ip_fw2.c revision 222582
1/*- 2 * Copyright (c) 2002-2009 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 222582 2011-06-01 19:44:52Z ae $"); 28 29/* 30 * The FreeBSD IP packet firewall, main file 31 */ 32 33#if !defined(KLD_MODULE) 34#include "opt_ipfw.h" 35#include "opt_ipdivert.h" 36#include "opt_ipdn.h" 37#include "opt_inet.h" 38#ifndef INET 39#error IPFIREWALL requires INET. 40#endif /* INET */ 41#endif 42#include "opt_inet6.h" 43#include "opt_ipsec.h" 44 45#include <sys/param.h> 46#include <sys/systm.h> 47#include <sys/condvar.h> 48#include <sys/eventhandler.h> 49#include <sys/malloc.h> 50#include <sys/mbuf.h> 51#include <sys/kernel.h> 52#include <sys/lock.h> 53#include <sys/jail.h> 54#include <sys/module.h> 55#include <sys/priv.h> 56#include <sys/proc.h> 57#include <sys/rwlock.h> 58#include <sys/socket.h> 59#include <sys/socketvar.h> 60#include <sys/sysctl.h> 61#include <sys/syslog.h> 62#include <sys/ucred.h> 63#include <net/ethernet.h> /* for ETHERTYPE_IP */ 64#include <net/if.h> 65#include <net/route.h> 66#include <net/pf_mtag.h> 67#include <net/vnet.h> 68 69#include <netinet/in.h> 70#include <netinet/in_var.h> 71#include <netinet/in_pcb.h> 72#include <netinet/ip.h> 73#include <netinet/ip_var.h> 74#include <netinet/ip_icmp.h> 75#include <netinet/ip_fw.h> 76#include <netinet/ipfw/ip_fw_private.h> 77#include <netinet/ip_carp.h> 78#include <netinet/pim.h> 79#include <netinet/tcp_var.h> 80#include <netinet/udp.h> 81#include <netinet/udp_var.h> 82#include <netinet/sctp.h> 83 84#include <netinet/ip6.h> 85#include <netinet/icmp6.h> 86#ifdef INET6 87#include <netinet6/scope6_var.h> 88#include <netinet6/ip6_var.h> 89#endif 90 91#include <machine/in_cksum.h> /* XXX for in_cksum */ 92 93#ifdef MAC 94#include <security/mac/mac_framework.h> 95#endif 96 97/* 98 * static variables followed by global ones. 99 * All ipfw global variables are here. 100 */ 101 102/* ipfw_vnet_ready controls when we are open for business */ 103static VNET_DEFINE(int, ipfw_vnet_ready) = 0; 104#define V_ipfw_vnet_ready VNET(ipfw_vnet_ready) 105 106static VNET_DEFINE(int, fw_deny_unknown_exthdrs); 107#define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs) 108 109#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 110static int default_to_accept = 1; 111#else 112static int default_to_accept; 113#endif 114 115VNET_DEFINE(int, autoinc_step); 116VNET_DEFINE(int, fw_one_pass) = 1; 117 118/* 119 * Each rule belongs to one of 32 different sets (0..31). 120 * The variable set_disable contains one bit per set. 121 * If the bit is set, all rules in the corresponding set 122 * are disabled. Set RESVD_SET(31) is reserved for the default rule 123 * and rules that are not deleted by the flush command, 124 * and CANNOT be disabled. 125 * Rules in set RESVD_SET can only be deleted individually. 126 */ 127VNET_DEFINE(u_int32_t, set_disable); 128#define V_set_disable VNET(set_disable) 129 130VNET_DEFINE(int, fw_verbose); 131/* counter for ipfw_log(NULL...) */ 132VNET_DEFINE(u_int64_t, norule_counter); 133VNET_DEFINE(int, verbose_limit); 134 135/* layer3_chain contains the list of rules for layer 3 */ 136VNET_DEFINE(struct ip_fw_chain, layer3_chain); 137 138ipfw_nat_t *ipfw_nat_ptr = NULL; 139struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int); 140ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; 141ipfw_nat_cfg_t *ipfw_nat_del_ptr; 142ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; 143ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; 144 145#ifdef SYSCTL_NODE 146uint32_t dummy_def = IPFW_DEFAULT_RULE; 147uint32_t dummy_tables_max = IPFW_TABLES_MAX; 148 149SYSBEGIN(f3) 150 151SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 152SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 153 CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0, 154 "Only do a single pass through ipfw when using dummynet(4)"); 155SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, 156 CTLFLAG_RW, &VNET_NAME(autoinc_step), 0, 157 "Rule number auto-increment step"); 158SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose, 159 CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0, 160 "Log matches to ipfw rules"); 161SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, 162 CTLFLAG_RW, &VNET_NAME(verbose_limit), 0, 163 "Set upper limit of matches of ipfw rules logged"); 164SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, 165 &dummy_def, 0, 166 "The default/max possible rule number."); 167SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, tables_max, CTLFLAG_RD, 168 &dummy_tables_max, 0, 169 "The maximum number of tables."); 170SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN, 171 &default_to_accept, 0, 172 "Make the default rule accept all packets."); 173TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept); 174SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, static_count, 175 CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0, 176 "Number of static rules"); 177 178#ifdef INET6 179SYSCTL_DECL(_net_inet6_ip6); 180SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 181SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs, 182 CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0, 183 "Deny packets with unknown IPv6 Extension Headers"); 184#endif /* INET6 */ 185 186SYSEND 187 188#endif /* SYSCTL_NODE */ 189 190 191/* 192 * Some macros used in the various matching options. 193 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 194 * Other macros just cast void * into the appropriate type 195 */ 196#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 197#define TCP(p) ((struct tcphdr *)(p)) 198#define SCTP(p) ((struct sctphdr *)(p)) 199#define UDP(p) ((struct udphdr *)(p)) 200#define ICMP(p) ((struct icmphdr *)(p)) 201#define ICMP6(p) ((struct icmp6_hdr *)(p)) 202 203static __inline int 204icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 205{ 206 int type = icmp->icmp_type; 207 208 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 209} 210 211#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 212 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 213 214static int 215is_icmp_query(struct icmphdr *icmp) 216{ 217 int type = icmp->icmp_type; 218 219 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 220} 221#undef TT 222 223/* 224 * The following checks use two arrays of 8 or 16 bits to store the 225 * bits that we want set or clear, respectively. They are in the 226 * low and high half of cmd->arg1 or cmd->d[0]. 