ip_fw2.c revision 159635
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 159635 2006-06-15 08:54:29Z 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( 1427 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1428 __func__, cmd->o.opcode, 1429 (args->f_id.src_ip), (args->f_id.src_port), 1430 (args->f_id.dst_ip), (args->f_id.dst_port)); 1431 ) 1432 1433 IPFW_DYN_LOCK(); 1434 1435 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1436 1437 if (q != NULL) { /* should never occur */ 1438 if (last_log != time_uptime) { 1439 last_log = time_uptime; 1440 printf("ipfw: %s: entry already present, done\n", 1441 __func__); 1442 } 1443 IPFW_DYN_UNLOCK(); 1444 return (0); 1445 } 1446 1447 if (dyn_count >= dyn_max) 1448 /* Run out of slots, try to remove any expired rule. */ 1449 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1450 1451 if (dyn_count >= dyn_max) { 1452 if (last_log != time_uptime) { 1453 last_log = time_uptime; 1454 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1455 } 1456 IPFW_DYN_UNLOCK(); 1457 return (1); /* cannot install, notify caller */ 1458 } 1459 1460 switch (cmd->o.opcode) { 1461 case O_KEEP_STATE: /* bidir rule */ 1462 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1463 break; 1464 1465 case O_LIMIT: { /* limit number of sessions */ 1466 struct ipfw_flow_id id; 1467 ipfw_dyn_rule *parent; 1468 uint16_t limit_mask = cmd->limit_mask; 1469 1470 DEB( 1471 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1472 __func__, cmd->conn_limit); 1473 ) 1474 1475 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1476 id.proto = args->f_id.proto; 1477 id.addr_type = args->f_id.addr_type; 1478 1479 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1480 if (limit_mask & DYN_SRC_ADDR) 1481 id.src_ip6 = args->f_id.src_ip6; 1482 if (limit_mask & DYN_DST_ADDR) 1483 id.dst_ip6 = args->f_id.dst_ip6; 1484 } else { 1485 if (limit_mask & DYN_SRC_ADDR) 1486 id.src_ip = args->f_id.src_ip; 1487 if (limit_mask & DYN_DST_ADDR) 1488 id.dst_ip = args->f_id.dst_ip; 1489 } 1490 if (limit_mask & DYN_SRC_PORT) 1491 id.src_port = args->f_id.src_port; 1492 if (limit_mask & DYN_DST_PORT) 1493 id.dst_port = args->f_id.dst_port; 1494 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1495 printf("ipfw: %s: add parent failed\n", __func__); 1496 IPFW_DYN_UNLOCK(); 1497 return (1); 1498 } 1499 1500 if (parent->count >= cmd->conn_limit) { 1501 /* See if we can remove some expired rule. */ 1502 remove_dyn_rule(rule, parent); 1503 if (parent->count >= cmd->conn_limit) { 1504 if (fw_verbose && last_log != time_uptime) { 1505 last_log = time_uptime; 1506 log(LOG_SECURITY | LOG_DEBUG, 1507 "drop session, too many entries\n"); 1508 } 1509 IPFW_DYN_UNLOCK(); 1510 return (1); 1511 } 1512 } 1513 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1514 break; 1515 } 1516 default: 1517 printf("ipfw: %s: unknown dynamic rule type %u\n", 1518 __func__, cmd->o.opcode); 1519 IPFW_DYN_UNLOCK(); 1520 return (1); 1521 } 1522 1523 /* XXX just set lifetime */ 1524 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1525 1526 IPFW_DYN_UNLOCK(); 1527 return (0); 1528} 1529 1530/* 1531 * Generate a TCP packet, containing either a RST or a keepalive. 1532 * When flags & TH_RST, we are sending a RST packet, because of a 1533 * "reset" action matched the packet. 1534 * Otherwise we are sending a keepalive, and flags & TH_ 1535 */ 1536static struct mbuf * 1537send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags) 1538{ 1539 struct mbuf *m; 1540 struct ip *ip; 1541 struct tcphdr *tcp; 1542 1543 MGETHDR(m, M_DONTWAIT, MT_DATA); 1544 if (m == 0) 1545 return (NULL); 1546 m->m_pkthdr.rcvif = (struct ifnet *)0; 1547 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1548 m->m_data += max_linkhdr; 1549 1550 ip = mtod(m, struct ip *); 1551 bzero(ip, m->m_len); 1552 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1553 ip->ip_p = IPPROTO_TCP; 1554 tcp->th_off = 5; 1555 /* 1556 * Assume we are sending a RST (or a keepalive in the reverse 1557 * direction), swap src and destination addresses and ports. 1558 */ 1559 ip->ip_src.s_addr = htonl(id->dst_ip); 1560 ip->ip_dst.s_addr = htonl(id->src_ip); 1561 tcp->th_sport = htons(id->dst_port); 1562 tcp->th_dport = htons(id->src_port); 1563 if (flags & TH_RST) { /* we are sending a RST */ 1564 if (flags & TH_ACK) { 1565 tcp->th_seq = htonl(ack); 1566 tcp->th_ack = htonl(0); 1567 tcp->th_flags = TH_RST; 1568 } else { 1569 if (flags & TH_SYN) 1570 seq++; 1571 tcp->th_seq = htonl(0); 1572 tcp->th_ack = htonl(seq); 1573 tcp->th_flags = TH_RST | TH_ACK; 1574 } 1575 } else { 1576 /* 1577 * We are sending a keepalive. flags & TH_SYN determines 1578 * the direction, forward if set, reverse if clear. 1579 * NOTE: seq and ack are always assumed to be correct 1580 * as set by the caller. This may be confusing... 1581 */ 1582 if (flags & TH_SYN) { 1583 /* 1584 * we have to rewrite the correct addresses! 1585 */ 1586 ip->ip_dst.s_addr = htonl(id->dst_ip); 1587 ip->ip_src.s_addr = htonl(id->src_ip); 1588 tcp->th_dport = htons(id->dst_port); 1589 tcp->th_sport = htons(id->src_port); 1590 } 1591 tcp->th_seq = htonl(seq); 1592 tcp->th_ack = htonl(ack); 1593 tcp->th_flags = TH_ACK; 1594 } 1595 /* 1596 * set ip_len to the payload size so we can compute 1597 * the tcp checksum on the pseudoheader 1598 * XXX check this, could save a couple of words ? 1599 */ 1600 ip->ip_len = htons(sizeof(struct tcphdr)); 1601 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1602 /* 1603 * now fill fields left out earlier 1604 */ 1605 ip->ip_ttl = ip_defttl; 1606 ip->ip_len = m->m_pkthdr.len; 1607 m->m_flags |= M_SKIP_FIREWALL; 1608 return (m); 1609} 1610 1611/* 1612 * sends a reject message, consuming the mbuf passed as an argument. 1613 */ 1614static void 1615send_reject(struct ip_fw_args *args, int code, u_short offset, int ip_len) 1616{ 1617 1618 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1619 /* We need the IP header in host order for icmp_error(). */ 1620 if (args->eh != NULL) { 1621 struct ip *ip = mtod(args->m, struct ip *); 1622 ip->ip_len = ntohs(ip->ip_len); 1623 ip->ip_off = ntohs(ip->ip_off); 1624 } 1625 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1626 } else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) { 1627 struct tcphdr *const tcp = 1628 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1629 if ( (tcp->th_flags & TH_RST) == 0) { 1630 struct mbuf *m; 1631 m = send_pkt(&(args->f_id), ntohl(tcp->th_seq), 1632 ntohl(tcp->th_ack), 1633 tcp->th_flags | TH_RST); 1634 if (m != NULL) 1635 ip_output(m, NULL, NULL, 0, NULL, NULL); 1636 } 1637 m_freem(args->m); 1638 } else 1639 m_freem(args->m); 1640 args->m = NULL; 1641} 1642 1643/** 1644 * 1645 * Given an ip_fw *, lookup_next_rule will return a pointer 1646 * to the next rule, which can be either the jump 1647 * target (for skipto instructions) or the next one in the list (in 1648 * all other cases including a missing jump target). 1649 * The result is also written in the "next_rule" field of the rule. 1650 * Backward jumps are not allowed, so start looking from the next 1651 * rule... 1652 * 1653 * This never returns NULL -- in case we do not have an exact match, 1654 * the next rule is returned. When the ruleset is changed, 1655 * pointers are flushed so we are always correct. 1656 */ 1657 1658static struct ip_fw * 1659lookup_next_rule(struct ip_fw *me) 1660{ 1661 struct ip_fw *rule = NULL; 1662 ipfw_insn *cmd; 1663 1664 /* look for action, in case it is a skipto */ 1665 cmd = ACTION_PTR(me); 1666 if (cmd->opcode == O_LOG) 1667 cmd += F_LEN(cmd); 1668 if (cmd->opcode == O_ALTQ) 1669 cmd += F_LEN(cmd); 1670 if (cmd->opcode == O_TAG) 1671 cmd += F_LEN(cmd); 1672 if ( cmd->opcode == O_SKIPTO ) 1673 for (rule = me->next; rule ; rule = rule->next) 1674 if (rule->rulenum >= cmd->arg1) 1675 break; 1676 if (rule == NULL) /* failure or not a skipto */ 1677 rule = me->next; 1678 me->next_rule = rule; 1679 return rule; 1680} 1681 1682static int 1683add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1684 uint8_t mlen, uint32_t value) 1685{ 1686 struct radix_node_head *rnh; 1687 struct table_entry *ent; 1688 1689 if (tbl >= IPFW_TABLES_MAX) 1690 return (EINVAL); 1691 rnh = ch->tables[tbl]; 1692 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1693 if (ent == NULL) 1694 return (ENOMEM); 1695 ent->value = value; 1696 ent->addr.sin_len = ent->mask.sin_len = 8; 1697 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1698 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1699 IPFW_WLOCK(&layer3_chain); 1700 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1701 NULL) { 1702 IPFW_WUNLOCK(&layer3_chain); 1703 free(ent, M_IPFW_TBL); 1704 return (EEXIST); 1705 } 1706 IPFW_WUNLOCK(&layer3_chain); 1707 return (0); 1708} 1709 1710static int 1711del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1712 uint8_t mlen) 1713{ 1714 struct radix_node_head *rnh; 1715 struct table_entry *ent; 1716 struct sockaddr_in sa, mask; 1717 1718 if (tbl >= IPFW_TABLES_MAX) 1719 return (EINVAL); 1720 rnh = ch->tables[tbl]; 1721 sa.sin_len = mask.sin_len = 8; 1722 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1723 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1724 IPFW_WLOCK(ch); 1725 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1726 if (ent == NULL) { 1727 IPFW_WUNLOCK(ch); 1728 return (ESRCH); 1729 } 1730 IPFW_WUNLOCK(ch); 1731 free(ent, M_IPFW_TBL); 1732 return (0); 1733} 1734 1735static int 1736flush_table_entry(struct radix_node *rn, void *arg) 1737{ 1738 struct radix_node_head * const rnh = arg; 1739 struct table_entry *ent; 1740 1741 ent = (struct table_entry *) 1742 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1743 if (ent != NULL) 1744 free(ent, M_IPFW_TBL); 1745 return (0); 1746} 1747 1748static int 1749flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1750{ 