ip_fw2.c revision 225030
1146297Sobrien/*- 2240373Sdelphij * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa 3146297Sobrien * 4146297Sobrien * Redistribution and use in source and binary forms, with or without 5146297Sobrien * modification, are permitted provided that the following conditions 6146297Sobrien * are met: 7146297Sobrien * 1. Redistributions of source code must retain the above copyright 8146297Sobrien * notice, this list of conditions and the following disclaimer. 9146297Sobrien * 2. Redistributions in binary form must reproduce the above copyright 10146297Sobrien * notice, this list of conditions and the following disclaimer in the 11146297Sobrien * documentation and/or other materials provided with the distribution. 12146297Sobrien * 13146297Sobrien * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 14146297Sobrien * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 15146297Sobrien * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 16240373Sdelphij * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 17146297Sobrien * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 18146297Sobrien * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 19146297Sobrien * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 20240373Sdelphij * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 21146297Sobrien * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 22146297Sobrien * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 23146297Sobrien * SUCH DAMAGE. 24146297Sobrien */ 25146297Sobrien 26146297Sobrien#include <sys/cdefs.h> 27__FBSDID("$FreeBSD: head/sys/netinet/ipfw/ip_fw2.c 225030 2011-08-20 12:40:17Z bz $"); 28 29/* 30 * The FreeBSD IP packet firewall, main file 31 */ 32 33#if !defined(KLD_MODULE) 34#include "opt_ipfw.h" 35#include "opt_ipdivert.h" 36#include "opt_ipdn.h" 37#include "opt_inet.h" 38#ifndef INET 39#error IPFIREWALL requires INET. 40#endif /* INET */ 41#endif 42#include "opt_inet6.h" 43#include "opt_ipsec.h" 44 45#include <sys/param.h> 46#include <sys/systm.h> 47#include <sys/condvar.h> 48#include <sys/eventhandler.h> 49#include <sys/malloc.h> 50#include <sys/mbuf.h> 51#include <sys/kernel.h> 52#include <sys/lock.h> 53#include <sys/jail.h> 54#include <sys/module.h> 55#include <sys/priv.h> 56#include <sys/proc.h> 57#include <sys/rwlock.h> 58#include <sys/socket.h> 59#include <sys/socketvar.h> 60#include <sys/sysctl.h> 61#include <sys/syslog.h> 62#include <sys/ucred.h> 63#include <net/ethernet.h> /* for ETHERTYPE_IP */ 64#include <net/if.h> 65#include <net/route.h> 66#include <net/pf_mtag.h> 67#include <net/vnet.h> 68 69#include <netinet/in.h> 70#include <netinet/in_var.h> 71#include <netinet/in_pcb.h> 72#include <netinet/ip.h> 73#include <netinet/ip_var.h> 74#include <netinet/ip_icmp.h> 75#include <netinet/ip_fw.h> 76#include <netinet/ipfw/ip_fw_private.h> 77#include <netinet/ip_carp.h> 78#include <netinet/pim.h> 79#include <netinet/tcp_var.h> 80#include <netinet/udp.h> 81#include <netinet/udp_var.h> 82#include <netinet/sctp.h> 83 84#include <netinet/ip6.h> 85#include <netinet/icmp6.h> 86#ifdef INET6 87#include <netinet6/in6_pcb.h> 88#include <netinet6/scope6_var.h> 89#include <netinet6/ip6_var.h> 90#endif 91 92#include <machine/in_cksum.h> /* XXX for in_cksum */ 93 94#ifdef MAC 95#include <security/mac/mac_framework.h> 96#endif 97 98/* 99 * static variables followed by global ones. 100 * All ipfw global variables are here. 101 */ 102 103/* ipfw_vnet_ready controls when we are open for business */ 104static VNET_DEFINE(int, ipfw_vnet_ready) = 0; 105#define V_ipfw_vnet_ready VNET(ipfw_vnet_ready) 106 107static VNET_DEFINE(int, fw_deny_unknown_exthdrs); 108#define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs) 109 110static VNET_DEFINE(int, fw_permit_single_frag6) = 1; 111#define V_fw_permit_single_frag6 VNET(fw_permit_single_frag6) 112 113#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 114static int default_to_accept = 1; 115#else 116static int default_to_accept; 117#endif 118 119VNET_DEFINE(int, autoinc_step); 120VNET_DEFINE(int, fw_one_pass) = 1; 121 122/* 123 * Each rule belongs to one of 32 different sets (0..31). 124 * The variable set_disable contains one bit per set. 125 * If the bit is set, all rules in the corresponding set 126 * are disabled. Set RESVD_SET(31) is reserved for the default rule 127 * and rules that are not deleted by the flush command, 128 * and CANNOT be disabled. 129 * Rules in set RESVD_SET can only be deleted individually. 130 */ 131VNET_DEFINE(u_int32_t, set_disable); 132#define V_set_disable VNET(set_disable) 133 134VNET_DEFINE(int, fw_verbose); 135/* counter for ipfw_log(NULL...) */ 136VNET_DEFINE(u_int64_t, norule_counter); 137VNET_DEFINE(int, verbose_limit); 138 139/* layer3_chain contains the list of rules for layer 3 */ 140VNET_DEFINE(struct ip_fw_chain, layer3_chain); 141 142ipfw_nat_t *ipfw_nat_ptr = NULL; 143struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int); 144ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; 145ipfw_nat_cfg_t *ipfw_nat_del_ptr; 146ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; 147ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; 148 149#ifdef SYSCTL_NODE 150uint32_t dummy_def = IPFW_DEFAULT_RULE; 151uint32_t dummy_tables_max = IPFW_TABLES_MAX; 152 153SYSBEGIN(f3) 154 155SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 156SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 157 CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0, 158 "Only do a single pass through ipfw when using dummynet(4)"); 159SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, 160 CTLFLAG_RW, &VNET_NAME(autoinc_step), 0, 161 "Rule number auto-increment step"); 162SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose, 163 CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0, 164 "Log matches to ipfw rules"); 165SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, 166 CTLFLAG_RW, &VNET_NAME(verbose_limit), 0, 167 "Set upper limit of matches of ipfw rules logged"); 168SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, 169 &dummy_def, 0, 170 "The default/max possible rule number."); 171SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, tables_max, CTLFLAG_RD, 172 &dummy_tables_max, 0, 173 "The maximum number of tables."); 174SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN, 175 &default_to_accept, 0, 176 "Make the default rule accept all packets."); 177TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept); 178SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, static_count, 179 CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0, 180 "Number of static rules"); 181 182#ifdef INET6 183SYSCTL_DECL(_net_inet6_ip6); 184SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 185SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs, 186 CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0, 187 "Deny packets with unknown IPv6 Extension Headers"); 188SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6, 189 CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_permit_single_frag6), 0, 190 "Permit single packet IPv6 fragments"); 191#endif /* INET6 */ 192 193SYSEND 194 195#endif /* SYSCTL_NODE */ 196 197 198/* 199 * Some macros used in the various matching options. 200 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 201 * Other macros just cast void * into the appropriate type 202 */ 203#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 204#define TCP(p) ((struct tcphdr *)(p)) 205#define SCTP(p) ((struct sctphdr *)(p)) 206#define UDP(p) ((struct udphdr *)(p)) 207#define ICMP(p) ((struct icmphdr *)(p)) 208#define ICMP6(p) ((struct icmp6_hdr *)(p)) 209 210static __inline int 211icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 212{ 213 int type = icmp->icmp_type; 214 215 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 216} 217 218#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 219 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 220 221static int 222is_icmp_query(struct icmphdr *icmp) 223{ 224 int type = icmp->icmp_type; 225 226 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 227} 228#undef TT 229 230/* 231 * The following checks use two arrays of 8 or 16 bits to store the 232 * bits that we want set or clear, respectively. They are in the 233 * low and high half of cmd->arg1 or cmd->d[0]. 234 * 235 * We scan options and store the bits we find set. We succeed if 236 * 237 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 238 * 239 * The code is sometimes optimized not to store additional variables. 