227 * 228 * We scan options and store the bits we find set. We succeed if 229 * 230 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 231 * 232 * The code is sometimes optimized not to store additional variables. 233 */ 234 235static int 236flags_match(ipfw_insn *cmd, u_int8_t bits) 237{ 238 u_char want_clear; 239 bits = ~bits; 240 241 if ( ((cmd->arg1 & 0xff) & bits) != 0) 242 return 0; /* some bits we want set were clear */ 243 want_clear = (cmd->arg1 >> 8) & 0xff; 244 if ( (want_clear & bits) != want_clear) 245 return 0; /* some bits we want clear were set */ 246 return 1; 247} 248 249static int 250ipopts_match(struct ip *ip, ipfw_insn *cmd) 251{ 252 int optlen, bits = 0; 253 u_char *cp = (u_char *)(ip + 1); 254 int x = (ip->ip_hl << 2) - sizeof (struct ip); 255 256 for (; x > 0; x -= optlen, cp += optlen) { 257 int opt = cp[IPOPT_OPTVAL]; 258 259 if (opt == IPOPT_EOL) 260 break; 261 if (opt == IPOPT_NOP) 262 optlen = 1; 263 else { 264 optlen = cp[IPOPT_OLEN]; 265 if (optlen <= 0 || optlen > x) 266 return 0; /* invalid or truncated */ 267 } 268 switch (opt) { 269 270 default: 271 break; 272 273 case IPOPT_LSRR: 274 bits |= IP_FW_IPOPT_LSRR; 275 break; 276 277 case IPOPT_SSRR: 278 bits |= IP_FW_IPOPT_SSRR; 279 break; 280 281 case IPOPT_RR: 282 bits |= IP_FW_IPOPT_RR; 283 break; 284 285 case IPOPT_TS: 286 bits |= IP_FW_IPOPT_TS; 287 break; 288 } 289 } 290 return (flags_match(cmd, bits)); 291} 292 293static int 294tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 295{ 296 int optlen, bits = 0; 297 u_char *cp = (u_char *)(tcp + 1); 298 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 299 300 for (; x > 0; x -= optlen, cp += optlen) { 301 int opt = cp[0]; 302 if (opt == TCPOPT_EOL) 303 break; 304 if (opt == TCPOPT_NOP) 305 optlen = 1; 306 else { 307 optlen = cp[1]; 308 if (optlen <= 0) 309 break; 310 } 311 312 switch (opt) { 313 314 default: 315 break; 316 317 case TCPOPT_MAXSEG: 318 bits |= IP_FW_TCPOPT_MSS; 319 break; 320 321 case TCPOPT_WINDOW: 322 bits |= IP_FW_TCPOPT_WINDOW; 323 break; 324 325 case TCPOPT_SACK_PERMITTED: 326 case TCPOPT_SACK: 327 bits |= IP_FW_TCPOPT_SACK; 328 break; 329 330 case TCPOPT_TIMESTAMP: 331 bits |= IP_FW_TCPOPT_TS; 332 break; 333 334 } 335 } 336 return (flags_match(cmd, bits)); 337} 338 339static int 340iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 341{ 342 if (ifp == NULL) /* no iface with this packet, match fails */ 343 return 0; 344 /* Check by name or by IP address */ 345 if (cmd->name[0] != '\0') { /* match by name */ 346 /* Check name */ 347 if (cmd->p.glob) { 348 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 349 return(1); 350 } else { 351 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 352 return(1); 353 } 354 } else { 355#ifdef __FreeBSD__ /* and OSX too ? */ 356 struct ifaddr *ia; 357 358 if_addr_rlock(ifp); 359 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 360 if (ia->ifa_addr->sa_family != AF_INET) 361 continue; 362 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 363 (ia->ifa_addr))->sin_addr.s_addr) { 364 if_addr_runlock(ifp); 365 return(1); /* match */ 366 } 367 } 368 if_addr_runlock(ifp); 369#endif /* __FreeBSD__ */ 370 } 371 return(0); /* no match, fail ... */ 372} 373 374/* 375 * The verify_path function checks if a route to the src exists and 376 * if it is reachable via ifp (when provided). 377 * 378 * The 'verrevpath' option checks that the interface that an IP packet 379 * arrives on is the same interface that traffic destined for the 380 * packet's source address would be routed out of. 381 * The 'versrcreach' option just checks that the source address is 382 * reachable via any route (except default) in the routing table. 383 * These two are a measure to block forged packets. This is also 384 * commonly known as "anti-spoofing" or Unicast Reverse Path 385 * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 386 * is purposely reminiscent of the Cisco IOS command, 387 * 388 * ip verify unicast reverse-path 389 * ip verify unicast source reachable-via any 390 * 391 * which implements the same functionality. But note that the syntax 392 * is misleading, and the check may be performed on all IP packets 393 * whether unicast, multicast, or broadcast. 394 */ 395static int 396verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) 397{ 398#ifndef __FreeBSD__ 399 return 0; 400#else 401 struct route ro; 402 struct sockaddr_in *dst; 403 404 bzero(&ro, sizeof(ro)); 405 406 dst = (struct sockaddr_in *)&(ro.ro_dst); 407 dst->sin_family = AF_INET; 408 dst->sin_len = sizeof(*dst); 409 dst->sin_addr = src; 410 in_rtalloc_ign(&ro, 0, fib); 411 412 if (ro.ro_rt == NULL) 413 return 0; 414 415 /* 416 * If ifp is provided, check for equality with rtentry. 417 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 418 * in order to pass packets injected back by if_simloop(): 419 * if useloopback == 1 routing entry (via lo0) for our own address 420 * may exist, so we need to handle routing assymetry. 421 */ 422 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 423 RTFREE(ro.ro_rt); 424 return 0; 425 } 426 427 /* if no ifp provided, check if rtentry is not default route */ 428 if (ifp == NULL && 429 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 430 RTFREE(ro.ro_rt); 431 return 0; 432 } 433 434 /* or if this is a blackhole/reject route */ 435 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 436 RTFREE(ro.ro_rt); 437 return 0; 438 } 439 440 /* found valid route */ 441 RTFREE(ro.ro_rt); 442 return 1; 443#endif /* __FreeBSD__ */ 444} 445 446#ifdef INET6 447/* 448 * ipv6 specific rules here... 449 */ 450static __inline int 451icmp6type_match (int type, ipfw_insn_u32 *cmd) 452{ 453 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 454} 455 456static int 457flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 458{ 459 int i; 460 for (i=0; i <= cmd->o.arg1; ++i ) 461 if (curr_flow == cmd->d[i] ) 462 return 1; 463 return 0; 464} 465 466/* support for IP6_*_ME opcodes */ 467static int 468search_ip6_addr_net (struct in6_addr * ip6_addr) 469{ 470 struct ifnet *mdc; 471 struct ifaddr *mdc2; 472 struct in6_ifaddr *fdm; 473 struct