ip_fw2.c revision 200629
1/*- 2 * Copyright (c) 2002 Luigi Rizzo, Universita` di Pisa 3 * 4 * Redistribution and use in source and binary forms, with or without 5 * modification, are permitted provided that the following conditions 6 * are met: 7 * 1. Redistributions of source code must retain the above copyright 8 * notice, this list of conditions and the following disclaimer. 9 * 2. Redistributions in binary form must reproduce the above copyright 10 * notice, this list of conditions and the following disclaimer in the 11 * documentation and/or other materials provided with the distribution. 12 * 13 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 14 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 15 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 16 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 17 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 18 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 19 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 20 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 21 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 22 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 23 * SUCH DAMAGE. 24 */ 25 26#include <sys/cdefs.h> 27__FBSDID("$FreeBSD: head/sys/netinet/ipfw/ip_fw2.c 200629 2009-12-17 12:27:54Z luigi $"); 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_divert.h> 78#include <netinet/ip_dummynet.h> 79#include <netinet/ip_carp.h> 80#include <netinet/pim.h> 81#include <netinet/tcp_var.h> 82#include <netinet/udp.h> 83#include <netinet/udp_var.h> 84#include <netinet/sctp.h> 85 86#include <netgraph/ng_ipfw.h> 87 88#include <netinet/ip6.h> 89#include <netinet/icmp6.h> 90#ifdef INET6 91#include <netinet6/scope6_var.h> 92#include <netinet6/ip6_var.h> 93#endif 94 95#include <machine/in_cksum.h> /* XXX for in_cksum */ 96 97#ifdef MAC 98#include <security/mac/mac_framework.h> 99#endif 100 101/* 102 * static variables followed by global ones. 103 * All ipfw global variables are here. 104 */ 105 106/* ipfw_vnet_ready controls when we are open for business */ 107static VNET_DEFINE(int, ipfw_vnet_ready) = 0; 108#define V_ipfw_vnet_ready VNET(ipfw_vnet_ready) 109 110static VNET_DEFINE(int, fw_deny_unknown_exthdrs); 111#define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs) 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); 120 121/* 122 * Each rule belongs to one of 32 different sets (0..31). 123 * The variable set_disable contains one bit per set. 124 * If the bit is set, all rules in the corresponding set 125 * are disabled. Set RESVD_SET(31) is reserved for the default rule 126 * and rules that are not deleted by the flush command, 127 * and CANNOT be disabled. 128 * Rules in set RESVD_SET can only be deleted individually. 129 */ 130VNET_DEFINE(u_int32_t, set_disable); 131#define V_set_disable VNET(set_disable) 132 133VNET_DEFINE(int, fw_verbose); 134//#define V_verbose_limit VNET(verbose_limit) 135/* counter for ipfw_log(NULL...) */ 136VNET_DEFINE(u_int64_t, norule_counter); 137VNET_DEFINE(int, verbose_limit); 138 139 140/* layer3_chain contains the list of rules for layer 3 */ 141VNET_DEFINE(struct ip_fw_chain, layer3_chain); 142 143ipfw_nat_t *ipfw_nat_ptr = NULL; 144struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int); 145ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; 146ipfw_nat_cfg_t *ipfw_nat_del_ptr; 147ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; 148ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; 149 150extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS); 151 152#ifdef SYSCTL_NODE 153SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 154SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 155 CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0, 156 "Only do a single pass through ipfw when using dummynet(4)"); 157SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, 158 CTLFLAG_RW, &VNET_NAME(autoinc_step), 0, 159 "Rule number auto-increment step"); 160SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose, 161 CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0, 162 "Log matches to ipfw rules"); 163SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, 164 CTLFLAG_RW, &VNET_NAME(verbose_limit), 0, 165 "Set upper limit of matches of ipfw rules logged"); 166SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, 167 NULL, IPFW_DEFAULT_RULE, 168 "The default/max possible rule number."); 169SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, tables_max, CTLFLAG_RD, 170 NULL, IPFW_TABLES_MAX, 171 "The maximum number of tables."); 172SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN, 173 &default_to_accept, 0, 174 "Make the default rule accept all packets."); 175TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept); 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 */ 184 185#endif /* SYSCTL_NODE */ 186 187 188/* 189 * Some macros used in the various matching options. 190 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 191 * Other macros just cast void * into the appropriate type 192 */ 193#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 194#define TCP(p) ((struct tcphdr *)(p)) 195#define SCTP(p) ((struct sctphdr *)(p)) 196#define UDP(p) ((struct udphdr *)(p)) 197#define ICMP(p) ((struct icmphdr *)(p)) 198#define ICMP6(p) ((struct icmp6_hdr *)(p)) 199 200static __inline int 201icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 202{ 203 int type = icmp->icmp_type; 204 205 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 206} 207 208#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 209 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 210 211static int 212is_icmp_query(struct icmphdr *icmp) 213{ 214 int type = icmp->icmp_type; 215 216 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 217} 218#undef TT 219 220/* 221 * The following checks use two arrays of 8 or 16 bits to store the 222 * bits that we want set or clear, respectively. They are in the 223 * low and high half of cmd->arg1 or cmd->d[0]. 224 * 225 * We scan options and store the bits we find set. We succeed if 226 * 227 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 228 * 229 * The code is sometimes optimized not to store additional variables. 230 */ 231 232static int 233flags_match(ipfw_insn *cmd, u_int8_t bits) 234{ 235 u_char want_clear; 236 bits = ~bits; 237 238 if ( ((cmd->arg1 & 0xff) & bits) != 0) 239 return 0; /* some bits we want set were clear */ 240 want_clear = (cmd->arg1 >> 8) & 0xff; 241 if ( (want_clear & bits) != want_clear) 242 return 0; /* some bits we want clear were set */ 243 return 1; 244} 245 246static int 247ipopts_match(struct ip *ip, ipfw_insn *cmd) 248{ 249 int optlen, bits = 0; 250 u_char *cp = (u_char *)(ip + 1); 251 int x = (ip->ip_hl << 2) - sizeof (struct ip); 252 253 for (; x > 0; x -= optlen, cp += optlen) { 254 int opt = cp[IPOPT_OPTVAL]; 255 256 if (opt == IPOPT_EOL) 257 break; 258 if (opt == IPOPT_NOP) 259 optlen = 1; 260 else { 261 optlen = cp[IPOPT_OLEN]; 262 if (optlen <= 0 || optlen > x) 263 return 0; /* invalid or truncated */ 264 } 265 switch (opt) { 266 267 default: 268 break; 269 270 case IPOPT_LSRR: 271 bits |= IP_FW_IPOPT_LSRR; 272 break; 273 274 case IPOPT_SSRR: 275 bits |= IP_FW_IPOPT_SSRR; 276 break; 277 278 case IPOPT_RR: 279 bits |= IP_FW_IPOPT_RR; 280 break; 281 282 case IPOPT_TS: 283 bits |= IP_FW_IPOPT_TS; 284 break; 285 } 286 } 287 return (flags_match(cmd, bits)); 288} 289 290static int 291tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 292{ 293 int optlen, bits = 0; 294 u_char *cp = (u_char *)(tcp + 1); 295 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 296 297 for (; x > 0; x -= optlen, cp += optlen) { 298 int opt = cp[0]; 299 if (opt == TCPOPT_EOL) 300 break; 301 if (opt == TCPOPT_NOP) 302 optlen = 1; 303 else { 304 optlen = cp[1]; 305 if (optlen <= 0) 306 break; 307 } 308 309 switch (opt) { 310 311 default: 312 break; 313 314 case TCPOPT_MAXSEG: 315 bits |= IP_FW_TCPOPT_MSS; 316 break; 317 318 case TCPOPT_WINDOW: 319 bits |= IP_FW_TCPOPT_WINDOW; 320 break; 321 322 case TCPOPT_SACK_PERMITTED: 323 case TCPOPT_SACK: 324 bits |= IP_FW_TCPOPT_SACK; 325 break; 326 327 case TCPOPT_TIMESTAMP: 328 bits |= IP_FW_TCPOPT_TS; 329 break; 330 331 } 332 } 333 return (flags_match(cmd, bits)); 334} 335 336static int 337iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 338{ 339 if (ifp == NULL) /* no iface with this packet, match fails */ 340 return 0; 341 /* Check by name or by IP address */ 342 if (cmd->name[0] != '\0') { /* match by name */ 343 /* Check name */ 344 if (cmd->p.glob) { 345 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 346 return(1); 347 } else { 348 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 349 return(1); 350 } 351 } else { 352 struct ifaddr *ia; 353 354 if_addr_rlock(ifp); 355 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 356 if (ia->ifa_addr->sa_family != AF_INET) 357 continue; 358 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 359 (ia->ifa_addr))->sin_addr.s_addr) { 360 if_addr_runlock(ifp); 361 return(1); /* match */ 362 } 363 } 364 if_addr_runlock(ifp); 365 } 366 return(0); /* no match, fail ... */ 367} 368 369/* 370 * The verify_path function checks if a route to the src exists and 371 * if it is reachable via ifp (when provided). 