ip_fw2.c revision 164258
113901Salanb/*- 213901Salanb * Copyright (c) 2002 Luigi Rizzo, Universita` di Pisa 313901Salanb * 413901Salanb * Redistribution and use in source and binary forms, with or without 513901Salanb * modification, are permitted provided that the following conditions 613901Salanb * are met: 713901Salanb * 1. Redistributions of source code must retain the above copyright 813901Salanb * notice, this list of conditions and the following disclaimer. 913901Salanb * 2. Redistributions in binary form must reproduce the above copyright 1013901Salanb * notice, this list of conditions and the following disclaimer in the 1113901Salanb * documentation and/or other materials provided with the distribution. 1213901Salanb * 1313901Salanb * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 1413901Salanb * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 1513901Salanb * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 1613901Salanb * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 1713901Salanb * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 1813901Salanb * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 1913901Salanb * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 2013901Salanb * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 2113901Salanb * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 2213901Salanb * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 2313901Salanb * SUCH DAMAGE. 2413901Salanb * 2513901Salanb * $FreeBSD: head/sys/netinet/ip_fw2.c 164258 2006-11-13 19:07:32Z bz $ 2616177Salanb */ 2713901Salanb 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/module.h> 56#include <sys/priv.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#include <netinet/sctp.h> 86 87#include <netgraph/ng_ipfw.h> 88 89#include <altq/if_altq.h> 90 91#ifdef IPSEC 92#include <netinet6/ipsec.h> 93#endif 94 95#include <netinet/ip6.h> 96#include <netinet/icmp6.h> 97#ifdef INET6 98#include <netinet6/scope6_var.h> 99#endif 100 101#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 102 103#include <machine/in_cksum.h> /* XXX for in_cksum */ 104 105#include <security/mac/mac_framework.h> 106 107/* 108 * set_disable contains one bit per set value (0..31). 109 * If the bit is set, all rules with the corresponding set 110 * are disabled. Set RESVD_SET(31) is reserved for the default rule 111 * and rules that are not deleted by the flush command, 112 * and CANNOT be disabled. 113 * Rules in set RESVD_SET can only be deleted explicitly. 114 */ 115static u_int32_t set_disable; 116 117static int fw_verbose; 118static int verbose_limit; 119 120static struct callout ipfw_timeout; 121static uma_zone_t ipfw_dyn_rule_zone; 122#define IPFW_DEFAULT_RULE 65535 123 124/* 125 * Data structure to cache our ucred related 126 * information. This structure only gets used if 127 * the user specified UID/GID based constraints in 128 * a firewall rule. 129 */ 130struct ip_fw_ugid { 131 gid_t fw_groups[NGROUPS]; 132 int fw_ngroups; 133 uid_t fw_uid; 134 int fw_prid; 135}; 136 137#define IPFW_TABLES_MAX 128 138struct ip_fw_chain { 139 struct ip_fw *rules; /* list of rules */ 140 struct ip_fw *reap; /* list of rules to reap */ 141 struct radix_node_head *tables[IPFW_TABLES_MAX]; 142 struct rwlock rwmtx; 143}; 144#define IPFW_LOCK_INIT(_chain) \ 145 rw_init(&(_chain)->rwmtx, "IPFW static rules") 146#define IPFW_LOCK_DESTROY(_chain) rw_destroy(&(_chain)->rwmtx) 147#define IPFW_WLOCK_ASSERT(_chain) do { \ 148 rw_assert(&(_chain)->rwmtx, RA_WLOCKED); \ 149 NET_ASSERT_GIANT(); \ 150} while (0) 151 152#define IPFW_RLOCK(p) rw_rlock(&(p)->rwmtx) 153#define IPFW_RUNLOCK(p) rw_runlock(&(p)->rwmtx) 154#define IPFW_WLOCK(p) rw_wlock(&(p)->rwmtx) 155#define IPFW_WUNLOCK(p) rw_wunlock(&(p)->rwmtx) 156 157/* 158 * list of rules for layer 3 159 */ 160static struct ip_fw_chain layer3_chain; 161 162MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 163MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 164 165struct table_entry { 166 struct radix_node rn[2]; 167 struct sockaddr_in addr, mask; 168 u_int32_t value; 169}; 170 171static int fw_debug = 1; 172static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 173 174extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS); 175 176#ifdef SYSCTL_NODE 177SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 178SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable, 179 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &fw_enable, 0, 180 ipfw_chg_hook, "I", "Enable ipfw"); 181SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 182 &autoinc_step, 0, "Rule number autincrement step"); 183SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 184 CTLFLAG_RW | CTLFLAG_SECURE3, 185 &fw_one_pass, 0, 186 "Only do a single pass through ipfw when using dummynet(4)"); 187SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 188 &fw_debug, 0, "Enable printing of debug ip_fw statements"); 189SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 190 CTLFLAG_RW | CTLFLAG_SECURE3, 191 &fw_verbose, 0, "Log matches to ipfw rules"); 192SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 193 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 194 195/* 196 * Description of dynamic rules. 197 * 198 * Dynamic rules are stored in lists accessed through a hash table 199 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 200 * be modified through the sysctl variable dyn_buckets which is 201 * updated when the table becomes empty. 202 * 203 * XXX currently there is only one list, ipfw_dyn. 204 * 205 * When a packet is received, its address fields are first masked 206 * with the mask defined for the rule, then hashed, then matched 207 * against the entries in the corresponding list. 208 * Dynamic rules can be used for different purposes: 209 * + stateful rules; 210 * + enforcing limits on the number of sessions; 211 * + in-kernel NAT (not implemented yet) 212 * 213 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 214 * measured in seconds and depending on the flags. 215 * 216 * The total number of dynamic rules is stored in dyn_count. 217 * The max number of dynamic rules is dyn_max. When we reach 218 * the maximum number of rules we do not create anymore. This is 219 * done to avoid consuming too much memory, but also too much 220 * time when searching on each packet (ideally, we should try instead 221 * to put a limit on the length of the list on each bucket...). 222 * 223 * Each dynamic rule holds a pointer to the parent ipfw rule so 224 * we know what action to perform. Dynamic rules are removed when 225 * the parent rule is deleted. XXX we should make them survive. 226 * 227 * There are some limitations with dynamic rules -- we do not 228 * obey the 'randomized match', and we do not do multiple 229 * passes through the firewall. XXX check the latter!!! 230 */ 231static ipfw_dyn_rule **ipfw_dyn_v = NULL; 232static u_int32_t dyn_buckets = 256; /* must be power of 2 */ 233static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 234 235static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 236#define IPFW_DYN_LOCK_INIT() \ 237 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 238#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 239#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 240#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 241#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 242 243/* 244 * Timeouts for various events in handing dynamic rules. 245 */ 246static u_int32_t dyn_ack_lifetime = 300; 247static u_int32_t dyn_syn_lifetime = 20; 248static u_int32_t dyn_fin_lifetime = 1; 249static u_int32_t dyn_rst_lifetime = 1; 250static u_int32_t dyn_udp_lifetime = 10; 251static u_int32_t dyn_short_lifetime = 5; 252 253/* 254 * Keepalives are sent if dyn_keepalive is set. They are sent every 255 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 256 * seconds of lifetime of a rule. 257 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 258 * than dyn_keepalive_period. 259 */ 260 261static u_int32_t dyn_keepalive_interval = 20; 262static u_int32_t dyn_keepalive_period = 5; 263static u_int32_t dyn_keepalive = 1; /* do send keepalives */ 264 265static u_int32_t static_count; /* # of static rules */ 266static u_int32_t static_len; /* size in bytes of static rules */ 267static u_int32_t dyn_count; /* # of dynamic rules */ 268static u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 269 270SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 271 &dyn_buckets, 0, "Number of dyn. buckets"); 272SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 273 &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 274SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 275 &dyn_count, 0, "Number of dyn. rules"); 276SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 277 &dyn_max, 0, "Max number of dyn. rules"); 278SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 279 &static_count, 0, "Number of static rules"); 280SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 281 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 282SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 283 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 284SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 285 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 286SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 287 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 288SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 289 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 290SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 291 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 292SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 293 &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 294 295#ifdef INET6 296/* 297 * IPv6 specific variables 298 */ 299SYSCTL_DECL(_net_inet6_ip6); 300 301static struct sysctl_ctx_list ip6_fw_sysctl_ctx; 302static struct sysctl_oid *ip6_fw_sysctl_tree; 303#endif /* INET6 */ 304#endif /* SYSCTL_NODE */ 305 306static int fw_deny_unknown_exthdrs = 1; 307 308 309/* 310 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 311 * Other macros just cast void * into the appropriate type 312 */ 313#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 314#define TCP(p) ((struct tcphdr *)(p)) 315#define SCTP(p) ((struct sctphdr *)(p)) 316#define UDP(p) ((struct udphdr *)(p)) 317#define ICMP(p) ((struct icmphdr *)(p)) 318#define ICMP6(p) ((struct icmp6_hdr *)(p)) 319 320static __inline int 321icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 322{ 323 int type = icmp->icmp_type; 324 325 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 326} 327 328#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 329 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 330 331static int 332is_icmp_query(struct icmphdr *icmp) 333{ 334 int type = icmp->icmp_type; 335 336 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 337} 338#undef TT 339 340/* 341 * The following checks use two arrays of 8 or 16 bits to store the 342 * bits that we want set or clear, respectively. They are in the 343 * low and high half of cmd->arg1 or cmd->d[0]. 344 * 345 * We scan options and store the bits we find set. We succeed if 346 * 347 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 348 * 349 * The code is sometimes optimized not to store additional variables. 350 */ 351 352static int 353flags_match(ipfw_insn *cmd, u_int8_t bits) 354{ 355 u_char want_clear; 356 bits = ~bits; 357 358 if ( ((cmd->arg1 & 0xff) & bits) != 0) 359 return 0; /* some bits we want set were clear */ 360 want_clear = (cmd->arg1 >> 8) & 0xff; 361 if ( (want_clear & bits) != want_clear) 362 return 0; /* some bits we want clear were set */ 363 return 1; 364} 365 366static int 367ipopts_match(struct ip *ip, ipfw_insn *cmd) 368{ 369 int optlen, bits = 0; 370 u_char *cp = (u_char *)(ip + 1); 371 int x = (ip->ip_hl << 2) - sizeof (struct ip); 372 373 for (; x > 0; x -= optlen, cp += optlen) { 374 int opt = cp[IPOPT_OPTVAL]; 375 376 if (opt == IPOPT_EOL) 377 break; 378 if (opt == IPOPT_NOP) 379 optlen = 1; 380 else { 381 optlen = cp[IPOPT_OLEN]; 382 if (optlen <= 0 || optlen > x) 383 return 0; /* invalid or truncated */ 384 } 385 switch (opt) { 386 387 default: 388 break; 389 390 case IPOPT_LSRR: 391 bits |= IP_FW_IPOPT_LSRR; 392 break; 393 394 case IPOPT_SSRR: 395 bits |= IP_FW_IPOPT_SSRR; 396 break; 397 398 case IPOPT_RR: 399 bits |= IP_FW_IPOPT_RR; 400 break; 401 402 case IPOPT_TS: 403 bits |= IP_FW_IPOPT_TS; 404 break; 405 } 406 } 407 return (flags_match(cmd, bits)); 408} 409 410static int 411tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 412{ 413 int optlen, bits = 0; 414 u_char *cp = (u_char *)(tcp + 1); 415 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 416 417 for (; x > 0; x -= optlen, cp += optlen) { 418 int opt = cp[0]; 419 if (opt == TCPOPT_EOL) 420 break; 421 if (opt == TCPOPT_NOP) 422 optlen = 1; 423 else { 424 optlen = cp[1]; 425 if (optlen <= 0) 426 break; 427 } 428 429 switch (opt) { 430 431 default: 432 break; 433 434 case TCPOPT_MAXSEG: 435 bits |= IP_FW_TCPOPT_MSS; 436 break; 437 438 case TCPOPT_WINDOW: 439 bits |= IP_FW_TCPOPT_WINDOW; 440 break; 441 442 case TCPOPT_SACK_PERMITTED: 443 case TCPOPT_SACK: 444 bits |= IP_FW_TCPOPT_SACK; 445 break; 446 447 case TCPOPT_TIMESTAMP: 448 bits |= IP_FW_TCPOPT_TS; 449 break; 450 451 } 452 } 453 return (flags_match(cmd, bits)); 454} 455 456static int 457iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 458{ 459 if (ifp == NULL) /* no iface with this packet, match fails */ 460 return 0; 461 /* Check by name or by IP address */ 462 if (cmd->name[0] != '\0') { /* match by name */ 463 /* Check name */ 464 if (cmd->p.glob) { 465 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 466 return(1); 467 } else { 468 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 469 return(1); 470 } 471 } else { 472 struct ifaddr *ia; 473 474 /* XXX lock? */ 475 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 476 if (ia->ifa_addr->sa_family != AF_INET) 477 continue; 478 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 479 (ia->ifa_addr))->sin_addr.s_addr) 480 return(1); /* match */ 481 } 482 } 483 return(0); /* no match, fail ... */ 484} 485 486/* 487 * The verify_path function checks if a route to the src exists and 488 * if it is reachable via ifp (when provided). 489 * 490 * The 'verrevpath' option checks that the interface that an IP packet 491 * arrives on is the same interface that traffic destined for the 492 * packet's source address would be routed out of. The 'versrcreach' 493 * option just checks that the source address is reachable via any route 494 * (except default) in the routing table. These two are a measure to block 495 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 496 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 497 * is purposely reminiscent of the Cisco IOS command, 498 * 499 * ip verify unicast reverse-path 500 * ip verify unicast source reachable-via any 501 * 502 * which implements the same functionality. But note that syntax is 503 * misleading. The check may be performed on all IP packets whether unicast, 504 * multicast, or broadcast. 