1751 struct radix_node_head *rnh; 1752 1753 IPFW_WLOCK_ASSERT(ch); 1754 1755 if (tbl >= IPFW_TABLES_MAX) 1756 return (EINVAL); 1757 rnh = ch->tables[tbl]; 1758 KASSERT(rnh != NULL, ("NULL IPFW table")); 1759 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1760 return (0); 1761} 1762 1763static void 1764flush_tables(struct ip_fw_chain *ch) 1765{ 1766 uint16_t tbl; 1767 1768 IPFW_WLOCK_ASSERT(ch); 1769 1770 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1771 flush_table(ch, tbl); 1772} 1773 1774static int 1775init_tables(struct ip_fw_chain *ch) 1776{ 1777 int i; 1778 uint16_t j; 1779 1780 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1781 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1782 for (j = 0; j < i; j++) { 1783 (void) flush_table(ch, j); 1784 } 1785 return (ENOMEM); 1786 } 1787 } 1788 return (0); 1789} 1790 1791static int 1792lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1793 uint32_t *val) 1794{ 1795 struct radix_node_head *rnh; 1796 struct table_entry *ent; 1797 struct sockaddr_in sa; 1798 1799 if (tbl >= IPFW_TABLES_MAX) 1800 return (0); 1801 rnh = ch->tables[tbl]; 1802 sa.sin_len = 8; 1803 sa.sin_addr.s_addr = addr; 1804 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1805 if (ent != NULL) { 1806 *val = ent->value; 1807 return (1); 1808 } 1809 return (0); 1810} 1811 1812static int 1813count_table_entry(struct radix_node *rn, void *arg) 1814{ 1815 u_int32_t * const cnt = arg; 1816 1817 (*cnt)++; 1818 return (0); 1819} 1820 1821static int 1822count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1823{ 1824 struct radix_node_head *rnh; 1825 1826 if (tbl >= IPFW_TABLES_MAX) 1827 return (EINVAL); 1828 rnh = ch->tables[tbl]; 1829 *cnt = 0; 1830 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1831 return (0); 1832} 1833 1834static int 1835dump_table_entry(struct radix_node *rn, void *arg) 1836{ 1837 struct table_entry * const n = (struct table_entry *)rn; 1838 ipfw_table * const tbl = arg; 1839 ipfw_table_entry *ent; 1840 1841 if (tbl->cnt == tbl->size) 1842 return (1); 1843 ent = &tbl->ent[tbl->cnt]; 1844 ent->tbl = tbl->tbl; 1845 if (in_nullhost(n->mask.sin_addr)) 1846 ent->masklen = 0; 1847 else 1848 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1849 ent->addr = n->addr.sin_addr.s_addr; 1850 ent->value = n->value; 1851 tbl->cnt++; 1852 return (0); 1853} 1854 1855static int 1856dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1857{ 1858 struct radix_node_head *rnh; 1859 1860 if (tbl->tbl >= IPFW_TABLES_MAX) 1861 return (EINVAL); 1862 rnh = ch->tables[tbl->tbl]; 1863 tbl->cnt = 0; 1864 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1865 return (0); 1866} 1867 1868static void 1869fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1870{ 1871 struct ucred *cr; 1872 1873 if (inp->inp_socket != NULL) { 1874 cr = inp->inp_socket->so_cred; 1875 ugp->fw_prid = jailed(cr) ? 1876 cr->cr_prison->pr_id : -1; 1877 ugp->fw_uid = cr->cr_uid; 1878 ugp->fw_ngroups = cr->cr_ngroups; 1879 bcopy(cr->cr_groups, ugp->fw_groups, 1880 sizeof(ugp->fw_groups)); 1881 } 1882} 1883 1884static int 1885check_uidgid(ipfw_insn_u32 *insn, 1886 int proto, struct ifnet *oif, 1887 struct in_addr dst_ip, u_int16_t dst_port, 1888 struct in_addr src_ip, u_int16_t src_port, 1889 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 1890{ 1891 struct inpcbinfo *pi; 1892 int wildcard; 1893 struct inpcb *pcb; 1894 int match; 1895 gid_t *gp; 1896 1897 /* 1898 * Check to see if the UDP or TCP stack supplied us with 1899 * the PCB. If so, rather then holding a lock and looking 1900 * up the PCB, we can use the one that was supplied. 1901 */ 1902 if (inp && *lookup == 0) { 1903 INP_LOCK_ASSERT(inp); 1904 if (inp->inp_socket != NULL) { 1905 fill_ugid_cache(inp, ugp); 1906 *lookup = 1; 1907 } 1908 } 1909 /* 1910 * If we have already been here and the packet has no 1911 * PCB entry associated with it, then we can safely 1912 * assume that this is a no match. 1913 */ 1914 if (*lookup == -1) 1915 return (0); 1916 if (proto == IPPROTO_TCP) { 1917 wildcard = 0; 1918 pi = &tcbinfo; 1919 } else if (proto == IPPROTO_UDP) { 1920 wildcard = 1; 1921 pi = &udbinfo; 1922 } else 1923 return 0; 1924 match = 0; 1925 if (*lookup == 0) { 1926 INP_INFO_RLOCK(pi); 1927 pcb = (oif) ? 1928 in_pcblookup_hash(pi, 1929 dst_ip, htons(dst_port), 1930 src_ip, htons(src_port), 1931 wildcard, oif) : 1932 in_pcblookup_hash(pi, 1933 src_ip, htons(src_port), 1934 dst_ip, htons(dst_port), 1935 wildcard, NULL); 1936 if (pcb != NULL) { 1937 INP_LOCK(pcb); 1938 if (pcb->inp_socket != NULL) { 1939 fill_ugid_cache(pcb, ugp); 1940 *lookup = 1; 1941 } 1942 INP_UNLOCK(pcb); 1943 } 1944 INP_INFO_RUNLOCK(pi); 1945 if (*lookup == 0) { 1946 /* 1947 * If the lookup did not yield any results, there 1948 * is no sense in coming back and trying again. So 1949 * we can set lookup to -1 and ensure that we wont 1950 * bother the pcb system again. 1951 */ 1952 *lookup = -1; 1953 return (0); 1954 } 1955 } 1956 if (insn->o.opcode == O_UID) 1957 match = (ugp->fw_uid == (uid_t)insn->d[0]); 1958 else if (insn->o.opcode == O_GID) { 1959 for (gp = ugp->fw_groups; 1960 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 1961 if (*gp == (gid_t)insn->d[0]) { 1962 match = 1; 1963 break; 1964 } 1965 } else if (insn->o.opcode == O_JAIL) 1966 match = (ugp->fw_prid == (int)insn->d[0]); 1967 return match; 1968} 1969 1970/* 1971 * The main check routine for the firewall. 1972 * 1973 * All arguments are in args so we can modify them and return them 1974 * back to the caller. 1975 * 1976 * Parameters: 1977 * 1978 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1979 * Starts with the IP header. 1980 * args->eh (in) Mac header if present, or NULL for layer3 packet. 1981 * args->oif Outgoing interface, or NULL if packet is incoming. 1982 * The incoming interface is in the mbuf. (in) 1983 * args->divert_rule (in/out) 1984 * Skip up to the first rule past this rule number; 1985 * upon return, non-zero port number for divert or tee. 1986 * 1987 * args->rule Pointer to the last matching rule (in/out) 1988 * args->next_hop Socket we are forwarding to (out). 1989 * args->f_id Addresses grabbed from the packet (out) 1990 * args->cookie a cookie depending on rule action 1991 * 1992 * Return value: 1993 * 1994 * IP_FW_PASS the packet must be accepted 1995 * IP_FW_DENY the packet must be dropped 1996 * IP_FW_DIVERT divert packet, port in m_tag 1997 * IP_FW_TEE tee packet, port in m_tag 1998 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 1999 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2000 * 2001 */ 2002 2003int 2004ipfw_chk(struct ip_fw_args *args) 2005{ 2006 /* 2007 * Local variables hold state during the processing of a packet. 2008 * 2009 * IMPORTANT NOTE: to speed up the processing of rules, there 2010 * are some assumption on the values of the variables, which 2011 * are documented here. Should you change them, please check 2012 * the implementation of the various instructions to make sure 2013 * that they still work. 2014 * 2015 * args->eh The MAC header. It is non-null for a layer2 2016 * packet, it is NULL for a layer-3 packet. 2017 * 2018 * m | args->m Pointer to the mbuf, as received from the caller. 2019 * It may change if ipfw_chk() does an m_pullup, or if it 2020 * consumes the packet because it calls send_reject(). 2021 * XXX This has to change, so that ipfw_chk() never modifies 2022 * or consumes the buffer. 2023 * ip is simply an alias of the value of m, and it is kept 2024 * in sync with it (the packet is supposed to start with 2025 * the ip header). 2026 */ 2027 struct mbuf *m = args->m; 2028 struct ip *ip = mtod(m, struct ip *); 2029 2030 /* 2031 * For rules which contain uid/gid or jail constraints, cache 2032 * a copy of the users credentials after the pcb lookup has been 2033 * executed. This will speed up the processing of rules with 2034 * these types of constraints, as well as decrease contention 2035 * on pcb related locks. 2036 */ 2037 struct ip_fw_ugid fw_ugid_cache; 2038 int ugid_lookup = 0; 2039 2040 /* 2041 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2042 * associated with a packet input on a divert socket. This 2043 * will allow to distinguish traffic and its direction when 2044 * it originates from a divert socket. 2045 */ 2046 u_int divinput_flags = 0; 2047 2048 /* 2049 * oif | args->oif If NULL, ipfw_chk has been called on the 2050 * inbound path (ether_input, ip_input). 2051 * If non-NULL, ipfw_chk has been called on the outbound path 2052 * (ether_output, ip_output). 2053 */ 2054 struct ifnet *oif = args->oif; 2055 2056 struct ip_fw *f = NULL; /* matching rule */ 2057 int retval = 0; 2058 2059 /* 2060 * hlen The length of the IP header. 2061 */ 2062 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2063 2064 /* 2065 * offset The offset of a fragment. offset != 0 means that 2066 * we have a fragment at this offset of an IPv4 packet. 2067 * offset == 0 means that (if this is an IPv4 packet) 2068 * this is the first or only fragment. 2069 * For IPv6 offset == 0 means there is no Fragment Header. 2070 * If offset != 0 for IPv6 always use correct mask to 2071 * get the correct offset because we add IP6F_MORE_FRAG 2072 * to be able to dectect the first fragment which would 2073 * otherwise have offset = 0. 2074 */ 2075 u_short offset = 0; 2076 2077 /* 2078 * Local copies of addresses. They are only valid if we have 2079 * an IP packet. 2080 * 2081 * proto The protocol. Set to 0 for non-ip packets, 2082 * or to the protocol read from the packet otherwise. 2083 * proto != 0 means that we have an IPv4 packet. 2084 * 2085 * src_port, dst_port port numbers, in HOST format. Only 2086 * valid for TCP and UDP packets. 2087 * 2088 * src_ip, dst_ip ip addresses, in NETWORK format. 2089 * Only valid for IPv4 packets. 