372 * 373 * The 'verrevpath' option checks that the interface that an IP packet 374 * arrives on is the same interface that traffic destined for the 375 * packet's source address would be routed out of. 376 * The 'versrcreach' option just checks that the source address is 377 * reachable via any route (except default) in the routing table. 378 * These two are a measure to block forged packets. This is also 379 * commonly known as "anti-spoofing" or Unicast Reverse Path 380 * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 381 * is purposely reminiscent of the Cisco IOS command, 382 * 383 * ip verify unicast reverse-path 384 * ip verify unicast source reachable-via any 385 * 386 * which implements the same functionality. But note that the syntax 387 * is misleading, and the check may be performed on all IP packets 388 * whether unicast, multicast, or broadcast. 389 */ 390static int 391verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) 392{ 393 struct route ro; 394 struct sockaddr_in *dst; 395 396 bzero(&ro, sizeof(ro)); 397 398 dst = (struct sockaddr_in *)&(ro.ro_dst); 399 dst->sin_family = AF_INET; 400 dst->sin_len = sizeof(*dst); 401 dst->sin_addr = src; 402 in_rtalloc_ign(&ro, 0, fib); 403 404 if (ro.ro_rt == NULL) 405 return 0; 406 407 /* 408 * If ifp is provided, check for equality with rtentry. 409 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 410 * in order to pass packets injected back by if_simloop(): 411 * if useloopback == 1 routing entry (via lo0) for our own address 412 * may exist, so we need to handle routing assymetry. 413 */ 414 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 415 RTFREE(ro.ro_rt); 416 return 0; 417 } 418 419 /* if no ifp provided, check if rtentry is not default route */ 420 if (ifp == NULL && 421 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 422 RTFREE(ro.ro_rt); 423 return 0; 424 } 425 426 /* or if this is a blackhole/reject route */ 427 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 428 RTFREE(ro.ro_rt); 429 return 0; 430 } 431 432 /* found valid route */ 433 RTFREE(ro.ro_rt); 434 return 1; 435} 436 437#ifdef INET6 438/* 439 * ipv6 specific rules here... 440 */ 441static __inline int 442icmp6type_match (int type, ipfw_insn_u32 *cmd) 443{ 444 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 445} 446 447static int 448flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 449{ 450 int i; 451 for (i=0; i <= cmd->o.arg1; ++i ) 452 if (curr_flow == cmd->d[i] ) 453 return 1; 454 return 0; 455} 456 457/* support for IP6_*_ME opcodes */ 458static int 459search_ip6_addr_net (struct in6_addr * ip6_addr) 460{ 461 struct ifnet *mdc; 462 struct ifaddr *mdc2; 463 struct in6_ifaddr *fdm; 464 struct in6_addr copia; 465 466 TAILQ_FOREACH(mdc, &V_ifnet, if_link) { 467 if_addr_rlock(mdc); 468 TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) { 469 if (mdc2->ifa_addr->sa_family == AF_INET6) { 470 fdm = (struct in6_ifaddr *)mdc2; 471 copia = fdm->ia_addr.sin6_addr; 472 /* need for leaving scope_id in the sock_addr */ 473 in6_clearscope(&copia); 474 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) { 475 if_addr_runlock(mdc); 476 return 1; 477 } 478 } 479 } 480 if_addr_runlock(mdc); 481 } 482 return 0; 483} 484 485static int 486verify_path6(struct in6_addr *src, struct ifnet *ifp) 487{ 488 struct route_in6 ro; 489 struct sockaddr_in6 *dst; 490 491 bzero(&ro, sizeof(ro)); 492 493 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 494 dst->sin6_family = AF_INET6; 495 dst->sin6_len = sizeof(*dst); 496 dst->sin6_addr = *src; 497 /* XXX MRT 0 for ipv6 at this time */ 498 rtalloc_ign((struct route *)&ro, 0); 499 500 if (ro.ro_rt == NULL) 501 return 0; 502 503 /* 504 * if ifp is provided, check for equality with rtentry 505 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 506 * to support the case of sending packets to an address of our own. 507 * (where the former interface is the first argument of if_simloop() 508 * (=ifp), the latter is lo0) 509 */ 510 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 511 RTFREE(ro.ro_rt); 512 return 0; 513 } 514 515 /* if no ifp provided, check if rtentry is not default route */ 516 if (ifp == NULL && 517 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 518 RTFREE(ro.ro_rt); 519 return 0; 520 } 521 522 /* or if this is a blackhole/reject route */ 523 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 524 RTFREE(ro.ro_rt); 525 return 0; 526 } 527 528 /* found valid route */ 529 RTFREE(ro.ro_rt); 530 return 1; 531 532} 533 534static int 535is_icmp6_query(int icmp6_type) 536{ 537 if ((icmp6_type <= ICMP6_MAXTYPE) && 538 (icmp6_type == ICMP6_ECHO_REQUEST || 539 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 540 icmp6_type == ICMP6_WRUREQUEST || 541 icmp6_type == ICMP6_FQDN_QUERY || 542 icmp6_type == ICMP6_NI_QUERY)) 543 return (1); 544 545 return (0); 546} 547 548static void 549send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 550{ 551 struct mbuf *m; 552 553 m = args->m; 554 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 555 struct tcphdr *tcp; 556 tcp = (struct tcphdr *)((char *)ip6 + hlen); 557 558 if ((tcp->th_flags & TH_RST) == 0) { 559 struct mbuf *m0; 560 m0 = ipfw_send_pkt(args->m, &(args->f_id), 561 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 562 tcp->th_flags | TH_RST); 563 if (m0 != NULL) 564 ip6_output(m0, NULL, NULL, 0, NULL, NULL, 565 NULL); 566 } 567 m_freem(m); 568 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 569#if 0 570 /* 571 * Unlike above, the mbufs need to line up with the ip6 hdr, 572 * as the contents are read. We need to m_adj() the 573 * needed amount. 574 * The mbuf will however be thrown away so we can adjust it. 575 * Remember we did an m_pullup on it already so we 576 * can make some assumptions about contiguousness. 577 */ 578 if (args->L3offset) 579 m_adj(m, args->L3offset); 580#endif 581 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 582 } else 583 m_freem(m); 584 585 args->m = NULL; 586} 587 588#endif /* INET6 */ 589 590 591/* 592 * sends a reject message, consuming the mbuf passed as an argument. 593 */ 594static void 595send_reject(struct ip_fw_args *args, int code, int ip_len, struct ip *ip) 596{ 597 598#if 0 599 /* XXX When ip is not guaranteed to be at mtod() we will 600 * need to account for this */ 601 * The mbuf will however be thrown away so we can adjust it. 602 * Remember we did an m_pullup on it already so we 603 * can make some assumptions about contiguousness. 604 */ 605 if (args->L3offset) 606 m_adj(m, args->L3offset); 607#endif 608 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 609 /* We need the IP header in host order for icmp_error(). */ 610 if (args->eh != NULL) { 611 ip->ip_len = ntohs(ip->ip_len); 612 ip->ip_off = ntohs(ip->ip_off); 613 } 614 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 615 } else if (args->f_id.proto == IPPROTO_TCP) { 616 struct tcphdr *const tcp = 617 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 618 if ( (tcp->th_flags & TH_RST) == 0) { 619 struct mbuf *m; 620 m = ipfw_send_pkt(args->m, &(args->f_id), 621 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 622 tcp->th_flags | TH_RST); 623 if (m != NULL) 624 ip_output(m, NULL, NULL, 0, NULL, NULL); 625 } 626 m_freem(args->m); 627 } else 628 m_freem(args->m); 629 args->m = NULL; 630} 631 632/** 633 * Given an ip_fw *, lookup_next_rule will return a pointer 634 * to the next rule, which can be either the jump 635 * target (for skipto instructions) or the next one in the list (in 636 * all other cases including a missing jump target). 637 * The result is also written in the "next_rule" field of the rule. 638 * Backward jumps are not allowed, so we start the search from the 639 * rule following the current one. 640 * 641 * The function never returns NULL: if the requested rule is not 642 * present, it returns the next rule in the chain. 