in6_addr copia; 474 475 TAILQ_FOREACH(mdc, &V_ifnet, if_link) { 476 if_addr_rlock(mdc); 477 TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) { 478 if (mdc2->ifa_addr->sa_family == AF_INET6) { 479 fdm = (struct in6_ifaddr *)mdc2; 480 copia = fdm->ia_addr.sin6_addr; 481 /* need for leaving scope_id in the sock_addr */ 482 in6_clearscope(&copia); 483 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) { 484 if_addr_runlock(mdc); 485 return 1; 486 } 487 } 488 } 489 if_addr_runlock(mdc); 490 } 491 return 0; 492} 493 494static int 495verify_path6(struct in6_addr *src, struct ifnet *ifp) 496{ 497 struct route_in6 ro; 498 struct sockaddr_in6 *dst; 499 500 bzero(&ro, sizeof(ro)); 501 502 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 503 dst->sin6_family = AF_INET6; 504 dst->sin6_len = sizeof(*dst); 505 dst->sin6_addr = *src; 506 /* XXX MRT 0 for ipv6 at this time */ 507 rtalloc_ign((struct route *)&ro, 0); 508 509 if (ro.ro_rt == NULL) 510 return 0; 511 512 /* 513 * if ifp is provided, check for equality with rtentry 514 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 515 * to support the case of sending packets to an address of our own. 516 * (where the former interface is the first argument of if_simloop() 517 * (=ifp), the latter is lo0) 518 */ 519 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 520 RTFREE(ro.ro_rt); 521 return 0; 522 } 523 524 /* if no ifp provided, check if rtentry is not default route */ 525 if (ifp == NULL && 526 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 527 RTFREE(ro.ro_rt); 528 return 0; 529 } 530 531 /* or if this is a blackhole/reject route */ 532 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 533 RTFREE(ro.ro_rt); 534 return 0; 535 } 536 537 /* found valid route */ 538 RTFREE(ro.ro_rt); 539 return 1; 540 541} 542 543static int 544is_icmp6_query(int icmp6_type) 545{ 546 if ((icmp6_type <= ICMP6_MAXTYPE) && 547 (icmp6_type == ICMP6_ECHO_REQUEST || 548 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 549 icmp6_type == ICMP6_WRUREQUEST || 550 icmp6_type == ICMP6_FQDN_QUERY || 551 icmp6_type == ICMP6_NI_QUERY)) 552 return (1); 553 554 return (0); 555} 556 557static void 558send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 559{ 560 struct mbuf *m; 561 562 m = args->m; 563 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 564 struct tcphdr *tcp; 565 tcp = (struct tcphdr *)((char *)ip6 + hlen); 566 567 if ((tcp->th_flags & TH_RST) == 0) { 568 struct mbuf *m0; 569 m0 = ipfw_send_pkt(args->m, &(args->f_id), 570 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 571 tcp->th_flags | TH_RST); 572 if (m0 != NULL) 573 ip6_output(m0, NULL, NULL, 0, NULL, NULL, 574 NULL); 575 } 576 FREE_PKT(m); 577 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 578#if 0 579 /* 580 * Unlike above, the mbufs need to line up with the ip6 hdr, 581 * as the contents are read. We need to m_adj() the 582 * needed amount. 583 * The mbuf will however be thrown away so we can adjust it. 584 * Remember we did an m_pullup on it already so we 585 * can make some assumptions about contiguousness. 586 */ 587 if (args->L3offset) 588 m_adj(m, args->L3offset); 589#endif 590 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 591 } else 592 FREE_PKT(m); 593 594 args->m = NULL; 595} 596 597#endif /* INET6 */ 598 599 600/* 601 * sends a reject message, consuming the mbuf passed as an argument. 602 */ 603static void 604send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip) 605{ 606 607#if 0 608 /* XXX When ip is not guaranteed to be at mtod() we will 609 * need to account for this */ 610 * The mbuf will however be thrown away so we can adjust it. 611 * Remember we did an m_pullup on it already so we 612 * can make some assumptions about contiguousness. 613 */ 614 if (args->L3offset) 615 m_adj(m, args->L3offset); 616#endif 617 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 618 /* We need the IP header in host order for icmp_error(). */ 619 SET_HOST_IPLEN(ip); 620 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 621 } else if (args->f_id.proto == IPPROTO_TCP) { 622 struct tcphdr *const tcp = 623 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 624 if ( (tcp->th_flags & TH_RST) == 0) { 625 struct mbuf *m; 626 m = ipfw_send_pkt(args->m, &(args->f_id), 627 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 628 tcp->th_flags | TH_RST); 629 if (m != NULL) 630 ip_output(m, NULL, NULL, 0, NULL, NULL); 631 } 632 FREE_PKT(args->m); 633 } else 634 FREE_PKT(args->m); 635 args->m = NULL; 636} 637 638/* 639 * Support for uid/gid/jail lookup. These tests are expensive 640 * (because we may need to look into the list of active sockets) 641 * so we cache the results. ugid_lookupp is 0 if we have not 642 * yet done a lookup, 1 if we succeeded, and -1 if we tried 643 * and failed. The function always returns the match value. 644 * We could actually spare the variable and use *uc, setting 645 * it to '(void *)check_uidgid if we have no info, NULL if 646 * we tried and failed, or any other value if successful. 647 */ 648static int 649check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif, 650 struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip, 651 u_int16_t src_port, int *ugid_lookupp, 652 struct ucred **uc, struct inpcb *inp) 653{ 654#ifndef __FreeBSD__ 655 return cred_check(insn, proto, oif, 656 dst_ip, dst_port, src_ip, src_port, 657 (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb); 658#else /* FreeBSD */ 659 struct inpcbinfo *pi; 660 int lookupflags; 661 struct inpcb *pcb; 662 int match; 663 664 /* 665 * Check to see if the UDP or TCP stack supplied us with 666 * the PCB. If so, rather then holding a lock and looking 667 * up the PCB, we can use the one that was supplied. 668 */ 669 if (inp && *ugid_lookupp == 0) { 670 INP_LOCK_ASSERT(inp); 671 if (inp->inp_socket != NULL) { 672 *uc = crhold(inp->inp_cred); 673 *ugid_lookupp = 1; 674 } else 675 *ugid_lookupp = -1; 676 } 677 /* 678 * If we have already been here and the packet has no 679 * PCB entry associated with it, then we can safely 680 * assume that this is a no match. 681 */ 682 if (*ugid_lookupp == -1) 683 return (0); 684 if (proto == IPPROTO_TCP) { 685 lookupflags = 0; 686 pi = &V_tcbinfo; 687 } else if (proto == IPPROTO_UDP) { 688 lookupflags = INPLOOKUP_WILDCARD; 689 pi = &V_udbinfo; 690 } else 691 return 0; 692 lookupflags |= INPLOOKUP_RLOCKPCB; 693 match = 0; 694 if (*ugid_lookupp == 0) { 695 pcb = (oif) ? 