ip_fw2.c revision 165118
1/*- 2 * Copyright (c) 2002 Luigi Rizzo, Universita` di Pisa 3 * 4 * Redistribution and use in source and binary forms, with or without 5 * modification, are permitted provided that the following conditions 6 * are met: 7 * 1. Redistributions of source code must retain the above copyright 8 * notice, this list of conditions and the following disclaimer. 9 * 2. Redistributions in binary form must reproduce the above copyright 10 * notice, this list of conditions and the following disclaimer in the 11 * documentation and/or other materials provided with the distribution. 12 * 13 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 14 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 15 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 16 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 17 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 18 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 19 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 20 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 21 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 22 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 23 * SUCH DAMAGE. 24 * 25 * $FreeBSD: head/sys/netinet/ip_fw2.c 165118 2006-12-12 12:17:58Z bz $ 26 */ 27 28#define DEB(x) 29#define DDB(x) x 30 31/* 32 * Implement IP packet firewall (new version) 33 */ 34 35#if !defined(KLD_MODULE) 36#include "opt_ipfw.h" 37#include "opt_ipdn.h" 38#include "opt_inet.h" 39#ifndef INET 40#error IPFIREWALL requires INET. 41#endif /* INET */ 42#endif 43#include "opt_inet6.h" 44#include "opt_ipsec.h" 45#include "opt_mac.h" 46 47#include <sys/param.h> 48#include <sys/systm.h> 49#include <sys/condvar.h> 50#include <sys/malloc.h> 51#include <sys/mbuf.h> 52#include <sys/kernel.h> 53#include <sys/lock.h> 54#include <sys/jail.h> 55#include <sys/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 char ip6buf[INET6_ADDRSTRLEN]; 890 snprintf(src, sizeof(src), "[%s]", 891 ip6_sprintf(ip6buf, &args->f_id.src_ip6)); 892 snprintf(dst, sizeof(dst), "[%s]", 893 ip6_sprintf(ip6buf, &args->f_id.dst_ip6)); 894 895 ip6 = (struct ip6_hdr *)mtod(m, struct ip6_hdr *); 896 tcp = (struct tcphdr *)(mtod(args->m, char *) + hlen); 897 udp = (struct udphdr *)(mtod(args->m, char *) + hlen); 898 } else 899#endif 900 { 901 ip = mtod(m, struct ip *); 902 tcp = L3HDR(struct tcphdr, ip); 903 udp = L3HDR(struct udphdr, ip); 904 905 inet_ntoa_r(ip->ip_src, src); 906 inet_ntoa_r(ip->ip_dst, dst); 907 } 908 909 switch (args->f_id.proto) { 910 case IPPROTO_TCP: 911 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 912 if (offset == 0) 913 snprintf(SNPARGS(proto, len), ":%d %s:%d", 914 ntohs(tcp->th_sport), 915 dst, 916 ntohs(tcp->th_dport)); 917 else 918 snprintf(SNPARGS(proto, len), " %s", dst); 919 break; 920 921 case IPPROTO_UDP: 922 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 923 if (offset == 0) 924 snprintf(SNPARGS(proto, len), ":%d %s:%d", 925 ntohs(udp->uh_sport), 926 dst, 927 ntohs(udp->uh_dport)); 928 else 929 snprintf(SNPARGS(proto, len), " %s", dst); 930 break; 931 932 case IPPROTO_ICMP: 933 icmp = L3HDR(struct icmphdr, ip); 934 if (offset == 0) 935 len = snprintf(SNPARGS(proto, 0), 936 "ICMP:%u.%u ", 937 icmp->icmp_type, icmp->icmp_code); 938 else 939 len = snprintf(SNPARGS(proto, 0), "ICMP "); 940 len += snprintf(SNPARGS(proto, len), "%s", src); 941 snprintf(SNPARGS(proto, len), " %s", dst); 942 break; 943#ifdef INET6 944 case IPPROTO_ICMPV6: 945 icmp6 = (struct icmp6_hdr *)(mtod(args->m, char *) + hlen); 946 if (offset == 0) 947 len = snprintf(SNPARGS(proto, 0), 948 "ICMPv6:%u.%u ", 949 icmp6->icmp6_type, icmp6->icmp6_code); 950 else 951 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 952 len += snprintf(SNPARGS(proto, len), "%s", src); 953 snprintf(SNPARGS(proto, len), " %s", dst); 954 break; 955#endif 956 default: 957 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 958 args->f_id.proto, src); 959 snprintf(SNPARGS(proto, len), " %s", dst); 960 break; 961 } 962 963#ifdef INET6 964 if (IS_IP6_FLOW_ID(&(args->f_id))) { 965 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 966 snprintf(SNPARGS(fragment, 0), 967 " (frag %08x:%d@%d%s)", 968 args->f_id.frag_id6, 969 ntohs(ip6->ip6_plen) - hlen, 970 ntohs(offset & IP6F_OFF_MASK) << 3, 971 (offset & IP6F_MORE_FRAG) ? "+" : ""); 972 } else 973#endif 974 { 975 int ip_off, ip_len; 976 if (eh != NULL) { /* layer 2 packets are as on the wire */ 977 ip_off = ntohs(ip->ip_off); 978 ip_len = ntohs(ip->ip_len); 979 } else { 980 ip_off = ip->ip_off; 981 ip_len = ip->ip_len; 982 } 983 if (ip_off & (IP_MF | IP_OFFMASK)) 984 snprintf(SNPARGS(fragment, 0), 985 " (frag %d:%d@%d%s)", 986 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 987 offset << 3, 988 (ip_off & IP_MF) ? "+" : ""); 989 } 990 } 991 if (oif || m->m_pkthdr.rcvif) 992 log(LOG_SECURITY | LOG_INFO, 993 "ipfw: %d %s %s %s via %s%s\n", 994 f ? f->rulenum : -1, 995 action, proto, oif ? "out" : "in", 996 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 997 fragment); 998 else 999 log(LOG_SECURITY | LOG_INFO, 1000 "ipfw: %d %s %s [no if info]%s\n", 1001 f ? f->rulenum : -1, 1002 action, proto, fragment); 1003 if (limit_reached) 1004 log(LOG_SECURITY | LOG_NOTICE, 1005 "ipfw: limit %d reached on entry %d\n", 1006 limit_reached, f ? f->rulenum : -1); 1007} 1008 1009/* 1010 * IMPORTANT: the hash function for dynamic rules must be commutative 1011 * in source and destination (ip,port), because rules are bidirectional 1012 * and we want to find both in the same bucket. 1013 */ 1014static __inline int 1015hash_packet(struct ipfw_flow_id *id) 1016{ 1017 u_int32_t i; 1018 1019#ifdef INET6 1020 if (IS_IP6_FLOW_ID(id)) 1021 i = hash_packet6(id); 1022 else 1023#endif /* INET6 */ 1024 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1025 i &= (curr_dyn_buckets - 1); 1026 return i; 1027} 1028 1029/** 1030 * unlink a dynamic rule from a chain. prev is a pointer to 1031 * the previous one, q is a pointer to the rule to delete, 1032 * head is a pointer to the head of the queue. 1033 * Modifies q and potentially also head. 1034 */ 1035#define UNLINK_DYN_RULE(prev, head, q) { \ 1036 ipfw_dyn_rule *old_q = q; \ 1037 \ 1038 /* remove a refcount to the parent */ \ 1039 if (q->dyn_type == O_LIMIT) \ 1040 q->parent->count--; \ 1041 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1042 (q->id.src_ip), (q->id.src_port), \ 1043 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 1044 if (prev != NULL) \ 1045 prev->next = q = q->next; \ 1046 else \ 1047 head = q = q->next; \ 1048 dyn_count--; \ 1049 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1050 1051#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1052 1053/** 1054 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1055 * 1056 * If keep_me == NULL, rules are deleted even if not expired, 1057 * otherwise only expired rules are removed. 1058 * 1059 * The value of the second parameter is also used to point to identify 1060 * a rule we absolutely do not want to remove (e.g. because we are 1061 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1062 * rules). The pointer is only used for comparison, so any non-null 1063 * value will do. 1064 */ 1065static void 1066remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1067{ 1068 static u_int32_t last_remove = 0; 1069 1070#define FORCE (keep_me == NULL) 1071 1072 ipfw_dyn_rule *prev, *q; 1073 int i, pass = 0, max_pass = 0; 1074 1075 IPFW_DYN_LOCK_ASSERT(); 1076 1077 if (ipfw_dyn_v == NULL || dyn_count == 0) 1078 return; 1079 /* do not expire more than once per second, it is useless */ 1080 if (!FORCE && last_remove == time_uptime) 1081 return; 1082 last_remove = time_uptime; 1083 1084 /* 1085 * because O_LIMIT refer to parent rules, during the first pass only 1086 * remove child and mark any pending LIMIT_PARENT, and remove 1087 * them in a second pass. 1088 */ 1089next_pass: 1090 for (i = 0 ; i < curr_dyn_buckets ; i++) { 1091 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 1092 /* 1093 * Logic can become complex here, so we split tests. 1094 */ 1095 if (q == keep_me) 1096 goto next; 1097 if (rule != NULL && rule != q->rule) 1098 goto next; /* not the one we are looking for */ 1099 if (q->dyn_type == O_LIMIT_PARENT) { 1100 /* 1101 * handle parent in the second pass, 1102 * record we need one. 1103 */ 1104 max_pass = 1; 1105 if (pass == 0) 1106 goto next; 1107 if (FORCE && q->count != 0 ) { 1108 /* XXX should not happen! */ 1109 printf("ipfw: OUCH! cannot remove rule," 1110 " count %d\n", q->count); 1111 } 1112 } else { 1113 if (!FORCE && 1114 !TIME_LEQ( q->expire, time_uptime )) 1115 goto next; 1116 } 1117 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1118 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1119 continue; 1120 } 1121next: 1122 prev=q; 1123 q=q->next; 1124 } 1125 } 1126 if (pass++ < max_pass) 1127 goto next_pass; 1128} 1129 1130 1131/** 1132 * lookup a dynamic rule. 1133 */ 1134static ipfw_dyn_rule * 1135lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1136 struct tcphdr *tcp) 1137{ 1138 /* 1139 * stateful ipfw extensions. 1140 * Lookup into dynamic session queue 1141 */ 1142#define MATCH_REVERSE 0 1143#define MATCH_FORWARD 1 1144#define MATCH_NONE 2 1145#define MATCH_UNKNOWN 3 1146 int i, dir = MATCH_NONE; 1147 ipfw_dyn_rule *prev, *q=NULL; 1148 1149 IPFW_DYN_LOCK_ASSERT(); 1150 1151 if (ipfw_dyn_v == NULL) 1152 goto done; /* not found */ 1153 i = hash_packet( pkt ); 1154 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 1155 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1156 goto next; 1157 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1158 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 1159 continue; 1160 } 1161 if (pkt->proto == q->id.proto && 1162 q->dyn_type != O_LIMIT_PARENT) { 1163 if (IS_IP6_FLOW_ID(pkt)) { 1164 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1165 &(q->id.src_ip6)) && 1166 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1167 &(q->id.dst_ip6)) && 1168 pkt->src_port == q->id.src_port && 1169 pkt->dst_port == q->id.dst_port ) { 1170 dir = MATCH_FORWARD; 1171 break; 1172 } 1173 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1174 &(q->id.dst_ip6)) && 1175 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1176 &(q->id.src_ip6)) && 1177 pkt->src_port == q->id.dst_port && 1178 pkt->dst_port == q->id.src_port ) { 1179 dir = MATCH_REVERSE; 1180 break; 1181 } 1182 } else { 1183 if (pkt->src_ip == q->id.src_ip && 1184 pkt->dst_ip == q->id.dst_ip && 1185 pkt->src_port == q->id.src_port && 1186 pkt->dst_port == q->id.dst_port ) { 1187 dir = MATCH_FORWARD; 1188 break; 1189 } 1190 if (pkt->src_ip == q->id.dst_ip && 1191 pkt->dst_ip == q->id.src_ip && 1192 pkt->src_port == q->id.dst_port && 1193 pkt->dst_port == q->id.src_port ) { 1194 dir = MATCH_REVERSE; 1195 break; 1196 } 1197 } 1198 } 1199next: 1200 prev = q; 1201 q = q->next; 1202 } 1203 if (q == NULL) 1204 goto done; /* q = NULL, not found */ 1205 1206 if ( prev != NULL) { /* found and not in front */ 1207 prev->next = q->next; 1208 q->next = ipfw_dyn_v[i]; 1209 ipfw_dyn_v[i] = q; 1210 } 1211 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1212 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1213 1214#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1215#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1216 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1217 switch (q->state) { 1218 case TH_SYN: /* opening */ 1219 q->expire = time_uptime + dyn_syn_lifetime; 1220 break; 1221 1222 case BOTH_SYN: /* move to established */ 1223 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1224 case BOTH_SYN | (TH_FIN << 8) : 1225 if (tcp) { 1226#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1227 u_int32_t ack = ntohl(tcp->th_ack); 1228 if (dir == MATCH_FORWARD) { 1229 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1230 q->ack_fwd = ack; 1231 else { /* ignore out-of-sequence */ 1232 break; 1233 } 1234 } else { 1235 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1236 q->ack_rev = ack; 1237 else { /* ignore out-of-sequence */ 1238 break; 1239 } 1240 } 1241 } 1242 q->expire = time_uptime + dyn_ack_lifetime; 1243 break; 1244 1245 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1246 if (dyn_fin_lifetime >= dyn_keepalive_period) 1247 dyn_fin_lifetime = dyn_keepalive_period - 1; 1248 q->expire = time_uptime + dyn_fin_lifetime; 1249 break; 1250 1251 default: 1252#if 0 1253 /* 1254 * reset or some invalid combination, but can also 1255 * occur if we use keep-state the wrong way. 1256 */ 1257 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1258 printf("invalid state: 0x%x\n", q->state); 1259#endif 1260 if (dyn_rst_lifetime >= dyn_keepalive_period) 1261 dyn_rst_lifetime = dyn_keepalive_period - 1; 1262 q->expire = time_uptime + dyn_rst_lifetime; 1263 break; 1264 } 1265 } else if (pkt->proto == IPPROTO_UDP) { 1266 q->expire = time_uptime + dyn_udp_lifetime; 1267 } else { 1268 /* other protocols */ 1269 q->expire = time_uptime + dyn_short_lifetime; 1270 } 1271done: 1272 if (match_direction) 1273 *match_direction = dir; 1274 return q; 1275} 1276 1277static ipfw_dyn_rule * 1278lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1279 struct tcphdr *tcp) 1280{ 1281 ipfw_dyn_rule *q; 1282 1283 IPFW_DYN_LOCK(); 1284 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1285 if (q == NULL) 1286 IPFW_DYN_UNLOCK(); 1287 /* NB: return table locked when q is not NULL */ 1288 return q; 1289} 1290 1291static void 1292realloc_dynamic_table(void) 1293{ 1294 IPFW_DYN_LOCK_ASSERT(); 1295 1296 /* 1297 * Try reallocation, make sure we have a power of 2 and do 1298 * not allow more than 64k entries. In case of overflow, 1299 * default to 1024. 