643 * As a side effect, the rule pointer is also set so next time 644 * the jump will not require a scan of the list. 645 */ 646 647static struct ip_fw * 648lookup_next_rule(struct ip_fw *me, u_int32_t tablearg) 649{ 650 struct ip_fw *rule = NULL; 651 ipfw_insn *cmd; 652 u_int16_t rulenum; 653 654 /* look for action, in case it is a skipto */ 655 cmd = ACTION_PTR(me); 656 if (cmd->opcode == O_LOG) 657 cmd += F_LEN(cmd); 658 if (cmd->opcode == O_ALTQ) 659 cmd += F_LEN(cmd); 660 if (cmd->opcode == O_TAG) 661 cmd += F_LEN(cmd); 662 if (cmd->opcode == O_SKIPTO ) { 663 if (tablearg != 0) { 664 rulenum = (u_int16_t)tablearg; 665 } else { 666 rulenum = cmd->arg1; 667 } 668 for (rule = me->next; rule ; rule = rule->next) { 669 if (rule->rulenum >= rulenum) { 670 break; 671 } 672 } 673 } 674 if (rule == NULL) /* failure or not a skipto */ 675 rule = me->next; 676 me->next_rule = rule; 677 return rule; 678} 679 680/* 681 * Support for uid/gid/jail lookup. These tests are expensive 682 * (because we may need to look into the list of active sockets) 683 * so we cache the results. ugid_lookupp is 0 if we have not 684 * yet done a lookup, 1 if we succeeded, and -1 if we tried 685 * and failed. The function always returns the match value. 686 * We could actually spare the variable and use *uc, setting 687 * it to '(void *)check_uidgid if we have no info, NULL if 688 * we tried and failed, or any other value if successful. 689 */ 690static int 691check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif, 692 struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip, 693 u_int16_t src_port, struct ucred **uc, int *ugid_lookupp, 694 struct inpcb *inp) 695{ 696 struct inpcbinfo *pi; 697 int wildcard; 698 struct inpcb *pcb; 699 int match; 700 701 /* 702 * Check to see if the UDP or TCP stack supplied us with 703 * the PCB. If so, rather then holding a lock and looking 704 * up the PCB, we can use the one that was supplied. 705 */ 706 if (inp && *ugid_lookupp == 0) { 707 INP_LOCK_ASSERT(inp); 708 if (inp->inp_socket != NULL) { 709 *uc = crhold(inp->inp_cred); 710 *ugid_lookupp = 1; 711 } else 712 *ugid_lookupp = -1; 713 } 714 /* 715 * If we have already been here and the packet has no 716 * PCB entry associated with it, then we can safely 717 * assume that this is a no match. 718 */ 719 if (*ugid_lookupp == -1) 720 return (0); 721 if (proto == IPPROTO_TCP) { 722 wildcard = 0; 723 pi = &V_tcbinfo; 724 } else if (proto == IPPROTO_UDP) { 725 wildcard = INPLOOKUP_WILDCARD; 726 pi = &V_udbinfo; 727 } else 728 return 0; 729 match = 0; 730 if (*ugid_lookupp == 0) { 731 INP_INFO_RLOCK(pi); 732 pcb = (oif) ? 733 in_pcblookup_hash(pi, 734 dst_ip, htons(dst_port), 735 src_ip, htons(src_port), 736 wildcard, oif) : 737 in_pcblookup_hash(pi, 738 src_ip, htons(src_port), 739 dst_ip, htons(dst_port), 740 wildcard, NULL); 741 if (pcb != NULL) { 742 *uc = crhold(pcb->inp_cred); 743 *ugid_lookupp = 1; 744 } 745 INP_INFO_RUNLOCK(pi); 746 if (*ugid_lookupp == 0) { 747 /* 748 * We tried and failed, set the variable to -1 749 * so we will not try again on this packet. 750 */ 751 *ugid_lookupp = -1; 752 return (0); 753 } 754 } 755 if (insn->o.opcode == O_UID) 756 match = ((*uc)->cr_uid == (uid_t)insn->d[0]); 757 else if (insn->o.opcode == O_GID) 758 match = groupmember((gid_t)insn->d[0], *uc); 759 else if (insn->o.opcode == O_JAIL) 760 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]); 761 return match; 762} 763 764/* 765 * The main check routine for the firewall. 766 * 767 * All arguments are in args so we can modify them and return them 768 * back to the caller. 769 * 770 * Parameters: 771 * 772 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 773 * Starts with the IP header. 774 * args->eh (in) Mac header if present, NULL for layer3 packet. 775 * args->L3offset Number of bytes bypassed if we came from L2. 776 * e.g. often sizeof(eh) ** NOTYET ** 777 * args->oif Outgoing interface, NULL if packet is incoming. 778 * The incoming interface is in the mbuf. (in) 779 * args->divert_rule (in/out) 780 * Skip up to the first rule past this rule number; 781 * upon return, non-zero port number for divert or tee. 782 * 783 * args->rule Pointer to the last matching rule (in/out) 784 * args->next_hop Socket we are forwarding to (out). 785 * args->f_id Addresses grabbed from the packet (out) 786 * args->cookie a cookie depending on rule action 787 * 788 * Return value: 789 * 790 * IP_FW_PASS the packet must be accepted 791 * IP_FW_DENY the packet must be dropped 792 * IP_FW_DIVERT divert packet, port in m_tag 793 * IP_FW_TEE tee packet, port in m_tag 794 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 795 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 796 * 797 */ 798int 799ipfw_chk(struct ip_fw_args *args) 800{ 801 802 /* 803 * Local variables holding state while processing a packet: 804 * 805 * IMPORTANT NOTE: to speed up the processing of rules, there 806 * are some assumption on the values of the variables, which 807 * are documented here. Should you change them, please check 808 * the implementation of the various instructions to make sure 809 * that they still work. 810 * 811 * args->eh The MAC header. It is non-null for a layer2 812 * packet, it is NULL for a layer-3 packet. 813 * **notyet** 814 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. 815 * 816 * m | args->m Pointer to the mbuf, as received from the caller. 817 * It may change if ipfw_chk() does an m_pullup, or if it 818 * consumes the packet because it calls send_reject(). 819 * XXX This has to change, so that ipfw_chk() never modifies 820 * or consumes the buffer. 821 * ip is the beginning of the ip(4 or 6) header. 822 * Calculated by adding the L3offset to the start of data. 823 * (Until we start using L3offset, the packet is 824 * supposed to start with the ip header). 825 */ 826 struct mbuf *m = args->m; 827 struct ip *ip = mtod(m, struct ip *); 828 829 /* 830 * For rules which contain uid/gid or jail constraints, cache 831 * a copy of the users credentials after the pcb lookup has been 832 * executed. This will speed up the processing of rules with 833 * these types of constraints, as well as decrease contention 834 * on pcb related locks. 835 */ 836 struct ucred *ucred_cache = NULL; 837 int ucred_lookup = 0; 838 839 /* 840 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 841 * associated with a packet input on a divert socket. This 842 * will allow to distinguish traffic and its direction when 843 * it originates from a divert socket. 844 */ 845 u_int divinput_flags = 0; 846 847 /* 848 * oif | args->oif If NULL, ipfw_chk has been called on the 849 * inbound path (ether_input, ip_input). 850 * If non-NULL, ipfw_chk has been called on the outbound path 851 * (ether_output, ip_output). 852 */ 853 struct ifnet *oif = args->oif; 854 855 struct ip_fw *f = NULL; /* matching rule */ 856 int retval = 0; 857 858 /* 859 * hlen The length of the IP header. 860 */ 861 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 862 863 /* 864 * offset The offset of a fragment. offset != 0 means that 865 * we have a fragment at this offset of an IPv4 packet. 866 * offset == 0 means that (if this is an IPv4 packet) 867 * this is the first or only fragment. 868 * For IPv6 offset == 0 means there is no Fragment Header. 869 * If offset != 0 for IPv6 always use correct mask to 870 * get the correct offset because we add IP6F_MORE_FRAG 871 * to be able to dectect the first fragment which would 872 * otherwise have offset = 0. 873 */ 874 u_short offset = 0; 875 876 /* 877 * Local copies of addresses. They are only valid if we have 878 * an IP packet. 879 * 880 * proto The protocol. Set to 0 for non-ip packets, 881 * or to the protocol read from the packet otherwise. 882 * proto != 0 means that we have an IPv4 packet. 883 * 884 * src_port, dst_port port numbers, in HOST format. Only 885 * valid for TCP and UDP packets. 886 * 887 * src_ip, dst_ip ip addresses, in NETWORK format. 888 * Only valid for IPv4 packets. 889 */ 890 u_int8_t proto; 891 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 892 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 893 u_int16_t ip_len=0; 894 int pktlen; 895 u_int16_t etype = 0; /* Host order stored ether type */ 896 897 /* 898 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 899 * MATCH_NONE when checked and not matched (q = NULL), 900 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 901 */ 902 int dyn_dir = MATCH_UNKNOWN; 903 ipfw_dyn_rule *q = NULL; 904 struct ip_fw_chain *chain = &V_layer3_chain; 905 struct m_tag *mtag; 906 907 /* 908 * We store in ulp a pointer to the upper layer protocol header. 