696 in_pcblookup(pi, 697 dst_ip, htons(dst_port), 698 src_ip, htons(src_port), 699 lookupflags, oif) : 700 in_pcblookup(pi, 701 src_ip, htons(src_port), 702 dst_ip, htons(dst_port), 703 lookupflags, NULL); 704 if (pcb != NULL) { 705 INP_RLOCK_ASSERT(pcb); 706 *uc = crhold(pcb->inp_cred); 707 *ugid_lookupp = 1; 708 INP_RUNLOCK(pcb); 709 } 710 if (*ugid_lookupp == 0) { 711 /* 712 * We tried and failed, set the variable to -1 713 * so we will not try again on this packet. 714 */ 715 *ugid_lookupp = -1; 716 return (0); 717 } 718 } 719 if (insn->o.opcode == O_UID) 720 match = ((*uc)->cr_uid == (uid_t)insn->d[0]); 721 else if (insn->o.opcode == O_GID) 722 match = groupmember((gid_t)insn->d[0], *uc); 723 else if (insn->o.opcode == O_JAIL) 724 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]); 725 return match; 726#endif /* __FreeBSD__ */ 727} 728 729/* 730 * Helper function to set args with info on the rule after the matching 731 * one. slot is precise, whereas we guess rule_id as they are 732 * assigned sequentially. 733 */ 734static inline void 735set_match(struct ip_fw_args *args, int slot, 736 struct ip_fw_chain *chain) 737{ 738 args->rule.chain_id = chain->id; 739 args->rule.slot = slot + 1; /* we use 0 as a marker */ 740 args->rule.rule_id = 1 + chain->map[slot]->id; 741 args->rule.rulenum = chain->map[slot]->rulenum; 742} 743 744/* 745 * The main check routine for the firewall. 746 * 747 * All arguments are in args so we can modify them and return them 748 * back to the caller. 749 * 750 * Parameters: 751 * 752 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 753 * Starts with the IP header. 754 * args->eh (in) Mac header if present, NULL for layer3 packet. 755 * args->L3offset Number of bytes bypassed if we came from L2. 756 * e.g. often sizeof(eh) ** NOTYET ** 757 * args->oif Outgoing interface, NULL if packet is incoming. 758 * The incoming interface is in the mbuf. (in) 759 * args->divert_rule (in/out) 760 * Skip up to the first rule past this rule number; 761 * upon return, non-zero port number for divert or tee. 762 * 763 * args->rule Pointer to the last matching rule (in/out) 764 * args->next_hop Socket we are forwarding to (out). 765 * args->f_id Addresses grabbed from the packet (out) 766 * args->rule.info a cookie depending on rule action 767 * 768 * Return value: 769 * 770 * IP_FW_PASS the packet must be accepted 771 * IP_FW_DENY the packet must be dropped 772 * IP_FW_DIVERT divert packet, port in m_tag 773 * IP_FW_TEE tee packet, port in m_tag 774 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 775 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 776 * args->rule contains the matching rule, 777 * args->rule.info has additional information. 778 * 779 */ 780int 781ipfw_chk(struct ip_fw_args *args) 782{ 783 784 /* 785 * Local variables holding state while processing a packet: 786 * 787 * IMPORTANT NOTE: to speed up the processing of rules, there 788 * are some assumption on the values of the variables, which 789 * are documented here. Should you change them, please check 790 * the implementation of the various instructions to make sure 791 * that they still work. 792 * 793 * args->eh The MAC header. It is non-null for a layer2 794 * packet, it is NULL for a layer-3 packet. 795 * **notyet** 796 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. 797 * 798 * m | args->m Pointer to the mbuf, as received from the caller. 799 * It may change if ipfw_chk() does an m_pullup, or if it 800 * consumes the packet because it calls send_reject(). 801 * XXX This has to change, so that ipfw_chk() never modifies 802 * or consumes the buffer. 803 * ip is the beginning of the ip(4 or 6) header. 804 * Calculated by adding the L3offset to the start of data. 805 * (Until we start using L3offset, the packet is 806 * supposed to start with the ip header). 807 */ 808 struct mbuf *m = args->m; 809 struct ip *ip = mtod(m, struct ip *); 810 811 /* 812 * For rules which contain uid/gid or jail constraints, cache 813 * a copy of the users credentials after the pcb lookup has been 814 * executed. This will speed up the processing of rules with 815 * these types of constraints, as well as decrease contention 816 * on pcb related locks. 817 */ 818#ifndef __FreeBSD__ 819 struct bsd_ucred ucred_cache; 820#else 821 struct ucred *ucred_cache = NULL; 822#endif 823 int ucred_lookup = 0; 824 825 /* 826 * oif | args->oif If NULL, ipfw_chk has been called on the 827 * inbound path (ether_input, ip_input). 828 * If non-NULL, ipfw_chk has been called on the outbound path 829 * (ether_output, ip_output). 830 */ 831 struct ifnet *oif = args->oif; 832 833 int f_pos = 0; /* index of current rule in the array */ 834 int retval = 0; 835 836 /* 837 * hlen The length of the IP header. 838 */ 839 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 840 841 /* 842 * offset The offset of a fragment. offset != 0 means that 843 * we have a fragment at this offset of an IPv4 packet. 844 * offset == 0 means that (if this is an IPv4 packet) 845 * this is the first or only fragment. 846 * For IPv6 offset == 0 means there is no Fragment Header. 847 * If offset != 0 for IPv6 always use correct mask to 848 * get the correct offset because we add IP6F_MORE_FRAG 849 * to be able to dectect the first fragment which would 850 * otherwise have offset = 0. 851 */ 852 u_short offset = 0; 853 854 /* 855 * Local copies of addresses. They are only valid if we have 856 * an IP packet. 857 * 858 * proto The protocol. Set to 0 for non-ip packets, 859 * or to the protocol read from the packet otherwise. 860 * proto != 0 means that we have an IPv4 packet. 861 * 862 * src_port, dst_port port numbers, in HOST format. Only 863 * valid for TCP and UDP packets. 864 * 865 * src_ip, dst_ip ip addresses, in NETWORK format. 866 * Only valid for IPv4 packets. 867 */ 868 uint8_t proto; 869 uint16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 870 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 871 uint16_t iplen=0; 872 int pktlen; 873 uint16_t etype = 0; /* Host order stored ether type */ 874 875 /* 876 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 877 * MATCH_NONE when checked and not matched (q = NULL), 878 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 879 */ 880 int dyn_dir = MATCH_UNKNOWN; 881 ipfw_dyn_rule *q = NULL; 882 struct ip_fw_chain *chain = &V_layer3_chain; 883 884 /* 885 * We store in ulp a pointer to the upper layer protocol header. 