ip_fw2.c revision 215701
196191Sdes/*- 296191Sdes * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa 396191Sdes * 496191Sdes * Redistribution and use in source and binary forms, with or without 596191Sdes * modification, are permitted provided that the following conditions 696191Sdes * are met: 796191Sdes * 1. Redistributions of source code must retain the above copyright 896191Sdes * notice, this list of conditions and the following disclaimer. 996191Sdes * 2. Redistributions in binary form must reproduce the above copyright 1096191Sdes * notice, this list of conditions and the following disclaimer in the 1196191Sdes * documentation and/or other materials provided with the distribution. 1296191Sdes * 1396191Sdes * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 1496191Sdes * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 1596191Sdes * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 1696191Sdes * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 1796191Sdes * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 1896191Sdes * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 1996191Sdes * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 2096191Sdes * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 2196191Sdes * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 2296191Sdes * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 2396191Sdes * SUCH DAMAGE. 2496191Sdes */ 2596191Sdes 2696191Sdes#include <sys/cdefs.h> 2796191Sdes__FBSDID("$FreeBSD: head/sys/netinet/ipfw/ip_fw2.c 215701 2010-11-22 19:32:54Z dim $"); 2896191Sdes 2996191Sdes/* 3096191Sdes * The FreeBSD IP packet firewall, main file 3196191Sdes */ 3296191Sdes 3396191Sdes#if !defined(KLD_MODULE) 3496191Sdes#include "opt_ipfw.h" 3596191Sdes#include "opt_ipdivert.h" 3696191Sdes#include "opt_ipdn.h" 3796191Sdes#include "opt_inet.h" 3896191Sdes#ifndef INET 3996191Sdes#error IPFIREWALL requires INET. 4096191Sdes#endif /* INET */ 4196191Sdes#endif 4296191Sdes#include "opt_inet6.h" 4396191Sdes#include "opt_ipsec.h" 4496191Sdes 4596191Sdes#include <sys/param.h> 4696191Sdes#include <sys/systm.h> 4796191Sdes#include <sys/condvar.h> 4896191Sdes#include <sys/eventhandler.h> 4996191Sdes#include <sys/malloc.h> 5096191Sdes#include <sys/mbuf.h> 5196191Sdes#include <sys/kernel.h> 5296191Sdes#include <sys/lock.h> 5396191Sdes#include <sys/jail.h> 5496191Sdes#include <sys/module.h> 5596191Sdes#include <sys/priv.h> 5696191Sdes#include <sys/proc.h> 5796191Sdes#include <sys/rwlock.h> 5896191Sdes#include <sys/socket.h> 5996191Sdes#include <sys/socketvar.h> 6096191Sdes#include <sys/sysctl.h> 6196191Sdes#include <sys/syslog.h> 6296191Sdes#include <sys/ucred.h> 6396191Sdes#include <net/ethernet.h> /* for ETHERTYPE_IP */ 6496191Sdes#include <net/if.h> 6596191Sdes#include <net/route.h> 6696191Sdes#include <net/pf_mtag.h> 6796191Sdes#include <net/vnet.h> 6896191Sdes 6996191Sdes#include <netinet/in.h> 7096191Sdes#include <netinet/in_var.h> 7196191Sdes#include <netinet/in_pcb.h> 7296191Sdes#include <netinet/ip.h> 7396191Sdes#include <netinet/ip_var.h> 7496191Sdes#include <netinet/ip_icmp.h> 7596191Sdes#include <netinet/ip_fw.h> 7696191Sdes#include <netinet/ipfw/ip_fw_private.h> 7796191Sdes#include <netinet/ip_carp.h> 7896191Sdes#include <netinet/pim.h> 7996191Sdes#include <netinet/tcp_var.h> 8096191Sdes#include <netinet/udp.h> 8196191Sdes#include <netinet/udp_var.h> 8296191Sdes#include <netinet/sctp.h> 8396191Sdes 8496191Sdes#include <netinet/ip6.h> 8596191Sdes#include <netinet/icmp6.h> 8696191Sdes#ifdef INET6 8796191Sdes#include <netinet6/scope6_var.h> 8896191Sdes#include <netinet6/ip6_var.h> 8996191Sdes#endif 9096191Sdes 9196191Sdes#include <machine/in_cksum.h> /* XXX for in_cksum */ 9296191Sdes 9396191Sdes#ifdef MAC 9496191Sdes#include <security/mac/mac_framework.h> 9596191Sdes#endif 9696191Sdes 9796191Sdes/* 9896191Sdes * static variables followed by global ones. 9996191Sdes * All ipfw global variables are here. 10096191Sdes */ 10196191Sdes 10296191Sdes/* ipfw_vnet_ready controls when we are open for business */ 10396191Sdesstatic VNET_DEFINE(int, ipfw_vnet_ready) = 0; 10496191Sdes#define V_ipfw_vnet_ready VNET(ipfw_vnet_ready) 10596191Sdes 10696191Sdesstatic VNET_DEFINE(int, fw_deny_unknown_exthdrs); 10796191Sdes#define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs) 10896191Sdes 10996191Sdes#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 11096191Sdesstatic int default_to_accept = 1; 11196191Sdes#else 11296191Sdesstatic int default_to_accept; 11396191Sdes#endif 11496191Sdes 11596191SdesVNET_DEFINE(int, autoinc_step); 116 117/* 118 * Each rule belongs to one of 32 different sets (0..31). 119 * The variable set_disable contains one bit per set. 120 * If the bit is set, all rules in the corresponding set 121 * are disabled. Set RESVD_SET(31) is reserved for the default rule 122 * and rules that are not deleted by the flush command, 123 * and CANNOT be disabled. 124 * Rules in set RESVD_SET can only be deleted individually. 125 */ 126VNET_DEFINE(u_int32_t, set_disable); 127#define V_set_disable VNET(set_disable) 128 129VNET_DEFINE(int, fw_verbose); 130/* counter for ipfw_log(NULL...) */ 131VNET_DEFINE(u_int64_t, norule_counter); 132VNET_DEFINE(int, verbose_limit); 133 134/* layer3_chain contains the list of rules for layer 3 */ 135VNET_DEFINE(struct ip_fw_chain, layer3_chain); 136 137ipfw_nat_t *ipfw_nat_ptr = NULL; 138struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int); 139ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; 140ipfw_nat_cfg_t *ipfw_nat_del_ptr; 141ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; 142ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; 143 144#ifdef SYSCTL_NODE 145uint32_t dummy_def = IPFW_DEFAULT_RULE; 146uint32_t dummy_tables_max = IPFW_TABLES_MAX; 147 148SYSBEGIN(f3) 149 150SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 151SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 152 CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0, 153 "Only do a single pass through ipfw when using dummynet(4)"); 154SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, 155 CTLFLAG_RW, &VNET_NAME(autoinc_step), 0, 156 "Rule number auto-increment step"); 157SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose, 158 CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0, 159 "Log matches to ipfw rules"); 160SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, 161 CTLFLAG_RW, &VNET_NAME(verbose_limit), 0, 162 "Set upper limit of matches of ipfw rules logged"); 163SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, 164 &dummy_def, 0, 165 "The default/max possible rule number."); 166SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, tables_max, CTLFLAG_RD, 167 &dummy_tables_max, 0, 168 "The maximum number of tables."); 169SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN, 170 &default_to_accept, 0, 171 "Make the default rule accept all packets."); 172TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept); 173SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, static_count, 174 CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0, 175 "Number of static rules"); 176 177#ifdef INET6 178SYSCTL_DECL(_net_inet6_ip6); 179SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 180SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs, 181 CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0, 182 "Deny packets with unknown IPv6 Extension Headers"); 183#endif /* INET6 */ 184 185SYSEND 186 187#endif /* SYSCTL_NODE */ 188 189 190/* 191 * Some macros used in the various matching options. 192 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 193 * Other macros just cast void * into the appropriate type 194 */ 195#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 196#define TCP(p) ((struct tcphdr *)(p)) 197#define SCTP(p) ((struct sctphdr *)(p)) 198#define UDP(p) ((struct udphdr *)(p)) 199#define ICMP(p) ((struct icmphdr *)(p)) 200#define ICMP6(p) ((struct icmp6_hdr *)(p)) 201 202static __inline int 203icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 204{ 205 int type = icmp->icmp_type; 206 207 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 208} 209 210#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 211 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 212 213static int 214is_icmp_query(struct icmphdr *icmp) 215{ 216 int type = icmp->icmp_type; 217 218 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 219} 220#undef TT 221 222/* 223 * The following checks use two arrays of 8 or 16 bits to store the 224 * bits that we want set or clear, respectively. They are in the 225 * low and high half of cmd->arg1 or cmd->d[0]. 226 * 227 * We scan options and store the bits we find set. We succeed if 228 * 229 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 230 * 231 * The code is sometimes optimized not to store additional variables. 