1300 */ 1301 1302 if (dyn_buckets > 65536) 1303 dyn_buckets = 1024; 1304 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1305 dyn_buckets = curr_dyn_buckets; /* reset */ 1306 return; 1307 } 1308 curr_dyn_buckets = dyn_buckets; 1309 if (ipfw_dyn_v != NULL) 1310 free(ipfw_dyn_v, M_IPFW); 1311 for (;;) { 1312 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1313 M_IPFW, M_NOWAIT | M_ZERO); 1314 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 1315 break; 1316 curr_dyn_buckets /= 2; 1317 } 1318} 1319 1320/** 1321 * Install state of type 'type' for a dynamic session. 1322 * The hash table contains two type of rules: 1323 * - regular rules (O_KEEP_STATE) 1324 * - rules for sessions with limited number of sess per user 1325 * (O_LIMIT). When they are created, the parent is 1326 * increased by 1, and decreased on delete. In this case, 1327 * the third parameter is the parent rule and not the chain. 1328 * - "parent" rules for the above (O_LIMIT_PARENT). 1329 */ 1330static ipfw_dyn_rule * 1331add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1332{ 1333 ipfw_dyn_rule *r; 1334 int i; 1335 1336 IPFW_DYN_LOCK_ASSERT(); 1337 1338 if (ipfw_dyn_v == NULL || 1339 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1340 realloc_dynamic_table(); 1341 if (ipfw_dyn_v == NULL) 1342 return NULL; /* failed ! */ 1343 } 1344 i = hash_packet(id); 1345 1346 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1347 if (r == NULL) { 1348 printf ("ipfw: sorry cannot allocate state\n"); 1349 return NULL; 1350 } 1351 1352 /* increase refcount on parent, and set pointer */ 1353 if (dyn_type == O_LIMIT) { 1354 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1355 if ( parent->dyn_type != O_LIMIT_PARENT) 1356 panic("invalid parent"); 1357 parent->count++; 1358 r->parent = parent; 1359 rule = parent->rule; 1360 } 1361 1362 r->id = *id; 1363 r->expire = time_uptime + dyn_syn_lifetime; 1364 r->rule = rule; 1365 r->dyn_type = dyn_type; 1366 r->pcnt = r->bcnt = 0; 1367 r->count = 0; 1368 1369 r->bucket = i; 1370 r->next = ipfw_dyn_v[i]; 1371 ipfw_dyn_v[i] = r; 1372 dyn_count++; 1373 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1374 dyn_type, 1375 (r->id.src_ip), (r->id.src_port), 1376 (r->id.dst_ip), (r->id.dst_port), 1377 dyn_count ); ) 1378 return r; 1379} 1380 1381/** 1382 * lookup dynamic parent rule using pkt and rule as search keys. 1383 * If the lookup fails, then install one. 1384 */ 1385static ipfw_dyn_rule * 1386lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1387{ 1388 ipfw_dyn_rule *q; 1389 int i; 1390 1391 IPFW_DYN_LOCK_ASSERT(); 1392 1393 if (ipfw_dyn_v) { 1394 int is_v6 = IS_IP6_FLOW_ID(pkt); 1395 i = hash_packet( pkt ); 1396 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1397 if (q->dyn_type == O_LIMIT_PARENT && 1398 rule== q->rule && 1399 pkt->proto == q->id.proto && 1400 pkt->src_port == q->id.src_port && 1401 pkt->dst_port == q->id.dst_port && 1402 ( 1403 (is_v6 && 1404 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1405 &(q->id.src_ip6)) && 1406 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1407 &(q->id.dst_ip6))) || 1408 (!is_v6 && 1409 pkt->src_ip == q->id.src_ip && 1410 pkt->dst_ip == q->id.dst_ip) 1411 ) 1412 ) { 1413 q->expire = time_uptime + dyn_short_lifetime; 1414 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1415 return q; 1416 } 1417 } 1418 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1419} 1420 1421/** 1422 * Install dynamic state for rule type cmd->o.opcode 1423 * 1424 * Returns 1 (failure) if state is not installed because of errors or because 1425 * session limitations are enforced. 1426 */ 1427static int 1428install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1429 struct ip_fw_args *args, uint32_t tablearg) 1430{ 1431 static int last_log; 1432 ipfw_dyn_rule *q; 1433 struct in_addr da; 1434 char src[48], dst[48]; 1435 1436 src[0] = '\0'; 1437 dst[0] = '\0'; 1438 1439 DEB( 1440 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1441 __func__, cmd->o.opcode, 1442 (args->f_id.src_ip), (args->f_id.src_port), 1443 (args->f_id.dst_ip), (args->f_id.dst_port)); 1444 ) 1445 1446 IPFW_DYN_LOCK(); 1447 1448 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1449 1450 if (q != NULL) { /* should never occur */ 1451 if (last_log != time_uptime) { 1452 last_log = time_uptime; 1453 printf("ipfw: %s: entry already present, done\n", 1454 __func__); 1455 } 1456 IPFW_DYN_UNLOCK(); 1457 return (0); 1458 } 1459 1460 if (dyn_count >= dyn_max) 1461 /* Run out of slots, try to remove any expired rule. */ 1462 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1463 1464 if (dyn_count >= dyn_max) { 1465 if (last_log != time_uptime) { 1466 last_log = time_uptime; 1467 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1468 } 1469 IPFW_DYN_UNLOCK(); 1470 return (1); /* cannot install, notify caller */ 1471 } 1472 1473 switch (cmd->o.opcode) { 1474 case O_KEEP_STATE: /* bidir rule */ 1475 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1476 break; 1477 1478 case O_LIMIT: { /* limit number of sessions */ 1479 struct ipfw_flow_id id; 1480 ipfw_dyn_rule *parent; 1481 uint32_t conn_limit; 1482 uint16_t limit_mask = cmd->limit_mask; 1483 1484 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1485 tablearg : cmd->conn_limit; 1486 1487 DEB( 1488 if (cmd->conn_limit == IP_FW_TABLEARG) 1489 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1490 "(tablearg)\n", __func__, conn_limit); 1491 else 1492 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1493 __func__, conn_limit); 1494 ) 1495 1496 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1497 id.proto = args->f_id.proto; 1498 id.addr_type = args->f_id.addr_type; 1499 1500 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1501 if (limit_mask & DYN_SRC_ADDR) 1502 id.src_ip6 = args->f_id.src_ip6; 1503 if (limit_mask & DYN_DST_ADDR) 1504 id.dst_ip6 = args->f_id.dst_ip6; 1505 } else { 1506 if (limit_mask & DYN_SRC_ADDR) 1507 id.src_ip = args->f_id.src_ip; 1508 if (limit_mask & DYN_DST_ADDR) 1509 id.dst_ip = args->f_id.dst_ip; 1510 } 1511 if (limit_mask & DYN_SRC_PORT) 1512 id.src_port = args->f_id.src_port; 1513 if (limit_mask & DYN_DST_PORT) 1514 id.dst_port = args->f_id.dst_port; 1515 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1516 printf("ipfw: %s: add parent failed\n", __func__); 1517 IPFW_DYN_UNLOCK(); 1518 return (1); 1519 } 1520 1521 if (parent->count >= conn_limit) { 1522 /* See if we can remove some expired rule. */ 1523 remove_dyn_rule(rule, parent); 1524 if (parent->count >= conn_limit) { 1525 if (fw_verbose && last_log != time_uptime) { 1526 last_log = time_uptime; 1527#ifdef INET6 1528 /* 1529 * XXX IPv6 flows are not 1530 * supported yet. 1531 */ 1532 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1533 char ip6buf[INET6_ADDRSTRLEN]; 1534 snprintf(src, sizeof(src), 1535 "[%s]", ip6_sprintf(ip6buf, 1536 &args->f_id.src_ip6)); 1537 snprintf(dst, sizeof(dst), 1538 "[%s]", ip6_sprintf(ip6buf, 1539 &args->f_id.dst_ip6)); 1540 } else 1541#endif 1542 { 1543 da.s_addr = 1544 htonl(args->f_id.src_ip); 1545 inet_ntoa_r(da, src); 1546 da.s_addr = 1547 htonl(args->f_id.dst_ip); 1548 inet_ntoa_r(da, dst); 1549 } 1550 log(LOG_SECURITY | LOG_DEBUG, 1551 "%s %s:%u -> %s:%u, %s\n", 1552 "drop session", 1553 src, (args->f_id.src_port), 1554 dst, (args->f_id.dst_port), 1555 "too many entries"); 1556 } 1557 IPFW_DYN_UNLOCK(); 1558 return (1); 1559 } 1560 } 1561 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1562 break; 1563 } 1564 default: 1565 printf("ipfw: %s: unknown dynamic rule type %u\n", 1566 __func__, cmd->o.opcode); 1567 IPFW_DYN_UNLOCK(); 1568 return (1); 1569 } 1570 1571 /* XXX just set lifetime */ 1572 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1573 1574 IPFW_DYN_UNLOCK(); 1575 return (0); 1576} 1577 1578/* 1579 * Generate a TCP packet, containing either a RST or a keepalive. 1580 * When flags & TH_RST, we are sending a RST packet, because of a 1581 * "reset" action matched the packet. 1582 * Otherwise we are sending a keepalive, and flags & TH_ 1583 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 1584 * so that MAC can label the reply appropriately. 1585 */ 1586static struct mbuf * 1587send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 1588 u_int32_t ack, int flags) 1589{ 1590 struct mbuf *m; 1591 struct ip *ip; 1592 struct tcphdr *tcp; 1593 1594 MGETHDR(m, M_DONTWAIT, MT_DATA); 1595 if (m == 0) 1596 return (NULL); 1597 m->m_pkthdr.rcvif = (struct ifnet *)0; 1598 1599#ifdef MAC 1600 if (replyto != NULL) 1601 mac_create_mbuf_netlayer(replyto, m); 1602 else 1603 mac_create_mbuf_from_firewall(m); 1604#else 1605 (void)replyto; /* don't warn about unused arg */ 1606#endif 1607 1608 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1609 m->m_data += max_linkhdr; 1610 1611 ip = mtod(m, struct ip *); 1612 bzero(ip, m->m_len); 1613 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1614 ip->ip_p = IPPROTO_TCP; 1615 tcp->th_off = 5; 1616 /* 1617 * Assume we are sending a RST (or a keepalive in the reverse 1618 * direction), swap src and destination addresses and ports. 1619 */ 1620 ip->ip_src.s_addr = htonl(id->dst_ip); 1621 ip->ip_dst.s_addr = htonl(id->src_ip); 1622 tcp->th_sport = htons(id->dst_port); 1623 tcp->th_dport = htons(id->src_port); 1624 if (flags & TH_RST) { /* we are sending a RST */ 1625 if (flags & TH_ACK) { 1626 tcp->th_seq = htonl(ack); 1627 tcp->th_ack = htonl(0); 1628 tcp->th_flags = TH_RST; 1629 } else { 1630 if (flags & TH_SYN) 1631 seq++; 1632 tcp->th_seq = htonl(0); 1633 tcp->th_ack = htonl(seq); 1634 tcp->th_flags = TH_RST | TH_ACK; 1635 } 1636 } else { 1637 /* 1638 * We are sending a keepalive. flags & TH_SYN determines 1639 * the direction, forward if set, reverse if clear. 1640 * NOTE: seq and ack are always assumed to be correct 1641 * as set by the caller. This may be confusing... 1642 */ 1643 if (flags & TH_SYN) { 1644 /* 1645 * we have to rewrite the correct addresses! 1646 */ 1647 ip->ip_dst.s_addr = htonl(id->dst_ip); 1648 ip->ip_src.s_addr = htonl(id->src_ip); 1649 tcp->th_dport = htons(id->dst_port); 1650 tcp->th_sport = htons(id->src_port); 1651 } 1652 tcp->th_seq = htonl(seq); 1653 tcp->th_ack = htonl(ack); 1654 tcp->th_flags = TH_ACK; 1655 } 1656 /* 1657 * set ip_len to the payload size so we can compute 1658 * the tcp checksum on the pseudoheader 1659 * XXX check this, could save a couple of words ? 1660 */ 1661 ip->ip_len = htons(sizeof(struct tcphdr)); 1662 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1663 /* 1664 * now fill fields left out earlier 1665 */ 1666 ip->ip_ttl = ip_defttl; 1667 ip->ip_len = m->m_pkthdr.len; 1668 m->m_flags |= M_SKIP_FIREWALL; 1669 return (m); 1670} 1671 1672/* 1673 * sends a reject message, consuming the mbuf passed as an argument. 