886 * In the ipv4 case this is easy to determine from the header, 887 * but for ipv6 we might have some additional headers in the middle. 888 * ulp is NULL if not found. 889 */ 890 void *ulp = NULL; /* upper layer protocol pointer. */ 891 892 /* XXX ipv6 variables */ 893 int is_ipv6 = 0; 894 uint8_t icmp6_type = 0; 895 uint16_t ext_hd = 0; /* bits vector for extension header filtering */ 896 /* end of ipv6 variables */ 897 898 int is_ipv4 = 0; 899 900 int done = 0; /* flag to exit the outer loop */ 901 902 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready)) 903 return (IP_FW_PASS); /* accept */ 904 905 dst_ip.s_addr = 0; /* make sure it is initialized */ 906 src_ip.s_addr = 0; /* make sure it is initialized */ 907 pktlen = m->m_pkthdr.len; 908 args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */ 909 proto = args->f_id.proto = 0; /* mark f_id invalid */ 910 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 911 912/* 913 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 914 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 915 * pointer might become stale after other pullups (but we never use it 916 * this way). 917 */ 918#define PULLUP_TO(_len, p, T) PULLUP_LEN(_len, p, sizeof(T)) 919#define PULLUP_LEN(_len, p, T) \ 920do { \ 921 int x = (_len) + T; \ 922 if ((m)->m_len < x) { \ 923 args->m = m = m_pullup(m, x); \ 924 if (m == NULL) \ 925 goto pullup_failed; \ 926 } \ 927 p = (mtod(m, char *) + (_len)); \ 928} while (0) 929 930 /* 931 * if we have an ether header, 932 */ 933 if (args->eh) 934 etype = ntohs(args->eh->ether_type); 935 936 /* Identify IP packets and fill up variables. */ 937 if (pktlen >= sizeof(struct ip6_hdr) && 938 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { 939 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 940 is_ipv6 = 1; 941 args->f_id.addr_type = 6; 942 hlen = sizeof(struct ip6_hdr); 943 proto = ip6->ip6_nxt; 944 945 /* Search extension headers to find upper layer protocols */ 946 while (ulp == NULL) { 947 switch (proto) { 948 case IPPROTO_ICMPV6: 949 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 950 icmp6_type = ICMP6(ulp)->icmp6_type; 951 break; 952 953 case IPPROTO_TCP: 954 PULLUP_TO(hlen, ulp, struct tcphdr); 955 dst_port = TCP(ulp)->th_dport; 956 src_port = TCP(ulp)->th_sport; 957 /* save flags for dynamic rules */ 958 args->f_id._flags = TCP(ulp)->th_flags; 959 break; 960 961 case IPPROTO_SCTP: 962 PULLUP_TO(hlen, ulp, struct sctphdr); 963 src_port = SCTP(ulp)->src_port; 964 dst_port = SCTP(ulp)->dest_port; 965 break; 966 967 case IPPROTO_UDP: 968 PULLUP_TO(hlen, ulp, struct udphdr); 969 dst_port = UDP(ulp)->uh_dport; 970 src_port = UDP(ulp)->uh_sport; 971 break; 972 973 case IPPROTO_HOPOPTS: /* RFC 2460 */ 974 PULLUP_TO(hlen, ulp, struct ip6_hbh); 975 ext_hd |= EXT_HOPOPTS; 976 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 977 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 978 ulp = NULL; 979 break; 980 981 case IPPROTO_ROUTING: /* RFC 2460 */ 982 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 983 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 984 case 0: 985 ext_hd |= EXT_RTHDR0; 986 break; 987 case 2: 988 ext_hd |= EXT_RTHDR2; 989 break; 990 default: 991 printf("IPFW2: IPV6 - Unknown Routing " 992 "Header type(%d)\n", 993 ((struct ip6_rthdr *)ulp)->ip6r_type); 994 if (V_fw_deny_unknown_exthdrs) 995 return (IP_FW_DENY); 996 break; 997 } 998 ext_hd |= EXT_ROUTING; 999 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 1000 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 1001 ulp = NULL; 1002 break; 1003 1004 case IPPROTO_FRAGMENT: /* RFC 2460 */ 1005 PULLUP_TO(hlen, ulp, struct ip6_frag); 1006 ext_hd |= EXT_FRAGMENT; 1007 hlen += sizeof (struct ip6_frag); 1008 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 1009 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 1010 IP6F_OFF_MASK; 1011 /* Add IP6F_MORE_FRAG for offset of first 1012 * fragment to be != 0. */ 1013 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 1014 IP6F_MORE_FRAG; 1015 if (offset == 0) { 1016 printf("IPFW2: IPV6 - Invalid Fragment " 1017 "Header\n"); 1018 if (V_fw_deny_unknown_exthdrs) 1019 return (IP_FW_DENY); 1020 break; 1021 } 1022 args->f_id.extra = 1023 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 1024 ulp = NULL; 1025 break; 1026 1027 case IPPROTO_DSTOPTS: /* RFC 2460 */ 1028 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1029 ext_hd |= EXT_DSTOPTS; 1030 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1031 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1032 ulp = NULL; 1033 break; 1034 1035 case IPPROTO_AH: /* RFC 2402 */ 1036 PULLUP_TO(hlen, ulp, struct ip6_ext); 1037 ext_hd |= EXT_AH; 1038 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 1039 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 1040 ulp = NULL; 1041 break; 1042 1043 case IPPROTO_ESP: /* RFC 2406 */ 1044 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 1045 /* Anything past Seq# is variable length and 1046 * data past this ext. header is encrypted. */ 1047 ext_hd |= EXT_ESP; 1048 break; 1049 1050 case IPPROTO_NONE: /* RFC 2460 */ 1051 /* 1052 * Packet ends here, and IPv6 header has 1053 * already been pulled up. If ip6e_len!=0 1054 * then octets must be ignored. 