if (size == sizeof(u_int32_t)) /* delete or reassign */ 3977 error = del_entry(&layer3_chain, rulenum[0]); 3978 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 3979 set_disable = 3980 (set_disable | rulenum[0]) & ~rulenum[1] & 3981 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 3982 else 3983 error = EINVAL; 3984 break; 3985 3986 case IP_FW_ZERO: 3987 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 3988 rule_num = 0; 3989 if (sopt->sopt_val != 0) { 3990 error = sooptcopyin(sopt, &rule_num, 3991 sizeof(int), sizeof(int)); 3992 if (error) 3993 break; 3994 } 3995 error = zero_entry(&layer3_chain, rule_num, 3996 sopt->sopt_name == IP_FW_RESETLOG); 3997 break; 3998 3999 case IP_FW_TABLE_ADD: 4000 { 4001 ipfw_table_entry ent; 4002 4003 error = sooptcopyin(sopt, &ent, 4004 sizeof(ent), sizeof(ent)); 4005 if (error) 4006 break; 4007 error = add_table_entry(&layer3_chain, ent.tbl, 4008 ent.addr, ent.masklen, ent.value); 4009 } 4010 break; 4011 4012 case IP_FW_TABLE_DEL: 4013 { 4014 ipfw_table_entry ent; 4015 4016 error = sooptcopyin(sopt, &ent, 4017 sizeof(ent), sizeof(ent)); 4018 if (error) 4019 break; 4020 error = del_table_entry(&layer3_chain, ent.tbl, 4021 ent.addr, ent.masklen); 4022 } 4023 break; 4024 4025 case IP_FW_TABLE_FLUSH: 4026 { 4027 u_int16_t tbl; 4028 4029 error = sooptcopyin(sopt, &tbl, 4030 sizeof(tbl), sizeof(tbl)); 4031 if (error) 4032 break; 4033 IPFW_WLOCK(&layer3_chain); 4034 error = flush_table(&layer3_chain, tbl); 4035 IPFW_WUNLOCK(&layer3_chain); 4036 } 4037 break; 4038 4039 case IP_FW_TABLE_GETSIZE: 4040 { 4041 u_int32_t tbl, cnt; 4042 4043 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4044 sizeof(tbl)))) 4045 break; 4046 IPFW_RLOCK(&layer3_chain); 4047 error = count_table(&layer3_chain, tbl, &cnt); 4048 IPFW_RUNLOCK(&layer3_chain); 4049 if (error) 4050 break; 4051 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4052 } 4053 break; 4054 4055 case IP_FW_TABLE_LIST: 4056 { 4057 ipfw_table *tbl; 4058 4059 if (sopt->sopt_valsize < sizeof(*tbl)) { 4060 error = EINVAL; 4061 break; 4062 } 4063 size = sopt->sopt_valsize; 4064 tbl = malloc(size, M_TEMP, M_WAITOK); 4065 if (tbl == NULL) { 4066 error = ENOMEM; 4067 break; 4068 } 4069 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4070 if (error) { 4071 free(tbl, M_TEMP); 4072 break; 4073 } 4074 tbl->size = (size - sizeof(*tbl)) / 4075 sizeof(ipfw_table_entry); 4076 IPFW_RLOCK(&layer3_chain); 4077 error = dump_table(&layer3_chain, tbl); 4078 IPFW_RUNLOCK(&layer3_chain); 4079 if (error) { 4080 free(tbl, M_TEMP); 4081 break; 4082 } 4083 error = sooptcopyout(sopt, tbl, size); 4084 free(tbl, M_TEMP); 4085 } 4086 break; 4087 4088 default: 4089 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4090 error = EINVAL; 4091 } 4092 4093 return (error); 4094#undef RULE_MAXSIZE 4095} 4096 4097/** 4098 * dummynet needs a reference to the default rule, because rules can be 4099 * deleted while packets hold a reference to them. When this happens, 4100 * dummynet changes the reference to the default rule (it could well be a 4101 * NULL pointer, but this way we do not need to check for the special 4102 * case, plus here he have info on the default behaviour). 4103 */ 4104struct ip_fw *ip_fw_default_rule; 4105 4106/* 4107 * This procedure is only used to handle keepalives. It is invoked 4108 * every dyn_keepalive_period 4109 */ 4110static void 4111ipfw_tick(void * __unused unused) 4112{ 4113 struct mbuf *m0, *m, *mnext, **mtailp; 4114 int i; 4115 ipfw_dyn_rule *q; 4116 4117 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 4118 goto done; 4119 4120 /* 4121 * We make a chain of packets to go out here -- not deferring 4122 * until after we drop the IPFW dynamic rule lock would result 4123 * in a lock order reversal with the normal packet input -> ipfw 4124 * call stack. 