232 */ 233 234static int 235flags_match(ipfw_insn *cmd, u_int8_t bits) 236{ 237 u_char want_clear; 238 bits = ~bits; 239 240 if ( ((cmd->arg1 & 0xff) & bits) != 0) 241 return 0; /* some bits we want set were clear */ 242 want_clear = (cmd->arg1 >> 8) & 0xff; 243 if ( (want_clear & bits) != want_clear) 244 return 0; /* some bits we want clear were set */ 245 return 1; 246} 247 248static int 249ipopts_match(struct ip *ip, ipfw_insn *cmd) 250{ 251 int optlen, bits = 0; 252 u_char *cp = (u_char *)(ip + 1); 253 int x = (ip->ip_hl << 2) - sizeof (struct ip); 254 255 for (; x > 0; x -= optlen, cp += optlen) { 256 int opt = cp[IPOPT_OPTVAL]; 257 258 if (opt == IPOPT_EOL) 259 break; 260 if (opt == IPOPT_NOP) 261 optlen = 1; 262 else { 263 optlen = cp[IPOPT_OLEN]; 264 if (optlen <= 0 || optlen > x) 265 return 0; /* invalid or truncated */ 266 } 267 switch (opt) { 268 269 default: 270 break; 271 272 case IPOPT_LSRR: 273 bits |= IP_FW_IPOPT_LSRR; 274 break; 275 276 case IPOPT_SSRR: 277 bits |= IP_FW_IPOPT_SSRR; 278 break; 279 280 case IPOPT_RR: 281 bits |= IP_FW_IPOPT_RR; 282 break; 283 284 case IPOPT_TS: 285 bits |= IP_FW_IPOPT_TS; 286 break; 287 } 288 } 289 return (flags_match(cmd, bits)); 290} 291 292static int 293tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 294{ 295 int optlen, bits = 0; 296 u_char *cp = (u_char *)(tcp + 1); 297 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 298 299 for (; x > 0; x -= optlen, cp += optlen) { 300 int opt = cp[0]; 301 if (opt == TCPOPT_EOL) 302 break; 303 if (opt == TCPOPT_NOP) 304 optlen = 1; 305 else { 306 optlen = cp[1]; 307 if (optlen <= 0) 308 break; 309 } 310 311 switch (opt) { 312 313 default: 314 break; 315 316 case TCPOPT_MAXSEG: 317 bits |= IP_FW_TCPOPT_MSS; 318 break; 319 320 case TCPOPT_WINDOW: 321 bits |= IP_FW_TCPOPT_WINDOW; 322 break; 323 324 case TCPOPT_SACK_PERMITTED: 325 case TCPOPT_SACK: 326 bits |= IP_FW_TCPOPT_SACK; 327 break; 328 329 case TCPOPT_TIMESTAMP: 330 bits |= IP_FW_TCPOPT_TS; 331 break; 332 333 } 334 } 335 return (flags_match(cmd, bits)); 336} 337 338static int 339iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 340{ 341 if (ifp == NULL) /* no iface with this packet, match fails */ 342 return 0; 343 /* Check by name or by IP address */ 344 if (cmd->name[0] != '\0') { /* match by name */ 345 /* Check name */ 346 if (cmd->p.glob) { 347 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 348 return(1); 349 } else { 350 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 351 return(1); 352 } 353 } else { 354#ifdef __FreeBSD__ /* and OSX too ? */ 355 struct ifaddr *ia; 356 357 if_addr_rlock(ifp); 358 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 359 if (ia->ifa_addr->sa_family != AF_INET) 360 continue; 361 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 362 (ia->ifa_addr))->sin_addr.s_addr) { 363 if_addr_runlock(ifp); 364 return(1); /* match */ 365 } 366 } 367 if_addr_runlock(ifp); 368#endif /* __FreeBSD__ */ 369 } 370 return(0); /* no match, fail ... */ 371} 372 373/* 374 * The verify_path function checks if a route to the src exists and 375 * if it is reachable via ifp (when provided). 376 * 377 * The 'verrevpath' option checks that the interface that an IP packet 378 * arrives on is the same interface that traffic destined for the 379 * packet's source address would be routed out of. 380 * The 'versrcreach' option just checks that the source address is 381 * reachable via any route (except default) in the routing table. 382 * These two are a measure to block forged packets. This is also 383 * commonly known as "anti-spoofing" or Unicast Reverse Path 384 * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 385 * is purposely reminiscent of the Cisco IOS command, 386 * 387 * ip verify unicast reverse-path 388 * ip verify unicast source reachable-via any 389 * 390 * which implements the same functionality. But note that the syntax 391 * is misleading, and the check may be performed on all IP packets 392 * whether unicast, multicast, or broadcast. 393 */ 394static int 395verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) 396{ 397#ifndef __FreeBSD__ 398 return 0; 399#else 400 struct route ro; 401 struct sockaddr_in *dst; 402 403 bzero(&ro, sizeof(ro)); 404 405 dst = (struct sockaddr_in *)&(ro.ro_dst); 406 dst->sin_family = AF_INET; 407 dst->sin_len = sizeof(*dst); 408 dst->sin_addr = src; 409 in_rtalloc_ign(&ro, 0, fib); 410 411 if (ro.ro_rt == NULL) 412 return 0; 413 414 /* 415 * If ifp is provided, check for equality with rtentry. 416 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 417 * in order to pass packets injected back by if_simloop(): 418 * if useloopback == 1 routing entry (via lo0) for our own address 419 * may exist, so we need to handle routing assymetry. 420 */ 421 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 422 RTFREE(ro.ro_rt); 423 return 0; 424 } 425 426 /* if no ifp provided, check if rtentry is not default route */ 427 if (ifp == NULL && 428 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 429 RTFREE(ro.ro_rt); 430 return 0; 431 } 432 433 /* or if this is a blackhole/reject route */ 434 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 435 RTFREE(ro.ro_rt); 436 return 0; 437 } 438 439 /* found valid route */ 440 RTFREE(ro.ro_rt); 441 return 1; 442#endif /* __FreeBSD__ */ 443} 444 445#ifdef INET6 446/* 447 * ipv6 specific rules here... 448 */ 449static __inline int 450icmp6type_match (int type, ipfw_insn_u32 *cmd) 451{ 452 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 453} 454 455static int 456flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 457{ 458 int i; 459 for (i=0; i <= cmd->o.arg1; ++i ) 460 if (curr_flow == cmd->d[i] ) 461 return 1; 462 return 0; 463} 464 465/* support for IP6_*_ME opcodes */ 466static int 467search_ip6_addr_net (struct in6_addr * ip6_addr) 468{ 469 struct ifnet *mdc; 470 struct ifaddr *mdc2; 471 struct in6_ifaddr *fdm; 472 struct in6_addr copia; 473 474 TAILQ_FOREACH(mdc, &V_ifnet, if_link) { 475 if_addr_rlock(mdc); 476 TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) { 477 if (mdc2->ifa_addr->sa_family == AF_INET6) { 478 fdm = (struct in6_ifaddr *)mdc2; 479 copia = fdm->ia_addr.sin6_addr; 480 /* need for leaving scope_id in the sock_addr */ 481 in6_clearscope(&copia); 482 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) { 483 if_addr_runlock(mdc); 484 return 1; 485 } 486 } 487 } 488 if_addr_runlock(mdc); 489 } 490 return 0; 491} 492 493static int 494verify_path6(struct in6_addr *src, struct ifnet *ifp) 495{ 496 struct route_in6 ro; 497 struct sockaddr_in6 *dst; 498 499 bzero(&ro, sizeof(ro)); 500 501 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 502 dst->sin6_family = AF_INET6; 503 dst->sin6_len = sizeof(*dst); 504 dst->sin6_addr = *src; 505 /* XXX MRT 0 for ipv6 at this time */ 506 rtalloc_ign((struct route *)&ro, 0); 507 508 if (ro.ro_rt == NULL) 509 return 0; 510 511 /* 512 * if ifp is provided, check for equality with rtentry 513 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 514 * to support the case of sending packets to an address of our own. 515 * (where the former interface is the first argument of if_simloop() 516 * (=ifp), the latter is lo0) 517 */ 518 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 519 RTFREE(ro.ro_rt); 520 return 0; 521 } 522 523 /* if no ifp provided, check if rtentry is not default route */ 524 if (ifp == NULL && 525 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 526 RTFREE(ro.ro_rt); 527 return 0; 528 } 529 530 /* or if this is a blackhole/reject route */ 531 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 532 RTFREE(ro.ro_rt); 533 return 0; 534 } 535 536 /* found valid route */ 537 RTFREE(ro.ro_rt); 538 return 1; 539 540} 541 542static int 543is_icmp6_query(int icmp6_type) 544{ 545 if ((icmp6_type <= ICMP6_MAXTYPE) && 546 (icmp6_type == ICMP6_ECHO_REQUEST || 547 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 548 icmp6_type == ICMP6_WRUREQUEST || 549 icmp6_type == ICMP6_FQDN_QUERY || 550 icmp6_type == ICMP6_NI_QUERY)) 551 return (1); 552 553 return (0); 554} 555 556static void 557send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 558{ 559 struct mbuf *m; 560 561 m = args->m; 562 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 563 struct tcphdr *tcp; 564 tcp = (struct tcphdr *)((char *)ip6 + hlen); 565 566 if ((tcp->th_flags & TH_RST) == 0) { 567 struct mbuf *m0; 568 m0 = ipfw_send_pkt(args->m, &(args->f_id), 569 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 570 tcp->th_flags | TH_RST); 571 if (m0 != NULL) 572 ip6_output(m0, NULL, NULL, 0, NULL, NULL, 573 NULL); 574 } 575 FREE_PKT(m); 576 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 577#if 0 578 /* 579 * Unlike above, the mbufs need to line up with the ip6 hdr, 580 * as the contents are read. We need to m_adj() the 581 * needed amount. 582 * The mbuf will however be thrown away so we can adjust it. 583 * Remember we did an m_pullup on it already so we 584 * can make some assumptions about contiguousness. 