1055 */ 1056 ulp = ip; /* non-NULL to get out of loop. */ 1057 break; 1058 1059 case IPPROTO_OSPFIGP: 1060 /* XXX OSPF header check? */ 1061 PULLUP_TO(hlen, ulp, struct ip6_ext); 1062 break; 1063 1064 case IPPROTO_PIM: 1065 /* XXX PIM header check? */ 1066 PULLUP_TO(hlen, ulp, struct pim); 1067 break; 1068 1069 case IPPROTO_CARP: 1070 PULLUP_TO(hlen, ulp, struct carp_header); 1071 if (((struct carp_header *)ulp)->carp_version != 1072 CARP_VERSION) 1073 return (IP_FW_DENY); 1074 if (((struct carp_header *)ulp)->carp_type != 1075 CARP_ADVERTISEMENT) 1076 return (IP_FW_DENY); 1077 break; 1078 1079 case IPPROTO_IPV6: /* RFC 2893 */ 1080 PULLUP_TO(hlen, ulp, struct ip6_hdr); 1081 break; 1082 1083 case IPPROTO_IPV4: /* RFC 2893 */ 1084 PULLUP_TO(hlen, ulp, struct ip); 1085 break; 1086 1087 default: 1088 printf("IPFW2: IPV6 - Unknown Extension " 1089 "Header(%d), ext_hd=%x\n", proto, ext_hd); 1090 if (V_fw_deny_unknown_exthdrs) 1091 return (IP_FW_DENY); 1092 PULLUP_TO(hlen, ulp, struct ip6_ext); 1093 break; 1094 } /*switch */ 1095 } 1096 ip = mtod(m, struct ip *); 1097 ip6 = (struct ip6_hdr *)ip; 1098 args->f_id.src_ip6 = ip6->ip6_src; 1099 args->f_id.dst_ip6 = ip6->ip6_dst; 1100 args->f_id.src_ip = 0; 1101 args->f_id.dst_ip = 0; 1102 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 1103 } else if (pktlen >= sizeof(struct ip) && 1104 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { 1105 is_ipv4 = 1; 1106 hlen = ip->ip_hl << 2; 1107 args->f_id.addr_type = 4; 1108 1109 /* 1110 * Collect parameters into local variables for faster matching. 1111 */ 1112 proto = ip->ip_p; 1113 src_ip = ip->ip_src; 1114 dst_ip = ip->ip_dst; 1115 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1116 iplen = ntohs(ip->ip_len); 1117 pktlen = iplen < pktlen ? iplen : pktlen; 1118 1119 if (offset == 0) { 1120 switch (proto) { 1121 case IPPROTO_TCP: 1122 PULLUP_TO(hlen, ulp, struct tcphdr); 1123 dst_port = TCP(ulp)->th_dport; 1124 src_port = TCP(ulp)->th_sport; 1125 /* save flags for dynamic rules */ 1126 args->f_id._flags = TCP(ulp)->th_flags; 1127 break; 1128 1129 case IPPROTO_SCTP: 1130 PULLUP_TO(hlen, ulp, struct sctphdr); 1131 src_port = SCTP(ulp)->src_port; 1132 dst_port = SCTP(ulp)->dest_port; 1133 break; 1134 1135 case IPPROTO_UDP: 1136 PULLUP_TO(hlen, ulp, struct udphdr); 1137 dst_port = UDP(ulp)->uh_dport; 1138 src_port = UDP(ulp)->uh_sport; 1139 break; 1140 1141 case IPPROTO_ICMP: 1142 PULLUP_TO(hlen, ulp, struct icmphdr); 1143 //args->f_id.flags = ICMP(ulp)->icmp_type; 1144 break; 1145 1146 default: 1147 break; 1148 } 1149 } 1150 1151 ip = mtod(m, struct ip *); 1152 args->f_id.src_ip = ntohl(src_ip.s_addr); 1153 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1154 } 1155#undef PULLUP_TO 1156 if (proto) { /* we may have port numbers, store them */ 1157 args->f_id.proto = proto; 1158 args->f_id.src_port = src_port = ntohs(src_port); 1159 args->f_id.dst_port = dst_port = ntohs(dst_port); 1160 } 1161 1162 IPFW_RLOCK(chain); 1163 if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */ 1164 IPFW_RUNLOCK(chain); 1165 return (IP_FW_PASS); /* accept */ 1166 } 1167 if (args->rule.slot) { 1168 /* 1169 * Packet has already been tagged as a result of a previous 1170 * match on rule args->rule aka args->rule_id (PIPE, QUEUE, 1171 * REASS, NETGRAPH, DIVERT/TEE...) 1172 * Validate the slot and continue from the next one 1173 * if still present, otherwise do a lookup. 1174 */ 1175 f_pos = (args->rule.chain_id == chain->id) ? 1176 args->rule.slot : 1177 ipfw_find_rule(chain, args->rule.rulenum, 1178 args->rule.rule_id); 1179 } else { 1180 f_pos = 0; 1181 } 1182 1183 /* 1184 * Now scan the rules, and parse microinstructions for each rule. 1185 * We have two nested loops and an inner switch. Sometimes we 1186 * need to break out of one or both loops, or re-enter one of 1187 * the loops with updated variables. Loop variables are: 1188 * 1189 * f_pos (outer loop) points to the current rule. 1190 * On output it points to the matching rule. 1191 * done (outer loop) is used as a flag to break the loop. 1192 * l (inner loop) residual length of current rule. 1193 * cmd points to the current microinstruction. 1194 * 1195 * We break the inner loop by setting l=0 and possibly 1196 * cmdlen=0 if we don't want to advance cmd. 1197 * We break the outer loop by setting done=1 1198 * We can restart the inner loop by setting l>0 and f_pos, f, cmd 1199 * as needed. 1200 */ 1201 for (; f_pos < chain->n_rules; f_pos++) { 1202 ipfw_insn *cmd; 1203 uint32_t tablearg = 0; 1204 int l, cmdlen, skip_or; /* skip rest of OR block */ 1205 struct ip_fw *f; 1206 1207 f = chain->map[f_pos]; 1208 if (V_set_disable & (1 << f->set) ) 1209 continue; 1210 1211 skip_or = 0; 1212 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1213 l -= cmdlen, cmd += cmdlen) { 1214 int match; 1215 1216 /* 1217 * check_body is a jump target used when we find a 1218 * CHECK_STATE, and need to jump to the body of 1219 * the target rule. 1220 */ 1221 1222/* check_body: */ 1223 cmdlen = F_LEN(cmd); 1224 /* 1225 * An OR block (insn_1 || .. || insn_n) has the 1226 * F_OR bit set in all but the last instruction. 1227 * The first match will set "skip_or", and cause 1228 * the following instructions to be skipped until 1229 * past the one with the F_OR bit clear. 1230 */ 1231 if (skip_or) { /* skip this instruction */ 1232 if ((cmd->len & F_OR) == 0) 1233 skip_or = 0; /* next one is good */ 1234 continue; 1235 } 1236 match = 0; /* set to 1 if we succeed */ 1237 1238 switch (cmd->opcode) { 1239 /* 1240 * The first set of opcodes compares the packet's 1241 * fields with some pattern, setting 'match' if a 1242 * match is found. At the end of the loop there is 1243 * logic to deal with F_NOT and F_OR flags associated 1244 * with the opcode. 1245 */ 1246 case O_NOP: 1247 match = 1; 1248 break; 1249 1250 case O_FORWARD_MAC: 1251 printf("ipfw: opcode %d unimplemented\n", 1252 cmd->opcode); 1253 break; 1254 1255 case O_GID: 1256 case O_UID: 1257 case O_JAIL: 1258 /* 1259 * We only check offset == 0 && proto != 0, 1260 * as this ensures that we have a 1261 * packet with the ports info. 1262 */ 1263 if (offset!=0) 1264 break; 1265 if (is_ipv6) /* XXX to be fixed later */ 1266 break; 1267 if (proto == IPPROTO_TCP || 1268 proto == IPPROTO_UDP) 1269 match = check_uidgid( 1270 (ipfw_insn_u32 *)cmd, 1271 proto, oif, 1272 dst_ip, dst_port, 1273 src_ip, src_port, &ucred_lookup, 1274#ifdef __FreeBSD__ 1275 &ucred_cache, args->inp); 1276#else 1277 (void *)&ucred_cache, 1278 (struct inpcb *)args->m); 1279#endif 1280 break; 1281 1282 case O_RECV: 1283 match = iface_match(m->m_pkthdr.rcvif, 1284 (ipfw_insn_if *)cmd); 1285 break; 1286 1287 case O_XMIT: 1288 match = iface_match(oif, (ipfw_insn_if *)cmd); 1289 break; 1290 1291 case O_VIA: 1292 match = iface_match(oif ? oif : 1293 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 1294 break; 1295 1296 case O_MACADDR2: 1297 if (args->eh != NULL) { /* have MAC header */ 1298 u_int32_t *want = (u_int32_t *) 1299 ((ipfw_insn_mac *)cmd)->addr; 1300 u_int32_t *mask = (u_int32_t *) 1301 ((ipfw_insn_mac *)cmd)->mask; 1302 u_int32_t *hdr = (u_int32_t *)args->eh; 1303 1304 match = 1305 ( want[0] == (hdr[0] & mask[0]) && 1306 want[1] == (hdr[1] & mask[1]) && 1307 want[2] == (hdr[2] & mask[2]) ); 1308 } 1309 break; 1310 1311 case O_MAC_TYPE: 1312 if (args->eh != NULL) { 1313 u_int16_t *p = 1314 ((ipfw_insn_u16 *)cmd)->ports; 1315 int i; 1316 1317 for (i = cmdlen - 1; !match && i>0; 1318 i--, p += 2) 1319 match = (etype >= p[0] && 1320 etype <= p[1]); 1321 } 1322 break; 1323 1324 case O_FRAG: 1325 match = (offset != 0); 1326 break; 1327 1328 case O_IN: /* "out" is "not in" */ 1329 match = (oif == NULL); 1330 break; 1331 1332 case O_LAYER2: 1333 match = (args->eh != NULL); 1334 break; 1335 1336 case O_DIVERTED: 1337 { 1338 /* For diverted packets, args->rule.info 1339 * contains the divert port (in host format) 1340 * reason and direction. 