505 */ 506static int 507verify_path(struct in_addr src, struct ifnet *ifp) 508{ 509 struct route ro; 510 struct sockaddr_in *dst; 511 512 bzero(&ro, sizeof(ro)); 513 514 dst = (struct sockaddr_in *)&(ro.ro_dst); 515 dst->sin_family = AF_INET; 516 dst->sin_len = sizeof(*dst); 517 dst->sin_addr = src; 518 rtalloc_ign(&ro, RTF_CLONING); 519 520 if (ro.ro_rt == NULL) 521 return 0; 522 523 /* 524 * If ifp is provided, check for equality with rtentry. 525 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 526 * in order to pass packets injected back by if_simloop(): 527 * if useloopback == 1 routing entry (via lo0) for our own address 528 * may exist, so we need to handle routing assymetry. 529 */ 530 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 531 RTFREE(ro.ro_rt); 532 return 0; 533 } 534 535 /* if no ifp provided, check if rtentry is not default route */ 536 if (ifp == NULL && 537 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 538 RTFREE(ro.ro_rt); 539 return 0; 540 } 541 542 /* or if this is a blackhole/reject route */ 543 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 544 RTFREE(ro.ro_rt); 545 return 0; 546 } 547 548 /* found valid route */ 549 RTFREE(ro.ro_rt); 550 return 1; 551} 552 553#ifdef INET6 554/* 555 * ipv6 specific rules here... 556 */ 557static __inline int 558icmp6type_match (int type, ipfw_insn_u32 *cmd) 559{ 560 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 561} 562 563static int 564flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 565{ 566 int i; 567 for (i=0; i <= cmd->o.arg1; ++i ) 568 if (curr_flow == cmd->d[i] ) 569 return 1; 570 return 0; 571} 572 573/* support for IP6_*_ME opcodes */ 574static int 575search_ip6_addr_net (struct in6_addr * ip6_addr) 576{ 577 struct ifnet *mdc; 578 struct ifaddr *mdc2; 579 struct in6_ifaddr *fdm; 580 struct in6_addr copia; 581 582 TAILQ_FOREACH(mdc, &ifnet, if_link) 583 TAILQ_FOREACH(mdc2, &mdc->if_addrlist, ifa_list) { 584 if (mdc2->ifa_addr->sa_family == AF_INET6) { 585 fdm = (struct in6_ifaddr *)mdc2; 586 copia = fdm->ia_addr.sin6_addr; 587 /* need for leaving scope_id in the sock_addr */ 588 in6_clearscope(&copia); 589 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) 590 return 1; 591 } 592 } 593 return 0; 594} 595 596static int 597verify_path6(struct in6_addr *src, struct ifnet *ifp) 598{ 599 struct route_in6 ro; 600 struct sockaddr_in6 *dst; 601 602 bzero(&ro, sizeof(ro)); 603 604 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 605 dst->sin6_family = AF_INET6; 606 dst->sin6_len = sizeof(*dst); 607 dst->sin6_addr = *src; 608 rtalloc_ign((struct route *)&ro, RTF_CLONING); 609 610 if (ro.ro_rt == NULL) 611 return 0; 612 613 /* 614 * if ifp is provided, check for equality with rtentry 615 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 616 * to support the case of sending packets to an address of our own. 617 * (where the former interface is the first argument of if_simloop() 618 * (=ifp), the latter is lo0) 619 */ 620 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 621 RTFREE(ro.ro_rt); 622 return 0; 623 } 624 625 /* if no ifp provided, check if rtentry is not default route */ 626 if (ifp == NULL && 627 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 628 RTFREE(ro.ro_rt); 629 return 0; 630 } 631 632 /* or if this is a blackhole/reject route */ 633 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 634 RTFREE(ro.ro_rt); 635 return 0; 636 } 637 638 /* found valid route */ 639 RTFREE(ro.ro_rt); 640 return 1; 641 642} 643static __inline int 644hash_packet6(struct ipfw_flow_id *id) 645{ 646 u_int32_t i; 647 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 648 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 649 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 650 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 651 (id->dst_port) ^ (id->src_port); 652 return i; 653} 654 655static int 656is_icmp6_query(int icmp6_type) 657{ 658 if ((icmp6_type <= ICMP6_MAXTYPE) && 659 (icmp6_type == ICMP6_ECHO_REQUEST || 660 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 661 icmp6_type == ICMP6_WRUREQUEST || 662 icmp6_type == ICMP6_FQDN_QUERY || 663 icmp6_type == ICMP6_NI_QUERY)) 664 return (1); 665 666 return (0); 667} 668 669static void 670send_reject6(struct ip_fw_args *args, int code, u_int hlen) 671{ 672 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 673 struct ip6_hdr *ip6; 674 struct tcphdr *tcp; 675 tcp_seq ack, seq; 676 int flags; 677 struct { 678 struct ip6_hdr ip6; 679 struct tcphdr th; 680 } ti; 681 682 if (args->m->m_len < (hlen+sizeof(struct tcphdr))) { 683 args->m = m_pullup(args->m, hlen+sizeof(struct tcphdr)); 684 if (args->m == NULL) 685 return; 686 } 687 688 ip6 = mtod(args->m, struct ip6_hdr *); 689 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 690 691 if ((tcp->th_flags & TH_RST) != 0) { 692 m_freem(args->m); 693 return; 694 } 695 696 ti.ip6 = *ip6; 697 ti.th = *tcp; 698 ti.th.th_seq = ntohl(ti.th.th_seq); 699 ti.th.th_ack = ntohl(ti.th.th_ack); 700 ti.ip6.ip6_nxt = IPPROTO_TCP; 701 702 if (ti.th.th_flags & TH_ACK) { 703 ack = 0; 704 seq = ti.th.th_ack; 705 flags = TH_RST; 706 } else { 707 ack = ti.th.th_seq; 708 if (((args->m)->m_flags & M_PKTHDR) != 0) { 709 ack += (args->m)->m_pkthdr.len - hlen 710 - (ti.th.th_off << 2); 711 } else if (ip6->ip6_plen) { 712 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) 713 - hlen - (ti.th.th_off << 2); 714 } else { 715 m_freem(args->m); 716 return; 717 } 718 if (tcp->th_flags & TH_SYN) 719 ack++; 720 seq = 0; 721 flags = TH_RST|TH_ACK; 722 } 723 bcopy(&ti, ip6, sizeof(ti)); 724 tcp_respond(NULL, ip6, (struct tcphdr *)(ip6 + 1), 725 args->m, ack, seq, flags); 726 727 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 728 icmp6_error(args->m, ICMP6_DST_UNREACH, code, 0); 729 730 } else 731 m_freem(args->m); 732 733 args->m = NULL; 734} 735 736#endif /* INET6 */ 737 738static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 739 740#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 741#define SNP(buf) buf, sizeof(buf) 742 743/* 744 * We enter here when we have a rule with O_LOG. 745 * XXX this function alone takes about 2Kbytes of code! 746 */ 747static void 748ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 749 struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg) 750{ 751 struct ether_header *eh = args->eh; 752 char *action; 753 int limit_reached = 0; 754 char action2[40], proto[128], fragment[32]; 755 756 fragment[0] = '\0'; 757 proto[0] = '\0'; 758 759 if (f == NULL) { /* bogus pkt */ 760 if (verbose_limit != 0 && norule_counter >= verbose_limit) 761 return; 762 norule_counter++; 763 if (norule_counter == verbose_limit) 764 limit_reached = verbose_limit; 765 action = "Refuse"; 766 } else { /* O_LOG is the first action, find the real one */ 767 ipfw_insn *cmd = ACTION_PTR(f); 768 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 769 770 if (l->max_log != 0 && l->log_left == 0) 771 return; 772 l->log_left--; 773 if (l->log_left == 0) 774 limit_reached = l->max_log; 775 cmd += F_LEN(cmd); /* point to first action */ 776 if (cmd->opcode == O_ALTQ) { 777 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 778 779 snprintf(SNPARGS(action2, 0), "Altq %d", 780 altq->qid); 781 cmd += F_LEN(cmd); 782 } 783 if (cmd->opcode == O_PROB) 784 cmd += F_LEN(cmd); 785 786 if (cmd->opcode == O_TAG) 787 cmd += F_LEN(cmd); 788 789 action = action2; 790 switch (cmd->opcode) { 791 case O_DENY: 792 action = "Deny"; 793 break; 794 795 case O_REJECT: 796 if (cmd->arg1==ICMP_REJECT_RST) 797 action = "Reset"; 798 else if (cmd->arg1==ICMP_UNREACH_HOST) 799 action = "Reject"; 800 else 801 snprintf(SNPARGS(action2, 0), "Unreach %d", 802 cmd->arg1); 803 break; 804 805 case O_UNREACH6: 806 if (cmd->arg1==ICMP6_UNREACH_RST) 807 action = "Reset"; 808 else 809 snprintf(SNPARGS(action2, 0), "Unreach %d", 810 cmd->arg1); 811 break; 812 813 case O_ACCEPT: 814 action = "Accept"; 815 break; 816 case O_COUNT: 817 action = "Count"; 818 break; 819 case O_DIVERT: 820 snprintf(SNPARGS(action2, 0), "Divert %d", 821 cmd->arg1); 822 break; 823 case O_TEE: 824 snprintf(SNPARGS(action2, 0), "Tee %d", 825 cmd->arg1); 826 break; 827 case O_SKIPTO: 828 snprintf(SNPARGS(action2, 0), "SkipTo %d", 829 cmd->arg1); 830 break; 831 case O_PIPE: 832 snprintf(SNPARGS(action2, 0), "Pipe %d", 833 cmd->arg1); 834 break; 835 case O_QUEUE: 836 snprintf(SNPARGS(action2, 0), "Queue %d", 837 cmd->arg1); 838 break; 839 case O_FORWARD_IP: { 840 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 841 int len; 842 struct in_addr dummyaddr; 843 if (sa->sa.sin_addr.s_addr == INADDR_ANY) 844 dummyaddr.s_addr = htonl(tablearg); 845 else 846 dummyaddr.s_addr = sa->sa.sin_addr.s_addr; 847 848 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 849 inet_ntoa(dummyaddr)); 850 851 if (sa->sa.sin_port) 852 snprintf(SNPARGS(action2, len), ":%d", 853 sa->sa.sin_port); 854 } 855 break; 856 case O_NETGRAPH: 857 snprintf(SNPARGS(action2, 0), "Netgraph %d", 858 cmd->arg1); 859 break; 860 case O_NGTEE: 861 snprintf(SNPARGS(action2, 0), "Ngtee %d", 862 cmd->arg1); 863 break; 864 default: 865 action = "UNKNOWN"; 866 break; 867 } 868 } 869 870 if (hlen == 0) { /* non-ip */ 871 snprintf(SNPARGS(proto, 0), "MAC"); 872 873 } else { 874 int len; 875 char src[48], dst[48]; 876 struct icmphdr *icmp; 877 struct tcphdr *tcp; 878 struct udphdr *udp; 879 /* Initialize to make compiler happy. */ 880 struct ip *ip = NULL; 881#ifdef INET6 882 struct ip6_hdr *ip6 = NULL; 883 struct icmp6_hdr *icmp6; 884#endif 885 src[0] = '\0'; 886 dst[0] = '\0'; 887#ifdef INET6 888 if (IS_IP6_FLOW_ID(&(args->f_id))) { 889 snprintf(src, sizeof(src), "[%s]", 890 ip6_sprintf(&args->f_id.src_ip6)); 891 snprintf(dst, sizeof(dst), "[%s]", 892 ip6_sprintf(&args->f_id.dst_ip6)); 893 894 ip6 = (struct ip6_hdr *)mtod(m, struct ip6_hdr *); 895 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 896 udp = (struct udphdr *)(mtod(args->m, char *) + hlen); 897 } else 898#endif 899 { 900 ip = mtod(m, struct ip *); 901 tcp = L3HDR(struct tcphdr, ip); 902 udp = L3HDR(struct udphdr, ip); 903 904 inet_ntoa_r(ip->ip_src, src); 905 inet_ntoa_r(ip->ip_dst, dst); 906 } 907 908 switch (args->f_id.proto) { 909 case IPPROTO_TCP: 910 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 911 if (offset == 0) 912 snprintf(SNPARGS(proto, len), ":%d %s:%d", 913 ntohs(tcp->th_sport), 914 dst, 915 ntohs(tcp->th_dport)); 916 else 917 snprintf(SNPARGS(proto, len), " %s", dst); 918 break; 919 920 case IPPROTO_UDP: 921 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 922 if (offset == 0) 923 snprintf(SNPARGS(proto, len), ":%d %s:%d", 924 ntohs(udp->uh_sport), 925 dst, 926 ntohs(udp->uh_dport)); 927 else 928 snprintf(SNPARGS(proto, len), " %s", dst); 929 break; 930 931 case IPPROTO_ICMP: 932 icmp = L3HDR(struct icmphdr, ip); 933 if (offset == 0) 934 len = snprintf(SNPARGS(proto, 0), 935 "ICMP:%u.%u ", 936 icmp->icmp_type, icmp->icmp_code); 937 else 938 len = snprintf(SNPARGS(proto, 0), "ICMP "); 939 len += snprintf(SNPARGS(proto, len), "%s", src); 940 snprintf(SNPARGS(proto, len), " %s", dst); 941 break; 942#ifdef INET6 943 case IPPROTO_ICMPV6: 944 icmp6 = (struct icmp6_hdr *)(mtod(args->m, char *) + hlen); 945 if (offset == 0) 946 len = snprintf(SNPARGS(proto, 0), 947 "ICMPv6:%u.%u ", 948 icmp6->icmp6_type, icmp6->icmp6_code); 949 else 950 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 951 len += snprintf(SNPARGS(proto, len), "%s", src); 952 snprintf(SNPARGS(proto, len), " %s", dst); 953 break; 954#endif 955 default: 956 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 957 args->f_id.proto, src); 958 snprintf(SNPARGS(proto, len), " %s", dst); 959 break; 960 } 961 962#ifdef INET6 963 if (IS_IP6_FLOW_ID(&(args->f_id))) { 964 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 965 snprintf(SNPARGS(fragment, 0), 966 " (frag %08x:%d@%d%s)", 967 args->f_id.frag_id6, 968 ntohs(ip6->ip6_plen) - hlen, 969 ntohs(offset & IP6F_OFF_MASK) << 3, 970 (offset & IP6F_MORE_FRAG) ? "+" : ""); 971 } else 972#endif 973 { 974 int ip_off, ip_len; 975 if (eh != NULL) { /* layer 2 packets are as on the wire */ 976 ip_off = ntohs(ip->ip_off); 977 ip_len = ntohs(ip->ip_len); 978 } else { 979 ip_off = ip->ip_off; 980 ip_len = ip->ip_len; 981 } 982 if (ip_off & (IP_MF | IP_OFFMASK)) 983 snprintf(SNPARGS(fragment, 0), 984 " (frag %d:%d@%d%s)", 985 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 986 offset << 3, 987 (ip_off & IP_MF) ? "+" : ""); 988 } 989 } 990 if (oif || m->m_pkthdr.rcvif) 991 log(LOG_SECURITY | LOG_INFO, 992 "ipfw: %d %s %s %s via %s%s\n", 993 f ? f->rulenum : -1, 994 action, proto, oif ? "out" : "in", 995 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 996 fragment); 997 else 998 log(LOG_SECURITY | LOG_INFO, 999 "ipfw: %d %s %s [no if info]%s\n", 1000 f ? f->rulenum : -1, 1001 action, proto, fragment); 1002 if (limit_reached) 1003 log(LOG_SECURITY | LOG_NOTICE, 1004 "ipfw: limit %d reached on entry %d\n", 1005 limit_reached, f ? f->rulenum : -1); 1006} 1007 1008/* 1009 * IMPORTANT: the hash function for dynamic rules must be commutative 1010 * in source and destination (ip,port), because rules are bidirectional 1011 * and we want to find both in the same bucket. 1012 */ 1013static __inline int 1014hash_packet(struct ipfw_flow_id *id) 1015{ 1016 u_int32_t i; 1017 1018#ifdef INET6 1019 if (IS_IP6_FLOW_ID(id)) 1020 i = hash_packet6(id); 1021 else 1022#endif /* INET6 */ 1023 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1024 i &= (curr_dyn_buckets - 1); 1025 return i; 1026} 1027 1028/** 1029 * unlink a dynamic rule from a chain. prev is a pointer to 1030 * the previous one, q is a pointer to the rule to delete, 1031 * head is a pointer to the head of the queue. 1032 * Modifies q and potentially also head. 1033 */ 1034#define UNLINK_DYN_RULE(prev, head, q) { \ 1035 ipfw_dyn_rule *old_q = q; \ 1036 \ 1037 /* remove a refcount to the parent */ \ 1038 if (q->dyn_type == O_LIMIT) \ 1039 q->parent->count--; \ 1040 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1041 (q->id.src_ip), (q->id.src_port), \ 1042 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 1043 if (prev != NULL) \ 1044 prev->next = q = q->next; \ 1045 else \ 1046 head = q = q->next; \ 1047 dyn_count--; \ 1048 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1049 1050#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1051 1052/** 1053 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1054 * 1055 * If keep_me == NULL, rules are deleted even if not expired, 1056 * otherwise only expired rules are removed. 1057 * 1058 * The value of the second parameter is also used to point to identify 1059 * a rule we absolutely do not want to remove (e.g. because we are 1060 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1061 * rules). The pointer is only used for comparison, so any non-null 1062 * value will do. 1063 */ 1064static void 1065remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1066{ 1067 static u_int32_t last_remove = 0; 1068 1069#define FORCE (keep_me == NULL) 1070 1071 ipfw_dyn_rule *prev, *q; 1072 int i, pass = 0, max_pass = 0; 1073 1074 IPFW_DYN_LOCK_ASSERT(); 1075 1076 if (ipfw_dyn_v == NULL || dyn_count == 0) 1077 return; 1078 /* do not expire more than once per second, it is useless */ 1079 if (!FORCE && last_remove == time_uptime) 1080 return; 1081 last_remove = time_uptime; 1082 1083 /* 1084 * because O_LIMIT refer to parent rules, during the first pass only 1085 * remove child and mark any pending LIMIT_PARENT, and remove 1086 * them in a second pass. 1087 */ 1088next_pass: 1089 for (i = 0 ; i < curr_dyn_buckets ; i++) { 1090 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 1091 /* 1092 * Logic can become complex here, so we split tests. 