585 */ 586 if (args->L3offset) 587 m_adj(m, args->L3offset); 588#endif 589 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 590 } else 591 FREE_PKT(m); 592 593 args->m = NULL; 594} 595 596#endif /* INET6 */ 597 598 599/* 600 * sends a reject message, consuming the mbuf passed as an argument. 601 */ 602static void 603send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip) 604{ 605 606#if 0 607 /* XXX When ip is not guaranteed to be at mtod() we will 608 * need to account for this */ 609 * The mbuf will however be thrown away so we can adjust it. 610 * Remember we did an m_pullup on it already so we 611 * can make some assumptions about contiguousness. 612 */ 613 if (args->L3offset) 614 m_adj(m, args->L3offset); 615#endif 616 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 617 /* We need the IP header in host order for icmp_error(). */ 618 SET_HOST_IPLEN(ip); 619 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 620 } else if (args->f_id.proto == IPPROTO_TCP) { 621 struct tcphdr *const tcp = 622 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 623 if ( (tcp->th_flags & TH_RST) == 0) { 624 struct mbuf *m; 625 m = ipfw_send_pkt(args->m, &(args->f_id), 626 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 627 tcp->th_flags | TH_RST); 628 if (m != NULL) 629 ip_output(m, NULL, NULL, 0, NULL, NULL); 630 } 631 FREE_PKT(args->m); 632 } else 633 FREE_PKT(args->m); 634 args->m = NULL; 635} 636 637/* 638 * Support for uid/gid/jail lookup. These tests are expensive 639 * (because we may need to look into the list of active sockets) 640 * so we cache the results. ugid_lookupp is 0 if we have not 641 * yet done a lookup, 1 if we succeeded, and -1 if we tried 642 * and failed. The function always returns the match value. 643 * We could actually spare the variable and use *uc, setting 644 * it to '(void *)check_uidgid if we have no info, NULL if 645 * we tried and failed, or any other value if successful. 646 */ 647static int 648check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif, 649 struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip, 650 u_int16_t src_port, int *ugid_lookupp, 651 struct ucred **uc, struct inpcb *inp) 652{ 653#ifndef __FreeBSD__ 654 return cred_check(insn, proto, oif, 655 dst_ip, dst_port, src_ip, src_port, 656 (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb); 657#else /* FreeBSD */ 658 struct inpcbinfo *pi; 659 int wildcard; 660 struct inpcb *pcb; 661 int match; 662 663 /* 664 * Check to see if the UDP or TCP stack supplied us with 665 * the PCB. If so, rather then holding a lock and looking 666 * up the PCB, we can use the one that was supplied. 667 */ 668 if (inp && *ugid_lookupp == 0) { 669 INP_LOCK_ASSERT(inp); 670 if (inp->inp_socket != NULL) { 671 *uc = crhold(inp->inp_cred); 672 *ugid_lookupp = 1; 673 } else 674 *ugid_lookupp = -1; 675 } 676 /* 677 * If we have already been here and the packet has no 678 * PCB entry associated with it, then we can safely 679 * assume that this is a no match. 680 */ 681 if (*ugid_lookupp == -1) 682 return (0); 683 if (proto == IPPROTO_TCP) { 684 wildcard = 0; 685 pi = &V_tcbinfo; 686 } else if (proto == IPPROTO_UDP) { 687 wildcard = INPLOOKUP_WILDCARD; 688 pi = &V_udbinfo; 689 } else 690 return 0; 691 match = 0; 692 if (*ugid_lookupp == 0) { 693 INP_INFO_RLOCK(pi); 694 pcb = (oif) ? 695 in_pcblookup_hash(pi, 696 dst_ip, htons(dst_port), 697 src_ip, htons(src_port), 698 wildcard, oif) : 699 in_pcblookup_hash(pi, 700 src_ip, htons(src_port), 701 dst_ip, htons(dst_port), 702 wildcard, NULL); 703 if (pcb != NULL) { 704 *uc = crhold(pcb->inp_cred); 705 *ugid_lookupp = 1; 706 } 707 INP_INFO_RUNLOCK(pi); 708 if (*ugid_lookupp == 0) { 709 /* 710 * We tried and failed, set the variable to -1 711 * so we will not try again on this packet. 712 */ 713 *ugid_lookupp = -1; 714 return (0); 715 } 716 } 717 if (insn->o.opcode == O_UID) 718 match = ((*uc)->cr_uid == (uid_t)insn->d[0]); 719 else if (insn->o.opcode == O_GID) 720 match = groupmember((gid_t)insn->d[0], *uc); 721 else if (insn->o.opcode == O_JAIL) 722 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]); 723 return match; 724#endif /* __FreeBSD__ */ 725} 726 727/* 728 * Helper function to set args with info on the rule after the matching 729 * one. slot is precise, whereas we guess rule_id as they are 730 * assigned sequentially. 731 */ 732static inline void 733set_match(struct ip_fw_args *args, int slot, 734 struct ip_fw_chain *chain) 735{ 736 args->rule.chain_id = chain->id; 737 args->rule.slot = slot + 1; /* we use 0 as a marker */ 738 args->rule.rule_id = 1 + chain->map[slot]->id; 739 args->rule.rulenum = chain->map[slot]->rulenum; 740} 741 742/* 743 * The main check routine for the firewall. 744 * 745 * All arguments are in args so we can modify them and return them 746 * back to the caller. 747 * 748 * Parameters: 749 * 750 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 751 * Starts with the IP header. 752 * args->eh (in) Mac header if present, NULL for layer3 packet. 753 * args->L3offset Number of bytes bypassed if we came from L2. 754 * e.g. often sizeof(eh) ** NOTYET ** 755 * args->oif Outgoing interface, NULL if packet is incoming. 756 * The incoming interface is in the mbuf. (in) 757 * args->divert_rule (in/out) 758 * Skip up to the first rule past this rule number; 759 * upon return, non-zero port number for divert or tee. 760 * 761 * args->rule Pointer to the last matching rule (in/out) 762 * args->next_hop Socket we are forwarding to (out). 763 * args->f_id Addresses grabbed from the packet (out) 764 * args->rule.info a cookie depending on rule action 765 * 766 * Return value: 767 * 768 * IP_FW_PASS the packet must be accepted 769 * IP_FW_DENY the packet must be dropped 770 * IP_FW_DIVERT divert packet, port in m_tag 771 * IP_FW_TEE tee packet, port in m_tag 772 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 773 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 774 * args->rule contains the matching rule, 775 * args->rule.info has additional information. 776 * 777 */ 778int 779ipfw_chk(struct ip_fw_args *args) 780{ 781 782 /* 783 * Local variables holding state while processing a packet: 784 * 785 * IMPORTANT NOTE: to speed up the processing of rules, there 786 * are some assumption on the values of the variables, which 787 * are documented here. Should you change them, please check 788 * the implementation of the various instructions to make sure 789 * that they still work. 790 * 791 * args->eh The MAC header. It is non-null for a layer2 792 * packet, it is NULL for a layer-3 packet. 793 * **notyet** 794 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. 795 * 796 * m | args->m Pointer to the mbuf, as received from the caller. 797 * It may change if ipfw_chk() does an m_pullup, or if it 798 * consumes the packet because it calls send_reject(). 799 * XXX This has to change, so that ipfw_chk() never modifies 800 * or consumes the buffer. 801 * ip is the beginning of the ip(4 or 6) header. 802 * Calculated by adding the L3offset to the start of data. 803 * (Until we start using L3offset, the packet is 804 * supposed to start with the ip header). 805 */ 806 struct mbuf *m = args->m; 807 struct ip *ip = mtod(m, struct ip *); 808 809 /* 810 * For rules which contain uid/gid or jail constraints, cache 811 * a copy of the users credentials after the pcb lookup has been 812 * executed. This will speed up the processing of rules with 813 * these types of constraints, as well as decrease contention 814 * on pcb related locks. 815 */ 816#ifndef __FreeBSD__ 817 struct bsd_ucred ucred_cache; 818#else 819 struct ucred *ucred_cache = NULL; 820#endif 821 int ucred_lookup = 0; 822 823 /* 824 * oif | args->oif If NULL, ipfw_chk has been called on the 825 * inbound path (ether_input, ip_input). 826 * If non-NULL, ipfw_chk has been called on the outbound path 827 * (ether_output, ip_output). 828 */ 829 struct ifnet *oif = args->oif; 830 831 int f_pos = 0; /* index of current rule in the array */ 832 int retval = 0; 833 834 /* 835 * hlen The length of the IP header. 836 */ 837 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 838 839 /* 840 * offset The offset of a fragment. offset != 0 means that 841 * we have a fragment at this offset of an IPv4 packet. 842 * offset == 0 means that (if this is an IPv4 packet) 843 * this is the first or only fragment. 844 * For IPv6 offset == 0 means there is no Fragment Header. 845 * If offset != 0 for IPv6 always use correct mask to 846 * get the correct offset because we add IP6F_MORE_FRAG 847 * to be able to dectect the first fragment which would 848 * otherwise have offset = 0. 849 */ 850 u_short offset = 0; 851 852 /* 853 * Local copies of addresses. They are only valid if we have 854 * an IP packet. 855 * 856 * proto The protocol. Set to 0 for non-ip packets, 857 * or to the protocol read from the packet otherwise. 858 * proto != 0 means that we have an IPv4 packet. 859 * 860 * src_port, dst_port port numbers, in HOST format. Only 861 * valid for TCP and UDP packets. 