909 * In the ipv4 case this is easy to determine from the header, 910 * but for ipv6 we might have some additional headers in the middle. 911 * ulp is NULL if not found. 912 */ 913 void *ulp = NULL; /* upper layer protocol pointer. */ 914 /* XXX ipv6 variables */ 915 int is_ipv6 = 0; 916 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 917 /* end of ipv6 variables */ 918 int is_ipv4 = 0; 919 920 int done = 0; /* flag to exit the outer loop */ 921 922 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready)) 923 return (IP_FW_PASS); /* accept */ 924 925 dst_ip.s_addr = 0; /* make sure it is initialized */ 926 src_ip.s_addr = 0; /* make sure it is initialized */ 927 pktlen = m->m_pkthdr.len; 928 args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */ 929 proto = args->f_id.proto = 0; /* mark f_id invalid */ 930 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 931 932/* 933 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 934 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 935 * pointer might become stale after other pullups (but we never use it 936 * this way). 937 */ 938#define PULLUP_TO(_len, p, T) \ 939do { \ 940 int x = (_len) + sizeof(T); \ 941 if ((m)->m_len < x) { \ 942 args->m = m = m_pullup(m, x); \ 943 if (m == NULL) \ 944 goto pullup_failed; \ 945 } \ 946 p = (mtod(m, char *) + (_len)); \ 947} while (0) 948 949 /* 950 * if we have an ether header, 951 */ 952 if (args->eh) 953 etype = ntohs(args->eh->ether_type); 954 955 /* Identify IP packets and fill up variables. */ 956 if (pktlen >= sizeof(struct ip6_hdr) && 957 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { 958 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 959 is_ipv6 = 1; 960 args->f_id.addr_type = 6; 961 hlen = sizeof(struct ip6_hdr); 962 proto = ip6->ip6_nxt; 963 964 /* Search extension headers to find upper layer protocols */ 965 while (ulp == NULL) { 966 switch (proto) { 967 case IPPROTO_ICMPV6: 968 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 969 args->f_id.flags = ICMP6(ulp)->icmp6_type; 970 break; 971 972 case IPPROTO_TCP: 973 PULLUP_TO(hlen, ulp, struct tcphdr); 974 dst_port = TCP(ulp)->th_dport; 975 src_port = TCP(ulp)->th_sport; 976 args->f_id.flags = TCP(ulp)->th_flags; 977 break; 978 979 case IPPROTO_SCTP: 980 PULLUP_TO(hlen, ulp, struct sctphdr); 981 src_port = SCTP(ulp)->src_port; 982 dst_port = SCTP(ulp)->dest_port; 983 break; 984 985 case IPPROTO_UDP: 986 PULLUP_TO(hlen, ulp, struct udphdr); 987 dst_port = UDP(ulp)->uh_dport; 988 src_port = UDP(ulp)->uh_sport; 989 break; 990 991 case IPPROTO_HOPOPTS: /* RFC 2460 */ 992 PULLUP_TO(hlen, ulp, struct ip6_hbh); 993 ext_hd |= EXT_HOPOPTS; 994 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 995 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 996 ulp = NULL; 997 break; 998 999 case IPPROTO_ROUTING: /* RFC 2460 */ 1000 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 1001 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 1002 case 0: 1003 ext_hd |= EXT_RTHDR0; 1004 break; 1005 case 2: 1006 ext_hd |= EXT_RTHDR2; 1007 break; 1008 default: 1009 printf("IPFW2: IPV6 - Unknown Routing " 1010 "Header type(%d)\n", 1011 ((struct ip6_rthdr *)ulp)->ip6r_type); 1012 if (V_fw_deny_unknown_exthdrs) 1013 return (IP_FW_DENY); 1014 break; 1015 } 1016 ext_hd |= EXT_ROUTING; 1017 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 1018 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 1019 ulp = NULL; 1020 break; 1021 1022 case IPPROTO_FRAGMENT: /* RFC 2460 */ 1023 PULLUP_TO(hlen, ulp, struct ip6_frag); 1024 ext_hd |= EXT_FRAGMENT; 1025 hlen += sizeof (struct ip6_frag); 1026 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 1027 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 1028 IP6F_OFF_MASK; 1029 /* Add IP6F_MORE_FRAG for offset of first 1030 * fragment to be != 0. */ 1031 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 1032 IP6F_MORE_FRAG; 1033 if (offset == 0) { 1034 printf("IPFW2: IPV6 - Invalid Fragment " 1035 "Header\n"); 1036 if (V_fw_deny_unknown_exthdrs) 1037 return (IP_FW_DENY); 1038 break; 1039 } 1040 args->f_id.frag_id6 = 1041 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 1042 ulp = NULL; 1043 break; 1044 1045 case IPPROTO_DSTOPTS: /* RFC 2460 */ 1046 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1047 ext_hd |= EXT_DSTOPTS; 1048 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1049 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1050 ulp = NULL; 1051 break; 1052 1053 case IPPROTO_AH: /* RFC 2402 */ 1054 PULLUP_TO(hlen, ulp, struct ip6_ext); 1055 ext_hd |= EXT_AH; 1056 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 1057 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 1058 ulp = NULL; 1059 break; 1060 1061 case IPPROTO_ESP: /* RFC 2406 */ 1062 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 1063 /* Anything past Seq# is variable length and 1064 * data past this ext. header is encrypted. */ 1065 ext_hd |= EXT_ESP; 1066 break; 1067 1068 case IPPROTO_NONE: /* RFC 2460 */ 1069 /* 1070 * Packet ends here, and IPv6 header has 1071 * already been pulled up. If ip6e_len!=0 1072 * then octets must be ignored. 1073 */ 1074 ulp = ip; /* non-NULL to get out of loop. */ 1075 break; 1076 1077 case IPPROTO_OSPFIGP: 1078 /* XXX OSPF header check? */ 1079 PULLUP_TO(hlen, ulp, struct ip6_ext); 1080 break; 1081 1082 case IPPROTO_PIM: 1083 /* XXX PIM header check? */ 1084 PULLUP_TO(hlen, ulp, struct pim); 1085 break; 1086 1087 case IPPROTO_CARP: 1088 PULLUP_TO(hlen, ulp, struct carp_header); 1089 if (((struct carp_header *)ulp)->carp_version != 1090 CARP_VERSION) 1091 return (IP_FW_DENY); 1092 if (((struct carp_header *)ulp)->carp_type != 1093 CARP_ADVERTISEMENT) 1094 return (IP_FW_DENY); 1095 break; 1096 1097 case IPPROTO_IPV6: /* RFC 2893 */ 1098 PULLUP_TO(hlen, ulp, struct ip6_hdr); 1099 break; 1100 1101 case IPPROTO_IPV4: /* RFC 2893 */ 1102 PULLUP_TO(hlen, ulp, struct ip); 1103 break; 1104 1105 default: 1106 printf("IPFW2: IPV6 - Unknown Extension " 1107 "Header(%d), ext_hd=%x\n", proto, ext_hd); 1108 if (V_fw_deny_unknown_exthdrs) 1109 return (IP_FW_DENY); 1110 PULLUP_TO(hlen, ulp, struct ip6_ext); 1111 break; 1112 } /*switch */ 1113 } 1114 ip = mtod(m, struct ip *); 1115 ip6 = (struct ip6_hdr *)ip; 1116 args->f_id.src_ip6 = ip6->ip6_src; 1117 args->f_id.dst_ip6 = ip6->ip6_dst; 1118 args->f_id.src_ip = 0; 1119 args->f_id.dst_ip = 0; 1120 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 1121 } else if (pktlen >= sizeof(struct ip) && 1122 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { 1123 is_ipv4 = 1; 1124 hlen = ip->ip_hl << 2; 1125 args->f_id.addr_type = 4; 1126 1127 /* 1128 * Collect parameters into local variables for faster matching. 1129 */ 1130 proto = ip->ip_p; 1131 src_ip = ip->ip_src; 1132 dst_ip = ip->ip_dst; 1133 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 1134 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1135 ip_len = ntohs(ip->ip_len); 1136 } else { 1137 offset = ip->ip_off & IP_OFFMASK; 1138 ip_len = ip->ip_len; 1139 } 1140 pktlen = ip_len < pktlen ? ip_len : pktlen; 1141 1142 if (offset == 0) { 1143 switch (proto) { 1144 case IPPROTO_TCP: 1145 PULLUP_TO(hlen, ulp, struct tcphdr); 1146 dst_port = TCP(ulp)->th_dport; 1147 src_port = TCP(ulp)->th_sport; 1148 args->f_id.flags = TCP(ulp)->th_flags; 1149 break; 1150 1151 case IPPROTO_UDP: 1152 PULLUP_TO(hlen, ulp, struct udphdr); 1153 dst_port = UDP(ulp)->uh_dport; 1154 src_port = UDP(ulp)->uh_sport; 1155 break; 1156 1157 case IPPROTO_ICMP: 1158 PULLUP_TO(hlen, ulp, struct icmphdr); 1159 args->f_id.flags = ICMP(ulp)->icmp_type; 1160 break; 1161 1162 default: 1163 break; 1164 } 1165 } 1166 1167 ip = mtod(m, struct ip *); 1168 args->f_id.src_ip = ntohl(src_ip.s_addr); 1169 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1170 } 1171#undef PULLUP_TO 1172 if (proto) { /* we may have port numbers, store them */ 1173 args->f_id.proto = proto; 1174 args->f_id.src_port = src_port = ntohs(src_port); 1175 args->f_id.dst_port = dst_port = ntohs(dst_port); 1176 } 1177 1178 IPFW_RLOCK(chain); 1179 if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */ 1180 IPFW_RUNLOCK(chain); 1181 return (IP_FW_PASS); /* accept */ 1182 } 1183 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 1184 if (args->rule) { 1185 /* 1186 * Packet has already been tagged as a result of a previous 1187 * match on rule args->rule aka args->rule_id (PIPE, QUEUE, 1188 * REASS, NETGRAPH and similar, never a skipto). 