ip_fw2.c revision 163237
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 163237 2006-10-11 12:14:28Z maxim $ 26 */ 27 28#define DEB(x) 29#define DDB(x) x 30 31/* 32 * Implement IP packet firewall (new version) 33 */ 34 35#if !defined(KLD_MODULE) 36#include "opt_ipfw.h" 37#include "opt_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#include "opt_mac.h" 46 47#include <sys/param.h> 48#include <sys/systm.h> 49#include <sys/condvar.h> 50#include <sys/malloc.h> 51#include <sys/mbuf.h> 52#include <sys/kernel.h> 53#include <sys/lock.h> 54#include <sys/jail.h> 55#include <sys/mac.h> 56#include <sys/module.h> 57#include <sys/proc.h> 58#include <sys/rwlock.h> 59#include <sys/socket.h> 60#include <sys/socketvar.h> 61#include <sys/sysctl.h> 62#include <sys/syslog.h> 63#include <sys/ucred.h> 64#include <net/if.h> 65#include <net/radix.h> 66#include <net/route.h> 67#include <netinet/in.h> 68#include <netinet/in_systm.h> 69#include <netinet/in_var.h> 70#include <netinet/in_pcb.h> 71#include <netinet/ip.h> 72#include <netinet/ip_var.h> 73#include <netinet/ip_icmp.h> 74#include <netinet/ip_fw.h> 75#include <netinet/ip_divert.h> 76#include <netinet/ip_dummynet.h> 77#include <netinet/ip_carp.h> 78#include <netinet/pim.h> 79#include <netinet/tcp.h> 80#include <netinet/tcp_timer.h> 81#include <netinet/tcp_var.h> 82#include <netinet/tcpip.h> 83#include <netinet/udp.h> 84#include <netinet/udp_var.h> 85 86#include <netgraph/ng_ipfw.h> 87 88#include <altq/if_altq.h> 89 90#ifdef IPSEC 91#include <netinet6/ipsec.h> 92#endif 93 94#include <netinet/ip6.h> 95#include <netinet/icmp6.h> 96#ifdef INET6 97#include <netinet6/scope6_var.h> 98#endif 99 100#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 101 102#include <machine/in_cksum.h> /* XXX for in_cksum */ 103 104/* 105 * set_disable contains one bit per set value (0..31). 106 * If the bit is set, all rules with the corresponding set 107 * are disabled. Set RESVD_SET(31) is reserved for the default rule 108 * and rules that are not deleted by the flush command, 109 * and CANNOT be disabled. 110 * Rules in set RESVD_SET can only be deleted explicitly. 111 */ 112static u_int32_t set_disable; 113 114static int fw_verbose; 115static int verbose_limit; 116 117static struct callout ipfw_timeout; 118static uma_zone_t ipfw_dyn_rule_zone; 119#define IPFW_DEFAULT_RULE 65535 120 121/* 122 * Data structure to cache our ucred related 123 * information. This structure only gets used if 124 * the user specified UID/GID based constraints in 125 * a firewall rule. 126 */ 127struct ip_fw_ugid { 128 gid_t fw_groups[NGROUPS]; 129 int fw_ngroups; 130 uid_t fw_uid; 131 int fw_prid; 132}; 133 134#define IPFW_TABLES_MAX 128 135struct ip_fw_chain { 136 struct ip_fw *rules; /* list of rules */ 137 struct ip_fw *reap; /* list of rules to reap */ 138 struct radix_node_head *tables[IPFW_TABLES_MAX]; 139 struct rwlock rwmtx; 140}; 141#define IPFW_LOCK_INIT(_chain) \ 142 rw_init(&(_chain)->rwmtx, "IPFW static rules") 143#define IPFW_LOCK_DESTROY(_chain) rw_destroy(&(_chain)->rwmtx) 144#define IPFW_WLOCK_ASSERT(_chain) do { \ 145 rw_assert(&(_chain)->rwmtx, RA_WLOCKED); \ 146 NET_ASSERT_GIANT(); \ 147} while (0) 148 149#define IPFW_RLOCK(p) rw_rlock(&(p)->rwmtx) 150#define IPFW_RUNLOCK(p) rw_runlock(&(p)->rwmtx) 151#define IPFW_WLOCK(p) rw_wlock(&(p)->rwmtx) 152#define IPFW_WUNLOCK(p) rw_wunlock(&(p)->rwmtx) 153 154/* 155 * list of rules for layer 3 156 */ 157static struct ip_fw_chain layer3_chain; 158 159MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 160MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 161 162struct table_entry { 163 struct radix_node rn[2]; 164 struct sockaddr_in addr, mask; 165 u_int32_t value; 166}; 167 168static int fw_debug = 1; 169static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 170 171extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS); 172 173#ifdef SYSCTL_NODE 174SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 175SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable, 176 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &fw_enable, 0, 177 ipfw_chg_hook, "I", "Enable ipfw"); 178SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 179 &autoinc_step, 0, "Rule number autincrement step"); 180SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 181 CTLFLAG_RW | CTLFLAG_SECURE3, 182 &fw_one_pass, 0, 183 "Only do a single pass through ipfw when using dummynet(4)"); 184SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 185 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 186SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 187 CTLFLAG_RW | CTLFLAG_SECURE3, 188 &fw_verbose, 0, "Log matches to ipfw rules"); 189SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 190 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 191 192/* 193 * Description of dynamic rules. 194 * 195 * Dynamic rules are stored in lists accessed through a hash table 196 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 197 * be modified through the sysctl variable dyn_buckets which is 198 * updated when the table becomes empty. 199 * 200 * XXX currently there is only one list, ipfw_dyn. 201 * 202 * When a packet is received, its address fields are first masked 203 * with the mask defined for the rule, then hashed, then matched 204 * against the entries in the corresponding list. 205 * Dynamic rules can be used for different purposes: 206 * + stateful rules; 207 * + enforcing limits on the number of sessions; 208 * + in-kernel NAT (not implemented yet) 209 * 210 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 211 * measured in seconds and depending on the flags. 212 * 213 * The total number of dynamic rules is stored in dyn_count. 214 * The max number of dynamic rules is dyn_max. When we reach 215 * the maximum number of rules we do not create anymore. This is 216 * done to avoid consuming too much memory, but also too much 217 * time when searching on each packet (ideally, we should try instead 218 * to put a limit on the length of the list on each bucket...). 219 * 220 * Each dynamic rule holds a pointer to the parent ipfw rule so 221 * we know what action to perform. Dynamic rules are removed when 222 * the parent rule is deleted. XXX we should make them survive. 223 * 224 * There are some limitations with dynamic rules -- we do not 225 * obey the 'randomized match', and we do not do multiple 226 * passes through the firewall. XXX check the latter!!! 227 */ 228static ipfw_dyn_rule **ipfw_dyn_v = NULL; 229static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 230static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 231 232static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 233#define IPFW_DYN_LOCK_INIT() \ 234 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 235#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 236#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 237#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 238#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 239 240/* 241 * Timeouts for various events in handing dynamic rules. 242 */ 243static u_int32_t dyn_ack_lifetime = 300; 244static u_int32_t dyn_syn_lifetime = 20; 245static u_int32_t dyn_fin_lifetime = 1; 246static u_int32_t dyn_rst_lifetime = 1; 247static u_int32_t dyn_udp_lifetime = 10; 248static u_int32_t dyn_short_lifetime = 5; 249 250/* 251 * Keepalives are sent if dyn_keepalive is set. They are sent every 252 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 253 * seconds of lifetime of a rule. 254 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 255 * than dyn_keepalive_period. 256 */ 257 258static u_int32_t dyn_keepalive_interval = 20; 259static u_int32_t dyn_keepalive_period = 5; 260static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 261 262static u_int32_t static_count; /* # of static rules */ 263static u_int32_t static_len; /* size in bytes of static rules */ 264static u_int32_t dyn_count; /* # of dynamic rules */ 265static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 266 267SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 268 &dyn_buckets, 0, "Number of dyn. buckets"); 269SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 270 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 271SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 272 &dyn_count, 0, "Number of dyn. rules"); 273SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 274 &dyn_max, 0, "Max number of dyn. rules"); 275SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 276 &static_count, 0, "Number of static rules"); 277SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 278 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 279SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 280 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 281SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 282 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 283SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 284 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 285SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 286 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 287SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 288 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 289SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 290 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 291 292#ifdef INET6 293/* 294 * IPv6 specific variables 295 */ 296SYSCTL_DECL(_net_inet6_ip6); 297 298static struct sysctl_ctx_list ip6_fw_sysctl_ctx; 299static struct sysctl_oid *ip6_fw_sysctl_tree; 300#endif /* INET6 */ 301#endif /* SYSCTL_NODE */ 302 303static int fw_deny_unknown_exthdrs = 1; 304 305 306/* 307 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 308 * Other macros just cast void * into the appropriate type 309 */ 310#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 311#define TCP(p) ((struct tcphdr *)(p)) 312#define UDP(p) ((struct udphdr *)(p)) 313#define ICMP(p) ((struct icmphdr *)(p)) 314#define ICMP6(p) ((struct icmp6_hdr *)(p)) 315 316static __inline int 317icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 318{ 319 int type = icmp->icmp_type; 320 321 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 322} 323 324#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 325 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 326 327static int 328is_icmp_query(struct icmphdr *icmp) 329{ 330 int type = icmp->icmp_type; 331 332 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 333} 334#undef TT 335 336/* 337 * The following checks use two arrays of 8 or 16 bits to store the 338 * bits that we want set or clear, respectively. They are in the 339 * low and high half of cmd->arg1 or cmd->d[0]. 340 * 341 * We scan options and store the bits we find set. We succeed if 342 * 343 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 344 * 345 * The code is sometimes optimized not to store additional variables. 346 */ 347 348static int 349flags_match(ipfw_insn *cmd, u_int8_t bits) 350{ 351 u_char want_clear; 352 bits = ~bits; 353 354 if ( ((cmd->arg1 & 0xff) & bits) != 0) 355 return 0; /* some bits we want set were clear */ 356 want_clear = (cmd->arg1 >> 8) & 0xff; 357 if ( (want_clear & bits) != want_clear) 358 return 0; /* some bits we want clear were set */ 359 return 1; 360} 361 362static int 363ipopts_match(struct ip *ip, ipfw_insn *cmd) 364{ 365 int optlen, bits = 0; 366 u_char *cp = (u_char *)(ip + 1); 367 int x = (ip->ip_hl << 2) - sizeof (struct ip); 368 369 for (; x > 0; x -= optlen, cp += optlen) { 370 int opt = cp[IPOPT_OPTVAL]; 371 372 if (opt == IPOPT_EOL) 373 break; 374 if (opt == IPOPT_NOP) 375 optlen = 1; 376 else { 377 optlen = cp[IPOPT_OLEN]; 378 if (optlen <= 0 || optlen > x) 379 return 0; /* invalid or truncated */ 380 } 381 switch (opt) { 382 383 default: 384 break; 385 386 case IPOPT_LSRR: 387 bits |= IP_FW_IPOPT_LSRR; 388 break; 389 390 case IPOPT_SSRR: 391 bits |= IP_FW_IPOPT_SSRR; 392 break; 393 394 case IPOPT_RR: 395 bits |= IP_FW_IPOPT_RR; 396 break; 397 398 case IPOPT_TS: 399 bits |= IP_FW_IPOPT_TS; 400 break; 401 } 402 } 403 return (flags_match(cmd, bits)); 404} 405 406static int 407tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 408{ 409 int optlen, bits = 0; 410 u_char *cp = (u_char *)(tcp + 1); 411 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 412 413 for (; x > 0; x -= optlen, cp += optlen) { 414 int opt = cp[0]; 415 if (opt == TCPOPT_EOL) 416 break; 417 if (opt == TCPOPT_NOP) 418 optlen = 1; 419 else { 420 optlen = cp[1]; 421 if (optlen <= 0) 422 break; 423 } 424 425 switch (opt) { 426 427 default: 428 break; 429 430 case TCPOPT_MAXSEG: 431 bits |= IP_FW_TCPOPT_MSS; 432 break; 433 434 case TCPOPT_WINDOW: 435 bits |= IP_FW_TCPOPT_WINDOW; 436 break; 437 438 case TCPOPT_SACK_PERMITTED: 439 case TCPOPT_SACK: 440 bits |= IP_FW_TCPOPT_SACK; 441 break; 442 443 case TCPOPT_TIMESTAMP: 444 bits |= IP_FW_TCPOPT_TS; 445 break; 446 447 } 448 } 449 return (flags_match(cmd, bits)); 450} 451 452static int 453iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 454{ 455 if (ifp == NULL) /* no iface with this packet, match fails */ 456 return 0; 457 /* Check by name or by IP address */ 458 if (cmd->name[0] != '\0') { /* match by name */ 459 /* Check name */ 460 if (cmd->p.glob) { 461 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 462 return(1); 463 } else { 464 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 465 return(1); 466 } 467 } else { 468 struct ifaddr *ia; 469 470 /* XXX lock? */ 471 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 472 if (ia->ifa_addr->sa_family != AF_INET) 473 continue; 474 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 475 (ia->ifa_addr))->sin_addr.s_addr) 476 return(1); /* match */ 477 } 478 } 479 return(0); /* no match, fail ... */ 480} 481 482/* 483 * The verify_path function checks if a route to the src exists and 484 * if it is reachable via ifp (when provided). 485 * 486 * The 'verrevpath' option checks that the interface that an IP packet 487 * arrives on is the same interface that traffic destined for the 488 * packet's source address would be routed out of. The 'versrcreach' 489 * option just checks that the source address is reachable via any route 490 * (except default) in the routing table. These two are a measure to block 491 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 492 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 493 * is purposely reminiscent of the Cisco IOS command, 494 * 495 * ip verify unicast reverse-path 496 * ip verify unicast source reachable-via any 497 * 498 * which implements the same functionality. But note that syntax is 499 * misleading. The check may be performed on all IP packets whether unicast, 500 * multicast, or broadcast. 501 */ 502static int 503verify_path(struct in_addr src, struct ifnet *ifp) 504{ 505 struct route ro; 506 struct sockaddr_in *dst; 507 508 bzero(&ro, sizeof(ro)); 509 510 dst = (struct sockaddr_in *)&(ro.ro_dst); 511 dst->sin_family = AF_INET; 512 dst->sin_len = sizeof(*dst); 513 dst->sin_addr = src; 514 rtalloc_ign(&ro, RTF_CLONING); 515 516 if (ro.ro_rt == NULL) 517 return 0; 518 519 /* 520 * If ifp is provided, check for equality with rtentry. 