862 * 863 * src_ip, dst_ip ip addresses, in NETWORK format. 864 * Only valid for IPv4 packets. 865 */ 866 uint8_t proto; 867 uint16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 868 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 869 uint16_t iplen=0; 870 int pktlen; 871 uint16_t etype = 0; /* Host order stored ether type */ 872 873 /* 874 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 875 * MATCH_NONE when checked and not matched (q = NULL), 876 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 877 */ 878 int dyn_dir = MATCH_UNKNOWN; 879 ipfw_dyn_rule *q = NULL; 880 struct ip_fw_chain *chain = &V_layer3_chain; 881 882 /* 883 * We store in ulp a pointer to the upper layer protocol header. 884 * In the ipv4 case this is easy to determine from the header, 885 * but for ipv6 we might have some additional headers in the middle. 886 * ulp is NULL if not found. 887 */ 888 void *ulp = NULL; /* upper layer protocol pointer. */ 889 890 /* XXX ipv6 variables */ 891 int is_ipv6 = 0; 892 uint8_t icmp6_type = 0; 893 uint16_t ext_hd = 0; /* bits vector for extension header filtering */ 894 /* end of ipv6 variables */ 895 896 int is_ipv4 = 0; 897 898 int done = 0; /* flag to exit the outer loop */ 899 900 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready)) 901 return (IP_FW_PASS); /* accept */ 902 903 dst_ip.s_addr = 0; /* make sure it is initialized */ 904 src_ip.s_addr = 0; /* make sure it is initialized */ 905 pktlen = m->m_pkthdr.len; 906 args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */ 907 proto = args->f_id.proto = 0; /* mark f_id invalid */ 908 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 909 910/* 911 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 912 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 913 * pointer might become stale after other pullups (but we never use it 914 * this way). 915 */ 916#define PULLUP_TO(_len, p, T) \ 917do { \ 918 int x = (_len) + sizeof(T); \ 919 if ((m)->m_len < x) { \ 920 args->m = m = m_pullup(m, x); \ 921 if (m == NULL) \ 922 goto pullup_failed; \ 923 } \ 924 p = (mtod(m, char *) + (_len)); \ 925} while (0) 926 927 /* 928 * if we have an ether header, 929 */ 930 if (args->eh) 931 etype = ntohs(args->eh->ether_type); 932 933 /* Identify IP packets and fill up variables. */ 934 if (pktlen >= sizeof(struct ip6_hdr) && 935 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { 936 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 937 is_ipv6 = 1; 938 args->f_id.addr_type = 6; 939 hlen = sizeof(struct ip6_hdr); 940 proto = ip6->ip6_nxt; 941 942 /* Search extension headers to find upper layer protocols */ 943 while (ulp == NULL) { 944 switch (proto) { 945 case IPPROTO_ICMPV6: 946 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 947 icmp6_type = ICMP6(ulp)->icmp6_type; 948 break; 949 950 case IPPROTO_TCP: 951 PULLUP_TO(hlen, ulp, struct tcphdr); 952 dst_port = TCP(ulp)->th_dport; 953 src_port = TCP(ulp)->th_sport; 954 /* save flags for dynamic rules */ 955 args->f_id._flags = TCP(ulp)->th_flags; 956 break; 957 958 case IPPROTO_SCTP: 959 PULLUP_TO(hlen, ulp, struct sctphdr); 960 src_port = SCTP(ulp)->src_port; 961 dst_port = SCTP(ulp)->dest_port; 962 break; 963 964 case IPPROTO_UDP: 965 PULLUP_TO(hlen, ulp, struct udphdr); 966 dst_port = UDP(ulp)->uh_dport; 967 src_port = UDP(ulp)->uh_sport; 968 break; 969 970 case IPPROTO_HOPOPTS: /* RFC 2460 */ 971 PULLUP_TO(hlen, ulp, struct ip6_hbh); 972 ext_hd |= EXT_HOPOPTS; 973 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 974 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 975 ulp = NULL; 976 break; 977 978 case IPPROTO_ROUTING: /* RFC 2460 */ 979 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 980 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 981 case 0: 982 ext_hd |= EXT_RTHDR0; 983 break; 984 case 2: 985 ext_hd |= EXT_RTHDR2; 986 break; 987 default: 988 printf("IPFW2: IPV6 - Unknown Routing " 989 "Header type(%d)\n", 990 ((struct ip6_rthdr *)ulp)->ip6r_type); 991 if (V_fw_deny_unknown_exthdrs) 992 return (IP_FW_DENY); 993 break; 994 } 995 ext_hd |= EXT_ROUTING; 996 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 997 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 998 ulp = NULL; 999 break; 1000 1001 case IPPROTO_FRAGMENT: /* RFC 2460 */ 1002 PULLUP_TO(hlen, ulp, struct ip6_frag); 1003 ext_hd |= EXT_FRAGMENT; 1004 hlen += sizeof (struct ip6_frag); 1005 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 1006 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 1007 IP6F_OFF_MASK; 1008 /* Add IP6F_MORE_FRAG for offset of first 1009 * fragment to be != 0. */ 1010 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 1011 IP6F_MORE_FRAG; 1012 if (offset == 0) { 1013 printf("IPFW2: IPV6 - Invalid Fragment " 1014 "Header\n"); 1015 if (V_fw_deny_unknown_exthdrs) 1016 return (IP_FW_DENY); 1017 break; 1018 } 1019 args->f_id.extra = 1020 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 1021 ulp = NULL; 1022 break; 1023 1024 case IPPROTO_DSTOPTS: /* RFC 2460 */ 1025 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1026 ext_hd |= EXT_DSTOPTS; 1027 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1028 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1029 ulp = NULL; 1030 break; 1031 1032 case IPPROTO_AH: /* RFC 2402 */ 1033 PULLUP_TO(hlen, ulp, struct ip6_ext); 1034 ext_hd |= EXT_AH; 1035 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 1036 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 1037 ulp = NULL; 1038 break; 1039 1040 case IPPROTO_ESP: /* RFC 2406 */ 1041 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 1042 /* Anything past Seq# is variable length and 1043 * data past this ext. header is encrypted. */ 1044 ext_hd |= EXT_ESP; 1045 break; 1046 1047 case IPPROTO_NONE: /* RFC 2460 */ 1048 /* 1049 * Packet ends here, and IPv6 header has 1050 * already been pulled up. If ip6e_len!=0 1051 * then octets must be ignored. 1052 */ 1053 ulp = ip; /* non-NULL to get out of loop. */ 1054 break; 1055 1056 case IPPROTO_OSPFIGP: 1057 /* XXX OSPF header check? */ 1058 PULLUP_TO(hlen, ulp, struct ip6_ext); 1059 break; 1060 1061 case IPPROTO_PIM: 1062 /* XXX PIM header check? */ 1063 PULLUP_TO(hlen, ulp, struct pim); 1064 break; 1065 1066 case IPPROTO_CARP: 1067 PULLUP_TO(hlen, ulp, struct carp_header); 1068 if (((struct carp_header *)ulp)->carp_version != 1069 CARP_VERSION) 1070 return (IP_FW_DENY); 1071 if (((struct carp_header *)ulp)->carp_type != 1072 CARP_ADVERTISEMENT) 1073 return (IP_FW_DENY); 1074 break; 1075 1076 case IPPROTO_IPV6: /* RFC 2893 */ 1077 PULLUP_TO(hlen, ulp, struct ip6_hdr); 1078 break; 1079 1080 case IPPROTO_IPV4: /* RFC 2893 */ 1081 PULLUP_TO(hlen, ulp, struct ip); 1082 break; 1083 1084 default: 1085 printf("IPFW2: IPV6 - Unknown Extension " 1086 "Header(%d), ext_hd=%x\n", proto, ext_hd); 1087 if (V_fw_deny_unknown_exthdrs) 1088 return (IP_FW_DENY); 1089 PULLUP_TO(hlen, ulp, struct ip6_ext); 1090 break; 1091 } /*switch */ 1092 } 1093 ip = mtod(m, struct ip *); 1094 ip6 = (struct ip6_hdr *)ip; 1095 args->f_id.src_ip6 = ip6->ip6_src; 1096 args->f_id.dst_ip6 = ip6->ip6_dst; 1097 args->f_id.src_ip = 0; 1098 args->f_id.dst_ip = 0; 1099 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 1100 } else if (pktlen >= sizeof(struct ip) && 1101 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { 1102 is_ipv4 = 1; 1103 hlen = ip->ip_hl << 2; 1104 args->f_id.addr_type = 4; 1105 1106 /* 1107 * Collect parameters into local variables for faster matching. 1108 */ 1109 proto = ip->ip_p; 1110 src_ip = ip->ip_src; 1111 dst_ip = ip->ip_dst; 1112 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1113 iplen = ntohs(ip->ip_len); 1114 pktlen = iplen < pktlen ? iplen : pktlen; 1115 1116 if (offset == 0) { 1117 switch (proto) { 1118 case IPPROTO_TCP: 1119 PULLUP_TO(hlen, ulp, struct tcphdr); 1120 dst_port = TCP(ulp)->th_dport; 1121 src_port = TCP(ulp)->th_sport; 1122 /* save flags for dynamic rules */ 1123 args->f_id._flags = TCP(ulp)->th_flags; 1124 break; 1125 1126 case IPPROTO_UDP: 1127 PULLUP_TO(hlen, ulp, struct udphdr); 1128 dst_port = UDP(ulp)->uh_dport; 1129 src_port = UDP(ulp)->uh_sport; 1130 break; 1131 1132 case IPPROTO_ICMP: 1133 PULLUP_TO(hlen, ulp, struct icmphdr); 1134 //args->f_id.flags = ICMP(ulp)->icmp_type; 1135 break; 1136 1137 default: 1138 break; 1139 } 1140 } 1141 1142 ip = mtod(m, struct ip *); 1143 args->f_id.src_ip = ntohl(src_ip.s_addr); 1144 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1145 } 1146#undef PULLUP_TO 1147 if (proto) { /* we may have port numbers, store them */ 1148 args->f_id.proto = proto; 1149 args->f_id.src_port = src_port = ntohs(src_port); 1150 args->f_id.dst_port = dst_port = ntohs(dst_port); 1151 } 1152 1153 IPFW_RLOCK(chain); 1154 if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */ 1155 IPFW_RUNLOCK(chain); 1156 return (IP_FW_PASS); /* accept */ 1157 } 1158 if (args->rule.slot) { 1159 /* 1160 * Packet has already been tagged as a result of a previous 1161 * match on rule args->rule aka args->rule_id (PIPE, QUEUE, 1162 * REASS, NETGRAPH, DIVERT/TEE...) 1163 * Validate the slot and continue from the next one 1164 * if still present, otherwise do a lookup. 