1189 * Validate the pointer and continue from args->rule->next 1190 * if still present, otherwise use the default rule. 1191 * XXX If fw_one_pass != 0 then just accept it, though 1192 * the caller should never pass us such packets. 1193 */ 1194 if (V_fw_one_pass) { 1195 IPFW_RUNLOCK(chain); 1196 return (IP_FW_PASS); 1197 } 1198 if (chain->id == args->chain_id) { /* pointer still valid */ 1199 f = args->rule->next; 1200 } else { /* must revalidate the pointer */ 1201 for (f = chain->rules; f != NULL; f = f->next) 1202 if (f == args->rule && f->id == args->rule_id) { 1203 f = args->rule->next; 1204 break; 1205 } 1206 } 1207 if (f == NULL) /* in case of errors, use default; */ 1208 f = chain->default_rule; 1209 } else { 1210 /* 1211 * Find the starting rule. It can be either the first 1212 * one, or the one after divert_rule if asked so. 1213 */ 1214 int skipto = mtag ? divert_cookie(mtag) : 0; 1215 1216 f = chain->rules; 1217 if (args->eh == NULL && skipto != 0) { 1218 if (skipto >= IPFW_DEFAULT_RULE) { 1219 IPFW_RUNLOCK(chain); 1220 return (IP_FW_DENY); /* invalid */ 1221 } 1222 while (f && f->rulenum <= skipto) 1223 f = f->next; 1224 if (f == NULL) { /* drop packet */ 1225 IPFW_RUNLOCK(chain); 1226 return (IP_FW_DENY); 1227 } 1228 } 1229 } 1230 /* reset divert rule to avoid confusion later */ 1231 if (mtag) { 1232 divinput_flags = divert_info(mtag) & 1233 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 1234 m_tag_delete(m, mtag); 1235 } 1236 1237 /* 1238 * Now scan the rules, and parse microinstructions for each rule. 1239 * We have two nested loops and an inner switch. Sometimes we 1240 * need to break out of one or both loops, or re-enter one of 1241 * the loops with updated variables. Loop variables are: 1242 * 1243 * f (outer loop) points to the current rule. 1244 * On output it points to the matching rule. 1245 * done (outer loop) is used as a flag to break the loop. 1246 * l (inner loop) residual length of current rule. 1247 * cmd points to the current microinstruction. 1248 * 1249 * We break the inner loop by setting l=0 and possibly 1250 * cmdlen=0 if we don't want to advance cmd. 1251 * We break the outer loop by setting done=1 1252 * We can restart the inner loop by setting l>0 and f, cmd 1253 * as needed. 1254 */ 1255 for (; f; f = f->next) { 1256 ipfw_insn *cmd; 1257 uint32_t tablearg = 0; 1258 int l, cmdlen, skip_or; /* skip rest of OR block */ 1259 1260/* again: */ 1261 if (V_set_disable & (1 << f->set) ) 1262 continue; 1263 1264 skip_or = 0; 1265 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1266 l -= cmdlen, cmd += cmdlen) { 1267 int match; 1268 1269 /* 1270 * check_body is a jump target used when we find a 1271 * CHECK_STATE, and need to jump to the body of 1272 * the target rule. 1273 */ 1274 1275/* check_body: */ 1276 cmdlen = F_LEN(cmd); 1277 /* 1278 * An OR block (insn_1 || .. || insn_n) has the 1279 * F_OR bit set in all but the last instruction. 1280 * The first match will set "skip_or", and cause 1281 * the following instructions to be skipped until 1282 * past the one with the F_OR bit clear. 1283 */ 1284 if (skip_or) { /* skip this instruction */ 1285 if ((cmd->len & F_OR) == 0) 1286 skip_or = 0; /* next one is good */ 1287 continue; 1288 } 1289 match = 0; /* set to 1 if we succeed */ 1290 1291 switch (cmd->opcode) { 1292 /* 1293 * The first set of opcodes compares the packet's 1294 * fields with some pattern, setting 'match' if a 1295 * match is found. At the end of the loop there is 1296 * logic to deal with F_NOT and F_OR flags associated 1297 * with the opcode. 1298 */ 1299 case O_NOP: 1300 match = 1; 1301 break; 1302 1303 case O_FORWARD_MAC: 1304 printf("ipfw: opcode %d unimplemented\n", 1305 cmd->opcode); 1306 break; 1307 1308 case O_GID: 1309 case O_UID: 1310 case O_JAIL: 1311 /* 1312 * We only check offset == 0 && proto != 0, 1313 * as this ensures that we have a 1314 * packet with the ports info. 1315 */ 1316 if (offset!=0) 1317 break; 1318 if (is_ipv6) /* XXX to be fixed later */ 1319 break; 1320 if (proto == IPPROTO_TCP || 1321 proto == IPPROTO_UDP) 1322 match = check_uidgid( 1323 (ipfw_insn_u32 *)cmd, 1324 proto, oif, 1325 dst_ip, dst_port, 1326 src_ip, src_port, &ucred_cache, 1327 &ucred_lookup, args->inp); 1328 break; 1329 1330 case O_RECV: 1331 match = iface_match(m->m_pkthdr.rcvif, 1332 (ipfw_insn_if *)cmd); 1333 break; 1334 1335 case O_XMIT: 1336 match = iface_match(oif, (ipfw_insn_if *)cmd); 1337 break; 1338 1339 case O_VIA: 1340 match = iface_match(oif ? oif : 1341 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 1342 break; 1343 1344 case O_MACADDR2: 1345 if (args->eh != NULL) { /* have MAC header */ 1346 u_int32_t *want = (u_int32_t *) 1347 ((ipfw_insn_mac *)cmd)->addr; 1348 u_int32_t *mask = (u_int32_t *) 1349 ((ipfw_insn_mac *)cmd)->mask; 1350 u_int32_t *hdr = (u_int32_t *)args->eh; 1351 1352 match = 1353 ( want[0] == (hdr[0] & mask[0]) && 1354 want[1] == (hdr[1] & mask[1]) && 1355 want[2] == (hdr[2] & mask[2]) ); 1356 } 1357 break; 1358 1359 case O_MAC_TYPE: 1360 if (args->eh != NULL) { 1361 u_int16_t *p = 1362 ((ipfw_insn_u16 *)cmd)->ports; 1363 int i; 1364 1365 for (i = cmdlen - 1; !match && i>0; 1366 i--, p += 2) 1367 match = (etype >= p[0] && 1368 etype <= p[1]); 1369 } 1370 break; 1371 1372 case O_FRAG: 1373 match = (offset != 0); 1374 break; 1375 1376 case O_IN: /* "out" is "not in" */ 1377 match = (oif == NULL); 1378 break; 1379 1380 case O_LAYER2: 1381 match = (args->eh != NULL); 1382 break; 1383 1384 case O_DIVERTED: 1385 match = (cmd->arg1 & 1 && divinput_flags & 1386 IP_FW_DIVERT_LOOPBACK_FLAG) || 1387 (cmd->arg1 & 2 && divinput_flags & 1388 IP_FW_DIVERT_OUTPUT_FLAG); 1389 break; 1390 1391 case O_PROTO: 1392 /* 1393 * We do not allow an arg of 0 so the 1394 * check of "proto" only suffices. 1395 */ 1396 match = (proto == cmd->arg1); 1397 break; 1398 1399 case O_IP_SRC: 1400 match = is_ipv4 && 1401 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1402 src_ip.s_addr); 1403 break; 1404 1405 case O_IP_SRC_LOOKUP: 1406 case O_IP_DST_LOOKUP: 1407 if (is_ipv4) { 1408 uint32_t a = 1409 (cmd->opcode == O_IP_DST_LOOKUP) ? 1410 dst_ip.s_addr : src_ip.s_addr; 1411 uint32_t v = 0; 1412 1413 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) { 1414 /* generic lookup */ 1415 v = ((ipfw_insn_u32 *)cmd)->d[1]; 1416 if (v == 0) 1417 a = dst_ip.s_addr; 1418 else if (v == 1) 1419 a = src_ip.s_addr; 1420 else if (offset != 0) 1421 break; 1422 else if (proto != IPPROTO_TCP && 1423 proto != IPPROTO_UDP) 1424 break; 1425 else if (v == 2) 1426 a = dst_port; 1427 else if (v == 3) 1428 a = src_port; 1429 else if (v == 4 || v == 5) { 1430 check_uidgid( 1431 (ipfw_insn_u32 *)cmd, 1432 proto, oif, 1433 dst_ip, dst_port, 1434 src_ip, src_port, &ucred_cache, 1435 &ucred_lookup, args->inp); 1436 if (v == 4 /* O_UID */) 1437 a = ucred_cache->cr_uid; 1438 else if (v == 5 /* O_JAIL */) 1439 a = ucred_cache->cr_prison->pr_id; 1440 } else 1441 break; 1442 } 1443 match = ipfw_lookup_table(chain, cmd->arg1, a, 1444 &v); 1445 if (!match) 1446 break; 1447 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 1448 match = 1449 ((ipfw_insn_u32 *)cmd)->d[0] == v; 1450 else 1451 tablearg = v; 1452 } 1453 break; 1454 1455 case O_IP_SRC_MASK: 1456 case O_IP_DST_MASK: 1457 if (is_ipv4) { 1458 uint32_t a = 1459 (cmd->opcode == O_IP_DST_MASK) ? 1460 dst_ip.s_addr : src_ip.s_addr; 1461 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 1462 int i = cmdlen-1; 1463 1464 for (; !match && i>0; i-= 2, p+= 2) 1465 match = (p[0] == (a & p[1])); 1466 } 1467 break; 1468 1469 case O_IP_SRC_ME: 1470 if (is_ipv4) { 1471 struct ifnet *tif; 1472 1473 INADDR_TO_IFP(src_ip, tif); 1474 match = (tif != NULL); 1475 } 1476 break; 1477 1478 case O_IP_DST_SET: 1479 case O_IP_SRC_SET: 1480 if (is_ipv4) { 1481 u_int32_t *d = (u_int32_t *)(cmd+1); 1482 u_int32_t addr = 1483 cmd->opcode == O_IP_DST_SET ? 1484 args->f_id.dst_ip : 1485 args->f_id.src_ip; 1486 1487 if (addr < d[0]) 1488 break; 1489 addr -= d[0]; /* subtract base */ 1490 match = (addr < cmd->arg1) && 1491 ( d[ 1 + (addr>>5)] & 1492 (1<<(addr & 0x1f)) ); 1493 } 1494 break; 1495 1496 case O_IP_DST: 1497 match = is_ipv4 && 1498 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1499 dst_ip.s_addr); 1500 break; 1501 1502 case O_IP_DST_ME: 1503 if (is_ipv4) { 1504 struct ifnet *tif; 1505 1506 INADDR_TO_IFP(dst_ip, tif); 1507 match = (tif != NULL); 1508 } 1509 break; 1510 1511 case O_IP_SRCPORT: 1512 case O_IP_DSTPORT: 1513 /* 1514 * offset == 0 && proto != 0 is enough 1515 * to guarantee that we have a 1516 * packet with port info. 1517 */ 1518 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 1519 && offset == 0) { 1520 u_int16_t x = 1521 (cmd->opcode == O_IP_SRCPORT) ? 