521 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 522 * in order to pass packets injected back by if_simloop(): 523 * if useloopback == 1 routing entry (via lo0) for our own address 524 * may exist, so we need to handle routing assymetry. 525 */ 526 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 527 RTFREE(ro.ro_rt); 528 return 0; 529 } 530 531 /* if no ifp provided, check if rtentry is not default route */ 532 if (ifp == NULL && 533 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 534 RTFREE(ro.ro_rt); 535 return 0; 536 } 537 538 /* or if this is a blackhole/reject route */ 539 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 540 RTFREE(ro.ro_rt); 541 return 0; 542 } 543 544 /* found valid route */ 545 RTFREE(ro.ro_rt); 546 return 1; 547} 548 549#ifdef INET6 550/* 551 * ipv6 specific rules here... 552 */ 553static __inline int 554icmp6type_match (int type, ipfw_insn_u32 *cmd) 555{ 556 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 557} 558 559static int 560flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 561{ 562 int i; 563 for (i=0; i <= cmd->o.arg1; ++i ) 564 if (curr_flow == cmd->d[i] ) 565 return 1; 566 return 0; 567} 568 569/* support for IP6_*_ME opcodes */ 570static int 571search_ip6_addr_net (struct in6_addr * ip6_addr) 572{ 573 struct ifnet *mdc; 574 struct ifaddr *mdc2; 575 struct in6_ifaddr *fdm; 576 struct in6_addr copia; 577 578 TAILQ_FOREACH(mdc, &ifnet, if_link) 579 TAILQ_FOREACH(mdc2, &mdc->if_addrlist, ifa_list) { 580 if (mdc2->ifa_addr->sa_family == AF_INET6) { 581 fdm = (struct in6_ifaddr *)mdc2; 582 copia = fdm->ia_addr.sin6_addr; 583 /* need for leaving scope_id in the sock_addr */ 584 in6_clearscope(&copia); 585 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) 586 return 1; 587 } 588 } 589 return 0; 590} 591 592static int 593verify_path6(struct in6_addr *src, struct ifnet *ifp) 594{ 595 struct route_in6 ro; 596 struct sockaddr_in6 *dst; 597 598 bzero(&ro, sizeof(ro)); 599 600 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 601 dst->sin6_family = AF_INET6; 602 dst->sin6_len = sizeof(*dst); 603 dst->sin6_addr = *src; 604 rtalloc_ign((struct route *)&ro, RTF_CLONING); 605 606 if (ro.ro_rt == NULL) 607 return 0; 608 609 /* 610 * if ifp is provided, check for equality with rtentry 611 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 612 * to support the case of sending packets to an address of our own. 613 * (where the former interface is the first argument of if_simloop() 614 * (=ifp), the latter is lo0) 615 */ 616 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 617 RTFREE(ro.ro_rt); 618 return 0; 619 } 620 621 /* if no ifp provided, check if rtentry is not default route */ 622 if (ifp == NULL && 623 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 624 RTFREE(ro.ro_rt); 625 return 0; 626 } 627 628 /* or if this is a blackhole/reject route */ 629 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 630 RTFREE(ro.ro_rt); 631 return 0; 632 } 633 634 /* found valid route */ 635 RTFREE(ro.ro_rt); 636 return 1; 637 638} 639static __inline int 640hash_packet6(struct ipfw_flow_id *id) 641{ 642 u_int32_t i; 643 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 644 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 645 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 646 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 647 (id->dst_port) ^ (id->src_port); 648 return i; 649} 650 651static int 652is_icmp6_query(int icmp6_type) 653{ 654 if ((icmp6_type <= ICMP6_MAXTYPE) && 655 (icmp6_type == ICMP6_ECHO_REQUEST || 656 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 657 icmp6_type == ICMP6_WRUREQUEST || 658 icmp6_type == ICMP6_FQDN_QUERY || 659 icmp6_type == ICMP6_NI_QUERY)) 660 return (1); 661 662 return (0); 663} 664 665static void 666send_reject6(struct ip_fw_args *args, int code, u_int hlen) 667{ 668 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 669 struct ip6_hdr *ip6; 670 struct tcphdr *tcp; 671 tcp_seq ack, seq; 672 int flags; 673 struct { 674 struct ip6_hdr ip6; 675 struct tcphdr th; 676 } ti; 677 678 if (args->m->m_len < (hlen+sizeof(struct tcphdr))) { 679 args->m = m_pullup(args->m, hlen+sizeof(struct tcphdr)); 680 if (args->m == NULL) 681 return; 682 } 683 684 ip6 = mtod(args->m, struct ip6_hdr *); 685 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 686 687 if ((tcp->th_flags & TH_RST) != 0) { 688 m_freem(args->m); 689 return; 690 } 691 692 ti.ip6 = *ip6; 693 ti.th = *tcp; 694 ti.th.th_seq = ntohl(ti.th.th_seq); 695 ti.th.th_ack = ntohl(ti.th.th_ack); 696 ti.ip6.ip6_nxt = IPPROTO_TCP; 697 698 if (ti.th.th_flags & TH_ACK) { 699 ack = 0; 700 seq = ti.th.th_ack; 701 flags = TH_RST; 702 } else { 703 ack = ti.th.th_seq; 704 if (((args->m)->m_flags & M_PKTHDR) != 0) { 705 ack += (args->m)->m_pkthdr.len - hlen 706 - (ti.th.th_off << 2); 707 } else if (ip6->ip6_plen) { 708 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) 709 - hlen - (ti.th.th_off << 2); 710 } else { 711 m_freem(args->m); 712 return; 713 } 714 if (tcp->th_flags & TH_SYN) 715 ack++; 716 seq = 0; 717 flags = TH_RST|TH_ACK; 718 } 719 bcopy(&ti, ip6, sizeof(ti)); 720 tcp_respond(NULL, ip6, (struct tcphdr *)(ip6 + 1), 721 args->m, ack, seq, flags); 722 723 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 724 icmp6_error(args->m, ICMP6_DST_UNREACH, code, 0); 725 726 } else 727 m_freem(args->m); 728 729 args->m = NULL; 730} 731 732#endif /* INET6 */ 733 734static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 735 736#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 737#define SNP(buf) buf, sizeof(buf) 738 739/* 740 * We enter here when we have a rule with O_LOG. 741 * XXX this function alone takes about 2Kbytes of code! 742 */ 743static void 744ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 745 struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg) 746{ 747 struct ether_header *eh = args->eh; 748 char *action; 749 int limit_reached = 0; 750 char action2[40], proto[128], fragment[32]; 751 752 fragment[0] = '\0'; 753 proto[0] = '\0'; 754 755 if (f == NULL) { /* bogus pkt */ 756 if (verbose_limit != 0 && norule_counter >= verbose_limit) 757 return; 758 norule_counter++; 759 if (norule_counter == verbose_limit) 760 limit_reached = verbose_limit; 761 action = "Refuse"; 762 } else { /* O_LOG is the first action, find the real one */ 763 ipfw_insn *cmd = ACTION_PTR(f); 764 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 765 766 if (l->max_log != 0 && l->log_left == 0) 767 return; 768 l->log_left--; 769 if (l->log_left == 0) 770 limit_reached = l->max_log; 771 cmd += F_LEN(cmd); /* point to first action */ 772 if (cmd->opcode == O_ALTQ) { 773 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 774 775 snprintf(SNPARGS(action2, 0), "Altq %d", 776 altq->qid); 777 cmd += F_LEN(cmd); 778 } 779 if (cmd->opcode == O_PROB) 780 cmd += F_LEN(cmd); 781 782 if (cmd->opcode == O_TAG) 783 cmd += F_LEN(cmd); 784 785 action = action2; 786 switch (cmd->opcode) { 787 case O_DENY: 788 action = "Deny"; 789 break; 790 791 case O_REJECT: 792 if (cmd->arg1==ICMP_REJECT_RST) 793 action = "Reset"; 794 else if (cmd->arg1==ICMP_UNREACH_HOST) 795 action = "Reject"; 796 else 797 snprintf(SNPARGS(action2, 0), "Unreach %d", 798 cmd->arg1); 799 break; 800 801 case O_UNREACH6: 802 if (cmd->arg1==ICMP6_UNREACH_RST) 803 action = "Reset"; 804 else 805 snprintf(SNPARGS(action2, 0), "Unreach %d", 806 cmd->arg1); 807 break; 808 809 case O_ACCEPT: 810 action = "Accept"; 811 break; 812 case O_COUNT: 813 action = "Count"; 814 break; 815 case O_DIVERT: 816 snprintf(SNPARGS(action2, 0), "Divert %d", 817 cmd->arg1); 818 break; 819 case O_TEE: 820 snprintf(SNPARGS(action2, 0), "Tee %d", 821 cmd->arg1); 822 break; 823 case O_SKIPTO: 824 snprintf(SNPARGS(action2, 0), "SkipTo %d", 825 cmd->arg1); 826 break; 827 case O_PIPE: 828 snprintf(SNPARGS(action2, 0), "Pipe %d", 829 cmd->arg1); 830 break; 831 case O_QUEUE: 832 snprintf(SNPARGS(action2, 0), "Queue %d", 833 cmd->arg1); 834 break; 835 case O_FORWARD_IP: { 836 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 837 int len; 838 struct in_addr dummyaddr; 839 if (sa->sa.sin_addr.s_addr == INADDR_ANY) 840 dummyaddr.s_addr = htonl(tablearg); 841 else 842 dummyaddr.s_addr = sa->sa.sin_addr.s_addr; 843 844 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 845 inet_ntoa(dummyaddr)); 846 847 if (sa->sa.sin_port) 848 snprintf(SNPARGS(action2, len), ":%d", 849 sa->sa.sin_port); 850 } 851 break; 852 case O_NETGRAPH: 853 snprintf(SNPARGS(action2, 0), "Netgraph %d", 854 cmd->arg1); 855 break; 856 case O_NGTEE: 857 snprintf(SNPARGS(action2, 0), "Ngtee %d", 858 cmd->arg1); 859 break; 860 default: 861 action = "UNKNOWN"; 862 break; 863 } 864 } 865 866 if (hlen == 0) { /* non-ip */ 867 snprintf(SNPARGS(proto, 0), "MAC"); 868 869 } else { 870 int len; 871 char src[48], dst[48]; 872 struct icmphdr *icmp; 873 struct tcphdr *tcp; 874 struct udphdr *udp; 875 /* Initialize to make compiler happy. */ 876 struct ip *ip = NULL; 877#ifdef INET6 878 struct ip6_hdr *ip6 = NULL; 879 struct icmp6_hdr *icmp6; 880#endif 881 src[0] = '\0'; 882 dst[0] = '\0'; 883#ifdef INET6 884 if (IS_IP6_FLOW_ID(&(args->f_id))) { 885 snprintf(src, sizeof(src), "[%s]", 886 ip6_sprintf(&args->f_id.src_ip6)); 887 snprintf(dst, sizeof(dst), "[%s]", 888 ip6_sprintf(&args->f_id.dst_ip6)); 889 890 ip6 = (struct ip6_hdr *)mtod(m, struct ip6_hdr *); 891 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 892 udp = (struct udphdr *)(mtod(args->m, char *) + hlen); 893 } else 894#endif 895 { 896 ip = mtod(m, struct ip *); 897 tcp = L3HDR(struct tcphdr, ip); 898 udp = L3HDR(struct udphdr, ip); 899 900 inet_ntoa_r(ip->ip_src, src); 901 inet_ntoa_r(ip->ip_dst, dst); 902 } 903 904 switch (args->f_id.proto) { 905 case IPPROTO_TCP: 906 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 907 if (offset == 0) 908 snprintf(SNPARGS(proto, len), ":%d %s:%d", 909 ntohs(tcp->th_sport), 910 dst, 911 ntohs(tcp->th_dport)); 912 else 913 snprintf(SNPARGS(proto, len), " %s", dst); 914 break; 915 916 case IPPROTO_UDP: 917 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 918 if (offset == 0) 919 snprintf(SNPARGS(proto, len), ":%d %s:%d", 920 ntohs(udp->uh_sport), 921 dst, 922 ntohs(udp->uh_dport)); 923 else 924 snprintf(SNPARGS(proto, len), " %s", dst); 925 break; 926 927 case IPPROTO_ICMP: 928 icmp = L3HDR(struct icmphdr, ip); 929 if (offset == 0) 930 len = snprintf(SNPARGS(proto, 0), 931 "ICMP:%u.%u ", 932 icmp->icmp_type, icmp->icmp_code); 933 else 934 len = snprintf(SNPARGS(proto, 0), "ICMP "); 935 len += snprintf(SNPARGS(proto, len), "%s", src); 936 snprintf(SNPARGS(proto, len), " %s", dst); 937 break; 938#ifdef INET6 939 case IPPROTO_ICMPV6: 940 icmp6 = (struct icmp6_hdr *)(mtod(args->m, char *) + hlen); 941 if (offset == 0) 942 len = snprintf(SNPARGS(proto, 0), 943 "ICMPv6:%u.%u ", 944 icmp6->icmp6_type, icmp6->icmp6_code); 945 else 946 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 947 len += snprintf(SNPARGS(proto, len), "%s", src); 948 snprintf(SNPARGS(proto, len), " %s", dst); 949 break; 950#endif 951 default: 952 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 953 args->f_id.proto, src); 954 snprintf(SNPARGS(proto, len), " %s", dst); 955 break; 956 } 957 958#ifdef INET6 959 if (IS_IP6_FLOW_ID(&(args->f_id))) { 960 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 961 snprintf(SNPARGS(fragment, 0), 962 " (frag %08x:%d@%d%s)", 963 args->f_id.frag_id6, 964 ntohs(ip6->ip6_plen) - hlen, 965 ntohs(offset & IP6F_OFF_MASK) << 3, 966 (offset & IP6F_MORE_FRAG) ? "+" : ""); 967 } else 968#endif 969 { 970 int ip_off, ip_len; 971 if (eh != NULL) { /* layer 2 packets are as on the wire */ 972 ip_off = ntohs(ip->ip_off); 973 ip_len = ntohs(ip->ip_len); 974 } else { 975 ip_off = ip->ip_off; 976 ip_len = ip->ip_len; 977 } 978 if (ip_off & (IP_MF | IP_OFFMASK)) 979 snprintf(SNPARGS(fragment, 0), 980 " (frag %d:%d@%d%s)", 981 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 982 offset << 3, 983 (ip_off & IP_MF) ? "+" : ""); 984 } 985 } 986 if (oif || m->m_pkthdr.rcvif) 987 log(LOG_SECURITY | LOG_INFO, 988 "ipfw: %d %s %s %s via %s%s\n", 989 f ? f->rulenum : -1, 990 action, proto, oif ? "out" : "in", 991 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 992 fragment); 993 else 994 log(LOG_SECURITY | LOG_INFO, 995 "ipfw: %d %s %s [no if info]%s\n", 996 f ? f->rulenum : -1, 997 action, proto, fragment); 998 if (limit_reached) 999 log(LOG_SECURITY | LOG_NOTICE, 1000 "ipfw: limit %d reached on entry %d\n", 1001 limit_reached, f ? f->rulenum : -1); 1002} 1003 1004/* 1005 * IMPORTANT: the hash function for dynamic rules must be commutative 1006 * in source and destination (ip,port), because rules are bidirectional 1007 * and we want to find both in the same bucket. 1008 */ 1009static __inline int 1010hash_packet(struct ipfw_flow_id *id) 1011{ 1012 u_int32_t i; 1013 1014#ifdef INET6 1015 if (IS_IP6_FLOW_ID(id)) 1016 i = hash_packet6(id); 1017 else 1018#endif /* INET6 */ 1019 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1020 i &= (curr_dyn_buckets - 1); 1021 return i; 1022} 1023 1024/** 1025 * unlink a dynamic rule from a chain. prev is a pointer to 1026 * the previous one, q is a pointer to the rule to delete, 1027 * head is a pointer to the head of the queue. 1028 * Modifies q and potentially also head. 1029 */ 1030#define UNLINK_DYN_RULE(prev, head, q) { \ 1031 ipfw_dyn_rule *old_q = q; \ 1032 \ 1033 /* remove a refcount to the parent */ \ 1034 if (q->dyn_type == O_LIMIT) \ 1035 q->parent->count--; \ 1036 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1037 (q->id.src_ip), (q->id.src_port), \ 1038 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 1039 if (prev != NULL) \ 1040 prev->next = q = q->next; \ 1041 else \ 1042 head = q = q->next; \ 1043 dyn_count--; \ 1044 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1045 1046#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1047 1048/** 1049 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1050 * 1051 * If keep_me == NULL, rules are deleted even if not expired, 1052 * otherwise only expired rules are removed. 1053 * 1054 * The value of the second parameter is also used to point to identify 1055 * a rule we absolutely do not want to remove (e.g. because we are 1056 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1057 * rules). The pointer is only used for comparison, so any non-null 1058 * value will do. 1059 */ 1060static void 1061remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1062{ 1063 static u_int32_t last_remove = 0; 1064 1065#define FORCE (keep_me == NULL) 1066 1067 ipfw_dyn_rule *prev, *q; 1068 int i, pass = 0, max_pass = 0; 1069 1070 IPFW_DYN_LOCK_ASSERT(); 1071 1072 if (ipfw_dyn_v == NULL || dyn_count == 0) 1073 return; 1074 /* do not expire more than once per second, it is useless */ 1075 if (!FORCE && last_remove == time_uptime) 1076 return; 1077 last_remove = time_uptime; 1078 1079 /* 1080 * because O_LIMIT refer to parent rules, during the first pass only 1081 * remove child and mark any pending LIMIT_PARENT, and remove 1082 * them in a second pass. 1083 */ 1084next_pass: 1085 for (i = 0 ; i < curr_dyn_buckets ; i++) { 1086 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 1087 /* 1088 * Logic can become complex here, so we split tests. 