1165 */ 1166 f_pos = (args->rule.chain_id == chain->id) ? 1167 args->rule.slot : 1168 ipfw_find_rule(chain, args->rule.rulenum, 1169 args->rule.rule_id); 1170 } else { 1171 f_pos = 0; 1172 } 1173 1174 /* 1175 * Now scan the rules, and parse microinstructions for each rule. 1176 * We have two nested loops and an inner switch. Sometimes we 1177 * need to break out of one or both loops, or re-enter one of 1178 * the loops with updated variables. Loop variables are: 1179 * 1180 * f_pos (outer loop) points to the current rule. 1181 * On output it points to the matching rule. 1182 * done (outer loop) is used as a flag to break the loop. 1183 * l (inner loop) residual length of current rule. 1184 * cmd points to the current microinstruction. 1185 * 1186 * We break the inner loop by setting l=0 and possibly 1187 * cmdlen=0 if we don't want to advance cmd. 1188 * We break the outer loop by setting done=1 1189 * We can restart the inner loop by setting l>0 and f_pos, f, cmd 1190 * as needed. 1191 */ 1192 for (; f_pos < chain->n_rules; f_pos++) { 1193 ipfw_insn *cmd; 1194 uint32_t tablearg = 0; 1195 int l, cmdlen, skip_or; /* skip rest of OR block */ 1196 struct ip_fw *f; 1197 1198 f = chain->map[f_pos]; 1199 if (V_set_disable & (1 << f->set) ) 1200 continue; 1201 1202 skip_or = 0; 1203 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1204 l -= cmdlen, cmd += cmdlen) { 1205 int match; 1206 1207 /* 1208 * check_body is a jump target used when we find a 1209 * CHECK_STATE, and need to jump to the body of 1210 * the target rule. 1211 */ 1212 1213/* check_body: */ 1214 cmdlen = F_LEN(cmd); 1215 /* 1216 * An OR block (insn_1 || .. || insn_n) has the 1217 * F_OR bit set in all but the last instruction. 1218 * The first match will set "skip_or", and cause 1219 * the following instructions to be skipped until 1220 * past the one with the F_OR bit clear. 1221 */ 1222 if (skip_or) { /* skip this instruction */ 1223 if ((cmd->len & F_OR) == 0) 1224 skip_or = 0; /* next one is good */ 1225 continue; 1226 } 1227 match = 0; /* set to 1 if we succeed */ 1228 1229 switch (cmd->opcode) { 1230 /* 1231 * The first set of opcodes compares the packet's 1232 * fields with some pattern, setting 'match' if a 1233 * match is found. At the end of the loop there is 1234 * logic to deal with F_NOT and F_OR flags associated 1235 * with the opcode. 1236 */ 1237 case O_NOP: 1238 match = 1; 1239 break; 1240 1241 case O_FORWARD_MAC: 1242 printf("ipfw: opcode %d unimplemented\n", 1243 cmd->opcode); 1244 break; 1245 1246 case O_GID: 1247 case O_UID: 1248 case O_JAIL: 1249 /* 1250 * We only check offset == 0 && proto != 0, 1251 * as this ensures that we have a 1252 * packet with the ports info. 1253 */ 1254 if (offset!=0) 1255 break; 1256 if (is_ipv6) /* XXX to be fixed later */ 1257 break; 1258 if (proto == IPPROTO_TCP || 1259 proto == IPPROTO_UDP) 1260 match = check_uidgid( 1261 (ipfw_insn_u32 *)cmd, 1262 proto, oif, 1263 dst_ip, dst_port, 1264 src_ip, src_port, &ucred_lookup, 1265#ifdef __FreeBSD__ 1266 &ucred_cache, args->inp); 1267#else 1268 (void *)&ucred_cache, 1269 (struct inpcb *)args->m); 1270#endif 1271 break; 1272 1273 case O_RECV: 1274 match = iface_match(m->m_pkthdr.rcvif, 1275 (ipfw_insn_if *)cmd); 1276 break; 1277 1278 case O_XMIT: 1279 match = iface_match(oif, (ipfw_insn_if *)cmd); 1280 break; 1281 1282 case O_VIA: 1283 match = iface_match(oif ? oif : 1284 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 1285 break; 1286 1287 case O_MACADDR2: 1288 if (args->eh != NULL) { /* have MAC header */ 1289 u_int32_t *want = (u_int32_t *) 1290 ((ipfw_insn_mac *)cmd)->addr; 1291 u_int32_t *mask = (u_int32_t *) 1292 ((ipfw_insn_mac *)cmd)->mask; 1293 u_int32_t *hdr = (u_int32_t *)args->eh; 1294 1295 match = 1296 ( want[0] == (hdr[0] & mask[0]) && 1297 want[1] == (hdr[1] & mask[1]) && 1298 want[2] == (hdr[2] & mask[2]) ); 1299 } 1300 break; 1301 1302 case O_MAC_TYPE: 1303 if (args->eh != NULL) { 1304 u_int16_t *p = 1305 ((ipfw_insn_u16 *)cmd)->ports; 1306 int i; 1307 1308 for (i = cmdlen - 1; !match && i>0; 1309 i--, p += 2) 1310 match = (etype >= p[0] && 1311 etype <= p[1]); 1312 } 1313 break; 1314 1315 case O_FRAG: 1316 match = (offset != 0); 1317 break; 1318 1319 case O_IN: /* "out" is "not in" */ 1320 match = (oif == NULL); 1321 break; 1322 1323 case O_LAYER2: 1324 match = (args->eh != NULL); 1325 break; 1326 1327 case O_DIVERTED: 1328 { 1329 /* For diverted packets, args->rule.info 1330 * contains the divert port (in host format) 1331 * reason and direction. 1332 */ 1333 uint32_t i = args->rule.info; 1334 match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT && 1335 cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2); 1336 } 1337 break; 1338 1339 case O_PROTO: 1340 /* 1341 * We do not allow an arg of 0 so the 1342 * check of "proto" only suffices. 1343 */ 1344 match = (proto == cmd->arg1); 1345 break; 1346 1347 case O_IP_SRC: 1348 match = is_ipv4 && 1349 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1350 src_ip.s_addr); 1351 break; 1352 1353 case O_IP_SRC_LOOKUP: 1354 case O_IP_DST_LOOKUP: 1355 if (is_ipv4) { 1356 uint32_t key = 1357 (cmd->opcode == O_IP_DST_LOOKUP) ? 1358 dst_ip.s_addr : src_ip.s_addr; 1359 uint32_t v = 0; 1360 1361 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) { 1362 /* generic lookup. The key must be 1363 * in 32bit big-endian format. 1364 */ 1365 v = ((ipfw_insn_u32 *)cmd)->d[1]; 1366 if (v == 0) 1367 key = dst_ip.s_addr; 1368 else if (v == 1) 1369 key = src_ip.s_addr; 1370 else if (v == 6) /* dscp */ 1371 key = (ip->ip_tos >> 2) & 0x3f; 1372 else if (offset != 0) 1373 break; 1374 else if (proto != IPPROTO_TCP && 1375 proto != IPPROTO_UDP) 1376 break; 1377 else if (v == 2) 1378 key = htonl(dst_port); 1379 else if (v == 3) 1380 key = htonl(src_port); 1381 else if (v == 4 || v == 5) { 1382 check_uidgid( 1383 (ipfw_insn_u32 *)cmd, 1384 proto, oif, 1385 dst_ip, dst_port, 1386 src_ip, src_port, &ucred_lookup, 1387#ifdef __FreeBSD__ 1388 &ucred_cache, args->inp); 1389 if (v == 4 /* O_UID */) 1390 key = ucred_cache->cr_uid; 1391 else if (v == 5 /* O_JAIL */) 1392 key = ucred_cache->cr_prison->pr_id; 1393#else /* !__FreeBSD__ */ 1394 (void *)&ucred_cache, 1395 (struct inpcb *)args->m); 1396 if (v ==4 /* O_UID */) 1397 key = ucred_cache.uid; 1398 else if (v == 5 /* O_JAIL */) 1399 key = ucred_cache.xid; 1400#endif /* !__FreeBSD__ */ 1401 key = htonl(key); 1402 } else 1403 break; 1404 } 1405 match = ipfw_lookup_table(chain, 1406 cmd->arg1, key, &v); 1407 if (!match) 1408 break; 1409 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 1410 match = 1411 ((ipfw_insn_u32 *)cmd)->d[0] == v; 1412 else 1413 tablearg = v; 1414 } 1415 break; 1416 1417 case O_IP_SRC_MASK: 1418 case O_IP_DST_MASK: 1419 if (is_ipv4) { 1420 uint32_t a = 1421 (cmd->opcode == O_IP_DST_MASK) ? 1422 dst_ip.s_addr : src_ip.s_addr; 1423 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 1424 int i = cmdlen-1; 1425 1426 for (; !match && i>0; i-= 2, p+= 2) 1427 match = (p[0] == (a & p[1])); 1428 } 1429 break; 1430 1431 case O_IP_SRC_ME: 1432 if (is_ipv4) { 1433 struct ifnet *tif; 1434 1435 INADDR_TO_IFP(src_ip, tif); 1436 match = (tif != NULL); 1437 break; 1438 } 1439#ifdef INET6 1440 /* FALLTHROUGH */ 1441 case O_IP6_SRC_ME: 1442 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 1443#endif 1444 break; 1445 1446 case O_IP_DST_SET: 1447 case O_IP_SRC_SET: 1448 if (is_ipv4) { 1449 u_int32_t *d = (u_int32_t *)(cmd+1); 1450 u_int32_t addr = 1451 cmd->opcode == O_IP_DST_SET ? 1452 args->f_id.dst_ip : 1453 args->f_id.src_ip; 1454 1455 if (addr < d[0]) 1456 break; 1457 addr -= d[0]; /* subtract base */ 1458 match = (addr < cmd->arg1) && 1459 ( d[ 1 + (addr>>5)] & 1460 (1<<(addr & 0x1f)) ); 1461 } 1462 break; 1463 1464 case O_IP_DST: 1465 match = is_ipv4 && 1466 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1467 dst_ip.s_addr); 1468 break; 1469 1470 case O_IP_DST_ME: 1471 if (is_ipv4) { 1472 struct ifnet *tif; 1473 1474 INADDR_TO_IFP(dst_ip, tif); 1475 match = (tif != NULL); 1476 break; 1477 } 1478#ifdef INET6 1479 /* FALLTHROUGH */ 1480 case O_IP6_DST_ME: 1481 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 1482#endif 1483 break; 1484 1485 1486 case O_IP_SRCPORT: 1487 case O_IP_DSTPORT: 1488 /* 1489 * offset == 0 && proto != 0 is enough 1490 * to guarantee that we have a 1491 * packet with port info. 1492 */ 1493 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 1494 && offset == 0) { 1495 u_int16_t x = 1496 (cmd->opcode == O_IP_SRCPORT) ? 