1522 src_port : dst_port ; 1523 u_int16_t *p = 1524 ((ipfw_insn_u16 *)cmd)->ports; 1525 int i; 1526 1527 for (i = cmdlen - 1; !match && i>0; 1528 i--, p += 2) 1529 match = (x>=p[0] && x<=p[1]); 1530 } 1531 break; 1532 1533 case O_ICMPTYPE: 1534 match = (offset == 0 && proto==IPPROTO_ICMP && 1535 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 1536 break; 1537 1538#ifdef INET6 1539 case O_ICMP6TYPE: 1540 match = is_ipv6 && offset == 0 && 1541 proto==IPPROTO_ICMPV6 && 1542 icmp6type_match( 1543 ICMP6(ulp)->icmp6_type, 1544 (ipfw_insn_u32 *)cmd); 1545 break; 1546#endif /* INET6 */ 1547 1548 case O_IPOPT: 1549 match = (is_ipv4 && 1550 ipopts_match(ip, cmd) ); 1551 break; 1552 1553 case O_IPVER: 1554 match = (is_ipv4 && 1555 cmd->arg1 == ip->ip_v); 1556 break; 1557 1558 case O_IPID: 1559 case O_IPLEN: 1560 case O_IPTTL: 1561 if (is_ipv4) { /* only for IP packets */ 1562 uint16_t x; 1563 uint16_t *p; 1564 int i; 1565 1566 if (cmd->opcode == O_IPLEN) 1567 x = ip_len; 1568 else if (cmd->opcode == O_IPTTL) 1569 x = ip->ip_ttl; 1570 else /* must be IPID */ 1571 x = ntohs(ip->ip_id); 1572 if (cmdlen == 1) { 1573 match = (cmd->arg1 == x); 1574 break; 1575 } 1576 /* otherwise we have ranges */ 1577 p = ((ipfw_insn_u16 *)cmd)->ports; 1578 i = cmdlen - 1; 1579 for (; !match && i>0; i--, p += 2) 1580 match = (x >= p[0] && x <= p[1]); 1581 } 1582 break; 1583 1584 case O_IPPRECEDENCE: 1585 match = (is_ipv4 && 1586 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 1587 break; 1588 1589 case O_IPTOS: 1590 match = (is_ipv4 && 1591 flags_match(cmd, ip->ip_tos)); 1592 break; 1593 1594 case O_TCPDATALEN: 1595 if (proto == IPPROTO_TCP && offset == 0) { 1596 struct tcphdr *tcp; 1597 uint16_t x; 1598 uint16_t *p; 1599 int i; 1600 1601 tcp = TCP(ulp); 1602 x = ip_len - 1603 ((ip->ip_hl + tcp->th_off) << 2); 1604 if (cmdlen == 1) { 1605 match = (cmd->arg1 == x); 1606 break; 1607 } 1608 /* otherwise we have ranges */ 1609 p = ((ipfw_insn_u16 *)cmd)->ports; 1610 i = cmdlen - 1; 1611 for (; !match && i>0; i--, p += 2) 1612 match = (x >= p[0] && x <= p[1]); 1613 } 1614 break; 1615 1616 case O_TCPFLAGS: 1617 match = (proto == IPPROTO_TCP && offset == 0 && 1618 flags_match(cmd, TCP(ulp)->th_flags)); 1619 break; 1620 1621 case O_TCPOPTS: 1622 match = (proto == IPPROTO_TCP && offset == 0 && 1623 tcpopts_match(TCP(ulp), cmd)); 1624 break; 1625 1626 case O_TCPSEQ: 1627 match = (proto == IPPROTO_TCP && offset == 0 && 1628 ((ipfw_insn_u32 *)cmd)->d[0] == 1629 TCP(ulp)->th_seq); 1630 break; 1631 1632 case O_TCPACK: 1633 match = (proto == IPPROTO_TCP && offset == 0 && 1634 ((ipfw_insn_u32 *)cmd)->d[0] == 1635 TCP(ulp)->th_ack); 1636 break; 1637 1638 case O_TCPWIN: 1639 match = (proto == IPPROTO_TCP && offset == 0 && 1640 cmd->arg1 == TCP(ulp)->th_win); 1641 break; 1642 1643 case O_ESTAB: 1644 /* reject packets which have SYN only */ 1645 /* XXX should i also check for TH_ACK ? */ 1646 match = (proto == IPPROTO_TCP && offset == 0 && 1647 (TCP(ulp)->th_flags & 1648 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 1649 break; 1650 1651 case O_ALTQ: { 1652 struct pf_mtag *at; 1653 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 1654 1655 match = 1; 1656 at = pf_find_mtag(m); 1657 if (at != NULL && at->qid != 0) 1658 break; 1659 at = pf_get_mtag(m); 1660 if (at == NULL) { 1661 /* 1662 * Let the packet fall back to the 1663 * default ALTQ. 1664 */ 1665 break; 1666 } 1667 at->qid = altq->qid; 1668 if (is_ipv4) 1669 at->af = AF_INET; 1670 else 1671 at->af = AF_LINK; 1672 at->hdr = ip; 1673 break; 1674 } 1675 1676 case O_LOG: 1677 if (V_fw_verbose) 1678 ipfw_log(f, hlen, args, m, 1679 oif, offset, tablearg, ip); 1680 match = 1; 1681 break; 1682 1683 case O_PROB: 1684 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 1685 break; 1686 1687 case O_VERREVPATH: 1688 /* Outgoing packets automatically pass/match */ 1689 match = ((oif != NULL) || 1690 (m->m_pkthdr.rcvif == NULL) || 1691 ( 1692#ifdef INET6 1693 is_ipv6 ? 1694 verify_path6(&(args->f_id.src_ip6), 1695 m->m_pkthdr.rcvif) : 1696#endif 1697 verify_path(src_ip, m->m_pkthdr.rcvif, 1698 args->f_id.fib))); 1699 break; 1700 1701 case O_VERSRCREACH: 1702 /* Outgoing packets automatically pass/match */ 1703 match = (hlen > 0 && ((oif != NULL) || 1704#ifdef INET6 1705 is_ipv6 ? 1706 verify_path6(&(args->f_id.src_ip6), 1707 NULL) : 1708#endif 1709 verify_path(src_ip, NULL, args->f_id.fib))); 1710 break; 1711 1712 case O_ANTISPOOF: 1713 /* Outgoing packets automatically pass/match */ 1714 if (oif == NULL && hlen > 0 && 1715 ( (is_ipv4 && in_localaddr(src_ip)) 1716#ifdef INET6 1717 || (is_ipv6 && 1718 in6_localaddr(&(args->f_id.src_ip6))) 1719#endif 1720 )) 1721 match = 1722#ifdef INET6 1723 is_ipv6 ? verify_path6( 1724 &(args->f_id.src_ip6), 1725 m->m_pkthdr.rcvif) : 1726#endif 1727 verify_path(src_ip, 1728 m->m_pkthdr.rcvif, 1729 args->f_id.fib); 1730 else 1731 match = 1; 1732 break; 1733 1734 case O_IPSEC: 1735#ifdef IPSEC 1736 match = (m_tag_find(m, 1737 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 1738#endif 1739 /* otherwise no match */ 1740 break; 1741 1742#ifdef INET6 1743 case O_IP6_SRC: 1744 match = is_ipv6 && 1745 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 1746 &((ipfw_insn_ip6 *)cmd)->addr6); 1747 break; 1748 1749 case O_IP6_DST: 1750 match = is_ipv6 && 1751 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 1752 &((ipfw_insn_ip6 *)cmd)->addr6); 1753 break; 1754 case O_IP6_SRC_MASK: 1755 case O_IP6_DST_MASK: 1756 if (is_ipv6) { 1757 int i = cmdlen - 1; 1758 struct in6_addr p; 1759 struct in6_addr *d = 1760 &((ipfw_insn_ip6 *)cmd)->addr6; 1761 1762 for (; !match && i > 0; d += 2, 1763 i -= F_INSN_SIZE(struct in6_addr) 1764 * 2) { 1765 p = (cmd->opcode == 1766 O_IP6_SRC_MASK) ? 1767 args->f_id.src_ip6: 1768 args->f_id.dst_ip6; 1769 APPLY_MASK(&p, &d[1]); 1770 match = 1771 IN6_ARE_ADDR_EQUAL(&d[0], 1772 &p); 1773 } 1774 } 1775 break; 1776 1777 case O_IP6_SRC_ME: 1778 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 1779 break; 1780 1781 case O_IP6_DST_ME: 1782 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 1783 break; 1784 1785 case O_FLOW6ID: 1786 match = is_ipv6 && 1787 flow6id_match(args->f_id.flow_id6, 1788 (ipfw_insn_u32 *) cmd); 1789 break; 1790 1791 case O_EXT_HDR: 1792 match = is_ipv6 && 1793 (ext_hd & ((ipfw_insn *) cmd)->arg1); 1794 break; 1795 1796 case O_IP6: 1797 match = is_ipv6; 1798 break; 1799#endif 1800 1801 case O_IP4: 1802 match = is_ipv4; 1803 break; 1804 1805 case O_TAG: { 1806 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 1807 tablearg : cmd->arg1; 1808 1809 /* Packet is already tagged with this tag? */ 1810 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 1811 1812 /* We have `untag' action when F_NOT flag is 1813 * present. And we must remove this mtag from 1814 * mbuf and reset `match' to zero (`match' will 1815 * be inversed later). 1816 * Otherwise we should allocate new mtag and 1817 * push it into mbuf. 1818 */ 1819 if (cmd->len & F_NOT) { /* `untag' action */ 1820 if (mtag != NULL) 1821 m_tag_delete(m, mtag); 1822 } else if (mtag == NULL) { 1823 if ((mtag = m_tag_alloc(MTAG_IPFW, 1824 tag, 0, M_NOWAIT)) != NULL) 1825 m_tag_prepend(m, mtag); 1826 } 1827 match = (cmd->len & F_NOT) ? 0: 1; 1828 break; 1829 } 1830 1831 case O_FIB: /* try match the specified fib */ 1832 if (args->f_id.fib == cmd->arg1) 1833 match = 1; 1834 break; 1835 1836 case O_TAGGED: { 1837 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 1838 tablearg : cmd->arg1; 1839 1840 if (cmdlen == 1) { 1841 match = m_tag_locate(m, MTAG_IPFW, 1842 tag, NULL) != NULL; 1843 break; 1844 } 1845 1846 /* we have ranges */ 1847 for (mtag = m_tag_first(m); 1848 mtag != NULL && !match; 1849 mtag = m_tag_next(m, mtag)) { 1850 uint16_t *p; 1851 int i; 1852 1853 if (mtag->m_tag_cookie != MTAG_IPFW) 1854 continue; 1855 1856 p = ((ipfw_insn_u16 *)cmd)->ports; 1857 i = cmdlen - 1; 1858 for(; !match && i > 0; i--, p += 2) 1859 match = 1860 mtag->m_tag_id >= p[0] && 1861 mtag->m_tag_id <= p[1]; 1862 } 1863 break; 1864 } 1865 1866 /* 1867 * The second set of opcodes represents 'actions', 1868 * i.e. the terminal part of a rule once the packet 1869 * matches all previous patterns. 1870 * Typically there is only one action for each rule, 1871 * and the opcode is stored at the end of the rule 1872 * (but there are exceptions -- see below). 