240 */ 241 242static int 243flags_match(ipfw_insn *cmd, u_int8_t bits) 244{ 245 u_char want_clear; 246 bits = ~bits; 247 248 if ( ((cmd->arg1 & 0xff) & bits) != 0) 249 return 0; /* some bits we want set were clear */ 250 want_clear = (cmd->arg1 >> 8) & 0xff; 251 if ( (want_clear & bits) != want_clear) 252 return 0; /* some bits we want clear were set */ 253 return 1; 254} 255 256static int 257ipopts_match(struct ip *ip, ipfw_insn *cmd) 258{ 259 int optlen, bits = 0; 260 u_char *cp = (u_char *)(ip + 1); 261 int x = (ip->ip_hl << 2) - sizeof (struct ip); 262 263 for (; x > 0; x -= optlen, cp += optlen) { 264 int opt = cp[IPOPT_OPTVAL]; 265 266 if (opt == IPOPT_EOL) 267 break; 268 if (opt == IPOPT_NOP) 269 optlen = 1; 270 else { 271 optlen = cp[IPOPT_OLEN]; 272 if (optlen <= 0 || optlen > x) 273 return 0; /* invalid or truncated */ 274 } 275 switch (opt) { 276 277 default: 278 break; 279 280 case IPOPT_LSRR: 281 bits |= IP_FW_IPOPT_LSRR; 282 break; 283 284 case IPOPT_SSRR: 285 bits |= IP_FW_IPOPT_SSRR; 286 break; 287 288 case IPOPT_RR: 289 bits |= IP_FW_IPOPT_RR; 290 break; 291 292 case IPOPT_TS: 293 bits |= IP_FW_IPOPT_TS; 294 break; 295 } 296 } 297 return (flags_match(cmd, bits)); 298} 299 300static int 301tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 302{ 303 int optlen, bits = 0; 304 u_char *cp = (u_char *)(tcp + 1); 305 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 306 307 for (; x > 0; x -= optlen, cp += optlen) { 308 int opt = cp[0]; 309 if (opt == TCPOPT_EOL) 310 break; 311 if (opt == TCPOPT_NOP) 312 optlen = 1; 313 else { 314 optlen = cp[1]; 315 if (optlen <= 0) 316 break; 317 } 318 319 switch (opt) { 320 321 default: 322 break; 323 324 case TCPOPT_MAXSEG: 325 bits |= IP_FW_TCPOPT_MSS; 326 break; 327 328 case TCPOPT_WINDOW: 329 bits |= IP_FW_TCPOPT_WINDOW; 330 break; 331 332 case TCPOPT_SACK_PERMITTED: 333 case TCPOPT_SACK: 334 bits |= IP_FW_TCPOPT_SACK; 335 break; 336 337 case TCPOPT_TIMESTAMP: 338 bits |= IP_FW_TCPOPT_TS; 339 break; 340 341 } 342 } 343 return (flags_match(cmd, bits)); 344} 345 346static int 347iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 348{ 349 if (ifp == NULL) /* no iface with this packet, match fails */ 350 return 0; 351 /* Check by name or by IP address */ 352 if (cmd->name[0] != '\0') { /* match by name */ 353 /* Check name */ 354 if (cmd->p.glob) { 355 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 356 return(1); 357 } else { 358 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 359 return(1); 360 } 361 } else { 362#ifdef __FreeBSD__ /* and OSX too ? */ 363 struct ifaddr *ia; 364 365 if_addr_rlock(ifp); 366 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 367 if (ia->ifa_addr->sa_family != AF_INET) 368 continue; 369 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 370 (ia->ifa_addr))->sin_addr.s_addr) { 371 if_addr_runlock(ifp); 372 return(1); /* match */ 373 } 374 } 375 if_addr_runlock(ifp); 376#endif /* __FreeBSD__ */ 377 } 378 return(0); /* no match, fail ... */ 379} 380 381/* 382 * The verify_path function checks if a route to the src exists and 383 * if it is reachable via ifp (when provided). 384 * 385 * The 'verrevpath' option checks that the interface that an IP packet 386 * arrives on is the same interface that traffic destined for the 387 * packet's source address would be routed out of. 388 * The 'versrcreach' option just checks that the source address is 389 * reachable via any route (except default) in the routing table. 390 * These two are a measure to block forged packets. This is also 391 * commonly known as "anti-spoofing" or Unicast Reverse Path 392 * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 393 * is purposely reminiscent of the Cisco IOS command, 394 * 395 * ip verify unicast reverse-path 396 * ip verify unicast source reachable-via any 397 * 398 * which implements the same functionality. But note that the syntax 399 * is misleading, and the check may be performed on all IP packets 400 * whether unicast, multicast, or broadcast. 401 */ 402static int 403verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) 404{ 405#ifndef __FreeBSD__ 406 return 0; 407#else 408 struct route ro; 409 struct sockaddr_in *dst; 410 411 bzero(&ro, sizeof(ro)); 412 413 dst = (struct sockaddr_in *)&(ro.ro_dst); 414 dst->sin_family = AF_INET; 415 dst->sin_len = sizeof(*dst); 416 dst->sin_addr = src; 417 in_rtalloc_ign(&ro, 0, fib); 418 419 if (ro.ro_rt == NULL) 420 return 0; 421 422 /* 423 * If ifp is provided, check for equality with rtentry. 424 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 425 * in order to pass packets injected back by if_simloop(): 426 * if useloopback == 1 routing entry (via lo0) for our own address 427 * may exist, so we need to handle routing assymetry. 428 */ 429 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 430 RTFREE(ro.ro_rt); 431 return 0; 432 } 433 434 /* if no ifp provided, check if rtentry is not default route */ 435 if (ifp == NULL && 436 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 437 RTFREE(ro.ro_rt); 438 return 0; 439 } 440 441 /* or if this is a blackhole/reject route */ 442 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 443 RTFREE(ro.ro_rt); 444 return 0; 445 } 446 447 /* found valid route */ 448 RTFREE(ro.ro_rt); 449 return 1; 450#endif /* __FreeBSD__ */ 451} 452 453#ifdef INET6 454/* 455 * ipv6 specific rules here... 456 */ 457static __inline int 458icmp6type_match (int type, ipfw_insn_u32 *cmd) 459{ 460 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 461} 462 463static int 464flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 465{ 466 int i; 467 for (i=0; i <= cmd->o.arg1; ++i ) 468 if (curr_flow == cmd->d[i] ) 469 return 1; 470 return 0; 471} 472 473/* support for IP6_*_ME opcodes */ 474static int 475search_ip6_addr_net (struct in6_addr * ip6_addr) 476{ 477 struct ifnet *mdc; 478 struct ifaddr *mdc2; 479 struct in6_ifaddr *fdm; 480 struct in6_addr copia; 481 482 TAILQ_FOREACH(mdc, &V_ifnet, if_link) { 483 if_addr_rlock(mdc); 484 TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) { 485 if (mdc2->ifa_addr->sa_family == AF_INET6) { 486 fdm = (struct in6_ifaddr *)mdc2; 487 copia = fdm->ia_addr.sin6_addr; 488 /* need for leaving scope_id in the sock_addr */ 489 in6_clearscope(&copia); 490 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) { 491 if_addr_runlock(mdc); 492 return 1; 493 } 494 } 495 } 496 if_addr_runlock(mdc); 497 } 498 return 0; 499} 500 501static int 502verify_path6(struct in6_addr *src, struct ifnet *ifp) 503{ 504 struct route_in6 ro; 505 struct sockaddr_in6 *dst; 506 507 bzero(&ro, sizeof(ro)); 508 509 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 510 dst->sin6_family = AF_INET6; 511 dst->sin6_len = sizeof(*dst); 512 dst->sin6_addr = *src; 513 /* XXX MRT 0 for ipv6 at this time */ 514 rtalloc_ign((struct route *)&ro, 0); 515 516 if (ro.ro_rt == NULL) 517 return 0; 518 519 /* 520 * if ifp is provided, check for equality with rtentry 521 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 522 * to support the case of sending packets to an address of our own. 523 * (where the former interface is the first argument of if_simloop() 524 * (=ifp), the latter is lo0) 525 */ 526 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 527 RTFREE(ro.ro_rt); 528 return 0; 529 } 530 531 /* if no ifp provided, check if rtentry is not default route */ 532 if (ifp == NULL && 533 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 534 RTFREE(ro.ro_rt); 535 return 0; 536 } 537 538 /* or if this is a blackhole/reject route */ 539 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 540 RTFREE(ro.ro_rt); 541 return 0; 542 } 543 544 /* found valid route */ 545 RTFREE(ro.ro_rt); 546 return 1; 547 548} 549 550static int 551is_icmp6_query(int icmp6_type) 552{ 553 if ((icmp6_type <= ICMP6_MAXTYPE) && 554 (icmp6_type == ICMP6_ECHO_REQUEST || 555 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 556 icmp6_type == ICMP6_WRUREQUEST || 557 icmp6_type == ICMP6_FQDN_QUERY || 558 icmp6_type == ICMP6_NI_QUERY)) 559 return (1); 560 561 return (0); 562} 563 564static void 565send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 566{ 567 struct mbuf *m; 568 569 m = args->m; 570 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 571 struct tcphdr *tcp; 572 tcp = (struct tcphdr *)((char *)ip6 + hlen); 573 574 if ((tcp->th_flags & TH_RST) == 0) { 575 struct mbuf *m0; 576 m0 = ipfw_send_pkt(args->m, &(args->f_id), 577 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 578 tcp->th_flags | TH_RST); 579 if (m0 != NULL) 580 ip6_output(m0, NULL, NULL, 0, NULL, NULL, 581 NULL); 582 } 583 FREE_PKT(m); 584 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 585#if 0 586 /* 587 * Unlike above, the mbufs need to line up with the ip6 hdr, 588 * as the contents are read. We need to m_adj() the 589 * needed amount. 590 * The mbuf will however be thrown away so we can adjust it. 591 * Remember we did an m_pullup on it already so we 592 * can make some assumptions about contiguousness. 593 */ 594 if (args->L3offset) 595 m_adj(m, args->L3offset); 596#endif 597 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 598 } else 599 FREE_PKT(m); 600 601 args->m = NULL; 602} 603 604#endif /* INET6 */ 605 606 607/* 608 * sends a reject message, consuming the mbuf passed as an argument. 609 */ 610static void 611send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip) 612{ 613 614#if 0 615 /* XXX When ip is not guaranteed to be at mtod() we will 616 * need to account for this */ 617 * The mbuf will however be thrown away so we can adjust it. 618 * Remember we did an m_pullup on it already so we 619 * can make some assumptions about contiguousness. 620 */ 621 if (args->L3offset) 622 m_adj(m, args->L3offset); 623#endif 624 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 625 /* We need the IP header in host order for icmp_error(). */ 626 SET_HOST_IPLEN(ip); 627 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 628 } else if (args->f_id.proto == IPPROTO_TCP) { 629 struct tcphdr *const tcp = 630 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 631 if ( (tcp->th_flags & TH_RST) == 0) { 632 struct mbuf *m; 633 m = ipfw_send_pkt(args->m, &(args->f_id), 634 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 635 tcp->th_flags | TH_RST); 636 if (m != NULL) 637 ip_output(m, NULL, NULL, 0, NULL, NULL); 638 } 639 FREE_PKT(args->m); 640 } else 641 FREE_PKT(args->m); 642 args->m = NULL; 643} 644 645/* 646 * Support for uid/gid/jail lookup. These tests are expensive 647 * (because we may need to look into the list of active sockets) 648 * so we cache the results. ugid_lookupp is 0 if we have not 649 * yet done a lookup, 1 if we succeeded, and -1 if we tried 650 * and failed. The function always returns the match value. 