1093 */ 1094 if (q == keep_me) 1095 goto next; 1096 if (rule != NULL && rule != q->rule) 1097 goto next; /* not the one we are looking for */ 1098 if (q->dyn_type == O_LIMIT_PARENT) { 1099 /* 1100 * handle parent in the second pass, 1101 * record we need one. 1102 */ 1103 max_pass = 1; 1104 if (pass == 0) 1105 goto next; 1106 if (FORCE && q->count != 0 ) { 1107 /* XXX should not happen! */ 1108 printf("ipfw: OUCH! cannot remove rule," 1109 " count %d\n", q->count); 1110 } 1111 } else { 1112 if (!FORCE && 1113 !TIME_LEQ( q->expire, time_uptime )) 1114 goto next; 1115 } 1116 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1117 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1118 continue; 1119 } 1120next: 1121 prev=q; 1122 q=q->next; 1123 } 1124 } 1125 if (pass++ < max_pass) 1126 goto next_pass; 1127} 1128 1129 1130/** 1131 * lookup a dynamic rule. 1132 */ 1133static ipfw_dyn_rule * 1134lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1135 struct tcphdr *tcp) 1136{ 1137 /* 1138 * stateful ipfw extensions. 1139 * Lookup into dynamic session queue 1140 */ 1141#define MATCH_REVERSE 0 1142#define MATCH_FORWARD 1 1143#define MATCH_NONE 2 1144#define MATCH_UNKNOWN 3 1145 int i, dir = MATCH_NONE; 1146 ipfw_dyn_rule *prev, *q=NULL; 1147 1148 IPFW_DYN_LOCK_ASSERT(); 1149 1150 if (ipfw_dyn_v == NULL) 1151 goto done; /* not found */ 1152 i = hash_packet( pkt ); 1153 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 1154 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1155 goto next; 1156 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1157 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1158 continue; 1159 } 1160 if (pkt->proto == q->id.proto && 1161 q->dyn_type != O_LIMIT_PARENT) { 1162 if (IS_IP6_FLOW_ID(pkt)) { 1163 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1164 &(q->id.src_ip6)) && 1165 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1166 &(q->id.dst_ip6)) && 1167 pkt->src_port == q->id.src_port && 1168 pkt->dst_port == q->id.dst_port ) { 1169 dir = MATCH_FORWARD; 1170 break; 1171 } 1172 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1173 &(q->id.dst_ip6)) && 1174 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1175 &(q->id.src_ip6)) && 1176 pkt->src_port == q->id.dst_port && 1177 pkt->dst_port == q->id.src_port ) { 1178 dir = MATCH_REVERSE; 1179 break; 1180 } 1181 } else { 1182 if (pkt->src_ip == q->id.src_ip && 1183 pkt->dst_ip == q->id.dst_ip && 1184 pkt->src_port == q->id.src_port && 1185 pkt->dst_port == q->id.dst_port ) { 1186 dir = MATCH_FORWARD; 1187 break; 1188 } 1189 if (pkt->src_ip == q->id.dst_ip && 1190 pkt->dst_ip == q->id.src_ip && 1191 pkt->src_port == q->id.dst_port && 1192 pkt->dst_port == q->id.src_port ) { 1193 dir = MATCH_REVERSE; 1194 break; 1195 } 1196 } 1197 } 1198next: 1199 prev = q; 1200 q = q->next; 1201 } 1202 if (q == NULL) 1203 goto done; /* q = NULL, not found */ 1204 1205 if ( prev != NULL) { /* found and not in front */ 1206 prev->next = q->next; 1207 q->next = ipfw_dyn_v[i]; 1208 ipfw_dyn_v[i] = q; 1209 } 1210 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1211 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1212 1213#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1214#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1215 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1216 switch (q->state) { 1217 case TH_SYN: /* opening */ 1218 q->expire = time_uptime + dyn_syn_lifetime; 1219 break; 1220 1221 case BOTH_SYN: /* move to established */ 1222 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1223 case BOTH_SYN | (TH_FIN << 8) : 1224 if (tcp) { 1225#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1226 u_int32_t ack = ntohl(tcp->th_ack); 1227 if (dir == MATCH_FORWARD) { 1228 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1229 q->ack_fwd = ack; 1230 else { /* ignore out-of-sequence */ 1231 break; 1232 } 1233 } else { 1234 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1235 q->ack_rev = ack; 1236 else { /* ignore out-of-sequence */ 1237 break; 1238 } 1239 } 1240 } 1241 q->expire = time_uptime + dyn_ack_lifetime; 1242 break; 1243 1244 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1245 if (dyn_fin_lifetime >= dyn_keepalive_period) 1246 dyn_fin_lifetime = dyn_keepalive_period - 1; 1247 q->expire = time_uptime + dyn_fin_lifetime; 1248 break; 1249 1250 default: 1251#if 0 1252 /* 1253 * reset or some invalid combination, but can also 1254 * occur if we use keep-state the wrong way. 1255 */ 1256 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1257 printf("invalid state: 0x%x\n", q->state); 1258#endif 1259 if (dyn_rst_lifetime >= dyn_keepalive_period) 1260 dyn_rst_lifetime = dyn_keepalive_period - 1; 1261 q->expire = time_uptime + dyn_rst_lifetime; 1262 break; 1263 } 1264 } else if (pkt->proto == IPPROTO_UDP) { 1265 q->expire = time_uptime + dyn_udp_lifetime; 1266 } else { 1267 /* other protocols */ 1268 q->expire = time_uptime + dyn_short_lifetime; 1269 } 1270done: 1271 if (match_direction) 1272 *match_direction = dir; 1273 return q; 1274} 1275 1276static ipfw_dyn_rule * 1277lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1278 struct tcphdr *tcp) 1279{ 1280 ipfw_dyn_rule *q; 1281 1282 IPFW_DYN_LOCK(); 1283 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1284 if (q == NULL) 1285 IPFW_DYN_UNLOCK(); 1286 /* NB: return table locked when q is not NULL */ 1287 return q; 1288} 1289 1290static void 1291realloc_dynamic_table(void) 1292{ 1293 IPFW_DYN_LOCK_ASSERT(); 1294 1295 /* 1296 * Try reallocation, make sure we have a power of 2 and do 1297 * not allow more than 64k entries. In case of overflow, 1298 * default to 1024. 1299 */ 1300 1301 if (dyn_buckets > 65536) 1302 dyn_buckets = 1024; 1303 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1304 dyn_buckets = curr_dyn_buckets; /* reset */ 1305 return; 1306 } 1307 curr_dyn_buckets = dyn_buckets; 1308 if (ipfw_dyn_v != NULL) 1309 free(ipfw_dyn_v, M_IPFW); 1310 for (;;) { 1311 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1312 M_IPFW, M_NOWAIT | M_ZERO); 1313 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 1314 break; 1315 curr_dyn_buckets /= 2; 1316 } 1317} 1318 1319/** 1320 * Install state of type 'type' for a dynamic session. 1321 * The hash table contains two type of rules: 1322 * - regular rules (O_KEEP_STATE) 1323 * - rules for sessions with limited number of sess per user 1324 * (O_LIMIT). When they are created, the parent is 1325 * increased by 1, and decreased on delete. In this case, 1326 * the third parameter is the parent rule and not the chain. 1327 * - "parent" rules for the above (O_LIMIT_PARENT). 1328 */ 1329static ipfw_dyn_rule * 1330add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1331{ 1332 ipfw_dyn_rule *r; 1333 int i; 1334 1335 IPFW_DYN_LOCK_ASSERT(); 1336 1337 if (ipfw_dyn_v == NULL || 1338 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1339 realloc_dynamic_table(); 1340 if (ipfw_dyn_v == NULL) 1341 return NULL; /* failed ! */ 1342 } 1343 i = hash_packet(id); 1344 1345 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1346 if (r == NULL) { 1347 printf ("ipfw: sorry cannot allocate state\n"); 1348 return NULL; 1349 } 1350 1351 /* increase refcount on parent, and set pointer */ 1352 if (dyn_type == O_LIMIT) { 1353 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1354 if ( parent->dyn_type != O_LIMIT_PARENT) 1355 panic("invalid parent"); 1356 parent->count++; 1357 r->parent = parent; 1358 rule = parent->rule; 1359 } 1360 1361 r->id = *id; 1362 r->expire = time_uptime + dyn_syn_lifetime; 1363 r->rule = rule; 1364 r->dyn_type = dyn_type; 1365 r->pcnt = r->bcnt = 0; 1366 r->count = 0; 1367 1368 r->bucket = i; 1369 r->next = ipfw_dyn_v[i]; 1370 ipfw_dyn_v[i] = r; 1371 dyn_count++; 1372 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1373 dyn_type, 1374 (r->id.src_ip), (r->id.src_port), 1375 (r->id.dst_ip), (r->id.dst_port), 1376 dyn_count ); ) 1377 return r; 1378} 1379 1380/** 1381 * lookup dynamic parent rule using pkt and rule as search keys. 1382 * If the lookup fails, then install one. 1383 */ 1384static ipfw_dyn_rule * 1385lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1386{ 1387 ipfw_dyn_rule *q; 1388 int i; 1389 1390 IPFW_DYN_LOCK_ASSERT(); 1391 1392 if (ipfw_dyn_v) { 1393 int is_v6 = IS_IP6_FLOW_ID(pkt); 1394 i = hash_packet( pkt ); 1395 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1396 if (q->dyn_type == O_LIMIT_PARENT && 1397 rule== q->rule && 1398 pkt->proto == q->id.proto && 1399 pkt->src_port == q->id.src_port && 1400 pkt->dst_port == q->id.dst_port && 1401 ( 1402 (is_v6 && 1403 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1404 &(q->id.src_ip6)) && 1405 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1406 &(q->id.dst_ip6))) || 1407 (!is_v6 && 1408 pkt->src_ip == q->id.src_ip && 1409 pkt->dst_ip == q->id.dst_ip) 1410 ) 1411 ) { 1412 q->expire = time_uptime + dyn_short_lifetime; 1413 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1414 return q; 1415 } 1416 } 1417 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1418} 1419 1420/** 1421 * Install dynamic state for rule type cmd->o.opcode 1422 * 1423 * Returns 1 (failure) if state is not installed because of errors or because 1424 * session limitations are enforced. 1425 */ 1426static int 1427install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1428 struct ip_fw_args *args, uint32_t tablearg) 1429{ 1430 static int last_log; 1431 ipfw_dyn_rule *q; 1432 struct in_addr da; 1433 char src[48], dst[48]; 1434 1435 src[0] = '\0'; 1436 dst[0] = '\0'; 1437 1438 DEB( 1439 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1440 __func__, cmd->o.opcode, 1441 (args->f_id.src_ip), (args->f_id.src_port), 1442 (args->f_id.dst_ip), (args->f_id.dst_port)); 1443 ) 1444 1445 IPFW_DYN_LOCK(); 1446 1447 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1448 1449 if (q != NULL) { /* should never occur */ 1450 if (last_log != time_uptime) { 1451 last_log = time_uptime; 1452 printf("ipfw: %s: entry already present, done\n", 1453 __func__); 1454 } 1455 IPFW_DYN_UNLOCK(); 1456 return (0); 1457 } 1458 1459 if (dyn_count >= dyn_max) 1460 /* Run out of slots, try to remove any expired rule. */ 1461 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1462 1463 if (dyn_count >= dyn_max) { 1464 if (last_log != time_uptime) { 1465 last_log = time_uptime; 1466 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1467 } 1468 IPFW_DYN_UNLOCK(); 1469 return (1); /* cannot install, notify caller */ 1470 } 1471 1472 switch (cmd->o.opcode) { 1473 case O_KEEP_STATE: /* bidir rule */ 1474 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1475 break; 1476 1477 case O_LIMIT: { /* limit number of sessions */ 1478 struct ipfw_flow_id id; 1479 ipfw_dyn_rule *parent; 1480 uint32_t conn_limit; 1481 uint16_t limit_mask = cmd->limit_mask; 1482 1483 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1484 tablearg : cmd->conn_limit; 1485 1486 DEB( 1487 if (cmd->conn_limit == IP_FW_TABLEARG) 1488 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1489 "(tablearg)\n", __func__, conn_limit); 1490 else 1491 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1492 __func__, conn_limit); 1493 ) 1494 1495 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1496 id.proto = args->f_id.proto; 1497 id.addr_type = args->f_id.addr_type; 1498 1499 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1500 if (limit_mask & DYN_SRC_ADDR) 1501 id.src_ip6 = args->f_id.src_ip6; 1502 if (limit_mask & DYN_DST_ADDR) 1503 id.dst_ip6 = args->f_id.dst_ip6; 1504 } else { 1505 if (limit_mask & DYN_SRC_ADDR) 1506 id.src_ip = args->f_id.src_ip; 1507 if (limit_mask & DYN_DST_ADDR) 1508 id.dst_ip = args->f_id.dst_ip; 1509 } 1510 if (limit_mask & DYN_SRC_PORT) 1511 id.src_port = args->f_id.src_port; 1512 if (limit_mask & DYN_DST_PORT) 1513 id.dst_port = args->f_id.dst_port; 1514 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1515 printf("ipfw: %s: add parent failed\n", __func__); 1516 IPFW_DYN_UNLOCK(); 1517 return (1); 1518 } 1519 1520 if (parent->count >= conn_limit) { 1521 /* See if we can remove some expired rule. */ 1522 remove_dyn_rule(rule, parent); 1523 if (parent->count >= conn_limit) { 1524 if (fw_verbose && last_log != time_uptime) { 1525 last_log = time_uptime; 1526#ifdef INET6 1527 /* 1528 * XXX IPv6 flows are not 1529 * supported yet. 1530 */ 1531 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1532 snprintf(src, sizeof(src), 1533 "[%s]", ip6_sprintf( 1534 &args->f_id.src_ip6)); 1535 snprintf(dst, sizeof(dst), 1536 "[%s]", ip6_sprintf( 1537 &args->f_id.dst_ip6)); 1538 } else 1539#endif 1540 { 1541 da.s_addr = 1542 htonl(args->f_id.src_ip); 1543 inet_ntoa_r(da, src); 1544 da.s_addr = 1545 htonl(args->f_id.dst_ip); 1546 inet_ntoa_r(da, dst); 1547 } 1548 log(LOG_SECURITY | LOG_DEBUG, 1549 "%s %s:%u -> %s:%u, %s\n", 1550 "drop session", 1551 src, (args->f_id.src_port), 1552 dst, (args->f_id.dst_port), 1553 "too many entries"); 1554 } 1555 IPFW_DYN_UNLOCK(); 1556 return (1); 1557 } 1558 } 1559 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1560 break; 1561 } 1562 default: 1563 printf("ipfw: %s: unknown dynamic rule type %u\n", 1564 __func__, cmd->o.opcode); 1565 IPFW_DYN_UNLOCK(); 1566 return (1); 1567 } 1568 1569 /* XXX just set lifetime */ 1570 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1571 1572 IPFW_DYN_UNLOCK(); 1573 return (0); 1574} 1575 1576/* 1577 * Generate a TCP packet, containing either a RST or a keepalive. 1578 * When flags & TH_RST, we are sending a RST packet, because of a 1579 * "reset" action matched the packet. 1580 * Otherwise we are sending a keepalive, and flags & TH_ 1581 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 1582 * so that MAC can label the reply appropriately. 1583 */ 1584static struct mbuf * 1585send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 1586 u_int32_t ack, int flags) 1587{ 1588 struct mbuf *m; 1589 struct ip *ip; 1590 struct tcphdr *tcp; 1591 1592 MGETHDR(m, M_DONTWAIT, MT_DATA); 1593 if (m == 0) 1594 return (NULL); 1595 m->m_pkthdr.rcvif = (struct ifnet *)0; 1596 1597#ifdef MAC 1598 if (replyto != NULL) 1599 mac_create_mbuf_netlayer(replyto, m); 1600 else 1601 mac_create_mbuf_from_firewall(m); 1602#else 1603 (void)replyto; /* don't warn about unused arg */ 1604#endif 1605 1606 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1607 m->m_data += max_linkhdr; 1608 1609 ip = mtod(m, struct ip *); 1610 bzero(ip, m->m_len); 1611 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1612 ip->ip_p = IPPROTO_TCP; 1613 tcp->th_off = 5; 1614 /* 1615 * Assume we are sending a RST (or a keepalive in the reverse 1616 * direction), swap src and destination addresses and ports. 