1674 */ 1675static void 1676send_reject(struct ip_fw_args *args, int code, int ip_len) 1677{ 1678 1679 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1680 /* We need the IP header in host order for icmp_error(). */ 1681 if (args->eh != NULL) { 1682 struct ip *ip = mtod(args->m, struct ip *); 1683 ip->ip_len = ntohs(ip->ip_len); 1684 ip->ip_off = ntohs(ip->ip_off); 1685 } 1686 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1687 } else if (args->f_id.proto == IPPROTO_TCP) { 1688 struct tcphdr *const tcp = 1689 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1690 if ( (tcp->th_flags & TH_RST) == 0) { 1691 struct mbuf *m; 1692 m = send_pkt(args->m, &(args->f_id), 1693 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1694 tcp->th_flags | TH_RST); 1695 if (m != NULL) 1696 ip_output(m, NULL, NULL, 0, NULL, NULL); 1697 } 1698 m_freem(args->m); 1699 } else 1700 m_freem(args->m); 1701 args->m = NULL; 1702} 1703 1704/** 1705 * 1706 * Given an ip_fw *, lookup_next_rule will return a pointer 1707 * to the next rule, which can be either the jump 1708 * target (for skipto instructions) or the next one in the list (in 1709 * all other cases including a missing jump target). 1710 * The result is also written in the "next_rule" field of the rule. 1711 * Backward jumps are not allowed, so start looking from the next 1712 * rule... 1713 * 1714 * This never returns NULL -- in case we do not have an exact match, 1715 * the next rule is returned. When the ruleset is changed, 1716 * pointers are flushed so we are always correct. 1717 */ 1718 1719static struct ip_fw * 1720lookup_next_rule(struct ip_fw *me) 1721{ 1722 struct ip_fw *rule = NULL; 1723 ipfw_insn *cmd; 1724 1725 /* look for action, in case it is a skipto */ 1726 cmd = ACTION_PTR(me); 1727 if (cmd->opcode == O_LOG) 1728 cmd += F_LEN(cmd); 1729 if (cmd->opcode == O_ALTQ) 1730 cmd += F_LEN(cmd); 1731 if (cmd->opcode == O_TAG) 1732 cmd += F_LEN(cmd); 1733 if ( cmd->opcode == O_SKIPTO ) 1734 for (rule = me->next; rule ; rule = rule->next) 1735 if (rule->rulenum >= cmd->arg1) 1736 break; 1737 if (rule == NULL) /* failure or not a skipto */ 1738 rule = me->next; 1739 me->next_rule = rule; 1740 return rule; 1741} 1742 1743static int 1744add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1745 uint8_t mlen, uint32_t value) 1746{ 1747 struct radix_node_head *rnh; 1748 struct table_entry *ent; 1749 1750 if (tbl >= IPFW_TABLES_MAX) 1751 return (EINVAL); 1752 rnh = ch->tables[tbl]; 1753 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1754 if (ent == NULL) 1755 return (ENOMEM); 1756 ent->value = value; 1757 ent->addr.sin_len = ent->mask.sin_len = 8; 1758 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1759 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1760 IPFW_WLOCK(&layer3_chain); 1761 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1762 NULL) { 1763 IPFW_WUNLOCK(&layer3_chain); 1764 free(ent, M_IPFW_TBL); 1765 return (EEXIST); 1766 } 1767 IPFW_WUNLOCK(&layer3_chain); 1768 return (0); 1769} 1770 1771static int 1772del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1773 uint8_t mlen) 1774{ 1775 struct radix_node_head *rnh; 1776 struct table_entry *ent; 1777 struct sockaddr_in sa, mask; 1778 1779 if (tbl >= IPFW_TABLES_MAX) 1780 return (EINVAL); 1781 rnh = ch->tables[tbl]; 1782 sa.sin_len = mask.sin_len = 8; 1783 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1784 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1785 IPFW_WLOCK(ch); 1786 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1787 if (ent == NULL) { 1788 IPFW_WUNLOCK(ch); 1789 return (ESRCH); 1790 } 1791 IPFW_WUNLOCK(ch); 1792 free(ent, M_IPFW_TBL); 1793 return (0); 1794} 1795 1796static int 1797flush_table_entry(struct radix_node *rn, void *arg) 1798{ 1799 struct radix_node_head * const rnh = arg; 1800 struct table_entry *ent; 1801 1802 ent = (struct table_entry *) 1803 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1804 if (ent != NULL) 1805 free(ent, M_IPFW_TBL); 1806 return (0); 1807} 1808 1809static int 1810flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1811{ 1812 struct radix_node_head *rnh; 1813 1814 IPFW_WLOCK_ASSERT(ch); 1815 1816 if (tbl >= IPFW_TABLES_MAX) 1817 return (EINVAL); 1818 rnh = ch->tables[tbl]; 1819 KASSERT(rnh != NULL, ("NULL IPFW table")); 1820 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1821 return (0); 1822} 1823 1824static void 1825flush_tables(struct ip_fw_chain *ch) 1826{ 1827 uint16_t tbl; 1828 1829 IPFW_WLOCK_ASSERT(ch); 1830 1831 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1832 flush_table(ch, tbl); 1833} 1834 1835static int 1836init_tables(struct ip_fw_chain *ch) 1837{ 1838 int i; 1839 uint16_t j; 1840 1841 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1842 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1843 for (j = 0; j < i; j++) { 1844 (void) flush_table(ch, j); 1845 } 1846 return (ENOMEM); 1847 } 1848 } 1849 return (0); 1850} 1851 1852static int 1853lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1854 uint32_t *val) 1855{ 1856 struct radix_node_head *rnh; 1857 struct table_entry *ent; 1858 struct sockaddr_in sa; 1859 1860 if (tbl >= IPFW_TABLES_MAX) 1861 return (0); 1862 rnh = ch->tables[tbl]; 1863 sa.sin_len = 8; 1864 sa.sin_addr.s_addr = addr; 1865 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1866 if (ent != NULL) { 1867 *val = ent->value; 1868 return (1); 1869 } 1870 return (0); 1871} 1872 1873static int 1874count_table_entry(struct radix_node *rn, void *arg) 1875{ 1876 u_int32_t * const cnt = arg; 1877 1878 (*cnt)++; 1879 return (0); 1880} 1881 1882static int 1883count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1884{ 1885 struct radix_node_head *rnh; 1886 1887 if (tbl >= IPFW_TABLES_MAX) 1888 return (EINVAL); 1889 rnh = ch->tables[tbl]; 1890 *cnt = 0; 1891 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1892 return (0); 1893} 1894 1895static int 1896dump_table_entry(struct radix_node *rn, void *arg) 1897{ 1898 struct table_entry * const n = (struct table_entry *)rn; 1899 ipfw_table * const tbl = arg; 1900 ipfw_table_entry *ent; 1901 1902 if (tbl->cnt == tbl->size) 1903 return (1); 1904 ent = &tbl->ent[tbl->cnt]; 1905 ent->tbl = tbl->tbl; 1906 if (in_nullhost(n->mask.sin_addr)) 1907 ent->masklen = 0; 1908 else 1909 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1910 ent->addr = n->addr.sin_addr.s_addr; 1911 ent->value = n->value; 1912 tbl->cnt++; 1913 return (0); 1914} 1915 1916static int 1917dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1918{ 1919 struct radix_node_head *rnh; 1920 1921 if (tbl->tbl >= IPFW_TABLES_MAX) 1922 return (EINVAL); 1923 rnh = ch->tables[tbl->tbl]; 1924 tbl->cnt = 0; 1925 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1926 return (0); 1927} 1928 1929static void 1930fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1931{ 1932 struct ucred *cr; 1933 1934 if (inp->inp_socket != NULL) { 1935 cr = inp->inp_socket->so_cred; 1936 ugp->fw_prid = jailed(cr) ? 1937 cr->cr_prison->pr_id : -1; 1938 ugp->fw_uid = cr->cr_uid; 1939 ugp->fw_ngroups = cr->cr_ngroups; 1940 bcopy(cr->cr_groups, ugp->fw_groups, 1941 sizeof(ugp->fw_groups)); 1942 } 1943} 1944 1945static int 1946check_uidgid(ipfw_insn_u32 *insn, 1947 int proto, struct ifnet *oif, 1948 struct in_addr dst_ip, u_int16_t dst_port, 1949 struct in_addr src_ip, u_int16_t src_port, 1950 struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 1951{ 1952 struct inpcbinfo *pi; 1953 int wildcard; 1954 struct inpcb *pcb; 1955 int match; 1956 gid_t *gp; 1957 1958 /* 1959 * Check to see if the UDP or TCP stack supplied us with 1960 * the PCB. If so, rather then holding a lock and looking 1961 * up the PCB, we can use the one that was supplied. 1962 */ 1963 if (inp && *lookup == 0) { 1964 INP_LOCK_ASSERT(inp); 1965 if (inp->inp_socket != NULL) { 1966 fill_ugid_cache(inp, ugp); 1967 *lookup = 1; 1968 } 1969 } 1970 /* 1971 * If we have already been here and the packet has no 1972 * PCB entry associated with it, then we can safely 1973 * assume that this is a no match. 1974 */ 1975 if (*lookup == -1) 1976 return (0); 1977 if (proto == IPPROTO_TCP) { 1978 wildcard = 0; 1979 pi = &tcbinfo; 1980 } else if (proto == IPPROTO_UDP) { 1981 wildcard = INPLOOKUP_WILDCARD; 1982 pi = &udbinfo; 1983 } else 1984 return 0; 1985 match = 0; 1986 if (*lookup == 0) { 1987 INP_INFO_RLOCK(pi); 1988 pcb = (oif) ? 1989 in_pcblookup_hash(pi, 1990 dst_ip, htons(dst_port), 1991 src_ip, htons(src_port), 1992 wildcard, oif) : 1993 in_pcblookup_hash(pi, 1994 src_ip, htons(src_port), 1995 dst_ip, htons(dst_port), 1996 wildcard, NULL); 1997 if (pcb != NULL) { 1998 INP_LOCK(pcb); 1999 if (pcb->inp_socket != NULL) { 2000 fill_ugid_cache(pcb, ugp); 2001 *lookup = 1; 2002 } 2003 INP_UNLOCK(pcb); 2004 } 2005 INP_INFO_RUNLOCK(pi); 2006 if (*lookup == 0) { 2007 /* 2008 * If the lookup did not yield any results, there 2009 * is no sense in coming back and trying again. So 2010 * we can set lookup to -1 and ensure that we wont 2011 * bother the pcb system again. 2012 */ 2013 *lookup = -1; 2014 return (0); 2015 } 2016 } 2017 if (insn->o.opcode == O_UID) 2018 match = (ugp->fw_uid == (uid_t)insn->d[0]); 2019 else if (insn->o.opcode == O_GID) { 2020 for (gp = ugp->fw_groups; 2021 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 2022 if (*gp == (gid_t)insn->d[0]) { 2023 match = 1; 2024 break; 2025 } 2026 } else if (insn->o.opcode == O_JAIL) 2027 match = (ugp->fw_prid == (int)insn->d[0]); 2028 return match; 2029} 2030 2031/* 2032 * The main check routine for the firewall. 2033 * 2034 * All arguments are in args so we can modify them and return them 2035 * back to the caller. 2036 * 2037 * Parameters: 2038 * 2039 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 2040 * Starts with the IP header. 2041 * args->eh (in) Mac header if present, or NULL for layer3 packet. 2042 * args->oif Outgoing interface, or NULL if packet is incoming. 2043 * The incoming interface is in the mbuf. (in) 2044 * args->divert_rule (in/out) 2045 * Skip up to the first rule past this rule number; 2046 * upon return, non-zero port number for divert or tee. 2047 * 2048 * args->rule Pointer to the last matching rule (in/out) 2049 * args->next_hop Socket we are forwarding to (out). 2050 * args->f_id Addresses grabbed from the packet (out) 2051 * args->cookie a cookie depending on rule action 2052 * 2053 * Return value: 2054 * 2055 * IP_FW_PASS the packet must be accepted 2056 * IP_FW_DENY the packet must be dropped 2057 * IP_FW_DIVERT divert packet, port in m_tag 2058 * IP_FW_TEE tee packet, port in m_tag 2059 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2060 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2061 * 2062 */ 2063 2064int 2065ipfw_chk(struct ip_fw_args *args) 2066{ 2067 /* 2068 * Local variables hold state during the processing of a packet. 2069 * 2070 * IMPORTANT NOTE: to speed up the processing of rules, there 2071 * are some assumption on the values of the variables, which 2072 * are documented here. Should you change them, please check 2073 * the implementation of the various instructions to make sure 2074 * that they still work. 2075 * 2076 * args->eh The MAC header. It is non-null for a layer2 2077 * packet, it is NULL for a layer-3 packet. 2078 * 2079 * m | args->m Pointer to the mbuf, as received from the caller. 2080 * It may change if ipfw_chk() does an m_pullup, or if it 2081 * consumes the packet because it calls send_reject(). 2082 * XXX This has to change, so that ipfw_chk() never modifies 2083 * or consumes the buffer. 2084 * ip is simply an alias of the value of m, and it is kept 2085 * in sync with it (the packet is supposed to start with 2086 * the ip header). 2087 */ 2088 struct mbuf *m = args->m; 2089 struct ip *ip = mtod(m, struct ip *); 2090 2091 /* 2092 * For rules which contain uid/gid or jail constraints, cache 2093 * a copy of the users credentials after the pcb lookup has been 2094 * executed. This will speed up the processing of rules with 2095 * these types of constraints, as well as decrease contention 2096 * on pcb related locks. 2097 */ 2098 struct ip_fw_ugid fw_ugid_cache; 2099 int ugid_lookup = 0; 2100 2101 /* 2102 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2103 * associated with a packet input on a divert socket. This 2104 * will allow to distinguish traffic and its direction when 2105 * it originates from a divert socket. 2106 */ 2107 u_int divinput_flags = 0; 2108 2109 /* 2110 * oif | args->oif If NULL, ipfw_chk has been called on the 2111 * inbound path (ether_input, ip_input). 