651 * We could actually spare the variable and use *uc, setting 652 * it to '(void *)check_uidgid if we have no info, NULL if 653 * we tried and failed, or any other value if successful. 654 */ 655static int 656check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp, 657 struct ucred **uc) 658{ 659#ifndef __FreeBSD__ 660 /* XXX */ 661 return cred_check(insn, proto, oif, 662 dst_ip, dst_port, src_ip, src_port, 663 (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb); 664#else /* FreeBSD */ 665 struct in_addr src_ip, dst_ip; 666 struct inpcbinfo *pi; 667 struct ipfw_flow_id *id; 668 struct inpcb *pcb, *inp; 669 struct ifnet *oif; 670 int lookupflags; 671 int match; 672 673 id = &args->f_id; 674 inp = args->inp; 675 oif = args->oif; 676 677 /* 678 * Check to see if the UDP or TCP stack supplied us with 679 * the PCB. If so, rather then holding a lock and looking 680 * up the PCB, we can use the one that was supplied. 681 */ 682 if (inp && *ugid_lookupp == 0) { 683 INP_LOCK_ASSERT(inp); 684 if (inp->inp_socket != NULL) { 685 *uc = crhold(inp->inp_cred); 686 *ugid_lookupp = 1; 687 } else 688 *ugid_lookupp = -1; 689 } 690 /* 691 * If we have already been here and the packet has no 692 * PCB entry associated with it, then we can safely 693 * assume that this is a no match. 694 */ 695 if (*ugid_lookupp == -1) 696 return (0); 697 if (id->proto == IPPROTO_TCP) { 698 lookupflags = 0; 699 pi = &V_tcbinfo; 700 } else if (id->proto == IPPROTO_UDP) { 701 lookupflags = INPLOOKUP_WILDCARD; 702 pi = &V_udbinfo; 703 } else 704 return 0; 705 lookupflags |= INPLOOKUP_RLOCKPCB; 706 match = 0; 707 if (*ugid_lookupp == 0) { 708 if (id->addr_type == 6) { 709#ifdef INET6 710 if (oif == NULL) 711 pcb = in6_pcblookup_mbuf(pi, 712 &id->src_ip6, htons(id->src_port), 713 &id->dst_ip6, htons(id->dst_port), 714 lookupflags, oif, args->m); 715 else 716 pcb = in6_pcblookup_mbuf(pi, 717 &id->dst_ip6, htons(id->dst_port), 718 &id->src_ip6, htons(id->src_port), 719 lookupflags, oif, args->m); 720#else 721 *ugid_lookupp = -1; 722 return (0); 723#endif 724 } else { 725 src_ip.s_addr = htonl(id->src_ip); 726 dst_ip.s_addr = htonl(id->dst_ip); 727 if (oif == NULL) 728 pcb = in_pcblookup_mbuf(pi, 729 src_ip, htons(id->src_port), 730 dst_ip, htons(id->dst_port), 731 lookupflags, oif, args->m); 732 else 733 pcb = in_pcblookup_mbuf(pi, 734 dst_ip, htons(id->dst_port), 735 src_ip, htons(id->src_port), 736 lookupflags, oif, args->m); 737 } 738 if (pcb != NULL) { 739 INP_RLOCK_ASSERT(pcb); 740 *uc = crhold(pcb->inp_cred); 741 *ugid_lookupp = 1; 742 INP_RUNLOCK(pcb); 743 } 744 if (*ugid_lookupp == 0) { 745 /* 746 * We tried and failed, set the variable to -1 747 * so we will not try again on this packet. 748 */ 749 *ugid_lookupp = -1; 750 return (0); 751 } 752 } 753 if (insn->o.opcode == O_UID) 754 match = ((*uc)->cr_uid == (uid_t)insn->d[0]); 755 else if (insn->o.opcode == O_GID) 756 match = groupmember((gid_t)insn->d[0], *uc); 757 else if (insn->o.opcode == O_JAIL) 758 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]); 759 return (match); 760#endif /* __FreeBSD__ */ 761} 762 763/* 764 * Helper function to set args with info on the rule after the matching 765 * one. slot is precise, whereas we guess rule_id as they are 766 * assigned sequentially. 767 */ 768static inline void 769set_match(struct ip_fw_args *args, int slot, 770 struct ip_fw_chain *chain) 771{ 772 args->rule.chain_id = chain->id; 773 args->rule.slot = slot + 1; /* we use 0 as a marker */ 774 args->rule.rule_id = 1 + chain->map[slot]->id; 775 args->rule.rulenum = chain->map[slot]->rulenum; 776} 777 778/* 779 * The main check routine for the firewall. 780 * 781 * All arguments are in args so we can modify them and return them 782 * back to the caller. 783 * 784 * Parameters: 785 * 786 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 787 * Starts with the IP header. 788 * args->eh (in) Mac header if present, NULL for layer3 packet. 789 * args->L3offset Number of bytes bypassed if we came from L2. 790 * e.g. often sizeof(eh) ** NOTYET ** 791 * args->oif Outgoing interface, NULL if packet is incoming. 792 * The incoming interface is in the mbuf. (in) 793 * args->divert_rule (in/out) 794 * Skip up to the first rule past this rule number; 795 * upon return, non-zero port number for divert or tee. 796 * 797 * args->rule Pointer to the last matching rule (in/out) 798 * args->next_hop Socket we are forwarding to (out). 799 * args->f_id Addresses grabbed from the packet (out) 800 * args->rule.info a cookie depending on rule action 801 * 802 * Return value: 803 * 804 * IP_FW_PASS the packet must be accepted 805 * IP_FW_DENY the packet must be dropped 806 * IP_FW_DIVERT divert packet, port in m_tag 807 * IP_FW_TEE tee packet, port in m_tag 808 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 809 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 810 * args->rule contains the matching rule, 811 * args->rule.info has additional information. 812 * 813 */ 814int 815ipfw_chk(struct ip_fw_args *args) 816{ 817 818 /* 819 * Local variables holding state while processing a packet: 820 * 821 * IMPORTANT NOTE: to speed up the processing of rules, there 822 * are some assumption on the values of the variables, which 823 * are documented here. Should you change them, please check 824 * the implementation of the various instructions to make sure 825 * that they still work. 826 * 827 * args->eh The MAC header. It is non-null for a layer2 828 * packet, it is NULL for a layer-3 packet. 829 * **notyet** 830 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. 831 * 832 * m | args->m Pointer to the mbuf, as received from the caller. 833 * It may change if ipfw_chk() does an m_pullup, or if it 834 * consumes the packet because it calls send_reject(). 835 * XXX This has to change, so that ipfw_chk() never modifies 836 * or consumes the buffer. 837 * ip is the beginning of the ip(4 or 6) header. 838 * Calculated by adding the L3offset to the start of data. 839 * (Until we start using L3offset, the packet is 840 * supposed to start with the ip header). 841 */ 842 struct mbuf *m = args->m; 843 struct ip *ip = mtod(m, struct ip *); 844 845 /* 846 * For rules which contain uid/gid or jail constraints, cache 847 * a copy of the users credentials after the pcb lookup has been 848 * executed. This will speed up the processing of rules with 849 * these types of constraints, as well as decrease contention 850 * on pcb related locks. 851 */ 852#ifndef __FreeBSD__ 853 struct bsd_ucred ucred_cache; 854#else 855 struct ucred *ucred_cache = NULL; 856#endif 857 int ucred_lookup = 0; 858 859 /* 860 * oif | args->oif If NULL, ipfw_chk has been called on the 861 * inbound path (ether_input, ip_input). 862 * If non-NULL, ipfw_chk has been called on the outbound path 863 * (ether_output, ip_output). 864 */ 865 struct ifnet *oif = args->oif; 866 867 int f_pos = 0; /* index of current rule in the array */ 868 int retval = 0; 869 870 /* 871 * hlen The length of the IP header. 872 */ 873 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 874 875 /* 876 * offset The offset of a fragment. offset != 0 means that 877 * we have a fragment at this offset of an IPv4 packet. 878 * offset == 0 means that (if this is an IPv4 packet) 879 * this is the first or only fragment. 880 * For IPv6 offset == 0 means there is no Fragment Header or there 881 * is a single packet fragement (fragement header added without 882 * needed). We will treat a single packet fragment as if there 883 * was no fragment header (or log/block depending on the 884 * V_fw_permit_single_frag6 sysctl setting). 885 * If offset != 0 for IPv6 always use correct mask to 886 * get the correct offset because we add IP6F_MORE_FRAG to be able 887 * to dectect the first of multiple fragments which would 888 * otherwise have offset = 0. 889 */ 890 u_short offset = 0; 891 892 /* 893 * Local copies of addresses. They are only valid if we have 894 * an IP packet. 895 * 896 * proto The protocol. Set to 0 for non-ip packets, 897 * or to the protocol read from the packet otherwise. 898 * proto != 0 means that we have an IPv4 packet. 899 * 900 * src_port, dst_port port numbers, in HOST format. Only 901 * valid for TCP and UDP packets. 902 * 903 * src_ip, dst_ip ip addresses, in NETWORK format. 904 * Only valid for IPv4 packets. 905 */ 906 uint8_t proto; 907 uint16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 908 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 909 uint16_t iplen=0; 910 int pktlen; 911 uint16_t etype = 0; /* Host order stored ether type */ 912 913 /* 914 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 915 * MATCH_NONE when checked and not matched (q = NULL), 916 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 917 */ 918 int dyn_dir = MATCH_UNKNOWN; 919 ipfw_dyn_rule *q = NULL; 920 struct ip_fw_chain *chain = &V_layer3_chain; 921 922 /* 923 * We store in ulp a pointer to the upper layer protocol header. 924 * In the ipv4 case this is easy to determine from the header, 925 * but for ipv6 we might have some additional headers in the middle. 926 * ulp is NULL if not found. 