1089 */ 1090 if (q == keep_me) 1091 goto next; 1092 if (rule != NULL && rule != q->rule) 1093 goto next; /* not the one we are looking for */ 1094 if (q->dyn_type == O_LIMIT_PARENT) { 1095 /* 1096 * handle parent in the second pass, 1097 * record we need one. 1098 */ 1099 max_pass = 1; 1100 if (pass == 0) 1101 goto next; 1102 if (FORCE && q->count != 0 ) { 1103 /* XXX should not happen! */ 1104 printf("ipfw: OUCH! cannot remove rule," 1105 " count %d\n", q->count); 1106 } 1107 } else { 1108 if (!FORCE && 1109 !TIME_LEQ( q->expire, time_uptime )) 1110 goto next; 1111 } 1112 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1113 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1114 continue; 1115 } 1116next: 1117 prev=q; 1118 q=q->next; 1119 } 1120 } 1121 if (pass++ < max_pass) 1122 goto next_pass; 1123} 1124 1125 1126/** 1127 * lookup a dynamic rule. 1128 */ 1129static ipfw_dyn_rule * 1130lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1131 struct tcphdr *tcp) 1132{ 1133 /* 1134 * stateful ipfw extensions. 1135 * Lookup into dynamic session queue 1136 */ 1137#define MATCH_REVERSE 0 1138#define MATCH_FORWARD 1 1139#define MATCH_NONE 2 1140#define MATCH_UNKNOWN 3 1141 int i, dir = MATCH_NONE; 1142 ipfw_dyn_rule *prev, *q=NULL; 1143 1144 IPFW_DYN_LOCK_ASSERT(); 1145 1146 if (ipfw_dyn_v == NULL) 1147 goto done; /* not found */ 1148 i = hash_packet( pkt ); 1149 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 1150 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1151 goto next; 1152 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1153 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1154 continue; 1155 } 1156 if (pkt->proto == q->id.proto && 1157 q->dyn_type != O_LIMIT_PARENT) { 1158 if (IS_IP6_FLOW_ID(pkt)) { 1159 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1160 &(q->id.src_ip6)) && 1161 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1162 &(q->id.dst_ip6)) && 1163 pkt->src_port == q->id.src_port && 1164 pkt->dst_port == q->id.dst_port ) { 1165 dir = MATCH_FORWARD; 1166 break; 1167 } 1168 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1169 &(q->id.dst_ip6)) && 1170 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1171 &(q->id.src_ip6)) && 1172 pkt->src_port == q->id.dst_port && 1173 pkt->dst_port == q->id.src_port ) { 1174 dir = MATCH_REVERSE; 1175 break; 1176 } 1177 } else { 1178 if (pkt->src_ip == q->id.src_ip && 1179 pkt->dst_ip == q->id.dst_ip && 1180 pkt->src_port == q->id.src_port && 1181 pkt->dst_port == q->id.dst_port ) { 1182 dir = MATCH_FORWARD; 1183 break; 1184 } 1185 if (pkt->src_ip == q->id.dst_ip && 1186 pkt->dst_ip == q->id.src_ip && 1187 pkt->src_port == q->id.dst_port && 1188 pkt->dst_port == q->id.src_port ) { 1189 dir = MATCH_REVERSE; 1190 break; 1191 } 1192 } 1193 } 1194next: 1195 prev = q; 1196 q = q->next; 1197 } 1198 if (q == NULL) 1199 goto done; /* q = NULL, not found */ 1200 1201 if ( prev != NULL) { /* found and not in front */ 1202 prev->next = q->next; 1203 q->next = ipfw_dyn_v[i]; 1204 ipfw_dyn_v[i] = q; 1205 } 1206 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1207 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1208 1209#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1210#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1211 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1212 switch (q->state) { 1213 case TH_SYN: /* opening */ 1214 q->expire = time_uptime + dyn_syn_lifetime; 1215 break; 1216 1217 case BOTH_SYN: /* move to established */ 1218 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1219 case BOTH_SYN | (TH_FIN << 8) : 1220 if (tcp) { 1221#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1222 u_int32_t ack = ntohl(tcp->th_ack); 1223 if (dir == MATCH_FORWARD) { 1224 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1225 q->ack_fwd = ack; 1226 else { /* ignore out-of-sequence */ 1227 break; 1228 } 1229 } else { 1230 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1231 q->ack_rev = ack; 1232 else { /* ignore out-of-sequence */ 1233 break; 1234 } 1235 } 1236 } 1237 q->expire = time_uptime + dyn_ack_lifetime; 1238 break; 1239 1240 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1241 if (dyn_fin_lifetime >= dyn_keepalive_period) 1242 dyn_fin_lifetime = dyn_keepalive_period - 1; 1243 q->expire = time_uptime + dyn_fin_lifetime; 1244 break; 1245 1246 default: 1247#if 0 1248 /* 1249 * reset or some invalid combination, but can also 1250 * occur if we use keep-state the wrong way. 1251 */ 1252 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1253 printf("invalid state: 0x%x\n", q->state); 1254#endif 1255 if (dyn_rst_lifetime >= dyn_keepalive_period) 1256 dyn_rst_lifetime = dyn_keepalive_period - 1; 1257 q->expire = time_uptime + dyn_rst_lifetime; 1258 break; 1259 } 1260 } else if (pkt->proto == IPPROTO_UDP) { 1261 q->expire = time_uptime + dyn_udp_lifetime; 1262 } else { 1263 /* other protocols */ 1264 q->expire = time_uptime + dyn_short_lifetime; 1265 } 1266done: 1267 if (match_direction) 1268 *match_direction = dir; 1269 return q; 1270} 1271 1272static ipfw_dyn_rule * 1273lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1274 struct tcphdr *tcp) 1275{ 1276 ipfw_dyn_rule *q; 1277 1278 IPFW_DYN_LOCK(); 1279 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1280 if (q == NULL) 1281 IPFW_DYN_UNLOCK(); 1282 /* NB: return table locked when q is not NULL */ 1283 return q; 1284} 1285 1286static void 1287realloc_dynamic_table(void) 1288{ 1289 IPFW_DYN_LOCK_ASSERT(); 1290 1291 /* 1292 * Try reallocation, make sure we have a power of 2 and do 1293 * not allow more than 64k entries. In case of overflow, 1294 * default to 1024. 1295 */ 1296 1297 if (dyn_buckets > 65536) 1298 dyn_buckets = 1024; 1299 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1300 dyn_buckets = curr_dyn_buckets; /* reset */ 1301 return; 1302 } 1303 curr_dyn_buckets = dyn_buckets; 1304 if (ipfw_dyn_v != NULL) 1305 free(ipfw_dyn_v, M_IPFW); 1306 for (;;) { 1307 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1308 M_IPFW, M_NOWAIT | M_ZERO); 1309 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 1310 break; 1311 curr_dyn_buckets /= 2; 1312 } 1313} 1314 1315/** 1316 * Install state of type 'type' for a dynamic session. 1317 * The hash table contains two type of rules: 1318 * - regular rules (O_KEEP_STATE) 1319 * - rules for sessions with limited number of sess per user 1320 * (O_LIMIT). When they are created, the parent is 1321 * increased by 1, and decreased on delete. In this case, 1322 * the third parameter is the parent rule and not the chain. 1323 * - "parent" rules for the above (O_LIMIT_PARENT). 1324 */ 1325static ipfw_dyn_rule * 1326add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1327{ 1328 ipfw_dyn_rule *r; 1329 int i; 1330 1331 IPFW_DYN_LOCK_ASSERT(); 1332 1333 if (ipfw_dyn_v == NULL || 1334 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1335 realloc_dynamic_table(); 1336 if (ipfw_dyn_v == NULL) 1337 return NULL; /* failed ! */ 1338 } 1339 i = hash_packet(id); 1340 1341 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1342 if (r == NULL) { 1343 printf ("ipfw: sorry cannot allocate state\n"); 1344 return NULL; 1345 } 1346 1347 /* increase refcount on parent, and set pointer */ 1348 if (dyn_type == O_LIMIT) { 1349 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1350 if ( parent->dyn_type != O_LIMIT_PARENT) 1351 panic("invalid parent"); 1352 parent->count++; 1353 r->parent = parent; 1354 rule = parent->rule; 1355 } 1356 1357 r->id = *id; 1358 r->expire = time_uptime + dyn_syn_lifetime; 1359 r->rule = rule; 1360 r->dyn_type = dyn_type; 1361 r->pcnt = r->bcnt = 0; 1362 r->count = 0; 1363 1364 r->bucket = i; 1365 r->next = ipfw_dyn_v[i]; 1366 ipfw_dyn_v[i] = r; 1367 dyn_count++; 1368 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1369 dyn_type, 1370 (r->id.src_ip), (r->id.src_port), 1371 (r->id.dst_ip), (r->id.dst_port), 1372 dyn_count ); ) 1373 return r; 1374} 1375 1376/** 1377 * lookup dynamic parent rule using pkt and rule as search keys. 1378 * If the lookup fails, then install one. 1379 */ 1380static ipfw_dyn_rule * 1381lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1382{ 1383 ipfw_dyn_rule *q; 1384 int i; 1385 1386 IPFW_DYN_LOCK_ASSERT(); 1387 1388 if (ipfw_dyn_v) { 1389 int is_v6 = IS_IP6_FLOW_ID(pkt); 1390 i = hash_packet( pkt ); 1391 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1392 if (q->dyn_type == O_LIMIT_PARENT && 1393 rule== q->rule && 1394 pkt->proto == q->id.proto && 1395 pkt->src_port == q->id.src_port && 1396 pkt->dst_port == q->id.dst_port && 1397 ( 1398 (is_v6 && 1399 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1400 &(q->id.src_ip6)) && 1401 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1402 &(q->id.dst_ip6))) || 1403 (!is_v6 && 1404 pkt->src_ip == q->id.src_ip && 1405 pkt->dst_ip == q->id.dst_ip) 1406 ) 1407 ) { 1408 q->expire = time_uptime + dyn_short_lifetime; 1409 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1410 return q; 1411 } 1412 } 1413 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1414} 1415 1416/** 1417 * Install dynamic state for rule type cmd->o.opcode 1418 * 1419 * Returns 1 (failure) if state is not installed because of errors or because 1420 * session limitations are enforced. 1421 */ 1422static int 1423install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1424 struct ip_fw_args *args, uint32_t tablearg) 1425{ 1426 static int last_log; 1427 ipfw_dyn_rule *q; 1428 struct in_addr da; 1429 char src[48], dst[48]; 1430 1431 src[0] = '\0'; 1432 dst[0] = '\0'; 1433 1434 DEB( 1435 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1436 __func__, cmd->o.opcode, 1437 (args->f_id.src_ip), (args->f_id.src_port), 1438 (args->f_id.dst_ip), (args->f_id.dst_port)); 1439 ) 1440 1441 IPFW_DYN_LOCK(); 1442 1443 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1444 1445 if (q != NULL) { /* should never occur */ 1446 if (last_log != time_uptime) { 1447 last_log = time_uptime; 1448 printf("ipfw: %s: entry already present, done\n", 1449 __func__); 1450 } 1451 IPFW_DYN_UNLOCK(); 1452 return (0); 1453 } 1454 1455 if (dyn_count >= dyn_max) 1456 /* Run out of slots, try to remove any expired rule. */ 1457 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1458 1459 if (dyn_count >= dyn_max) { 1460 if (last_log != time_uptime) { 1461 last_log = time_uptime; 1462 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1463 } 1464 IPFW_DYN_UNLOCK(); 1465 return (1); /* cannot install, notify caller */ 1466 } 1467 1468 switch (cmd->o.opcode) { 1469 case O_KEEP_STATE: /* bidir rule */ 1470 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1471 break; 1472 1473 case O_LIMIT: { /* limit number of sessions */ 1474 struct ipfw_flow_id id; 1475 ipfw_dyn_rule *parent; 1476 uint32_t conn_limit; 1477 uint16_t limit_mask = cmd->limit_mask; 1478 1479 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1480 tablearg : cmd->conn_limit; 1481 1482 DEB( 1483 if (cmd->conn_limit == IP_FW_TABLEARG) 1484 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1485 "(tablearg)\n", __func__, conn_limit); 1486 else 1487 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1488 __func__, conn_limit); 1489 ) 1490 1491 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1492 id.proto = args->f_id.proto; 1493 id.addr_type = args->f_id.addr_type; 1494 1495 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1496 if (limit_mask & DYN_SRC_ADDR) 1497 id.src_ip6 = args->f_id.src_ip6; 1498 if (limit_mask & DYN_DST_ADDR) 1499 id.dst_ip6 = args->f_id.dst_ip6; 1500 } else { 1501 if (limit_mask & DYN_SRC_ADDR) 1502 id.src_ip = args->f_id.src_ip; 1503 if (limit_mask & DYN_DST_ADDR) 1504 id.dst_ip = args->f_id.dst_ip; 1505 } 1506 if (limit_mask & DYN_SRC_PORT) 1507 id.src_port = args->f_id.src_port; 1508 if (limit_mask & DYN_DST_PORT) 1509 id.dst_port = args->f_id.dst_port; 1510 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1511 printf("ipfw: %s: add parent failed\n", __func__); 1512 IPFW_DYN_UNLOCK(); 1513 return (1); 1514 } 1515 1516 if (parent->count >= conn_limit) { 1517 /* See if we can remove some expired rule. */ 1518 remove_dyn_rule(rule, parent); 1519 if (parent->count >= conn_limit) { 1520 if (fw_verbose && last_log != time_uptime) { 1521 last_log = time_uptime; 1522#ifdef INET6 1523 /* 1524 * XXX IPv6 flows are not 1525 * supported yet. 1526 */ 1527 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1528 snprintf(src, sizeof(src), 1529 "[%s]", ip6_sprintf( 1530 &args->f_id.src_ip6)); 1531 snprintf(dst, sizeof(dst), 1532 "[%s]", ip6_sprintf( 1533 &args->f_id.dst_ip6)); 1534 } else 1535#endif 1536 { 1537 da.s_addr = 1538 htonl(args->f_id.src_ip); 1539 inet_ntoa_r(da, src); 1540 da.s_addr = 1541 htonl(args->f_id.dst_ip); 1542 inet_ntoa_r(da, dst); 1543 } 1544 log(LOG_SECURITY | LOG_DEBUG, 1545 "%s %s:%u -> %s:%u, %s\n", 1546 "drop session", 1547 src, (args->f_id.src_port), 1548 dst, (args->f_id.dst_port), 1549 "too many entries"); 1550 } 1551 IPFW_DYN_UNLOCK(); 1552 return (1); 1553 } 1554 } 1555 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1556 break; 1557 } 1558 default: 1559 printf("ipfw: %s: unknown dynamic rule type %u\n", 1560 __func__, cmd->o.opcode); 1561 IPFW_DYN_UNLOCK(); 1562 return (1); 1563 } 1564 1565 /* XXX just set lifetime */ 1566 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1567 1568 IPFW_DYN_UNLOCK(); 1569 return (0); 1570} 1571 1572/* 1573 * Generate a TCP packet, containing either a RST or a keepalive. 1574 * When flags & TH_RST, we are sending a RST packet, because of a 1575 * "reset" action matched the packet. 1576 * Otherwise we are sending a keepalive, and flags & TH_ 1577 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 1578 * so that MAC can label the reply appropriately. 1579 */ 1580static struct mbuf * 1581send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 1582 u_int32_t ack, int flags) 1583{ 1584 struct mbuf *m; 1585 struct ip *ip; 1586 struct tcphdr *tcp; 1587 1588 MGETHDR(m, M_DONTWAIT, MT_DATA); 1589 if (m == 0) 1590 return (NULL); 1591 m->m_pkthdr.rcvif = (struct ifnet *)0; 1592 1593#ifdef MAC 1594 if (replyto != NULL) 1595 mac_create_mbuf_netlayer(replyto, m); 1596 else 1597 mac_create_mbuf_from_firewall(m); 1598#else 1599 (void)replyto; /* don't warn about unused arg */ 1600#endif 1601 1602 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1603 m->m_data += max_linkhdr; 1604 1605 ip = mtod(m, struct ip *); 1606 bzero(ip, m->m_len); 1607 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1608 ip->ip_p = IPPROTO_TCP; 1609 tcp->th_off = 5; 1610 /* 1611 * Assume we are sending a RST (or a keepalive in the reverse 1612 * direction), swap src and destination addresses and ports. 