2112 * If non-NULL, ipfw_chk has been called on the outbound path 2113 * (ether_output, ip_output). 2114 */ 2115 struct ifnet *oif = args->oif; 2116 2117 struct ip_fw *f = NULL; /* matching rule */ 2118 int retval = 0; 2119 2120 /* 2121 * hlen The length of the IP header. 2122 */ 2123 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2124 2125 /* 2126 * offset The offset of a fragment. offset != 0 means that 2127 * we have a fragment at this offset of an IPv4 packet. 2128 * offset == 0 means that (if this is an IPv4 packet) 2129 * this is the first or only fragment. 2130 * For IPv6 offset == 0 means there is no Fragment Header. 2131 * If offset != 0 for IPv6 always use correct mask to 2132 * get the correct offset because we add IP6F_MORE_FRAG 2133 * to be able to dectect the first fragment which would 2134 * otherwise have offset = 0. 2135 */ 2136 u_short offset = 0; 2137 2138 /* 2139 * Local copies of addresses. They are only valid if we have 2140 * an IP packet. 2141 * 2142 * proto The protocol. Set to 0 for non-ip packets, 2143 * or to the protocol read from the packet otherwise. 2144 * proto != 0 means that we have an IPv4 packet. 2145 * 2146 * src_port, dst_port port numbers, in HOST format. Only 2147 * valid for TCP and UDP packets. 2148 * 2149 * src_ip, dst_ip ip addresses, in NETWORK format. 2150 * Only valid for IPv4 packets. 2151 */ 2152 u_int8_t proto; 2153 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2154 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2155 u_int16_t ip_len=0; 2156 int pktlen; 2157 2158 /* 2159 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2160 * MATCH_NONE when checked and not matched (q = NULL), 2161 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2162 */ 2163 int dyn_dir = MATCH_UNKNOWN; 2164 ipfw_dyn_rule *q = NULL; 2165 struct ip_fw_chain *chain = &layer3_chain; 2166 struct m_tag *mtag; 2167 2168 /* 2169 * We store in ulp a pointer to the upper layer protocol header. 2170 * In the ipv4 case this is easy to determine from the header, 2171 * but for ipv6 we might have some additional headers in the middle. 2172 * ulp is NULL if not found. 2173 */ 2174 void *ulp = NULL; /* upper layer protocol pointer. */ 2175 /* XXX ipv6 variables */ 2176 int is_ipv6 = 0; 2177 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2178 /* end of ipv6 variables */ 2179 int is_ipv4 = 0; 2180 2181 if (m->m_flags & M_SKIP_FIREWALL) 2182 return (IP_FW_PASS); /* accept */ 2183 2184 pktlen = m->m_pkthdr.len; 2185 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2186 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2187 2188/* 2189 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2190 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2191 * pointer might become stale after other pullups (but we never use it 2192 * this way). 2193 */ 2194#define PULLUP_TO(len, p, T) \ 2195do { \ 2196 int x = (len) + sizeof(T); \ 2197 if ((m)->m_len < x) { \ 2198 args->m = m = m_pullup(m, x); \ 2199 if (m == NULL) \ 2200 goto pullup_failed; \ 2201 } \ 2202 p = (mtod(m, char *) + (len)); \ 2203} while (0) 2204 2205 /* Identify IP packets and fill up variables. */ 2206 if (pktlen >= sizeof(struct ip6_hdr) && 2207 (args->eh == NULL || ntohs(args->eh->ether_type)==ETHERTYPE_IPV6) && 2208 mtod(m, struct ip *)->ip_v == 6) { 2209 is_ipv6 = 1; 2210 args->f_id.addr_type = 6; 2211 hlen = sizeof(struct ip6_hdr); 2212 proto = mtod(m, struct ip6_hdr *)->ip6_nxt; 2213 2214 /* Search extension headers to find upper layer protocols */ 2215 while (ulp == NULL) { 2216 switch (proto) { 2217 case IPPROTO_ICMPV6: 2218 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2219 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2220 break; 2221 2222 case IPPROTO_TCP: 2223 PULLUP_TO(hlen, ulp, struct tcphdr); 2224 dst_port = TCP(ulp)->th_dport; 2225 src_port = TCP(ulp)->th_sport; 2226 args->f_id.flags = TCP(ulp)->th_flags; 2227 break; 2228 2229 case IPPROTO_SCTP: 2230 PULLUP_TO(hlen, ulp, struct sctphdr); 2231 src_port = SCTP(ulp)->src_port; 2232 dst_port = SCTP(ulp)->dest_port; 2233 break; 2234 2235 case IPPROTO_UDP: 2236 PULLUP_TO(hlen, ulp, struct udphdr); 2237 dst_port = UDP(ulp)->uh_dport; 2238 src_port = UDP(ulp)->uh_sport; 2239 break; 2240 2241 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2242 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2243 ext_hd |= EXT_HOPOPTS; 2244 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2245 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2246 ulp = NULL; 2247 break; 2248 2249 case IPPROTO_ROUTING: /* RFC 2460 */ 2250 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2251 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2252 case 0: 2253 break; 2254 default: 2255 printf("IPFW2: IPV6 - Unknown Routing " 2256 "Header type(%d)\n", 2257 ((struct ip6_rthdr *)ulp)->ip6r_type); 2258 if (fw_deny_unknown_exthdrs) 2259 return (IP_FW_DENY); 2260 break; 2261 } 2262 ext_hd |= EXT_ROUTING; 2263 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2264 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2265 ulp = NULL; 2266 break; 2267 2268 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2269 PULLUP_TO(hlen, ulp, struct ip6_frag); 2270 ext_hd |= EXT_FRAGMENT; 2271 hlen += sizeof (struct ip6_frag); 2272 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2273 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2274 IP6F_OFF_MASK; 2275 /* Add IP6F_MORE_FRAG for offset of first 2276 * fragment to be != 0. */ 2277 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2278 IP6F_MORE_FRAG; 2279 if (offset == 0) { 2280 printf("IPFW2: IPV6 - Invalid Fragment " 2281 "Header\n"); 2282 if (fw_deny_unknown_exthdrs) 2283 return (IP_FW_DENY); 2284 break; 2285 } 2286 args->f_id.frag_id6 = 2287 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2288 ulp = NULL; 2289 break; 2290 2291 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2292 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2293 ext_hd |= EXT_DSTOPTS; 2294 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2295 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2296 ulp = NULL; 2297 break; 2298 2299 case IPPROTO_AH: /* RFC 2402 */ 2300 PULLUP_TO(hlen, ulp, struct ip6_ext); 2301 ext_hd |= EXT_AH; 2302 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2303 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2304 ulp = NULL; 2305 break; 2306 2307 case IPPROTO_ESP: /* RFC 2406 */ 2308 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2309 /* Anything past Seq# is variable length and 2310 * data past this ext. header is encrypted. */ 2311 ext_hd |= EXT_ESP; 2312 break; 2313 2314 case IPPROTO_NONE: /* RFC 2460 */ 2315 PULLUP_TO(hlen, ulp, struct ip6_ext); 2316 /* Packet ends here. if ip6e_len!=0 octets 2317 * must be ignored. */ 2318 break; 2319 2320 case IPPROTO_OSPFIGP: 2321 /* XXX OSPF header check? */ 2322 PULLUP_TO(hlen, ulp, struct ip6_ext); 2323 break; 2324 2325 case IPPROTO_PIM: 2326 /* XXX PIM header check? */ 2327 PULLUP_TO(hlen, ulp, struct pim); 2328 break; 2329 2330 case IPPROTO_CARP: 2331 PULLUP_TO(hlen, ulp, struct carp_header); 2332 if (((struct carp_header *)ulp)->carp_version != 2333 CARP_VERSION) 2334 return (IP_FW_DENY); 2335 if (((struct carp_header *)ulp)->carp_type != 2336 CARP_ADVERTISEMENT) 2337 return (IP_FW_DENY); 2338 break; 2339 2340 case IPPROTO_IPV6: /* RFC 2893 */ 2341 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2342 break; 2343 2344 case IPPROTO_IPV4: /* RFC 2893 */ 2345 PULLUP_TO(hlen, ulp, struct ip); 2346 break; 2347 2348 default: 2349 printf("IPFW2: IPV6 - Unknown Extension " 2350 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2351 if (fw_deny_unknown_exthdrs) 2352 return (IP_FW_DENY); 2353 PULLUP_TO(hlen, ulp, struct ip6_ext); 2354 break; 2355 } /*switch */ 2356 } 2357 args->f_id.src_ip6 = mtod(m,struct ip6_hdr *)->ip6_src; 2358 args->f_id.dst_ip6 = mtod(m,struct ip6_hdr *)->ip6_dst; 2359 args->f_id.src_ip = 0; 2360 args->f_id.dst_ip = 0; 2361 args->f_id.flow_id6 = ntohl(mtod(m, struct ip6_hdr *)->ip6_flow); 2362 } else if (pktlen >= sizeof(struct ip) && 2363 (args->eh == NULL || ntohs(args->eh->ether_type) == ETHERTYPE_IP) && 2364 mtod(m, struct ip *)->ip_v == 4) { 2365 is_ipv4 = 1; 2366 ip = mtod(m, struct ip *); 2367 hlen = ip->ip_hl << 2; 2368 args->f_id.addr_type = 4; 2369 2370 /* 2371 * Collect parameters into local variables for faster matching. 2372 */ 2373 proto = ip->ip_p; 2374 src_ip = ip->ip_src; 2375 dst_ip = ip->ip_dst; 2376 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2377 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2378 ip_len = ntohs(ip->ip_len); 2379 } else { 2380 offset = ip->ip_off & IP_OFFMASK; 2381 ip_len = ip->ip_len; 2382 } 2383 pktlen = ip_len < pktlen ? ip_len : pktlen; 2384 2385 if (offset == 0) { 2386 switch (proto) { 2387 case IPPROTO_TCP: 2388 PULLUP_TO(hlen, ulp, struct tcphdr); 2389 dst_port = TCP(ulp)->th_dport; 2390 src_port = TCP(ulp)->th_sport; 2391 args->f_id.flags = TCP(ulp)->th_flags; 2392 break; 2393 2394 case IPPROTO_UDP: 2395 PULLUP_TO(hlen, ulp, struct udphdr); 2396 dst_port = UDP(ulp)->uh_dport; 2397 src_port = UDP(ulp)->uh_sport; 2398 break; 2399 2400 case IPPROTO_ICMP: 2401 PULLUP_TO(hlen, ulp, struct icmphdr); 2402 args->f_id.flags = ICMP(ulp)->icmp_type; 2403 break; 2404 2405 default: 2406 break; 2407 } 2408 } 2409 2410 args->f_id.src_ip = ntohl(src_ip.s_addr); 2411 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2412 } 2413#undef PULLUP_TO 2414 if (proto) { /* we may have port numbers, store them */ 2415 args->f_id.proto = proto; 2416 args->f_id.src_port = src_port = ntohs(src_port); 2417 args->f_id.dst_port = dst_port = ntohs(dst_port); 2418 } 2419 2420 IPFW_RLOCK(chain); 2421 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2422 if (args->rule) { 2423 /* 2424 * Packet has already been tagged. Look for the next rule 2425 * to restart processing. 2426 * 2427 * If fw_one_pass != 0 then just accept it. 2428 * XXX should not happen here, but optimized out in 2429 * the caller. 2430 */ 2431 if (fw_one_pass) { 2432 IPFW_RUNLOCK(chain); 2433 return (IP_FW_PASS); 2434 } 2435 2436 f = args->rule->next_rule; 2437 if (f == NULL) 2438 f = lookup_next_rule(args->rule); 2439 } else { 2440 /* 2441 * Find the starting rule. It can be either the first 2442 * one, or the one after divert_rule if asked so. 2443 */ 2444 int skipto = mtag ? divert_cookie(mtag) : 0; 2445 2446 f = chain->rules; 2447 if (args->eh == NULL && skipto != 0) { 2448 if (skipto >= IPFW_DEFAULT_RULE) { 2449 IPFW_RUNLOCK(chain); 2450 return (IP_FW_DENY); /* invalid */ 2451 } 2452 while (f && f->rulenum <= skipto) 2453 f = f->next; 2454 if (f == NULL) { /* drop packet */ 2455 IPFW_RUNLOCK(chain); 2456 return (IP_FW_DENY); 2457 } 2458 } 2459 } 2460 /* reset divert rule to avoid confusion later */ 2461 if (mtag) { 2462 divinput_flags = divert_info(mtag) & 2463 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2464 m_tag_delete(m, mtag); 2465 } 2466 2467 /* 2468 * Now scan the rules, and parse microinstructions for each rule. 2469 */ 2470 for (; f; f = f->next) { 2471 ipfw_insn *cmd; 2472 uint32_t tablearg = 0; 2473 int l, cmdlen, skip_or; /* skip rest of OR block */ 2474 2475again: 2476 if (set_disable & (1 << f->set) ) 2477 continue; 2478 2479 skip_or = 0; 2480 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2481 l -= cmdlen, cmd += cmdlen) { 2482 int match; 2483 2484 /* 2485 * check_body is a jump target used when we find a 2486 * CHECK_STATE, and need to jump to the body of 2487 * the target rule. 2488 */ 2489 2490check_body: 2491 cmdlen = F_LEN(cmd); 2492 /* 2493 * An OR block (insn_1 || .. || insn_n) has the 2494 * F_OR bit set in all but the last instruction. 2495 * The first match will set "skip_or", and cause 2496 * the following instructions to be skipped until 2497 * past the one with the F_OR bit clear. 2498 */ 2499 if (skip_or) { /* skip this instruction */ 2500 if ((cmd->len & F_OR) == 0) 2501 skip_or = 0; /* next one is good */ 2502 continue; 2503 } 2504 match = 0; /* set to 1 if we succeed */ 2505 2506 switch (cmd->opcode) { 2507 /* 2508 * The first set of opcodes compares the packet's 2509 * fields with some pattern, setting 'match' if a 2510 * match is found. At the end of the loop there is 2511 * logic to deal with F_NOT and F_OR flags associated 2512 * with the opcode. 