2090 */ 2091 u_int8_t proto; 2092 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2093 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2094 u_int16_t ip_len=0; 2095 int pktlen; 2096 2097 /* 2098 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2099 * MATCH_NONE when checked and not matched (q = NULL), 2100 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2101 */ 2102 int dyn_dir = MATCH_UNKNOWN; 2103 ipfw_dyn_rule *q = NULL; 2104 struct ip_fw_chain *chain = &layer3_chain; 2105 struct m_tag *mtag; 2106 2107 /* 2108 * We store in ulp a pointer to the upper layer protocol header. 2109 * In the ipv4 case this is easy to determine from the header, 2110 * but for ipv6 we might have some additional headers in the middle. 2111 * ulp is NULL if not found. 2112 */ 2113 void *ulp = NULL; /* upper layer protocol pointer. */ 2114 /* XXX ipv6 variables */ 2115 int is_ipv6 = 0; 2116 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2117 /* end of ipv6 variables */ 2118 int is_ipv4 = 0; 2119 2120 if (m->m_flags & M_SKIP_FIREWALL) 2121 return (IP_FW_PASS); /* accept */ 2122 2123 pktlen = m->m_pkthdr.len; 2124 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2125 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2126 2127/* 2128 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2129 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2130 * pointer might become stale after other pullups (but we never use it 2131 * this way). 2132 */ 2133#define PULLUP_TO(len, p, T) \ 2134do { \ 2135 int x = (len) + sizeof(T); \ 2136 if ((m)->m_len < x) { \ 2137 args->m = m = m_pullup(m, x); \ 2138 if (m == NULL) \ 2139 goto pullup_failed; \ 2140 } \ 2141 p = (mtod(m, char *) + (len)); \ 2142} while (0) 2143 2144 /* Identify IP packets and fill up variables. */ 2145 if (pktlen >= sizeof(struct ip6_hdr) && 2146 (args->eh == NULL || ntohs(args->eh->ether_type)==ETHERTYPE_IPV6) && 2147 mtod(m, struct ip *)->ip_v == 6) { 2148 is_ipv6 = 1; 2149 args->f_id.addr_type = 6; 2150 hlen = sizeof(struct ip6_hdr); 2151 proto = mtod(m, struct ip6_hdr *)->ip6_nxt; 2152 2153 /* Search extension headers to find upper layer protocols */ 2154 while (ulp == NULL) { 2155 switch (proto) { 2156 case IPPROTO_ICMPV6: 2157 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2158 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2159 break; 2160 2161 case IPPROTO_TCP: 2162 PULLUP_TO(hlen, ulp, struct tcphdr); 2163 dst_port = TCP(ulp)->th_dport; 2164 src_port = TCP(ulp)->th_sport; 2165 args->f_id.flags = TCP(ulp)->th_flags; 2166 break; 2167 2168 case IPPROTO_UDP: 2169 PULLUP_TO(hlen, ulp, struct udphdr); 2170 dst_port = UDP(ulp)->uh_dport; 2171 src_port = UDP(ulp)->uh_sport; 2172 break; 2173 2174 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2175 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2176 ext_hd |= EXT_HOPOPTS; 2177 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2178 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2179 ulp = NULL; 2180 break; 2181 2182 case IPPROTO_ROUTING: /* RFC 2460 */ 2183 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2184 if (((struct ip6_rthdr *)ulp)->ip6r_type != 0) { 2185 printf("IPFW2: IPV6 - Unknown Routing " 2186 "Header type(%d)\n", 2187 ((struct ip6_rthdr *)ulp)->ip6r_type); 2188 if (fw_deny_unknown_exthdrs) 2189 return (IP_FW_DENY); 2190 break; 2191 } 2192 ext_hd |= EXT_ROUTING; 2193 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2194 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2195 ulp = NULL; 2196 break; 2197 2198 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2199 PULLUP_TO(hlen, ulp, struct ip6_frag); 2200 ext_hd |= EXT_FRAGMENT; 2201 hlen += sizeof (struct ip6_frag); 2202 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2203 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2204 IP6F_OFF_MASK; 2205 /* Add IP6F_MORE_FRAG for offset of first 2206 * fragment to be != 0. */ 2207 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2208 IP6F_MORE_FRAG; 2209 if (offset == 0) { 2210 printf("IPFW2: IPV6 - Invalid Fragment " 2211 "Header\n"); 2212 if (fw_deny_unknown_exthdrs) 2213 return (IP_FW_DENY); 2214 break; 2215 } 2216 args->f_id.frag_id6 = 2217 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2218 ulp = NULL; 2219 break; 2220 2221 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2222 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2223 ext_hd |= EXT_DSTOPTS; 2224 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2225 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2226 ulp = NULL; 2227 break; 2228 2229 case IPPROTO_AH: /* RFC 2402 */ 2230 PULLUP_TO(hlen, ulp, struct ip6_ext); 2231 ext_hd |= EXT_AH; 2232 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2233 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2234 ulp = NULL; 2235 break; 2236 2237 case IPPROTO_ESP: /* RFC 2406 */ 2238 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2239 /* Anything past Seq# is variable length and 2240 * data past this ext. header is encrypted. */ 2241 ext_hd |= EXT_ESP; 2242 break; 2243 2244 case IPPROTO_NONE: /* RFC 2460 */ 2245 PULLUP_TO(hlen, ulp, struct ip6_ext); 2246 /* Packet ends here. if ip6e_len!=0 octets 2247 * must be ignored. */ 2248 break; 2249 2250 case IPPROTO_OSPFIGP: 2251 /* XXX OSPF header check? */ 2252 PULLUP_TO(hlen, ulp, struct ip6_ext); 2253 break; 2254 2255 default: 2256 printf("IPFW2: IPV6 - Unknown Extension " 2257 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2258 if (fw_deny_unknown_exthdrs) 2259 return (IP_FW_DENY); 2260 break; 2261 } /*switch */ 2262 } 2263 args->f_id.src_ip6 = mtod(m,struct ip6_hdr *)->ip6_src; 2264 args->f_id.dst_ip6 = mtod(m,struct ip6_hdr *)->ip6_dst; 2265 args->f_id.src_ip = 0; 2266 args->f_id.dst_ip = 0; 2267 args->f_id.flow_id6 = ntohl(mtod(m, struct ip6_hdr *)->ip6_flow); 2268 } else if (pktlen >= sizeof(struct ip) && 2269 (args->eh == NULL || ntohs(args->eh->ether_type) == ETHERTYPE_IP) && 2270 mtod(m, struct ip *)->ip_v == 4) { 2271 is_ipv4 = 1; 2272 ip = mtod(m, struct ip *); 2273 hlen = ip->ip_hl << 2; 2274 args->f_id.addr_type = 4; 2275 2276 /* 2277 * Collect parameters into local variables for faster matching. 2278 */ 2279 proto = ip->ip_p; 2280 src_ip = ip->ip_src; 2281 dst_ip = ip->ip_dst; 2282 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2283 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2284 ip_len = ntohs(ip->ip_len); 2285 } else { 2286 offset = ip->ip_off & IP_OFFMASK; 2287 ip_len = ip->ip_len; 2288 } 2289 pktlen = ip_len < pktlen ? ip_len : pktlen; 2290 2291 if (offset == 0) { 2292 switch (proto) { 2293 case IPPROTO_TCP: 2294 PULLUP_TO(hlen, ulp, struct tcphdr); 2295 dst_port = TCP(ulp)->th_dport; 2296 src_port = TCP(ulp)->th_sport; 2297 args->f_id.flags = TCP(ulp)->th_flags; 2298 break; 2299 2300 case IPPROTO_UDP: 2301 PULLUP_TO(hlen, ulp, struct udphdr); 2302 dst_port = UDP(ulp)->uh_dport; 2303 src_port = UDP(ulp)->uh_sport; 2304 break; 2305 2306 case IPPROTO_ICMP: 2307 PULLUP_TO(hlen, ulp, struct icmphdr); 2308 args->f_id.flags = ICMP(ulp)->icmp_type; 2309 break; 2310 2311 default: 2312 break; 2313 } 2314 } 2315 2316 args->f_id.src_ip = ntohl(src_ip.s_addr); 2317 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2318 } 2319#undef PULLUP_TO 2320 if (proto) { /* we may have port numbers, store them */ 2321 args->f_id.proto = proto; 2322 args->f_id.src_port = src_port = ntohs(src_port); 2323 args->f_id.dst_port = dst_port = ntohs(dst_port); 2324 } 2325 2326 IPFW_RLOCK(chain); 2327 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2328 if (args->rule) { 2329 /* 2330 * Packet has already been tagged. Look for the next rule 2331 * to restart processing. 2332 * 2333 * If fw_one_pass != 0 then just accept it. 2334 * XXX should not happen here, but optimized out in 2335 * the caller. 2336 */ 2337 if (fw_one_pass) { 2338 IPFW_RUNLOCK(chain); 2339 return (IP_FW_PASS); 2340 } 2341 2342 f = args->rule->next_rule; 2343 if (f == NULL) 2344 f = lookup_next_rule(args->rule); 2345 } else { 2346 /* 2347 * Find the starting rule. It can be either the first 2348 * one, or the one after divert_rule if asked so. 2349 */ 2350 int skipto = mtag ? divert_cookie(mtag) : 0; 2351 2352 f = chain->rules; 2353 if (args->eh == NULL && skipto != 0) { 2354 if (skipto >= IPFW_DEFAULT_RULE) { 2355 IPFW_RUNLOCK(chain); 2356 return (IP_FW_DENY); /* invalid */ 2357 } 2358 while (f && f->rulenum <= skipto) 2359 f = f->next; 2360 if (f == NULL) { /* drop packet */ 2361 IPFW_RUNLOCK(chain); 2362 return (IP_FW_DENY); 2363 } 2364 } 2365 } 2366 /* reset divert rule to avoid confusion later */ 2367 if (mtag) { 2368 divinput_flags = divert_info(mtag) & 2369 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2370 m_tag_delete(m, mtag); 2371 } 2372 2373 /* 2374 * Now scan the rules, and parse microinstructions for each rule. 2375 */ 2376 for (; f; f = f->next) { 2377 ipfw_insn *cmd; 2378 uint32_t tablearg = 0; 2379 int l, cmdlen, skip_or; /* skip rest of OR block */ 2380 2381again: 2382 if (set_disable & (1 << f->set) ) 2383 continue; 2384 2385 skip_or = 0; 2386 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2387 l -= cmdlen, cmd += cmdlen) { 2388 int match; 2389 2390 /* 2391 * check_body is a jump target used when we find a 2392 * CHECK_STATE, and need to jump to the body of 2393 * the target rule. 2394 */ 2395 2396check_body: 2397 cmdlen = F_LEN(cmd); 2398 /* 2399 * An OR block (insn_1 || .. || insn_n) has the 2400 * F_OR bit set in all but the last instruction. 2401 * The first match will set "skip_or", and cause 2402 * the following instructions to be skipped until 2403 * past the one with the F_OR bit clear. 