1341 */ 1342 uint32_t i = args->rule.info; 1343 match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT && 1344 cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2); 1345 } 1346 break; 1347 1348 case O_PROTO: 1349 /* 1350 * We do not allow an arg of 0 so the 1351 * check of "proto" only suffices. 1352 */ 1353 match = (proto == cmd->arg1); 1354 break; 1355 1356 case O_IP_SRC: 1357 match = is_ipv4 && 1358 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1359 src_ip.s_addr); 1360 break; 1361 1362 case O_IP_SRC_LOOKUP: 1363 case O_IP_DST_LOOKUP: 1364 if (is_ipv4) { 1365 uint32_t key = 1366 (cmd->opcode == O_IP_DST_LOOKUP) ? 1367 dst_ip.s_addr : src_ip.s_addr; 1368 uint32_t v = 0; 1369 1370 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) { 1371 /* generic lookup. The key must be 1372 * in 32bit big-endian format. 1373 */ 1374 v = ((ipfw_insn_u32 *)cmd)->d[1]; 1375 if (v == 0) 1376 key = dst_ip.s_addr; 1377 else if (v == 1) 1378 key = src_ip.s_addr; 1379 else if (v == 6) /* dscp */ 1380 key = (ip->ip_tos >> 2) & 0x3f; 1381 else if (offset != 0) 1382 break; 1383 else if (proto != IPPROTO_TCP && 1384 proto != IPPROTO_UDP) 1385 break; 1386 else if (v == 2) 1387 key = htonl(dst_port); 1388 else if (v == 3) 1389 key = htonl(src_port); 1390 else if (v == 4 || v == 5) { 1391 check_uidgid( 1392 (ipfw_insn_u32 *)cmd, 1393 proto, oif, 1394 dst_ip, dst_port, 1395 src_ip, src_port, &ucred_lookup, 1396#ifdef __FreeBSD__ 1397 &ucred_cache, args->inp); 1398 if (v == 4 /* O_UID */) 1399 key = ucred_cache->cr_uid; 1400 else if (v == 5 /* O_JAIL */) 1401 key = ucred_cache->cr_prison->pr_id; 1402#else /* !__FreeBSD__ */ 1403 (void *)&ucred_cache, 1404 (struct inpcb *)args->m); 1405 if (v ==4 /* O_UID */) 1406 key = ucred_cache.uid; 1407 else if (v == 5 /* O_JAIL */) 1408 key = ucred_cache.xid; 1409#endif /* !__FreeBSD__ */ 1410 key = htonl(key); 1411 } else 1412 break; 1413 } 1414 match = ipfw_lookup_table(chain, 1415 cmd->arg1, key, &v); 1416 if (!match) 1417 break; 1418 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 1419 match = 1420 ((ipfw_insn_u32 *)cmd)->d[0] == v; 1421 else 1422 tablearg = v; 1423 } 1424 break; 1425 1426 case O_IP_SRC_MASK: 1427 case O_IP_DST_MASK: 1428 if (is_ipv4) { 1429 uint32_t a = 1430 (cmd->opcode == O_IP_DST_MASK) ? 1431 dst_ip.s_addr : src_ip.s_addr; 1432 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 1433 int i = cmdlen-1; 1434 1435 for (; !match && i>0; i-= 2, p+= 2) 1436 match = (p[0] == (a & p[1])); 1437 } 1438 break; 1439 1440 case O_IP_SRC_ME: 1441 if (is_ipv4) { 1442 struct ifnet *tif; 1443 1444 INADDR_TO_IFP(src_ip, tif); 1445 match = (tif != NULL); 1446 break; 1447 } 1448#ifdef INET6 1449 /* FALLTHROUGH */ 1450 case O_IP6_SRC_ME: 1451 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 1452#endif 1453 break; 1454 1455 case O_IP_DST_SET: 1456 case O_IP_SRC_SET: 1457 if (is_ipv4) { 1458 u_int32_t *d = (u_int32_t *)(cmd+1); 1459 u_int32_t addr = 1460 cmd->opcode == O_IP_DST_SET ? 1461 args->f_id.dst_ip : 1462 args->f_id.src_ip; 1463 1464 if (addr < d[0]) 1465 break; 1466 addr -= d[0]; /* subtract base */ 1467 match = (addr < cmd->arg1) && 1468 ( d[ 1 + (addr>>5)] & 1469 (1<<(addr & 0x1f)) ); 1470 } 1471 break; 1472 1473 case O_IP_DST: 1474 match = is_ipv4 && 1475 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1476 dst_ip.s_addr); 1477 break; 1478 1479 case O_IP_DST_ME: 1480 if (is_ipv4) { 1481 struct ifnet *tif; 1482 1483 INADDR_TO_IFP(dst_ip, tif); 1484 match = (tif != NULL); 1485 break; 1486 } 1487#ifdef INET6 1488 /* FALLTHROUGH */ 1489 case O_IP6_DST_ME: 1490 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 1491#endif 1492 break; 1493 1494 1495 case O_IP_SRCPORT: 1496 case O_IP_DSTPORT: 1497 /* 1498 * offset == 0 && proto != 0 is enough 1499 * to guarantee that we have a 1500 * packet with port info. 1501 */ 1502 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 1503 && offset == 0) { 1504 u_int16_t x = 1505 (cmd->opcode == O_IP_SRCPORT) ? 1506 src_port : dst_port ; 1507 u_int16_t *p = 1508 ((ipfw_insn_u16 *)cmd)->ports; 1509 int i; 1510 1511 for (i = cmdlen - 1; !match && i>0; 1512 i--, p += 2) 1513 match = (x>=p[0] && x<=p[1]); 1514 } 1515 break; 1516 1517 case O_ICMPTYPE: 1518 match = (offset == 0 && proto==IPPROTO_ICMP && 1519 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 1520 break; 1521 1522#ifdef INET6 1523 case O_ICMP6TYPE: 1524 match = is_ipv6 && offset == 0 && 1525 proto==IPPROTO_ICMPV6 && 1526 icmp6type_match( 1527 ICMP6(ulp)->icmp6_type, 1528 (ipfw_insn_u32 *)cmd); 1529 break; 1530#endif /* INET6 */ 1531 1532 case O_IPOPT: 1533 match = (is_ipv4 && 1534 ipopts_match(ip, cmd) ); 1535 break; 1536 1537 case O_IPVER: 1538 match = (is_ipv4 && 1539 cmd->arg1 == ip->ip_v); 1540 break; 1541 1542 case O_IPID: 1543 case O_IPLEN: 1544 case O_IPTTL: 1545 if (is_ipv4) { /* only for IP packets */ 1546 uint16_t x; 1547 uint16_t *p; 1548 int i; 1549 1550 if (cmd->opcode == O_IPLEN) 1551 x = iplen; 1552 else if (cmd->opcode == O_IPTTL) 1553 x = ip->ip_ttl; 1554 else /* must be IPID */ 1555 x = ntohs(ip->ip_id); 1556 if (cmdlen == 1) { 1557 match = (cmd->arg1 == x); 1558 break; 1559 } 1560 /* otherwise we have ranges */ 1561 p = ((ipfw_insn_u16 *)cmd)->ports; 1562 i = cmdlen - 1; 1563 for (; !match && i>0; i--, p += 2) 1564 match = (x >= p[0] && x <= p[1]); 1565 } 1566 break; 1567 1568 case O_IPPRECEDENCE: 1569 match = (is_ipv4 && 1570 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 