4125 */ 4126 m0 = NULL; 4127 mtailp = &m0; 4128 IPFW_DYN_LOCK(); 4129 for (i = 0 ; i < curr_dyn_buckets ; i++) { 4130 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 4131 if (q->dyn_type == O_LIMIT_PARENT) 4132 continue; 4133 if (q->id.proto != IPPROTO_TCP) 4134 continue; 4135 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4136 continue; 4137 if (TIME_LEQ( time_uptime+dyn_keepalive_interval, 4138 q->expire)) 4139 continue; /* too early */ 4140 if (TIME_LEQ(q->expire, time_uptime)) 4141 continue; /* too late, rule expired */ 4142 4143 *mtailp = send_pkt(&(q->id), q->ack_rev - 1, 4144 q->ack_fwd, TH_SYN); 4145 if (*mtailp != NULL) 4146 mtailp = &(*mtailp)->m_nextpkt; 4147 *mtailp = send_pkt(&(q->id), q->ack_fwd - 1, 4148 q->ack_rev, 0); 4149 if (*mtailp != NULL) 4150 mtailp = &(*mtailp)->m_nextpkt; 4151 } 4152 } 4153 IPFW_DYN_UNLOCK(); 4154 for (m = mnext = m0; m != NULL; m = mnext) { 4155 mnext = m->m_nextpkt; 4156 m->m_nextpkt = NULL; 4157 ip_output(m, NULL, NULL, 0, NULL, NULL); 4158 } 4159done: 4160 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 4161} 4162 4163int 4164ipfw_init(void) 4165{ 4166 struct ip_fw default_rule; 4167 int error; 4168 4169#ifdef INET6 4170 /* Setup IPv6 fw sysctl tree. */ 4171 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4172 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4173 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4174 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4175 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4176 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4177 &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4178 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4179 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4180 &fw_deny_unknown_exthdrs, 0, 4181 "Deny packets with unknown IPv6 Extension Headers"); 4182#endif 4183 4184 layer3_chain.rules = NULL; 4185 IPFW_LOCK_INIT(&layer3_chain); 4186 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule zone", 4187 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4188 UMA_ALIGN_PTR, 0); 4189 IPFW_DYN_LOCK_INIT(); 4190 callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE); 4191 4192 bzero(&default_rule, sizeof default_rule); 4193 4194 default_rule.act_ofs = 0; 4195 default_rule.rulenum = IPFW_DEFAULT_RULE; 4196 default_rule.cmd_len = 1; 4197 default_rule.set = RESVD_SET; 4198 4199 default_rule.cmd[0].len = 1; 4200 default_rule.cmd[0].opcode = 4201#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4202 1 ? O_ACCEPT : 4203#endif 4204 O_DENY; 4205 4206 error = add_rule(&layer3_chain, &default_rule); 4207 if (error != 0) { 4208 printf("ipfw2: error %u initializing default rule " 4209 "(support disabled)\n", error); 4210 IPFW_DYN_LOCK_DESTROY(); 4211 IPFW_LOCK_DESTROY(&layer3_chain); 4212 uma_zdestroy(ipfw_dyn_rule_zone); 4213 return (error); 4214 } 4215 4216 ip_fw_default_rule = layer3_chain.rules; 4217 printf("ipfw2 " 4218#ifdef INET6 4219 "(+ipv6) " 4220#endif 4221 "initialized, divert %s, " 4222 "rule-based forwarding " 4223#ifdef IPFIREWALL_FORWARD 4224 "enabled, " 4225#else 4226 "disabled, " 4227#endif 4228 "default to %s, logging ", 4229#ifdef IPDIVERT 4230 "enabled", 4231#else 4232 "loadable", 4233#endif 4234 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4235 4236#ifdef IPFIREWALL_VERBOSE 4237 fw_verbose = 1; 4238#endif 4239#ifdef IPFIREWALL_VERBOSE_LIMIT 4240 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4241#endif 4242 if (fw_verbose == 0) 4243 printf("disabled\n"); 4244 else if (verbose_limit == 0) 4245 printf("unlimited\n"); 4246 else 4247 printf("limited to %d packets/entry by default\n", 4248 verbose_limit); 4249 4250 error = init_tables(&layer3_chain); 4251 if (error) { 4252 IPFW_DYN_LOCK_DESTROY(); 4253 IPFW_LOCK_DESTROY(&layer3_chain); 4254 uma_zdestroy(ipfw_dyn_rule_zone); 4255 return (error); 4256 } 4257 ip_fw_ctl_ptr = ipfw_ctl; 4258 ip_fw_chk_ptr = ipfw_chk; 4259 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 4260 4261 return (0); 4262} 4263 4264void 4265ipfw_destroy(void) 4266{ 4267 struct ip_fw *reap; 4268 4269 ip_fw_chk_ptr = NULL; 4270 ip_fw_ctl_ptr = NULL; 4271 callout_drain(&ipfw_timeout); 4272 IPFW_WLOCK(&layer3_chain); 4273 flush_tables(&layer3_chain); 4274 layer3_chain.reap = NULL; 4275 free_chain(&layer3_chain, 1 /* kill default rule */); 4276 reap = layer3_chain.reap, layer3_chain.reap = NULL; 4277 IPFW_WUNLOCK(&layer3_chain); 4278 if (reap != NULL) 4279 reap_rules(reap); 4280 IPFW_DYN_LOCK_DESTROY(); 4281 uma_zdestroy(ipfw_dyn_rule_zone); 4282 IPFW_LOCK_DESTROY(&layer3_chain); 4283 4284#ifdef INET6 4285 /* Free IPv6 fw sysctl tree. */ 4286 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4287#endif 4288 4289 printf("IP firewall unloaded\n"); 4290} 4291