1497 src_port : dst_port ; 1498 u_int16_t *p = 1499 ((ipfw_insn_u16 *)cmd)->ports; 1500 int i; 1501 1502 for (i = cmdlen - 1; !match && i>0; 1503 i--, p += 2) 1504 match = (x>=p[0] && x<=p[1]); 1505 } 1506 break; 1507 1508 case O_ICMPTYPE: 1509 match = (offset == 0 && proto==IPPROTO_ICMP && 1510 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 1511 break; 1512 1513#ifdef INET6 1514 case O_ICMP6TYPE: 1515 match = is_ipv6 && offset == 0 && 1516 proto==IPPROTO_ICMPV6 && 1517 icmp6type_match( 1518 ICMP6(ulp)->icmp6_type, 1519 (ipfw_insn_u32 *)cmd); 1520 break; 1521#endif /* INET6 */ 1522 1523 case O_IPOPT: 1524 match = (is_ipv4 && 1525 ipopts_match(ip, cmd) ); 1526 break; 1527 1528 case O_IPVER: 1529 match = (is_ipv4 && 1530 cmd->arg1 == ip->ip_v); 1531 break; 1532 1533 case O_IPID: 1534 case O_IPLEN: 1535 case O_IPTTL: 1536 if (is_ipv4) { /* only for IP packets */ 1537 uint16_t x; 1538 uint16_t *p; 1539 int i; 1540 1541 if (cmd->opcode == O_IPLEN) 1542 x = iplen; 1543 else if (cmd->opcode == O_IPTTL) 1544 x = ip->ip_ttl; 1545 else /* must be IPID */ 1546 x = ntohs(ip->ip_id); 1547 if (cmdlen == 1) { 1548 match = (cmd->arg1 == x); 1549 break; 1550 } 1551 /* otherwise we have ranges */ 1552 p = ((ipfw_insn_u16 *)cmd)->ports; 1553 i = cmdlen - 1; 1554 for (; !match && i>0; i--, p += 2) 1555 match = (x >= p[0] && x <= p[1]); 1556 } 1557 break; 1558 1559 case O_IPPRECEDENCE: 1560 match = (is_ipv4 && 1561 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 1562 break; 1563 1564 case O_IPTOS: 1565 match = (is_ipv4 && 1566 flags_match(cmd, ip->ip_tos)); 1567 break; 1568 1569 case O_TCPDATALEN: 1570 if (proto == IPPROTO_TCP && offset == 0) { 1571 struct tcphdr *tcp; 1572 uint16_t x; 1573 uint16_t *p; 1574 int i; 1575 1576 tcp = TCP(ulp); 1577 x = iplen - 1578 ((ip->ip_hl + tcp->th_off) << 2); 1579 if (cmdlen == 1) { 1580 match = (cmd->arg1 == x); 1581 break; 1582 } 1583 /* otherwise we have ranges */ 1584 p = ((ipfw_insn_u16 *)cmd)->ports; 1585 i = cmdlen - 1; 1586 for (; !match && i>0; i--, p += 2) 1587 match = (x >= p[0] && x <= p[1]); 1588 } 1589 break; 1590 1591 case O_TCPFLAGS: 1592 match = (proto == IPPROTO_TCP && offset == 0 && 1593 flags_match(cmd, TCP(ulp)->th_flags)); 1594 break; 1595 1596 case O_TCPOPTS: 1597 match = (proto == IPPROTO_TCP && offset == 0 && 1598 tcpopts_match(TCP(ulp), cmd)); 1599 break; 1600 1601 case O_TCPSEQ: 1602 match = (proto == IPPROTO_TCP && offset == 0 && 1603 ((ipfw_insn_u32 *)cmd)->d[0] == 1604 TCP(ulp)->th_seq); 1605 break; 1606 1607 case O_TCPACK: 1608 match = (proto == IPPROTO_TCP && offset == 0 && 1609 ((ipfw_insn_u32 *)cmd)->d[0] == 1610 TCP(ulp)->th_ack); 1611 break; 1612 1613 case O_TCPWIN: 1614 match = (proto == IPPROTO_TCP && offset == 0 && 1615 cmd->arg1 == TCP(ulp)->th_win); 1616 break; 1617 1618 case O_ESTAB: 1619 /* reject packets which have SYN only */ 1620 /* XXX should i also check for TH_ACK ? */ 1621 match = (proto == IPPROTO_TCP && offset == 0 && 1622 (TCP(ulp)->th_flags & 1623 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 1624 break; 1625 1626 case O_ALTQ: { 1627 struct pf_mtag *at; 1628 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 1629 1630 match = 1; 1631 at = pf_find_mtag(m); 1632 if (at != NULL && at->qid != 0) 1633 break; 1634 at = pf_get_mtag(m); 1635 if (at == NULL) { 1636 /* 1637 * Let the packet fall back to the 1638 * default ALTQ. 1639 */ 1640 break; 1641 } 1642 at->qid = altq->qid; 1643 if (is_ipv4) 1644 at->af = AF_INET; 1645 else 1646 at->af = AF_LINK; 1647 at->hdr = ip; 1648 break; 1649 } 1650 1651 case O_LOG: 1652 ipfw_log(f, hlen, args, m, 1653 oif, offset, tablearg, ip); 1654 match = 1; 1655 break; 1656 1657 case O_PROB: 1658 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 1659 break; 1660 1661 case O_VERREVPATH: 1662 /* Outgoing packets automatically pass/match */ 1663 match = ((oif != NULL) || 1664 (m->m_pkthdr.rcvif == NULL) || 1665 ( 1666#ifdef INET6 1667 is_ipv6 ? 1668 verify_path6(&(args->f_id.src_ip6), 1669 m->m_pkthdr.rcvif) : 1670#endif 1671 verify_path(src_ip, m->m_pkthdr.rcvif, 1672 args->f_id.fib))); 1673 break; 1674 1675 case O_VERSRCREACH: 1676 /* Outgoing packets automatically pass/match */ 1677 match = (hlen > 0 && ((oif != NULL) || 1678#ifdef INET6 1679 is_ipv6 ? 1680 verify_path6(&(args->f_id.src_ip6), 1681 NULL) : 1682#endif 1683 verify_path(src_ip, NULL, args->f_id.fib))); 1684 break; 1685 1686 case O_ANTISPOOF: 1687 /* Outgoing packets automatically pass/match */ 1688 if (oif == NULL && hlen > 0 && 1689 ( (is_ipv4 && in_localaddr(src_ip)) 1690#ifdef INET6 1691 || (is_ipv6 && 1692 in6_localaddr(&(args->f_id.src_ip6))) 1693#endif 1694 )) 1695 match = 1696#ifdef INET6 1697 is_ipv6 ? verify_path6( 1698 &(args->f_id.src_ip6), 1699 m->m_pkthdr.rcvif) : 1700#endif 1701 verify_path(src_ip, 1702 m->m_pkthdr.rcvif, 1703 args->f_id.fib); 1704 else 1705 match = 1; 1706 break; 1707 1708 case O_IPSEC: 1709#ifdef IPSEC 1710 match = (m_tag_find(m, 1711 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 1712#endif 1713 /* otherwise no match */ 1714 break; 1715 1716#ifdef INET6 1717 case O_IP6_SRC: 1718 match = is_ipv6 && 1719 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 1720 &((ipfw_insn_ip6 *)cmd)->addr6); 1721 break; 1722 1723 case O_IP6_DST: 1724 match = is_ipv6 && 1725 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 1726 &((ipfw_insn_ip6 *)cmd)->addr6); 1727 break; 1728 case O_IP6_SRC_MASK: 1729 case O_IP6_DST_MASK: 1730 if (is_ipv6) { 1731 int i = cmdlen - 1; 1732 struct in6_addr p; 1733 struct in6_addr *d = 1734 &((ipfw_insn_ip6 *)cmd)->addr6; 1735 1736 for (; !match && i > 0; d += 2, 1737 i -= F_INSN_SIZE(struct in6_addr) 1738 * 2) { 1739 p = (cmd->opcode == 1740 O_IP6_SRC_MASK) ? 1741 args->f_id.src_ip6: 1742 args->f_id.dst_ip6; 1743 APPLY_MASK(&p, &d[1]); 1744 match = 1745 IN6_ARE_ADDR_EQUAL(&d[0], 1746 &p); 1747 } 1748 } 1749 break; 1750 1751 case O_FLOW6ID: 1752 match = is_ipv6 && 1753 flow6id_match(args->f_id.flow_id6, 1754 (ipfw_insn_u32 *) cmd); 1755 break; 1756 1757 case O_EXT_HDR: 1758 match = is_ipv6 && 1759 (ext_hd & ((ipfw_insn *) cmd)->arg1); 1760 break; 1761 1762 case O_IP6: 1763 match = is_ipv6; 1764 break; 1765#endif 1766 1767 case O_IP4: 1768 match = is_ipv4; 1769 break; 1770 1771 case O_TAG: { 1772 struct m_tag *mtag; 1773 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 1774 tablearg : cmd->arg1; 1775 1776 /* Packet is already tagged with this tag? */ 1777 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 1778 1779 /* We have `untag' action when F_NOT flag is 1780 * present. And we must remove this mtag from 1781 * mbuf and reset `match' to zero (`match' will 1782 * be inversed later). 1783 * Otherwise we should allocate new mtag and 1784 * push it into mbuf. 1785 */ 1786 if (cmd->len & F_NOT) { /* `untag' action */ 1787 if (mtag != NULL) 1788 m_tag_delete(m, mtag); 1789 match = 0; 1790 } else if (mtag == NULL) { 1791 if ((mtag = m_tag_alloc(MTAG_IPFW, 1792 tag, 0, M_NOWAIT)) != NULL) 1793 m_tag_prepend(m, mtag); 1794 match = 1; 1795 } 1796 break; 1797 } 1798 1799 case O_FIB: /* try match the specified fib */ 1800 if (args->f_id.fib == cmd->arg1) 1801 match = 1; 1802 break; 1803 1804 case O_SOCKARG: { 1805 struct inpcb *inp = args->inp; 1806 struct inpcbinfo *pi; 1807 1808 if (is_ipv6) /* XXX can we remove this ? */ 1809 break; 1810 1811 if (proto == IPPROTO_TCP) 1812 pi = &V_tcbinfo; 1813 else if (proto == IPPROTO_UDP) 1814 pi = &V_udbinfo; 1815 else 1816 break; 1817 1818 /* For incomming packet, lookup up the 1819 inpcb using the src/dest ip/port tuple */ 1820 if (inp == NULL) { 1821 INP_INFO_RLOCK(pi); 1822 inp = in_pcblookup_hash(pi, 1823 src_ip, htons(src_port), 1824 dst_ip, htons(dst_port), 1825 0, NULL); 1826 INP_INFO_RUNLOCK(pi); 1827 } 1828 1829 if (inp && inp->inp_socket) { 1830 tablearg = inp->inp_socket->so_user_cookie; 1831 if (tablearg) 1832 match = 1; 1833 } 1834 break; 1835 } 1836 1837 case O_TAGGED: { 1838 struct m_tag *mtag; 1839 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 1840 tablearg : cmd->arg1; 1841 1842 if (cmdlen == 1) { 1843 match = m_tag_locate(m, MTAG_IPFW, 1844 tag, NULL) != NULL; 1845 break; 1846 } 1847 1848 /* we have ranges */ 1849 for (mtag = m_tag_first(m); 1850 mtag != NULL && !match; 1851 mtag = m_tag_next(m, mtag)) { 1852 uint16_t *p; 1853 int i; 1854 1855 if (mtag->m_tag_cookie != MTAG_IPFW) 1856 continue; 1857 1858 p = ((ipfw_insn_u16 *)cmd)->ports; 1859 i = cmdlen - 1; 1860 for(; !match && i > 0; i--, p += 2) 1861 match = 1862 mtag->m_tag_id >= p[0] && 1863 mtag->m_tag_id <= p[1]; 1864 } 1865 break; 1866 } 1867 1868 /* 1869 * The second set of opcodes represents 'actions', 1870 * i.e. the terminal part of a rule once the packet 1871 * matches all previous patterns. 1872 * Typically there is only one action for each rule, 1873 * and the opcode is stored at the end of the rule 1874 * (but there are exceptions -- see below). 1875 * 1876 * In general, here we set retval and terminate the 1877 * outer loop (would be a 'break 3' in some language, 1878 * but we need to set l=0, done=1) 1879 * 1880 * Exceptions: 1881 * O_COUNT and O_SKIPTO actions: 1882 * instead of terminating, we jump to the next rule 1883 * (setting l=0), or to the SKIPTO target (setting 1884 * f/f_len, cmd and l as needed), respectively. 