1617 */ 1618 ip->ip_src.s_addr = htonl(id->dst_ip); 1619 ip->ip_dst.s_addr = htonl(id->src_ip); 1620 tcp->th_sport = htons(id->dst_port); 1621 tcp->th_dport = htons(id->src_port); 1622 if (flags & TH_RST) { /* we are sending a RST */ 1623 if (flags & TH_ACK) { 1624 tcp->th_seq = htonl(ack); 1625 tcp->th_ack = htonl(0); 1626 tcp->th_flags = TH_RST; 1627 } else { 1628 if (flags & TH_SYN) 1629 seq++; 1630 tcp->th_seq = htonl(0); 1631 tcp->th_ack = htonl(seq); 1632 tcp->th_flags = TH_RST | TH_ACK; 1633 } 1634 } else { 1635 /* 1636 * We are sending a keepalive. flags & TH_SYN determines 1637 * the direction, forward if set, reverse if clear. 1638 * NOTE: seq and ack are always assumed to be correct 1639 * as set by the caller. This may be confusing... 1640 */ 1641 if (flags & TH_SYN) { 1642 /* 1643 * we have to rewrite the correct addresses! 1644 */ 1645 ip->ip_dst.s_addr = htonl(id->dst_ip); 1646 ip->ip_src.s_addr = htonl(id->src_ip); 1647 tcp->th_dport = htons(id->dst_port); 1648 tcp->th_sport = htons(id->src_port); 1649 } 1650 tcp->th_seq = htonl(seq); 1651 tcp->th_ack = htonl(ack); 1652 tcp->th_flags = TH_ACK; 1653 } 1654 /* 1655 * set ip_len to the payload size so we can compute 1656 * the tcp checksum on the pseudoheader 1657 * XXX check this, could save a couple of words ? 1658 */ 1659 ip->ip_len = htons(sizeof(struct tcphdr)); 1660 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1661 /* 1662 * now fill fields left out earlier 1663 */ 1664 ip->ip_ttl = ip_defttl; 1665 ip->ip_len = m->m_pkthdr.len; 1666 m->m_flags |= M_SKIP_FIREWALL; 1667 return (m); 1668} 1669 1670/* 1671 * sends a reject message, consuming the mbuf passed as an argument. 1672 */ 1673static void 1674send_reject(struct ip_fw_args *args, int code, int ip_len) 1675{ 1676 1677 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1678 /* We need the IP header in host order for icmp_error(). */ 1679 if (args->eh != NULL) { 1680 struct ip *ip = mtod(args->m, struct ip *); 1681 ip->ip_len = ntohs(ip->ip_len); 1682 ip->ip_off = ntohs(ip->ip_off); 1683 } 1684 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1685 } else if (args->f_id.proto == IPPROTO_TCP) { 1686 struct tcphdr *const tcp = 1687 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1688 if ( (tcp->th_flags & TH_RST) == 0) { 1689 struct mbuf *m; 1690 m = send_pkt(args->m, &(args->f_id), 1691 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1692 tcp->th_flags | TH_RST); 1693 if (m != NULL) 1694 ip_output(m, NULL, NULL, 0, NULL, NULL); 1695 } 1696 m_freem(args->m); 1697 } else 1698 m_freem(args->m); 1699 args->m = NULL; 1700} 1701 1702/** 1703 * 1704 * Given an ip_fw *, lookup_next_rule will return a pointer 1705 * to the next rule, which can be either the jump 1706 * target (for skipto instructions) or the next one in the list (in 1707 * all other cases including a missing jump target). 1708 * The result is also written in the "next_rule" field of the rule. 1709 * Backward jumps are not allowed, so start looking from the next 1710 * rule... 1711 * 1712 * This never returns NULL -- in case we do not have an exact match, 1713 * the next rule is returned. When the ruleset is changed, 1714 * pointers are flushed so we are always correct. 1715 */ 1716 1717static struct ip_fw * 1718lookup_next_rule(struct ip_fw *me) 1719{ 1720 struct ip_fw *rule = NULL; 1721 ipfw_insn *cmd; 1722 1723 /* look for action, in case it is a skipto */ 1724 cmd = ACTION_PTR(me); 1725 if (cmd->opcode == O_LOG) 1726 cmd += F_LEN(cmd); 1727 if (cmd->opcode == O_ALTQ) 1728 cmd += F_LEN(cmd); 1729 if (cmd->opcode == O_TAG) 1730 cmd += F_LEN(cmd); 1731 if ( cmd->opcode == O_SKIPTO ) 1732 for (rule = me->next; rule ; rule = rule->next) 1733 if (rule->rulenum >= cmd->arg1) 1734 break; 1735 if (rule == NULL) /* failure or not a skipto */ 1736 rule = me->next; 1737 me->next_rule = rule; 1738 return rule; 1739} 1740 1741static int 1742add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1743 uint8_t mlen, uint32_t value) 1744{ 1745 struct radix_node_head *rnh; 1746 struct table_entry *ent; 1747 1748 if (tbl >= IPFW_TABLES_MAX) 1749 return (EINVAL); 1750 rnh = ch->tables[tbl]; 1751 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1752 if (ent == NULL) 1753 return (ENOMEM); 1754 ent->value = value; 1755 ent->addr.sin_len = ent->mask.sin_len = 8; 1756 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1757 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1758 IPFW_WLOCK(&layer3_chain); 1759 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1760 NULL) { 1761 IPFW_WUNLOCK(&layer3_chain); 1762 free(ent, M_IPFW_TBL); 1763 return (EEXIST); 1764 } 1765 IPFW_WUNLOCK(&layer3_chain); 1766 return (0); 1767} 1768 1769static int 1770del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1771 uint8_t mlen) 1772{ 1773 struct radix_node_head *rnh; 1774 struct table_entry *ent; 1775 struct sockaddr_in sa, mask; 1776 1777 if (tbl >= IPFW_TABLES_MAX) 1778 return (EINVAL); 1779 rnh = ch->tables[tbl]; 1780 sa.sin_len = mask.sin_len = 8; 1781 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1782 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1783 IPFW_WLOCK(ch); 1784 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1785 if (ent == NULL) { 1786 IPFW_WUNLOCK(ch); 1787 return (ESRCH); 1788 } 1789 IPFW_WUNLOCK(ch); 1790 free(ent, M_IPFW_TBL); 1791 return (0); 1792} 1793 1794static int 1795flush_table_entry(struct radix_node *rn, void *arg) 1796{ 1797 struct radix_node_head * const rnh = arg; 1798 struct table_entry *ent; 1799 1800 ent = (struct table_entry *) 1801 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1802 if (ent != NULL) 1803 free(ent, M_IPFW_TBL); 1804 return (0); 1805} 1806 1807static int 1808flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1809{ 1810 struct radix_node_head *rnh; 1811 1812 IPFW_WLOCK_ASSERT(ch); 1813 1814 if (tbl >= IPFW_TABLES_MAX) 1815 return (EINVAL); 1816 rnh = ch->tables[tbl]; 1817 KASSERT(rnh != NULL, ("NULL IPFW table")); 1818 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1819 return (0); 1820} 1821 1822static void 1823flush_tables(struct ip_fw_chain *ch) 1824{ 1825 uint16_t tbl; 1826 1827 IPFW_WLOCK_ASSERT(ch); 1828 1829 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1830 flush_table(ch, tbl); 1831} 1832 1833static int 1834init_tables(struct ip_fw_chain *ch) 1835{ 1836 int i; 1837 uint16_t j; 1838 1839 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1840 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1841 for (j = 0; j < i; j++) { 1842 (void) flush_table(ch, j); 1843 } 1844 return (ENOMEM); 1845 } 1846 } 1847 return (0); 1848} 1849 1850static int 1851lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1852 uint32_t *val) 1853{ 1854 struct radix_node_head *rnh; 1855 struct table_entry *ent; 1856 struct sockaddr_in sa; 1857 1858 if (tbl >= IPFW_TABLES_MAX) 1859 return (0); 1860 rnh = ch->tables[tbl]; 1861 sa.sin_len = 8; 1862 sa.sin_addr.s_addr = addr; 1863 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1864 if (ent != NULL) { 1865 *val = ent->value; 1866 return (1); 1867 } 1868 return (0); 1869} 1870 1871static int 1872count_table_entry(struct radix_node *rn, void *arg) 1873{ 1874 u_int32_t * const cnt = arg; 1875 1876 (*cnt)++; 1877 return (0); 1878} 1879 1880static int 1881count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1882{ 1883 struct radix_node_head *rnh; 1884 1885 if (tbl >= IPFW_TABLES_MAX) 1886 return (EINVAL); 1887 rnh = ch->tables[tbl]; 1888 *cnt = 0; 1889 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1890 return (0); 1891} 1892 1893static int 1894dump_table_entry(struct radix_node *rn, void *arg) 1895{ 1896 struct table_entry * const n = (struct table_entry *)rn; 1897 ipfw_table * const tbl = arg; 1898 ipfw_table_entry *ent; 1899 1900 if (tbl->cnt == tbl->size) 1901 return (1); 1902 ent = &tbl->ent[tbl->cnt]; 1903 ent->tbl = tbl->tbl; 1904 if (in_nullhost(n->mask.sin_addr)) 1905 ent->masklen = 0; 1906 else 1907 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1908 ent->addr = n->addr.sin_addr.s_addr; 1909 ent->value = n->value; 1910 tbl->cnt++; 1911 return (0); 1912} 1913 1914static int 1915dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1916{ 1917 struct radix_node_head *rnh; 1918 1919 if (tbl->tbl >= IPFW_TABLES_MAX) 1920 return (EINVAL); 1921 rnh = ch->tables[tbl->tbl]; 1922 tbl->cnt = 0; 1923 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1924 return (0); 1925} 1926 1927static void 1928fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1929{ 1930 struct ucred *cr; 1931 1932 if (inp->inp_socket != NULL) { 1933 cr = inp->inp_socket->so_cred; 1934 ugp->fw_prid = jailed(cr) ? 1935 cr->cr_prison->pr_id : -1; 1936 ugp->fw_uid = cr->cr_uid; 1937 ugp->fw_ngroups = cr->cr_ngroups; 1938 bcopy(cr->cr_groups, ugp->fw_groups, 1939 sizeof(ugp->fw_groups)); 1940 } 1941} 1942 1943static int 1944check_uidgid(ipfw_insn_u32 *insn, 1945 int proto, struct ifnet *oif, 1946 struct in_addr dst_ip, u_int16_t dst_port, 1947 struct in_addr src_ip, u_int16_t src_port, 1948 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 1949{ 1950 struct inpcbinfo *pi; 1951 int wildcard; 1952 struct inpcb *pcb; 1953 int match; 1954 gid_t *gp; 1955 1956 /* 1957 * Check to see if the UDP or TCP stack supplied us with 1958 * the PCB. If so, rather then holding a lock and looking 1959 * up the PCB, we can use the one that was supplied. 1960 */ 1961 if (inp && *lookup == 0) { 1962 INP_LOCK_ASSERT(inp); 1963 if (inp->inp_socket != NULL) { 1964 fill_ugid_cache(inp, ugp); 1965 *lookup = 1; 1966 } 1967 } 1968 /* 1969 * If we have already been here and the packet has no 1970 * PCB entry associated with it, then we can safely 1971 * assume that this is a no match. 1972 */ 1973 if (*lookup == -1) 1974 return (0); 1975 if (proto == IPPROTO_TCP) { 1976 wildcard = 0; 1977 pi = &tcbinfo; 1978 } else if (proto == IPPROTO_UDP) { 1979 wildcard = INPLOOKUP_WILDCARD; 1980 pi = &udbinfo; 1981 } else 1982 return 0; 1983 match = 0; 1984 if (*lookup == 0) { 1985 INP_INFO_RLOCK(pi); 1986 pcb = (oif) ? 1987 in_pcblookup_hash(pi, 1988 dst_ip, htons(dst_port), 1989 src_ip, htons(src_port), 1990 wildcard, oif) : 1991 in_pcblookup_hash(pi, 1992 src_ip, htons(src_port), 1993 dst_ip, htons(dst_port), 1994 wildcard, NULL); 1995 if (pcb != NULL) { 1996 INP_LOCK(pcb); 1997 if (pcb->inp_socket != NULL) { 1998 fill_ugid_cache(pcb, ugp); 1999 *lookup = 1; 2000 } 2001 INP_UNLOCK(pcb); 2002 } 2003 INP_INFO_RUNLOCK(pi); 2004 if (*lookup == 0) { 2005 /* 2006 * If the lookup did not yield any results, there 2007 * is no sense in coming back and trying again. So 2008 * we can set lookup to -1 and ensure that we wont 2009 * bother the pcb system again. 2010 */ 2011 *lookup = -1; 2012 return (0); 2013 } 2014 } 2015 if (insn->o.opcode == O_UID) 2016 match = (ugp->fw_uid == (uid_t)insn->d[0]); 2017 else if (insn->o.opcode == O_GID) { 2018 for (gp = ugp->fw_groups; 2019 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 2020 if (*gp == (gid_t)insn->d[0]) { 2021 match = 1; 2022 break; 2023 } 2024 } else if (insn->o.opcode == O_JAIL) 2025 match = (ugp->fw_prid == (int)insn->d[0]); 2026 return match; 2027} 2028 2029/* 2030 * The main check routine for the firewall. 2031 * 2032 * All arguments are in args so we can modify them and return them 2033 * back to the caller. 2034 * 2035 * Parameters: 2036 * 2037 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 2038 * Starts with the IP header. 2039 * args->eh (in) Mac header if present, or NULL for layer3 packet. 2040 * args->oif Outgoing interface, or NULL if packet is incoming. 2041 * The incoming interface is in the mbuf. (in) 2042 * args->divert_rule (in/out) 2043 * Skip up to the first rule past this rule number; 2044 * upon return, non-zero port number for divert or tee. 2045 * 2046 * args->rule Pointer to the last matching rule (in/out) 2047 * args->next_hop Socket we are forwarding to (out). 2048 * args->f_id Addresses grabbed from the packet (out) 2049 * args->cookie a cookie depending on rule action 2050 * 2051 * Return value: 2052 * 2053 * IP_FW_PASS the packet must be accepted 2054 * IP_FW_DENY the packet must be dropped 2055 * IP_FW_DIVERT divert packet, port in m_tag 2056 * IP_FW_TEE tee packet, port in m_tag 2057 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2058 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2059 * 2060 */ 2061 2062int 2063ipfw_chk(struct ip_fw_args *args) 2064{ 2065 /* 2066 * Local variables hold state during the processing of a packet. 2067 * 2068 * IMPORTANT NOTE: to speed up the processing of rules, there 2069 * are some assumption on the values of the variables, which 2070 * are documented here. Should you change them, please check 2071 * the implementation of the various instructions to make sure 2072 * that they still work. 2073 * 2074 * args->eh The MAC header. It is non-null for a layer2 2075 * packet, it is NULL for a layer-3 packet. 2076 * 2077 * m | args->m Pointer to the mbuf, as received from the caller. 2078 * It may change if ipfw_chk() does an m_pullup, or if it 2079 * consumes the packet because it calls send_reject(). 2080 * XXX This has to change, so that ipfw_chk() never modifies 2081 * or consumes the buffer. 2082 * ip is simply an alias of the value of m, and it is kept 2083 * in sync with it (the packet is supposed to start with 2084 * the ip header). 2085 */ 2086 struct mbuf *m = args->m; 2087 struct ip *ip = mtod(m, struct ip *); 2088 2089 /* 2090 * For rules which contain uid/gid or jail constraints, cache 2091 * a copy of the users credentials after the pcb lookup has been 2092 * executed. This will speed up the processing of rules with 2093 * these types of constraints, as well as decrease contention 2094 * on pcb related locks. 2095 */ 2096 struct ip_fw_ugid fw_ugid_cache; 2097 int ugid_lookup = 0; 2098 2099 /* 2100 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2101 * associated with a packet input on a divert socket. This 2102 * will allow to distinguish traffic and its direction when 2103 * it originates from a divert socket. 2104 */ 2105 u_int divinput_flags = 0; 2106 2107 /* 2108 * oif | args->oif If NULL, ipfw_chk has been called on the 2109 * inbound path (ether_input, ip_input). 2110 * If non-NULL, ipfw_chk has been called on the outbound path 2111 * (ether_output, ip_output). 2112 */ 2113 struct ifnet *oif = args->oif; 2114 2115 struct ip_fw *f = NULL; /* matching rule */ 2116 int retval = 0; 2117 2118 /* 2119 * hlen The length of the IP header. 2120 */ 2121 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2122 2123 /* 2124 * offset The offset of a fragment. offset != 0 means that 2125 * we have a fragment at this offset of an IPv4 packet. 2126 * offset == 0 means that (if this is an IPv4 packet) 2127 * this is the first or only fragment. 2128 * For IPv6 offset == 0 means there is no Fragment Header. 2129 * If offset != 0 for IPv6 always use correct mask to 2130 * get the correct offset because we add IP6F_MORE_FRAG 2131 * to be able to dectect the first fragment which would 2132 * otherwise have offset = 0. 