2513 */ 2514 case O_NOP: 2515 match = 1; 2516 break; 2517 2518 case O_FORWARD_MAC: 2519 printf("ipfw: opcode %d unimplemented\n", 2520 cmd->opcode); 2521 break; 2522 2523 case O_GID: 2524 case O_UID: 2525 case O_JAIL: 2526 /* 2527 * We only check offset == 0 && proto != 0, 2528 * as this ensures that we have a 2529 * packet with the ports info. 2530 */ 2531 if (offset!=0) 2532 break; 2533 if (is_ipv6) /* XXX to be fixed later */ 2534 break; 2535 if (proto == IPPROTO_TCP || 2536 proto == IPPROTO_UDP) 2537 match = check_uidgid( 2538 (ipfw_insn_u32 *)cmd, 2539 proto, oif, 2540 dst_ip, dst_port, 2541 src_ip, src_port, &fw_ugid_cache, 2542 &ugid_lookup, args->inp); 2543 break; 2544 2545 case O_RECV: 2546 match = iface_match(m->m_pkthdr.rcvif, 2547 (ipfw_insn_if *)cmd); 2548 break; 2549 2550 case O_XMIT: 2551 match = iface_match(oif, (ipfw_insn_if *)cmd); 2552 break; 2553 2554 case O_VIA: 2555 match = iface_match(oif ? oif : 2556 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2557 break; 2558 2559 case O_MACADDR2: 2560 if (args->eh != NULL) { /* have MAC header */ 2561 u_int32_t *want = (u_int32_t *) 2562 ((ipfw_insn_mac *)cmd)->addr; 2563 u_int32_t *mask = (u_int32_t *) 2564 ((ipfw_insn_mac *)cmd)->mask; 2565 u_int32_t *hdr = (u_int32_t *)args->eh; 2566 2567 match = 2568 ( want[0] == (hdr[0] & mask[0]) && 2569 want[1] == (hdr[1] & mask[1]) && 2570 want[2] == (hdr[2] & mask[2]) ); 2571 } 2572 break; 2573 2574 case O_MAC_TYPE: 2575 if (args->eh != NULL) { 2576 u_int16_t t = 2577 ntohs(args->eh->ether_type); 2578 u_int16_t *p = 2579 ((ipfw_insn_u16 *)cmd)->ports; 2580 int i; 2581 2582 for (i = cmdlen - 1; !match && i>0; 2583 i--, p += 2) 2584 match = (t>=p[0] && t<=p[1]); 2585 } 2586 break; 2587 2588 case O_FRAG: 2589 match = (offset != 0); 2590 break; 2591 2592 case O_IN: /* "out" is "not in" */ 2593 match = (oif == NULL); 2594 break; 2595 2596 case O_LAYER2: 2597 match = (args->eh != NULL); 2598 break; 2599 2600 case O_DIVERTED: 2601 match = (cmd->arg1 & 1 && divinput_flags & 2602 IP_FW_DIVERT_LOOPBACK_FLAG) || 2603 (cmd->arg1 & 2 && divinput_flags & 2604 IP_FW_DIVERT_OUTPUT_FLAG); 2605 break; 2606 2607 case O_PROTO: 2608 /* 2609 * We do not allow an arg of 0 so the 2610 * check of "proto" only suffices. 2611 */ 2612 match = (proto == cmd->arg1); 2613 break; 2614 2615 case O_IP_SRC: 2616 match = is_ipv4 && 2617 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2618 src_ip.s_addr); 2619 break; 2620 2621 case O_IP_SRC_LOOKUP: 2622 case O_IP_DST_LOOKUP: 2623 if (is_ipv4) { 2624 uint32_t a = 2625 (cmd->opcode == O_IP_DST_LOOKUP) ? 2626 dst_ip.s_addr : src_ip.s_addr; 2627 uint32_t v; 2628 2629 match = lookup_table(chain, cmd->arg1, a, 2630 &v); 2631 if (!match) 2632 break; 2633 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2634 match = 2635 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2636 else 2637 tablearg = v; 2638 } 2639 break; 2640 2641 case O_IP_SRC_MASK: 2642 case O_IP_DST_MASK: 2643 if (is_ipv4) { 2644 uint32_t a = 2645 (cmd->opcode == O_IP_DST_MASK) ? 2646 dst_ip.s_addr : src_ip.s_addr; 2647 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2648 int i = cmdlen-1; 2649 2650 for (; !match && i>0; i-= 2, p+= 2) 2651 match = (p[0] == (a & p[1])); 2652 } 2653 break; 2654 2655 case O_IP_SRC_ME: 2656 if (is_ipv4) { 2657 struct ifnet *tif; 2658 2659 INADDR_TO_IFP(src_ip, tif); 2660 match = (tif != NULL); 2661 } 2662 break; 2663 2664 case O_IP_DST_SET: 2665 case O_IP_SRC_SET: 2666 if (is_ipv4) { 2667 u_int32_t *d = (u_int32_t *)(cmd+1); 2668 u_int32_t addr = 2669 cmd->opcode == O_IP_DST_SET ? 2670 args->f_id.dst_ip : 2671 args->f_id.src_ip; 2672 2673 if (addr < d[0]) 2674 break; 2675 addr -= d[0]; /* subtract base */ 2676 match = (addr < cmd->arg1) && 2677 ( d[ 1 + (addr>>5)] & 2678 (1<<(addr & 0x1f)) ); 2679 } 2680 break; 2681 2682 case O_IP_DST: 2683 match = is_ipv4 && 2684 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2685 dst_ip.s_addr); 2686 break; 2687 2688 case O_IP_DST_ME: 2689 if (is_ipv4) { 2690 struct ifnet *tif; 2691 2692 INADDR_TO_IFP(dst_ip, tif); 2693 match = (tif != NULL); 2694 } 2695 break; 2696 2697 case O_IP_SRCPORT: 2698 case O_IP_DSTPORT: 2699 /* 2700 * offset == 0 && proto != 0 is enough 2701 * to guarantee that we have a 2702 * packet with port info. 2703 */ 2704 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2705 && offset == 0) { 2706 u_int16_t x = 2707 (cmd->opcode == O_IP_SRCPORT) ? 2708 src_port : dst_port ; 2709 u_int16_t *p = 2710 ((ipfw_insn_u16 *)cmd)->ports; 2711 int i; 2712 2713 for (i = cmdlen - 1; !match && i>0; 2714 i--, p += 2) 2715 match = (x>=p[0] && x<=p[1]); 2716 } 2717 break; 2718 2719 case O_ICMPTYPE: 2720 match = (offset == 0 && proto==IPPROTO_ICMP && 2721 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2722 break; 2723 2724#ifdef INET6 2725 case O_ICMP6TYPE: 2726 match = is_ipv6 && offset == 0 && 2727 proto==IPPROTO_ICMPV6 && 2728 icmp6type_match( 2729 ICMP6(ulp)->icmp6_type, 2730 (ipfw_insn_u32 *)cmd); 2731 break; 2732#endif /* INET6 */ 2733 2734 case O_IPOPT: 2735 match = (is_ipv4 && 2736 ipopts_match(mtod(m, struct ip *), cmd) ); 2737 break; 2738 2739 case O_IPVER: 2740 match = (is_ipv4 && 2741 cmd->arg1 == mtod(m, struct ip *)->ip_v); 2742 break; 2743 2744 case O_IPID: 2745 case O_IPLEN: 2746 case O_IPTTL: 2747 if (is_ipv4) { /* only for IP packets */ 2748 uint16_t x; 2749 uint16_t *p; 2750 int i; 2751 2752 if (cmd->opcode == O_IPLEN) 2753 x = ip_len; 2754 else if (cmd->opcode == O_IPTTL) 2755 x = mtod(m, struct ip *)->ip_ttl; 2756 else /* must be IPID */ 2757 x = ntohs(mtod(m, struct ip *)->ip_id); 2758 if (cmdlen == 1) { 2759 match = (cmd->arg1 == x); 2760 break; 2761 } 2762 /* otherwise we have ranges */ 2763 p = ((ipfw_insn_u16 *)cmd)->ports; 2764 i = cmdlen - 1; 2765 for (; !match && i>0; i--, p += 2) 2766 match = (x >= p[0] && x <= p[1]); 2767 } 2768 break; 2769 2770 case O_IPPRECEDENCE: 2771 match = (is_ipv4 && 2772 (cmd->arg1 == (mtod(m, struct ip *)->ip_tos & 0xe0)) ); 2773 break; 2774 2775 case O_IPTOS: 2776 match = (is_ipv4 && 2777 flags_match(cmd, mtod(m, struct ip *)->ip_tos)); 2778 break; 2779 2780 case O_TCPDATALEN: 2781 if (proto == IPPROTO_TCP && offset == 0) { 2782 struct tcphdr *tcp; 2783 uint16_t x; 2784 uint16_t *p; 2785 int i; 2786 2787 tcp = TCP(ulp); 2788 x = ip_len - 2789 ((ip->ip_hl + tcp->th_off) << 2); 2790 if (cmdlen == 1) { 2791 match = (cmd->arg1 == x); 2792 break; 2793 } 2794 /* otherwise we have ranges */ 2795 p = ((ipfw_insn_u16 *)cmd)->ports; 2796 i = cmdlen - 1; 2797 for (; !match && i>0; i--, p += 2) 2798 match = (x >= p[0] && x <= p[1]); 2799 } 2800 break; 2801 2802 case O_TCPFLAGS: 2803 match = (proto == IPPROTO_TCP && offset == 0 && 2804 flags_match(cmd, TCP(ulp)->th_flags)); 2805 break; 2806 2807 case O_TCPOPTS: 2808 match = (proto == IPPROTO_TCP && offset == 0 && 2809 tcpopts_match(TCP(ulp), cmd)); 2810 break; 2811 2812 case O_TCPSEQ: 2813 match = (proto == IPPROTO_TCP && offset == 0 && 2814 ((ipfw_insn_u32 *)cmd)->d[0] == 2815 TCP(ulp)->th_seq); 2816 break; 2817 2818 case O_TCPACK: 2819 match = (proto == IPPROTO_TCP && offset == 0 && 2820 ((ipfw_insn_u32 *)cmd)->d[0] == 2821 TCP(ulp)->th_ack); 2822 break; 2823 2824 case O_TCPWIN: 2825 match = (proto == IPPROTO_TCP && offset == 0 && 2826 cmd->arg1 == TCP(ulp)->th_win); 2827 break; 2828 2829 case O_ESTAB: 2830 /* reject packets which have SYN only */ 2831 /* XXX should i also check for TH_ACK ? */ 2832 match = (proto == IPPROTO_TCP && offset == 0 && 2833 (TCP(ulp)->th_flags & 2834 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2835 break; 2836 2837 case O_ALTQ: { 2838 struct altq_tag *at; 2839 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2840 2841 match = 1; 2842 mtag = m_tag_find(m, PACKET_TAG_PF_QID, NULL); 2843 if (mtag != NULL) 2844 break; 2845 mtag = m_tag_get(PACKET_TAG_PF_QID, 2846 sizeof(struct altq_tag), 2847 M_NOWAIT); 2848 if (mtag == NULL) { 2849 /* 2850 * Let the packet fall back to the 2851 * default ALTQ. 2852 */ 2853 break; 2854 } 2855 at = (struct altq_tag *)(mtag+1); 2856 at->qid = altq->qid; 2857 if (is_ipv4) 2858 at->af = AF_INET; 2859 else 2860 at->af = AF_LINK; 2861 at->hdr = ip; 2862 m_tag_prepend(m, mtag); 2863 break; 2864 } 2865 2866 case O_LOG: 2867 if (fw_verbose) 2868 ipfw_log(f, hlen, args, m, 2869 oif, offset, tablearg); 2870 match = 1; 2871 break; 2872 2873 case O_PROB: 2874 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2875 break; 2876 2877 case O_VERREVPATH: 2878 /* Outgoing packets automatically pass/match */ 2879 match = ((oif != NULL) || 2880 (m->m_pkthdr.rcvif == NULL) || 2881 ( 2882#ifdef INET6 2883 is_ipv6 ? 2884 verify_path6(&(args->f_id.src_ip6), 2885 m->m_pkthdr.rcvif) : 2886#endif 2887 verify_path(src_ip, m->m_pkthdr.rcvif))); 2888 break; 2889 2890 case O_VERSRCREACH: 2891 /* Outgoing packets automatically pass/match */ 2892 match = (hlen > 0 && ((oif != NULL) || 2893#ifdef INET6 2894 is_ipv6 ? 2895 verify_path6(&(args->f_id.src_ip6), 2896 NULL) : 2897#endif 2898 verify_path(src_ip, NULL))); 2899 break; 2900 2901 case O_ANTISPOOF: 2902 /* Outgoing packets automatically pass/match */ 2903 if (oif == NULL && hlen > 0 && 2904 ( (is_ipv4 && in_localaddr(src_ip)) 2905#ifdef INET6 2906 || (is_ipv6 && 2907 in6_localaddr(&(args->f_id.src_ip6))) 2908#endif 2909 )) 2910 match = 2911#ifdef INET6 2912 is_ipv6 ? verify_path6( 2913 &(args->f_id.src_ip6), 2914 m->m_pkthdr.rcvif) : 2915#endif 2916 verify_path(src_ip, 2917 m->m_pkthdr.rcvif); 2918 else 2919 match = 1; 2920 break; 2921 2922 case O_IPSEC: 2923#ifdef FAST_IPSEC 2924 match = (m_tag_find(m, 2925 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2926#endif 2927#ifdef IPSEC 2928 match = (ipsec_getnhist(m) != 0); 2929#endif 2930 /* otherwise no match */ 2931 break; 2932 2933#ifdef INET6 2934 case O_IP6_SRC: 2935 match = is_ipv6 && 2936 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2937 &((ipfw_insn_ip6 *)cmd)->addr6); 2938 break; 2939 2940 case O_IP6_DST: 2941 match = is_ipv6 && 2942 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2943 &((ipfw_insn_ip6 *)cmd)->addr6); 2944 break; 2945 case O_IP6_SRC_MASK: 2946 case O_IP6_DST_MASK: 2947 if (is_ipv6) { 2948 int i = cmdlen - 1; 2949 struct in6_addr p; 2950 struct in6_addr *d = 2951 &((ipfw_insn_ip6 *)cmd)->addr6; 2952 2953 for (; !match && i > 0; d += 2, 2954 i -= F_INSN_SIZE(struct in6_addr) 2955 * 2) { 2956 p = (cmd->opcode == 2957 O_IP6_SRC_MASK) ? 2958 args->f_id.src_ip6: 2959 args->f_id.dst_ip6; 2960 APPLY_MASK(&p, &d[1]); 2961 match = 2962 IN6_ARE_ADDR_EQUAL(&d[0], 2963 &p); 2964 } 2965 } 2966 break; 2967 2968 case O_IP6_SRC_ME: 2969 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 2970 break; 2971 2972 case O_IP6_DST_ME: 2973 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 2974 break; 2975 2976 case O_FLOW6ID: 2977 match = is_ipv6 && 2978 flow6id_match(args->f_id.flow_id6, 2979 (ipfw_insn_u32 *) cmd); 2980 break; 2981 2982 case O_EXT_HDR: 2983 match = is_ipv6 && 2984 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2985 break; 2986 2987 case O_IP6: 2988 match = is_ipv6; 2989 break; 2990#endif 2991 2992 case O_IP4: 2993 match = is_ipv4; 2994 break; 2995 2996 case O_TAG: { 2997 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 2998 tablearg : cmd->arg1; 2999 3000 /* Packet is already tagged with this tag? */ 3001 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 3002 3003 /* We have `untag' action when F_NOT flag is 3004 * present. And we must remove this mtag from 3005 * mbuf and reset `match' to zero (`match' will 3006 * be inversed later). 3007 * Otherwise we should allocate new mtag and 3008 * push it into mbuf. 3009 */ 3010 if (cmd->len & F_NOT) { /* `untag' action */ 3011 if (mtag != NULL) 3012 m_tag_delete(m, mtag); 3013 } else if (mtag == NULL) { 3014 if ((mtag = m_tag_alloc(MTAG_IPFW, 3015 tag, 0, M_NOWAIT)) != NULL) 3016 m_tag_prepend(m, mtag); 3017 } 3018 match = (cmd->len & F_NOT) ? 