927 */ 928 void *ulp = NULL; /* upper layer protocol pointer. */ 929 930 /* XXX ipv6 variables */ 931 int is_ipv6 = 0; 932 uint8_t icmp6_type = 0; 933 uint16_t ext_hd = 0; /* bits vector for extension header filtering */ 934 /* end of ipv6 variables */ 935 936 int is_ipv4 = 0; 937 938 int done = 0; /* flag to exit the outer loop */ 939 940 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready)) 941 return (IP_FW_PASS); /* accept */ 942 943 dst_ip.s_addr = 0; /* make sure it is initialized */ 944 src_ip.s_addr = 0; /* make sure it is initialized */ 945 pktlen = m->m_pkthdr.len; 946 args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */ 947 proto = args->f_id.proto = 0; /* mark f_id invalid */ 948 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 949 950/* 951 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 952 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 953 * pointer might become stale after other pullups (but we never use it 954 * this way). 955 */ 956#define PULLUP_TO(_len, p, T) PULLUP_LEN(_len, p, sizeof(T)) 957#define PULLUP_LEN(_len, p, T) \ 958do { \ 959 int x = (_len) + T; \ 960 if ((m)->m_len < x) { \ 961 args->m = m = m_pullup(m, x); \ 962 if (m == NULL) \ 963 goto pullup_failed; \ 964 } \ 965 p = (mtod(m, char *) + (_len)); \ 966} while (0) 967 968 /* 969 * if we have an ether header, 970 */ 971 if (args->eh) 972 etype = ntohs(args->eh->ether_type); 973 974 /* Identify IP packets and fill up variables. */ 975 if (pktlen >= sizeof(struct ip6_hdr) && 976 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { 977 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 978 is_ipv6 = 1; 979 args->f_id.addr_type = 6; 980 hlen = sizeof(struct ip6_hdr); 981 proto = ip6->ip6_nxt; 982 983 /* Search extension headers to find upper layer protocols */ 984 while (ulp == NULL) { 985 switch (proto) { 986 case IPPROTO_ICMPV6: 987 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 988 icmp6_type = ICMP6(ulp)->icmp6_type; 989 break; 990 991 case IPPROTO_TCP: 992 PULLUP_TO(hlen, ulp, struct tcphdr); 993 dst_port = TCP(ulp)->th_dport; 994 src_port = TCP(ulp)->th_sport; 995 /* save flags for dynamic rules */ 996 args->f_id._flags = TCP(ulp)->th_flags; 997 break; 998 999 case IPPROTO_SCTP: 1000 PULLUP_TO(hlen, ulp, struct sctphdr); 1001 src_port = SCTP(ulp)->src_port; 1002 dst_port = SCTP(ulp)->dest_port; 1003 break; 1004 1005 case IPPROTO_UDP: 1006 PULLUP_TO(hlen, ulp, struct udphdr); 1007 dst_port = UDP(ulp)->uh_dport; 1008 src_port = UDP(ulp)->uh_sport; 1009 break; 1010 1011 case IPPROTO_HOPOPTS: /* RFC 2460 */ 1012 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1013 ext_hd |= EXT_HOPOPTS; 1014 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1015 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1016 ulp = NULL; 1017 break; 1018 1019 case IPPROTO_ROUTING: /* RFC 2460 */ 1020 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 1021 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 1022 case 0: 1023 ext_hd |= EXT_RTHDR0; 1024 break; 1025 case 2: 1026 ext_hd |= EXT_RTHDR2; 1027 break; 1028 default: 1029 printf("IPFW2: IPV6 - Unknown Routing " 1030 "Header type(%d)\n", 1031 ((struct ip6_rthdr *)ulp)->ip6r_type); 1032 if (V_fw_deny_unknown_exthdrs) 1033 return (IP_FW_DENY); 1034 break; 1035 } 1036 ext_hd |= EXT_ROUTING; 1037 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 1038 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 1039 ulp = NULL; 1040 break; 1041 1042 case IPPROTO_FRAGMENT: /* RFC 2460 */ 1043 PULLUP_TO(hlen, ulp, struct ip6_frag); 1044 ext_hd |= EXT_FRAGMENT; 1045 hlen += sizeof (struct ip6_frag); 1046 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 1047 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 1048 IP6F_OFF_MASK; 1049 /* Add IP6F_MORE_FRAG for offset of first 1050 * fragment to be != 0 if there shall be more. */ 1051 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 1052 IP6F_MORE_FRAG; 1053 if (V_fw_permit_single_frag6 == 0 && 1054 offset == 0) { 1055 printf("IPFW2: IPV6 - Invalid Fragment " 1056 "Header\n"); 1057 if (V_fw_deny_unknown_exthdrs) 1058 return (IP_FW_DENY); 1059 break; 1060 } 1061 args->f_id.extra = 1062 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 1063 ulp = NULL; 1064 break; 1065 1066 case IPPROTO_DSTOPTS: /* RFC 2460 */ 1067 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1068 ext_hd |= EXT_DSTOPTS; 1069 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1070 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1071 ulp = NULL; 1072 break; 1073 1074 case IPPROTO_AH: /* RFC 2402 */ 1075 PULLUP_TO(hlen, ulp, struct ip6_ext); 1076 ext_hd |= EXT_AH; 1077 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 1078 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 1079 ulp = NULL; 1080 break; 1081 1082 case IPPROTO_ESP: /* RFC 2406 */ 1083 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 1084 /* Anything past Seq# is variable length and 1085 * data past this ext. header is encrypted. */ 1086 ext_hd |= EXT_ESP; 1087 break; 1088 1089 case IPPROTO_NONE: /* RFC 2460 */ 1090 /* 1091 * Packet ends here, and IPv6 header has 1092 * already been pulled up. If ip6e_len!=0 1093 * then octets must be ignored. 1094 */ 1095 ulp = ip; /* non-NULL to get out of loop. */ 1096 break; 1097 1098 case IPPROTO_OSPFIGP: 1099 /* XXX OSPF header check? */ 1100 PULLUP_TO(hlen, ulp, struct ip6_ext); 1101 break; 1102 1103 case IPPROTO_PIM: 1104 /* XXX PIM header check? */ 1105 PULLUP_TO(hlen, ulp, struct pim); 1106 break; 1107 1108 case IPPROTO_CARP: 1109 PULLUP_TO(hlen, ulp, struct carp_header); 1110 if (((struct carp_header *)ulp)->carp_version != 1111 CARP_VERSION) 1112 return (IP_FW_DENY); 1113 if (((struct carp_header *)ulp)->carp_type != 1114 CARP_ADVERTISEMENT) 1115 return (IP_FW_DENY); 1116 break; 1117 1118 case IPPROTO_IPV6: /* RFC 2893 */ 1119 PULLUP_TO(hlen, ulp, struct ip6_hdr); 1120 break; 1121 1122 case IPPROTO_IPV4: /* RFC 2893 */ 1123 PULLUP_TO(hlen, ulp, struct ip); 1124 break; 1125 1126 default: 1127 printf("IPFW2: IPV6 - Unknown Extension " 1128 "Header(%d), ext_hd=%x\n", proto, ext_hd); 1129 if (V_fw_deny_unknown_exthdrs) 1130 return (IP_FW_DENY); 1131 PULLUP_TO(hlen, ulp, struct ip6_ext); 1132 break; 1133 } /*switch */ 1134 } 1135 ip = mtod(m, struct ip *); 1136 ip6 = (struct ip6_hdr *)ip; 1137 args->f_id.src_ip6 = ip6->ip6_src; 1138 args->f_id.dst_ip6 = ip6->ip6_dst; 1139 args->f_id.src_ip = 0; 1140 args->f_id.dst_ip = 0; 1141 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 1142 } else if (pktlen >= sizeof(struct ip) && 1143 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { 1144 is_ipv4 = 1; 1145 hlen = ip->ip_hl << 2; 1146 args->f_id.addr_type = 4; 1147 1148 /* 1149 * Collect parameters into local variables for faster matching. 1150 */ 1151 proto = ip->ip_p; 1152 src_ip = ip->ip_src; 1153 dst_ip = ip->ip_dst; 1154 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1155 iplen = ntohs(ip->ip_len); 1156 pktlen = iplen < pktlen ? iplen : pktlen; 1157 1158 if (offset == 0) { 1159 switch (proto) { 1160 case IPPROTO_TCP: 1161 PULLUP_TO(hlen, ulp, struct tcphdr); 1162 dst_port = TCP(ulp)->th_dport; 1163 src_port = TCP(ulp)->th_sport; 1164 /* save flags for dynamic rules */ 1165 args->f_id._flags = TCP(ulp)->th_flags; 1166 break; 1167 1168 case IPPROTO_SCTP: 1169 PULLUP_TO(hlen, ulp, struct sctphdr); 1170 src_port = SCTP(ulp)->src_port; 1171 dst_port = SCTP(ulp)->dest_port; 1172 break; 1173 1174 case IPPROTO_UDP: 1175 PULLUP_TO(hlen, ulp, struct udphdr); 1176 dst_port = UDP(ulp)->uh_dport; 1177 src_port = UDP(ulp)->uh_sport; 1178 break; 1179 1180 case IPPROTO_ICMP: 1181 PULLUP_TO(hlen, ulp, struct icmphdr); 1182 //args->f_id.flags = ICMP(ulp)->icmp_type; 1183 break; 1184 1185 default: 1186 break; 1187 } 1188 } 1189 1190 ip = mtod(m, struct ip *); 1191 args->f_id.src_ip = ntohl(src_ip.s_addr); 1192 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1193 } 1194#undef PULLUP_TO 1195 if (proto) { /* we may have port numbers, store them */ 1196 args->f_id.proto = proto; 1197 args->f_id.src_port = src_port = ntohs(src_port); 1198 args->f_id.dst_port = dst_port = ntohs(dst_port); 1199 } 1200 1201 IPFW_RLOCK(chain); 1202 if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */ 1203 IPFW_RUNLOCK(chain); 1204 return (IP_FW_PASS); /* accept */ 1205 } 1206 if (args->rule.slot) { 1207 /* 1208 * Packet has already been tagged as a result of a previous 1209 * match on rule args->rule aka args->rule_id (PIPE, QUEUE, 1210 * REASS, NETGRAPH, DIVERT/TEE...) 1211 * Validate the slot and continue from the next one 1212 * if still present, otherwise do a lookup. 1213 */ 1214 f_pos = (args->rule.chain_id == chain->id) ? 1215 args->rule.slot : 1216 ipfw_find_rule(chain, args->rule.rulenum, 1217 args->rule.rule_id); 1218 } else { 1219 f_pos = 0; 1220 } 1221 1222 /* 1223 * Now scan the rules, and parse microinstructions for each rule. 1224 * We have two nested loops and an inner switch. Sometimes we 1225 * need to break out of one or both loops, or re-enter one of 1226 * the loops with updated variables. Loop variables are: 1227 * 1228 * f_pos (outer loop) points to the current rule. 1229 * On output it points to the matching rule. 1230 * done (outer loop) is used as a flag to break the loop. 1231 * l (inner loop) residual length of current rule. 1232 * cmd points to the current microinstruction. 1233 * 1234 * We break the inner loop by setting l=0 and possibly 1235 * cmdlen=0 if we don't want to advance cmd. 1236 * We break the outer loop by setting done=1 1237 * We can restart the inner loop by setting l>0 and f_pos, f, cmd 1238 * as needed. 