1873 * 1874 * In general, here we set retval and terminate the 1875 * outer loop (would be a 'break 3' in some language, 1876 * but we need to set l=0, done=1) 1877 * 1878 * Exceptions: 1879 * O_COUNT and O_SKIPTO actions: 1880 * instead of terminating, we jump to the next rule 1881 * (setting l=0), or to the SKIPTO target (by 1882 * setting f, cmd and l as needed), respectively. 1883 * 1884 * O_TAG, O_LOG and O_ALTQ action parameters: 1885 * perform some action and set match = 1; 1886 * 1887 * O_LIMIT and O_KEEP_STATE: these opcodes are 1888 * not real 'actions', and are stored right 1889 * before the 'action' part of the rule. 1890 * These opcodes try to install an entry in the 1891 * state tables; if successful, we continue with 1892 * the next opcode (match=1; break;), otherwise 1893 * the packet must be dropped (set retval, 1894 * break loops with l=0, done=1) 1895 * 1896 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 1897 * cause a lookup of the state table, and a jump 1898 * to the 'action' part of the parent rule 1899 * if an entry is found, or 1900 * (CHECK_STATE only) a jump to the next rule if 1901 * the entry is not found. 1902 * The result of the lookup is cached so that 1903 * further instances of these opcodes become NOPs. 1904 * The jump to the next rule is done by setting 1905 * l=0, cmdlen=0. 1906 */ 1907 case O_LIMIT: 1908 case O_KEEP_STATE: 1909 if (ipfw_install_state(f, 1910 (ipfw_insn_limit *)cmd, args, tablearg)) { 1911 /* error or limit violation */ 1912 retval = IP_FW_DENY; 1913 l = 0; /* exit inner loop */ 1914 done = 1; /* exit outer loop */ 1915 } 1916 match = 1; 1917 break; 1918 1919 case O_PROBE_STATE: 1920 case O_CHECK_STATE: 1921 /* 1922 * dynamic rules are checked at the first 1923 * keep-state or check-state occurrence, 1924 * with the result being stored in dyn_dir. 1925 * The compiler introduces a PROBE_STATE 1926 * instruction for us when we have a 1927 * KEEP_STATE (because PROBE_STATE needs 1928 * to be run first). 1929 */ 1930 if (dyn_dir == MATCH_UNKNOWN && 1931 (q = ipfw_lookup_dyn_rule(&args->f_id, 1932 &dyn_dir, proto == IPPROTO_TCP ? 1933 TCP(ulp) : NULL)) 1934 != NULL) { 1935 /* 1936 * Found dynamic entry, update stats 1937 * and jump to the 'action' part of 1938 * the parent rule by setting 1939 * f, cmd, l and clearing cmdlen. 1940 */ 1941 q->pcnt++; 1942 q->bcnt += pktlen; 1943 f = q->rule; 1944 cmd = ACTION_PTR(f); 1945 l = f->cmd_len - f->act_ofs; 1946 ipfw_dyn_unlock(); 1947 cmdlen = 0; 1948 match = 1; 1949 break; 1950 } 1951 /* 1952 * Dynamic entry not found. If CHECK_STATE, 1953 * skip to next rule, if PROBE_STATE just 1954 * ignore and continue with next opcode. 1955 */ 1956 if (cmd->opcode == O_CHECK_STATE) 1957 l = 0; /* exit inner loop */ 1958 match = 1; 1959 break; 1960 1961 case O_ACCEPT: 1962 retval = 0; /* accept */ 1963 l = 0; /* exit inner loop */ 1964 done = 1; /* exit outer loop */ 1965 break; 1966 1967 case O_PIPE: 1968 case O_QUEUE: 1969 args->rule = f; /* report matching rule */ 1970 args->rule_id = f->id; 1971 args->chain_id = chain->id; 1972 if (cmd->arg1 == IP_FW_TABLEARG) 1973 args->cookie = tablearg; 1974 else 1975 args->cookie = cmd->arg1; 1976 retval = IP_FW_DUMMYNET; 1977 l = 0; /* exit inner loop */ 1978 done = 1; /* exit outer loop */ 1979 break; 1980 1981 case O_DIVERT: 1982 case O_TEE: 1983 if (args->eh) /* not on layer 2 */ 1984 break; 1985 /* otherwise this is terminal */ 1986 l = 0; /* exit inner loop */ 1987 done = 1; /* exit outer loop */ 1988 mtag = m_tag_get(PACKET_TAG_DIVERT, 1989 sizeof(struct divert_tag), 1990 M_NOWAIT); 1991 if (mtag == NULL) { 1992 retval = IP_FW_DENY; 1993 } else { 1994 struct divert_tag *dt; 1995 dt = (struct divert_tag *)(mtag+1); 1996 dt->cookie = f->rulenum; 1997 if (cmd->arg1 == IP_FW_TABLEARG) 1998 dt->info = tablearg; 1999 else 2000 dt->info = cmd->arg1; 2001 m_tag_prepend(m, mtag); 2002 retval = (cmd->opcode == O_DIVERT) ? 2003 IP_FW_DIVERT : IP_FW_TEE; 2004 } 2005 break; 2006 2007 case O_COUNT: 2008 case O_SKIPTO: 2009 f->pcnt++; /* update stats */ 2010 f->bcnt += pktlen; 2011 f->timestamp = time_uptime; 2012 if (cmd->opcode == O_COUNT) { 2013 l = 0; /* exit inner loop */ 2014 break; 2015 } 2016 /* handle skipto */ 2017 if (cmd->arg1 == IP_FW_TABLEARG) { 2018 f = lookup_next_rule(f, tablearg); 2019 } else { 2020 if (f->next_rule == NULL) 2021 lookup_next_rule(f, 0); 2022 f = f->next_rule; 2023 } 2024 /* 2025 * Skip disabled rules, and 2026 * re-enter the inner loop 2027 * with the correct f, l and cmd. 2028 * Also clear cmdlen and skip_or 2029 */ 2030 while (f && (V_set_disable & (1 << f->set))) 2031 f = f->next; 2032 if (f) { /* found a valid rule */ 2033 l = f->cmd_len; 2034 cmd = f->cmd; 2035 } else { 2036 l = 0; /* exit inner loop */ 2037 } 2038 match = 1; 2039 cmdlen = 0; 2040 skip_or = 0; 2041 break; 2042 2043 case O_REJECT: 2044 /* 2045 * Drop the packet and send a reject notice 2046 * if the packet is not ICMP (or is an ICMP 2047 * query), and it is not multicast/broadcast. 2048 */ 2049 if (hlen > 0 && is_ipv4 && offset == 0 && 2050 (proto != IPPROTO_ICMP || 2051 is_icmp_query(ICMP(ulp))) && 2052 !(m->m_flags & (M_BCAST|M_MCAST)) && 2053 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 2054 send_reject(args, cmd->arg1, ip_len, ip); 2055 m = args->m; 2056 } 2057 /* FALLTHROUGH */ 2058#ifdef INET6 2059 case O_UNREACH6: 2060 if (hlen > 0 && is_ipv6 && 2061 ((offset & IP6F_OFF_MASK) == 0) && 2062 (proto != IPPROTO_ICMPV6 || 2063 (is_icmp6_query(args->f_id.flags) == 1)) && 2064 !(m->m_flags & (M_BCAST|M_MCAST)) && 2065 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 2066 send_reject6( 2067 args, cmd->arg1, hlen, 2068 (struct ip6_hdr *)ip); 2069 m = args->m; 2070 } 2071 /* FALLTHROUGH */ 2072#endif 2073 case O_DENY: 2074 retval = IP_FW_DENY; 2075 l = 0; /* exit inner loop */ 2076 done = 1; /* exit outer loop */ 2077 break; 2078 2079 case O_FORWARD_IP: 2080 if (args->eh) /* not valid on layer2 pkts */ 2081 break; 2082 if (!q || dyn_dir == MATCH_FORWARD) { 2083 struct sockaddr_in *sa; 2084 sa = &(((ipfw_insn_sa *)cmd)->sa); 2085 if (sa->sin_addr.s_addr == INADDR_ANY) { 2086 bcopy(sa, &args->hopstore, 2087 sizeof(*sa)); 2088 args->hopstore.sin_addr.s_addr = 2089 htonl(tablearg); 2090 args->next_hop = &args->hopstore; 2091 } else { 2092 args->next_hop = sa; 2093 } 2094 } 2095 retval = IP_FW_PASS; 2096 l = 0; /* exit inner loop */ 2097 done = 1; /* exit outer loop */ 2098 break; 2099 2100 case O_NETGRAPH: 2101 case O_NGTEE: 2102 args->rule = f; /* report matching rule */ 2103 args->rule_id = f->id; 2104 args->chain_id = chain->id; 2105 if (cmd->arg1 == IP_FW_TABLEARG) 2106 args->cookie = tablearg; 2107 else 2108 args->cookie = cmd->arg1; 2109 retval = (cmd->opcode == O_NETGRAPH) ? 2110 IP_FW_NETGRAPH : IP_FW_NGTEE; 2111 l = 0; /* exit inner loop */ 2112 done = 1; /* exit outer loop */ 2113 break; 2114 2115 case O_SETFIB: 2116 f->pcnt++; /* update stats */ 2117 f->bcnt += pktlen; 2118 f->timestamp = time_uptime; 2119 M_SETFIB(m, cmd->arg1); 2120 args->f_id.fib = cmd->arg1; 2121 l = 0; /* exit inner loop */ 2122 break; 2123 2124 case O_NAT: 2125 if (!IPFW_NAT_LOADED) { 2126 retval = IP_FW_DENY; 2127 } else { 2128 struct cfg_nat *t; 2129 int nat_id; 2130 2131 args->rule = f; /* Report matching rule. */ 2132 args->rule_id = f->id; 2133 args->chain_id = chain->id; 2134 t = ((ipfw_insn_nat *)cmd)->nat; 2135 if (t == NULL) { 2136 nat_id = (cmd->arg1 == IP_FW_TABLEARG) ? 2137 tablearg : cmd->arg1; 2138 t = (*lookup_nat_ptr)(&V_layer3_chain.nat, nat_id); 2139 2140 if (t == NULL) { 2141 retval = IP_FW_DENY; 2142 l = 0; /* exit inner loop */ 2143 done = 1; /* exit outer loop */ 2144 break; 2145 } 2146 if (cmd->arg1 != IP_FW_TABLEARG) 2147 ((ipfw_insn_nat *)cmd)->nat = t; 2148 } 2149 retval = ipfw_nat_ptr(args, t, m); 2150 } 2151 l = 0; /* exit inner loop */ 2152 done = 1; /* exit outer loop */ 2153 break; 2154 2155 case O_REASS: { 2156 int ip_off; 2157 2158 f->pcnt++; 2159 f->bcnt += pktlen; 2160 l = 0; /* in any case exit inner loop */ 2161 2162 ip_off = (args->eh != NULL) ? 2163 ntohs(ip->ip_off) : ip->ip_off; 2164 /* if not fragmented, go to next rule */ 2165 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0) 2166 break; 2167 /* 2168 * ip_reass() expects len & off in host 2169 * byte order: fix them in case we come 2170 * from layer2. 2171 */ 2172 if (args->eh != NULL) { 2173 ip->ip_len = ntohs(ip->ip_len); 2174 ip->ip_off = ntohs(ip->ip_off); 2175 } 2176 2177 args->m = m = ip_reass(m); 2178 2179 /* 2180 * IP header checksum fixup after 2181 * reassembly and leave header 2182 * in network byte order. 