2133 */ 2134 u_short offset = 0; 2135 2136 /* 2137 * Local copies of addresses. They are only valid if we have 2138 * an IP packet. 2139 * 2140 * proto The protocol. Set to 0 for non-ip packets, 2141 * or to the protocol read from the packet otherwise. 2142 * proto != 0 means that we have an IPv4 packet. 2143 * 2144 * src_port, dst_port port numbers, in HOST format. Only 2145 * valid for TCP and UDP packets. 2146 * 2147 * src_ip, dst_ip ip addresses, in NETWORK format. 2148 * Only valid for IPv4 packets. 2149 */ 2150 u_int8_t proto; 2151 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2152 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2153 u_int16_t ip_len=0; 2154 int pktlen; 2155 2156 /* 2157 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2158 * MATCH_NONE when checked and not matched (q = NULL), 2159 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2160 */ 2161 int dyn_dir = MATCH_UNKNOWN; 2162 ipfw_dyn_rule *q = NULL; 2163 struct ip_fw_chain *chain = &layer3_chain; 2164 struct m_tag *mtag; 2165 2166 /* 2167 * We store in ulp a pointer to the upper layer protocol header. 2168 * In the ipv4 case this is easy to determine from the header, 2169 * but for ipv6 we might have some additional headers in the middle. 2170 * ulp is NULL if not found. 2171 */ 2172 void *ulp = NULL; /* upper layer protocol pointer. */ 2173 /* XXX ipv6 variables */ 2174 int is_ipv6 = 0; 2175 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2176 /* end of ipv6 variables */ 2177 int is_ipv4 = 0; 2178 2179 if (m->m_flags & M_SKIP_FIREWALL) 2180 return (IP_FW_PASS); /* accept */ 2181 2182 pktlen = m->m_pkthdr.len; 2183 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2184 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2185 2186/* 2187 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2188 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2189 * pointer might become stale after other pullups (but we never use it 2190 * this way). 2191 */ 2192#define PULLUP_TO(len, p, T) \ 2193do { \ 2194 int x = (len) + sizeof(T); \ 2195 if ((m)->m_len < x) { \ 2196 args->m = m = m_pullup(m, x); \ 2197 if (m == NULL) \ 2198 goto pullup_failed; \ 2199 } \ 2200 p = (mtod(m, char *) + (len)); \ 2201} while (0) 2202 2203 /* Identify IP packets and fill up variables. */ 2204 if (pktlen >= sizeof(struct ip6_hdr) && 2205 (args->eh == NULL || ntohs(args->eh->ether_type)==ETHERTYPE_IPV6) && 2206 mtod(m, struct ip *)->ip_v == 6) { 2207 is_ipv6 = 1; 2208 args->f_id.addr_type = 6; 2209 hlen = sizeof(struct ip6_hdr); 2210 proto = mtod(m, struct ip6_hdr *)->ip6_nxt; 2211 2212 /* Search extension headers to find upper layer protocols */ 2213 while (ulp == NULL) { 2214 switch (proto) { 2215 case IPPROTO_ICMPV6: 2216 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2217 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2218 break; 2219 2220 case IPPROTO_TCP: 2221 PULLUP_TO(hlen, ulp, struct tcphdr); 2222 dst_port = TCP(ulp)->th_dport; 2223 src_port = TCP(ulp)->th_sport; 2224 args->f_id.flags = TCP(ulp)->th_flags; 2225 break; 2226 2227 case IPPROTO_SCTP: 2228 PULLUP_TO(hlen, ulp, struct sctphdr); 2229 src_port = SCTP(ulp)->src_port; 2230 dst_port = SCTP(ulp)->dest_port; 2231 break; 2232 2233 case IPPROTO_UDP: 2234 PULLUP_TO(hlen, ulp, struct udphdr); 2235 dst_port = UDP(ulp)->uh_dport; 2236 src_port = UDP(ulp)->uh_sport; 2237 break; 2238 2239 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2240 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2241 ext_hd |= EXT_HOPOPTS; 2242 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2243 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2244 ulp = NULL; 2245 break; 2246 2247 case IPPROTO_ROUTING: /* RFC 2460 */ 2248 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2249 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2250 case 0: 2251 break; 2252 default: 2253 printf("IPFW2: IPV6 - Unknown Routing " 2254 "Header type(%d)\n", 2255 ((struct ip6_rthdr *)ulp)->ip6r_type); 2256 if (fw_deny_unknown_exthdrs) 2257 return (IP_FW_DENY); 2258 break; 2259 } 2260 ext_hd |= EXT_ROUTING; 2261 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2262 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2263 ulp = NULL; 2264 break; 2265 2266 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2267 PULLUP_TO(hlen, ulp, struct ip6_frag); 2268 ext_hd |= EXT_FRAGMENT; 2269 hlen += sizeof (struct ip6_frag); 2270 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2271 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2272 IP6F_OFF_MASK; 2273 /* Add IP6F_MORE_FRAG for offset of first 2274 * fragment to be != 0. */ 2275 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2276 IP6F_MORE_FRAG; 2277 if (offset == 0) { 2278 printf("IPFW2: IPV6 - Invalid Fragment " 2279 "Header\n"); 2280 if (fw_deny_unknown_exthdrs) 2281 return (IP_FW_DENY); 2282 break; 2283 } 2284 args->f_id.frag_id6 = 2285 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2286 ulp = NULL; 2287 break; 2288 2289 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2290 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2291 ext_hd |= EXT_DSTOPTS; 2292 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2293 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2294 ulp = NULL; 2295 break; 2296 2297 case IPPROTO_AH: /* RFC 2402 */ 2298 PULLUP_TO(hlen, ulp, struct ip6_ext); 2299 ext_hd |= EXT_AH; 2300 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2301 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2302 ulp = NULL; 2303 break; 2304 2305 case IPPROTO_ESP: /* RFC 2406 */ 2306 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2307 /* Anything past Seq# is variable length and 2308 * data past this ext. header is encrypted. */ 2309 ext_hd |= EXT_ESP; 2310 break; 2311 2312 case IPPROTO_NONE: /* RFC 2460 */ 2313 PULLUP_TO(hlen, ulp, struct ip6_ext); 2314 /* Packet ends here. if ip6e_len!=0 octets 2315 * must be ignored. */ 2316 break; 2317 2318 case IPPROTO_OSPFIGP: 2319 /* XXX OSPF header check? */ 2320 PULLUP_TO(hlen, ulp, struct ip6_ext); 2321 break; 2322 2323 case IPPROTO_PIM: 2324 /* XXX PIM header check? */ 2325 PULLUP_TO(hlen, ulp, struct pim); 2326 break; 2327 2328 case IPPROTO_CARP: 2329 PULLUP_TO(hlen, ulp, struct carp_header); 2330 if (((struct carp_header *)ulp)->carp_version != 2331 CARP_VERSION) 2332 return (IP_FW_DENY); 2333 if (((struct carp_header *)ulp)->carp_type != 2334 CARP_ADVERTISEMENT) 2335 return (IP_FW_DENY); 2336 break; 2337 2338 case IPPROTO_IPV6: /* RFC 2893 */ 2339 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2340 break; 2341 2342 case IPPROTO_IPV4: /* RFC 2893 */ 2343 PULLUP_TO(hlen, ulp, struct ip); 2344 break; 2345 2346 default: 2347 printf("IPFW2: IPV6 - Unknown Extension " 2348 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2349 if (fw_deny_unknown_exthdrs) 2350 return (IP_FW_DENY); 2351 PULLUP_TO(hlen, ulp, struct ip6_ext); 2352 break; 2353 } /*switch */ 2354 } 2355 args->f_id.src_ip6 = mtod(m,struct ip6_hdr *)->ip6_src; 2356 args->f_id.dst_ip6 = mtod(m,struct ip6_hdr *)->ip6_dst; 2357 args->f_id.src_ip = 0; 2358 args->f_id.dst_ip = 0; 2359 args->f_id.flow_id6 = ntohl(mtod(m, struct ip6_hdr *)->ip6_flow); 2360 } else if (pktlen >= sizeof(struct ip) && 2361 (args->eh == NULL || ntohs(args->eh->ether_type) == ETHERTYPE_IP) && 2362 mtod(m, struct ip *)->ip_v == 4) { 2363 is_ipv4 = 1; 2364 ip = mtod(m, struct ip *); 2365 hlen = ip->ip_hl << 2; 2366 args->f_id.addr_type = 4; 2367 2368 /* 2369 * Collect parameters into local variables for faster matching. 2370 */ 2371 proto = ip->ip_p; 2372 src_ip = ip->ip_src; 2373 dst_ip = ip->ip_dst; 2374 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2375 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2376 ip_len = ntohs(ip->ip_len); 2377 } else { 2378 offset = ip->ip_off & IP_OFFMASK; 2379 ip_len = ip->ip_len; 2380 } 2381 pktlen = ip_len < pktlen ? ip_len : pktlen; 2382 2383 if (offset == 0) { 2384 switch (proto) { 2385 case IPPROTO_TCP: 2386 PULLUP_TO(hlen, ulp, struct tcphdr); 2387 dst_port = TCP(ulp)->th_dport; 2388 src_port = TCP(ulp)->th_sport; 2389 args->f_id.flags = TCP(ulp)->th_flags; 2390 break; 2391 2392 case IPPROTO_UDP: 2393 PULLUP_TO(hlen, ulp, struct udphdr); 2394 dst_port = UDP(ulp)->uh_dport; 2395 src_port = UDP(ulp)->uh_sport; 2396 break; 2397 2398 case IPPROTO_ICMP: 2399 PULLUP_TO(hlen, ulp, struct icmphdr); 2400 args->f_id.flags = ICMP(ulp)->icmp_type; 2401 break; 2402 2403 default: 2404 break; 2405 } 2406 } 2407 2408 args->f_id.src_ip = ntohl(src_ip.s_addr); 2409 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2410 } 2411#undef PULLUP_TO 2412 if (proto) { /* we may have port numbers, store them */ 2413 args->f_id.proto = proto; 2414 args->f_id.src_port = src_port = ntohs(src_port); 2415 args->f_id.dst_port = dst_port = ntohs(dst_port); 2416 } 2417 2418 IPFW_RLOCK(chain); 2419 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2420 if (args->rule) { 2421 /* 2422 * Packet has already been tagged. Look for the next rule 2423 * to restart processing. 2424 * 2425 * If fw_one_pass != 0 then just accept it. 2426 * XXX should not happen here, but optimized out in 2427 * the caller. 2428 */ 2429 if (fw_one_pass) { 2430 IPFW_RUNLOCK(chain); 2431 return (IP_FW_PASS); 2432 } 2433 2434 f = args->rule->next_rule; 2435 if (f == NULL) 2436 f = lookup_next_rule(args->rule); 2437 } else { 2438 /* 2439 * Find the starting rule. It can be either the first 2440 * one, or the one after divert_rule if asked so. 2441 */ 2442 int skipto = mtag ? divert_cookie(mtag) : 0; 2443 2444 f = chain->rules; 2445 if (args->eh == NULL && skipto != 0) { 2446 if (skipto >= IPFW_DEFAULT_RULE) { 2447 IPFW_RUNLOCK(chain); 2448 return (IP_FW_DENY); /* invalid */ 2449 } 2450 while (f && f->rulenum <= skipto) 2451 f = f->next; 2452 if (f == NULL) { /* drop packet */ 2453 IPFW_RUNLOCK(chain); 2454 return (IP_FW_DENY); 2455 } 2456 } 2457 } 2458 /* reset divert rule to avoid confusion later */ 2459 if (mtag) { 2460 divinput_flags = divert_info(mtag) & 2461 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2462 m_tag_delete(m, mtag); 2463 } 2464 2465 /* 2466 * Now scan the rules, and parse microinstructions for each rule. 2467 */ 2468 for (; f; f = f->next) { 2469 ipfw_insn *cmd; 2470 uint32_t tablearg = 0; 2471 int l, cmdlen, skip_or; /* skip rest of OR block */ 2472 2473again: 2474 if (set_disable & (1 << f->set) ) 2475 continue; 2476 2477 skip_or = 0; 2478 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2479 l -= cmdlen, cmd += cmdlen) { 2480 int match; 2481 2482 /* 2483 * check_body is a jump target used when we find a 2484 * CHECK_STATE, and need to jump to the body of 2485 * the target rule. 2486 */ 2487 2488check_body: 2489 cmdlen = F_LEN(cmd); 2490 /* 2491 * An OR block (insn_1 || .. || insn_n) has the 2492 * F_OR bit set in all but the last instruction. 2493 * The first match will set "skip_or", and cause 2494 * the following instructions to be skipped until 2495 * past the one with the F_OR bit clear. 2496 */ 2497 if (skip_or) { /* skip this instruction */ 2498 if ((cmd->len & F_OR) == 0) 2499 skip_or = 0; /* next one is good */ 2500 continue; 2501 } 2502 match = 0; /* set to 1 if we succeed */ 2503 2504 switch (cmd->opcode) { 2505 /* 2506 * The first set of opcodes compares the packet's 2507 * fields with some pattern, setting 'match' if a 2508 * match is found. At the end of the loop there is 2509 * logic to deal with F_NOT and F_OR flags associated 2510 * with the opcode. 2511 */ 2512 case O_NOP: 2513 match = 1; 2514 break; 2515 2516 case O_FORWARD_MAC: 2517 printf("ipfw: opcode %d unimplemented\n", 2518 cmd->opcode); 2519 break; 2520 2521 case O_GID: 2522 case O_UID: 2523 case O_JAIL: 2524 /* 2525 * We only check offset == 0 && proto != 0, 2526 * as this ensures that we have a 2527 * packet with the ports info. 2528 */ 2529 if (offset!=0) 2530 break; 2531 if (is_ipv6) /* XXX to be fixed later */ 2532 break; 2533 if (proto == IPPROTO_TCP || 2534 proto == IPPROTO_UDP) 2535 match = check_uidgid( 2536 (ipfw_insn_u32 *)cmd, 2537 proto, oif, 2538 dst_ip, dst_port, 2539 src_ip, src_port, &fw_ugid_cache, 2540 &ugid_lookup, args->inp); 2541 break; 2542 2543 case O_RECV: 2544 match = iface_match(m->m_pkthdr.rcvif, 2545 (ipfw_insn_if *)cmd); 2546 break; 2547 2548 case O_XMIT: 2549 match = iface_match(oif, (ipfw_insn_if *)cmd); 2550 break; 2551 2552 case O_VIA: 2553 match = iface_match(oif ? oif : 2554 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2555 break; 2556 2557 case O_MACADDR2: 2558 if (args->eh != NULL) { /* have MAC header */ 2559 u_int32_t *want = (u_int32_t *) 2560 ((ipfw_insn_mac *)cmd)->addr; 2561 u_int32_t *mask = (u_int32_t *) 2562 ((ipfw_insn_mac *)cmd)->mask; 2563 u_int32_t *hdr = (u_int32_t *)args->eh; 2564 2565 match = 2566 ( want[0] == (hdr[0] & mask[0]) && 2567 want[1] == (hdr[1] & mask[1]) && 2568 want[2] == (hdr[2] & mask[2]) ); 2569 } 2570 break; 2571 2572 case O_MAC_TYPE: 2573 if (args->eh != NULL) { 2574 u_int16_t t = 2575 ntohs(args->eh->ether_type); 2576 u_int16_t *p = 2577 ((ipfw_insn_u16 *)cmd)->ports; 2578 int i; 2579 2580 for (i = cmdlen - 1; !match && i>0; 2581 i--, p += 2) 2582 match = (t>=p[0] && t<=p[1]); 2583 } 2584 break; 2585 2586 case O_FRAG: 2587 match = (offset != 0); 2588 break; 2589 2590 case O_IN: /* "out" is "not in" */ 2591 match = (oif == NULL); 2592 break; 2593 2594 case O_LAYER2: 2595 match = (args->eh != NULL); 2596 break; 2597 2598 case O_DIVERTED: 2599 match = (cmd->arg1 & 1 && divinput_flags & 2600 IP_FW_DIVERT_LOOPBACK_FLAG) || 2601 (cmd->arg1 & 2 && divinput_flags & 2602 IP_FW_DIVERT_OUTPUT_FLAG); 2603 break; 2604 2605 case O_PROTO: 2606 /* 2607 * We do not allow an arg of 0 so the 2608 * check of "proto" only suffices. 2609 */ 2610 match = (proto == cmd->arg1); 2611 break; 2612 2613 case O_IP_SRC: 2614 match = is_ipv4 && 2615 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2616 src_ip.s_addr); 2617 break; 2618 2619 case O_IP_SRC_LOOKUP: 2620 case O_IP_DST_LOOKUP: 2621 if (is_ipv4) { 2622 uint32_t a = 2623 (cmd->opcode == O_IP_DST_LOOKUP) ? 2624 dst_ip.s_addr : src_ip.s_addr; 2625 uint32_t v; 2626 2627 match = lookup_table(chain, cmd->arg1, a, 2628 &v); 2629 if (!match) 2630 break; 2631 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2632 match = 2633 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2634 else 2635 tablearg = v; 2636 } 2637 break; 2638 2639 case O_IP_SRC_MASK: 2640 case O_IP_DST_MASK: 2641 if (is_ipv4) { 2642 uint32_t a = 2643 (cmd->opcode == O_IP_DST_MASK) ? 2644 dst_ip.s_addr : src_ip.s_addr; 2645 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2646 int i = cmdlen-1; 2647 2648 for (; !match && i>0; i-= 2, p+= 2) 2649 match = (p[0] == (a & p[1])); 2650 } 2651 break; 2652 2653 case O_IP_SRC_ME: 2654 if (is_ipv4) { 2655 struct ifnet *tif; 2656 2657 INADDR_TO_IFP(src_ip, tif); 2658 match = (tif != NULL); 2659 } 2660 break; 2661 2662 case O_IP_DST_SET: 2663 case O_IP_SRC_SET: 2664 if (is_ipv4) { 2665 u_int32_t *d = (u_int32_t *)(cmd+1); 2666 u_int32_t addr = 2667 cmd->opcode == O_IP_DST_SET ? 