1239 */ 1240 for (; f_pos < chain->n_rules; f_pos++) { 1241 ipfw_insn *cmd; 1242 uint32_t tablearg = 0; 1243 int l, cmdlen, skip_or; /* skip rest of OR block */ 1244 struct ip_fw *f; 1245 1246 f = chain->map[f_pos]; 1247 if (V_set_disable & (1 << f->set) ) 1248 continue; 1249 1250 skip_or = 0; 1251 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1252 l -= cmdlen, cmd += cmdlen) { 1253 int match; 1254 1255 /* 1256 * check_body is a jump target used when we find a 1257 * CHECK_STATE, and need to jump to the body of 1258 * the target rule. 1259 */ 1260 1261/* check_body: */ 1262 cmdlen = F_LEN(cmd); 1263 /* 1264 * An OR block (insn_1 || .. || insn_n) has the 1265 * F_OR bit set in all but the last instruction. 1266 * The first match will set "skip_or", and cause 1267 * the following instructions to be skipped until 1268 * past the one with the F_OR bit clear. 1269 */ 1270 if (skip_or) { /* skip this instruction */ 1271 if ((cmd->len & F_OR) == 0) 1272 skip_or = 0; /* next one is good */ 1273 continue; 1274 } 1275 match = 0; /* set to 1 if we succeed */ 1276 1277 switch (cmd->opcode) { 1278 /* 1279 * The first set of opcodes compares the packet's 1280 * fields with some pattern, setting 'match' if a 1281 * match is found. At the end of the loop there is 1282 * logic to deal with F_NOT and F_OR flags associated 1283 * with the opcode. 1284 */ 1285 case O_NOP: 1286 match = 1; 1287 break; 1288 1289 case O_FORWARD_MAC: 1290 printf("ipfw: opcode %d unimplemented\n", 1291 cmd->opcode); 1292 break; 1293 1294 case O_GID: 1295 case O_UID: 1296 case O_JAIL: 1297 /* 1298 * We only check offset == 0 && proto != 0, 1299 * as this ensures that we have a 1300 * packet with the ports info. 1301 */ 1302 if (offset != 0) 1303 break; 1304 if (proto == IPPROTO_TCP || 1305 proto == IPPROTO_UDP) 1306 match = check_uidgid( 1307 (ipfw_insn_u32 *)cmd, 1308 args, &ucred_lookup, 1309#ifdef __FreeBSD__ 1310 &ucred_cache); 1311#else 1312 (void *)&ucred_cache); 1313#endif 1314 break; 1315 1316 case O_RECV: 1317 match = iface_match(m->m_pkthdr.rcvif, 1318 (ipfw_insn_if *)cmd); 1319 break; 1320 1321 case O_XMIT: 1322 match = iface_match(oif, (ipfw_insn_if *)cmd); 1323 break; 1324 1325 case O_VIA: 1326 match = iface_match(oif ? oif : 1327 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 1328 break; 1329 1330 case O_MACADDR2: 1331 if (args->eh != NULL) { /* have MAC header */ 1332 u_int32_t *want = (u_int32_t *) 1333 ((ipfw_insn_mac *)cmd)->addr; 1334 u_int32_t *mask = (u_int32_t *) 1335 ((ipfw_insn_mac *)cmd)->mask; 1336 u_int32_t *hdr = (u_int32_t *)args->eh; 1337 1338 match = 1339 ( want[0] == (hdr[0] & mask[0]) && 1340 want[1] == (hdr[1] & mask[1]) && 1341 want[2] == (hdr[2] & mask[2]) ); 1342 } 1343 break; 1344 1345 case O_MAC_TYPE: 1346 if (args->eh != NULL) { 1347 u_int16_t *p = 1348 ((ipfw_insn_u16 *)cmd)->ports; 1349 int i; 1350 1351 for (i = cmdlen - 1; !match && i>0; 1352 i--, p += 2) 1353 match = (etype >= p[0] && 1354 etype <= p[1]); 1355 } 1356 break; 1357 1358 case O_FRAG: 1359 match = (offset != 0); 1360 break; 1361 1362 case O_IN: /* "out" is "not in" */ 1363 match = (oif == NULL); 1364 break; 1365 1366 case O_LAYER2: 1367 match = (args->eh != NULL); 1368 break; 1369 1370 case O_DIVERTED: 1371 { 1372 /* For diverted packets, args->rule.info 1373 * contains the divert port (in host format) 1374 * reason and direction. 1375 */ 1376 uint32_t i = args->rule.info; 1377 match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT && 1378 cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2); 1379 } 1380 break; 1381 1382 case O_PROTO: 1383 /* 1384 * We do not allow an arg of 0 so the 1385 * check of "proto" only suffices. 1386 */ 1387 match = (proto == cmd->arg1); 1388 break; 1389 1390 case O_IP_SRC: 1391 match = is_ipv4 && 1392 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1393 src_ip.s_addr); 1394 break; 1395 1396 case O_IP_SRC_LOOKUP: 1397 case O_IP_DST_LOOKUP: 1398 if (is_ipv4) { 1399 uint32_t key = 1400 (cmd->opcode == O_IP_DST_LOOKUP) ? 1401 dst_ip.s_addr : src_ip.s_addr; 1402 uint32_t v = 0; 1403 1404 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) { 1405 /* generic lookup. The key must be 1406 * in 32bit big-endian format. 1407 */ 1408 v = ((ipfw_insn_u32 *)cmd)->d[1]; 1409 if (v == 0) 1410 key = dst_ip.s_addr; 1411 else if (v == 1) 1412 key = src_ip.s_addr; 1413 else if (v == 6) /* dscp */ 1414 key = (ip->ip_tos >> 2) & 0x3f; 1415 else if (offset != 0) 1416 break; 1417 else if (proto != IPPROTO_TCP && 1418 proto != IPPROTO_UDP) 1419 break; 1420 else if (v == 2) 1421 key = htonl(dst_port); 1422 else if (v == 3) 1423 key = htonl(src_port); 1424 else if (v == 4 || v == 5) { 1425 check_uidgid( 1426 (ipfw_insn_u32 *)cmd, 1427 args, &ucred_lookup, 1428#ifdef __FreeBSD__ 1429 &ucred_cache); 1430 if (v == 4 /* O_UID */) 1431 key = ucred_cache->cr_uid; 1432 else if (v == 5 /* O_JAIL */) 1433 key = ucred_cache->cr_prison->pr_id; 1434#else /* !__FreeBSD__ */ 1435 (void *)&ucred_cache); 1436 if (v ==4 /* O_UID */) 1437 key = ucred_cache.uid; 1438 else if (v == 5 /* O_JAIL */) 1439 key = ucred_cache.xid; 1440#endif /* !__FreeBSD__ */ 1441 key = htonl(key); 1442 } else 1443 break; 1444 } 1445 match = ipfw_lookup_table(chain, 1446 cmd->arg1, key, &v); 1447 if (!match) 1448 break; 1449 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 1450 match = 1451 ((ipfw_insn_u32 *)cmd)->d[0] == v; 1452 else 1453 tablearg = v; 1454 } 1455 break; 1456 1457 case O_IP_SRC_MASK: 1458 case O_IP_DST_MASK: 1459 if (is_ipv4) { 1460 uint32_t a = 1461 (cmd->opcode == O_IP_DST_MASK) ? 1462 dst_ip.s_addr : src_ip.s_addr; 1463 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 1464 int i = cmdlen-1; 1465 1466 for (; !match && i>0; i-= 2, p+= 2) 1467 match = (p[0] == (a & p[1])); 1468 } 1469 break; 1470 1471 case O_IP_SRC_ME: 1472 if (is_ipv4) { 1473 struct ifnet *tif; 1474 1475 INADDR_TO_IFP(src_ip, tif); 1476 match = (tif != NULL); 1477 break; 1478 } 1479#ifdef INET6 1480 /* FALLTHROUGH */ 1481 case O_IP6_SRC_ME: 1482 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 1483#endif 1484 break; 1485 1486 case O_IP_DST_SET: 1487 case O_IP_SRC_SET: 1488 if (is_ipv4) { 1489 u_int32_t *d = (u_int32_t *)(cmd+1); 1490 u_int32_t addr = 1491 cmd->opcode == O_IP_DST_SET ? 1492 args->f_id.dst_ip : 1493 args->f_id.src_ip; 1494 1495 if (addr < d[0]) 1496 break; 1497 addr -= d[0]; /* subtract base */ 1498 match = (addr < cmd->arg1) && 1499 ( d[ 1 + (addr>>5)] & 1500 (1<<(addr & 0x1f)) ); 1501 } 1502 break; 1503 1504 case O_IP_DST: 1505 match = is_ipv4 && 1506 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1507 dst_ip.s_addr); 1508 break; 1509 1510 case O_IP_DST_ME: 1511 if (is_ipv4) { 1512 struct ifnet *tif; 1513 1514 INADDR_TO_IFP(dst_ip, tif); 1515 match = (tif != NULL); 1516 break; 1517 } 1518#ifdef INET6 1519 /* FALLTHROUGH */ 1520 case O_IP6_DST_ME: 1521 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 1522#endif 1523 break; 1524 1525 1526 case O_IP_SRCPORT: 1527 case O_IP_DSTPORT: 1528 /* 1529 * offset == 0 && proto != 0 is enough 1530 * to guarantee that we have a 1531 * packet with port info. 1532 */ 1533 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 1534 && offset == 0) { 1535 u_int16_t x = 1536 (cmd->opcode == O_IP_SRCPORT) ? 1537 src_port : dst_port ; 1538 u_int16_t *p = 1539 ((ipfw_insn_u16 *)cmd)->ports; 1540 int i; 1541 1542 for (i = cmdlen - 1; !match && i>0; 1543 i--, p += 2) 1544 match = (x>=p[0] && x<=p[1]); 1545 } 1546 break; 1547 1548 case O_ICMPTYPE: 1549 match = (offset == 0 && proto==IPPROTO_ICMP && 1550 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 1551 break; 1552 1553#ifdef INET6 1554 case O_ICMP6TYPE: 1555 match = is_ipv6 && offset == 0 && 1556 proto==IPPROTO_ICMPV6 && 1557 icmp6type_match( 1558 ICMP6(ulp)->icmp6_type, 1559 (ipfw_insn_u32 *)cmd); 1560 break; 1561#endif /* INET6 */ 1562 1563 case O_IPOPT: 1564 match = (is_ipv4 && 1565 ipopts_match(ip, cmd) ); 1566 break; 1567 1568 case O_IPVER: 1569 match = (is_ipv4 && 1570 cmd->arg1 == ip->ip_v); 1571 break; 1572 1573 case O_IPID: 1574 case O_IPLEN: 1575 case O_IPTTL: 1576 if (is_ipv4) { /* only for IP packets */ 1577 uint16_t x; 1578 uint16_t *p; 1579 int i; 1580 1581 if (cmd->opcode == O_IPLEN) 1582 x = iplen; 1583 else if (cmd->opcode == O_IPTTL) 1584 x = ip->ip_ttl; 1585 else /* must be IPID */ 1586 x = ntohs(ip->ip_id); 1587 if (cmdlen == 1) { 1588 match = (cmd->arg1 == x); 1589 break; 1590 } 1591 /* otherwise we have ranges */ 1592 p = ((ipfw_insn_u16 *)cmd)->ports; 1593 i = cmdlen - 1; 1594 for (; !match && i>0; i--, p += 2) 1595 match = (x >= p[0] && x <= p[1]); 1596 } 1597 break; 1598 1599 case O_IPPRECEDENCE: 1600 match = (is_ipv4 && 1601 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 1602 break; 1603 1604 case O_IPTOS: 1605 match = (is_ipv4 && 1606 flags_match(cmd, ip->ip_tos)); 1607 break; 1608 1609 case O_TCPDATALEN: 1610 if (proto == IPPROTO_TCP && offset == 0) { 1611 struct tcphdr *tcp; 1612 uint16_t x; 1613 uint16_t *p; 1614 int i; 1615 1616 tcp = TCP(ulp); 1617 x = iplen - 1618 ((ip->ip_hl + tcp->th_off) << 2); 1619 if (cmdlen == 1) { 1620 match = (cmd->arg1 == x); 1621 break; 1622 } 1623 /* otherwise we have ranges */ 1624 p = ((ipfw_insn_u16 *)cmd)->ports; 1625 i = cmdlen - 1; 1626 for (; !match && i>0; i--, p += 2) 1627 match = (x >= p[0] && x <= p[1]); 1628 } 1629 break; 1630 1631 case O_TCPFLAGS: 1632 match = (proto == IPPROTO_TCP && offset == 0 && 1633 flags_match(cmd, TCP(ulp)->th_flags)); 1634 break; 1635 1636 case O_TCPOPTS: 1637 PULLUP_LEN(hlen, ulp, (TCP(ulp)->th_off << 2)); 1638 match = (proto == IPPROTO_TCP && offset == 0 && 1639 tcpopts_match(TCP(ulp), cmd)); 1640 break; 1641 1642 case O_TCPSEQ: 1643 match = (proto == IPPROTO_TCP && offset == 0 && 1644 ((ipfw_insn_u32 *)cmd)->d[0] == 1645 TCP(ulp)->th_seq); 1646 break; 1647 1648 case O_TCPACK: 1649 match = (proto == IPPROTO_TCP && offset == 0 && 1650 ((ipfw_insn_u32 *)cmd)->d[0] == 1651 TCP(ulp)->th_ack); 1652 break; 1653 1654 case O_TCPWIN: 1655 match = (proto == IPPROTO_TCP && offset == 0 && 1656 cmd->arg1 == TCP(ulp)->th_win); 1657 break; 1658 1659 case O_ESTAB: 1660 /* reject packets which have SYN only */ 1661 /* XXX should i also check for TH_ACK ? */ 1662 match = (proto == IPPROTO_TCP && offset == 0 && 1663 (TCP(ulp)->th_flags & 1664 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 1665 break; 1666 1667 case O_ALTQ: { 1668 struct pf_mtag *at; 1669 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 1670 1671 match = 1; 1672 at = pf_find_mtag(m); 1673 if (at != NULL && at->qid != 0) 1674 break; 1675 at = pf_get_mtag(m); 1676 if (at == NULL) { 1677 /* 1678 * Let the packet fall back to the 1679 * default ALTQ. 1680 */ 1681 break; 1682 } 1683 at->qid = altq->qid; 1684 at->hdr = ip; 1685 break; 1686 } 1687 1688 case O_LOG: 1689 ipfw_log(f, hlen, args, m, 1690 oif, offset, tablearg, ip); 1691 match = 1; 1692 break; 1693 1694 case O_PROB: 1695 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 1696 break; 1697 1698 case O_VERREVPATH: 1699 /* Outgoing packets automatically pass/match */ 1700 match = ((oif != NULL) || 1701 (m->m_pkthdr.rcvif == NULL) || 1702 ( 1703#ifdef INET6 1704 is_ipv6 ? 1705 verify_path6(&(args->f_id.src_ip6), 1706 m->m_pkthdr.rcvif) : 1707#endif 1708 verify_path(src_ip, m->m_pkthdr.rcvif, 1709 args->f_id.fib))); 1710 break; 1711 1712 case O_VERSRCREACH: 1713 /* Outgoing packets automatically pass/match */ 1714 match = (hlen > 0 && ((oif != NULL) || 1715#ifdef INET6 1716 is_ipv6 ? 1717 verify_path6(&(args->f_id.src_ip6), 1718 NULL) : 1719#endif 1720 verify_path(src_ip, NULL, args->f_id.fib))); 1721 break; 1722 1723 case O_ANTISPOOF: 1724 /* Outgoing packets automatically pass/match */ 1725 if (oif == NULL && hlen > 0 && 1726 ( (is_ipv4 && in_localaddr(src_ip)) 1727#ifdef INET6 1728 || (is_ipv6 && 1729 in6_localaddr(&(args->f_id.src_ip6))) 1730#endif 1731 )) 1732 match = 1733#ifdef INET6 1734 is_ipv6 ? verify_path6( 1735 &(args->f_id.src_ip6), 1736 m->m_pkthdr.rcvif) : 1737#endif 1738 verify_path(src_ip, 1739 m->m_pkthdr.rcvif, 1740 args->f_id.fib); 1741 else 1742 match = 1; 1743 break; 1744 1745 case O_IPSEC: 1746#ifdef IPSEC 1747 match = (m_tag_find(m, 1748 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 1749#endif 1750 /* otherwise no match */ 1751 break; 1752 1753#ifdef INET6 1754 case O_IP6_SRC: 1755 match = is_ipv6 && 1756 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 1757 &((ipfw_insn_ip6 *)cmd)->addr6); 1758 break; 1759 1760 case O_IP6_DST: 1761 match = is_ipv6 && 1762 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 1763 &((ipfw_insn_ip6 *)cmd)->addr6); 1764 break; 1765 case O_IP6_SRC_MASK: 1766 case O_IP6_DST_MASK: 1767 if (is_ipv6) { 1768 int i = cmdlen - 1; 1769 struct in6_addr p; 1770 struct in6_addr *d = 1771 &((ipfw_insn_ip6 *)cmd)->addr6; 1772 1773 for (; !match && i > 0; d += 2, 1774 i -= F_INSN_SIZE(struct in6_addr) 1775 * 2) { 1776 p = (cmd->opcode == 1777 O_IP6_SRC_MASK) ? 1778 args->f_id.src_ip6: 1779 args->f_id.dst_ip6; 1780 APPLY_MASK(&p, &d[1]); 1781 match = 1782 IN6_ARE_ADDR_EQUAL(&d[0], 1783 &p); 1784 } 1785 } 1786 break; 1787 1788 case O_FLOW6ID: 1789 match = is_ipv6 && 1790 flow6id_match(args->f_id.flow_id6, 1791 (ipfw_insn_u32 *) cmd); 1792 break; 1793 1794 case O_EXT_HDR: 1795 match = is_ipv6 && 1796 (ext_hd & ((ipfw_insn *) cmd)->arg1); 1797 break; 1798 1799 case O_IP6: 1800 match = is_ipv6; 1801 break; 1802#endif 1803 1804 case O_IP4: 1805 match = is_ipv4; 1806 break; 1807 1808 case O_TAG: { 1809 struct m_tag *mtag; 1810 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 1811 tablearg : cmd->arg1; 1812 1813 /* Packet is already tagged with this tag? */ 1814 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 1815 1816 /* We have `untag' action when F_NOT flag is 1817 * present. And we must remove this mtag from 1818 * mbuf and reset `match' to zero (`match' will 1819 * be inversed later). 1820 * Otherwise we should allocate new mtag and 1821 * push it into mbuf. 1822 */ 1823 if (cmd->len & F_NOT) { /* `untag' action */ 1824 if (mtag != NULL) 1825 m_tag_delete(m, mtag); 1826 match = 0; 1827 } else { 1828 if (mtag == NULL) { 1829 mtag = m_tag_alloc( MTAG_IPFW, 1830 tag, 0, M_NOWAIT); 1831 if (mtag != NULL) 1832 m_tag_prepend(m, mtag); 1833 } 1834 match = 1; 1835 } 1836 break; 1837 } 1838 1839 case O_FIB: /* try match the specified fib */ 1840 if (args->f_id.fib == cmd->arg1) 1841 match = 1; 1842 break; 1843 1844 case O_SOCKARG: { 1845 struct inpcb *inp = args->inp; 1846 struct inpcbinfo *pi; 1847 1848 if (is_ipv6) /* XXX can we remove this ? */ 1849 break; 1850 1851 if (proto == IPPROTO_TCP) 1852 pi = &V_tcbinfo; 1853 else if (proto == IPPROTO_UDP) 1854 pi = &V_udbinfo; 1855 else 1856 break; 1857 1858 /* 1859 * XXXRW: so_user_cookie should almost 1860 * certainly be inp_user_cookie? 1861 */ 1862 1863 /* For incomming packet, lookup up the 1864 inpcb using the src/dest ip/port tuple */ 1865 if (inp == NULL) { 1866 inp = in_pcblookup(pi, 1867 src_ip, htons(src_port), 1868 dst_ip, htons(dst_port), 1869 INPLOOKUP_RLOCKPCB, NULL); 1870 if (inp != NULL) { 1871 tablearg = 1872 inp->inp_socket->so_user_cookie; 1873 if (tablearg) 1874 match = 1; 1875 INP_RUNLOCK(inp); 1876 } 1877 } else { 1878 if (inp->inp_socket) { 1879 tablearg = 1880 inp->inp_socket->so_user_cookie; 1881 if (tablearg) 1882 match = 1; 1883 } 1884 } 1885 break; 1886 } 1887 1888 case O_TAGGED: { 1889 struct m_tag *mtag; 1890 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 1891 tablearg : cmd->arg1; 1892 1893 if (cmdlen == 1) { 1894 match = m_tag_locate(m, MTAG_IPFW, 1895 tag, NULL) != NULL; 1896 break; 1897 } 1898 1899 /* we have ranges */ 1900 for (mtag = m_tag_first(m); 1901 mtag != NULL && !match; 1902 mtag = m_tag_next(m, mtag)) { 1903 uint16_t *p; 1904 int i; 1905 1906 if (mtag->m_tag_cookie != MTAG_IPFW) 1907 continue; 1908 1909 p = ((ipfw_insn_u16 *)cmd)->ports; 1910 i = cmdlen - 1; 1911 for(; !match && i > 0; i--, p += 2) 1912 match = 1913 mtag->m_tag_id >= p[0] && 1914 mtag->m_tag_id <= p[1]; 1915 } 1916 break; 1917 } 1918 1919 /* 1920 * The second set of opcodes represents 'actions', 1921 * i.e. the terminal part of a rule once the packet 1922 * matches all previous patterns. 1923 * Typically there is only one action for each rule, 1924 * and the opcode is stored at the end of the rule 1925 * (but there are exceptions -- see below). 1926 * 1927 * In general, here we set retval and terminate the 1928 * outer loop (would be a 'break 3' in some language, 1929 * but we need to set l=0, done=1) 1930 * 1931 * Exceptions: 1932 * O_COUNT and O_SKIPTO actions: 1933 * instead of terminating, we jump to the next rule 1934 * (setting l=0), or to the SKIPTO target (setting 1935 * f/f_len, cmd and l as needed), respectively. 1936 * 1937 * O_TAG, O_LOG and O_ALTQ action parameters: 1938 * perform some action and set match = 1; 1939 * 1940 * O_LIMIT and O_KEEP_STATE: these opcodes are 1941 * not real 'actions', and are stored right 1942 * before the 'action' part of the rule. 1943 * These opcodes try to install an entry in the 1944 * state tables; if successful, we continue with 1945 * the next opcode (match=1; break;), otherwise 1946 * the packet must be dropped (set retval, 1947 * break loops with l=0, done=1) 1948 * 1949 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 1950 * cause a lookup of the state table, and a jump 1951 * to the 'action' part of the parent rule 1952 * if an entry is found, or 1953 * (CHECK_STATE only) a jump to the next rule if 1954 * the entry is not found. 1955 * The result of the lookup is cached so that 1956 * further instances of these opcodes become NOPs. 1957 * The jump to the next rule is done by setting 1958 * l=0, cmdlen=0. 1959 */ 1960 case O_LIMIT: 1961 case O_KEEP_STATE: 1962 if (ipfw_install_state(f, 1963 (ipfw_insn_limit *)cmd, args, tablearg)) { 1964 /* error or limit violation */ 1965 retval = IP_FW_DENY; 1966 l = 0; /* exit inner loop */ 1967 done = 1; /* exit outer loop */ 1968 } 1969 match = 1; 1970 break; 1971 1972 case O_PROBE_STATE: 1973 case O_CHECK_STATE: 1974 /* 1975 * dynamic rules are checked at the first 1976 * keep-state or check-state occurrence, 1977 * with the result being stored in dyn_dir. 1978 * The compiler introduces a PROBE_STATE 1979 * instruction for us when we have a 1980 * KEEP_STATE (because PROBE_STATE needs 1981 * to be run first). 1982 */ 1983 if (dyn_dir == MATCH_UNKNOWN && 1984 (q = ipfw_lookup_dyn_rule(&args->f_id, 1985 &dyn_dir, proto == IPPROTO_TCP ? 1986 TCP(ulp) : NULL)) 1987 != NULL) { 1988 /* 1989 * Found dynamic entry, update stats 1990 * and jump to the 'action' part of 1991 * the parent rule by setting 1992 * f, cmd, l and clearing cmdlen. 1993 */ 1994 q->pcnt++; 1995 q->bcnt += pktlen; 1996 /* XXX we would like to have f_pos 1997 * readily accessible in the dynamic 1998 * rule, instead of having to 1999 * lookup q->rule. 2000 */ 2001 f = q->rule; 2002 f_pos = ipfw_find_rule(chain, 2003 f->rulenum, f->id); 2004 cmd = ACTION_PTR(f); 2005 l = f->cmd_len - f->act_ofs; 2006 ipfw_dyn_unlock(); 2007 cmdlen = 0; 2008 match = 1; 2009 break; 2010 } 2011 /* 2012 * Dynamic entry not found. If CHECK_STATE, 2013 * skip to next rule, if PROBE_STATE just 2014 * ignore and continue with next opcode. 