1885 * 1886 * O_TAG, O_LOG and O_ALTQ action parameters: 1887 * perform some action and set match = 1; 1888 * 1889 * O_LIMIT and O_KEEP_STATE: these opcodes are 1890 * not real 'actions', and are stored right 1891 * before the 'action' part of the rule. 1892 * These opcodes try to install an entry in the 1893 * state tables; if successful, we continue with 1894 * the next opcode (match=1; break;), otherwise 1895 * the packet must be dropped (set retval, 1896 * break loops with l=0, done=1) 1897 * 1898 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 1899 * cause a lookup of the state table, and a jump 1900 * to the 'action' part of the parent rule 1901 * if an entry is found, or 1902 * (CHECK_STATE only) a jump to the next rule if 1903 * the entry is not found. 1904 * The result of the lookup is cached so that 1905 * further instances of these opcodes become NOPs. 1906 * The jump to the next rule is done by setting 1907 * l=0, cmdlen=0. 1908 */ 1909 case O_LIMIT: 1910 case O_KEEP_STATE: 1911 if (ipfw_install_state(f, 1912 (ipfw_insn_limit *)cmd, args, tablearg)) { 1913 /* error or limit violation */ 1914 retval = IP_FW_DENY; 1915 l = 0; /* exit inner loop */ 1916 done = 1; /* exit outer loop */ 1917 } 1918 match = 1; 1919 break; 1920 1921 case O_PROBE_STATE: 1922 case O_CHECK_STATE: 1923 /* 1924 * dynamic rules are checked at the first 1925 * keep-state or check-state occurrence, 1926 * with the result being stored in dyn_dir. 1927 * The compiler introduces a PROBE_STATE 1928 * instruction for us when we have a 1929 * KEEP_STATE (because PROBE_STATE needs 1930 * to be run first). 1931 */ 1932 if (dyn_dir == MATCH_UNKNOWN && 1933 (q = ipfw_lookup_dyn_rule(&args->f_id, 1934 &dyn_dir, proto == IPPROTO_TCP ? 1935 TCP(ulp) : NULL)) 1936 != NULL) { 1937 /* 1938 * Found dynamic entry, update stats 1939 * and jump to the 'action' part of 1940 * the parent rule by setting 1941 * f, cmd, l and clearing cmdlen. 1942 */ 1943 q->pcnt++; 1944 q->bcnt += pktlen; 1945 /* XXX we would like to have f_pos 1946 * readily accessible in the dynamic 1947 * rule, instead of having to 1948 * lookup q->rule. 1949 */ 1950 f = q->rule; 1951 f_pos = ipfw_find_rule(chain, 1952 f->rulenum, f->id); 1953 cmd = ACTION_PTR(f); 1954 l = f->cmd_len - f->act_ofs; 1955 ipfw_dyn_unlock(); 1956 cmdlen = 0; 1957 match = 1; 1958 break; 1959 } 1960 /* 1961 * Dynamic entry not found. If CHECK_STATE, 1962 * skip to next rule, if PROBE_STATE just 1963 * ignore and continue with next opcode. 1964 */ 1965 if (cmd->opcode == O_CHECK_STATE) 1966 l = 0; /* exit inner loop */ 1967 match = 1; 1968 break; 1969 1970 case O_ACCEPT: 1971 retval = 0; /* accept */ 1972 l = 0; /* exit inner loop */ 1973 done = 1; /* exit outer loop */ 1974 break; 1975 1976 case O_PIPE: 1977 case O_QUEUE: 1978 set_match(args, f_pos, chain); 1979 args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ? 1980 tablearg : cmd->arg1; 1981 if (cmd->opcode == O_PIPE) 1982 args->rule.info |= IPFW_IS_PIPE; 1983 if (V_fw_one_pass) 1984 args->rule.info |= IPFW_ONEPASS; 1985 retval = IP_FW_DUMMYNET; 1986 l = 0; /* exit inner loop */ 1987 done = 1; /* exit outer loop */ 1988 break; 1989 1990 case O_DIVERT: 1991 case O_TEE: 1992 if (args->eh) /* not on layer 2 */ 1993 break; 1994 /* otherwise this is terminal */ 1995 l = 0; /* exit inner loop */ 1996 done = 1; /* exit outer loop */ 1997 retval = (cmd->opcode == O_DIVERT) ? 1998 IP_FW_DIVERT : IP_FW_TEE; 1999 set_match(args, f_pos, chain); 2000 args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ? 2001 tablearg : cmd->arg1; 2002 break; 2003 2004 case O_COUNT: 2005 f->pcnt++; /* update stats */ 2006 f->bcnt += pktlen; 2007 f->timestamp = time_uptime; 2008 l = 0; /* exit inner loop */ 2009 break; 2010 2011 case O_SKIPTO: 2012 f->pcnt++; /* update stats */ 2013 f->bcnt += pktlen; 2014 f->timestamp = time_uptime; 2015 /* If possible use cached f_pos (in f->next_rule), 2016 * whose version is written in f->next_rule 2017 * (horrible hacks to avoid changing the ABI). 2018 */ 2019 if (cmd->arg1 != IP_FW_TABLEARG && 2020 (uintptr_t)f->x_next == chain->id) { 2021 f_pos = (uintptr_t)f->next_rule; 2022 } else { 2023 int i = (cmd->arg1 == IP_FW_TABLEARG) ? 2024 tablearg : cmd->arg1; 2025 /* make sure we do not jump backward */ 2026 if (i <= f->rulenum) 2027 i = f->rulenum + 1; 2028 f_pos = ipfw_find_rule(chain, i, 0); 2029 /* update the cache */ 2030 if (cmd->arg1 != IP_FW_TABLEARG) { 2031 f->next_rule = 2032 (void *)(uintptr_t)f_pos; 2033 f->x_next = 2034 (void *)(uintptr_t)chain->id; 2035 } 2036 } 2037 /* 2038 * Skip disabled rules, and re-enter 2039 * the inner loop with the correct 2040 * f_pos, f, l and cmd. 2041 * Also clear cmdlen and skip_or 2042 */ 2043 for (; f_pos < chain->n_rules - 1 && 2044 (V_set_disable & 2045 (1 << chain->map[f_pos]->set)); 2046 f_pos++) 2047 ; 2048 /* Re-enter the inner loop at the skipto rule. */ 2049 f = chain->map[f_pos]; 2050 l = f->cmd_len; 2051 cmd = f->cmd; 2052 match = 1; 2053 cmdlen = 0; 2054 skip_or = 0; 2055 continue; 2056 break; /* not reached */ 2057 2058 case O_REJECT: 2059 /* 2060 * Drop the packet and send a reject notice 2061 * if the packet is not ICMP (or is an ICMP 2062 * query), and it is not multicast/broadcast. 2063 */ 2064 if (hlen > 0 && is_ipv4 && offset == 0 && 2065 (proto != IPPROTO_ICMP || 2066 is_icmp_query(ICMP(ulp))) && 2067 !(m->m_flags & (M_BCAST|M_MCAST)) && 2068 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 2069 send_reject(args, cmd->arg1, iplen, ip); 2070 m = args->m; 2071 } 2072 /* FALLTHROUGH */ 2073#ifdef INET6 2074 case O_UNREACH6: 2075 if (hlen > 0 && is_ipv6 && 2076 ((offset & IP6F_OFF_MASK) == 0) && 2077 (proto != IPPROTO_ICMPV6 || 2078 (is_icmp6_query(icmp6_type) == 1)) && 2079 !(m->m_flags & (M_BCAST|M_MCAST)) && 2080 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 2081 send_reject6( 2082 args, cmd->arg1, hlen, 2083 (struct ip6_hdr *)ip); 2084 m = args->m; 2085 } 2086 /* FALLTHROUGH */ 2087#endif 2088 case O_DENY: 2089 retval = IP_FW_DENY; 2090 l = 0; /* exit inner loop */ 2091 done = 1; /* exit outer loop */ 2092 break; 2093 2094 case O_FORWARD_IP: 2095 if (args->eh) /* not valid on layer2 pkts */ 2096 break; 2097 if (!q || dyn_dir == MATCH_FORWARD) { 2098 struct sockaddr_in *sa; 2099 sa = &(((ipfw_insn_sa *)cmd)->sa); 2100 if (sa->sin_addr.s_addr == INADDR_ANY) { 2101 bcopy(sa, &args->hopstore, 2102 sizeof(*sa)); 2103 args->hopstore.sin_addr.s_addr = 2104 htonl(tablearg); 2105 args->next_hop = &args->hopstore; 2106 } else { 2107 args->next_hop = sa; 2108 } 2109 } 2110 retval = IP_FW_PASS; 2111 l = 0; /* exit inner loop */ 2112 done = 1; /* exit outer loop */ 2113 break; 2114 2115 case O_NETGRAPH: 2116 case O_NGTEE: 2117 set_match(args, f_pos, chain); 2118 args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ? 2119 tablearg : cmd->arg1; 2120 if (V_fw_one_pass) 2121 args->rule.info |= IPFW_ONEPASS; 2122 retval = (cmd->opcode == O_NETGRAPH) ? 2123 IP_FW_NETGRAPH : IP_FW_NGTEE; 2124 l = 0; /* exit inner loop */ 2125 done = 1; /* exit outer loop */ 2126 break; 2127 2128 case O_SETFIB: 2129 f->pcnt++; /* update stats */ 2130 f->bcnt += pktlen; 2131 f->timestamp = time_uptime; 2132 M_SETFIB(m, cmd->arg1); 2133 args->f_id.fib = cmd->arg1; 2134 l = 0; /* exit inner loop */ 2135 break; 2136 2137 case O_NAT: 2138 if (!IPFW_NAT_LOADED) { 2139 retval = IP_FW_DENY; 2140 } else { 2141 struct cfg_nat *t; 2142 int nat_id; 2143 2144 set_match(args, f_pos, chain); 2145 t = ((ipfw_insn_nat *)cmd)->nat; 2146 if (t == NULL) { 2147 nat_id = (cmd->arg1 == IP_FW_TABLEARG) ? 2148 tablearg : cmd->arg1; 2149 t = (*lookup_nat_ptr)(&chain->nat, nat_id); 2150 2151 if (t == NULL) { 2152 retval = IP_FW_DENY; 2153 l = 0; /* exit inner loop */ 2154 done = 1; /* exit outer loop */ 2155 break; 2156 } 2157 if (cmd->arg1 != IP_FW_TABLEARG) 2158 ((ipfw_insn_nat *)cmd)->nat = t; 2159 } 2160 retval = ipfw_nat_ptr(args, t, m); 2161 } 2162 l = 0; /* exit inner loop */ 2163 done = 1; /* exit outer loop */ 2164 break; 2165 2166 case O_REASS: { 2167 int ip_off; 2168 2169 f->pcnt++; 2170 f->bcnt += pktlen; 2171 l = 0; /* in any case exit inner loop */ 2172 ip_off = ntohs(ip->ip_off); 2173 2174 /* if not fragmented, go to next rule */ 2175 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0) 2176 break; 2177 /* 2178 * ip_reass() expects len & off in host 2179 * byte order. 2180 */ 2181 SET_HOST_IPLEN(ip); 2182 2183 args->m = m = ip_reass(m); 2184 2185 /* 2186 * do IP header checksum fixup. 2187 */ 2188 if (m == NULL) { /* fragment got swallowed */ 2189 retval = IP_FW_DENY; 2190 } else { /* good, packet complete */ 2191 int hlen; 2192 2193 ip = mtod(m, struct ip *); 2194 hlen = ip->ip_hl << 2; 2195 SET_NET_IPLEN(ip); 2196 ip->ip_sum = 0; 2197 if (hlen == sizeof(struct ip)) 2198 ip->ip_sum = in_cksum_hdr(ip); 2199 else 2200 ip->ip_sum = in_cksum(m, hlen); 2201 retval = IP_FW_REASS; 2202 set_match(args, f_pos, chain); 2203 } 2204 done = 1; /* exit outer loop */ 2205 break; 2206 } 2207 2208 default: 2209 panic("-- unknown opcode %d\n", cmd->opcode); 2210 } /* end of switch() on opcodes */ 2211 /* 2212 * if we get here with l=0, then match is irrelevant. 