1613 */ 1614 ip->ip_src.s_addr = htonl(id->dst_ip); 1615 ip->ip_dst.s_addr = htonl(id->src_ip); 1616 tcp->th_sport = htons(id->dst_port); 1617 tcp->th_dport = htons(id->src_port); 1618 if (flags & TH_RST) { /* we are sending a RST */ 1619 if (flags & TH_ACK) { 1620 tcp->th_seq = htonl(ack); 1621 tcp->th_ack = htonl(0); 1622 tcp->th_flags = TH_RST; 1623 } else { 1624 if (flags & TH_SYN) 1625 seq++; 1626 tcp->th_seq = htonl(0); 1627 tcp->th_ack = htonl(seq); 1628 tcp->th_flags = TH_RST | TH_ACK; 1629 } 1630 } else { 1631 /* 1632 * We are sending a keepalive. flags & TH_SYN determines 1633 * the direction, forward if set, reverse if clear. 1634 * NOTE: seq and ack are always assumed to be correct 1635 * as set by the caller. This may be confusing... 1636 */ 1637 if (flags & TH_SYN) { 1638 /* 1639 * we have to rewrite the correct addresses! 1640 */ 1641 ip->ip_dst.s_addr = htonl(id->dst_ip); 1642 ip->ip_src.s_addr = htonl(id->src_ip); 1643 tcp->th_dport = htons(id->dst_port); 1644 tcp->th_sport = htons(id->src_port); 1645 } 1646 tcp->th_seq = htonl(seq); 1647 tcp->th_ack = htonl(ack); 1648 tcp->th_flags = TH_ACK; 1649 } 1650 /* 1651 * set ip_len to the payload size so we can compute 1652 * the tcp checksum on the pseudoheader 1653 * XXX check this, could save a couple of words ? 1654 */ 1655 ip->ip_len = htons(sizeof(struct tcphdr)); 1656 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1657 /* 1658 * now fill fields left out earlier 1659 */ 1660 ip->ip_ttl = ip_defttl; 1661 ip->ip_len = m->m_pkthdr.len; 1662 m->m_flags |= M_SKIP_FIREWALL; 1663 return (m); 1664} 1665 1666/* 1667 * sends a reject message, consuming the mbuf passed as an argument. 1668 */ 1669static void 1670send_reject(struct ip_fw_args *args, int code, int ip_len) 1671{ 1672 1673 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1674 /* We need the IP header in host order for icmp_error(). */ 1675 if (args->eh != NULL) { 1676 struct ip *ip = mtod(args->m, struct ip *); 1677 ip->ip_len = ntohs(ip->ip_len); 1678 ip->ip_off = ntohs(ip->ip_off); 1679 } 1680 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1681 } else if (args->f_id.proto == IPPROTO_TCP) { 1682 struct tcphdr *const tcp = 1683 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1684 if ( (tcp->th_flags & TH_RST) == 0) { 1685 struct mbuf *m; 1686 m = send_pkt(args->m, &(args->f_id), 1687 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1688 tcp->th_flags | TH_RST); 1689 if (m != NULL) 1690 ip_output(m, NULL, NULL, 0, NULL, NULL); 1691 } 1692 m_freem(args->m); 1693 } else 1694 m_freem(args->m); 1695 args->m = NULL; 1696} 1697 1698/** 1699 * 1700 * Given an ip_fw *, lookup_next_rule will return a pointer 1701 * to the next rule, which can be either the jump 1702 * target (for skipto instructions) or the next one in the list (in 1703 * all other cases including a missing jump target). 1704 * The result is also written in the "next_rule" field of the rule. 1705 * Backward jumps are not allowed, so start looking from the next 1706 * rule... 1707 * 1708 * This never returns NULL -- in case we do not have an exact match, 1709 * the next rule is returned. When the ruleset is changed, 1710 * pointers are flushed so we are always correct. 1711 */ 1712 1713static struct ip_fw * 1714lookup_next_rule(struct ip_fw *me) 1715{ 1716 struct ip_fw *rule = NULL; 1717 ipfw_insn *cmd; 1718 1719 /* look for action, in case it is a skipto */ 1720 cmd = ACTION_PTR(me); 1721 if (cmd->opcode == O_LOG) 1722 cmd += F_LEN(cmd); 1723 if (cmd->opcode == O_ALTQ) 1724 cmd += F_LEN(cmd); 1725 if (cmd->opcode == O_TAG) 1726 cmd += F_LEN(cmd); 1727 if ( cmd->opcode == O_SKIPTO ) 1728 for (rule = me->next; rule ; rule = rule->next) 1729 if (rule->rulenum >= cmd->arg1) 1730 break; 1731 if (rule == NULL) /* failure or not a skipto */ 1732 rule = me->next; 1733 me->next_rule = rule; 1734 return rule; 1735} 1736 1737static int 1738add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1739 uint8_t mlen, uint32_t value) 1740{ 1741 struct radix_node_head *rnh; 1742 struct table_entry *ent; 1743 1744 if (tbl >= IPFW_TABLES_MAX) 1745 return (EINVAL); 1746 rnh = ch->tables[tbl]; 1747 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1748 if (ent == NULL) 1749 return (ENOMEM); 1750 ent->value = value; 1751 ent->addr.sin_len = ent->mask.sin_len = 8; 1752 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1753 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1754 IPFW_WLOCK(&layer3_chain); 1755 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1756 NULL) { 1757 IPFW_WUNLOCK(&layer3_chain); 1758 free(ent, M_IPFW_TBL); 1759 return (EEXIST); 1760 } 1761 IPFW_WUNLOCK(&layer3_chain); 1762 return (0); 1763} 1764 1765static int 1766del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1767 uint8_t mlen) 1768{ 1769 struct radix_node_head *rnh; 1770 struct table_entry *ent; 1771 struct sockaddr_in sa, mask; 1772 1773 if (tbl >= IPFW_TABLES_MAX) 1774 return (EINVAL); 1775 rnh = ch->tables[tbl]; 1776 sa.sin_len = mask.sin_len = 8; 1777 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1778 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1779 IPFW_WLOCK(ch); 1780 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1781 if (ent == NULL) { 1782 IPFW_WUNLOCK(ch); 1783 return (ESRCH); 1784 } 1785 IPFW_WUNLOCK(ch); 1786 free(ent, M_IPFW_TBL); 1787 return (0); 1788} 1789 1790static int 1791flush_table_entry(struct radix_node *rn, void *arg) 1792{ 1793 struct radix_node_head * const rnh = arg; 1794 struct table_entry *ent; 1795 1796 ent = (struct table_entry *) 1797 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1798 if (ent != NULL) 1799 free(ent, M_IPFW_TBL); 1800 return (0); 1801} 1802 1803static int 1804flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1805{ 1806 struct radix_node_head *rnh; 1807 1808 IPFW_WLOCK_ASSERT(ch); 1809 1810 if (tbl >= IPFW_TABLES_MAX) 1811 return (EINVAL); 1812 rnh = ch->tables[tbl]; 1813 KASSERT(rnh != NULL, ("NULL IPFW table")); 1814 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1815 return (0); 1816} 1817 1818static void 1819flush_tables(struct ip_fw_chain *ch) 1820{ 1821 uint16_t tbl; 1822 1823 IPFW_WLOCK_ASSERT(ch); 1824 1825 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1826 flush_table(ch, tbl); 1827} 1828 1829static int 1830init_tables(struct ip_fw_chain *ch) 1831{ 1832 int i; 1833 uint16_t j; 1834 1835 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1836 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1837 for (j = 0; j < i; j++) { 1838 (void) flush_table(ch, j); 1839 } 1840 return (ENOMEM); 1841 } 1842 } 1843 return (0); 1844} 1845 1846static int 1847lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1848 uint32_t *val) 1849{ 1850 struct radix_node_head *rnh; 1851 struct table_entry *ent; 1852 struct sockaddr_in sa; 1853 1854 if (tbl >= IPFW_TABLES_MAX) 1855 return (0); 1856 rnh = ch->tables[tbl]; 1857 sa.sin_len = 8; 1858 sa.sin_addr.s_addr = addr; 1859 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1860 if (ent != NULL) { 1861 *val = ent->value; 1862 return (1); 1863 } 1864 return (0); 1865} 1866 1867static int 1868count_table_entry(struct radix_node *rn, void *arg) 1869{ 1870 u_int32_t * const cnt = arg; 1871 1872 (*cnt)++; 1873 return (0); 1874} 1875 1876static int 1877count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1878{ 1879 struct radix_node_head *rnh; 1880 1881 if (tbl >= IPFW_TABLES_MAX) 1882 return (EINVAL); 1883 rnh = ch->tables[tbl]; 1884 *cnt = 0; 1885 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1886 return (0); 1887} 1888 1889static int 1890dump_table_entry(struct radix_node *rn, void *arg) 1891{ 1892 struct table_entry * const n = (struct table_entry *)rn; 1893 ipfw_table * const tbl = arg; 1894 ipfw_table_entry *ent; 1895 1896 if (tbl->cnt == tbl->size) 1897 return (1); 1898 ent = &tbl->ent[tbl->cnt]; 1899 ent->tbl = tbl->tbl; 1900 if (in_nullhost(n->mask.sin_addr)) 1901 ent->masklen = 0; 1902 else 1903 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1904 ent->addr = n->addr.sin_addr.s_addr; 1905 ent->value = n->value; 1906 tbl->cnt++; 1907 return (0); 1908} 1909 1910static int 1911dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1912{ 1913 struct radix_node_head *rnh; 1914 1915 if (tbl->tbl >= IPFW_TABLES_MAX) 1916 return (EINVAL); 1917 rnh = ch->tables[tbl->tbl]; 1918 tbl->cnt = 0; 1919 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1920 return (0); 1921} 1922 1923static void 1924fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1925{ 1926 struct ucred *cr; 1927 1928 if (inp->inp_socket != NULL) { 1929 cr = inp->inp_socket->so_cred; 1930 ugp->fw_prid = jailed(cr) ? 1931 cr->cr_prison->pr_id : -1; 1932 ugp->fw_uid = cr->cr_uid; 1933 ugp->fw_ngroups = cr->cr_ngroups; 1934 bcopy(cr->cr_groups, ugp->fw_groups, 1935 sizeof(ugp->fw_groups)); 1936 } 1937} 1938 1939static int 1940check_uidgid(ipfw_insn_u32 *insn, 1941 int proto, struct ifnet *oif, 1942 struct in_addr dst_ip, u_int16_t dst_port, 1943 struct in_addr src_ip, u_int16_t src_port, 1944 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 1945{ 1946 struct inpcbinfo *pi; 1947 int wildcard; 1948 struct inpcb *pcb; 1949 int match; 1950 gid_t *gp; 1951 1952 /* 1953 * Check to see if the UDP or TCP stack supplied us with 1954 * the PCB. If so, rather then holding a lock and looking 1955 * up the PCB, we can use the one that was supplied. 1956 */ 1957 if (inp && *lookup == 0) { 1958 INP_LOCK_ASSERT(inp); 1959 if (inp->inp_socket != NULL) { 1960 fill_ugid_cache(inp, ugp); 1961 *lookup = 1; 1962 } 1963 } 1964 /* 1965 * If we have already been here and the packet has no 1966 * PCB entry associated with it, then we can safely 1967 * assume that this is a no match. 1968 */ 1969 if (*lookup == -1) 1970 return (0); 1971 if (proto == IPPROTO_TCP) { 1972 wildcard = 0; 1973 pi = &tcbinfo; 1974 } else if (proto == IPPROTO_UDP) { 1975 wildcard = INPLOOKUP_WILDCARD; 1976 pi = &udbinfo; 1977 } else 1978 return 0; 1979 match = 0; 1980 if (*lookup == 0) { 1981 INP_INFO_RLOCK(pi); 1982 pcb = (oif) ? 1983 in_pcblookup_hash(pi, 1984 dst_ip, htons(dst_port), 1985 src_ip, htons(src_port), 1986 wildcard, oif) : 1987 in_pcblookup_hash(pi, 1988 src_ip, htons(src_port), 1989 dst_ip, htons(dst_port), 1990 wildcard, NULL); 1991 if (pcb != NULL) { 1992 INP_LOCK(pcb); 1993 if (pcb->inp_socket != NULL) { 1994 fill_ugid_cache(pcb, ugp); 1995 *lookup = 1; 1996 } 1997 INP_UNLOCK(pcb); 1998 } 1999 INP_INFO_RUNLOCK(pi); 2000 if (*lookup == 0) { 2001 /* 2002 * If the lookup did not yield any results, there 2003 * is no sense in coming back and trying again. So 2004 * we can set lookup to -1 and ensure that we wont 2005 * bother the pcb system again. 2006 */ 2007 *lookup = -1; 2008 return (0); 2009 } 2010 } 2011 if (insn->o.opcode == O_UID) 2012 match = (ugp->fw_uid == (uid_t)insn->d[0]); 2013 else if (insn->o.opcode == O_GID) { 2014 for (gp = ugp->fw_groups; 2015 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 2016 if (*gp == (gid_t)insn->d[0]) { 2017 match = 1; 2018 break; 2019 } 2020 } else if (insn->o.opcode == O_JAIL) 2021 match = (ugp->fw_prid == (int)insn->d[0]); 2022 return match; 2023} 2024 2025/* 2026 * The main check routine for the firewall. 2027 * 2028 * All arguments are in args so we can modify them and return them 2029 * back to the caller. 2030 * 2031 * Parameters: 2032 * 2033 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 2034 * Starts with the IP header. 2035 * args->eh (in) Mac header if present, or NULL for layer3 packet. 2036 * args->oif Outgoing interface, or NULL if packet is incoming. 2037 * The incoming interface is in the mbuf. (in) 2038 * args->divert_rule (in/out) 2039 * Skip up to the first rule past this rule number; 2040 * upon return, non-zero port number for divert or tee. 2041 * 2042 * args->rule Pointer to the last matching rule (in/out) 2043 * args->next_hop Socket we are forwarding to (out). 2044 * args->f_id Addresses grabbed from the packet (out) 2045 * args->cookie a cookie depending on rule action 2046 * 2047 * Return value: 2048 * 2049 * IP_FW_PASS the packet must be accepted 2050 * IP_FW_DENY the packet must be dropped 2051 * IP_FW_DIVERT divert packet, port in m_tag 2052 * IP_FW_TEE tee packet, port in m_tag 2053 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2054 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2055 * 2056 */ 2057 2058int 2059ipfw_chk(struct ip_fw_args *args) 2060{ 2061 /* 2062 * Local variables hold state during the processing of a packet. 2063 * 2064 * IMPORTANT NOTE: to speed up the processing of rules, there 2065 * are some assumption on the values of the variables, which 2066 * are documented here. Should you change them, please check 2067 * the implementation of the various instructions to make sure 2068 * that they still work. 2069 * 2070 * args->eh The MAC header. It is non-null for a layer2 2071 * packet, it is NULL for a layer-3 packet. 2072 * 2073 * m | args->m Pointer to the mbuf, as received from the caller. 2074 * It may change if ipfw_chk() does an m_pullup, or if it 2075 * consumes the packet because it calls send_reject(). 2076 * XXX This has to change, so that ipfw_chk() never modifies 2077 * or consumes the buffer. 2078 * ip is simply an alias of the value of m, and it is kept 2079 * in sync with it (the packet is supposed to start with 2080 * the ip header). 2081 */ 2082 struct mbuf *m = args->m; 2083 struct ip *ip = mtod(m, struct ip *); 2084 2085 /* 2086 * For rules which contain uid/gid or jail constraints, cache 2087 * a copy of the users credentials after the pcb lookup has been 2088 * executed. This will speed up the processing of rules with 2089 * these types of constraints, as well as decrease contention 2090 * on pcb related locks. 2091 */ 2092 struct ip_fw_ugid fw_ugid_cache; 2093 int ugid_lookup = 0; 2094 2095 /* 2096 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2097 * associated with a packet input on a divert socket. This 2098 * will allow to distinguish traffic and its direction when 2099 * it originates from a divert socket. 2100 */ 2101 u_int divinput_flags = 0; 2102 2103 /* 2104 * oif | args->oif If NULL, ipfw_chk has been called on the 2105 * inbound path (ether_input, ip_input). 2106 * If non-NULL, ipfw_chk has been called on the outbound path 2107 * (ether_output, ip_output). 2108 */ 2109 struct ifnet *oif = args->oif; 2110 2111 struct ip_fw *f = NULL; /* matching rule */ 2112 int retval = 0; 2113 2114 /* 2115 * hlen The length of the IP header. 2116 */ 2117 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2118 2119 /* 2120 * offset The offset of a fragment. offset != 0 means that 2121 * we have a fragment at this offset of an IPv4 packet. 2122 * offset == 0 means that (if this is an IPv4 packet) 2123 * this is the first or only fragment. 2124 * For IPv6 offset == 0 means there is no Fragment Header. 2125 * If offset != 0 for IPv6 always use correct mask to 2126 * get the correct offset because we add IP6F_MORE_FRAG 2127 * to be able to dectect the first fragment which would 2128 * otherwise have offset = 0. 2129 */ 2130 u_short offset = 0; 2131 2132 /* 2133 * Local copies of addresses. They are only valid if we have 2134 * an IP packet. 