0: 1; 3019 break; 3020 } 3021 3022 case O_TAGGED: { 3023 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3024 tablearg : cmd->arg1; 3025 3026 if (cmdlen == 1) { 3027 match = m_tag_locate(m, MTAG_IPFW, 3028 tag, NULL) != NULL; 3029 break; 3030 } 3031 3032 /* we have ranges */ 3033 for (mtag = m_tag_first(m); 3034 mtag != NULL && !match; 3035 mtag = m_tag_next(m, mtag)) { 3036 uint16_t *p; 3037 int i; 3038 3039 if (mtag->m_tag_cookie != MTAG_IPFW) 3040 continue; 3041 3042 p = ((ipfw_insn_u16 *)cmd)->ports; 3043 i = cmdlen - 1; 3044 for(; !match && i > 0; i--, p += 2) 3045 match = 3046 mtag->m_tag_id >= p[0] && 3047 mtag->m_tag_id <= p[1]; 3048 } 3049 break; 3050 } 3051 3052 /* 3053 * The second set of opcodes represents 'actions', 3054 * i.e. the terminal part of a rule once the packet 3055 * matches all previous patterns. 3056 * Typically there is only one action for each rule, 3057 * and the opcode is stored at the end of the rule 3058 * (but there are exceptions -- see below). 3059 * 3060 * In general, here we set retval and terminate the 3061 * outer loop (would be a 'break 3' in some language, 3062 * but we need to do a 'goto done'). 3063 * 3064 * Exceptions: 3065 * O_COUNT and O_SKIPTO actions: 3066 * instead of terminating, we jump to the next rule 3067 * ('goto next_rule', equivalent to a 'break 2'), 3068 * or to the SKIPTO target ('goto again' after 3069 * having set f, cmd and l), respectively. 3070 * 3071 * O_TAG, O_LOG and O_ALTQ action parameters: 3072 * perform some action and set match = 1; 3073 * 3074 * O_LIMIT and O_KEEP_STATE: these opcodes are 3075 * not real 'actions', and are stored right 3076 * before the 'action' part of the rule. 3077 * These opcodes try to install an entry in the 3078 * state tables; if successful, we continue with 3079 * the next opcode (match=1; break;), otherwise 3080 * the packet * must be dropped 3081 * ('goto done' after setting retval); 3082 * 3083 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3084 * cause a lookup of the state table, and a jump 3085 * to the 'action' part of the parent rule 3086 * ('goto check_body') if an entry is found, or 3087 * (CHECK_STATE only) a jump to the next rule if 3088 * the entry is not found ('goto next_rule'). 3089 * The result of the lookup is cached to make 3090 * further instances of these opcodes are 3091 * effectively NOPs. 3092 */ 3093 case O_LIMIT: 3094 case O_KEEP_STATE: 3095 if (install_state(f, 3096 (ipfw_insn_limit *)cmd, args, tablearg)) { 3097 retval = IP_FW_DENY; 3098 goto done; /* error/limit violation */ 3099 } 3100 match = 1; 3101 break; 3102 3103 case O_PROBE_STATE: 3104 case O_CHECK_STATE: 3105 /* 3106 * dynamic rules are checked at the first 3107 * keep-state or check-state occurrence, 3108 * with the result being stored in dyn_dir. 3109 * The compiler introduces a PROBE_STATE 3110 * instruction for us when we have a 3111 * KEEP_STATE (because PROBE_STATE needs 3112 * to be run first). 3113 */ 3114 if (dyn_dir == MATCH_UNKNOWN && 3115 (q = lookup_dyn_rule(&args->f_id, 3116 &dyn_dir, proto == IPPROTO_TCP ? 3117 TCP(ulp) : NULL)) 3118 != NULL) { 3119 /* 3120 * Found dynamic entry, update stats 3121 * and jump to the 'action' part of 3122 * the parent rule. 3123 */ 3124 q->pcnt++; 3125 q->bcnt += pktlen; 3126 f = q->rule; 3127 cmd = ACTION_PTR(f); 3128 l = f->cmd_len - f->act_ofs; 3129 IPFW_DYN_UNLOCK(); 3130 goto check_body; 3131 } 3132 /* 3133 * Dynamic entry not found. If CHECK_STATE, 3134 * skip to next rule, if PROBE_STATE just 3135 * ignore and continue with next opcode. 3136 */ 3137 if (cmd->opcode == O_CHECK_STATE) 3138 goto next_rule; 3139 match = 1; 3140 break; 3141 3142 case O_ACCEPT: 3143 retval = 0; /* accept */ 3144 goto done; 3145 3146 case O_PIPE: 3147 case O_QUEUE: 3148 args->rule = f; /* report matching rule */ 3149 if (cmd->arg1 == IP_FW_TABLEARG) 3150 args->cookie = tablearg; 3151 else 3152 args->cookie = cmd->arg1; 3153 retval = IP_FW_DUMMYNET; 3154 goto done; 3155 3156 case O_DIVERT: 3157 case O_TEE: { 3158 struct divert_tag *dt; 3159 3160 if (args->eh) /* not on layer 2 */ 3161 break; 3162 mtag = m_tag_get(PACKET_TAG_DIVERT, 3163 sizeof(struct divert_tag), 3164 M_NOWAIT); 3165 if (mtag == NULL) { 3166 /* XXX statistic */ 3167 /* drop packet */ 3168 IPFW_RUNLOCK(chain); 3169 return (IP_FW_DENY); 3170 } 3171 dt = (struct divert_tag *)(mtag+1); 3172 dt->cookie = f->rulenum; 3173 if (cmd->arg1 == IP_FW_TABLEARG) 3174 dt->info = tablearg; 3175 else 3176 dt->info = cmd->arg1; 3177 m_tag_prepend(m, mtag); 3178 retval = (cmd->opcode == O_DIVERT) ? 3179 IP_FW_DIVERT : IP_FW_TEE; 3180 goto done; 3181 } 3182 3183 case O_COUNT: 3184 case O_SKIPTO: 3185 f->pcnt++; /* update stats */ 3186 f->bcnt += pktlen; 3187 f->timestamp = time_uptime; 3188 if (cmd->opcode == O_COUNT) 3189 goto next_rule; 3190 /* handle skipto */ 3191 if (f->next_rule == NULL) 3192 lookup_next_rule(f); 3193 f = f->next_rule; 3194 goto again; 3195 3196 case O_REJECT: 3197 /* 3198 * Drop the packet and send a reject notice 3199 * if the packet is not ICMP (or is an ICMP 3200 * query), and it is not multicast/broadcast. 3201 */ 3202 if (hlen > 0 && is_ipv4 && offset == 0 && 3203 (proto != IPPROTO_ICMP || 3204 is_icmp_query(ICMP(ulp))) && 3205 !(m->m_flags & (M_BCAST|M_MCAST)) && 3206 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3207 send_reject(args, cmd->arg1, ip_len); 3208 m = args->m; 3209 } 3210 /* FALLTHROUGH */ 3211#ifdef INET6 3212 case O_UNREACH6: 3213 if (hlen > 0 && is_ipv6 && 3214 ((offset & IP6F_OFF_MASK) == 0) && 3215 (proto != IPPROTO_ICMPV6 || 3216 (is_icmp6_query(args->f_id.flags) == 1)) && 3217 !(m->m_flags & (M_BCAST|M_MCAST)) && 3218 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3219 send_reject6(args, cmd->arg1, hlen); 3220 m = args->m; 3221 } 3222 /* FALLTHROUGH */ 3223#endif 3224 case O_DENY: 3225 retval = IP_FW_DENY; 3226 goto done; 3227 3228 case O_FORWARD_IP: { 3229 struct sockaddr_in *sa; 3230 sa = &(((ipfw_insn_sa *)cmd)->sa); 3231 if (args->eh) /* not valid on layer2 pkts */ 3232 break; 3233 if (!q || dyn_dir == MATCH_FORWARD) { 3234 if (sa->sin_addr.s_addr == INADDR_ANY) { 3235 bcopy(sa, &args->hopstore, 3236 sizeof(*sa)); 3237 args->hopstore.sin_addr.s_addr = 3238 htonl(tablearg); 3239 args->next_hop = 3240 &args->hopstore; 3241 } else { 3242 args->next_hop = sa; 3243 } 3244 } 3245 retval = IP_FW_PASS; 3246 } 3247 goto done; 3248 3249 case O_NETGRAPH: 3250 case O_NGTEE: 3251 args->rule = f; /* report matching rule */ 3252 if (cmd->arg1 == IP_FW_TABLEARG) 3253 args->cookie = tablearg; 3254 else 3255 args->cookie = cmd->arg1; 3256 retval = (cmd->opcode == O_NETGRAPH) ? 3257 IP_FW_NETGRAPH : IP_FW_NGTEE; 3258 goto done; 3259 3260 default: 3261 panic("-- unknown opcode %d\n", cmd->opcode); 3262 } /* end of switch() on opcodes */ 3263 3264 if (cmd->len & F_NOT) 3265 match = !match; 3266 3267 if (match) { 3268 if (cmd->len & F_OR) 3269 skip_or = 1; 3270 } else { 3271 if (!(cmd->len & F_OR)) /* not an OR block, */ 3272 break; /* try next rule */ 3273 } 3274 3275 } /* end of inner for, scan opcodes */ 3276 3277next_rule:; /* try next rule */ 3278 3279 } /* end of outer for, scan rules */ 3280 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3281 IPFW_RUNLOCK(chain); 3282 return (IP_FW_DENY); 3283 3284done: 3285 /* Update statistics */ 3286 f->pcnt++; 3287 f->bcnt += pktlen; 3288 f->timestamp = time_uptime; 3289 IPFW_RUNLOCK(chain); 3290 return (retval); 3291 3292pullup_failed: 3293 if (fw_verbose) 3294 printf("ipfw: pullup failed\n"); 3295 return (IP_FW_DENY); 3296} 3297 3298/* 3299 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3300 * These will be reconstructed on the fly as packets are matched. 3301 */ 3302static void 3303flush_rule_ptrs(struct ip_fw_chain *chain) 3304{ 3305 struct ip_fw *rule; 3306 3307 IPFW_WLOCK_ASSERT(chain); 3308 3309 for (rule = chain->rules; rule; rule = rule->next) 3310 rule->next_rule = NULL; 3311} 3312 3313/* 3314 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3315 * possibly create a rule number and add the rule to the list. 3316 * Update the rule_number in the input struct so the caller knows it as well. 3317 */ 3318static int 3319add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3320{ 3321 struct ip_fw *rule, *f, *prev; 3322 int l = RULESIZE(input_rule); 3323 3324 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3325 return (EINVAL); 3326 3327 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3328 if (rule == NULL) 3329 return (ENOSPC); 3330 3331 bcopy(input_rule, rule, l); 3332 3333 rule->next = NULL; 3334 rule->next_rule = NULL; 3335 3336 rule->pcnt = 0; 3337 rule->bcnt = 0; 3338 rule->timestamp = 0; 3339 3340 IPFW_WLOCK(chain); 3341 3342 if (chain->rules == NULL) { /* default rule */ 3343 chain->rules = rule; 3344 goto done; 3345 } 3346 3347 /* 3348 * If rulenum is 0, find highest numbered rule before the 3349 * default rule, and add autoinc_step 3350 */ 3351 if (autoinc_step < 1) 3352 autoinc_step = 1; 3353 else if (autoinc_step > 1000) 3354 autoinc_step = 1000; 3355 if (rule->rulenum == 0) { 3356 /* 3357 * locate the highest numbered rule before default 3358 */ 3359 for (f = chain->rules; f; f = f->next) { 3360 if (f->rulenum == IPFW_DEFAULT_RULE) 3361 break; 3362 rule->rulenum = f->rulenum; 3363 } 3364 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 3365 rule->rulenum += autoinc_step; 3366 input_rule->rulenum = rule->rulenum; 3367 } 3368 3369 /* 3370 * Now insert the new rule in the right place in the sorted list. 3371 */ 3372 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3373 if (f->rulenum > rule->rulenum) { /* found the location */ 3374 if (prev) { 3375 rule->next = f; 3376 prev->next = rule; 3377 } else { /* head insert */ 3378 rule->next = chain->rules; 3379 chain->rules = rule; 3380 } 3381 break; 3382 } 3383 } 3384 flush_rule_ptrs(chain); 3385done: 3386 static_count++; 3387 static_len += l; 3388 IPFW_WUNLOCK(chain); 3389 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3390 rule->rulenum, static_count);) 3391 return (0); 3392} 3393 3394/** 3395 * Remove a static rule (including derived * dynamic rules) 3396 * and place it on the ``reap list'' for later reclamation. 3397 * The caller is in charge of clearing rule pointers to avoid 3398 * dangling pointers. 3399 * @return a pointer to the next entry. 3400 * Arguments are not checked, so they better be correct. 3401 */ 3402static struct ip_fw * 3403remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 3404{ 3405 struct ip_fw *n; 3406 int l = RULESIZE(rule); 3407 3408 IPFW_WLOCK_ASSERT(chain); 3409 3410 n = rule->next; 3411 IPFW_DYN_LOCK(); 3412 remove_dyn_rule(rule, NULL /* force removal */); 3413 IPFW_DYN_UNLOCK(); 3414 if (prev == NULL) 3415 chain->rules = n; 3416 else 3417 prev->next = n; 3418 static_count--; 3419 static_len -= l; 3420 3421 rule->next = chain->reap; 3422 chain->reap = rule; 3423 3424 return n; 3425} 3426 3427/** 3428 * Reclaim storage associated with a list of rules. This is 3429 * typically the list created using remove_rule. 3430 */ 3431static void 3432reap_rules(struct ip_fw *head) 3433{ 3434 struct ip_fw *rule; 3435 3436 while ((rule = head) != NULL) { 3437 head = head->next; 3438 if (DUMMYNET_LOADED) 3439 ip_dn_ruledel_ptr(rule); 3440 free(rule, M_IPFW); 3441 } 3442} 3443 3444/* 3445 * Remove all rules from a chain (except rules in set RESVD_SET 3446 * unless kill_default = 1). The caller is responsible for 3447 * reclaiming storage for the rules left in chain->reap. 3448 */ 3449static void 3450free_chain(struct ip_fw_chain *chain, int kill_default) 3451{ 3452 struct ip_fw *prev, *rule; 3453 3454 IPFW_WLOCK_ASSERT(chain); 3455 3456 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3457 for (prev = NULL, rule = chain->rules; rule ; ) 3458 if (kill_default || rule->set != RESVD_SET) 3459 rule = remove_rule(chain, rule, prev); 3460 else { 3461 prev = rule; 3462 rule = rule->next; 3463 } 3464} 3465 3466/** 3467 * Remove all rules with given number, and also do set manipulation. 