2213 */ 2214 2215 if (cmd->len & F_NOT) 2216 match = !match; 2217 2218 if (match) { 2219 if (cmd->len & F_OR) 2220 skip_or = 1; 2221 } else { 2222 if (!(cmd->len & F_OR)) /* not an OR block, */ 2223 break; /* try next rule */ 2224 } 2225 2226 } /* end of inner loop, scan opcodes */ 2227 2228 if (done) 2229 break; 2230 2231/* next_rule:; */ /* try next rule */ 2232 2233 } /* end of outer for, scan rules */ 2234 2235 if (done) { 2236 struct ip_fw *rule = chain->map[f_pos]; 2237 /* Update statistics */ 2238 rule->pcnt++; 2239 rule->bcnt += pktlen; 2240 rule->timestamp = time_uptime; 2241 } else { 2242 retval = IP_FW_DENY; 2243 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 2244 } 2245 IPFW_RUNLOCK(chain); 2246#ifdef __FreeBSD__ 2247 if (ucred_cache != NULL) 2248 crfree(ucred_cache); 2249#endif 2250 return (retval); 2251 2252pullup_failed: 2253 if (V_fw_verbose) 2254 printf("ipfw: pullup failed\n"); 2255 return (IP_FW_DENY); 2256} 2257 2258/* 2259 * Module and VNET glue 2260 */ 2261 2262/* 2263 * Stuff that must be initialised only on boot or module load 2264 */ 2265static int 2266ipfw_init(void) 2267{ 2268 int error = 0; 2269 2270 ipfw_dyn_attach(); 2271 /* 2272 * Only print out this stuff the first time around, 2273 * when called from the sysinit code. 2274 */ 2275 printf("ipfw2 " 2276#ifdef INET6 2277 "(+ipv6) " 2278#endif 2279 "initialized, divert %s, nat %s, " 2280 "rule-based forwarding " 2281#ifdef IPFIREWALL_FORWARD 2282 "enabled, " 2283#else 2284 "disabled, " 2285#endif 2286 "default to %s, logging ", 2287#ifdef IPDIVERT 2288 "enabled", 2289#else 2290 "loadable", 2291#endif 2292#ifdef IPFIREWALL_NAT 2293 "enabled", 2294#else 2295 "loadable", 2296#endif 2297 default_to_accept ? "accept" : "deny"); 2298 2299 /* 2300 * Note: V_xxx variables can be accessed here but the vnet specific 2301 * initializer may not have been called yet for the VIMAGE case. 2302 * Tuneables will have been processed. We will print out values for 2303 * the default vnet. 2304 * XXX This should all be rationalized AFTER 8.0 2305 */ 2306 if (V_fw_verbose == 0) 2307 printf("disabled\n"); 2308 else if (V_verbose_limit == 0) 2309 printf("unlimited\n"); 2310 else 2311 printf("limited to %d packets/entry by default\n", 2312 V_verbose_limit); 2313 2314 ipfw_log_bpf(1); /* init */ 2315 return (error); 2316} 2317 2318/* 2319 * Called for the removal of the last instance only on module unload. 2320 */ 2321static void 2322ipfw_destroy(void) 2323{ 2324 2325 ipfw_log_bpf(0); /* uninit */ 2326 ipfw_dyn_detach(); 2327 printf("IP firewall unloaded\n"); 2328} 2329 2330/* 2331 * Stuff that must be initialized for every instance 2332 * (including the first of course). 2333 */ 2334static int 2335vnet_ipfw_init(const void *unused) 2336{ 2337 int error; 2338 struct ip_fw *rule = NULL; 2339 struct ip_fw_chain *chain; 2340 2341 chain = &V_layer3_chain; 2342 2343 /* First set up some values that are compile time options */ 2344 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 2345 V_fw_deny_unknown_exthdrs = 1; 2346#ifdef IPFIREWALL_VERBOSE 2347 V_fw_verbose = 1; 2348#endif 2349#ifdef IPFIREWALL_VERBOSE_LIMIT 2350 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 2351#endif 2352#ifdef IPFIREWALL_NAT 2353 LIST_INIT(&chain->nat); 2354#endif 2355 2356 /* insert the default rule and create the initial map */ 2357 chain->n_rules = 1; 2358 chain->static_len = sizeof(struct ip_fw); 2359 chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_NOWAIT | M_ZERO); 2360 if (chain->map) 2361 rule = malloc(chain->static_len, M_IPFW, M_NOWAIT | M_ZERO); 2362 if (rule == NULL) { 2363 if (chain->map) 2364 free(chain->map, M_IPFW); 2365 printf("ipfw2: ENOSPC initializing default rule " 2366 "(support disabled)\n"); 2367 return (ENOSPC); 2368 } 2369 error = ipfw_init_tables(chain); 2370 if (error) { 2371 panic("init_tables"); /* XXX Marko fix this ! */ 2372 } 2373 2374 /* fill and insert the default rule */ 2375 rule->act_ofs = 0; 2376 rule->rulenum = IPFW_DEFAULT_RULE; 2377 rule->cmd_len = 1; 2378 rule->set = RESVD_SET; 2379 rule->cmd[0].len = 1; 2380 rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; 2381 chain->rules = chain->default_rule = chain->map[0] = rule; 2382 chain->id = rule->id = 1; 2383 2384 IPFW_LOCK_INIT(chain); 2385 ipfw_dyn_init(); 2386 2387 /* First set up some values that are compile time options */ 2388 V_ipfw_vnet_ready = 1; /* Open for business */ 2389 2390 /* 2391 * Hook the sockopt handler, and the layer2 (V_ip_fw_chk_ptr) 2392 * and pfil hooks for ipv4 and ipv6. Even if the latter two fail 2393 * we still keep the module alive because the sockopt and 2394 * layer2 paths are still useful. 2395 * ipfw[6]_hook return 0 on success, ENOENT on failure, 2396 * so we can ignore the exact return value and just set a flag. 2397 * 2398 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so 2399 * changes in the underlying (per-vnet) variables trigger 2400 * immediate hook()/unhook() calls. 2401 * In layer2 we have the same behaviour, except that V_ether_ipfw 2402 * is checked on each packet because there are no pfil hooks. 2403 */ 2404 V_ip_fw_ctl_ptr = ipfw_ctl; 2405 V_ip_fw_chk_ptr = ipfw_chk; 2406 error = ipfw_attach_hooks(1); 2407 return (error); 2408} 2409 2410/* 2411 * Called for the removal of each instance. 2412 */ 2413static int 2414vnet_ipfw_uninit(const void *unused) 2415{ 2416 struct ip_fw *reap, *rule; 2417 struct ip_fw_chain *chain = &V_layer3_chain; 2418 int i; 2419 2420 V_ipfw_vnet_ready = 0; /* tell new callers to go away */ 2421 /* 2422 * disconnect from ipv4, ipv6, layer2 and sockopt. 2423 * Then grab, release and grab again the WLOCK so we make 2424 * sure the update is propagated and nobody will be in. 2425 */ 2426 (void)ipfw_attach_hooks(0 /* detach */); 2427 V_ip_fw_chk_ptr = NULL; 2428 V_ip_fw_ctl_ptr = NULL; 2429 IPFW_UH_WLOCK(chain); 2430 IPFW_UH_WUNLOCK(chain); 2431 IPFW_UH_WLOCK(chain); 2432 2433 IPFW_WLOCK(chain); 2434 IPFW_WUNLOCK(chain); 2435 IPFW_WLOCK(chain); 2436 2437 ipfw_dyn_uninit(0); /* run the callout_drain */ 2438 ipfw_destroy_tables(chain); 2439 reap = NULL; 2440 for (i = 0; i < chain->n_rules; i++) { 2441 rule = chain->map[i]; 2442 rule->x_next = reap; 2443 reap = rule; 2444 } 2445 if (chain->map) 2446 free(chain->map, M_IPFW); 2447 IPFW_WUNLOCK(chain); 2448 IPFW_UH_WUNLOCK(chain); 2449 if (reap != NULL) 2450 ipfw_reap_rules(reap); 2451 IPFW_LOCK_DESTROY(chain); 2452 ipfw_dyn_uninit(1); /* free the remaining parts */ 2453 return 0; 2454} 2455 2456/* 2457 * Module event handler. 2458 * In general we have the choice of handling most of these events by the 2459 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to 2460 * use the SYSINIT handlers as they are more capable of expressing the 2461 * flow of control during module and vnet operations, so this is just 2462 * a skeleton. Note there is no SYSINIT equivalent of the module 2463 * SHUTDOWN handler, but we don't have anything to do in that case anyhow. 2464 */ 2465static int 2466ipfw_modevent(module_t mod, int type, void *unused) 2467{ 2468 int err = 0; 2469 2470 switch (type) { 2471 case MOD_LOAD: 2472 /* Called once at module load or 2473 * system boot if compiled in. */ 2474 break; 2475 case MOD_QUIESCE: 2476 /* Called before unload. May veto unloading. */ 2477 break; 2478 case MOD_UNLOAD: 2479 /* Called during unload. */ 2480 break; 2481 case MOD_SHUTDOWN: 2482 /* Called during system shutdown. */ 2483 break; 2484 default: 2485 err = EOPNOTSUPP; 2486 break; 2487 } 2488 return err; 2489} 2490 2491static moduledata_t ipfwmod = { 2492 "ipfw", 2493 ipfw_modevent, 2494 0 2495}; 2496 2497/* Define startup order. */ 2498#define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_IFATTACHDOMAIN 2499#define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */ 2500#define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */ 2501#define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */ 2502 2503DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER); 2504MODULE_VERSION(ipfw, 2); 2505/* should declare some dependencies here */ 2506 2507/* 2508 * Starting up. Done in order after ipfwmod() has been called. 2509 * VNET_SYSINIT is also called for each existing vnet and each new vnet. 2510 */ 2511SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 2512 ipfw_init, NULL); 2513VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 2514 vnet_ipfw_init, NULL); 2515 2516/* 2517 * Closing up shop. These are done in REVERSE ORDER, but still 2518 * after ipfwmod() has been called. Not called on reboot. 2519 * VNET_SYSUNINIT is also called for each exiting vnet as it exits. 2520 * or when the module is unloaded. 2521 */ 2522SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 2523 ipfw_destroy, NULL); 2524VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 2525 vnet_ipfw_uninit, NULL); 2526/* end of file */ 2527