2404 */ 2405 if (skip_or) { /* skip this instruction */ 2406 if ((cmd->len & F_OR) == 0) 2407 skip_or = 0; /* next one is good */ 2408 continue; 2409 } 2410 match = 0; /* set to 1 if we succeed */ 2411 2412 switch (cmd->opcode) { 2413 /* 2414 * The first set of opcodes compares the packet's 2415 * fields with some pattern, setting 'match' if a 2416 * match is found. At the end of the loop there is 2417 * logic to deal with F_NOT and F_OR flags associated 2418 * with the opcode. 2419 */ 2420 case O_NOP: 2421 match = 1; 2422 break; 2423 2424 case O_FORWARD_MAC: 2425 printf("ipfw: opcode %d unimplemented\n", 2426 cmd->opcode); 2427 break; 2428 2429 case O_GID: 2430 case O_UID: 2431 case O_JAIL: 2432 /* 2433 * We only check offset == 0 && proto != 0, 2434 * as this ensures that we have a 2435 * packet with the ports info. 2436 */ 2437 if (offset!=0) 2438 break; 2439 if (is_ipv6) /* XXX to be fixed later */ 2440 break; 2441 if (proto == IPPROTO_TCP || 2442 proto == IPPROTO_UDP) 2443 match = check_uidgid( 2444 (ipfw_insn_u32 *)cmd, 2445 proto, oif, 2446 dst_ip, dst_port, 2447 src_ip, src_port, &fw_ugid_cache, 2448 &ugid_lookup, args->inp); 2449 break; 2450 2451 case O_RECV: 2452 match = iface_match(m->m_pkthdr.rcvif, 2453 (ipfw_insn_if *)cmd); 2454 break; 2455 2456 case O_XMIT: 2457 match = iface_match(oif, (ipfw_insn_if *)cmd); 2458 break; 2459 2460 case O_VIA: 2461 match = iface_match(oif ? oif : 2462 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2463 break; 2464 2465 case O_MACADDR2: 2466 if (args->eh != NULL) { /* have MAC header */ 2467 u_int32_t *want = (u_int32_t *) 2468 ((ipfw_insn_mac *)cmd)->addr; 2469 u_int32_t *mask = (u_int32_t *) 2470 ((ipfw_insn_mac *)cmd)->mask; 2471 u_int32_t *hdr = (u_int32_t *)args->eh; 2472 2473 match = 2474 ( want[0] == (hdr[0] & mask[0]) && 2475 want[1] == (hdr[1] & mask[1]) && 2476 want[2] == (hdr[2] & mask[2]) ); 2477 } 2478 break; 2479 2480 case O_MAC_TYPE: 2481 if (args->eh != NULL) { 2482 u_int16_t t = 2483 ntohs(args->eh->ether_type); 2484 u_int16_t *p = 2485 ((ipfw_insn_u16 *)cmd)->ports; 2486 int i; 2487 2488 for (i = cmdlen - 1; !match && i>0; 2489 i--, p += 2) 2490 match = (t>=p[0] && t<=p[1]); 2491 } 2492 break; 2493 2494 case O_FRAG: 2495 match = (offset != 0); 2496 break; 2497 2498 case O_IN: /* "out" is "not in" */ 2499 match = (oif == NULL); 2500 break; 2501 2502 case O_LAYER2: 2503 match = (args->eh != NULL); 2504 break; 2505 2506 case O_DIVERTED: 2507 match = (cmd->arg1 & 1 && divinput_flags & 2508 IP_FW_DIVERT_LOOPBACK_FLAG) || 2509 (cmd->arg1 & 2 && divinput_flags & 2510 IP_FW_DIVERT_OUTPUT_FLAG); 2511 break; 2512 2513 case O_PROTO: 2514 /* 2515 * We do not allow an arg of 0 so the 2516 * check of "proto" only suffices. 2517 */ 2518 match = (proto == cmd->arg1); 2519 break; 2520 2521 case O_IP_SRC: 2522 match = is_ipv4 && 2523 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2524 src_ip.s_addr); 2525 break; 2526 2527 case O_IP_SRC_LOOKUP: 2528 case O_IP_DST_LOOKUP: 2529 if (is_ipv4) { 2530 uint32_t a = 2531 (cmd->opcode == O_IP_DST_LOOKUP) ? 2532 dst_ip.s_addr : src_ip.s_addr; 2533 uint32_t v; 2534 2535 match = lookup_table(chain, cmd->arg1, a, 2536 &v); 2537 if (!match) 2538 break; 2539 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2540 match = 2541 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2542 else 2543 tablearg = v; 2544 } 2545 break; 2546 2547 case O_IP_SRC_MASK: 2548 case O_IP_DST_MASK: 2549 if (is_ipv4) { 2550 uint32_t a = 2551 (cmd->opcode == O_IP_DST_MASK) ? 2552 dst_ip.s_addr : src_ip.s_addr; 2553 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2554 int i = cmdlen-1; 2555 2556 for (; !match && i>0; i-= 2, p+= 2) 2557 match = (p[0] == (a & p[1])); 2558 } 2559 break; 2560 2561 case O_IP_SRC_ME: 2562 if (is_ipv4) { 2563 struct ifnet *tif; 2564 2565 INADDR_TO_IFP(src_ip, tif); 2566 match = (tif != NULL); 2567 } 2568 break; 2569 2570 case O_IP_DST_SET: 2571 case O_IP_SRC_SET: 2572 if (is_ipv4) { 2573 u_int32_t *d = (u_int32_t *)(cmd+1); 2574 u_int32_t addr = 2575 cmd->opcode == O_IP_DST_SET ? 2576 args->f_id.dst_ip : 2577 args->f_id.src_ip; 2578 2579 if (addr < d[0]) 2580 break; 2581 addr -= d[0]; /* subtract base */ 2582 match = (addr < cmd->arg1) && 2583 ( d[ 1 + (addr>>5)] & 2584 (1<<(addr & 0x1f)) ); 2585 } 2586 break; 2587 2588 case O_IP_DST: 2589 match = is_ipv4 && 2590 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2591 dst_ip.s_addr); 2592 break; 2593 2594 case O_IP_DST_ME: 2595 if (is_ipv4) { 2596 struct ifnet *tif; 2597 2598 INADDR_TO_IFP(dst_ip, tif); 2599 match = (tif != NULL); 2600 } 2601 break; 2602 2603 case O_IP_SRCPORT: 2604 case O_IP_DSTPORT: 2605 /* 2606 * offset == 0 && proto != 0 is enough 2607 * to guarantee that we have a 2608 * packet with port info. 2609 */ 2610 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2611 && offset == 0) { 2612 u_int16_t x = 2613 (cmd->opcode == O_IP_SRCPORT) ? 2614 src_port : dst_port ; 2615 u_int16_t *p = 2616 ((ipfw_insn_u16 *)cmd)->ports; 2617 int i; 2618 2619 for (i = cmdlen - 1; !match && i>0; 2620 i--, p += 2) 2621 match = (x>=p[0] && x<=p[1]); 2622 } 2623 break; 2624 2625 case O_ICMPTYPE: 2626 match = (offset == 0 && proto==IPPROTO_ICMP && 2627 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2628 break; 2629 2630#ifdef INET6 2631 case O_ICMP6TYPE: 2632 match = is_ipv6 && offset == 0 && 2633 proto==IPPROTO_ICMPV6 && 2634 icmp6type_match( 2635 ICMP6(ulp)->icmp6_type, 2636 (ipfw_insn_u32 *)cmd); 2637 break; 2638#endif /* INET6 */ 2639 2640 case O_IPOPT: 2641 match = (is_ipv4 && 2642 ipopts_match(mtod(m, struct ip *), cmd) ); 2643 break; 2644 2645 case O_IPVER: 2646 match = (is_ipv4 && 2647 cmd->arg1 == mtod(m, struct ip *)->ip_v); 2648 break; 2649 2650 case O_IPID: 2651 case O_IPLEN: 2652 case O_IPTTL: 2653 if (is_ipv4) { /* only for IP packets */ 2654 uint16_t x; 2655 uint16_t *p; 2656 int i; 2657 2658 if (cmd->opcode == O_IPLEN) 2659 x = ip_len; 2660 else if (cmd->opcode == O_IPTTL) 2661 x = mtod(m, struct ip *)->ip_ttl; 2662 else /* must be IPID */ 2663 x = ntohs(mtod(m, struct ip *)->ip_id); 2664 if (cmdlen == 1) { 2665 match = (cmd->arg1 == x); 2666 break; 2667 } 2668 /* otherwise we have ranges */ 2669 p = ((ipfw_insn_u16 *)cmd)->ports; 2670 i = cmdlen - 1; 2671 for (; !match && i>0; i--, p += 2) 2672 match = (x >= p[0] && x <= p[1]); 2673 } 2674 break; 2675 2676 case O_IPPRECEDENCE: 2677 match = (is_ipv4 && 2678 (cmd->arg1 == (mtod(m, struct ip *)->ip_tos & 0xe0)) ); 2679 break; 2680 2681 case O_IPTOS: 2682 match = (is_ipv4 && 2683 flags_match(cmd, mtod(m, struct ip *)->ip_tos)); 2684 break; 2685 2686 case O_TCPDATALEN: 2687 if (proto == IPPROTO_TCP && offset == 0) { 2688 struct tcphdr *tcp; 2689 uint16_t x; 2690 uint16_t *p; 2691 int i; 2692 2693 tcp = TCP(ulp); 2694 x = ip_len - 2695 ((ip->ip_hl + tcp->th_off) << 2); 2696 if (cmdlen == 1) { 2697 match = (cmd->arg1 == x); 2698 break; 2699 } 2700 /* otherwise we have ranges */ 2701 p = ((ipfw_insn_u16 *)cmd)->ports; 2702 i = cmdlen - 1; 2703 for (; !match && i>0; i--, p += 2) 2704 match = (x >= p[0] && x <= p[1]); 2705 } 2706 break; 2707 2708 case O_TCPFLAGS: 2709 match = (proto == IPPROTO_TCP && offset == 0 && 2710 flags_match(cmd, TCP(ulp)->th_flags)); 2711 break; 2712 2713 case O_TCPOPTS: 2714 match = (proto == IPPROTO_TCP && offset == 0 && 2715 tcpopts_match(TCP(ulp), cmd)); 2716 break; 2717 2718 case O_TCPSEQ: 2719 match = (proto == IPPROTO_TCP && offset == 0 && 2720 ((ipfw_insn_u32 *)cmd)->d[0] == 2721 TCP(ulp)->th_seq); 2722 break; 2723 2724 case O_TCPACK: 2725 match = (proto == IPPROTO_TCP && offset == 0 && 2726 ((ipfw_insn_u32 *)cmd)->d[0] == 2727 TCP(ulp)->th_ack); 2728 break; 2729 2730 case O_TCPWIN: 2731 match = (proto == IPPROTO_TCP && offset == 0 && 2732 cmd->arg1 == TCP(ulp)->th_win); 2733 break; 2734 2735 case O_ESTAB: 2736 /* reject packets which have SYN only */ 2737 /* XXX should i also check for TH_ACK ? */ 2738 match = (proto == IPPROTO_TCP && offset == 0 && 2739 (TCP(ulp)->th_flags & 2740 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2741 break; 2742 2743 case O_ALTQ: { 2744 struct altq_tag *at; 2745 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2746 2747 match = 1; 2748 mtag = m_tag_find(m, PACKET_TAG_PF_QID, NULL); 2749 if (mtag != NULL) 2750 break; 2751 mtag = m_tag_get(PACKET_TAG_PF_QID, 2752 sizeof(struct altq_tag), 2753 M_NOWAIT); 2754 if (mtag == NULL) { 2755 /* 2756 * Let the packet fall back to the 2757 * default ALTQ. 2758 */ 2759 break; 2760 } 2761 at = (struct altq_tag *)(mtag+1); 2762 at->qid = altq->qid; 2763 if (is_ipv4) 2764 at->af = AF_INET; 2765 else 2766 at->af = AF_LINK; 2767 at->hdr = ip; 2768 m_tag_prepend(m, mtag); 2769 break; 2770 } 2771 2772 case O_LOG: 2773 if (fw_verbose) 2774 ipfw_log(f, hlen, args, m, oif, offset); 2775 match = 1; 2776 break; 2777 2778 case O_PROB: 2779 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2780 break; 2781 2782 case O_VERREVPATH: 2783 /* Outgoing packets automatically pass/match */ 2784 match = ((oif != NULL) || 2785 (m->m_pkthdr.rcvif == NULL) || 2786 ( 2787#ifdef INET6 2788 is_ipv6 ? 2789 verify_path6(&(args->f_id.src_ip6), 2790 m->m_pkthdr.rcvif) : 2791#endif 2792 verify_path(src_ip, m->m_pkthdr.rcvif))); 2793 break; 2794 2795 case O_VERSRCREACH: 2796 /* Outgoing packets automatically pass/match */ 2797 match = (hlen > 0 && ((oif != NULL) || 2798#ifdef INET6 2799 is_ipv6 ? 