2668 args->f_id.dst_ip : 2669 args->f_id.src_ip; 2670 2671 if (addr < d[0]) 2672 break; 2673 addr -= d[0]; /* subtract base */ 2674 match = (addr < cmd->arg1) && 2675 ( d[ 1 + (addr>>5)] & 2676 (1<<(addr & 0x1f)) ); 2677 } 2678 break; 2679 2680 case O_IP_DST: 2681 match = is_ipv4 && 2682 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2683 dst_ip.s_addr); 2684 break; 2685 2686 case O_IP_DST_ME: 2687 if (is_ipv4) { 2688 struct ifnet *tif; 2689 2690 INADDR_TO_IFP(dst_ip, tif); 2691 match = (tif != NULL); 2692 } 2693 break; 2694 2695 case O_IP_SRCPORT: 2696 case O_IP_DSTPORT: 2697 /* 2698 * offset == 0 && proto != 0 is enough 2699 * to guarantee that we have a 2700 * packet with port info. 2701 */ 2702 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2703 && offset == 0) { 2704 u_int16_t x = 2705 (cmd->opcode == O_IP_SRCPORT) ? 2706 src_port : dst_port ; 2707 u_int16_t *p = 2708 ((ipfw_insn_u16 *)cmd)->ports; 2709 int i; 2710 2711 for (i = cmdlen - 1; !match && i>0; 2712 i--, p += 2) 2713 match = (x>=p[0] && x<=p[1]); 2714 } 2715 break; 2716 2717 case O_ICMPTYPE: 2718 match = (offset == 0 && proto==IPPROTO_ICMP && 2719 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2720 break; 2721 2722#ifdef INET6 2723 case O_ICMP6TYPE: 2724 match = is_ipv6 && offset == 0 && 2725 proto==IPPROTO_ICMPV6 && 2726 icmp6type_match( 2727 ICMP6(ulp)->icmp6_type, 2728 (ipfw_insn_u32 *)cmd); 2729 break; 2730#endif /* INET6 */ 2731 2732 case O_IPOPT: 2733 match = (is_ipv4 && 2734 ipopts_match(mtod(m, struct ip *), cmd) ); 2735 break; 2736 2737 case O_IPVER: 2738 match = (is_ipv4 && 2739 cmd->arg1 == mtod(m, struct ip *)->ip_v); 2740 break; 2741 2742 case O_IPID: 2743 case O_IPLEN: 2744 case O_IPTTL: 2745 if (is_ipv4) { /* only for IP packets */ 2746 uint16_t x; 2747 uint16_t *p; 2748 int i; 2749 2750 if (cmd->opcode == O_IPLEN) 2751 x = ip_len; 2752 else if (cmd->opcode == O_IPTTL) 2753 x = mtod(m, struct ip *)->ip_ttl; 2754 else /* must be IPID */ 2755 x = ntohs(mtod(m, struct ip *)->ip_id); 2756 if (cmdlen == 1) { 2757 match = (cmd->arg1 == x); 2758 break; 2759 } 2760 /* otherwise we have ranges */ 2761 p = ((ipfw_insn_u16 *)cmd)->ports; 2762 i = cmdlen - 1; 2763 for (; !match && i>0; i--, p += 2) 2764 match = (x >= p[0] && x <= p[1]); 2765 } 2766 break; 2767 2768 case O_IPPRECEDENCE: 2769 match = (is_ipv4 && 2770 (cmd->arg1 == (mtod(m, struct ip *)->ip_tos & 0xe0)) ); 2771 break; 2772 2773 case O_IPTOS: 2774 match = (is_ipv4 && 2775 flags_match(cmd, mtod(m, struct ip *)->ip_tos)); 2776 break; 2777 2778 case O_TCPDATALEN: 2779 if (proto == IPPROTO_TCP && offset == 0) { 2780 struct tcphdr *tcp; 2781 uint16_t x; 2782 uint16_t *p; 2783 int i; 2784 2785 tcp = TCP(ulp); 2786 x = ip_len - 2787 ((ip->ip_hl + tcp->th_off) << 2); 2788 if (cmdlen == 1) { 2789 match = (cmd->arg1 == x); 2790 break; 2791 } 2792 /* otherwise we have ranges */ 2793 p = ((ipfw_insn_u16 *)cmd)->ports; 2794 i = cmdlen - 1; 2795 for (; !match && i>0; i--, p += 2) 2796 match = (x >= p[0] && x <= p[1]); 2797 } 2798 break; 2799 2800 case O_TCPFLAGS: 2801 match = (proto == IPPROTO_TCP && offset == 0 && 2802 flags_match(cmd, TCP(ulp)->th_flags)); 2803 break; 2804 2805 case O_TCPOPTS: 2806 match = (proto == IPPROTO_TCP && offset == 0 && 2807 tcpopts_match(TCP(ulp), cmd)); 2808 break; 2809 2810 case O_TCPSEQ: 2811 match = (proto == IPPROTO_TCP && offset == 0 && 2812 ((ipfw_insn_u32 *)cmd)->d[0] == 2813 TCP(ulp)->th_seq); 2814 break; 2815 2816 case O_TCPACK: 2817 match = (proto == IPPROTO_TCP && offset == 0 && 2818 ((ipfw_insn_u32 *)cmd)->d[0] == 2819 TCP(ulp)->th_ack); 2820 break; 2821 2822 case O_TCPWIN: 2823 match = (proto == IPPROTO_TCP && offset == 0 && 2824 cmd->arg1 == TCP(ulp)->th_win); 2825 break; 2826 2827 case O_ESTAB: 2828 /* reject packets which have SYN only */ 2829 /* XXX should i also check for TH_ACK ? */ 2830 match = (proto == IPPROTO_TCP && offset == 0 && 2831 (TCP(ulp)->th_flags & 2832 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2833 break; 2834 2835 case O_ALTQ: { 2836 struct altq_tag *at; 2837 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2838 2839 match = 1; 2840 mtag = m_tag_find(m, PACKET_TAG_PF_QID, NULL); 2841 if (mtag != NULL) 2842 break; 2843 mtag = m_tag_get(PACKET_TAG_PF_QID, 2844 sizeof(struct altq_tag), 2845 M_NOWAIT); 2846 if (mtag == NULL) { 2847 /* 2848 * Let the packet fall back to the 2849 * default ALTQ. 2850 */ 2851 break; 2852 } 2853 at = (struct altq_tag *)(mtag+1); 2854 at->qid = altq->qid; 2855 if (is_ipv4) 2856 at->af = AF_INET; 2857 else 2858 at->af = AF_LINK; 2859 at->hdr = ip; 2860 m_tag_prepend(m, mtag); 2861 break; 2862 } 2863 2864 case O_LOG: 2865 if (fw_verbose) 2866 ipfw_log(f, hlen, args, m, 2867 oif, offset, tablearg); 2868 match = 1; 2869 break; 2870 2871 case O_PROB: 2872 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2873 break; 2874 2875 case O_VERREVPATH: 2876 /* Outgoing packets automatically pass/match */ 2877 match = ((oif != NULL) || 2878 (m->m_pkthdr.rcvif == NULL) || 2879 ( 2880#ifdef INET6 2881 is_ipv6 ? 2882 verify_path6(&(args->f_id.src_ip6), 2883 m->m_pkthdr.rcvif) : 2884#endif 2885 verify_path(src_ip, m->m_pkthdr.rcvif))); 2886 break; 2887 2888 case O_VERSRCREACH: 2889 /* Outgoing packets automatically pass/match */ 2890 match = (hlen > 0 && ((oif != NULL) || 2891#ifdef INET6 2892 is_ipv6 ? 2893 verify_path6(&(args->f_id.src_ip6), 2894 NULL) : 2895#endif 2896 verify_path(src_ip, NULL))); 2897 break; 2898 2899 case O_ANTISPOOF: 2900 /* Outgoing packets automatically pass/match */ 2901 if (oif == NULL && hlen > 0 && 2902 ( (is_ipv4 && in_localaddr(src_ip)) 2903#ifdef INET6 2904 || (is_ipv6 && 2905 in6_localaddr(&(args->f_id.src_ip6))) 2906#endif 2907 )) 2908 match = 2909#ifdef INET6 2910 is_ipv6 ? verify_path6( 2911 &(args->f_id.src_ip6), 2912 m->m_pkthdr.rcvif) : 2913#endif 2914 verify_path(src_ip, 2915 m->m_pkthdr.rcvif); 2916 else 2917 match = 1; 2918 break; 2919 2920 case O_IPSEC: 2921#ifdef FAST_IPSEC 2922 match = (m_tag_find(m, 2923 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2924#endif 2925#ifdef IPSEC 2926 match = (ipsec_getnhist(m) != 0); 2927#endif 2928 /* otherwise no match */ 2929 break; 2930 2931#ifdef INET6 2932 case O_IP6_SRC: 2933 match = is_ipv6 && 2934 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2935 &((ipfw_insn_ip6 *)cmd)->addr6); 2936 break; 2937 2938 case O_IP6_DST: 2939 match = is_ipv6 && 2940 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2941 &((ipfw_insn_ip6 *)cmd)->addr6); 2942 break; 2943 case O_IP6_SRC_MASK: 2944 case O_IP6_DST_MASK: 2945 if (is_ipv6) { 2946 int i = cmdlen - 1; 2947 struct in6_addr p; 2948 struct in6_addr *d = 2949 &((ipfw_insn_ip6 *)cmd)->addr6; 2950 2951 for (; !match && i > 0; d += 2, 2952 i -= F_INSN_SIZE(struct in6_addr) 2953 * 2) { 2954 p = (cmd->opcode == 2955 O_IP6_SRC_MASK) ? 2956 args->f_id.src_ip6: 2957 args->f_id.dst_ip6; 2958 APPLY_MASK(&p, &d[1]); 2959 match = 2960 IN6_ARE_ADDR_EQUAL(&d[0], 2961 &p); 2962 } 2963 } 2964 break; 2965 2966 case O_IP6_SRC_ME: 2967 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 2968 break; 2969 2970 case O_IP6_DST_ME: 2971 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 2972 break; 2973 2974 case O_FLOW6ID: 2975 match = is_ipv6 && 2976 flow6id_match(args->f_id.flow_id6, 2977 (ipfw_insn_u32 *) cmd); 2978 break; 2979 2980 case O_EXT_HDR: 2981 match = is_ipv6 && 2982 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2983 break; 2984 2985 case O_IP6: 2986 match = is_ipv6; 2987 break; 2988#endif 2989 2990 case O_IP4: 2991 match = is_ipv4; 2992 break; 2993 2994 case O_TAG: { 2995 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 2996 tablearg : cmd->arg1; 2997 2998 /* Packet is already tagged with this tag? */ 2999 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 3000 3001 /* We have `untag' action when F_NOT flag is 3002 * present. And we must remove this mtag from 3003 * mbuf and reset `match' to zero (`match' will 3004 * be inversed later). 3005 * Otherwise we should allocate new mtag and 3006 * push it into mbuf. 3007 */ 3008 if (cmd->len & F_NOT) { /* `untag' action */ 3009 if (mtag != NULL) 3010 m_tag_delete(m, mtag); 3011 } else if (mtag == NULL) { 3012 if ((mtag = m_tag_alloc(MTAG_IPFW, 3013 tag, 0, M_NOWAIT)) != NULL) 3014 m_tag_prepend(m, mtag); 3015 } 3016 match = (cmd->len & F_NOT) ? 0: 1; 3017 break; 3018 } 3019 3020 case O_TAGGED: { 3021 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3022 tablearg : cmd->arg1; 3023 3024 if (cmdlen == 1) { 3025 match = m_tag_locate(m, MTAG_IPFW, 3026 tag, NULL) != NULL; 3027 break; 3028 } 3029 3030 /* we have ranges */ 3031 for (mtag = m_tag_first(m); 3032 mtag != NULL && !match; 3033 mtag = m_tag_next(m, mtag)) { 3034 uint16_t *p; 3035 int i; 3036 3037 if (mtag->m_tag_cookie != MTAG_IPFW) 3038 continue; 3039 3040 p = ((ipfw_insn_u16 *)cmd)->ports; 3041 i = cmdlen - 1; 3042 for(; !match && i > 0; i--, p += 2) 3043 match = 3044 mtag->m_tag_id >= p[0] && 3045 mtag->m_tag_id <= p[1]; 3046 } 3047 break; 3048 } 3049 3050 /* 3051 * The second set of opcodes represents 'actions', 3052 * i.e. the terminal part of a rule once the packet 3053 * matches all previous patterns. 3054 * Typically there is only one action for each rule, 3055 * and the opcode is stored at the end of the rule 3056 * (but there are exceptions -- see below). 3057 * 3058 * In general, here we set retval and terminate the 3059 * outer loop (would be a 'break 3' in some language, 3060 * but we need to do a 'goto done'). 3061 * 3062 * Exceptions: 3063 * O_COUNT and O_SKIPTO actions: 3064 * instead of terminating, we jump to the next rule 3065 * ('goto next_rule', equivalent to a 'break 2'), 3066 * or to the SKIPTO target ('goto again' after 3067 * having set f, cmd and l), respectively. 3068 * 3069 * O_TAG, O_LOG and O_ALTQ action parameters: 3070 * perform some action and set match = 1; 3071 * 3072 * O_LIMIT and O_KEEP_STATE: these opcodes are 3073 * not real 'actions', and are stored right 3074 * before the 'action' part of the rule. 3075 * These opcodes try to install an entry in the 3076 * state tables; if successful, we continue with 3077 * the next opcode (match=1; break;), otherwise 3078 * the packet * must be dropped 3079 * ('goto done' after setting retval); 3080 * 3081 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3082 * cause a lookup of the state table, and a jump 3083 * to the 'action' part of the parent rule 3084 * ('goto check_body') if an entry is found, or 3085 * (CHECK_STATE only) a jump to the next rule if 3086 * the entry is not found ('goto next_rule'). 3087 * The result of the lookup is cached to make 3088 * further instances of these opcodes are 3089 * effectively NOPs. 3090 */ 3091 case O_LIMIT: 3092 case O_KEEP_STATE: 3093 if (install_state(f, 3094 (ipfw_insn_limit *)cmd, args, tablearg)) { 3095 retval = IP_FW_DENY; 3096 goto done; /* error/limit violation */ 3097 } 3098 match = 1; 3099 break; 3100 3101 case O_PROBE_STATE: 3102 case O_CHECK_STATE: 3103 /* 3104 * dynamic rules are checked at the first 3105 * keep-state or check-state occurrence, 3106 * with the result being stored in dyn_dir. 3107 * The compiler introduces a PROBE_STATE 3108 * instruction for us when we have a 3109 * KEEP_STATE (because PROBE_STATE needs 3110 * to be run first). 3111 */ 3112 if (dyn_dir == MATCH_UNKNOWN && 3113 (q = lookup_dyn_rule(&args->f_id, 3114 &dyn_dir, proto == IPPROTO_TCP ? 3115 TCP(ulp) : NULL)) 3116 != NULL) { 3117 /* 3118 * Found dynamic entry, update stats 3119 * and jump to the 'action' part of 3120 * the parent rule. 3121 */ 3122 q->pcnt++; 3123 q->bcnt += pktlen; 3124 f = q->rule; 3125 cmd = ACTION_PTR(f); 3126 l = f->cmd_len - f->act_ofs; 3127 IPFW_DYN_UNLOCK(); 3128 goto check_body; 3129 } 3130 /* 3131 * Dynamic entry not found. If CHECK_STATE, 3132 * skip to next rule, if PROBE_STATE just 3133 * ignore and continue with next opcode. 3134 */ 3135 if (cmd->opcode == O_CHECK_STATE) 3136 goto next_rule; 3137 match = 1; 3138 break; 3139 3140 case O_ACCEPT: 3141 retval = 0; /* accept */ 3142 goto done; 3143 3144 case O_PIPE: 3145 case O_QUEUE: 3146 args->rule = f; /* report matching rule */ 3147 if (cmd->arg1 == IP_FW_TABLEARG) 3148 args->cookie = tablearg; 3149 else 3150 args->cookie = cmd->arg1; 3151 retval = IP_FW_DUMMYNET; 3152 goto done; 3153 3154 case O_DIVERT: 3155 case O_TEE: { 3156 struct divert_tag *dt; 3157 3158 if (args->eh) /* not on layer 2 */ 3159 break; 3160 mtag = m_tag_get(PACKET_TAG_DIVERT, 3161 sizeof(struct divert_tag), 3162 M_NOWAIT); 3163 if (mtag == NULL) { 3164 /* XXX statistic */ 3165 /* drop packet */ 3166 IPFW_RUNLOCK(chain); 3167 return (IP_FW_DENY); 3168 } 3169 dt = (struct divert_tag *)(mtag+1); 3170 dt->cookie = f->rulenum; 3171 if (cmd->arg1 == IP_FW_TABLEARG) 3172 dt->info = tablearg; 3173 else 3174 dt->info = cmd->arg1; 3175 m_tag_prepend(m, mtag); 3176 retval = (cmd->opcode == O_DIVERT) ? 3177 IP_FW_DIVERT : IP_FW_TEE; 3178 goto done; 3179 } 3180 3181 case O_COUNT: 3182 case O_SKIPTO: 3183 f->pcnt++; /* update stats */ 3184 f->bcnt += pktlen; 3185 f->timestamp = time_uptime; 3186 if (cmd->opcode == O_COUNT) 3187 goto next_rule; 3188 /* handle skipto */ 3189 if (f->next_rule == NULL) 3190 lookup_next_rule(f); 3191 f = f->next_rule; 3192 goto again; 3193 3194 case O_REJECT: 3195 /* 3196 * Drop the packet and send a reject notice 3197 * if the packet is not ICMP (or is an ICMP 3198 * query), and it is not multicast/broadcast. 3199 */ 3200 if (hlen > 0 && is_ipv4 && offset == 0 && 3201 (proto != IPPROTO_ICMP || 3202 is_icmp_query(ICMP(ulp))) && 3203 !(m->m_flags & (M_BCAST|M_MCAST)) && 3204 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3205 send_reject(args, cmd->arg1, ip_len); 3206 m = args->m; 3207 } 3208 /* FALLTHROUGH */ 3209#ifdef INET6 3210 case O_UNREACH6: 3211 if (hlen > 0 && is_ipv6 && 3212 ((offset & IP6F_OFF_MASK) == 0) && 3213 (proto != IPPROTO_ICMPV6 || 3214 (is_icmp6_query(args->f_id.flags) == 1)) && 3215 !(m->m_flags & (M_BCAST|M_MCAST)) && 3216 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3217 send_reject6(args, cmd->arg1, hlen); 3218 m = args->m; 3219 } 3220 /* FALLTHROUGH */ 3221#endif 3222 case O_DENY: 3223 retval = IP_FW_DENY; 3224 goto done; 3225 3226 case O_FORWARD_IP: { 3227 struct sockaddr_in *sa; 3228 sa = &(((ipfw_insn_sa *)cmd)->sa); 3229 if (args->eh) /* not valid on layer2 pkts */ 3230 break; 3231 if (!q || dyn_dir == MATCH_FORWARD) { 3232 if (sa->sin_addr.s_addr == INADDR_ANY) { 3233 bcopy(sa, &args->hopstore, 3234 sizeof(*sa)); 3235 args->hopstore.sin_addr.s_addr = 3236 htonl(tablearg); 3237 args->next_hop = 3238 &args->hopstore; 3239 } else { 3240 args->next_hop = sa; 3241 } 3242 } 3243 retval = IP_FW_PASS; 3244 } 3245 goto done; 3246 3247 case O_NETGRAPH: 3248 case O_NGTEE: 3249 args->rule = f; /* report matching rule */ 3250 if (cmd->arg1 == IP_FW_TABLEARG) 3251 args->cookie = tablearg; 3252 else 3253 args->cookie = cmd->arg1; 3254 retval = (cmd->opcode == O_NETGRAPH) ? 3255 IP_FW_NETGRAPH : IP_FW_NGTEE; 3256 goto done; 3257 3258 default: 3259 panic("-- unknown opcode %d\n", cmd->opcode); 3260 } /* end of switch() on opcodes */ 3261 3262 if (cmd->len & F_NOT) 3263 match = !match; 3264 3265 if (match) { 3266 if (cmd->len & F_OR) 3267 skip_or = 1; 3268 } else { 3269 if (!