2015 */ 2016 if (cmd->opcode == O_CHECK_STATE) 2017 l = 0; /* exit inner loop */ 2018 match = 1; 2019 break; 2020 2021 case O_ACCEPT: 2022 retval = 0; /* accept */ 2023 l = 0; /* exit inner loop */ 2024 done = 1; /* exit outer loop */ 2025 break; 2026 2027 case O_PIPE: 2028 case O_QUEUE: 2029 set_match(args, f_pos, chain); 2030 args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ? 2031 tablearg : cmd->arg1; 2032 if (cmd->opcode == O_PIPE) 2033 args->rule.info |= IPFW_IS_PIPE; 2034 if (V_fw_one_pass) 2035 args->rule.info |= IPFW_ONEPASS; 2036 retval = IP_FW_DUMMYNET; 2037 l = 0; /* exit inner loop */ 2038 done = 1; /* exit outer loop */ 2039 break; 2040 2041 case O_DIVERT: 2042 case O_TEE: 2043 if (args->eh) /* not on layer 2 */ 2044 break; 2045 /* otherwise this is terminal */ 2046 l = 0; /* exit inner loop */ 2047 done = 1; /* exit outer loop */ 2048 retval = (cmd->opcode == O_DIVERT) ? 2049 IP_FW_DIVERT : IP_FW_TEE; 2050 set_match(args, f_pos, chain); 2051 args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ? 2052 tablearg : cmd->arg1; 2053 break; 2054 2055 case O_COUNT: 2056 f->pcnt++; /* update stats */ 2057 f->bcnt += pktlen; 2058 f->timestamp = time_uptime; 2059 l = 0; /* exit inner loop */ 2060 break; 2061 2062 case O_SKIPTO: 2063 f->pcnt++; /* update stats */ 2064 f->bcnt += pktlen; 2065 f->timestamp = time_uptime; 2066 /* If possible use cached f_pos (in f->next_rule), 2067 * whose version is written in f->next_rule 2068 * (horrible hacks to avoid changing the ABI). 2069 */ 2070 if (cmd->arg1 != IP_FW_TABLEARG && 2071 (uintptr_t)f->x_next == chain->id) { 2072 f_pos = (uintptr_t)f->next_rule; 2073 } else { 2074 int i = (cmd->arg1 == IP_FW_TABLEARG) ? 2075 tablearg : cmd->arg1; 2076 /* make sure we do not jump backward */ 2077 if (i <= f->rulenum) 2078 i = f->rulenum + 1; 2079 f_pos = ipfw_find_rule(chain, i, 0); 2080 /* update the cache */ 2081 if (cmd->arg1 != IP_FW_TABLEARG) { 2082 f->next_rule = 2083 (void *)(uintptr_t)f_pos; 2084 f->x_next = 2085 (void *)(uintptr_t)chain->id; 2086 } 2087 } 2088 /* 2089 * Skip disabled rules, and re-enter 2090 * the inner loop with the correct 2091 * f_pos, f, l and cmd. 2092 * Also clear cmdlen and skip_or 2093 */ 2094 for (; f_pos < chain->n_rules - 1 && 2095 (V_set_disable & 2096 (1 << chain->map[f_pos]->set)); 2097 f_pos++) 2098 ; 2099 /* Re-enter the inner loop at the skipto rule. */ 2100 f = chain->map[f_pos]; 2101 l = f->cmd_len; 2102 cmd = f->cmd; 2103 match = 1; 2104 cmdlen = 0; 2105 skip_or = 0; 2106 continue; 2107 break; /* not reached */ 2108 2109 case O_CALLRETURN: { 2110 /* 2111 * Implementation of `subroutine' call/return, 2112 * in the stack carried in an mbuf tag. This 2113 * is different from `skipto' in that any call 2114 * address is possible (`skipto' must prevent 2115 * backward jumps to avoid endless loops). 2116 * We have `return' action when F_NOT flag is 2117 * present. The `m_tag_id' field is used as 2118 * stack pointer. 2119 */ 2120 struct m_tag *mtag; 2121 uint16_t jmpto, *stack; 2122 2123#define IS_CALL ((cmd->len & F_NOT) == 0) 2124#define IS_RETURN ((cmd->len & F_NOT) != 0) 2125 /* 2126 * Hand-rolled version of m_tag_locate() with 2127 * wildcard `type'. 2128 * If not already tagged, allocate new tag. 2129 */ 2130 mtag = m_tag_first(m); 2131 while (mtag != NULL) { 2132 if (mtag->m_tag_cookie == 2133 MTAG_IPFW_CALL) 2134 break; 2135 mtag = m_tag_next(m, mtag); 2136 } 2137 if (mtag == NULL && IS_CALL) { 2138 mtag = m_tag_alloc(MTAG_IPFW_CALL, 0, 2139 IPFW_CALLSTACK_SIZE * 2140 sizeof(uint16_t), M_NOWAIT); 2141 if (mtag != NULL) 2142 m_tag_prepend(m, mtag); 2143 } 2144 2145 /* 2146 * On error both `call' and `return' just 2147 * continue with next rule. 2148 */ 2149 if (IS_RETURN && (mtag == NULL || 2150 mtag->m_tag_id == 0)) { 2151 l = 0; /* exit inner loop */ 2152 break; 2153 } 2154 if (IS_CALL && (mtag == NULL || 2155 mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) { 2156 printf("ipfw: call stack error, " 2157 "go to next rule\n"); 2158 l = 0; /* exit inner loop */ 2159 break; 2160 } 2161 2162 f->pcnt++; /* update stats */ 2163 f->bcnt += pktlen; 2164 f->timestamp = time_uptime; 2165 stack = (uint16_t *)(mtag + 1); 2166 2167 /* 2168 * The `call' action may use cached f_pos 2169 * (in f->next_rule), whose version is written 2170 * in f->next_rule. 2171 * The `return' action, however, doesn't have 2172 * fixed jump address in cmd->arg1 and can't use 2173 * cache. 2174 */ 2175 if (IS_CALL) { 2176 stack[mtag->m_tag_id] = f->rulenum; 2177 mtag->m_tag_id++; 2178 if (cmd->arg1 != IP_FW_TABLEARG && 2179 (uintptr_t)f->x_next == chain->id) { 2180 f_pos = (uintptr_t)f->next_rule; 2181 } else { 2182 jmpto = (cmd->arg1 == 2183 IP_FW_TABLEARG) ? tablearg: 2184 cmd->arg1; 2185 f_pos = ipfw_find_rule(chain, 2186 jmpto, 0); 2187 /* update the cache */ 2188 if (cmd->arg1 != 2189 IP_FW_TABLEARG) { 2190 f->next_rule = 2191 (void *)(uintptr_t) 2192 f_pos; 2193 f->x_next = 2194 (void *)(uintptr_t) 2195 chain->id; 2196 } 2197 } 2198 } else { /* `return' action */ 2199 mtag->m_tag_id--; 2200 jmpto = stack[mtag->m_tag_id] + 1; 2201 f_pos = ipfw_find_rule(chain, jmpto, 0); 2202 } 2203 2204 /* 2205 * Skip disabled rules, and re-enter 2206 * the inner loop with the correct 2207 * f_pos, f, l and cmd. 2208 * Also clear cmdlen and skip_or 2209 */ 2210 for (; f_pos < chain->n_rules - 1 && 2211 (V_set_disable & 2212 (1 << chain->map[f_pos]->set)); f_pos++) 2213 ; 2214 /* Re-enter the inner loop at the dest rule. */ 2215 f = chain->map[f_pos]; 2216 l = f->cmd_len; 2217 cmd = f->cmd; 2218 cmdlen = 0; 2219 skip_or = 0; 2220 continue; 2221 break; /* NOTREACHED */ 2222 } 2223#undef IS_CALL 2224#undef IS_RETURN 2225 2226 case O_REJECT: 2227 /* 2228 * Drop the packet and send a reject notice 2229 * if the packet is not ICMP (or is an ICMP 2230 * query), and it is not multicast/broadcast. 2231 */ 2232 if (hlen > 0 && is_ipv4 && offset == 0 && 2233 (proto != IPPROTO_ICMP || 2234 is_icmp_query(ICMP(ulp))) && 2235 !(m->m_flags & (M_BCAST|M_MCAST)) && 2236 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 2237 send_reject(args, cmd->arg1, iplen, ip); 2238 m = args->m; 2239 } 2240 /* FALLTHROUGH */ 2241#ifdef INET6 2242 case O_UNREACH6: 2243 if (hlen > 0 && is_ipv6 && 2244 ((offset & IP6F_OFF_MASK) == 0) && 2245 (proto != IPPROTO_ICMPV6 || 2246 (is_icmp6_query(icmp6_type) == 1)) && 2247 !(m->m_flags & (M_BCAST|M_MCAST)) && 2248 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 2249 send_reject6( 2250 args, cmd->arg1, hlen, 2251 (struct ip6_hdr *)ip); 2252 m = args->m; 2253 } 2254 /* FALLTHROUGH */ 2255#endif 2256 case O_DENY: 2257 retval = IP_FW_DENY; 2258 l = 0; /* exit inner loop */ 2259 done = 1; /* exit outer loop */ 2260 break; 2261 2262 case O_FORWARD_IP: 2263 if (args->eh) /* not valid on layer2 pkts */ 2264 break; 2265 if (q == NULL || q->rule != f || 2266 dyn_dir == MATCH_FORWARD) { 2267 struct sockaddr_in *sa; 2268 sa = &(((ipfw_insn_sa *)cmd)->sa); 2269 if (sa->sin_addr.s_addr == INADDR_ANY) { 2270 bcopy(sa, &args->hopstore, 2271 sizeof(*sa)); 2272 args->hopstore.sin_addr.s_addr = 2273 htonl(tablearg); 2274 args->next_hop = &args->hopstore; 2275 } else { 2276 args->next_hop = sa; 2277 } 2278 } 2279 retval = IP_FW_PASS; 2280 l = 0; /* exit inner loop */ 2281 done = 1; /* exit outer loop */ 2282 break; 2283 2284 case O_NETGRAPH: 2285 case O_NGTEE: 2286 set_match(args, f_pos, chain); 2287 args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ? 2288 tablearg : cmd->arg1; 2289 if (V_fw_one_pass) 2290 args->rule.info |= IPFW_ONEPASS; 2291 retval = (cmd->opcode == O_NETGRAPH) ? 2292 IP_FW_NETGRAPH : IP_FW_NGTEE; 2293 l = 0; /* exit inner loop */ 2294 done = 1; /* exit outer loop */ 2295 break; 2296 2297 case O_SETFIB: { 2298 uint32_t fib; 2299 2300 f->pcnt++; /* update stats */ 2301 f->bcnt += pktlen; 2302 f->timestamp = time_uptime; 2303 fib = (cmd->arg1 == IP_FW_TABLEARG) ? tablearg: 2304 cmd->arg1; 2305 if (fib >= rt_numfibs) 2306 fib = 0; 2307 M_SETFIB(m, fib); 2308 args->f_id.fib = fib; 2309 l = 0; /* exit inner loop */ 2310 break; 2311 } 2312 2313 case O_NAT: 2314 if (!IPFW_NAT_LOADED) { 2315 retval = IP_FW_DENY; 2316 } else { 2317 struct cfg_nat *t; 2318 int nat_id; 2319 2320 set_match(args, f_pos, chain); 2321 /* Check if this is 'global' nat rule */ 2322 if (cmd->arg1 == 0) { 2323 retval = ipfw_nat_ptr(args, NULL, m); 2324 l = 0; 2325 done = 1; 2326 break; 2327 } 2328 t = ((ipfw_insn_nat *)cmd)->nat; 2329 if (t == NULL) { 2330 nat_id = (cmd->arg1 == IP_FW_TABLEARG) ? 2331 tablearg : cmd->arg1; 2332 t = (*lookup_nat_ptr)(&chain->nat, nat_id); 2333 2334 if (t == NULL) { 2335 retval = IP_FW_DENY; 2336 l = 0; /* exit inner loop */ 2337 done = 1; /* exit outer loop */ 2338 break; 2339 } 2340 if (cmd->arg1 != IP_FW_TABLEARG) 2341 ((ipfw_insn_nat *)cmd)->nat = t; 2342 } 2343 retval = ipfw_nat_ptr(args, t, m); 2344 } 2345 l = 0; /* exit inner loop */ 2346 done = 1; /* exit outer loop */ 2347 break; 2348 2349 case O_REASS: { 2350 int ip_off; 2351 2352 f->pcnt++; 2353 f->bcnt += pktlen; 2354 l = 0; /* in any case exit inner loop */ 2355 ip_off = ntohs(ip->ip_off); 2356 2357 /* if not fragmented, go to next rule */ 2358 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0) 2359 break; 2360 /* 2361 * ip_reass() expects len & off in host 2362 * byte order. 2363 */ 2364 SET_HOST_IPLEN(ip); 2365 2366 args->m = m = ip_reass(m); 2367 2368 /* 2369 * do IP header checksum fixup. 