2800 verify_path6(&(args->f_id.src_ip6), 2801 NULL) : 2802#endif 2803 verify_path(src_ip, NULL))); 2804 break; 2805 2806 case O_ANTISPOOF: 2807 /* Outgoing packets automatically pass/match */ 2808 if (oif == NULL && hlen > 0 && 2809 ( (is_ipv4 && in_localaddr(src_ip)) 2810#ifdef INET6 2811 || (is_ipv6 && 2812 in6_localaddr(&(args->f_id.src_ip6))) 2813#endif 2814 )) 2815 match = 2816#ifdef INET6 2817 is_ipv6 ? verify_path6( 2818 &(args->f_id.src_ip6), 2819 m->m_pkthdr.rcvif) : 2820#endif 2821 verify_path(src_ip, 2822 m->m_pkthdr.rcvif); 2823 else 2824 match = 1; 2825 break; 2826 2827 case O_IPSEC: 2828#ifdef FAST_IPSEC 2829 match = (m_tag_find(m, 2830 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2831#endif 2832#ifdef IPSEC 2833 match = (ipsec_getnhist(m) != 0); 2834#endif 2835 /* otherwise no match */ 2836 break; 2837 2838#ifdef INET6 2839 case O_IP6_SRC: 2840 match = is_ipv6 && 2841 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2842 &((ipfw_insn_ip6 *)cmd)->addr6); 2843 break; 2844 2845 case O_IP6_DST: 2846 match = is_ipv6 && 2847 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2848 &((ipfw_insn_ip6 *)cmd)->addr6); 2849 break; 2850 case O_IP6_SRC_MASK: 2851 if (is_ipv6) { 2852 ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd; 2853 struct in6_addr p = args->f_id.src_ip6; 2854 2855 APPLY_MASK(&p, &te->mask6); 2856 match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p); 2857 } 2858 break; 2859 2860 case O_IP6_DST_MASK: 2861 if (is_ipv6) { 2862 ipfw_insn_ip6 *te = (ipfw_insn_ip6 *)cmd; 2863 struct in6_addr p = args->f_id.dst_ip6; 2864 2865 APPLY_MASK(&p, &te->mask6); 2866 match = IN6_ARE_ADDR_EQUAL(&te->addr6, &p); 2867 } 2868 break; 2869 2870 case O_IP6_SRC_ME: 2871 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 2872 break; 2873 2874 case O_IP6_DST_ME: 2875 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 2876 break; 2877 2878 case O_FLOW6ID: 2879 match = is_ipv6 && 2880 flow6id_match(args->f_id.flow_id6, 2881 (ipfw_insn_u32 *) cmd); 2882 break; 2883 2884 case O_EXT_HDR: 2885 match = is_ipv6 && 2886 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2887 break; 2888 2889 case O_IP6: 2890 match = is_ipv6; 2891 break; 2892#endif 2893 2894 case O_IP4: 2895 match = is_ipv4; 2896 break; 2897 2898 case O_TAG: 2899 /* Packet is already tagged with this tag? */ 2900 mtag = m_tag_locate(m, MTAG_IPFW, 2901 ((ipfw_insn *) cmd)->arg1, NULL); 2902 /* We have `untag' action when F_NOT flag is 2903 * present. And we must remove this mtag from 2904 * mbuf and reset `match' to zero (`match' will 2905 * be inversed later). 2906 * Otherwise we should allocate new mtag and 2907 * push it into mbuf. 2908 */ 2909 if (cmd->len & F_NOT) { /* `untag' action */ 2910 if (mtag != NULL) 2911 m_tag_delete(m, mtag); 2912 } else if (mtag == NULL) { 2913 mtag = m_tag_alloc(MTAG_IPFW, 2914 ((ipfw_insn *) cmd)->arg1, 0, M_NOWAIT); 2915 if (mtag != NULL) 2916 m_tag_prepend(m, mtag); 2917 } 2918 match = (cmd->len & F_NOT) ? 0: 1; 2919 break; 2920 2921 case O_TAGGED: 2922 if (cmdlen == 1) { 2923 match = (m_tag_locate(m, MTAG_IPFW, 2924 ((ipfw_insn *) cmd)->arg1, NULL) != NULL); 2925 break; 2926 } 2927 2928 /* we have ranges */ 2929 for (mtag = m_tag_first(m); 2930 mtag != NULL && !match; 2931 mtag = m_tag_next(m, mtag)) { 2932 uint16_t *p; 2933 int i; 2934 2935 if (mtag->m_tag_cookie != MTAG_IPFW) 2936 continue; 2937 2938 p = ((ipfw_insn_u16 *)cmd)->ports; 2939 i = cmdlen - 1; 2940 for(; !match && i > 0; i--, p += 2) 2941 match = 2942 mtag->m_tag_id >= p[0] && 2943 mtag->m_tag_id <= p[1]; 2944 } 2945 break; 2946 2947 /* 2948 * The second set of opcodes represents 'actions', 2949 * i.e. the terminal part of a rule once the packet 2950 * matches all previous patterns. 2951 * Typically there is only one action for each rule, 2952 * and the opcode is stored at the end of the rule 2953 * (but there are exceptions -- see below). 2954 * 2955 * In general, here we set retval and terminate the 2956 * outer loop (would be a 'break 3' in some language, 2957 * but we need to do a 'goto done'). 2958 * 2959 * Exceptions: 2960 * O_COUNT and O_SKIPTO actions: 2961 * instead of terminating, we jump to the next rule 2962 * ('goto next_rule', equivalent to a 'break 2'), 2963 * or to the SKIPTO target ('goto again' after 2964 * having set f, cmd and l), respectively. 2965 * 2966 * O_TAG, O_LOG and O_ALTQ action parameters: 2967 * perform some action and set match = 1; 2968 * 2969 * O_LIMIT and O_KEEP_STATE: these opcodes are 2970 * not real 'actions', and are stored right 2971 * before the 'action' part of the rule. 2972 * These opcodes try to install an entry in the 2973 * state tables; if successful, we continue with 2974 * the next opcode (match=1; break;), otherwise 2975 * the packet * must be dropped 2976 * ('goto done' after setting retval); 2977 * 2978 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2979 * cause a lookup of the state table, and a jump 2980 * to the 'action' part of the parent rule 2981 * ('goto check_body') if an entry is found, or 2982 * (CHECK_STATE only) a jump to the next rule if 2983 * the entry is not found ('goto next_rule'). 2984 * The result of the lookup is cached to make 2985 * further instances of these opcodes are 2986 * effectively NOPs. 2987 */ 2988 case O_LIMIT: 2989 case O_KEEP_STATE: 2990 if (install_state(f, 2991 (ipfw_insn_limit *)cmd, args)) { 2992 retval = IP_FW_DENY; 2993 goto done; /* error/limit violation */ 2994 } 2995 match = 1; 2996 break; 2997 2998 case O_PROBE_STATE: 2999 case O_CHECK_STATE: 3000 /* 3001 * dynamic rules are checked at the first 3002 * keep-state or check-state occurrence, 3003 * with the result being stored in dyn_dir. 3004 * The compiler introduces a PROBE_STATE 3005 * instruction for us when we have a 3006 * KEEP_STATE (because PROBE_STATE needs 3007 * to be run first). 3008 */ 3009 if (dyn_dir == MATCH_UNKNOWN && 3010 (q = lookup_dyn_rule(&args->f_id, 3011 &dyn_dir, proto == IPPROTO_TCP ? 3012 TCP(ulp) : NULL)) 3013 != NULL) { 3014 /* 3015 * Found dynamic entry, update stats 3016 * and jump to the 'action' part of 3017 * the parent rule. 3018 */ 3019 q->pcnt++; 3020 q->bcnt += pktlen; 3021 f = q->rule; 3022 cmd = ACTION_PTR(f); 3023 l = f->cmd_len - f->act_ofs; 3024 IPFW_DYN_UNLOCK(); 3025 goto check_body; 3026 } 3027 /* 3028 * Dynamic entry not found. If CHECK_STATE, 3029 * skip to next rule, if PROBE_STATE just 3030 * ignore and continue with next opcode. 3031 */ 3032 if (cmd->opcode == O_CHECK_STATE) 3033 goto next_rule; 3034 match = 1; 3035 break; 3036 3037 case O_ACCEPT: 3038 retval = 0; /* accept */ 3039 goto done; 3040 3041 case O_PIPE: 3042 case O_QUEUE: 3043 args->rule = f; /* report matching rule */ 3044 if (cmd->arg1 == IP_FW_TABLEARG) 3045 args->cookie = tablearg; 3046 else 3047 args->cookie = cmd->arg1; 3048 retval = IP_FW_DUMMYNET; 3049 goto done; 3050 3051 case O_DIVERT: 3052 case O_TEE: { 3053 struct divert_tag *dt; 3054 3055 if (args->eh) /* not on layer 2 */ 3056 break; 3057 mtag = m_tag_get(PACKET_TAG_DIVERT, 3058 sizeof(struct divert_tag), 3059 M_NOWAIT); 3060 if (mtag == NULL) { 3061 /* XXX statistic */ 3062 /* drop packet */ 3063 IPFW_RUNLOCK(chain); 3064 return (IP_FW_DENY); 3065 } 3066 dt = (struct divert_tag *)(mtag+1); 3067 dt->cookie = f->rulenum; 3068 if (cmd->arg1 == IP_FW_TABLEARG) 3069 dt->info = tablearg; 3070 else 3071 dt->info = cmd->arg1; 3072 m_tag_prepend(m, mtag); 3073 retval = (cmd->opcode == O_DIVERT) ? 3074 IP_FW_DIVERT : IP_FW_TEE; 3075 goto done; 3076 } 3077 3078 case O_COUNT: 3079 case O_SKIPTO: 3080 f->pcnt++; /* update stats */ 3081 f->bcnt += pktlen; 3082 f->timestamp = time_uptime; 3083 if (cmd->opcode == O_COUNT) 3084 goto next_rule; 3085 /* handle skipto */ 3086 if (f->next_rule == NULL) 3087 lookup_next_rule(f); 3088 f = f->next_rule; 3089 goto again; 3090 3091 case O_REJECT: 3092 /* 3093 * Drop the packet and send a reject notice 3094 * if the packet is not ICMP (or is an ICMP 3095 * query), and it is not multicast/broadcast. 3096 */ 3097 if (hlen > 0 && is_ipv4 && offset == 0 && 3098 (proto != IPPROTO_ICMP || 3099 is_icmp_query(ICMP(ulp))) && 3100 !(m->m_flags & (M_BCAST|M_MCAST)) && 3101 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3102 send_reject(args, cmd->arg1, 3103 offset,ip_len); 3104 m = args->m; 3105 } 3106 /* FALLTHROUGH */ 3107#ifdef INET6 3108 case O_UNREACH6: 3109 if (hlen > 0 && is_ipv6 && 3110 (proto != IPPROTO_ICMPV6 || 3111 (is_icmp6_query(args->f_id.flags) == 1)) && 3112 !(m->m_flags & (M_BCAST|M_MCAST)) && 3113 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3114 send_reject6(args, cmd->arg1, 3115 offset, hlen); 3116 m = args->m; 3117 } 3118 /* FALLTHROUGH */ 3119#endif 3120 case O_DENY: 3121 retval = IP_FW_DENY; 3122 goto done; 3123 3124 case O_FORWARD_IP: 3125 if (args->eh) /* not valid on layer2 pkts */ 3126 break; 3127 if (!q || dyn_dir == MATCH_FORWARD) 3128 args->next_hop = 3129 &((ipfw_insn_sa *)cmd)->sa; 3130 retval = IP_FW_PASS; 3131 goto done; 3132 3133 case O_NETGRAPH: 3134 case O_NGTEE: 3135 args->rule = f; /* report matching rule */ 3136 if (cmd->arg1 == IP_FW_TABLEARG) 3137 args->cookie = tablearg; 3138 else 3139 args->cookie = cmd->arg1; 3140 retval = (cmd->opcode == O_NETGRAPH) ? 3141 IP_FW_NETGRAPH : IP_FW_NGTEE; 3142 goto done; 3143 3144 default: 3145 panic("-- unknown opcode %d\n", cmd->opcode); 3146 } /* end of switch() on opcodes */ 3147 3148 if (cmd->len & F_NOT) 3149 match = !match; 3150 3151 if (match) { 3152 if (cmd->len & F_OR) 3153 skip_or = 1; 3154 } else { 3155 if (!(cmd->len & F_OR)) /* not an OR block, */ 3156 break; /* try next rule */ 3157 } 3158 3159 } /* end of inner for, scan opcodes */ 3160 3161next_rule:; /* try next rule */ 3162 3163 } /* end of outer for, scan rules */ 3164 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3165 IPFW_RUNLOCK(chain); 3166 return (IP_FW_DENY); 3167 3168done: 3169 /* Update statistics */ 3170 f->pcnt++; 3171 f->bcnt += pktlen; 3172 f->timestamp = time_uptime; 3173 IPFW_RUNLOCK(chain); 3174 return (retval); 3175 3176pullup_failed: 3177 if (fw_verbose) 3178 printf("ipfw: pullup failed\n"); 3179 return (IP_FW_DENY); 3180} 3181 3182/* 3183 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3184 * These will be reconstructed on the fly as packets are matched. 