2183 */ 2184 if (m == NULL) { /* fragment got swallowed */ 2185 retval = IP_FW_DENY; 2186 } else { /* good, packet complete */ 2187 int hlen; 2188 2189 ip = mtod(m, struct ip *); 2190 hlen = ip->ip_hl << 2; 2191 /* revert len & off for layer2 pkts */ 2192 if (args->eh != NULL) 2193 ip->ip_len = htons(ip->ip_len); 2194 ip->ip_sum = 0; 2195 if (hlen == sizeof(struct ip)) 2196 ip->ip_sum = in_cksum_hdr(ip); 2197 else 2198 ip->ip_sum = in_cksum(m, hlen); 2199 retval = IP_FW_REASS; 2200 args->rule = f; 2201 args->rule_id = f->id; 2202 args->chain_id = chain->id; 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 /* Update statistics */ 2237 f->pcnt++; 2238 f->bcnt += pktlen; 2239 f->timestamp = time_uptime; 2240 } else { 2241 retval = IP_FW_DENY; 2242 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 2243 } 2244 IPFW_RUNLOCK(chain); 2245 if (ucred_cache != NULL) 2246 crfree(ucred_cache); 2247 return (retval); 2248 2249pullup_failed: 2250 if (V_fw_verbose) 2251 printf("ipfw: pullup failed\n"); 2252 return (IP_FW_DENY); 2253} 2254 2255/* 2256 * Module and VNET glue 2257 */ 2258 2259/* 2260 * Stuff that must be initialised only on boot or module load 2261 */ 2262static int 2263ipfw_init(void) 2264{ 2265 int error = 0; 2266 2267 ipfw_dyn_attach(); 2268 /* 2269 * Only print out this stuff the first time around, 2270 * when called from the sysinit code. 2271 */ 2272 printf("ipfw2 " 2273#ifdef INET6 2274 "(+ipv6) " 2275#endif 2276 "initialized, divert %s, nat %s, " 2277 "rule-based forwarding " 2278#ifdef IPFIREWALL_FORWARD 2279 "enabled, " 2280#else 2281 "disabled, " 2282#endif 2283 "default to %s, logging ", 2284#ifdef IPDIVERT 2285 "enabled", 2286#else 2287 "loadable", 2288#endif 2289#ifdef IPFIREWALL_NAT 2290 "enabled", 2291#else 2292 "loadable", 2293#endif 2294 default_to_accept ? "accept" : "deny"); 2295 2296 /* 2297 * Note: V_xxx variables can be accessed here but the vnet specific 2298 * initializer may not have been called yet for the VIMAGE case. 2299 * Tuneables will have been processed. We will print out values for 2300 * the default vnet. 2301 * XXX This should all be rationalized AFTER 8.0 2302 */ 2303 if (V_fw_verbose == 0) 2304 printf("disabled\n"); 2305 else if (V_verbose_limit == 0) 2306 printf("unlimited\n"); 2307 else 2308 printf("limited to %d packets/entry by default\n", 2309 V_verbose_limit); 2310 2311 return (error); 2312} 2313 2314/* 2315 * Called for the removal of the last instance only on module unload. 2316 */ 2317static void 2318ipfw_destroy(void) 2319{ 2320 2321 ipfw_dyn_detach(); 2322 printf("IP firewall unloaded\n"); 2323} 2324 2325/* 2326 * Stuff that must be initialized for every instance 2327 * (including the first of course). 2328 */ 2329static int 2330vnet_ipfw_init(const void *unused) 2331{ 2332 int error; 2333 struct ip_fw default_rule; 2334 2335 /* First set up some values that are compile time options */ 2336#ifdef IPFIREWALL_VERBOSE 2337 V_fw_verbose = 1; 2338#endif 2339#ifdef IPFIREWALL_VERBOSE_LIMIT 2340 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 2341#endif 2342 2343 error = ipfw_init_tables(&V_layer3_chain); 2344 if (error) { 2345 panic("init_tables"); /* XXX Marko fix this ! */ 2346 } 2347#ifdef IPFIREWALL_NAT 2348 LIST_INIT(&V_layer3_chain.nat); 2349#endif 2350 2351 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 2352 2353 V_fw_deny_unknown_exthdrs = 1; 2354 2355 V_layer3_chain.rules = NULL; 2356 IPFW_LOCK_INIT(&V_layer3_chain); 2357 2358 bzero(&default_rule, sizeof default_rule); 2359 default_rule.act_ofs = 0; 2360 default_rule.rulenum = IPFW_DEFAULT_RULE; 2361 default_rule.cmd_len = 1; 2362 default_rule.set = RESVD_SET; 2363 default_rule.cmd[0].len = 1; 2364 default_rule.cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; 2365 error = ipfw_add_rule(&V_layer3_chain, &default_rule); 2366 2367 if (error != 0) { 2368 printf("ipfw2: error %u initializing default rule " 2369 "(support disabled)\n", error); 2370 IPFW_LOCK_DESTROY(&V_layer3_chain); 2371 printf("leaving ipfw_iattach (1) with error %d\n", error); 2372 return (error); 2373 } 2374 2375 V_layer3_chain.default_rule = V_layer3_chain.rules; 2376 2377 ipfw_dyn_init(); 2378 2379 /* First set up some values that are compile time options */ 2380 V_ipfw_vnet_ready = 1; /* Open for business */ 2381 2382 /* 2383 * Hook the sockopt handler, and the layer2 (V_ip_fw_chk_ptr) 2384 * and pfil hooks for ipv4 and ipv6. Even if the latter two fail 2385 * we still keep the module alive because the sockopt and 2386 * layer2 paths are still useful. 2387 * ipfw[6]_hook return 0 on success, ENOENT on failure, 2388 * so we can ignore the exact return value and just set a flag. 2389 * 2390 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so 2391 * changes in the underlying (per-vnet) variables trigger 2392 * immediate hook()/unhook() calls. 2393 * In layer2 we have the same behaviour, except that V_ether_ipfw 2394 * is checked on each packet because there are no pfil hooks. 2395 */ 2396 V_ip_fw_ctl_ptr = ipfw_ctl; 2397 V_ip_fw_chk_ptr = ipfw_chk; 2398 error = ipfw_attach_hooks(); 2399 return (error); 2400} 2401 2402/* 2403 * Called for the removal of each instance. 2404 */ 2405static int 2406vnet_ipfw_uninit(const void *unused) 2407{ 2408 struct ip_fw *reap; 2409 2410 V_ipfw_vnet_ready = 0; /* tell new callers to go away */ 2411 /* 2412 * disconnect from ipv4, ipv6, layer2 and sockopt. 2413 * Then grab, release and grab again the WLOCK so we make 2414 * sure the update is propagated and nobody will be in. 2415 */ 2416 ipfw_unhook(); 2417#ifdef INET6 2418 ipfw6_unhook(); 2419#endif 2420 V_ip_fw_chk_ptr = NULL; 2421 V_ip_fw_ctl_ptr = NULL; 2422 2423 IPFW_WLOCK(&V_layer3_chain); 2424 /* We wait on the wlock here until the last user leaves */ 2425 IPFW_WUNLOCK(&V_layer3_chain); 2426 IPFW_WLOCK(&V_layer3_chain); 2427 2428 ipfw_dyn_uninit(0); /* run the callout_drain */ 2429 ipfw_flush_tables(&V_layer3_chain); 2430 V_layer3_chain.reap = NULL; 2431 ipfw_free_chain(&V_layer3_chain, 1 /* kill default rule */); 2432 reap = V_layer3_chain.reap; 2433 V_layer3_chain.reap = NULL; 2434 IPFW_WUNLOCK(&V_layer3_chain); 2435 if (reap != NULL) 2436 ipfw_reap_rules(reap); 2437 IPFW_LOCK_DESTROY(&V_layer3_chain); 2438 ipfw_dyn_uninit(1); /* free the remaining parts */ 2439 return 0; 2440} 2441 2442/* 2443 * Module event handler. 2444 * In general we have the choice of handling most of these events by the 2445 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to 2446 * use the SYSINIT handlers as they are more capable of expressing the 2447 * flow of control during module and vnet operations, so this is just 2448 * a skeleton. Note there is no SYSINIT equivalent of the module 2449 * SHUTDOWN handler, but we don't have anything to do in that case anyhow. 2450 */ 2451static int 2452ipfw_modevent(module_t mod, int type, void *unused) 2453{ 2454 int err = 0; 2455 2456 switch (type) { 2457 case MOD_LOAD: 2458 /* Called once at module load or 2459 * system boot if compiled in. */ 2460 break; 2461 case MOD_QUIESCE: 2462 /* Called before unload. May veto unloading. */ 2463 break; 2464 case MOD_UNLOAD: 2465 /* Called during unload. */ 2466 break; 2467 case MOD_SHUTDOWN: 2468 /* Called during system shutdown. */ 2469 break; 2470 default: 2471 err = EOPNOTSUPP; 2472 break; 2473 } 2474 return err; 2475} 2476 2477static moduledata_t ipfwmod = { 2478 "ipfw", 2479 ipfw_modevent, 2480 0 2481}; 2482 2483/* Define startup order. */ 2484#define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_IFATTACHDOMAIN 2485#define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */ 2486#define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */ 2487#define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */ 2488 2489DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER); 2490MODULE_VERSION(ipfw, 2); 2491/* should declare some dependencies here */ 2492 2493/* 2494 * Starting up. Done in order after ipfwmod() has been called. 2495 * VNET_SYSINIT is also called for each existing vnet and each new vnet. 2496 */ 2497SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 2498 ipfw_init, NULL); 2499VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 2500 vnet_ipfw_init, NULL); 2501 2502/* 2503 * Closing up shop. These are done in REVERSE ORDER, but still 2504 * after ipfwmod() has been called. Not called on reboot. 2505 * VNET_SYSUNINIT is also called for each exiting vnet as it exits. 2506 * or when the module is unloaded. 2507 */ 2508SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 2509 ipfw_destroy, NULL); 2510VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 2511 vnet_ipfw_uninit, NULL); 2512/* end of file */ 2513