(cmd->len & F_OR)) /* not an OR block, */ 3270 break; /* try next rule */ 3271 } 3272 3273 } /* end of inner for, scan opcodes */ 3274 3275next_rule:; /* try next rule */ 3276 3277 } /* end of outer for, scan rules */ 3278 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3279 IPFW_RUNLOCK(chain); 3280 return (IP_FW_DENY); 3281 3282done: 3283 /* Update statistics */ 3284 f->pcnt++; 3285 f->bcnt += pktlen; 3286 f->timestamp = time_uptime; 3287 IPFW_RUNLOCK(chain); 3288 return (retval); 3289 3290pullup_failed: 3291 if (fw_verbose) 3292 printf("ipfw: pullup failed\n"); 3293 return (IP_FW_DENY); 3294} 3295 3296/* 3297 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3298 * These will be reconstructed on the fly as packets are matched. 3299 */ 3300static void 3301flush_rule_ptrs(struct ip_fw_chain *chain) 3302{ 3303 struct ip_fw *rule; 3304 3305 IPFW_WLOCK_ASSERT(chain); 3306 3307 for (rule = chain->rules; rule; rule = rule->next) 3308 rule->next_rule = NULL; 3309} 3310 3311/* 3312 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3313 * possibly create a rule number and add the rule to the list. 3314 * Update the rule_number in the input struct so the caller knows it as well. 3315 */ 3316static int 3317add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3318{ 3319 struct ip_fw *rule, *f, *prev; 3320 int l = RULESIZE(input_rule); 3321 3322 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3323 return (EINVAL); 3324 3325 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3326 if (rule == NULL) 3327 return (ENOSPC); 3328 3329 bcopy(input_rule, rule, l); 3330 3331 rule->next = NULL; 3332 rule->next_rule = NULL; 3333 3334 rule->pcnt = 0; 3335 rule->bcnt = 0; 3336 rule->timestamp = 0; 3337 3338 IPFW_WLOCK(chain); 3339 3340 if (chain->rules == NULL) { /* default rule */ 3341 chain->rules = rule; 3342 goto done; 3343 } 3344 3345 /* 3346 * If rulenum is 0, find highest numbered rule before the 3347 * default rule, and add autoinc_step 3348 */ 3349 if (autoinc_step < 1) 3350 autoinc_step = 1; 3351 else if (autoinc_step > 1000) 3352 autoinc_step = 1000; 3353 if (rule->rulenum == 0) { 3354 /* 3355 * locate the highest numbered rule before default 3356 */ 3357 for (f = chain->rules; f; f = f->next) { 3358 if (f->rulenum == IPFW_DEFAULT_RULE) 3359 break; 3360 rule->rulenum = f->rulenum; 3361 } 3362 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 3363 rule->rulenum += autoinc_step; 3364 input_rule->rulenum = rule->rulenum; 3365 } 3366 3367 /* 3368 * Now insert the new rule in the right place in the sorted list. 3369 */ 3370 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3371 if (f->rulenum > rule->rulenum) { /* found the location */ 3372 if (prev) { 3373 rule->next = f; 3374 prev->next = rule; 3375 } else { /* head insert */ 3376 rule->next = chain->rules; 3377 chain->rules = rule; 3378 } 3379 break; 3380 } 3381 } 3382 flush_rule_ptrs(chain); 3383done: 3384 static_count++; 3385 static_len += l; 3386 IPFW_WUNLOCK(chain); 3387 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3388 rule->rulenum, static_count);) 3389 return (0); 3390} 3391 3392/** 3393 * Remove a static rule (including derived * dynamic rules) 3394 * and place it on the ``reap list'' for later reclamation. 3395 * The caller is in charge of clearing rule pointers to avoid 3396 * dangling pointers. 3397 * @return a pointer to the next entry. 3398 * Arguments are not checked, so they better be correct. 3399 */ 3400static struct ip_fw * 3401remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 3402{ 3403 struct ip_fw *n; 3404 int l = RULESIZE(rule); 3405 3406 IPFW_WLOCK_ASSERT(chain); 3407 3408 n = rule->next; 3409 IPFW_DYN_LOCK(); 3410 remove_dyn_rule(rule, NULL /* force removal */); 3411 IPFW_DYN_UNLOCK(); 3412 if (prev == NULL) 3413 chain->rules = n; 3414 else 3415 prev->next = n; 3416 static_count--; 3417 static_len -= l; 3418 3419 rule->next = chain->reap; 3420 chain->reap = rule; 3421 3422 return n; 3423} 3424 3425/** 3426 * Reclaim storage associated with a list of rules. This is 3427 * typically the list created using remove_rule. 3428 */ 3429static void 3430reap_rules(struct ip_fw *head) 3431{ 3432 struct ip_fw *rule; 3433 3434 while ((rule = head) != NULL) { 3435 head = head->next; 3436 if (DUMMYNET_LOADED) 3437 ip_dn_ruledel_ptr(rule); 3438 free(rule, M_IPFW); 3439 } 3440} 3441 3442/* 3443 * Remove all rules from a chain (except rules in set RESVD_SET 3444 * unless kill_default = 1). The caller is responsible for 3445 * reclaiming storage for the rules left in chain->reap. 3446 */ 3447static void 3448free_chain(struct ip_fw_chain *chain, int kill_default) 3449{ 3450 struct ip_fw *prev, *rule; 3451 3452 IPFW_WLOCK_ASSERT(chain); 3453 3454 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3455 for (prev = NULL, rule = chain->rules; rule ; ) 3456 if (kill_default || rule->set != RESVD_SET) 3457 rule = remove_rule(chain, rule, prev); 3458 else { 3459 prev = rule; 3460 rule = rule->next; 3461 } 3462} 3463 3464/** 3465 * Remove all rules with given number, and also do set manipulation. 3466 * Assumes chain != NULL && *chain != NULL. 3467 * 3468 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3469 * the next 8 bits are the new set, the top 8 bits are the command: 3470 * 3471 * 0 delete rules with given number 3472 * 1 delete rules with given set number 3473 * 2 move rules with given number to new set 3474 * 3 move rules with given set number to new set 3475 * 4 swap sets with given numbers 3476 */ 3477static int 3478del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3479{ 3480 struct ip_fw *prev = NULL, *rule; 3481 u_int16_t rulenum; /* rule or old_set */ 3482 u_int8_t cmd, new_set; 3483 3484 rulenum = arg & 0xffff; 3485 cmd = (arg >> 24) & 0xff; 3486 new_set = (arg >> 16) & 0xff; 3487 3488 if (cmd > 4) 3489 return EINVAL; 3490 if (new_set > RESVD_SET) 3491 return EINVAL; 3492 if (cmd == 0 || cmd == 2) { 3493 if (rulenum >= IPFW_DEFAULT_RULE) 3494 return EINVAL; 3495 } else { 3496 if (rulenum > RESVD_SET) /* old_set */ 3497 return EINVAL; 3498 } 3499 3500 IPFW_WLOCK(chain); 3501 rule = chain->rules; 3502 chain->reap = NULL; 3503 switch (cmd) { 3504 case 0: /* delete rules with given number */ 3505 /* 3506 * locate first rule to delete 3507 */ 3508 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3509 ; 3510 if (rule->rulenum != rulenum) { 3511 IPFW_WUNLOCK(chain); 3512 return EINVAL; 3513 } 3514 3515 /* 3516 * flush pointers outside the loop, then delete all matching 3517 * rules. prev remains the same throughout the cycle. 3518 */ 3519 flush_rule_ptrs(chain); 3520 while (rule->rulenum == rulenum) 3521 rule = remove_rule(chain, rule, prev); 3522 break; 3523 3524 case 1: /* delete all rules with given set number */ 3525 flush_rule_ptrs(chain); 3526 rule = chain->rules; 3527 while (rule->rulenum < IPFW_DEFAULT_RULE) 3528 if (rule->set == rulenum) 3529 rule = remove_rule(chain, rule, prev); 3530 else { 3531 prev = rule; 3532 rule = rule->next; 3533 } 3534 break; 3535 3536 case 2: /* move rules with given number to new set */ 3537 rule = chain->rules; 3538 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3539 if (rule->rulenum == rulenum) 3540 rule->set = new_set; 3541 break; 3542 3543 case 3: /* move rules with given set number to new set */ 3544 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3545 if (rule->set == rulenum) 3546 rule->set = new_set; 3547 break; 3548 3549 case 4: /* swap two sets */ 3550 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3551 if (rule->set == rulenum) 3552 rule->set = new_set; 3553 else if (rule->set == new_set) 3554 rule->set = rulenum; 3555 break; 3556 } 3557 /* 3558 * Look for rules to reclaim. We grab the list before 3559 * releasing the lock then reclaim them w/o the lock to 3560 * avoid a LOR with dummynet. 3561 */ 3562 rule = chain->reap; 3563 chain->reap = NULL; 3564 IPFW_WUNLOCK(chain); 3565 if (rule) 3566 reap_rules(rule); 3567 return 0; 3568} 3569 3570/* 3571 * Clear counters for a specific rule. 3572 * The enclosing "table" is assumed locked. 3573 */ 3574static void 3575clear_counters(struct ip_fw *rule, int log_only) 3576{ 3577 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3578 3579 if (log_only == 0) { 3580 rule->bcnt = rule->pcnt = 0; 3581 rule->timestamp = 0; 3582 } 3583 if (l->o.opcode == O_LOG) 3584 l->log_left = l->max_log; 3585} 3586 3587/** 3588 * Reset some or all counters on firewall rules. 3589 * @arg frwl is null to clear all entries, or contains a specific 3590 * rule number. 3591 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 3592 */ 3593static int 3594zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 3595{ 3596 struct ip_fw *rule; 3597 char *msg; 3598 3599 IPFW_WLOCK(chain); 3600 if (rulenum == 0) { 3601 norule_counter = 0; 3602 for (rule = chain->rules; rule; rule = rule->next) 3603 clear_counters(rule, log_only); 3604 msg = log_only ? "ipfw: All logging counts reset.\n" : 3605 "ipfw: Accounting cleared.\n"; 3606 } else { 3607 int cleared = 0; 3608 /* 3609 * We can have multiple rules with the same number, so we 3610 * need to clear them all. 3611 */ 3612 for (rule = chain->rules; rule; rule = rule->next) 3613 if (rule->rulenum == rulenum) { 3614 while (rule && rule->rulenum == rulenum) { 3615 clear_counters(rule, log_only); 3616 rule = rule->next; 3617 } 3618 cleared = 1; 3619 break; 3620 } 3621 if (!cleared) { /* we did not find any matching rules */ 3622 IPFW_WUNLOCK(chain); 3623 return (EINVAL); 3624 } 3625 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 3626 "ipfw: Entry %d cleared.\n"; 3627 } 3628 IPFW_WUNLOCK(chain); 3629 3630 if (fw_verbose) 3631 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 3632 return (0); 3633} 3634 3635/* 3636 * Check validity of the structure before insert. 3637 * Fortunately rules are simple, so this mostly need to check rule sizes. 3638 */ 3639static int 3640check_ipfw_struct(struct ip_fw *rule, int size) 3641{ 3642 int l, cmdlen = 0; 3643 int have_action=0; 3644 ipfw_insn *cmd; 3645 3646 if (size < sizeof(*rule)) { 3647 printf("ipfw: rule too short\n"); 3648 return (EINVAL); 3649 } 3650 /* first, check for valid size */ 3651 l = RULESIZE(rule); 3652 if (l != size) { 3653 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3654 return (EINVAL); 3655 } 3656 if (rule->act_ofs >= rule->cmd_len) { 3657 printf("ipfw: bogus action offset (%u > %u)\n", 3658 rule->act_ofs, rule->cmd_len - 1); 3659 return (EINVAL); 3660 } 3661 /* 3662 * Now go for the individual checks. Very simple ones, basically only 3663 * instruction sizes. 3664 */ 3665 for (l = rule->cmd_len, cmd = rule->cmd ; 3666 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3667 cmdlen = F_LEN(cmd); 3668 if (cmdlen > l) { 3669 printf("ipfw: opcode %d size truncated\n", 3670 cmd->opcode); 3671 return EINVAL; 3672 } 3673 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3674 switch (cmd->opcode) { 3675 case O_PROBE_STATE: 3676 case O_KEEP_STATE: 3677 case O_PROTO: 3678 case O_IP_SRC_ME: 3679 case O_IP_DST_ME: 3680 case O_LAYER2: 3681 case O_IN: 3682 case O_FRAG: 3683 case O_DIVERTED: 3684 case O_IPOPT: 3685 case O_IPTOS: 3686 case O_IPPRECEDENCE: 3687 case O_IPVER: 3688 case O_TCPWIN: 3689 case O_TCPFLAGS: 3690 case O_TCPOPTS: 3691 case O_ESTAB: 3692 case O_VERREVPATH: 3693 case O_VERSRCREACH: 3694 case O_ANTISPOOF: 3695 case O_IPSEC: 3696#ifdef INET6 3697 case O_IP6_SRC_ME: 3698 case O_IP6_DST_ME: 3699 case O_EXT_HDR: 3700 case O_IP6: 3701#endif 3702 case O_IP4: 3703 case O_TAG: 3704 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3705 goto bad_size; 3706 break; 3707 3708 case O_UID: 3709 case O_GID: 3710 case O_JAIL: 3711 case O_IP_SRC: 3712 case O_IP_DST: 3713 case O_TCPSEQ: 3714 case O_TCPACK: 3715 case O_PROB: 3716 case O_ICMPTYPE: 3717 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3718 goto bad_size; 3719 break; 3720 3721 case O_LIMIT: 3722 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3723 goto bad_size; 3724 break; 3725 3726 case O_LOG: 3727 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3728 goto bad_size; 3729 3730 ((ipfw_insn_log *)cmd)->log_left = 3731 ((ipfw_insn_log *)cmd)->max_log; 3732 3733 break; 3734 3735 case O_IP_SRC_MASK: 3736 case O_IP_DST_MASK: 3737 /* only odd command lengths */ 3738 if ( !(cmdlen & 1) || cmdlen > 31) 3739 goto bad_size; 3740 break; 3741 3742 case O_IP_SRC_SET: 3743 case O_IP_DST_SET: 3744 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3745 printf("ipfw: invalid set size %d\n", 3746 cmd->arg1); 3747 return EINVAL; 3748 } 3749 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3750 (cmd->arg1+31)/32 ) 3751 goto bad_size; 3752 break; 3753 3754 case O_IP_SRC_LOOKUP: 3755 case O_IP_DST_LOOKUP: 3756 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3757 printf("ipfw: invalid table number %d\n", 3758 cmd->arg1); 3759 return (EINVAL); 3760 } 3761 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3762 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3763 goto bad_size; 3764 break; 3765 3766 case O_MACADDR2: 3767 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3768 goto bad_size; 3769 break; 3770 3771 case O_NOP: 3772 case O_IPID: 3773 case O_IPTTL: 3774 case O_IPLEN: 3775 case O_TCPDATALEN: 3776 case O_TAGGED: 3777 if (cmdlen < 1 || cmdlen > 31) 3778 goto bad_size; 3779 break; 3780 3781 case O_MAC_TYPE: 3782 case O_IP_SRCPORT: 3783 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3784 if (cmdlen < 2 || cmdlen > 31) 3785 goto bad_size; 3786 break; 3787 3788 case O_RECV: 3789 case O_XMIT: 3790 case O_VIA: 3791 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3792 goto bad_size; 3793 break; 3794 3795 case O_ALTQ: 3796 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3797 goto bad_size; 3798 break; 3799 3800 case O_PIPE: 3801 case O_QUEUE: 3802 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3803 goto bad_size; 3804 goto check_action; 3805 3806 case O_FORWARD_IP: 3807#ifdef IPFIREWALL_FORWARD 3808 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3809 goto bad_size; 3810 goto check_action; 3811#else 3812 return EINVAL; 3813#endif 3814 3815 case O_DIVERT: 3816 case O_TEE: 3817 if (ip_divert_ptr == NULL) 3818 return EINVAL; 3819 else 3820 goto check_size; 3821 case O_NETGRAPH: 3822 case O_NGTEE: 3823 if (!NG_IPFW_LOADED) 3824 return EINVAL; 3825 else 3826 goto check_size; 3827 case O_FORWARD_MAC: /* XXX not implemented yet */ 3828 case O_CHECK_STATE: 3829 case O_COUNT: 3830 case O_ACCEPT: 3831 case O_DENY: 3832 case O_REJECT: 3833#ifdef INET6 3834 case O_UNREACH6: 3835#endif 3836 case O_SKIPTO: 3837check_size: 3838 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3839 goto bad_size; 3840check_action: 3841 if (have_action) { 3842 printf("ipfw: opcode %d, multiple actions" 3843 " not allowed\n", 3844 cmd->opcode); 3845 return EINVAL; 3846 } 3847 have_action = 1; 3848 if (l != cmdlen) { 3849 printf("ipfw: opcode %d, action must be" 3850 " last opcode\n", 3851 cmd->opcode); 3852 return EINVAL; 3853 } 3854 break; 3855#ifdef INET6 3856 case O_IP6_SRC: 3857 case O_IP6_DST: 3858 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 3859 F_INSN_SIZE(ipfw_insn)) 3860 goto bad_size; 3861 break; 3862 3863 case O_FLOW6ID: 3864 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3865 ((ipfw_insn_u32 *)cmd)->o.arg1) 3866 goto bad_size; 3867 break; 3868 3869 case O_IP6_SRC_MASK: 3870 case O_IP6_DST_MASK: 3871 if ( !