3185 */ 3186static void 3187flush_rule_ptrs(struct ip_fw_chain *chain) 3188{ 3189 struct ip_fw *rule; 3190 3191 IPFW_WLOCK_ASSERT(chain); 3192 3193 for (rule = chain->rules; rule; rule = rule->next) 3194 rule->next_rule = NULL; 3195} 3196 3197/* 3198 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3199 * possibly create a rule number and add the rule to the list. 3200 * Update the rule_number in the input struct so the caller knows it as well. 3201 */ 3202static int 3203add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3204{ 3205 struct ip_fw *rule, *f, *prev; 3206 int l = RULESIZE(input_rule); 3207 3208 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3209 return (EINVAL); 3210 3211 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3212 if (rule == NULL) 3213 return (ENOSPC); 3214 3215 bcopy(input_rule, rule, l); 3216 3217 rule->next = NULL; 3218 rule->next_rule = NULL; 3219 3220 rule->pcnt = 0; 3221 rule->bcnt = 0; 3222 rule->timestamp = 0; 3223 3224 IPFW_WLOCK(chain); 3225 3226 if (chain->rules == NULL) { /* default rule */ 3227 chain->rules = rule; 3228 goto done; 3229 } 3230 3231 /* 3232 * If rulenum is 0, find highest numbered rule before the 3233 * default rule, and add autoinc_step 3234 */ 3235 if (autoinc_step < 1) 3236 autoinc_step = 1; 3237 else if (autoinc_step > 1000) 3238 autoinc_step = 1000; 3239 if (rule->rulenum == 0) { 3240 /* 3241 * locate the highest numbered rule before default 3242 */ 3243 for (f = chain->rules; f; f = f->next) { 3244 if (f->rulenum == IPFW_DEFAULT_RULE) 3245 break; 3246 rule->rulenum = f->rulenum; 3247 } 3248 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 3249 rule->rulenum += autoinc_step; 3250 input_rule->rulenum = rule->rulenum; 3251 } 3252 3253 /* 3254 * Now insert the new rule in the right place in the sorted list. 3255 */ 3256 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3257 if (f->rulenum > rule->rulenum) { /* found the location */ 3258 if (prev) { 3259 rule->next = f; 3260 prev->next = rule; 3261 } else { /* head insert */ 3262 rule->next = chain->rules; 3263 chain->rules = rule; 3264 } 3265 break; 3266 } 3267 } 3268 flush_rule_ptrs(chain); 3269done: 3270 static_count++; 3271 static_len += l; 3272 IPFW_WUNLOCK(chain); 3273 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3274 rule->rulenum, static_count);) 3275 return (0); 3276} 3277 3278/** 3279 * Remove a static rule (including derived * dynamic rules) 3280 * and place it on the ``reap list'' for later reclamation. 3281 * The caller is in charge of clearing rule pointers to avoid 3282 * dangling pointers. 3283 * @return a pointer to the next entry. 3284 * Arguments are not checked, so they better be correct. 3285 */ 3286static struct ip_fw * 3287remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 3288{ 3289 struct ip_fw *n; 3290 int l = RULESIZE(rule); 3291 3292 IPFW_WLOCK_ASSERT(chain); 3293 3294 n = rule->next; 3295 IPFW_DYN_LOCK(); 3296 remove_dyn_rule(rule, NULL /* force removal */); 3297 IPFW_DYN_UNLOCK(); 3298 if (prev == NULL) 3299 chain->rules = n; 3300 else 3301 prev->next = n; 3302 static_count--; 3303 static_len -= l; 3304 3305 rule->next = chain->reap; 3306 chain->reap = rule; 3307 3308 return n; 3309} 3310 3311/** 3312 * Reclaim storage associated with a list of rules. This is 3313 * typically the list created using remove_rule. 3314 */ 3315static void 3316reap_rules(struct ip_fw *head) 3317{ 3318 struct ip_fw *rule; 3319 3320 while ((rule = head) != NULL) { 3321 head = head->next; 3322 if (DUMMYNET_LOADED) 3323 ip_dn_ruledel_ptr(rule); 3324 free(rule, M_IPFW); 3325 } 3326} 3327 3328/* 3329 * Remove all rules from a chain (except rules in set RESVD_SET 3330 * unless kill_default = 1). The caller is responsible for 3331 * reclaiming storage for the rules left in chain->reap. 3332 */ 3333static void 3334free_chain(struct ip_fw_chain *chain, int kill_default) 3335{ 3336 struct ip_fw *prev, *rule; 3337 3338 IPFW_WLOCK_ASSERT(chain); 3339 3340 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3341 for (prev = NULL, rule = chain->rules; rule ; ) 3342 if (kill_default || rule->set != RESVD_SET) 3343 rule = remove_rule(chain, rule, prev); 3344 else { 3345 prev = rule; 3346 rule = rule->next; 3347 } 3348} 3349 3350/** 3351 * Remove all rules with given number, and also do set manipulation. 3352 * Assumes chain != NULL && *chain != NULL. 3353 * 3354 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3355 * the next 8 bits are the new set, the top 8 bits are the command: 3356 * 3357 * 0 delete rules with given number 3358 * 1 delete rules with given set number 3359 * 2 move rules with given number to new set 3360 * 3 move rules with given set number to new set 3361 * 4 swap sets with given numbers 3362 */ 3363static int 3364del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3365{ 3366 struct ip_fw *prev = NULL, *rule; 3367 u_int16_t rulenum; /* rule or old_set */ 3368 u_int8_t cmd, new_set; 3369 3370 rulenum = arg & 0xffff; 3371 cmd = (arg >> 24) & 0xff; 3372 new_set = (arg >> 16) & 0xff; 3373 3374 if (cmd > 4) 3375 return EINVAL; 3376 if (new_set > RESVD_SET) 3377 return EINVAL; 3378 if (cmd == 0 || cmd == 2) { 3379 if (rulenum >= IPFW_DEFAULT_RULE) 3380 return EINVAL; 3381 } else { 3382 if (rulenum > RESVD_SET) /* old_set */ 3383 return EINVAL; 3384 } 3385 3386 IPFW_WLOCK(chain); 3387 rule = chain->rules; 3388 chain->reap = NULL; 3389 switch (cmd) { 3390 case 0: /* delete rules with given number */ 3391 /* 3392 * locate first rule to delete 3393 */ 3394 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3395 ; 3396 if (rule->rulenum != rulenum) { 3397 IPFW_WUNLOCK(chain); 3398 return EINVAL; 3399 } 3400 3401 /* 3402 * flush pointers outside the loop, then delete all matching 3403 * rules. prev remains the same throughout the cycle. 3404 */ 3405 flush_rule_ptrs(chain); 3406 while (rule->rulenum == rulenum) 3407 rule = remove_rule(chain, rule, prev); 3408 break; 3409 3410 case 1: /* delete all rules with given set number */ 3411 flush_rule_ptrs(chain); 3412 rule = chain->rules; 3413 while (rule->rulenum < IPFW_DEFAULT_RULE) 3414 if (rule->set == rulenum) 3415 rule = remove_rule(chain, rule, prev); 3416 else { 3417 prev = rule; 3418 rule = rule->next; 3419 } 3420 break; 3421 3422 case 2: /* move rules with given number to new set */ 3423 rule = chain->rules; 3424 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3425 if (rule->rulenum == rulenum) 3426 rule->set = new_set; 3427 break; 3428 3429 case 3: /* move rules with given set number to new set */ 3430 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3431 if (rule->set == rulenum) 3432 rule->set = new_set; 3433 break; 3434 3435 case 4: /* swap two sets */ 3436 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3437 if (rule->set == rulenum) 3438 rule->set = new_set; 3439 else if (rule->set == new_set) 3440 rule->set = rulenum; 3441 break; 3442 } 3443 /* 3444 * Look for rules to reclaim. We grab the list before 3445 * releasing the lock then reclaim them w/o the lock to 3446 * avoid a LOR with dummynet. 3447 */ 3448 rule = chain->reap; 3449 chain->reap = NULL; 3450 IPFW_WUNLOCK(chain); 3451 if (rule) 3452 reap_rules(rule); 3453 return 0; 3454} 3455 3456/* 3457 * Clear counters for a specific rule. 3458 * The enclosing "table" is assumed locked. 3459 */ 3460static void 3461clear_counters(struct ip_fw *rule, int log_only) 3462{ 3463 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3464 3465 if (log_only == 0) { 3466 rule->bcnt = rule->pcnt = 0; 3467 rule->timestamp = 0; 3468 } 3469 if (l->o.opcode == O_LOG) 3470 l->log_left = l->max_log; 3471} 3472 3473/** 3474 * Reset some or all counters on firewall rules. 3475 * @arg frwl is null to clear all entries, or contains a specific 3476 * rule number. 3477 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 3478 */ 3479static int 3480zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 3481{ 3482 struct ip_fw *rule; 3483 char *msg; 3484 3485 IPFW_WLOCK(chain); 3486 if (rulenum == 0) { 3487 norule_counter = 0; 3488 for (rule = chain->rules; rule; rule = rule->next) 3489 clear_counters(rule, log_only); 3490 msg = log_only ? "ipfw: All logging counts reset.\n" : 3491 "ipfw: Accounting cleared.\n"; 3492 } else { 3493 int cleared = 0; 3494 /* 3495 * We can have multiple rules with the same number, so we 3496 * need to clear them all. 3497 */ 3498 for (rule = chain->rules; rule; rule = rule->next) 3499 if (rule->rulenum == rulenum) { 3500 while (rule && rule->rulenum == rulenum) { 3501 clear_counters(rule, log_only); 3502 rule = rule->next; 3503 } 3504 cleared = 1; 3505 break; 3506 } 3507 if (!cleared) { /* we did not find any matching rules */ 3508 IPFW_WUNLOCK(chain); 3509 return (EINVAL); 3510 } 3511 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 3512 "ipfw: Entry %d cleared.\n"; 3513 } 3514 IPFW_WUNLOCK(chain); 3515 3516 if (fw_verbose) 3517 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 3518 return (0); 3519} 3520 3521/* 3522 * Check validity of the structure before insert. 3523 * Fortunately rules are simple, so this mostly need to check rule sizes. 3524 */ 3525static int 3526check_ipfw_struct(struct ip_fw *rule, int size) 3527{ 3528 int l, cmdlen = 0; 3529 int have_action=0; 3530 ipfw_insn *cmd; 3531 3532 if (size < sizeof(*rule)) { 3533 printf("ipfw: rule too short\n"); 3534 return (EINVAL); 3535 } 3536 /* first, check for valid size */ 3537 l = RULESIZE(rule); 3538 if (l != size) { 3539 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3540 return (EINVAL); 3541 } 3542 if (rule->act_ofs >= rule->cmd_len) { 3543 printf("ipfw: bogus action offset (%u > %u)\n", 3544 rule->act_ofs, rule->cmd_len - 1); 3545 return (EINVAL); 3546 } 3547 /* 3548 * Now go for the individual checks. Very simple ones, basically only 3549 * instruction sizes. 