2135 * 2136 * proto The protocol. Set to 0 for non-ip packets, 2137 * or to the protocol read from the packet otherwise. 2138 * proto != 0 means that we have an IPv4 packet. 2139 * 2140 * src_port, dst_port port numbers, in HOST format. Only 2141 * valid for TCP and UDP packets. 2142 * 2143 * src_ip, dst_ip ip addresses, in NETWORK format. 2144 * Only valid for IPv4 packets. 2145 */ 2146 u_int8_t proto; 2147 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2148 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2149 u_int16_t ip_len=0; 2150 int pktlen; 2151 2152 /* 2153 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2154 * MATCH_NONE when checked and not matched (q = NULL), 2155 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2156 */ 2157 int dyn_dir = MATCH_UNKNOWN; 2158 ipfw_dyn_rule *q = NULL; 2159 struct ip_fw_chain *chain = &layer3_chain; 2160 struct m_tag *mtag; 2161 2162 /* 2163 * We store in ulp a pointer to the upper layer protocol header. 2164 * In the ipv4 case this is easy to determine from the header, 2165 * but for ipv6 we might have some additional headers in the middle. 2166 * ulp is NULL if not found. 2167 */ 2168 void *ulp = NULL; /* upper layer protocol pointer. */ 2169 /* XXX ipv6 variables */ 2170 int is_ipv6 = 0; 2171 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2172 /* end of ipv6 variables */ 2173 int is_ipv4 = 0; 2174 2175 if (m->m_flags & M_SKIP_FIREWALL) 2176 return (IP_FW_PASS); /* accept */ 2177 2178 pktlen = m->m_pkthdr.len; 2179 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2180 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2181 2182/* 2183 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2184 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2185 * pointer might become stale after other pullups (but we never use it 2186 * this way). 2187 */ 2188#define PULLUP_TO(len, p, T) \ 2189do { \ 2190 int x = (len) + sizeof(T); \ 2191 if ((m)->m_len < x) { \ 2192 args->m = m = m_pullup(m, x); \ 2193 if (m == NULL) \ 2194 goto pullup_failed; \ 2195 } \ 2196 p = (mtod(m, char *) + (len)); \ 2197} while (0) 2198 2199 /* Identify IP packets and fill up variables. */ 2200 if (pktlen >= sizeof(struct ip6_hdr) && 2201 (args->eh == NULL || ntohs(args->eh->ether_type)==ETHERTYPE_IPV6) && 2202 mtod(m, struct ip *)->ip_v == 6) { 2203 is_ipv6 = 1; 2204 args->f_id.addr_type = 6; 2205 hlen = sizeof(struct ip6_hdr); 2206 proto = mtod(m, struct ip6_hdr *)->ip6_nxt; 2207 2208 /* Search extension headers to find upper layer protocols */ 2209 while (ulp == NULL) { 2210 switch (proto) { 2211 case IPPROTO_ICMPV6: 2212 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2213 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2214 break; 2215 2216 case IPPROTO_TCP: 2217 PULLUP_TO(hlen, ulp, struct tcphdr); 2218 dst_port = TCP(ulp)->th_dport; 2219 src_port = TCP(ulp)->th_sport; 2220 args->f_id.flags = TCP(ulp)->th_flags; 2221 break; 2222 2223 case IPPROTO_UDP: 2224 PULLUP_TO(hlen, ulp, struct udphdr); 2225 dst_port = UDP(ulp)->uh_dport; 2226 src_port = UDP(ulp)->uh_sport; 2227 break; 2228 2229 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2230 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2231 ext_hd |= EXT_HOPOPTS; 2232 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2233 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2234 ulp = NULL; 2235 break; 2236 2237 case IPPROTO_ROUTING: /* RFC 2460 */ 2238 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2239 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2240 case 0: 2241 break; 2242 default: 2243 printf("IPFW2: IPV6 - Unknown Routing " 2244 "Header type(%d)\n", 2245 ((struct ip6_rthdr *)ulp)->ip6r_type); 2246 if (fw_deny_unknown_exthdrs) 2247 return (IP_FW_DENY); 2248 break; 2249 } 2250 ext_hd |= EXT_ROUTING; 2251 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2252 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2253 ulp = NULL; 2254 break; 2255 2256 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2257 PULLUP_TO(hlen, ulp, struct ip6_frag); 2258 ext_hd |= EXT_FRAGMENT; 2259 hlen += sizeof (struct ip6_frag); 2260 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2261 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2262 IP6F_OFF_MASK; 2263 /* Add IP6F_MORE_FRAG for offset of first 2264 * fragment to be != 0. */ 2265 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2266 IP6F_MORE_FRAG; 2267 if (offset == 0) { 2268 printf("IPFW2: IPV6 - Invalid Fragment " 2269 "Header\n"); 2270 if (fw_deny_unknown_exthdrs) 2271 return (IP_FW_DENY); 2272 break; 2273 } 2274 args->f_id.frag_id6 = 2275 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2276 ulp = NULL; 2277 break; 2278 2279 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2280 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2281 ext_hd |= EXT_DSTOPTS; 2282 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2283 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2284 ulp = NULL; 2285 break; 2286 2287 case IPPROTO_AH: /* RFC 2402 */ 2288 PULLUP_TO(hlen, ulp, struct ip6_ext); 2289 ext_hd |= EXT_AH; 2290 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2291 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2292 ulp = NULL; 2293 break; 2294 2295 case IPPROTO_ESP: /* RFC 2406 */ 2296 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2297 /* Anything past Seq# is variable length and 2298 * data past this ext. header is encrypted. */ 2299 ext_hd |= EXT_ESP; 2300 break; 2301 2302 case IPPROTO_NONE: /* RFC 2460 */ 2303 PULLUP_TO(hlen, ulp, struct ip6_ext); 2304 /* Packet ends here. if ip6e_len!=0 octets 2305 * must be ignored. */ 2306 break; 2307 2308 case IPPROTO_OSPFIGP: 2309 /* XXX OSPF header check? */ 2310 PULLUP_TO(hlen, ulp, struct ip6_ext); 2311 break; 2312 2313 case IPPROTO_PIM: 2314 /* XXX PIM header check? */ 2315 PULLUP_TO(hlen, ulp, struct pim); 2316 break; 2317 2318 case IPPROTO_CARP: 2319 PULLUP_TO(hlen, ulp, struct carp_header); 2320 if (((struct carp_header *)ulp)->carp_version != 2321 CARP_VERSION) 2322 return (IP_FW_DENY); 2323 if (((struct carp_header *)ulp)->carp_type != 2324 CARP_ADVERTISEMENT) 2325 return (IP_FW_DENY); 2326 break; 2327 2328 case IPPROTO_IPV6: /* RFC 2893 */ 2329 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2330 break; 2331 2332 case IPPROTO_IPV4: /* RFC 2893 */ 2333 PULLUP_TO(hlen, ulp, struct ip); 2334 break; 2335 2336 default: 2337 printf("IPFW2: IPV6 - Unknown Extension " 2338 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2339 if (fw_deny_unknown_exthdrs) 2340 return (IP_FW_DENY); 2341 PULLUP_TO(hlen, ulp, struct ip6_ext); 2342 break; 2343 } /*switch */ 2344 } 2345 args->f_id.src_ip6 = mtod(m,struct ip6_hdr *)->ip6_src; 2346 args->f_id.dst_ip6 = mtod(m,struct ip6_hdr *)->ip6_dst; 2347 args->f_id.src_ip = 0; 2348 args->f_id.dst_ip = 0; 2349 args->f_id.flow_id6 = ntohl(mtod(m, struct ip6_hdr *)->ip6_flow); 2350 } else if (pktlen >= sizeof(struct ip) && 2351 (args->eh == NULL || ntohs(args->eh->ether_type) == ETHERTYPE_IP) && 2352 mtod(m, struct ip *)->ip_v == 4) { 2353 is_ipv4 = 1; 2354 ip = mtod(m, struct ip *); 2355 hlen = ip->ip_hl << 2; 2356 args->f_id.addr_type = 4; 2357 2358 /* 2359 * Collect parameters into local variables for faster matching. 2360 */ 2361 proto = ip->ip_p; 2362 src_ip = ip->ip_src; 2363 dst_ip = ip->ip_dst; 2364 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2365 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2366 ip_len = ntohs(ip->ip_len); 2367 } else { 2368 offset = ip->ip_off & IP_OFFMASK; 2369 ip_len = ip->ip_len; 2370 } 2371 pktlen = ip_len < pktlen ? ip_len : pktlen; 2372 2373 if (offset == 0) { 2374 switch (proto) { 2375 case IPPROTO_TCP: 2376 PULLUP_TO(hlen, ulp, struct tcphdr); 2377 dst_port = TCP(ulp)->th_dport; 2378 src_port = TCP(ulp)->th_sport; 2379 args->f_id.flags = TCP(ulp)->th_flags; 2380 break; 2381 2382 case IPPROTO_UDP: 2383 PULLUP_TO(hlen, ulp, struct udphdr); 2384 dst_port = UDP(ulp)->uh_dport; 2385 src_port = UDP(ulp)->uh_sport; 2386 break; 2387 2388 case IPPROTO_ICMP: 2389 PULLUP_TO(hlen, ulp, struct icmphdr); 2390 args->f_id.flags = ICMP(ulp)->icmp_type; 2391 break; 2392 2393 default: 2394 break; 2395 } 2396 } 2397 2398 args->f_id.src_ip = ntohl(src_ip.s_addr); 2399 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2400 } 2401#undef PULLUP_TO 2402 if (proto) { /* we may have port numbers, store them */ 2403 args->f_id.proto = proto; 2404 args->f_id.src_port = src_port = ntohs(src_port); 2405 args->f_id.dst_port = dst_port = ntohs(dst_port); 2406 } 2407 2408 IPFW_RLOCK(chain); 2409 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2410 if (args->rule) { 2411 /* 2412 * Packet has already been tagged. Look for the next rule 2413 * to restart processing. 2414 * 2415 * If fw_one_pass != 0 then just accept it. 2416 * XXX should not happen here, but optimized out in 2417 * the caller. 2418 */ 2419 if (fw_one_pass) { 2420 IPFW_RUNLOCK(chain); 2421 return (IP_FW_PASS); 2422 } 2423 2424 f = args->rule->next_rule; 2425 if (f == NULL) 2426 f = lookup_next_rule(args->rule); 2427 } else { 2428 /* 2429 * Find the starting rule. It can be either the first 2430 * one, or the one after divert_rule if asked so. 2431 */ 2432 int skipto = mtag ? divert_cookie(mtag) : 0; 2433 2434 f = chain->rules; 2435 if (args->eh == NULL && skipto != 0) { 2436 if (skipto >= IPFW_DEFAULT_RULE) { 2437 IPFW_RUNLOCK(chain); 2438 return (IP_FW_DENY); /* invalid */ 2439 } 2440 while (f && f->rulenum <= skipto) 2441 f = f->next; 2442 if (f == NULL) { /* drop packet */ 2443 IPFW_RUNLOCK(chain); 2444 return (IP_FW_DENY); 2445 } 2446 } 2447 } 2448 /* reset divert rule to avoid confusion later */ 2449 if (mtag) { 2450 divinput_flags = divert_info(mtag) & 2451 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2452 m_tag_delete(m, mtag); 2453 } 2454 2455 /* 2456 * Now scan the rules, and parse microinstructions for each rule. 2457 */ 2458 for (; f; f = f->next) { 2459 ipfw_insn *cmd; 2460 uint32_t tablearg = 0; 2461 int l, cmdlen, skip_or; /* skip rest of OR block */ 2462 2463again: 2464 if (set_disable & (1 << f->set) ) 2465 continue; 2466 2467 skip_or = 0; 2468 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2469 l -= cmdlen, cmd += cmdlen) { 2470 int match; 2471 2472 /* 2473 * check_body is a jump target used when we find a 2474 * CHECK_STATE, and need to jump to the body of 2475 * the target rule. 2476 */ 2477 2478check_body: 2479 cmdlen = F_LEN(cmd); 2480 /* 2481 * An OR block (insn_1 || .. || insn_n) has the 2482 * F_OR bit set in all but the last instruction. 2483 * The first match will set "skip_or", and cause 2484 * the following instructions to be skipped until 2485 * past the one with the F_OR bit clear. 2486 */ 2487 if (skip_or) { /* skip this instruction */ 2488 if ((cmd->len & F_OR) == 0) 2489 skip_or = 0; /* next one is good */ 2490 continue; 2491 } 2492 match = 0; /* set to 1 if we succeed */ 2493 2494 switch (cmd->opcode) { 2495 /* 2496 * The first set of opcodes compares the packet's 2497 * fields with some pattern, setting 'match' if a 2498 * match is found. At the end of the loop there is 2499 * logic to deal with F_NOT and F_OR flags associated 2500 * with the opcode. 2501 */ 2502 case O_NOP: 2503 match = 1; 2504 break; 2505 2506 case O_FORWARD_MAC: 2507 printf("ipfw: opcode %d unimplemented\n", 2508 cmd->opcode); 2509 break; 2510 2511 case O_GID: 2512 case O_UID: 2513 case O_JAIL: 2514 /* 2515 * We only check offset == 0 && proto != 0, 2516 * as this ensures that we have a 2517 * packet with the ports info. 2518 */ 2519 if (offset!=0) 2520 break; 2521 if (is_ipv6) /* XXX to be fixed later */ 2522 break; 2523 if (proto == IPPROTO_TCP || 2524 proto == IPPROTO_UDP) 2525 match = check_uidgid( 2526 (ipfw_insn_u32 *)cmd, 2527 proto, oif, 2528 dst_ip, dst_port, 2529 src_ip, src_port, &fw_ugid_cache, 2530 &ugid_lookup, args->inp); 2531 break; 2532 2533 case O_RECV: 2534 match = iface_match(m->m_pkthdr.rcvif, 2535 (ipfw_insn_if *)cmd); 2536 break; 2537 2538 case O_XMIT: 2539 match = iface_match(oif, (ipfw_insn_if *)cmd); 2540 break; 2541 2542 case O_VIA: 2543 match = iface_match(oif ? oif : 2544 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2545 break; 2546 2547 case O_MACADDR2: 2548 if (args->eh != NULL) { /* have MAC header */ 2549 u_int32_t *want = (u_int32_t *) 2550 ((ipfw_insn_mac *)cmd)->addr; 2551 u_int32_t *mask = (u_int32_t *) 2552 ((ipfw_insn_mac *)cmd)->mask; 2553 u_int32_t *hdr = (u_int32_t *)args->eh; 2554 2555 match = 2556 ( want[0] == (hdr[0] & mask[0]) && 2557 want[1] == (hdr[1] & mask[1]) && 2558 want[2] == (hdr[2] & mask[2]) ); 2559 } 2560 break; 2561 2562 case O_MAC_TYPE: 2563 if (args->eh != NULL) { 2564 u_int16_t t = 2565 ntohs(args->eh->ether_type); 2566 u_int16_t *p = 2567 ((ipfw_insn_u16 *)cmd)->ports; 2568 int i; 2569 2570 for (i = cmdlen - 1; !match && i>0; 2571 i--, p += 2) 2572 match = (t>=p[0] && t<=p[1]); 2573 } 2574 break; 2575 2576 case O_FRAG: 2577 match = (offset != 0); 2578 break; 2579 2580 case O_IN: /* "out" is "not in" */ 2581 match = (oif == NULL); 2582 break; 2583 2584 case O_LAYER2: 2585 match = (args->eh != NULL); 2586 break; 2587 2588 case O_DIVERTED: 2589 match = (cmd->arg1 & 1 && divinput_flags & 2590 IP_FW_DIVERT_LOOPBACK_FLAG) || 2591 (cmd->arg1 & 2 && divinput_flags & 2592 IP_FW_DIVERT_OUTPUT_FLAG); 2593 break; 2594 2595 case O_PROTO: 2596 /* 2597 * We do not allow an arg of 0 so the 2598 * check of "proto" only suffices. 2599 */ 2600 match = (proto == cmd->arg1); 2601 break; 2602 2603 case O_IP_SRC: 2604 match = is_ipv4 && 2605 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2606 src_ip.s_addr); 2607 break; 2608 2609 case O_IP_SRC_LOOKUP: 2610 case O_IP_DST_LOOKUP: 2611 if (is_ipv4) { 2612 uint32_t a = 2613 (cmd->opcode == O_IP_DST_LOOKUP) ? 2614 dst_ip.s_addr : src_ip.s_addr; 2615 uint32_t v; 2616 2617 match = lookup_table(chain, cmd->arg1, a, 2618 &v); 2619 if (!match) 2620 break; 2621 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2622 match = 2623 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2624 else 2625 tablearg = v; 2626 } 2627 break; 2628 2629 case O_IP_SRC_MASK: 2630 case O_IP_DST_MASK: 2631 if (is_ipv4) { 2632 uint32_t a = 2633 (cmd->opcode == O_IP_DST_MASK) ? 2634 dst_ip.s_addr : src_ip.s_addr; 2635 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2636 int i = cmdlen-1; 2637 2638 for (; !match && i>0; i-= 2, p+= 2) 2639 match = (p[0] == (a & p[1])); 2640 } 2641 break; 2642 2643 case O_IP_SRC_ME: 2644 if (is_ipv4) { 2645 struct ifnet *tif; 2646 2647 INADDR_TO_IFP(src_ip, tif); 2648 match = (tif != NULL); 2649 } 2650 break; 2651 2652 case O_IP_DST_SET: 2653 case O_IP_SRC_SET: 2654 if (is_ipv4) { 2655 u_int32_t *d = (u_int32_t *)(cmd+1); 2656 u_int32_t addr = 2657 cmd->opcode == O_IP_DST_SET ? 