(cmdlen & 1) || cmdlen > 127) 3872 goto bad_size; 3873 break; 3874 case O_ICMP6TYPE: 3875 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 3876 goto bad_size; 3877 break; 3878#endif 3879 3880 default: 3881 switch (cmd->opcode) { 3882#ifndef INET6 3883 case O_IP6_SRC_ME: 3884 case O_IP6_DST_ME: 3885 case O_EXT_HDR: 3886 case O_IP6: 3887 case O_UNREACH6: 3888 case O_IP6_SRC: 3889 case O_IP6_DST: 3890 case O_FLOW6ID: 3891 case O_IP6_SRC_MASK: 3892 case O_IP6_DST_MASK: 3893 case O_ICMP6TYPE: 3894 printf("ipfw: no IPv6 support in kernel\n"); 3895 return EPROTONOSUPPORT; 3896#endif 3897 default: 3898 printf("ipfw: opcode %d, unknown opcode\n", 3899 cmd->opcode); 3900 return EINVAL; 3901 } 3902 } 3903 } 3904 if (have_action == 0) { 3905 printf("ipfw: missing action\n"); 3906 return EINVAL; 3907 } 3908 return 0; 3909 3910bad_size: 3911 printf("ipfw: opcode %d size %d wrong\n", 3912 cmd->opcode, cmdlen); 3913 return EINVAL; 3914} 3915 3916/* 3917 * Copy the static and dynamic rules to the supplied buffer 3918 * and return the amount of space actually used. 3919 */ 3920static size_t 3921ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 3922{ 3923 char *bp = buf; 3924 char *ep = bp + space; 3925 struct ip_fw *rule; 3926 int i; 3927 3928 /* XXX this can take a long time and locking will block packet flow */ 3929 IPFW_RLOCK(chain); 3930 for (rule = chain->rules; rule ; rule = rule->next) { 3931 /* 3932 * Verify the entry fits in the buffer in case the 3933 * rules changed between calculating buffer space and 3934 * now. This would be better done using a generation 3935 * number but should suffice for now. 3936 */ 3937 i = RULESIZE(rule); 3938 if (bp + i <= ep) { 3939 bcopy(rule, bp, i); 3940 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 3941 sizeof(set_disable)); 3942 bp += i; 3943 } 3944 } 3945 IPFW_RUNLOCK(chain); 3946 if (ipfw_dyn_v) { 3947 ipfw_dyn_rule *p, *last = NULL; 3948 3949 IPFW_DYN_LOCK(); 3950 for (i = 0 ; i < curr_dyn_buckets; i++) 3951 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 3952 if (bp + sizeof *p <= ep) { 3953 ipfw_dyn_rule *dst = 3954 (ipfw_dyn_rule *)bp; 3955 bcopy(p, dst, sizeof *p); 3956 bcopy(&(p->rule->rulenum), &(dst->rule), 3957 sizeof(p->rule->rulenum)); 3958 /* 3959 * store a non-null value in "next". 3960 * The userland code will interpret a 3961 * NULL here as a marker 3962 * for the last dynamic rule. 3963 */ 3964 bcopy(&dst, &dst->next, sizeof(dst)); 3965 last = dst; 3966 dst->expire = 3967 TIME_LEQ(dst->expire, time_uptime) ? 3968 0 : dst->expire - time_uptime ; 3969 bp += sizeof(ipfw_dyn_rule); 3970 } 3971 } 3972 IPFW_DYN_UNLOCK(); 3973 if (last != NULL) /* mark last dynamic rule */ 3974 bzero(&last->next, sizeof(last)); 3975 } 3976 return (bp - (char *)buf); 3977} 3978 3979 3980/** 3981 * {set|get}sockopt parser. 3982 */ 3983static int 3984ipfw_ctl(struct sockopt *sopt) 3985{ 3986#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3987 int error, rule_num; 3988 size_t size; 3989 struct ip_fw *buf, *rule; 3990 u_int32_t rulenum[2]; 3991 3992 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW); 3993 if (error) 3994 return (error); 3995 3996 /* 3997 * Disallow modifications in really-really secure mode, but still allow 3998 * the logging counters to be reset. 3999 */ 4000 if (sopt->sopt_name == IP_FW_ADD || 4001 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 4002 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 4003 if (error) 4004 return (error); 4005 } 4006 4007 error = 0; 4008 4009 switch (sopt->sopt_name) { 4010 case IP_FW_GET: 4011 /* 4012 * pass up a copy of the current rules. Static rules 4013 * come first (the last of which has number IPFW_DEFAULT_RULE), 4014 * followed by a possibly empty list of dynamic rule. 4015 * The last dynamic rule has NULL in the "next" field. 4016 * 4017 * Note that the calculated size is used to bound the 4018 * amount of data returned to the user. The rule set may 4019 * change between calculating the size and returning the 4020 * data in which case we'll just return what fits. 4021 */ 4022 size = static_len; /* size of static rules */ 4023 if (ipfw_dyn_v) /* add size of dyn.rules */ 4024 size += (dyn_count * sizeof(ipfw_dyn_rule)); 4025 4026 /* 4027 * XXX todo: if the user passes a short length just to know 4028 * how much room is needed, do not bother filling up the 4029 * buffer, just jump to the sooptcopyout. 4030 */ 4031 buf = malloc(size, M_TEMP, M_WAITOK); 4032 error = sooptcopyout(sopt, buf, 4033 ipfw_getrules(&layer3_chain, buf, size)); 4034 free(buf, M_TEMP); 4035 break; 4036 4037 case IP_FW_FLUSH: 4038 /* 4039 * Normally we cannot release the lock on each iteration. 4040 * We could do it here only because we start from the head all 4041 * the times so there is no risk of missing some entries. 4042 * On the other hand, the risk is that we end up with 4043 * a very inconsistent ruleset, so better keep the lock 4044 * around the whole cycle. 4045 * 4046 * XXX this code can be improved by resetting the head of 4047 * the list to point to the default rule, and then freeing 4048 * the old list without the need for a lock. 4049 */ 4050 4051 IPFW_WLOCK(&layer3_chain); 4052 layer3_chain.reap = NULL; 4053 free_chain(&layer3_chain, 0 /* keep default rule */); 4054 rule = layer3_chain.reap; 4055 layer3_chain.reap = NULL; 4056 IPFW_WUNLOCK(&layer3_chain); 4057 if (rule != NULL) 4058 reap_rules(rule); 4059 break; 4060 4061 case IP_FW_ADD: 4062 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 4063 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 4064 sizeof(struct ip_fw) ); 4065 if (error == 0) 4066 error = check_ipfw_struct(rule, sopt->sopt_valsize); 4067 if (error == 0) { 4068 error = add_rule(&layer3_chain, rule); 4069 size = RULESIZE(rule); 4070 if (!error && sopt->sopt_dir == SOPT_GET) 4071 error = sooptcopyout(sopt, rule, size); 4072 } 4073 free(rule, M_TEMP); 4074 break; 4075 4076 case IP_FW_DEL: 4077 /* 4078 * IP_FW_DEL is used for deleting single rules or sets, 4079 * and (ab)used to atomically manipulate sets. Argument size 4080 * is used to distinguish between the two: 4081 * sizeof(u_int32_t) 4082 * delete single rule or set of rules, 4083 * or reassign rules (or sets) to a different set. 4084 * 2*sizeof(u_int32_t) 4085 * atomic disable/enable sets. 4086 * first u_int32_t contains sets to be disabled, 4087 * second u_int32_t contains sets to be enabled. 4088 */ 4089 error = sooptcopyin(sopt, rulenum, 4090 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4091 if (error) 4092 break; 4093 size = sopt->sopt_valsize; 4094 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4095 error = del_entry(&layer3_chain, rulenum[0]); 4096 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4097 set_disable = 4098 (set_disable | rulenum[0]) & ~rulenum[1] & 4099 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4100 else 4101 error = EINVAL; 4102 break; 4103 4104 case IP_FW_ZERO: 4105 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 4106 rule_num = 0; 4107 if (sopt->sopt_val != 0) { 4108 error = sooptcopyin(sopt, &rule_num, 4109 sizeof(int), sizeof(int)); 4110 if (error) 4111 break; 4112 } 4113 error = zero_entry(&layer3_chain, rule_num, 4114 sopt->sopt_name == IP_FW_RESETLOG); 4115 break; 4116 4117 case IP_FW_TABLE_ADD: 4118 { 4119 ipfw_table_entry ent; 4120 4121 error = sooptcopyin(sopt, &ent, 4122 sizeof(ent), sizeof(ent)); 4123 if (error) 4124 break; 4125 error = add_table_entry(&layer3_chain, ent.tbl, 4126 ent.addr, ent.masklen, ent.value); 4127 } 4128 break; 4129 4130 case IP_FW_TABLE_DEL: 4131 { 4132 ipfw_table_entry ent; 4133 4134 error = sooptcopyin(sopt, &ent, 4135 sizeof(ent), sizeof(ent)); 4136 if (error) 4137 break; 4138 error = del_table_entry(&layer3_chain, ent.tbl, 4139 ent.addr, ent.masklen); 4140 } 4141 break; 4142 4143 case IP_FW_TABLE_FLUSH: 4144 { 4145 u_int16_t tbl; 4146 4147 error = sooptcopyin(sopt, &tbl, 4148 sizeof(tbl), sizeof(tbl)); 4149 if (error) 4150 break; 4151 IPFW_WLOCK(&layer3_chain); 4152 error = flush_table(&layer3_chain, tbl); 4153 IPFW_WUNLOCK(&layer3_chain); 4154 } 4155 break; 4156 4157 case IP_FW_TABLE_GETSIZE: 4158 { 4159 u_int32_t tbl, cnt; 4160 4161 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4162 sizeof(tbl)))) 4163 break; 4164 IPFW_RLOCK(&layer3_chain); 4165 error = count_table(&layer3_chain, tbl, &cnt); 4166 IPFW_RUNLOCK(&layer3_chain); 4167 if (error) 4168 break; 4169 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4170 } 4171 break; 4172 4173 case IP_FW_TABLE_LIST: 4174 { 4175 ipfw_table *tbl; 4176 4177 if (sopt->sopt_valsize < sizeof(*tbl)) { 4178 error = EINVAL; 4179 break; 4180 } 4181 size = sopt->sopt_valsize; 4182 tbl = malloc(size, M_TEMP, M_WAITOK); 4183 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4184 if (error) { 4185 free(tbl, M_TEMP); 4186 break; 4187 } 4188 tbl->size = (size - sizeof(*tbl)) / 4189 sizeof(ipfw_table_entry); 4190 IPFW_RLOCK(&layer3_chain); 4191 error = dump_table(&layer3_chain, tbl); 4192 IPFW_RUNLOCK(&layer3_chain); 4193 if (error) { 4194 free(tbl, M_TEMP); 4195 break; 4196 } 4197 error = sooptcopyout(sopt, tbl, size); 4198 free(tbl, M_TEMP); 4199 } 4200 break; 4201 4202 default: 4203 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4204 error = EINVAL; 4205 } 4206 4207 return (error); 4208#undef RULE_MAXSIZE 4209} 4210 4211/** 4212 * dummynet needs a reference to the default rule, because rules can be 4213 * deleted while packets hold a reference to them. When this happens, 4214 * dummynet changes the reference to the default rule (it could well be a 4215 * NULL pointer, but this way we do not need to check for the special 4216 * case, plus here he have info on the default behaviour). 4217 */ 4218struct ip_fw *ip_fw_default_rule; 4219 4220/* 4221 * This procedure is only used to handle keepalives. It is invoked 4222 * every dyn_keepalive_period 4223 */ 4224static void 4225ipfw_tick(void * __unused unused) 4226{ 4227 struct mbuf *m0, *m, *mnext, **mtailp; 4228 int i; 4229 ipfw_dyn_rule *q; 4230 4231 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 4232 goto done; 4233 4234 /* 4235 * We make a chain of packets to go out here -- not deferring 4236 * until after we drop the IPFW dynamic rule lock would result 4237 * in a lock order reversal with the normal packet input -> ipfw 4238 * call stack. 4239 */ 4240 m0 = NULL; 4241 mtailp = &m0; 4242 IPFW_DYN_LOCK(); 4243 for (i = 0 ; i < curr_dyn_buckets ; i++) { 4244 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 4245 if (q->dyn_type == O_LIMIT_PARENT) 4246 continue; 4247 if (q->id.proto != IPPROTO_TCP) 4248 continue; 4249 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4250 continue; 4251 if (TIME_LEQ( time_uptime+dyn_keepalive_interval, 4252 q->expire)) 4253 continue; /* too early */ 4254 if (TIME_LEQ(q->expire, time_uptime)) 4255 continue; /* too late, rule expired */ 4256 4257 *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1, 4258 q->ack_fwd, TH_SYN); 4259 if (*mtailp != NULL) 4260 mtailp = &(*mtailp)->m_nextpkt; 4261 *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1, 4262 q->ack_rev, 0); 4263 if (*mtailp != NULL) 4264 mtailp = &(*mtailp)->m_nextpkt; 4265 } 4266 } 4267 IPFW_DYN_UNLOCK(); 4268 for (m = mnext = m0; m != NULL; m = mnext) { 4269 mnext = m->m_nextpkt; 4270 m->m_nextpkt = NULL; 4271 ip_output(m, NULL, NULL, 0, NULL, NULL); 4272 } 4273done: 4274 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 4275} 4276 4277int 4278ipfw_init(void) 4279{ 4280 struct ip_fw default_rule; 4281 int error; 4282 4283#ifdef INET6 4284 /* Setup IPv6 fw sysctl tree. */ 4285 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4286 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4287 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4288 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4289 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4290 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4291 &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4292 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4293 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4294 &fw_deny_unknown_exthdrs, 0, 4295 "Deny packets with unknown IPv6 Extension Headers"); 4296#endif 4297 4298 layer3_chain.rules = NULL; 4299 IPFW_LOCK_INIT(&layer3_chain); 4300 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule zone", 4301 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4302 UMA_ALIGN_PTR, 0); 4303 IPFW_DYN_LOCK_INIT(); 4304 callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE); 4305 4306 bzero(&default_rule, sizeof default_rule); 4307 4308 default_rule.act_ofs = 0; 4309 default_rule.rulenum = IPFW_DEFAULT_RULE; 4310 default_rule.cmd_len = 1; 4311 default_rule.set = RESVD_SET; 4312 4313 default_rule.cmd[0].len = 1; 4314 default_rule.cmd[0].opcode = 4315#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4316 1 ? O_ACCEPT : 4317#endif 4318 O_DENY; 4319 4320 error = add_rule(&layer3_chain, &default_rule); 4321 if (error != 0) { 4322 printf("ipfw2: error %u initializing default rule " 4323 "(support disabled)\n", error); 4324 IPFW_DYN_LOCK_DESTROY(); 4325 IPFW_LOCK_DESTROY(&layer3_chain); 4326 uma_zdestroy(ipfw_dyn_rule_zone); 4327 return (error); 4328 } 4329 4330 ip_fw_default_rule = layer3_chain.rules; 4331 printf("ipfw2 " 4332#ifdef INET6 4333 "(+ipv6) " 4334#endif 4335 "initialized, divert %s, " 4336 "rule-based forwarding " 4337#ifdef IPFIREWALL_FORWARD 4338 "enabled, " 4339#else 4340 "disabled, " 4341#endif 4342 "default to %s, logging ", 4343#ifdef IPDIVERT 4344 "enabled", 4345#else 4346 "loadable", 4347#endif 4348 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4349 4350#ifdef IPFIREWALL_VERBOSE 4351 fw_verbose = 1; 4352#endif 4353#ifdef IPFIREWALL_VERBOSE_LIMIT 4354 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4355#endif 4356 if (fw_verbose == 0) 4357 printf("disabled\n"); 4358 else if (verbose_limit == 0) 4359 printf("unlimited\n"); 4360 else 4361 printf("limited to %d packets/entry by default\n", 4362 verbose_limit); 4363 4364 error = init_tables(&layer3_chain); 4365 if (error) { 4366 IPFW_DYN_LOCK_DESTROY(); 4367 IPFW_LOCK_DESTROY(&layer3_chain); 4368 uma_zdestroy(ipfw_dyn_rule_zone); 4369 return (error); 4370 } 4371 ip_fw_ctl_ptr = ipfw_ctl; 4372 ip_fw_chk_ptr = ipfw_chk; 4373 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 4374 4375 return (0); 4376} 4377 4378void 4379ipfw_destroy(void) 4380{ 4381 struct ip_fw *reap; 4382 4383 ip_fw_chk_ptr = NULL; 4384 ip_fw_ctl_ptr = NULL; 4385 callout_drain(&ipfw_timeout); 4386 IPFW_WLOCK(&layer3_chain); 4387 flush_tables(&layer3_chain); 4388 layer3_chain.reap = NULL; 4389 free_chain(&layer3_chain, 1 /* kill default rule */); 4390 reap = layer3_chain.reap, layer3_chain.reap = NULL; 4391 IPFW_WUNLOCK(&layer3_chain); 4392 if (reap != NULL) 4393 reap_rules(reap); 4394 IPFW_DYN_LOCK_DESTROY(); 4395 uma_zdestroy(ipfw_dyn_rule_zone); 4396 IPFW_LOCK_DESTROY(&layer3_chain); 4397 4398#ifdef INET6 4399 /* Free IPv6 fw sysctl tree. */ 4400 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4401#endif 4402 4403 printf("IP firewall unloaded\n"); 4404} 4405