2370 */ 2371 if (m == NULL) { /* fragment got swallowed */ 2372 retval = IP_FW_DENY; 2373 } else { /* good, packet complete */ 2374 int hlen; 2375 2376 ip = mtod(m, struct ip *); 2377 hlen = ip->ip_hl << 2; 2378 SET_NET_IPLEN(ip); 2379 ip->ip_sum = 0; 2380 if (hlen == sizeof(struct ip)) 2381 ip->ip_sum = in_cksum_hdr(ip); 2382 else 2383 ip->ip_sum = in_cksum(m, hlen); 2384 retval = IP_FW_REASS; 2385 set_match(args, f_pos, chain); 2386 } 2387 done = 1; /* exit outer loop */ 2388 break; 2389 } 2390 2391 default: 2392 panic("-- unknown opcode %d\n", cmd->opcode); 2393 } /* end of switch() on opcodes */ 2394 /* 2395 * if we get here with l=0, then match is irrelevant. 2396 */ 2397 2398 if (cmd->len & F_NOT) 2399 match = !match; 2400 2401 if (match) { 2402 if (cmd->len & F_OR) 2403 skip_or = 1; 2404 } else { 2405 if (!(cmd->len & F_OR)) /* not an OR block, */ 2406 break; /* try next rule */ 2407 } 2408 2409 } /* end of inner loop, scan opcodes */ 2410#undef PULLUP_LEN 2411 2412 if (done) 2413 break; 2414 2415/* next_rule:; */ /* try next rule */ 2416 2417 } /* end of outer for, scan rules */ 2418 2419 if (done) { 2420 struct ip_fw *rule = chain->map[f_pos]; 2421 /* Update statistics */ 2422 rule->pcnt++; 2423 rule->bcnt += pktlen; 2424 rule->timestamp = time_uptime; 2425 } else { 2426 retval = IP_FW_DENY; 2427 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 2428 } 2429 IPFW_RUNLOCK(chain); 2430#ifdef __FreeBSD__ 2431 if (ucred_cache != NULL) 2432 crfree(ucred_cache); 2433#endif 2434 return (retval); 2435 2436pullup_failed: 2437 if (V_fw_verbose) 2438 printf("ipfw: pullup failed\n"); 2439 return (IP_FW_DENY); 2440} 2441 2442/* 2443 * Module and VNET glue 2444 */ 2445 2446/* 2447 * Stuff that must be initialised only on boot or module load 2448 */ 2449static int 2450ipfw_init(void) 2451{ 2452 int error = 0; 2453 2454 ipfw_dyn_attach(); 2455 /* 2456 * Only print out this stuff the first time around, 2457 * when called from the sysinit code. 2458 */ 2459 printf("ipfw2 " 2460#ifdef INET6 2461 "(+ipv6) " 2462#endif 2463 "initialized, divert %s, nat %s, " 2464 "rule-based forwarding " 2465#ifdef IPFIREWALL_FORWARD 2466 "enabled, " 2467#else 2468 "disabled, " 2469#endif 2470 "default to %s, logging ", 2471#ifdef IPDIVERT 2472 "enabled", 2473#else 2474 "loadable", 2475#endif 2476#ifdef IPFIREWALL_NAT 2477 "enabled", 2478#else 2479 "loadable", 2480#endif 2481 default_to_accept ? "accept" : "deny"); 2482 2483 /* 2484 * Note: V_xxx variables can be accessed here but the vnet specific 2485 * initializer may not have been called yet for the VIMAGE case. 2486 * Tuneables will have been processed. We will print out values for 2487 * the default vnet. 2488 * XXX This should all be rationalized AFTER 8.0 2489 */ 2490 if (V_fw_verbose == 0) 2491 printf("disabled\n"); 2492 else if (V_verbose_limit == 0) 2493 printf("unlimited\n"); 2494 else 2495 printf("limited to %d packets/entry by default\n", 2496 V_verbose_limit); 2497 2498 ipfw_log_bpf(1); /* init */ 2499 return (error); 2500} 2501 2502/* 2503 * Called for the removal of the last instance only on module unload. 2504 */ 2505static void 2506ipfw_destroy(void) 2507{ 2508 2509 ipfw_log_bpf(0); /* uninit */ 2510 ipfw_dyn_detach(); 2511 printf("IP firewall unloaded\n"); 2512} 2513 2514/* 2515 * Stuff that must be initialized for every instance 2516 * (including the first of course). 2517 */ 2518static int 2519vnet_ipfw_init(const void *unused) 2520{ 2521 int error; 2522 struct ip_fw *rule = NULL; 2523 struct ip_fw_chain *chain; 2524 2525 chain = &V_layer3_chain; 2526 2527 /* First set up some values that are compile time options */ 2528 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 2529 V_fw_deny_unknown_exthdrs = 1; 2530#ifdef IPFIREWALL_VERBOSE 2531 V_fw_verbose = 1; 2532#endif 2533#ifdef IPFIREWALL_VERBOSE_LIMIT 2534 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 2535#endif 2536#ifdef IPFIREWALL_NAT 2537 LIST_INIT(&chain->nat); 2538#endif 2539 2540 /* insert the default rule and create the initial map */ 2541 chain->n_rules = 1; 2542 chain->static_len = sizeof(struct ip_fw); 2543 chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_NOWAIT | M_ZERO); 2544 if (chain->map) 2545 rule = malloc(chain->static_len, M_IPFW, M_NOWAIT | M_ZERO); 2546 if (rule == NULL) { 2547 if (chain->map) 2548 free(chain->map, M_IPFW); 2549 printf("ipfw2: ENOSPC initializing default rule " 2550 "(support disabled)\n"); 2551 return (ENOSPC); 2552 } 2553 error = ipfw_init_tables(chain); 2554 if (error) { 2555 panic("init_tables"); /* XXX Marko fix this ! */ 2556 } 2557 2558 /* fill and insert the default rule */ 2559 rule->act_ofs = 0; 2560 rule->rulenum = IPFW_DEFAULT_RULE; 2561 rule->cmd_len = 1; 2562 rule->set = RESVD_SET; 2563 rule->cmd[0].len = 1; 2564 rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; 2565 chain->rules = chain->default_rule = chain->map[0] = rule; 2566 chain->id = rule->id = 1; 2567 2568 IPFW_LOCK_INIT(chain); 2569 ipfw_dyn_init(); 2570 2571 /* First set up some values that are compile time options */ 2572 V_ipfw_vnet_ready = 1; /* Open for business */ 2573 2574 /* 2575 * Hook the sockopt handler, and the layer2 (V_ip_fw_chk_ptr) 2576 * and pfil hooks for ipv4 and ipv6. Even if the latter two fail 2577 * we still keep the module alive because the sockopt and 2578 * layer2 paths are still useful. 2579 * ipfw[6]_hook return 0 on success, ENOENT on failure, 2580 * so we can ignore the exact return value and just set a flag. 2581 * 2582 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so 2583 * changes in the underlying (per-vnet) variables trigger 2584 * immediate hook()/unhook() calls. 2585 * In layer2 we have the same behaviour, except that V_ether_ipfw 2586 * is checked on each packet because there are no pfil hooks. 2587 */ 2588 V_ip_fw_ctl_ptr = ipfw_ctl; 2589 V_ip_fw_chk_ptr = ipfw_chk; 2590 error = ipfw_attach_hooks(1); 2591 return (error); 2592} 2593 2594/* 2595 * Called for the removal of each instance. 2596 */ 2597static int 2598vnet_ipfw_uninit(const void *unused) 2599{ 2600 struct ip_fw *reap, *rule; 2601 struct ip_fw_chain *chain = &V_layer3_chain; 2602 int i; 2603 2604 V_ipfw_vnet_ready = 0; /* tell new callers to go away */ 2605 /* 2606 * disconnect from ipv4, ipv6, layer2 and sockopt. 2607 * Then grab, release and grab again the WLOCK so we make 2608 * sure the update is propagated and nobody will be in. 2609 */ 2610 (void)ipfw_attach_hooks(0 /* detach */); 2611 V_ip_fw_chk_ptr = NULL; 2612 V_ip_fw_ctl_ptr = NULL; 2613 IPFW_UH_WLOCK(chain); 2614 IPFW_UH_WUNLOCK(chain); 2615 IPFW_UH_WLOCK(chain); 2616 2617 IPFW_WLOCK(chain); 2618 IPFW_WUNLOCK(chain); 2619 IPFW_WLOCK(chain); 2620 2621 ipfw_dyn_uninit(0); /* run the callout_drain */ 2622 ipfw_destroy_tables(chain); 2623 reap = NULL; 2624 for (i = 0; i < chain->n_rules; i++) { 2625 rule = chain->map[i]; 2626 rule->x_next = reap; 2627 reap = rule; 2628 } 2629 if (chain->map) 2630 free(chain->map, M_IPFW); 2631 IPFW_WUNLOCK(chain); 2632 IPFW_UH_WUNLOCK(chain); 2633 if (reap != NULL) 2634 ipfw_reap_rules(reap); 2635 IPFW_LOCK_DESTROY(chain); 2636 ipfw_dyn_uninit(1); /* free the remaining parts */ 2637 return 0; 2638} 2639 2640/* 2641 * Module event handler. 2642 * In general we have the choice of handling most of these events by the 2643 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to 2644 * use the SYSINIT handlers as they are more capable of expressing the 2645 * flow of control during module and vnet operations, so this is just 2646 * a skeleton. Note there is no SYSINIT equivalent of the module 2647 * SHUTDOWN handler, but we don't have anything to do in that case anyhow. 2648 */ 2649static int 2650ipfw_modevent(module_t mod, int type, void *unused) 2651{ 2652 int err = 0; 2653 2654 switch (type) { 2655 case MOD_LOAD: 2656 /* Called once at module load or 2657 * system boot if compiled in. */ 2658 break; 2659 case MOD_QUIESCE: 2660 /* Called before unload. May veto unloading. */ 2661 break; 2662 case MOD_UNLOAD: 2663 /* Called during unload. */ 2664 break; 2665 case MOD_SHUTDOWN: 2666 /* Called during system shutdown. */ 2667 break; 2668 default: 2669 err = EOPNOTSUPP; 2670 break; 2671 } 2672 return err; 2673} 2674 2675static moduledata_t ipfwmod = { 2676 "ipfw", 2677 ipfw_modevent, 2678 0 2679}; 2680 2681/* Define startup order. */ 2682#define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_IFATTACHDOMAIN 2683#define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */ 2684#define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */ 2685#define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */ 2686 2687DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER); 2688MODULE_VERSION(ipfw, 2); 2689/* should declare some dependencies here */ 2690 2691/* 2692 * Starting up. Done in order after ipfwmod() has been called. 2693 * VNET_SYSINIT is also called for each existing vnet and each new vnet. 2694 */ 2695SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 2696 ipfw_init, NULL); 2697VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 2698 vnet_ipfw_init, NULL); 2699 2700/* 2701 * Closing up shop. These are done in REVERSE ORDER, but still 2702 * after ipfwmod() has been called. Not called on reboot. 2703 * VNET_SYSUNINIT is also called for each exiting vnet as it exits. 2704 * or when the module is unloaded. 2705 */ 2706SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 2707 ipfw_destroy, NULL); 2708VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 2709 vnet_ipfw_uninit, NULL); 2710/* end of file */ 2711