2658 args->f_id.dst_ip : 2659 args->f_id.src_ip; 2660 2661 if (addr < d[0]) 2662 break; 2663 addr -= d[0]; /* subtract base */ 2664 match = (addr < cmd->arg1) && 2665 ( d[ 1 + (addr>>5)] & 2666 (1<<(addr & 0x1f)) ); 2667 } 2668 break; 2669 2670 case O_IP_DST: 2671 match = is_ipv4 && 2672 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2673 dst_ip.s_addr); 2674 break; 2675 2676 case O_IP_DST_ME: 2677 if (is_ipv4) { 2678 struct ifnet *tif; 2679 2680 INADDR_TO_IFP(dst_ip, tif); 2681 match = (tif != NULL); 2682 } 2683 break; 2684 2685 case O_IP_SRCPORT: 2686 case O_IP_DSTPORT: 2687 /* 2688 * offset == 0 && proto != 0 is enough 2689 * to guarantee that we have a 2690 * packet with port info. 2691 */ 2692 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2693 && offset == 0) { 2694 u_int16_t x = 2695 (cmd->opcode == O_IP_SRCPORT) ? 2696 src_port : dst_port ; 2697 u_int16_t *p = 2698 ((ipfw_insn_u16 *)cmd)->ports; 2699 int i; 2700 2701 for (i = cmdlen - 1; !match && i>0; 2702 i--, p += 2) 2703 match = (x>=p[0] && x<=p[1]); 2704 } 2705 break; 2706 2707 case O_ICMPTYPE: 2708 match = (offset == 0 && proto==IPPROTO_ICMP && 2709 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2710 break; 2711 2712#ifdef INET6 2713 case O_ICMP6TYPE: 2714 match = is_ipv6 && offset == 0 && 2715 proto==IPPROTO_ICMPV6 && 2716 icmp6type_match( 2717 ICMP6(ulp)->icmp6_type, 2718 (ipfw_insn_u32 *)cmd); 2719 break; 2720#endif /* INET6 */ 2721 2722 case O_IPOPT: 2723 match = (is_ipv4 && 2724 ipopts_match(mtod(m, struct ip *), cmd) ); 2725 break; 2726 2727 case O_IPVER: 2728 match = (is_ipv4 && 2729 cmd->arg1 == mtod(m, struct ip *)->ip_v); 2730 break; 2731 2732 case O_IPID: 2733 case O_IPLEN: 2734 case O_IPTTL: 2735 if (is_ipv4) { /* only for IP packets */ 2736 uint16_t x; 2737 uint16_t *p; 2738 int i; 2739 2740 if (cmd->opcode == O_IPLEN) 2741 x = ip_len; 2742 else if (cmd->opcode == O_IPTTL) 2743 x = mtod(m, struct ip *)->ip_ttl; 2744 else /* must be IPID */ 2745 x = ntohs(mtod(m, struct ip *)->ip_id); 2746 if (cmdlen == 1) { 2747 match = (cmd->arg1 == x); 2748 break; 2749 } 2750 /* otherwise we have ranges */ 2751 p = ((ipfw_insn_u16 *)cmd)->ports; 2752 i = cmdlen - 1; 2753 for (; !match && i>0; i--, p += 2) 2754 match = (x >= p[0] && x <= p[1]); 2755 } 2756 break; 2757 2758 case O_IPPRECEDENCE: 2759 match = (is_ipv4 && 2760 (cmd->arg1 == (mtod(m, struct ip *)->ip_tos & 0xe0)) ); 2761 break; 2762 2763 case O_IPTOS: 2764 match = (is_ipv4 && 2765 flags_match(cmd, mtod(m, struct ip *)->ip_tos)); 2766 break; 2767 2768 case O_TCPDATALEN: 2769 if (proto == IPPROTO_TCP && offset == 0) { 2770 struct tcphdr *tcp; 2771 uint16_t x; 2772 uint16_t *p; 2773 int i; 2774 2775 tcp = TCP(ulp); 2776 x = ip_len - 2777 ((ip->ip_hl + tcp->th_off) << 2); 2778 if (cmdlen == 1) { 2779 match = (cmd->arg1 == x); 2780 break; 2781 } 2782 /* otherwise we have ranges */ 2783 p = ((ipfw_insn_u16 *)cmd)->ports; 2784 i = cmdlen - 1; 2785 for (; !match && i>0; i--, p += 2) 2786 match = (x >= p[0] && x <= p[1]); 2787 } 2788 break; 2789 2790 case O_TCPFLAGS: 2791 match = (proto == IPPROTO_TCP && offset == 0 && 2792 flags_match(cmd, TCP(ulp)->th_flags)); 2793 break; 2794 2795 case O_TCPOPTS: 2796 match = (proto == IPPROTO_TCP && offset == 0 && 2797 tcpopts_match(TCP(ulp), cmd)); 2798 break; 2799 2800 case O_TCPSEQ: 2801 match = (proto == IPPROTO_TCP && offset == 0 && 2802 ((ipfw_insn_u32 *)cmd)->d[0] == 2803 TCP(ulp)->th_seq); 2804 break; 2805 2806 case O_TCPACK: 2807 match = (proto == IPPROTO_TCP && offset == 0 && 2808 ((ipfw_insn_u32 *)cmd)->d[0] == 2809 TCP(ulp)->th_ack); 2810 break; 2811 2812 case O_TCPWIN: 2813 match = (proto == IPPROTO_TCP && offset == 0 && 2814 cmd->arg1 == TCP(ulp)->th_win); 2815 break; 2816 2817 case O_ESTAB: 2818 /* reject packets which have SYN only */ 2819 /* XXX should i also check for TH_ACK ? */ 2820 match = (proto == IPPROTO_TCP && offset == 0 && 2821 (TCP(ulp)->th_flags & 2822 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2823 break; 2824 2825 case O_ALTQ: { 2826 struct altq_tag *at; 2827 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2828 2829 match = 1; 2830 mtag = m_tag_find(m, PACKET_TAG_PF_QID, NULL); 2831 if (mtag != NULL) 2832 break; 2833 mtag = m_tag_get(PACKET_TAG_PF_QID, 2834 sizeof(struct altq_tag), 2835 M_NOWAIT); 2836 if (mtag == NULL) { 2837 /* 2838 * Let the packet fall back to the 2839 * default ALTQ. 2840 */ 2841 break; 2842 } 2843 at = (struct altq_tag *)(mtag+1); 2844 at->qid = altq->qid; 2845 if (is_ipv4) 2846 at->af = AF_INET; 2847 else 2848 at->af = AF_LINK; 2849 at->hdr = ip; 2850 m_tag_prepend(m, mtag); 2851 break; 2852 } 2853 2854 case O_LOG: 2855 if (fw_verbose) 2856 ipfw_log(f, hlen, args, m, 2857 oif, offset, tablearg); 2858 match = 1; 2859 break; 2860 2861 case O_PROB: 2862 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2863 break; 2864 2865 case O_VERREVPATH: 2866 /* Outgoing packets automatically pass/match */ 2867 match = ((oif != NULL) || 2868 (m->m_pkthdr.rcvif == NULL) || 2869 ( 2870#ifdef INET6 2871 is_ipv6 ? 2872 verify_path6(&(args->f_id.src_ip6), 2873 m->m_pkthdr.rcvif) : 2874#endif 2875 verify_path(src_ip, m->m_pkthdr.rcvif))); 2876 break; 2877 2878 case O_VERSRCREACH: 2879 /* Outgoing packets automatically pass/match */ 2880 match = (hlen > 0 && ((oif != NULL) || 2881#ifdef INET6 2882 is_ipv6 ? 2883 verify_path6(&(args->f_id.src_ip6), 2884 NULL) : 2885#endif 2886 verify_path(src_ip, NULL))); 2887 break; 2888 2889 case O_ANTISPOOF: 2890 /* Outgoing packets automatically pass/match */ 2891 if (oif == NULL && hlen > 0 && 2892 ( (is_ipv4 && in_localaddr(src_ip)) 2893#ifdef INET6 2894 || (is_ipv6 && 2895 in6_localaddr(&(args->f_id.src_ip6))) 2896#endif 2897 )) 2898 match = 2899#ifdef INET6 2900 is_ipv6 ? verify_path6( 2901 &(args->f_id.src_ip6), 2902 m->m_pkthdr.rcvif) : 2903#endif 2904 verify_path(src_ip, 2905 m->m_pkthdr.rcvif); 2906 else 2907 match = 1; 2908 break; 2909 2910 case O_IPSEC: 2911#ifdef FAST_IPSEC 2912 match = (m_tag_find(m, 2913 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2914#endif 2915#ifdef IPSEC 2916 match = (ipsec_getnhist(m) != 0); 2917#endif 2918 /* otherwise no match */ 2919 break; 2920 2921#ifdef INET6 2922 case O_IP6_SRC: 2923 match = is_ipv6 && 2924 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2925 &((ipfw_insn_ip6 *)cmd)->addr6); 2926 break; 2927 2928 case O_IP6_DST: 2929 match = is_ipv6 && 2930 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2931 &((ipfw_insn_ip6 *)cmd)->addr6); 2932 break; 2933 case O_IP6_SRC_MASK: 2934 case O_IP6_DST_MASK: 2935 if (is_ipv6) { 2936 int i = cmdlen - 1; 2937 struct in6_addr p; 2938 struct in6_addr *d = 2939 &((ipfw_insn_ip6 *)cmd)->addr6; 2940 2941 for (; !match && i > 0; d += 2, 2942 i -= F_INSN_SIZE(struct in6_addr) 2943 * 2) { 2944 p = (cmd->opcode == 2945 O_IP6_SRC_MASK) ? 2946 args->f_id.src_ip6: 2947 args->f_id.dst_ip6; 2948 APPLY_MASK(&p, &d[1]); 2949 match = 2950 IN6_ARE_ADDR_EQUAL(&d[0], 2951 &p); 2952 } 2953 } 2954 break; 2955 2956 case O_IP6_SRC_ME: 2957 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 2958 break; 2959 2960 case O_IP6_DST_ME: 2961 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 2962 break; 2963 2964 case O_FLOW6ID: 2965 match = is_ipv6 && 2966 flow6id_match(args->f_id.flow_id6, 2967 (ipfw_insn_u32 *) cmd); 2968 break; 2969 2970 case O_EXT_HDR: 2971 match = is_ipv6 && 2972 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2973 break; 2974 2975 case O_IP6: 2976 match = is_ipv6; 2977 break; 2978#endif 2979 2980 case O_IP4: 2981 match = is_ipv4; 2982 break; 2983 2984 case O_TAG: { 2985 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 2986 tablearg : cmd->arg1; 2987 2988 /* Packet is already tagged with this tag? */ 2989 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 2990 2991 /* We have `untag' action when F_NOT flag is 2992 * present. And we must remove this mtag from 2993 * mbuf and reset `match' to zero (`match' will 2994 * be inversed later). 2995 * Otherwise we should allocate new mtag and 2996 * push it into mbuf. 2997 */ 2998 if (cmd->len & F_NOT) { /* `untag' action */ 2999 if (mtag != NULL) 3000 m_tag_delete(m, mtag); 3001 } else if (mtag == NULL) { 3002 if ((mtag = m_tag_alloc(MTAG_IPFW, 3003 tag, 0, M_NOWAIT)) != NULL) 3004 m_tag_prepend(m, mtag); 3005 } 3006 match = (cmd->len & F_NOT) ? 0: 1; 3007 break; 3008 } 3009 3010 case O_TAGGED: { 3011 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3012 tablearg : cmd->arg1; 3013 3014 if (cmdlen == 1) { 3015 match = m_tag_locate(m, MTAG_IPFW, 3016 tag, NULL) != NULL; 3017 break; 3018 } 3019 3020 /* we have ranges */ 3021 for (mtag = m_tag_first(m); 3022 mtag != NULL && !match; 3023 mtag = m_tag_next(m, mtag)) { 3024 uint16_t *p; 3025 int i; 3026 3027 if (mtag->m_tag_cookie != MTAG_IPFW) 3028 continue; 3029 3030 p = ((ipfw_insn_u16 *)cmd)->ports; 3031 i = cmdlen - 1; 3032 for(; !match && i > 0; i--, p += 2) 3033 match = 3034 mtag->m_tag_id >= p[0] && 3035 mtag->m_tag_id <= p[1]; 3036 } 3037 break; 3038 } 3039 3040 /* 3041 * The second set of opcodes represents 'actions', 3042 * i.e. the terminal part of a rule once the packet 3043 * matches all previous patterns. 3044 * Typically there is only one action for each rule, 3045 * and the opcode is stored at the end of the rule 3046 * (but there are exceptions -- see below). 3047 * 3048 * In general, here we set retval and terminate the 3049 * outer loop (would be a 'break 3' in some language, 3050 * but we need to do a 'goto done'). 3051 * 3052 * Exceptions: 3053 * O_COUNT and O_SKIPTO actions: 3054 * instead of terminating, we jump to the next rule 3055 * ('goto next_rule', equivalent to a 'break 2'), 3056 * or to the SKIPTO target ('goto again' after 3057 * having set f, cmd and l), respectively. 3058 * 3059 * O_TAG, O_LOG and O_ALTQ action parameters: 3060 * perform some action and set match = 1; 3061 * 3062 * O_LIMIT and O_KEEP_STATE: these opcodes are 3063 * not real 'actions', and are stored right 3064 * before the 'action' part of the rule. 3065 * These opcodes try to install an entry in the 3066 * state tables; if successful, we continue with 3067 * the next opcode (match=1; break;), otherwise 3068 * the packet * must be dropped 3069 * ('goto done' after setting retval); 3070 * 3071 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3072 * cause a lookup of the state table, and a jump 3073 * to the 'action' part of the parent rule 3074 * ('goto check_body') if an entry is found, or 3075 * (CHECK_STATE only) a jump to the next rule if 3076 * the entry is not found ('goto next_rule'). 3077 * The result of the lookup is cached to make 3078 * further instances of these opcodes are 3079 * effectively NOPs. 3080 */ 3081 case O_LIMIT: 3082 case O_KEEP_STATE: 3083 if (install_state(f, 3084 (ipfw_insn_limit *)cmd, args, tablearg)) { 3085 retval = IP_FW_DENY; 3086 goto done; /* error/limit violation */ 3087 } 3088 match = 1; 3089 break; 3090 3091 case O_PROBE_STATE: 3092 case O_CHECK_STATE: 3093 /* 3094 * dynamic rules are checked at the first 3095 * keep-state or check-state occurrence, 3096 * with the result being stored in dyn_dir. 3097 * The compiler introduces a PROBE_STATE 3098 * instruction for us when we have a 3099 * KEEP_STATE (because PROBE_STATE needs 3100 * to be run first). 3101 */ 3102 if (dyn_dir == MATCH_UNKNOWN && 3103 (q = lookup_dyn_rule(&args->f_id, 3104 &dyn_dir, proto == IPPROTO_TCP ? 3105 TCP(ulp) : NULL)) 3106 != NULL) { 3107 /* 3108 * Found dynamic entry, update stats 3109 * and jump to the 'action' part of 3110 * the parent rule. 3111 */ 3112 q->pcnt++; 3113 q->bcnt += pktlen; 3114 f = q->rule; 3115 cmd = ACTION_PTR(f); 3116 l = f->cmd_len - f->act_ofs; 3117 IPFW_DYN_UNLOCK(); 3118 goto check_body; 3119 } 3120 /* 3121 * Dynamic entry not found. If CHECK_STATE, 3122 * skip to next rule, if PROBE_STATE just 3123 * ignore and continue with next opcode. 3124 */ 3125 if (cmd->opcode == O_CHECK_STATE) 3126 goto next_rule; 3127 match = 1; 3128 break; 3129 3130 case O_ACCEPT: 3131 retval = 0; /* accept */ 3132 goto done; 3133 3134 case O_PIPE: 3135 case O_QUEUE: 3136 args->rule = f; /* report matching rule */ 3137 if (cmd->arg1 == IP_FW_TABLEARG) 3138 args->cookie = tablearg; 3139 else 3140 args->cookie = cmd->arg1; 3141 retval = IP_FW_DUMMYNET; 3142 goto done; 3143 3144 case O_DIVERT: 3145 case O_TEE: { 3146 struct divert_tag *dt; 3147 3148 if (args->eh) /* not on layer 2 */ 3149 break; 3150 mtag = m_tag_get(PACKET_TAG_DIVERT, 3151 sizeof(struct divert_tag), 3152 M_NOWAIT); 3153 if (mtag == NULL) { 3154 /* XXX statistic */ 3155 /* drop packet */ 3156 IPFW_RUNLOCK(chain); 3157 return (IP_FW_DENY); 3158 } 3159 dt = (struct divert_tag *)(mtag+1); 3160 dt->cookie = f->rulenum; 3161 if (cmd->arg1 == IP_FW_TABLEARG) 3162 dt->info = tablearg; 3163 else 3164 dt->info = cmd->arg1; 3165 m_tag_prepend(m, mtag); 3166 retval = (cmd->opcode == O_DIVERT) ? 3167 IP_FW_DIVERT : IP_FW_TEE; 3168 goto done; 3169 } 3170 3171 case O_COUNT: 3172 case O_SKIPTO: 3173 f->pcnt++; /* update stats */ 3174 f->bcnt += pktlen; 3175 f->timestamp = time_uptime; 3176 if (cmd->opcode == O_COUNT) 3177 goto next_rule; 3178 /* handle skipto */ 3179 if (f->next_rule == NULL) 3180 lookup_next_rule(f); 3181 f = f->next_rule; 3182 goto again; 3183 3184 case O_REJECT: 3185 /* 3186 * Drop the packet and send a reject notice 3187 * if the packet is not ICMP (or is an ICMP 3188 * query), and it is not multicast/broadcast. 3189 */ 3190 if (hlen > 0 && is_ipv4 && offset == 0 && 3191 (proto != IPPROTO_ICMP || 3192 is_icmp_query(ICMP(ulp))) && 3193 !(m->m_flags & (M_BCAST|M_MCAST)) && 3194 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3195 send_reject(args, cmd->arg1, ip_len); 3196 m = args->m; 3197 } 3198 /* FALLTHROUGH */ 3199#ifdef INET6 3200 case O_UNREACH6: 3201 if (hlen > 0 && is_ipv6 && 3202 ((offset & IP6F_OFF_MASK) == 0) && 3203 (proto != IPPROTO_ICMPV6 || 3204 (is_icmp6_query(args->f_id.flags) == 1)) && 3205 !(m->m_flags & (M_BCAST|M_MCAST)) && 3206 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3207 send_reject6(args, cmd->arg1, hlen); 3208 m = args->m; 3209 } 3210 /* FALLTHROUGH */ 3211#endif 3212 case O_DENY: 3213 retval = IP_FW_DENY; 3214 goto done; 3215 3216 case O_FORWARD_IP: { 3217 struct sockaddr_in *sa; 3218 sa = &(((ipfw_insn_sa *)cmd)->sa); 3219 if (args->eh) /* not valid on layer2 pkts */ 3220 break; 3221 if (!q || dyn_dir == MATCH_FORWARD) { 3222 if (sa->sin_addr.s_addr == INADDR_ANY) { 3223 bcopy(sa, &args->hopstore, 3224 sizeof(*sa)); 3225 args->hopstore.sin_addr.s_addr = 3226 htonl(tablearg); 3227 args->next_hop = 3228 &args->hopstore; 3229 } else { 3230 args->next_hop = sa; 3231 } 3232 } 3233 retval = IP_FW_PASS; 3234 } 3235 goto done; 3236 3237 case O_NETGRAPH: 3238 case O_NGTEE: 3239 args->rule = f; /* report matching rule */ 3240 if (cmd->arg1 == IP_FW_TABLEARG) 3241 args->cookie = tablearg; 3242 else 3243 args->cookie = cmd->arg1; 3244 retval = (cmd->opcode == O_NETGRAPH) ? 3245 IP_FW_NETGRAPH : IP_FW_NGTEE; 3246 goto done; 3247 3248 default: 3249 panic("-- unknown opcode %d\n", cmd->opcode); 3250 } /* end of switch() on opcodes */ 3251 3252 if (cmd->len & F_NOT) 3253 match = !match; 3254 3255 if (match) { 3256 if (cmd->len & F_OR) 3257 skip_or = 1; 3258 } else { 3259 if (!