3550 */ 3551 for (l = rule->cmd_len, cmd = rule->cmd ; 3552 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3553 cmdlen = F_LEN(cmd); 3554 if (cmdlen > l) { 3555 printf("ipfw: opcode %d size truncated\n", 3556 cmd->opcode); 3557 return EINVAL; 3558 } 3559 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3560 switch (cmd->opcode) { 3561 case O_PROBE_STATE: 3562 case O_KEEP_STATE: 3563 case O_PROTO: 3564 case O_IP_SRC_ME: 3565 case O_IP_DST_ME: 3566 case O_LAYER2: 3567 case O_IN: 3568 case O_FRAG: 3569 case O_DIVERTED: 3570 case O_IPOPT: 3571 case O_IPTOS: 3572 case O_IPPRECEDENCE: 3573 case O_IPVER: 3574 case O_TCPWIN: 3575 case O_TCPFLAGS: 3576 case O_TCPOPTS: 3577 case O_ESTAB: 3578 case O_VERREVPATH: 3579 case O_VERSRCREACH: 3580 case O_ANTISPOOF: 3581 case O_IPSEC: 3582#ifdef INET6 3583 case O_IP6_SRC_ME: 3584 case O_IP6_DST_ME: 3585 case O_EXT_HDR: 3586 case O_IP6: 3587#endif 3588 case O_IP4: 3589 case O_TAG: 3590 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3591 goto bad_size; 3592 break; 3593 3594 case O_UID: 3595 case O_GID: 3596 case O_JAIL: 3597 case O_IP_SRC: 3598 case O_IP_DST: 3599 case O_TCPSEQ: 3600 case O_TCPACK: 3601 case O_PROB: 3602 case O_ICMPTYPE: 3603 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3604 goto bad_size; 3605 break; 3606 3607 case O_LIMIT: 3608 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3609 goto bad_size; 3610 break; 3611 3612 case O_LOG: 3613 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3614 goto bad_size; 3615 3616 ((ipfw_insn_log *)cmd)->log_left = 3617 ((ipfw_insn_log *)cmd)->max_log; 3618 3619 break; 3620 3621 case O_IP_SRC_MASK: 3622 case O_IP_DST_MASK: 3623 /* only odd command lengths */ 3624 if ( !(cmdlen & 1) || cmdlen > 31) 3625 goto bad_size; 3626 break; 3627 3628 case O_IP_SRC_SET: 3629 case O_IP_DST_SET: 3630 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3631 printf("ipfw: invalid set size %d\n", 3632 cmd->arg1); 3633 return EINVAL; 3634 } 3635 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3636 (cmd->arg1+31)/32 ) 3637 goto bad_size; 3638 break; 3639 3640 case O_IP_SRC_LOOKUP: 3641 case O_IP_DST_LOOKUP: 3642 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3643 printf("ipfw: invalid table number %d\n", 3644 cmd->arg1); 3645 return (EINVAL); 3646 } 3647 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3648 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3649 goto bad_size; 3650 break; 3651 3652 case O_MACADDR2: 3653 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3654 goto bad_size; 3655 break; 3656 3657 case O_NOP: 3658 case O_IPID: 3659 case O_IPTTL: 3660 case O_IPLEN: 3661 case O_TCPDATALEN: 3662 case O_TAGGED: 3663 if (cmdlen < 1 || cmdlen > 31) 3664 goto bad_size; 3665 break; 3666 3667 case O_MAC_TYPE: 3668 case O_IP_SRCPORT: 3669 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3670 if (cmdlen < 2 || cmdlen > 31) 3671 goto bad_size; 3672 break; 3673 3674 case O_RECV: 3675 case O_XMIT: 3676 case O_VIA: 3677 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3678 goto bad_size; 3679 break; 3680 3681 case O_ALTQ: 3682 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3683 goto bad_size; 3684 break; 3685 3686 case O_PIPE: 3687 case O_QUEUE: 3688 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3689 goto bad_size; 3690 goto check_action; 3691 3692 case O_FORWARD_IP: 3693#ifdef IPFIREWALL_FORWARD 3694 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3695 goto bad_size; 3696 goto check_action; 3697#else 3698 return EINVAL; 3699#endif 3700 3701 case O_DIVERT: 3702 case O_TEE: 3703 if (ip_divert_ptr == NULL) 3704 return EINVAL; 3705 else 3706 goto check_size; 3707 case O_NETGRAPH: 3708 case O_NGTEE: 3709 if (!NG_IPFW_LOADED) 3710 return EINVAL; 3711 else 3712 goto check_size; 3713 case O_FORWARD_MAC: /* XXX not implemented yet */ 3714 case O_CHECK_STATE: 3715 case O_COUNT: 3716 case O_ACCEPT: 3717 case O_DENY: 3718 case O_REJECT: 3719#ifdef INET6 3720 case O_UNREACH6: 3721#endif 3722 case O_SKIPTO: 3723check_size: 3724 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3725 goto bad_size; 3726check_action: 3727 if (have_action) { 3728 printf("ipfw: opcode %d, multiple actions" 3729 " not allowed\n", 3730 cmd->opcode); 3731 return EINVAL; 3732 } 3733 have_action = 1; 3734 if (l != cmdlen) { 3735 printf("ipfw: opcode %d, action must be" 3736 " last opcode\n", 3737 cmd->opcode); 3738 return EINVAL; 3739 } 3740 break; 3741#ifdef INET6 3742 case O_IP6_SRC: 3743 case O_IP6_DST: 3744 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 3745 F_INSN_SIZE(ipfw_insn)) 3746 goto bad_size; 3747 break; 3748 3749 case O_FLOW6ID: 3750 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3751 ((ipfw_insn_u32 *)cmd)->o.arg1) 3752 goto bad_size; 3753 break; 3754 3755 case O_IP6_SRC_MASK: 3756 case O_IP6_DST_MASK: 3757 if ( !(cmdlen & 1) || cmdlen > 127) 3758 goto bad_size; 3759 break; 3760 case O_ICMP6TYPE: 3761 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 3762 goto bad_size; 3763 break; 3764#endif 3765 3766 default: 3767 switch (cmd->opcode) { 3768#ifndef INET6 3769 case O_IP6_SRC_ME: 3770 case O_IP6_DST_ME: 3771 case O_EXT_HDR: 3772 case O_IP6: 3773 case O_UNREACH6: 3774 case O_IP6_SRC: 3775 case O_IP6_DST: 3776 case O_FLOW6ID: 3777 case O_IP6_SRC_MASK: 3778 case O_IP6_DST_MASK: 3779 case O_ICMP6TYPE: 3780 printf("ipfw: no IPv6 support in kernel\n"); 3781 return EPROTONOSUPPORT; 3782#endif 3783 default: 3784 printf("ipfw: opcode %d, unknown opcode\n", 3785 cmd->opcode); 3786 return EINVAL; 3787 } 3788 } 3789 } 3790 if (have_action == 0) { 3791 printf("ipfw: missing action\n"); 3792 return EINVAL; 3793 } 3794 return 0; 3795 3796bad_size: 3797 printf("ipfw: opcode %d size %d wrong\n", 3798 cmd->opcode, cmdlen); 3799 return EINVAL; 3800} 3801 3802/* 3803 * Copy the static and dynamic rules to the supplied buffer 3804 * and return the amount of space actually used. 3805 */ 3806static size_t 3807ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 3808{ 3809 char *bp = buf; 3810 char *ep = bp + space; 3811 struct ip_fw *rule; 3812 int i; 3813 3814 /* XXX this can take a long time and locking will block packet flow */ 3815 IPFW_RLOCK(chain); 3816 for (rule = chain->rules; rule ; rule = rule->next) { 3817 /* 3818 * Verify the entry fits in the buffer in case the 3819 * rules changed between calculating buffer space and 3820 * now. This would be better done using a generation 3821 * number but should suffice for now. 3822 */ 3823 i = RULESIZE(rule); 3824 if (bp + i <= ep) { 3825 bcopy(rule, bp, i); 3826 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 3827 sizeof(set_disable)); 3828 bp += i; 3829 } 3830 } 3831 IPFW_RUNLOCK(chain); 3832 if (ipfw_dyn_v) { 3833 ipfw_dyn_rule *p, *last = NULL; 3834 3835 IPFW_DYN_LOCK(); 3836 for (i = 0 ; i < curr_dyn_buckets; i++) 3837 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 3838 if (bp + sizeof *p <= ep) { 3839 ipfw_dyn_rule *dst = 3840 (ipfw_dyn_rule *)bp; 3841 bcopy(p, dst, sizeof *p); 3842 bcopy(&(p->rule->rulenum), &(dst->rule), 3843 sizeof(p->rule->rulenum)); 3844 /* 3845 * store a non-null value in "next". 3846 * The userland code will interpret a 3847 * NULL here as a marker 3848 * for the last dynamic rule. 3849 */ 3850 bcopy(&dst, &dst->next, sizeof(dst)); 3851 last = dst; 3852 dst->expire = 3853 TIME_LEQ(dst->expire, time_uptime) ? 3854 0 : dst->expire - time_uptime ; 3855 bp += sizeof(ipfw_dyn_rule); 3856 } 3857 } 3858 IPFW_DYN_UNLOCK(); 3859 if (last != NULL) /* mark last dynamic rule */ 3860 bzero(&last->next, sizeof(last)); 3861 } 3862 return (bp - (char *)buf); 3863} 3864 3865 3866/** 3867 * {set|get}sockopt parser. 3868 */ 3869static int 3870ipfw_ctl(struct sockopt *sopt) 3871{ 3872#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3873 int error, rule_num; 3874 size_t size; 3875 struct ip_fw *buf, *rule; 3876 u_int32_t rulenum[2]; 3877 3878 error = suser(sopt->sopt_td); 3879 if (error) 3880 return (error); 3881 3882 /* 3883 * Disallow modifications in really-really secure mode, but still allow 3884 * the logging counters to be reset. 3885 */ 3886 if (sopt->sopt_name == IP_FW_ADD || 3887 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 3888 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3889 if (error) 3890 return (error); 3891 } 3892 3893 error = 0; 3894 3895 switch (sopt->sopt_name) { 3896 case IP_FW_GET: 3897 /* 3898 * pass up a copy of the current rules. Static rules 3899 * come first (the last of which has number IPFW_DEFAULT_RULE), 3900 * followed by a possibly empty list of dynamic rule. 3901 * The last dynamic rule has NULL in the "next" field. 3902 * 3903 * Note that the calculated size is used to bound the 3904 * amount of data returned to the user. The rule set may 3905 * change between calculating the size and returning the 3906 * data in which case we'll just return what fits. 3907 */ 3908 size = static_len; /* size of static rules */ 3909 if (ipfw_dyn_v) /* add size of dyn.rules */ 3910 size += (dyn_count * sizeof(ipfw_dyn_rule)); 3911 3912 /* 3913 * XXX todo: if the user passes a short length just to know 3914 * how much room is needed, do not bother filling up the 3915 * buffer, just jump to the sooptcopyout. 3916 */ 3917 buf = malloc(size, M_TEMP, M_WAITOK); 3918 error = sooptcopyout(sopt, buf, 3919 ipfw_getrules(&layer3_chain, buf, size)); 3920 free(buf, M_TEMP); 3921 break; 3922 3923 case IP_FW_FLUSH: 3924 /* 3925 * Normally we cannot release the lock on each iteration. 3926 * We could do it here only because we start from the head all 3927 * the times so there is no risk of missing some entries. 3928 * On the other hand, the risk is that we end up with 3929 * a very inconsistent ruleset, so better keep the lock 3930 * around the whole cycle. 