3468 * Assumes chain != NULL && *chain != NULL. 3469 * 3470 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3471 * the next 8 bits are the new set, the top 8 bits are the command: 3472 * 3473 * 0 delete rules with given number 3474 * 1 delete rules with given set number 3475 * 2 move rules with given number to new set 3476 * 3 move rules with given set number to new set 3477 * 4 swap sets with given numbers 3478 */ 3479static int 3480del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3481{ 3482 struct ip_fw *prev = NULL, *rule; 3483 u_int16_t rulenum; /* rule or old_set */ 3484 u_int8_t cmd, new_set; 3485 3486 rulenum = arg & 0xffff; 3487 cmd = (arg >> 24) & 0xff; 3488 new_set = (arg >> 16) & 0xff; 3489 3490 if (cmd > 4) 3491 return EINVAL; 3492 if (new_set > RESVD_SET) 3493 return EINVAL; 3494 if (cmd == 0 || cmd == 2) { 3495 if (rulenum >= IPFW_DEFAULT_RULE) 3496 return EINVAL; 3497 } else { 3498 if (rulenum > RESVD_SET) /* old_set */ 3499 return EINVAL; 3500 } 3501 3502 IPFW_WLOCK(chain); 3503 rule = chain->rules; 3504 chain->reap = NULL; 3505 switch (cmd) { 3506 case 0: /* delete rules with given number */ 3507 /* 3508 * locate first rule to delete 3509 */ 3510 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3511 ; 3512 if (rule->rulenum != rulenum) { 3513 IPFW_WUNLOCK(chain); 3514 return EINVAL; 3515 } 3516 3517 /* 3518 * flush pointers outside the loop, then delete all matching 3519 * rules. prev remains the same throughout the cycle. 3520 */ 3521 flush_rule_ptrs(chain); 3522 while (rule->rulenum == rulenum) 3523 rule = remove_rule(chain, rule, prev); 3524 break; 3525 3526 case 1: /* delete all rules with given set number */ 3527 flush_rule_ptrs(chain); 3528 rule = chain->rules; 3529 while (rule->rulenum < IPFW_DEFAULT_RULE) 3530 if (rule->set == rulenum) 3531 rule = remove_rule(chain, rule, prev); 3532 else { 3533 prev = rule; 3534 rule = rule->next; 3535 } 3536 break; 3537 3538 case 2: /* move rules with given number to new set */ 3539 rule = chain->rules; 3540 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3541 if (rule->rulenum == rulenum) 3542 rule->set = new_set; 3543 break; 3544 3545 case 3: /* move rules with given set number to new set */ 3546 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3547 if (rule->set == rulenum) 3548 rule->set = new_set; 3549 break; 3550 3551 case 4: /* swap two sets */ 3552 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3553 if (rule->set == rulenum) 3554 rule->set = new_set; 3555 else if (rule->set == new_set) 3556 rule->set = rulenum; 3557 break; 3558 } 3559 /* 3560 * Look for rules to reclaim. We grab the list before 3561 * releasing the lock then reclaim them w/o the lock to 3562 * avoid a LOR with dummynet. 3563 */ 3564 rule = chain->reap; 3565 chain->reap = NULL; 3566 IPFW_WUNLOCK(chain); 3567 if (rule) 3568 reap_rules(rule); 3569 return 0; 3570} 3571 3572/* 3573 * Clear counters for a specific rule. 3574 * The enclosing "table" is assumed locked. 3575 */ 3576static void 3577clear_counters(struct ip_fw *rule, int log_only) 3578{ 3579 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3580 3581 if (log_only == 0) { 3582 rule->bcnt = rule->pcnt = 0; 3583 rule->timestamp = 0; 3584 } 3585 if (l->o.opcode == O_LOG) 3586 l->log_left = l->max_log; 3587} 3588 3589/** 3590 * Reset some or all counters on firewall rules. 3591 * @arg frwl is null to clear all entries, or contains a specific 3592 * rule number. 3593 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 3594 */ 3595static int 3596zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 3597{ 3598 struct ip_fw *rule; 3599 char *msg; 3600 3601 IPFW_WLOCK(chain); 3602 if (rulenum == 0) { 3603 norule_counter = 0; 3604 for (rule = chain->rules; rule; rule = rule->next) 3605 clear_counters(rule, log_only); 3606 msg = log_only ? "ipfw: All logging counts reset.\n" : 3607 "ipfw: Accounting cleared.\n"; 3608 } else { 3609 int cleared = 0; 3610 /* 3611 * We can have multiple rules with the same number, so we 3612 * need to clear them all. 3613 */ 3614 for (rule = chain->rules; rule; rule = rule->next) 3615 if (rule->rulenum == rulenum) { 3616 while (rule && rule->rulenum == rulenum) { 3617 clear_counters(rule, log_only); 3618 rule = rule->next; 3619 } 3620 cleared = 1; 3621 break; 3622 } 3623 if (!cleared) { /* we did not find any matching rules */ 3624 IPFW_WUNLOCK(chain); 3625 return (EINVAL); 3626 } 3627 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 3628 "ipfw: Entry %d cleared.\n"; 3629 } 3630 IPFW_WUNLOCK(chain); 3631 3632 if (fw_verbose) 3633 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 3634 return (0); 3635} 3636 3637/* 3638 * Check validity of the structure before insert. 3639 * Fortunately rules are simple, so this mostly need to check rule sizes. 3640 */ 3641static int 3642check_ipfw_struct(struct ip_fw *rule, int size) 3643{ 3644 int l, cmdlen = 0; 3645 int have_action=0; 3646 ipfw_insn *cmd; 3647 3648 if (size < sizeof(*rule)) { 3649 printf("ipfw: rule too short\n"); 3650 return (EINVAL); 3651 } 3652 /* first, check for valid size */ 3653 l = RULESIZE(rule); 3654 if (l != size) { 3655 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3656 return (EINVAL); 3657 } 3658 if (rule->act_ofs >= rule->cmd_len) { 3659 printf("ipfw: bogus action offset (%u > %u)\n", 3660 rule->act_ofs, rule->cmd_len - 1); 3661 return (EINVAL); 3662 } 3663 /* 3664 * Now go for the individual checks. Very simple ones, basically only 3665 * instruction sizes. 3666 */ 3667 for (l = rule->cmd_len, cmd = rule->cmd ; 3668 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3669 cmdlen = F_LEN(cmd); 3670 if (cmdlen > l) { 3671 printf("ipfw: opcode %d size truncated\n", 3672 cmd->opcode); 3673 return EINVAL; 3674 } 3675 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3676 switch (cmd->opcode) { 3677 case O_PROBE_STATE: 3678 case O_KEEP_STATE: 3679 case O_PROTO: 3680 case O_IP_SRC_ME: 3681 case O_IP_DST_ME: 3682 case O_LAYER2: 3683 case O_IN: 3684 case O_FRAG: 3685 case O_DIVERTED: 3686 case O_IPOPT: 3687 case O_IPTOS: 3688 case O_IPPRECEDENCE: 3689 case O_IPVER: 3690 case O_TCPWIN: 3691 case O_TCPFLAGS: 3692 case O_TCPOPTS: 3693 case O_ESTAB: 3694 case O_VERREVPATH: 3695 case O_VERSRCREACH: 3696 case O_ANTISPOOF: 3697 case O_IPSEC: 3698#ifdef INET6 3699 case O_IP6_SRC_ME: 3700 case O_IP6_DST_ME: 3701 case O_EXT_HDR: 3702 case O_IP6: 3703#endif 3704 case O_IP4: 3705 case O_TAG: 3706 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3707 goto bad_size; 3708 break; 3709 3710 case O_UID: 3711 case O_GID: 3712 case O_JAIL: 3713 case O_IP_SRC: 3714 case O_IP_DST: 3715 case O_TCPSEQ: 3716 case O_TCPACK: 3717 case O_PROB: 3718 case O_ICMPTYPE: 3719 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3720 goto bad_size; 3721 break; 3722 3723 case O_LIMIT: 3724 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3725 goto bad_size; 3726 break; 3727 3728 case O_LOG: 3729 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3730 goto bad_size; 3731 3732 ((ipfw_insn_log *)cmd)->log_left = 3733 ((ipfw_insn_log *)cmd)->max_log; 3734 3735 break; 3736 3737 case O_IP_SRC_MASK: 3738 case O_IP_DST_MASK: 3739 /* only odd command lengths */ 3740 if ( !(cmdlen & 1) || cmdlen > 31) 3741 goto bad_size; 3742 break; 3743 3744 case O_IP_SRC_SET: 3745 case O_IP_DST_SET: 3746 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3747 printf("ipfw: invalid set size %d\n", 3748 cmd->arg1); 3749 return EINVAL; 3750 } 3751 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3752 (cmd->arg1+31)/32 ) 3753 goto bad_size; 3754 break; 3755 3756 case O_IP_SRC_LOOKUP: 3757 case O_IP_DST_LOOKUP: 3758 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3759 printf("ipfw: invalid table number %d\n", 3760 cmd->arg1); 3761 return (EINVAL); 3762 } 3763 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3764 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3765 goto bad_size; 3766 break; 3767 3768 case O_MACADDR2: 3769 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3770 goto bad_size; 3771 break; 3772 3773 case O_NOP: 3774 case O_IPID: 3775 case O_IPTTL: 3776 case O_IPLEN: 3777 case O_TCPDATALEN: 3778 case O_TAGGED: 3779 if (cmdlen < 1 || cmdlen > 31) 3780 goto bad_size; 3781 break; 3782 3783 case O_MAC_TYPE: 3784 case O_IP_SRCPORT: 3785 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3786 if (cmdlen < 2 || cmdlen > 31) 3787 goto bad_size; 3788 break; 3789 3790 case O_RECV: 3791 case O_XMIT: 3792 case O_VIA: 3793 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3794 goto bad_size; 3795 break; 3796 3797 case O_ALTQ: 3798 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3799 goto bad_size; 3800 break; 3801 3802 case O_PIPE: 3803 case O_QUEUE: 3804 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3805 goto bad_size; 3806 goto check_action; 3807 3808 case O_FORWARD_IP: 3809#ifdef IPFIREWALL_FORWARD 3810 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3811 goto bad_size; 3812 goto check_action; 3813#else 3814 return EINVAL; 3815#endif 3816 3817 case O_DIVERT: 3818 case O_TEE: 3819 if (ip_divert_ptr == NULL) 3820 return EINVAL; 3821 else 3822 goto check_size; 3823 case O_NETGRAPH: 3824 case O_NGTEE: 3825 if (!NG_IPFW_LOADED) 3826 return EINVAL; 3827 else 3828 goto check_size; 3829 case O_FORWARD_MAC: /* XXX not implemented yet */ 3830 case O_CHECK_STATE: 3831 case O_COUNT: 3832 case O_ACCEPT: 3833 case O_DENY: 3834 case O_REJECT: 3835#ifdef INET6 3836 case O_UNREACH6: 3837#endif 3838 case O_SKIPTO: 3839check_size: 3840 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3841 goto bad_size; 3842check_action: 3843 if (have_action) { 3844 printf("ipfw: opcode %d, multiple actions" 3845 " not allowed\n", 3846 cmd->opcode); 3847 return EINVAL; 3848 } 3849 have_action = 1; 3850 if (l != cmdlen) { 3851 printf("ipfw: opcode %d, action must be" 3852 " last opcode\n", 3853 cmd->opcode); 3854 return EINVAL; 3855 } 3856 break; 3857#ifdef INET6 3858 case O_IP6_SRC: 3859 case O_IP6_DST: 3860 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 3861 F_INSN_SIZE(ipfw_insn)) 3862 goto bad_size; 3863 break; 3864 3865 case O_FLOW6ID: 3866 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3867 ((ipfw_insn_u32 *)cmd)->o.arg1) 3868 goto bad_size; 3869 break; 3870 3871 case O_IP6_SRC_MASK: 3872 case O_IP6_DST_MASK: 3873 if ( !(cmdlen & 1) || cmdlen > 127) 3874 goto bad_size; 3875 break; 3876 case O_ICMP6TYPE: 3877 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 3878 goto bad_size; 3879 break; 3880#endif 3881 3882 default: 3883 switch (cmd->opcode) { 3884#ifndef INET6 3885 case O_IP6_SRC_ME: 3886 case O_IP6_DST_ME: 3887 case O_EXT_HDR: 3888 case O_IP6: 3889 case O_UNREACH6: 3890 case O_IP6_SRC: 3891 case O_IP6_DST: 3892 case O_FLOW6ID: 3893 case O_IP6_SRC_MASK: 3894 case O_IP6_DST_MASK: 3895 case O_ICMP6TYPE: 3896 printf("ipfw: no IPv6 support in kernel\n"); 3897 return EPROTONOSUPPORT; 3898#endif 3899 default: 3900 printf("ipfw: opcode %d, unknown opcode\n", 3901 cmd->opcode); 3902 return EINVAL; 3903 } 3904 } 3905 } 3906 if (have_action == 0) { 3907 printf("ipfw: missing action\n"); 3908 return EINVAL; 3909 } 3910 return 0; 3911 3912bad_size: 3913 printf("ipfw: opcode %d size %d wrong\n", 3914 cmd->opcode, cmdlen); 3915 return EINVAL; 3916} 3917 3918/* 3919 * Copy the static and dynamic rules to the supplied buffer 3920 * and return the amount of space actually used. 3921 */ 3922static size_t 3923ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 3924{ 3925 char *bp = buf; 3926 char *ep = bp + space; 3927 struct ip_fw *rule; 3928 int i; 3929 3930 /* XXX this can take a long time and locking will block packet flow */ 3931 IPFW_RLOCK(chain); 3932 for (rule = chain->rules; rule ; rule = rule->next) { 3933 /* 3934 * Verify the entry fits in the buffer in case the 3935 * rules changed between calculating buffer space and 3936 * now. This would be better done using a generation 3937 * number but should suffice for now. 3938 */ 3939 i = RULESIZE(rule); 3940 if (bp + i <= ep) { 3941 bcopy(rule, bp, i); 3942 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 3943 sizeof(set_disable)); 3944 bp += i; 3945 } 3946 } 3947 IPFW_RUNLOCK(chain); 3948 if (ipfw_dyn_v) { 3949 ipfw_dyn_rule *p, *last = NULL; 3950 3951 IPFW_DYN_LOCK(); 3952 for (i = 0 ; i < curr_dyn_buckets; i++) 3953 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 3954 if (bp + sizeof *p <= ep) { 3955 ipfw_dyn_rule *dst = 3956 (ipfw_dyn_rule *)bp; 3957 bcopy(p, dst, sizeof *p); 3958 bcopy(&(p->rule->rulenum), &(dst->rule), 3959 sizeof(p->rule->rulenum)); 3960 /* 3961 * store a non-null value in "next". 3962 * The userland code will interpret a 3963 * NULL here as a marker 3964 * for the last dynamic rule. 3965 */ 3966 bcopy(&dst, &dst->next, sizeof(dst)); 3967 last = dst; 3968 dst->expire = 3969 TIME_LEQ(dst->expire, time_uptime) ? 3970 0 : dst->expire - time_uptime ; 3971 bp += sizeof(ipfw_dyn_rule); 3972 } 3973 } 3974 IPFW_DYN_UNLOCK(); 3975 if (last != NULL) /* mark last dynamic rule */ 3976 bzero(&last->next, sizeof(last)); 3977 } 3978 return (bp - (char *)buf); 3979} 3980 3981 3982/** 3983 * {set|get}sockopt parser. 