(cmd->len & F_OR)) /* not an OR block, */ 3260 break; /* try next rule */ 3261 } 3262 3263 } /* end of inner for, scan opcodes */ 3264 3265next_rule:; /* try next rule */ 3266 3267 } /* end of outer for, scan rules */ 3268 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3269 IPFW_RUNLOCK(chain); 3270 return (IP_FW_DENY); 3271 3272done: 3273 /* Update statistics */ 3274 f->pcnt++; 3275 f->bcnt += pktlen; 3276 f->timestamp = time_uptime; 3277 IPFW_RUNLOCK(chain); 3278 return (retval); 3279 3280pullup_failed: 3281 if (fw_verbose) 3282 printf("ipfw: pullup failed\n"); 3283 return (IP_FW_DENY); 3284} 3285 3286/* 3287 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3288 * These will be reconstructed on the fly as packets are matched. 3289 */ 3290static void 3291flush_rule_ptrs(struct ip_fw_chain *chain) 3292{ 3293 struct ip_fw *rule; 3294 3295 IPFW_WLOCK_ASSERT(chain); 3296 3297 for (rule = chain->rules; rule; rule = rule->next) 3298 rule->next_rule = NULL; 3299} 3300 3301/* 3302 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3303 * possibly create a rule number and add the rule to the list. 3304 * Update the rule_number in the input struct so the caller knows it as well. 3305 */ 3306static int 3307add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3308{ 3309 struct ip_fw *rule, *f, *prev; 3310 int l = RULESIZE(input_rule); 3311 3312 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3313 return (EINVAL); 3314 3315 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3316 if (rule == NULL) 3317 return (ENOSPC); 3318 3319 bcopy(input_rule, rule, l); 3320 3321 rule->next = NULL; 3322 rule->next_rule = NULL; 3323 3324 rule->pcnt = 0; 3325 rule->bcnt = 0; 3326 rule->timestamp = 0; 3327 3328 IPFW_WLOCK(chain); 3329 3330 if (chain->rules == NULL) { /* default rule */ 3331 chain->rules = rule; 3332 goto done; 3333 } 3334 3335 /* 3336 * If rulenum is 0, find highest numbered rule before the 3337 * default rule, and add autoinc_step 3338 */ 3339 if (autoinc_step < 1) 3340 autoinc_step = 1; 3341 else if (autoinc_step > 1000) 3342 autoinc_step = 1000; 3343 if (rule->rulenum == 0) { 3344 /* 3345 * locate the highest numbered rule before default 3346 */ 3347 for (f = chain->rules; f; f = f->next) { 3348 if (f->rulenum == IPFW_DEFAULT_RULE) 3349 break; 3350 rule->rulenum = f->rulenum; 3351 } 3352 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 3353 rule->rulenum += autoinc_step; 3354 input_rule->rulenum = rule->rulenum; 3355 } 3356 3357 /* 3358 * Now insert the new rule in the right place in the sorted list. 3359 */ 3360 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3361 if (f->rulenum > rule->rulenum) { /* found the location */ 3362 if (prev) { 3363 rule->next = f; 3364 prev->next = rule; 3365 } else { /* head insert */ 3366 rule->next = chain->rules; 3367 chain->rules = rule; 3368 } 3369 break; 3370 } 3371 } 3372 flush_rule_ptrs(chain); 3373done: 3374 static_count++; 3375 static_len += l; 3376 IPFW_WUNLOCK(chain); 3377 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3378 rule->rulenum, static_count);) 3379 return (0); 3380} 3381 3382/** 3383 * Remove a static rule (including derived * dynamic rules) 3384 * and place it on the ``reap list'' for later reclamation. 3385 * The caller is in charge of clearing rule pointers to avoid 3386 * dangling pointers. 3387 * @return a pointer to the next entry. 3388 * Arguments are not checked, so they better be correct. 3389 */ 3390static struct ip_fw * 3391remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 3392{ 3393 struct ip_fw *n; 3394 int l = RULESIZE(rule); 3395 3396 IPFW_WLOCK_ASSERT(chain); 3397 3398 n = rule->next; 3399 IPFW_DYN_LOCK(); 3400 remove_dyn_rule(rule, NULL /* force removal */); 3401 IPFW_DYN_UNLOCK(); 3402 if (prev == NULL) 3403 chain->rules = n; 3404 else 3405 prev->next = n; 3406 static_count--; 3407 static_len -= l; 3408 3409 rule->next = chain->reap; 3410 chain->reap = rule; 3411 3412 return n; 3413} 3414 3415/** 3416 * Reclaim storage associated with a list of rules. This is 3417 * typically the list created using remove_rule. 3418 */ 3419static void 3420reap_rules(struct ip_fw *head) 3421{ 3422 struct ip_fw *rule; 3423 3424 while ((rule = head) != NULL) { 3425 head = head->next; 3426 if (DUMMYNET_LOADED) 3427 ip_dn_ruledel_ptr(rule); 3428 free(rule, M_IPFW); 3429 } 3430} 3431 3432/* 3433 * Remove all rules from a chain (except rules in set RESVD_SET 3434 * unless kill_default = 1). The caller is responsible for 3435 * reclaiming storage for the rules left in chain->reap. 3436 */ 3437static void 3438free_chain(struct ip_fw_chain *chain, int kill_default) 3439{ 3440 struct ip_fw *prev, *rule; 3441 3442 IPFW_WLOCK_ASSERT(chain); 3443 3444 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3445 for (prev = NULL, rule = chain->rules; rule ; ) 3446 if (kill_default || rule->set != RESVD_SET) 3447 rule = remove_rule(chain, rule, prev); 3448 else { 3449 prev = rule; 3450 rule = rule->next; 3451 } 3452} 3453 3454/** 3455 * Remove all rules with given number, and also do set manipulation. 3456 * Assumes chain != NULL && *chain != NULL. 3457 * 3458 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3459 * the next 8 bits are the new set, the top 8 bits are the command: 3460 * 3461 * 0 delete rules with given number 3462 * 1 delete rules with given set number 3463 * 2 move rules with given number to new set 3464 * 3 move rules with given set number to new set 3465 * 4 swap sets with given numbers 3466 */ 3467static int 3468del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3469{ 3470 struct ip_fw *prev = NULL, *rule; 3471 u_int16_t rulenum; /* rule or old_set */ 3472 u_int8_t cmd, new_set; 3473 3474 rulenum = arg & 0xffff; 3475 cmd = (arg >> 24) & 0xff; 3476 new_set = (arg >> 16) & 0xff; 3477 3478 if (cmd > 4) 3479 return EINVAL; 3480 if (new_set > RESVD_SET) 3481 return EINVAL; 3482 if (cmd == 0 || cmd == 2) { 3483 if (rulenum >= IPFW_DEFAULT_RULE) 3484 return EINVAL; 3485 } else { 3486 if (rulenum > RESVD_SET) /* old_set */ 3487 return EINVAL; 3488 } 3489 3490 IPFW_WLOCK(chain); 3491 rule = chain->rules; 3492 chain->reap = NULL; 3493 switch (cmd) { 3494 case 0: /* delete rules with given number */ 3495 /* 3496 * locate first rule to delete 3497 */ 3498 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3499 ; 3500 if (rule->rulenum != rulenum) { 3501 IPFW_WUNLOCK(chain); 3502 return EINVAL; 3503 } 3504 3505 /* 3506 * flush pointers outside the loop, then delete all matching 3507 * rules. prev remains the same throughout the cycle. 3508 */ 3509 flush_rule_ptrs(chain); 3510 while (rule->rulenum == rulenum) 3511 rule = remove_rule(chain, rule, prev); 3512 break; 3513 3514 case 1: /* delete all rules with given set number */ 3515 flush_rule_ptrs(chain); 3516 rule = chain->rules; 3517 while (rule->rulenum < IPFW_DEFAULT_RULE) 3518 if (rule->set == rulenum) 3519 rule = remove_rule(chain, rule, prev); 3520 else { 3521 prev = rule; 3522 rule = rule->next; 3523 } 3524 break; 3525 3526 case 2: /* move rules with given number to new set */ 3527 rule = chain->rules; 3528 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3529 if (rule->rulenum == rulenum) 3530 rule->set = new_set; 3531 break; 3532 3533 case 3: /* move rules with given set number to new set */ 3534 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3535 if (rule->set == rulenum) 3536 rule->set = new_set; 3537 break; 3538 3539 case 4: /* swap two sets */ 3540 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3541 if (rule->set == rulenum) 3542 rule->set = new_set; 3543 else if (rule->set == new_set) 3544 rule->set = rulenum; 3545 break; 3546 } 3547 /* 3548 * Look for rules to reclaim. We grab the list before 3549 * releasing the lock then reclaim them w/o the lock to 3550 * avoid a LOR with dummynet. 3551 */ 3552 rule = chain->reap; 3553 chain->reap = NULL; 3554 IPFW_WUNLOCK(chain); 3555 if (rule) 3556 reap_rules(rule); 3557 return 0; 3558} 3559 3560/* 3561 * Clear counters for a specific rule. 3562 * The enclosing "table" is assumed locked. 3563 */ 3564static void 3565clear_counters(struct ip_fw *rule, int log_only) 3566{ 3567 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3568 3569 if (log_only == 0) { 3570 rule->bcnt = rule->pcnt = 0; 3571 rule->timestamp = 0; 3572 } 3573 if (l->o.opcode == O_LOG) 3574 l->log_left = l->max_log; 3575} 3576 3577/** 3578 * Reset some or all counters on firewall rules. 3579 * @arg frwl is null to clear all entries, or contains a specific 3580 * rule number. 3581 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 3582 */ 3583static int 3584zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 3585{ 3586 struct ip_fw *rule; 3587 char *msg; 3588 3589 IPFW_WLOCK(chain); 3590 if (rulenum == 0) { 3591 norule_counter = 0; 3592 for (rule = chain->rules; rule; rule = rule->next) 3593 clear_counters(rule, log_only); 3594 msg = log_only ? "ipfw: All logging counts reset.\n" : 3595 "ipfw: Accounting cleared.\n"; 3596 } else { 3597 int cleared = 0; 3598 /* 3599 * We can have multiple rules with the same number, so we 3600 * need to clear them all. 3601 */ 3602 for (rule = chain->rules; rule; rule = rule->next) 3603 if (rule->rulenum == rulenum) { 3604 while (rule && rule->rulenum == rulenum) { 3605 clear_counters(rule, log_only); 3606 rule = rule->next; 3607 } 3608 cleared = 1; 3609 break; 3610 } 3611 if (!cleared) { /* we did not find any matching rules */ 3612 IPFW_WUNLOCK(chain); 3613 return (EINVAL); 3614 } 3615 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 3616 "ipfw: Entry %d cleared.\n"; 3617 } 3618 IPFW_WUNLOCK(chain); 3619 3620 if (fw_verbose) 3621 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 3622 return (0); 3623} 3624 3625/* 3626 * Check validity of the structure before insert. 3627 * Fortunately rules are simple, so this mostly need to check rule sizes. 3628 */ 3629static int 3630check_ipfw_struct(struct ip_fw *rule, int size) 3631{ 3632 int l, cmdlen = 0; 3633 int have_action=0; 3634 ipfw_insn *cmd; 3635 3636 if (size < sizeof(*rule)) { 3637 printf("ipfw: rule too short\n"); 3638 return (EINVAL); 3639 } 3640 /* first, check for valid size */ 3641 l = RULESIZE(rule); 3642 if (l != size) { 3643 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3644 return (EINVAL); 3645 } 3646 if (rule->act_ofs >= rule->cmd_len) { 3647 printf("ipfw: bogus action offset (%u > %u)\n", 3648 rule->act_ofs, rule->cmd_len - 1); 3649 return (EINVAL); 3650 } 3651 /* 3652 * Now go for the individual checks. Very simple ones, basically only 3653 * instruction sizes. 3654 */ 3655 for (l = rule->cmd_len, cmd = rule->cmd ; 3656 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3657 cmdlen = F_LEN(cmd); 3658 if (cmdlen > l) { 3659 printf("ipfw: opcode %d size truncated\n", 3660 cmd->opcode); 3661 return EINVAL; 3662 } 3663 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3664 switch (cmd->opcode) { 3665 case O_PROBE_STATE: 3666 case O_KEEP_STATE: 3667 case O_PROTO: 3668 case O_IP_SRC_ME: 3669 case O_IP_DST_ME: 3670 case O_LAYER2: 3671 case O_IN: 3672 case O_FRAG: 3673 case O_DIVERTED: 3674 case O_IPOPT: 3675 case O_IPTOS: 3676 case O_IPPRECEDENCE: 3677 case O_IPVER: 3678 case O_TCPWIN: 3679 case O_TCPFLAGS: 3680 case O_TCPOPTS: 3681 case O_ESTAB: 3682 case O_VERREVPATH: 3683 case O_VERSRCREACH: 3684 case O_ANTISPOOF: 3685 case O_IPSEC: 3686#ifdef INET6 3687 case O_IP6_SRC_ME: 3688 case O_IP6_DST_ME: 3689 case O_EXT_HDR: 3690 case O_IP6: 3691#endif 3692 case O_IP4: 3693 case O_TAG: 3694 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3695 goto bad_size; 3696 break; 3697 3698 case O_UID: 3699 case O_GID: 3700 case O_JAIL: 3701 case O_IP_SRC: 3702 case O_IP_DST: 3703 case O_TCPSEQ: 3704 case O_TCPACK: 3705 case O_PROB: 3706 case O_ICMPTYPE: 3707 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3708 goto bad_size; 3709 break; 3710 3711 case O_LIMIT: 3712 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3713 goto bad_size; 3714 break; 3715 3716 case O_LOG: 3717 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3718 goto bad_size; 3719 3720 ((ipfw_insn_log *)cmd)->log_left = 3721 ((ipfw_insn_log *)cmd)->max_log; 3722 3723 break; 3724 3725 case O_IP_SRC_MASK: 3726 case O_IP_DST_MASK: 3727 /* only odd command lengths */ 3728 if ( !(cmdlen & 1) || cmdlen > 31) 3729 goto bad_size; 3730 break; 3731 3732 case O_IP_SRC_SET: 3733 case O_IP_DST_SET: 3734 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3735 printf("ipfw: invalid set size %d\n", 3736 cmd->arg1); 3737 return EINVAL; 3738 } 3739 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3740 (cmd->arg1+31)/32 ) 3741 goto bad_size; 3742 break; 3743 3744 case O_IP_SRC_LOOKUP: 3745 case O_IP_DST_LOOKUP: 3746 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3747 printf("ipfw: invalid table number %d\n", 3748 cmd->arg1); 3749 return (EINVAL); 3750 } 3751 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3752 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3753 goto bad_size; 3754 break; 3755 3756 case O_MACADDR2: 3757 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3758 goto bad_size; 3759 break; 3760 3761 case O_NOP: 3762 case O_IPID: 3763 case O_IPTTL: 3764 case O_IPLEN: 3765 case O_TCPDATALEN: 3766 case O_TAGGED: 3767 if (cmdlen < 1 || cmdlen > 31) 3768 goto bad_size; 3769 break; 3770 3771 case O_MAC_TYPE: 3772 case O_IP_SRCPORT: 3773 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3774 if (cmdlen < 2 || cmdlen > 31) 3775 goto bad_size; 3776 break; 3777 3778 case O_RECV: 3779 case O_XMIT: 3780 case O_VIA: 3781 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3782 goto bad_size; 3783 break; 3784 3785 case O_ALTQ: 3786 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3787 goto bad_size; 3788 break; 3789 3790 case O_PIPE: 3791 case O_QUEUE: 3792 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3793 goto bad_size; 3794 goto check_action; 3795 3796 case O_FORWARD_IP: 3797#ifdef IPFIREWALL_FORWARD 3798 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3799 goto bad_size; 3800 goto check_action; 3801#else 3802 return EINVAL; 3803#endif 3804 3805 case O_DIVERT: 3806 case O_TEE: 3807 if (ip_divert_ptr == NULL) 3808 return EINVAL; 3809 else 3810 goto check_size; 3811 case O_NETGRAPH: 3812 case O_NGTEE: 3813 if (!NG_IPFW_LOADED) 3814 return EINVAL; 3815 else 3816 goto check_size; 3817 case O_FORWARD_MAC: /* XXX not implemented yet */ 3818 case O_CHECK_STATE: 3819 case O_COUNT: 3820 case O_ACCEPT: 3821 case O_DENY: 3822 case O_REJECT: 3823#ifdef INET6 3824 case O_UNREACH6: 3825#endif 3826 case O_SKIPTO: 3827check_size: 3828 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3829 goto bad_size; 3830check_action: 3831 if (have_action) { 3832 printf("ipfw: opcode %d, multiple actions" 3833 " not allowed\n", 3834 cmd->opcode); 3835 return EINVAL; 3836 } 3837 have_action = 1; 3838 if (l != cmdlen) { 3839 printf("ipfw: opcode %d, action must be" 3840 " last opcode\n", 3841 cmd->opcode); 3842 return EINVAL; 3843 } 3844 break; 3845#ifdef INET6 3846 case O_IP6_SRC: 3847 case O_IP6_DST: 3848 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 3849 F_INSN_SIZE(ipfw_insn)) 3850 goto bad_size; 3851 break; 3852 3853 case O_FLOW6ID: 3854 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3855 ((ipfw_insn_u32 *)cmd)->o.arg1) 3856 goto bad_size; 3857 break; 3858 3859 case O_IP6_SRC_MASK: 3860 case O_IP6_DST_MASK: 3861 if ( !