3931 * 3932 * XXX this code can be improved by resetting the head of 3933 * the list to point to the default rule, and then freeing 3934 * the old list without the need for a lock. 3935 */ 3936 3937 IPFW_WLOCK(&layer3_chain); 3938 layer3_chain.reap = NULL; 3939 free_chain(&layer3_chain, 0 /* keep default rule */); 3940 rule = layer3_chain.reap, layer3_chain.reap = NULL; 3941 IPFW_WUNLOCK(&layer3_chain); 3942 if (layer3_chain.reap != NULL) 3943 reap_rules(rule); 3944 break; 3945 3946 case IP_FW_ADD: 3947 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 3948 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 3949 sizeof(struct ip_fw) ); 3950 if (error == 0) 3951 error = check_ipfw_struct(rule, sopt->sopt_valsize); 3952 if (error == 0) { 3953 error = add_rule(&layer3_chain, rule); 3954 size = RULESIZE(rule); 3955 if (!error && sopt->sopt_dir == SOPT_GET) 3956 error = sooptcopyout(sopt, rule, size); 3957 } 3958 free(rule, M_TEMP); 3959 break; 3960 3961 case IP_FW_DEL: 3962 /* 3963 * IP_FW_DEL is used for deleting single rules or sets, 3964 * and (ab)used to atomically manipulate sets. Argument size 3965 * is used to distinguish between the two: 3966 * sizeof(u_int32_t) 3967 * delete single rule or set of rules, 3968 * or reassign rules (or sets) to a different set. 3969 * 2*sizeof(u_int32_t) 3970 * atomic disable/enable sets. 3971 * first u_int32_t contains sets to be disabled, 3972 * second u_int32_t contains sets to be enabled. 3973 */ 3974 error = sooptcopyin(sopt, rulenum, 3975 2*sizeof(u_int32_t), sizeof(u_int32_t)); 3976 if (error) 3977 break; 3978 size = sopt->sopt_valsize; 3979 if (size == sizeof(u_int32_t)) /* delete or reassign */ 3980 error = del_entry(&layer3_chain, rulenum[0]); 3981 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 3982 set_disable = 3983 (set_disable | rulenum[0]) & ~rulenum[1] & 3984 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 3985 else 3986 error = EINVAL; 3987 break; 3988 3989 case IP_FW_ZERO: 3990 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 3991 rule_num = 0; 3992 if (sopt->sopt_val != 0) { 3993 error = sooptcopyin(sopt, &rule_num, 3994 sizeof(int), sizeof(int)); 3995 if (error) 3996 break; 3997 } 3998 error = zero_entry(&layer3_chain, rule_num, 3999 sopt->sopt_name == IP_FW_RESETLOG); 4000 break; 4001 4002 case IP_FW_TABLE_ADD: 4003 { 4004 ipfw_table_entry ent; 4005 4006 error = sooptcopyin(sopt, &ent, 4007 sizeof(ent), sizeof(ent)); 4008 if (error) 4009 break; 4010 error = add_table_entry(&layer3_chain, ent.tbl, 4011 ent.addr, ent.masklen, ent.value); 4012 } 4013 break; 4014 4015 case IP_FW_TABLE_DEL: 4016 { 4017 ipfw_table_entry ent; 4018 4019 error = sooptcopyin(sopt, &ent, 4020 sizeof(ent), sizeof(ent)); 4021 if (error) 4022 break; 4023 error = del_table_entry(&layer3_chain, ent.tbl, 4024 ent.addr, ent.masklen); 4025 } 4026 break; 4027 4028 case IP_FW_TABLE_FLUSH: 4029 { 4030 u_int16_t tbl; 4031 4032 error = sooptcopyin(sopt, &tbl, 4033 sizeof(tbl), sizeof(tbl)); 4034 if (error) 4035 break; 4036 IPFW_WLOCK(&layer3_chain); 4037 error = flush_table(&layer3_chain, tbl); 4038 IPFW_WUNLOCK(&layer3_chain); 4039 } 4040 break; 4041 4042 case IP_FW_TABLE_GETSIZE: 4043 { 4044 u_int32_t tbl, cnt; 4045 4046 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4047 sizeof(tbl)))) 4048 break; 4049 IPFW_RLOCK(&layer3_chain); 4050 error = count_table(&layer3_chain, tbl, &cnt); 4051 IPFW_RUNLOCK(&layer3_chain); 4052 if (error) 4053 break; 4054 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4055 } 4056 break; 4057 4058 case IP_FW_TABLE_LIST: 4059 { 4060 ipfw_table *tbl; 4061 4062 if (sopt->sopt_valsize < sizeof(*tbl)) { 4063 error = EINVAL; 4064 break; 4065 } 4066 size = sopt->sopt_valsize; 4067 tbl = malloc(size, M_TEMP, M_WAITOK); 4068 if (tbl == NULL) { 4069 error = ENOMEM; 4070 break; 4071 } 4072 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4073 if (error) { 4074 free(tbl, M_TEMP); 4075 break; 4076 } 4077 tbl->size = (size - sizeof(*tbl)) / 4078 sizeof(ipfw_table_entry); 4079 IPFW_RLOCK(&layer3_chain); 4080 error = dump_table(&layer3_chain, tbl); 4081 IPFW_RUNLOCK(&layer3_chain); 4082 if (error) { 4083 free(tbl, M_TEMP); 4084 break; 4085 } 4086 error = sooptcopyout(sopt, tbl, size); 4087 free(tbl, M_TEMP); 4088 } 4089 break; 4090 4091 default: 4092 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4093 error = EINVAL; 4094 } 4095 4096 return (error); 4097#undef RULE_MAXSIZE 4098} 4099 4100/** 4101 * dummynet needs a reference to the default rule, because rules can be 4102 * deleted while packets hold a reference to them. When this happens, 4103 * dummynet changes the reference to the default rule (it could well be a 4104 * NULL pointer, but this way we do not need to check for the special 4105 * case, plus here he have info on the default behaviour). 4106 */ 4107struct ip_fw *ip_fw_default_rule; 4108 4109/* 4110 * This procedure is only used to handle keepalives. It is invoked 4111 * every dyn_keepalive_period 4112 */ 4113static void 4114ipfw_tick(void * __unused unused) 4115{ 4116 struct mbuf *m0, *m, *mnext, **mtailp; 4117 int i; 4118 ipfw_dyn_rule *q; 4119 4120 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 4121 goto done; 4122 4123 /* 4124 * We make a chain of packets to go out here -- not deferring 4125 * until after we drop the IPFW dynamic rule lock would result 4126 * in a lock order reversal with the normal packet input -> ipfw 4127 * call stack. 4128 */ 4129 m0 = NULL; 4130 mtailp = &m0; 4131 IPFW_DYN_LOCK(); 4132 for (i = 0 ; i < curr_dyn_buckets ; i++) { 4133 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 4134 if (q->dyn_type == O_LIMIT_PARENT) 4135 continue; 4136 if (q->id.proto != IPPROTO_TCP) 4137 continue; 4138 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4139 continue; 4140 if (TIME_LEQ( time_uptime+dyn_keepalive_interval, 4141 q->expire)) 4142 continue; /* too early */ 4143 if (TIME_LEQ(q->expire, time_uptime)) 4144 continue; /* too late, rule expired */ 4145 4146 *mtailp = send_pkt(&(q->id), q->ack_rev - 1, 4147 q->ack_fwd, TH_SYN); 4148 if (*mtailp != NULL) 4149 mtailp = &(*mtailp)->m_nextpkt; 4150 *mtailp = send_pkt(&(q->id), q->ack_fwd - 1, 4151 q->ack_rev, 0); 4152 if (*mtailp != NULL) 4153 mtailp = &(*mtailp)->m_nextpkt; 4154 } 4155 } 4156 IPFW_DYN_UNLOCK(); 4157 for (m = mnext = m0; m != NULL; m = mnext) { 4158 mnext = m->m_nextpkt; 4159 m->m_nextpkt = NULL; 4160 ip_output(m, NULL, NULL, 0, NULL, NULL); 4161 } 4162done: 4163 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 4164} 4165 4166int 4167ipfw_init(void) 4168{ 4169 struct ip_fw default_rule; 4170 int error; 4171 4172#ifdef INET6 4173 /* Setup IPv6 fw sysctl tree. */ 4174 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4175 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4176 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4177 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4178 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4179 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4180 &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4181 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4182 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4183 &fw_deny_unknown_exthdrs, 0, 4184 "Deny packets with unknown IPv6 Extension Headers"); 4185#endif 4186 4187 layer3_chain.rules = NULL; 4188 IPFW_LOCK_INIT(&layer3_chain); 4189 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule zone", 4190 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4191 UMA_ALIGN_PTR, 0); 4192 IPFW_DYN_LOCK_INIT(); 4193 callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE); 4194 4195 bzero(&default_rule, sizeof default_rule); 4196 4197 default_rule.act_ofs = 0; 4198 default_rule.rulenum = IPFW_DEFAULT_RULE; 4199 default_rule.cmd_len = 1; 4200 default_rule.set = RESVD_SET; 4201 4202 default_rule.cmd[0].len = 1; 4203 default_rule.cmd[0].opcode = 4204#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4205 1 ? O_ACCEPT : 4206#endif 4207 O_DENY; 4208 4209 error = add_rule(&layer3_chain, &default_rule); 4210 if (error != 0) { 4211 printf("ipfw2: error %u initializing default rule " 4212 "(support disabled)\n", error); 4213 IPFW_DYN_LOCK_DESTROY(); 4214 IPFW_LOCK_DESTROY(&layer3_chain); 4215 uma_zdestroy(ipfw_dyn_rule_zone); 4216 return (error); 4217 } 4218 4219 ip_fw_default_rule = layer3_chain.rules; 4220 printf("ipfw2 " 4221#ifdef INET6 4222 "(+ipv6) " 4223#endif 4224 "initialized, divert %s, " 4225 "rule-based forwarding " 4226#ifdef IPFIREWALL_FORWARD 4227 "enabled, " 4228#else 4229 "disabled, " 4230#endif 4231 "default to %s, logging ", 4232#ifdef IPDIVERT 4233 "enabled", 4234#else 4235 "loadable", 4236#endif 4237 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4238 4239#ifdef IPFIREWALL_VERBOSE 4240 fw_verbose = 1; 4241#endif 4242#ifdef IPFIREWALL_VERBOSE_LIMIT 4243 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4244#endif 4245 if (fw_verbose == 0) 4246 printf("disabled\n"); 4247 else if (verbose_limit == 0) 4248 printf("unlimited\n"); 4249 else 4250 printf("limited to %d packets/entry by default\n", 4251 verbose_limit); 4252 4253 error = init_tables(&layer3_chain); 4254 if (error) { 4255 IPFW_DYN_LOCK_DESTROY(); 4256 IPFW_LOCK_DESTROY(&layer3_chain); 4257 uma_zdestroy(ipfw_dyn_rule_zone); 4258 return (error); 4259 } 4260 ip_fw_ctl_ptr = ipfw_ctl; 4261 ip_fw_chk_ptr = ipfw_chk; 4262 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 4263 4264 return (0); 4265} 4266 4267void 4268ipfw_destroy(void) 4269{ 4270 struct ip_fw *reap; 4271 4272 ip_fw_chk_ptr = NULL; 4273 ip_fw_ctl_ptr = NULL; 4274 callout_drain(&ipfw_timeout); 4275 IPFW_WLOCK(&layer3_chain); 4276 flush_tables(&layer3_chain); 4277 layer3_chain.reap = NULL; 4278 free_chain(&layer3_chain, 1 /* kill default rule */); 4279 reap = layer3_chain.reap, layer3_chain.reap = NULL; 4280 IPFW_WUNLOCK(&layer3_chain); 4281 if (reap != NULL) 4282 reap_rules(reap); 4283 IPFW_DYN_LOCK_DESTROY(); 4284 uma_zdestroy(ipfw_dyn_rule_zone); 4285 IPFW_LOCK_DESTROY(&layer3_chain); 4286 4287#ifdef INET6 4288 /* Free IPv6 fw sysctl tree. */ 4289 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4290#endif 4291 4292 printf("IP firewall unloaded\n"); 4293} 4294