3984 */ 3985static int 3986ipfw_ctl(struct sockopt *sopt) 3987{ 3988#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3989 int error, rule_num; 3990 size_t size; 3991 struct ip_fw *buf, *rule; 3992 u_int32_t rulenum[2]; 3993 3994 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW); 3995 if (error) 3996 return (error); 3997 3998 /* 3999 * Disallow modifications in really-really secure mode, but still allow 4000 * the logging counters to be reset. 4001 */ 4002 if (sopt->sopt_name == IP_FW_ADD || 4003 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 4004 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 4005 if (error) 4006 return (error); 4007 } 4008 4009 error = 0; 4010 4011 switch (sopt->sopt_name) { 4012 case IP_FW_GET: 4013 /* 4014 * pass up a copy of the current rules. Static rules 4015 * come first (the last of which has number IPFW_DEFAULT_RULE), 4016 * followed by a possibly empty list of dynamic rule. 4017 * The last dynamic rule has NULL in the "next" field. 4018 * 4019 * Note that the calculated size is used to bound the 4020 * amount of data returned to the user. The rule set may 4021 * change between calculating the size and returning the 4022 * data in which case we'll just return what fits. 4023 */ 4024 size = static_len; /* size of static rules */ 4025 if (ipfw_dyn_v) /* add size of dyn.rules */ 4026 size += (dyn_count * sizeof(ipfw_dyn_rule)); 4027 4028 /* 4029 * XXX todo: if the user passes a short length just to know 4030 * how much room is needed, do not bother filling up the 4031 * buffer, just jump to the sooptcopyout. 4032 */ 4033 buf = malloc(size, M_TEMP, M_WAITOK); 4034 error = sooptcopyout(sopt, buf, 4035 ipfw_getrules(&layer3_chain, buf, size)); 4036 free(buf, M_TEMP); 4037 break; 4038 4039 case IP_FW_FLUSH: 4040 /* 4041 * Normally we cannot release the lock on each iteration. 4042 * We could do it here only because we start from the head all 4043 * the times so there is no risk of missing some entries. 4044 * On the other hand, the risk is that we end up with 4045 * a very inconsistent ruleset, so better keep the lock 4046 * around the whole cycle. 4047 * 4048 * XXX this code can be improved by resetting the head of 4049 * the list to point to the default rule, and then freeing 4050 * the old list without the need for a lock. 4051 */ 4052 4053 IPFW_WLOCK(&layer3_chain); 4054 layer3_chain.reap = NULL; 4055 free_chain(&layer3_chain, 0 /* keep default rule */); 4056 rule = layer3_chain.reap; 4057 layer3_chain.reap = NULL; 4058 IPFW_WUNLOCK(&layer3_chain); 4059 if (rule != NULL) 4060 reap_rules(rule); 4061 break; 4062 4063 case IP_FW_ADD: 4064 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 4065 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 4066 sizeof(struct ip_fw) ); 4067 if (error == 0) 4068 error = check_ipfw_struct(rule, sopt->sopt_valsize); 4069 if (error == 0) { 4070 error = add_rule(&layer3_chain, rule); 4071 size = RULESIZE(rule); 4072 if (!error && sopt->sopt_dir == SOPT_GET) 4073 error = sooptcopyout(sopt, rule, size); 4074 } 4075 free(rule, M_TEMP); 4076 break; 4077 4078 case IP_FW_DEL: 4079 /* 4080 * IP_FW_DEL is used for deleting single rules or sets, 4081 * and (ab)used to atomically manipulate sets. Argument size 4082 * is used to distinguish between the two: 4083 * sizeof(u_int32_t) 4084 * delete single rule or set of rules, 4085 * or reassign rules (or sets) to a different set. 4086 * 2*sizeof(u_int32_t) 4087 * atomic disable/enable sets. 4088 * first u_int32_t contains sets to be disabled, 4089 * second u_int32_t contains sets to be enabled. 4090 */ 4091 error = sooptcopyin(sopt, rulenum, 4092 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4093 if (error) 4094 break; 4095 size = sopt->sopt_valsize; 4096 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4097 error = del_entry(&layer3_chain, rulenum[0]); 4098 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4099 set_disable = 4100 (set_disable | rulenum[0]) & ~rulenum[1] & 4101 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4102 else 4103 error = EINVAL; 4104 break; 4105 4106 case IP_FW_ZERO: 4107 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 4108 rule_num = 0; 4109 if (sopt->sopt_val != 0) { 4110 error = sooptcopyin(sopt, &rule_num, 4111 sizeof(int), sizeof(int)); 4112 if (error) 4113 break; 4114 } 4115 error = zero_entry(&layer3_chain, rule_num, 4116 sopt->sopt_name == IP_FW_RESETLOG); 4117 break; 4118 4119 case IP_FW_TABLE_ADD: 4120 { 4121 ipfw_table_entry ent; 4122 4123 error = sooptcopyin(sopt, &ent, 4124 sizeof(ent), sizeof(ent)); 4125 if (error) 4126 break; 4127 error = add_table_entry(&layer3_chain, ent.tbl, 4128 ent.addr, ent.masklen, ent.value); 4129 } 4130 break; 4131 4132 case IP_FW_TABLE_DEL: 4133 { 4134 ipfw_table_entry ent; 4135 4136 error = sooptcopyin(sopt, &ent, 4137 sizeof(ent), sizeof(ent)); 4138 if (error) 4139 break; 4140 error = del_table_entry(&layer3_chain, ent.tbl, 4141 ent.addr, ent.masklen); 4142 } 4143 break; 4144 4145 case IP_FW_TABLE_FLUSH: 4146 { 4147 u_int16_t tbl; 4148 4149 error = sooptcopyin(sopt, &tbl, 4150 sizeof(tbl), sizeof(tbl)); 4151 if (error) 4152 break; 4153 IPFW_WLOCK(&layer3_chain); 4154 error = flush_table(&layer3_chain, tbl); 4155 IPFW_WUNLOCK(&layer3_chain); 4156 } 4157 break; 4158 4159 case IP_FW_TABLE_GETSIZE: 4160 { 4161 u_int32_t tbl, cnt; 4162 4163 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4164 sizeof(tbl)))) 4165 break; 4166 IPFW_RLOCK(&layer3_chain); 4167 error = count_table(&layer3_chain, tbl, &cnt); 4168 IPFW_RUNLOCK(&layer3_chain); 4169 if (error) 4170 break; 4171 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4172 } 4173 break; 4174 4175 case IP_FW_TABLE_LIST: 4176 { 4177 ipfw_table *tbl; 4178 4179 if (sopt->sopt_valsize < sizeof(*tbl)) { 4180 error = EINVAL; 4181 break; 4182 } 4183 size = sopt->sopt_valsize; 4184 tbl = malloc(size, M_TEMP, M_WAITOK); 4185 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4186 if (error) { 4187 free(tbl, M_TEMP); 4188 break; 4189 } 4190 tbl->size = (size - sizeof(*tbl)) / 4191 sizeof(ipfw_table_entry); 4192 IPFW_RLOCK(&layer3_chain); 4193 error = dump_table(&layer3_chain, tbl); 4194 IPFW_RUNLOCK(&layer3_chain); 4195 if (error) { 4196 free(tbl, M_TEMP); 4197 break; 4198 } 4199 error = sooptcopyout(sopt, tbl, size); 4200 free(tbl, M_TEMP); 4201 } 4202 break; 4203 4204 default: 4205 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4206 error = EINVAL; 4207 } 4208 4209 return (error); 4210#undef RULE_MAXSIZE 4211} 4212 4213/** 4214 * dummynet needs a reference to the default rule, because rules can be 4215 * deleted while packets hold a reference to them. When this happens, 4216 * dummynet changes the reference to the default rule (it could well be a 4217 * NULL pointer, but this way we do not need to check for the special 4218 * case, plus here he have info on the default behaviour). 4219 */ 4220struct ip_fw *ip_fw_default_rule; 4221 4222/* 4223 * This procedure is only used to handle keepalives. It is invoked 4224 * every dyn_keepalive_period 4225 */ 4226static void 4227ipfw_tick(void * __unused unused) 4228{ 4229 struct mbuf *m0, *m, *mnext, **mtailp; 4230 int i; 4231 ipfw_dyn_rule *q; 4232 4233 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 4234 goto done; 4235 4236 /* 4237 * We make a chain of packets to go out here -- not deferring 4238 * until after we drop the IPFW dynamic rule lock would result 4239 * in a lock order reversal with the normal packet input -> ipfw 4240 * call stack. 4241 */ 4242 m0 = NULL; 4243 mtailp = &m0; 4244 IPFW_DYN_LOCK(); 4245 for (i = 0 ; i < curr_dyn_buckets ; i++) { 4246 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 4247 if (q->dyn_type == O_LIMIT_PARENT) 4248 continue; 4249 if (q->id.proto != IPPROTO_TCP) 4250 continue; 4251 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4252 continue; 4253 if (TIME_LEQ( time_uptime+dyn_keepalive_interval, 4254 q->expire)) 4255 continue; /* too early */ 4256 if (TIME_LEQ(q->expire, time_uptime)) 4257 continue; /* too late, rule expired */ 4258 4259 *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1, 4260 q->ack_fwd, TH_SYN); 4261 if (*mtailp != NULL) 4262 mtailp = &(*mtailp)->m_nextpkt; 4263 *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1, 4264 q->ack_rev, 0); 4265 if (*mtailp != NULL) 4266 mtailp = &(*mtailp)->m_nextpkt; 4267 } 4268 } 4269 IPFW_DYN_UNLOCK(); 4270 for (m = mnext = m0; m != NULL; m = mnext) { 4271 mnext = m->m_nextpkt; 4272 m->m_nextpkt = NULL; 4273 ip_output(m, NULL, NULL, 0, NULL, NULL); 4274 } 4275done: 4276 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 4277} 4278 4279int 4280ipfw_init(void) 4281{ 4282 struct ip_fw default_rule; 4283 int error; 4284 4285#ifdef INET6 4286 /* Setup IPv6 fw sysctl tree. */ 4287 sysctl_ctx_init(&ip6_fw_sysctl_ctx); 4288 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx, 4289 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw", 4290 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall"); 4291 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4292 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, 4293 &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6"); 4294 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree), 4295 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE, 4296 &fw_deny_unknown_exthdrs, 0, 4297 "Deny packets with unknown IPv6 Extension Headers"); 4298#endif 4299 4300 layer3_chain.rules = NULL; 4301 IPFW_LOCK_INIT(&layer3_chain); 4302 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule zone", 4303 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4304 UMA_ALIGN_PTR, 0); 4305 IPFW_DYN_LOCK_INIT(); 4306 callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE); 4307 4308 bzero(&default_rule, sizeof default_rule); 4309 4310 default_rule.act_ofs = 0; 4311 default_rule.rulenum = IPFW_DEFAULT_RULE; 4312 default_rule.cmd_len = 1; 4313 default_rule.set = RESVD_SET; 4314 4315 default_rule.cmd[0].len = 1; 4316 default_rule.cmd[0].opcode = 4317#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 4318 1 ? O_ACCEPT : 4319#endif 4320 O_DENY; 4321 4322 error = add_rule(&layer3_chain, &default_rule); 4323 if (error != 0) { 4324 printf("ipfw2: error %u initializing default rule " 4325 "(support disabled)\n", error); 4326 IPFW_DYN_LOCK_DESTROY(); 4327 IPFW_LOCK_DESTROY(&layer3_chain); 4328 uma_zdestroy(ipfw_dyn_rule_zone); 4329 return (error); 4330 } 4331 4332 ip_fw_default_rule = layer3_chain.rules; 4333 printf("ipfw2 " 4334#ifdef INET6 4335 "(+ipv6) " 4336#endif 4337 "initialized, divert %s, " 4338 "rule-based forwarding " 4339#ifdef IPFIREWALL_FORWARD 4340 "enabled, " 4341#else 4342 "disabled, " 4343#endif 4344 "default to %s, logging ", 4345#ifdef IPDIVERT 4346 "enabled", 4347#else 4348 "loadable", 4349#endif 4350 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 4351 4352#ifdef IPFIREWALL_VERBOSE 4353 fw_verbose = 1; 4354#endif 4355#ifdef IPFIREWALL_VERBOSE_LIMIT 4356 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4357#endif 4358 if (fw_verbose == 0) 4359 printf("disabled\n"); 4360 else if (verbose_limit == 0) 4361 printf("unlimited\n"); 4362 else 4363 printf("limited to %d packets/entry by default\n", 4364 verbose_limit); 4365 4366 error = init_tables(&layer3_chain); 4367 if (error) { 4368 IPFW_DYN_LOCK_DESTROY(); 4369 IPFW_LOCK_DESTROY(&layer3_chain); 4370 uma_zdestroy(ipfw_dyn_rule_zone); 4371 return (error); 4372 } 4373 ip_fw_ctl_ptr = ipfw_ctl; 4374 ip_fw_chk_ptr = ipfw_chk; 4375 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 4376 4377 return (0); 4378} 4379 4380void 4381ipfw_destroy(void) 4382{ 4383 struct ip_fw *reap; 4384 4385 ip_fw_chk_ptr = NULL; 4386 ip_fw_ctl_ptr = NULL; 4387 callout_drain(&ipfw_timeout); 4388 IPFW_WLOCK(&layer3_chain); 4389 flush_tables(&layer3_chain); 4390 layer3_chain.reap = NULL; 4391 free_chain(&layer3_chain, 1 /* kill default rule */); 4392 reap = layer3_chain.reap, layer3_chain.reap = NULL; 4393 IPFW_WUNLOCK(&layer3_chain); 4394 if (reap != NULL) 4395 reap_rules(reap); 4396 IPFW_DYN_LOCK_DESTROY(); 4397 uma_zdestroy(ipfw_dyn_rule_zone); 4398 IPFW_LOCK_DESTROY(&layer3_chain); 4399 4400#ifdef INET6 4401 /* Free IPv6 fw sysctl tree. */ 4402 sysctl_ctx_free(&ip6_fw_sysctl_ctx); 4403#endif 4404 4405 printf("IP firewall unloaded\n"); 4406} 4407