(cmdlen & 1) || cmdlen > 127) 3862 goto bad_size; 3863 break; 3864 case O_ICMP6TYPE: 3865 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 3866 goto bad_size; 3867 break; 3868#endif 3869 3870 default: 3871 switch (cmd->opcode) { 3872#ifndef INET6 3873 case O_IP6_SRC_ME: 3874 case O_IP6_DST_ME: 3875 case O_EXT_HDR: 3876 case O_IP6: 3877 case O_UNREACH6: 3878 case O_IP6_SRC: 3879 case O_IP6_DST: 3880 case O_FLOW6ID: 3881 case O_IP6_SRC_MASK: 3882 case O_IP6_DST_MASK: 3883 case O_ICMP6TYPE: 3884 printf("ipfw: no IPv6 support in kernel\n"); 3885 return EPROTONOSUPPORT; 3886#endif 3887 default: 3888 printf("ipfw: opcode %d, unknown opcode\n", 3889 cmd->opcode); 3890 return EINVAL; 3891 } 3892 } 3893 } 3894 if (have_action == 0) { 3895 printf("ipfw: missing action\n"); 3896 return EINVAL; 3897 } 3898 return 0; 3899 3900bad_size: 3901 printf("ipfw: opcode %d size %d wrong\n", 3902 cmd->opcode, cmdlen); 3903 return EINVAL; 3904} 3905 3906/* 3907 * Copy the static and dynamic rules to the supplied buffer 3908 * and return the amount of space actually used. 3909 */ 3910static size_t 3911ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 3912{ 3913 char *bp = buf; 3914 char *ep = bp + space; 3915 struct ip_fw *rule; 3916 int i; 3917 3918 /* XXX this can take a long time and locking will block packet flow */ 3919 IPFW_RLOCK(chain); 3920 for (rule = chain->rules; rule ; rule = rule->next) { 3921 /* 3922 * Verify the entry fits in the buffer in case the 3923 * rules changed between calculating buffer space and 3924 * now. This would be better done using a generation 3925 * number but should suffice for now. 3926 */ 3927 i = RULESIZE(rule); 3928 if (bp + i <= ep) { 3929 bcopy(rule, bp, i); 3930 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 3931 sizeof(set_disable)); 3932 bp += i; 3933 } 3934 } 3935 IPFW_RUNLOCK(chain); 3936 if (ipfw_dyn_v) { 3937 ipfw_dyn_rule *p, *last = NULL; 3938 3939 IPFW_DYN_LOCK(); 3940 for (i = 0 ; i < curr_dyn_buckets; i++) 3941 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 3942 if (bp + sizeof *p <= ep) { 3943 ipfw_dyn_rule *dst = 3944 (ipfw_dyn_rule *)bp; 3945 bcopy(p, dst, sizeof *p); 3946 bcopy(&(p->rule->rulenum), &(dst->rule), 3947 sizeof(p->rule->rulenum)); 3948 /* 3949 * store a non-null value in "next". 3950 * The userland code will interpret a 3951 * NULL here as a marker 3952 * for the last dynamic rule. 3953 */ 3954 bcopy(&dst, &dst->next, sizeof(dst)); 3955 last = dst; 3956 dst->expire = 3957 TIME_LEQ(dst->expire, time_uptime) ? 3958 0 : dst->expire - time_uptime ; 3959 bp += sizeof(ipfw_dyn_rule); 3960 } 3961 } 3962 IPFW_DYN_UNLOCK(); 3963 if (last != NULL) /* mark last dynamic rule */ 3964 bzero(&last->next, sizeof(last)); 3965 } 3966 return (bp - (char *)buf); 3967} 3968 3969 3970/** 3971 * {set|get}sockopt parser. 3972 */ 3973static int 3974ipfw_ctl(struct sockopt *sopt) 3975{ 3976#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3977 int error, rule_num; 3978 size_t size; 3979 struct ip_fw *buf, *rule; 3980 u_int32_t rulenum[2]; 3981 3982 error = suser(sopt->sopt_td); 3983 if (error) 3984 return (error); 3985 3986 /* 3987 * Disallow modifications in really-really secure mode, but still allow 3988 * the logging counters to be reset. 3989 */ 3990 if (sopt->sopt_name == IP_FW_ADD || 3991 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 3992 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3993 if (error) 3994 return (error); 3995 } 3996 3997 error = 0; 3998 3999 switch (sopt->sopt_name) { 4000 case IP_FW_GET: 4001 /* 4002 * pass up a copy of the current rules. Static rules 4003 * come first (the last of which has number IPFW_DEFAULT_RULE), 4004 * followed by a possibly empty list of dynamic rule. 4005 * The last dynamic rule has NULL in the "next" field. 4006 * 4007 * Note that the calculated size is used to bound the 4008 * amount of data returned to the user. The rule set may 4009 * change between calculating the size and returning the 4010 * data in which case we'll just return what fits. 4011 */ 4012 size = static_len; /* size of static rules */ 4013 if (ipfw_dyn_v) /* add size of dyn.rules */ 4014 size += (dyn_count * sizeof(ipfw_dyn_rule)); 4015 4016 /* 4017 * XXX todo: if the user passes a short length just to know 4018 * how much room is needed, do not bother filling up the 4019 * buffer, just jump to the sooptcopyout. 4020 */ 4021 buf = malloc(size, M_TEMP, M_WAITOK); 4022 error = sooptcopyout(sopt, buf, 4023 ipfw_getrules(&layer3_chain, buf, size)); 4024 free(buf, M_TEMP); 4025 break; 4026 4027 case IP_FW_FLUSH: 4028 /* 4029 * Normally we cannot release the lock on each iteration. 4030 * We could do it here only because we start from the head all 4031 * the times so there is no risk of missing some entries. 4032 * On the other hand, the risk is that we end up with 4033 * a very inconsistent ruleset, so better keep the lock 4034 * around the whole cycle. 4035 * 4036 * XXX this code can be improved by resetting the head of 4037 * the list to point to the default rule, and then freeing 4038 * the old list without the need for a lock. 4039 */ 4040 4041 IPFW_WLOCK(&layer3_chain); 4042 layer3_chain.reap = NULL; 4043 free_chain(&layer3_chain, 0 /* keep default rule */); 4044 rule = layer3_chain.reap; 4045 layer3_chain.reap = NULL; 4046 IPFW_WUNLOCK(&layer3_chain); 4047 if (rule != NULL) 4048 reap_rules(rule); 4049 break; 4050 4051 case IP_FW_ADD: 4052 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 4053 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 4054 sizeof(struct ip_fw) ); 4055 if (error == 0) 4056 error = check_ipfw_struct(rule, sopt->sopt_valsize); 4057 if (error == 0) { 4058 error = add_rule(&layer3_chain, rule); 4059 size = RULESIZE(rule); 4060 if (!error && sopt->sopt_dir == SOPT_GET) 4061 error = sooptcopyout(sopt, rule, size); 4062 } 4063 free(rule, M_TEMP); 4064 break; 4065 4066 case IP_FW_DEL: 4067 /* 4068 * IP_FW_DEL is used for deleting single rules or sets, 4069 * and (ab)used to atomically manipulate sets. Argument size 4070 * is used to distinguish between the two: 4071 * sizeof(u_int32_t) 4072 * delete single rule or set of rules, 4073 * or reassign rules (or sets) to a different set. 4074 * 2*sizeof(u_int32_t) 4075 * atomic disable/enable sets. 4076 * first u_int32_t contains sets to be disabled, 4077 * second u_int32_t contains sets to be enabled. 4078 */ 4079 error = sooptcopyin(sopt, rulenum, 4080 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4081 if (error) 4082 break; 4083 size = sopt->sopt_valsize; 4084 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4085 error = del_entry(&layer3_chain, rulenum[0]); 4086 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4087 set_disable = 4088 (set_disable | rulenum[0]) & ~rulenum[1] & 4089 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4090 else 4091 error = EINVAL; 4092 break; 4093 4094 case IP_FW_ZERO: 4095 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 4096 rule_num = 0; 4097 if (sopt->sopt_val != 0) { 4098 error = sooptcopyin(sopt, &rule_num, 4099 sizeof(int), sizeof(int)); 4100 if (error) 4101 break; 4102 } 4103 error = zero_entry(&layer3_chain, rule_num, 4104 sopt->sopt_name == IP_FW_RESETLOG); 4105 break; 4106 4107 case IP_FW_TABLE_ADD: 4108 { 4109 ipfw_table_entry ent; 4110 4111 error = sooptcopyin(sopt, &ent, 4112 sizeof(ent), sizeof(ent)); 4113 if (error) 4114 break; 4115 error = add_table_entry(&layer3_chain, ent.tbl, 4116 ent.addr, ent.masklen, ent.value); 4117 } 4118 break; 4119 4120 case IP_FW_TABLE_DEL: 4121 { 4122 ipfw_table_entry ent; 4123 4124 error = sooptcopyin(sopt, &ent, 4125 sizeof(ent), sizeof(ent)); 4126 if (error) 4127 break; 4128 error = del_table_entry(&layer3_chain, ent.tbl, 4129 ent.addr, ent.masklen); 4130 } 4131 break; 4132 4133 case IP_FW_TABLE_FLUSH: 4134 { 4135 u_int16_t tbl; 4136 4137 error = sooptcopyin(sopt, &tbl, 4138 sizeof(tbl), sizeof(tbl)); 4139 if (error) 4140 break; 4141 IPFW_WLOCK(&layer3_chain); 4142 error = flush_table(&layer3_chain, tbl); 4143 IPFW_WUNLOCK(&layer3_chain); 4144 } 4145 break; 4146 4147 case IP_FW_TABLE_GETSIZE: 4148 { 4149 u_int32_t tbl, cnt; 4150 4151 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4152 sizeof(tbl)))) 4153 break; 4154 IPFW_RLOCK(&layer3_chain); 4155 error = count_table(&layer3_chain, tbl, &cnt); 4156 IPFW_RUNLOCK(&layer3_chain); 4157 if (error) 4158 break; 4159 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4160 } 4161 break; 4162 4163 case IP_FW_TABLE_LIST: 4164 { 4165 ipfw_table *tbl; 4166 4167 if (sopt->sopt_valsize < sizeof(*tbl)) { 4168 error = EINVAL; 4169 break; 4170 } 4171 size = sopt->sopt_valsize; 4172 tbl = malloc(size, M_TEMP, M_WAITOK); 4173 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4174 if (error) { 4175 free(tbl, M_TEMP); 4176 break; 4177 } 4178 tbl->size = (size - sizeof(*tbl)) / 4179 sizeof(ipfw_table_entry); 4180 IPFW_RLOCK(&layer3_chain); 4181 error = dump_table(&layer3_chain, tbl); 4182 IPFW_RUNLOCK(&layer3_chain); 4183 if (error) { 4184 free(tbl, M_TEMP); 4185 break; 4186 } 4187 error = sooptcopyout(sopt, tbl, size); 4188 free(tbl, M_TEMP); 4189 } 4190 break; 4191 4192 default: 4193 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4194 error = EINVAL; 4195 } 4196 4197 return (error); 4198#undef RULE_MAXSIZE 4199} 4200 4201/** 4202 * dummynet needs a reference to the default rule, because rules can be 4203 * deleted while packets hold a reference to them. When this happens, 4204 * dummynet changes the reference to the default rule (it could well be a 4205 * NULL pointer, but this way we do not need to check for the special 4206 * case, plus here he have info on the default behaviour). 4207 */ 4208struct ip_fw *ip_fw_default_rule; 4209 4210/* 4211 * This procedure is only used to handle keepalives. It is invoked 4212 * every dyn_keepalive_period 4213 */ 4214static void 4215ipfw_tick(void * __unused unused) 4216{ 4217 struct mbuf *m0, *m, *mnext, **mtailp; 4218 int i; 4219 ipfw_dyn_rule *q; 4220 4221 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 4222 goto done; 4223 4224 /* 4225 * We make a chain of packets to go out here -- not deferring 4226 * until after we drop the IPFW dynamic rule lock would result 4227 * in a lock order reversal with the normal packet input -> ipfw 4228 * call stack. 4229 */ 4230 m0 = NULL; 4231 mtailp = &m0; 4232 IPFW_DYN_LOCK(); 4233 for (i = 0 ; i < curr_dyn_buckets ; i++) { 4234 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 4235 if (q->dyn_type == O_LIMIT_PARENT) 4236 continue; 4237 if (q->id.proto != IPPROTO_TCP) 4238 continue; 4239 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4240 continue; 4241 if (TIME_LEQ( time_uptime+dyn_keepalive_interval, 4242 q->expire)) 4243 continue; /* too early */ 4244 if (TIME_LEQ(q->expire, time_uptime)) 4245 continue; /* too late, rule expired */ 4246 4247 *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1, 4248 q->ack_fwd, TH_SYN); 4249 if (*mtailp != NULL) 4250 mtailp = &(*mtailp)->m_nextpkt; 4251 *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1, 4252 q->ack_rev, 0); 4253 if (*mtailp != NULL) 4254 mtailp = &(*mtailp)->m_nextpkt; 4255 } 4256 } 4257 IPFW_DYN_UNLOCK(); 4258 for (m = mnext = m0; m != NULL; m = mnext) { 4259 mnext = m->m_nextpkt; 4260 m->m_nextpkt = NULL; 4261 ip_output(m, NULL, NULL, 0, NULL, NULL); 4262 } 4263done: 4264 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 4265} 4266 4267int 4268ipfw_init(void) 4269{ 4270 struct ip_fw default_rule; 4271 int error; 4272 4273#ifdef INET6 4274 /* Setup IPv6 fw sysctl tree. */ 4275 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4276 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4277 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4278 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4279 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4280 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4281 &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4282 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4283 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4284 &fw_deny_unknown_exthdrs, 0, 4285 "Deny packets with unknown IPv6 Extension Headers"); 4286#endif 4287 4288 layer3_chain.rules = NULL; 4289 IPFW_LOCK_INIT(&layer3_chain); 4290 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule zone", 4291 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4292 UMA_ALIGN_PTR, 0); 4293 IPFW_DYN_LOCK_INIT(); 4294 callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE); 4295 4296 bzero(&default_rule, sizeof default_rule); 4297 4298 default_rule.act_ofs = 0; 4299 default_rule.rulenum = IPFW_DEFAULT_RULE; 4300 default_rule.cmd_len = 1; 4301 default_rule.set = RESVD_SET; 4302 4303 default_rule.cmd[0].len = 1; 4304 default_rule.cmd[0].opcode = 4305#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4306 1 ? O_ACCEPT : 4307#endif 4308 O_DENY; 4309 4310 error = add_rule(&layer3_chain, &default_rule); 4311 if (error != 0) { 4312 printf("ipfw2: error %u initializing default rule " 4313 "(support disabled)\n", error); 4314 IPFW_DYN_LOCK_DESTROY(); 4315 IPFW_LOCK_DESTROY(&layer3_chain); 4316 uma_zdestroy(ipfw_dyn_rule_zone); 4317 return (error); 4318 } 4319 4320 ip_fw_default_rule = layer3_chain.rules; 4321 printf("ipfw2 " 4322#ifdef INET6 4323 "(+ipv6) " 4324#endif 4325 "initialized, divert %s, " 4326 "rule-based forwarding " 4327#ifdef IPFIREWALL_FORWARD 4328 "enabled, " 4329#else 4330 "disabled, " 4331#endif 4332 "default to %s, logging ", 4333#ifdef IPDIVERT 4334 "enabled", 4335#else 4336 "loadable", 4337#endif 4338 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4339 4340#ifdef IPFIREWALL_VERBOSE 4341 fw_verbose = 1; 4342#endif 4343#ifdef IPFIREWALL_VERBOSE_LIMIT 4344 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4345#endif 4346 if (fw_verbose == 0) 4347 printf("disabled\n"); 4348 else if (verbose_limit == 0) 4349 printf("unlimited\n"); 4350 else 4351 printf("limited to %d packets/entry by default\n", 4352 verbose_limit); 4353 4354 error = init_tables(&layer3_chain); 4355 if (error) { 4356 IPFW_DYN_LOCK_DESTROY(); 4357 IPFW_LOCK_DESTROY(&layer3_chain); 4358 uma_zdestroy(ipfw_dyn_rule_zone); 4359 return (error); 4360 } 4361 ip_fw_ctl_ptr = ipfw_ctl; 4362 ip_fw_chk_ptr = ipfw_chk; 4363 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 4364 4365 return (0); 4366} 4367 4368void 4369ipfw_destroy(void) 4370{ 4371 struct ip_fw *reap; 4372 4373 ip_fw_chk_ptr = NULL; 4374 ip_fw_ctl_ptr = NULL; 4375 callout_drain(&ipfw_timeout); 4376 IPFW_WLOCK(&layer3_chain); 4377 flush_tables(&layer3_chain); 4378 layer3_chain.reap = NULL; 4379 free_chain(&layer3_chain, 1 /* kill default rule */); 4380 reap = layer3_chain.reap, layer3_chain.reap = NULL; 4381 IPFW_WUNLOCK(&layer3_chain); 4382 if (reap != NULL) 4383 reap_rules(reap); 4384 IPFW_DYN_LOCK_DESTROY(); 4385 uma_zdestroy(ipfw_dyn_rule_zone); 4386 IPFW_LOCK_DESTROY(&layer3_chain); 4387 4388#ifdef INET6 4389 /* Free IPv6 fw sysctl tree. */ 4390 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4391#endif 4392 4393 printf("IP firewall unloaded\n"); 4394} 4395