1571 break; 1572 1573 case O_IPTOS: 1574 match = (is_ipv4 && 1575 flags_match(cmd, ip->ip_tos)); 1576 break; 1577 1578 case O_TCPDATALEN: 1579 if (proto == IPPROTO_TCP && offset == 0) { 1580 struct tcphdr *tcp; 1581 uint16_t x; 1582 uint16_t *p; 1583 int i; 1584 1585 tcp = TCP(ulp); 1586 x = iplen - 1587 ((ip->ip_hl + tcp->th_off) << 2); 1588 if (cmdlen == 1) { 1589 match = (cmd->arg1 == x); 1590 break; 1591 } 1592 /* otherwise we have ranges */ 1593 p = ((ipfw_insn_u16 *)cmd)->ports; 1594 i = cmdlen - 1; 1595 for (; !match && i>0; i--, p += 2) 1596 match = (x >= p[0] && x <= p[1]); 1597 } 1598 break; 1599 1600 case O_TCPFLAGS: 1601 match = (proto == IPPROTO_TCP && offset == 0 && 1602 flags_match(cmd, TCP(ulp)->th_flags)); 1603 break; 1604 1605 case O_TCPOPTS: 1606 PULLUP_LEN(hlen, ulp, (TCP(ulp)->th_off << 2)); 1607 match = (proto == IPPROTO_TCP && offset == 0 && 1608 tcpopts_match(TCP(ulp), cmd)); 1609 break; 1610 1611 case O_TCPSEQ: 1612 match = (proto == IPPROTO_TCP && offset == 0 && 1613 ((ipfw_insn_u32 *)cmd)->d[0] == 1614 TCP(ulp)->th_seq); 1615 break; 1616 1617 case O_TCPACK: 1618 match = (proto == IPPROTO_TCP && offset == 0 && 1619 ((ipfw_insn_u32 *)cmd)->d[0] == 1620 TCP(ulp)->th_ack); 1621 break; 1622 1623 case O_TCPWIN: 1624 match = (proto == IPPROTO_TCP && offset == 0 && 1625 cmd->arg1 == TCP(ulp)->th_win); 1626 break; 1627 1628 case O_ESTAB: 1629 /* reject packets which have SYN only */ 1630 /* XXX should i also check for TH_ACK ? */ 1631 match = (proto == IPPROTO_TCP && offset == 0 && 1632 (TCP(ulp)->th_flags & 1633 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 1634 break; 1635 1636 case O_ALTQ: { 1637 struct pf_mtag *at; 1638 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 1639 1640 match = 1; 1641 at = pf_find_mtag(m); 1642 if (at != NULL && at->qid != 0) 1643 break; 1644 at = pf_get_mtag(m); 1645 if (at == NULL) { 1646 /* 1647 * Let the packet fall back to the 1648 * default ALTQ. 1649 */ 1650 break; 1651 } 1652 at->qid = altq->qid; 1653 if (is_ipv4) 1654 at->af = AF_INET; 1655 else 1656 at->af = AF_LINK; 1657 at->hdr = ip; 1658 break; 1659 } 1660 1661 case O_LOG: 1662 ipfw_log(f, hlen, args, m, 1663 oif, offset, tablearg, ip); 1664 match = 1; 1665 break; 1666 1667 case O_PROB: 1668 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 1669 break; 1670 1671 case O_VERREVPATH: 1672 /* Outgoing packets automatically pass/match */ 1673 match = ((oif != NULL) || 1674 (m->m_pkthdr.rcvif == NULL) || 1675 ( 1676#ifdef INET6 1677 is_ipv6 ? 1678 verify_path6(&(args->f_id.src_ip6), 1679 m->m_pkthdr.rcvif) : 1680#endif 1681 verify_path(src_ip, m->m_pkthdr.rcvif, 1682 args->f_id.fib))); 1683 break; 1684 1685 case O_VERSRCREACH: 1686 /* Outgoing packets automatically pass/match */ 1687 match = (hlen > 0 && ((oif != NULL) || 1688#ifdef INET6 1689 is_ipv6 ? 1690 verify_path6(&(args->f_id.src_ip6), 1691 NULL) : 1692#endif 1693 verify_path(src_ip, NULL, args->f_id.fib))); 1694 break; 1695 1696 case O_ANTISPOOF: 1697 /* Outgoing packets automatically pass/match */ 1698 if (oif == NULL && hlen > 0 && 1699 ( (is_ipv4 && in_localaddr(src_ip)) 1700#ifdef INET6 1701 || (is_ipv6 && 1702 in6_localaddr(&(args->f_id.src_ip6))) 1703#endif 1704 )) 1705 match = 1706#ifdef INET6 1707 is_ipv6 ? verify_path6( 1708 &(args->f_id.src_ip6), 1709 m->m_pkthdr.rcvif) : 1710#endif 1711 verify_path(src_ip, 1712 m->m_pkthdr.rcvif, 1713 args->f_id.fib); 1714 else 1715 match = 1; 1716 break; 1717 1718 case O_IPSEC: 1719#ifdef IPSEC 1720 match = (m_tag_find(m, 1721 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 1722#endif 1723 /* otherwise no match */ 1724 break; 1725 1726#ifdef INET6 1727 case O_IP6_SRC: 1728 match = is_ipv6 && 1729 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 1730 &((ipfw_insn_ip6 *)cmd)->addr6); 1731 break; 1732 1733 case O_IP6_DST: 1734 match = is_ipv6 && 1735 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 1736 &((ipfw_insn_ip6 *)cmd)->addr6); 1737 break; 1738 case O_IP6_SRC_MASK: 1739 case O_IP6_DST_MASK: 1740 if (is_ipv6) { 1741 int i = cmdlen - 1; 1742 struct in6_addr p; 1743 struct in6_addr *d = 1744 &((ipfw_insn_ip6 *)cmd)->addr6; 1745 1746 for (; !match && i > 0; d += 2, 1747 i -= F_INSN_SIZE(struct in6_addr) 1748 * 2) { 1749 p = (cmd->opcode == 1750 O_IP6_SRC_MASK) ? 1751 args->f_id.src_ip6: 1752 args->f_id.dst_ip6; 1753 APPLY_MASK(&p, &d[1]); 1754 match = 1755 IN6_ARE_ADDR_EQUAL(&d[0], 1756 &p); 1757 } 1758 } 1759 break; 1760 1761 case O_FLOW6ID: 1762 match = is_ipv6 && 1763 flow6id_match(args->f_id.flow_id6, 1764 (ipfw_insn_u32 *) cmd); 1765 break; 1766 1767 case O_EXT_HDR: 1768 match = is_ipv6 && 1769 (ext_hd & ((ipfw_insn *) cmd)->arg1); 1770 break; 1771 1772 case O_IP6: 1773 match = is_ipv6; 1774 break; 1775#endif 1776 1777 case O_IP4: 1778 match = is_ipv4; 1779 break; 1780 1781 case O_TAG: { 1782 struct m_tag *mtag; 1783 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 1784 tablearg : cmd->arg1; 1785 1786 /* Packet is already tagged with this tag? */ 1787 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 1788 1789 /* We have `untag' action when F_NOT flag is 1790 * present. And we must remove this mtag from 1791 * mbuf and reset `match' to zero (`match' will 1792 * be inversed later). 1793 * Otherwise we should allocate new mtag and 1794 * push it into mbuf. 1795 */ 1796 if (cmd->len & F_NOT) { /* `untag' action */ 1797 if (mtag != NULL) 1798 m_tag_delete(m, mtag); 1799 match = 0; 1800 } else { 1801 if (mtag == NULL) { 1802 mtag = m_tag_alloc( MTAG_IPFW, 1803 tag, 0, M_NOWAIT); 1804 if (mtag != NULL) 1805 m_tag_prepend(m, mtag); 1806 } 1807 match = 1; 1808 } 1809 break; 1810 } 1811 1812 case O_FIB: /* try match the specified fib */ 1813 if (args->f_id.fib == cmd->arg1) 1814 match = 1; 1815 break; 1816 1817 case O_SOCKARG: { 1818 struct inpcb *inp = args->inp; 1819 struct inpcbinfo *pi; 1820 1821 if (is_ipv6) /* XXX can we remove this ? */ 1822 break; 1823 1824 if (proto == IPPROTO_TCP) 1825 pi = &V_tcbinfo; 1826 else if (proto == IPPROTO_UDP) 1827 pi = &V_udbinfo; 1828 else 1829 break; 1830 1831 /* 1832 * XXXRW: so_user_cookie should almost 1833 * certainly be inp_user_cookie? 1834 */ 1835 1836 /* For incomming packet, lookup up the 1837 inpcb using the src/dest ip/port tuple */ 1838 if (inp == NULL) { 1839 inp = in_pcblookup(pi, 1840 src_ip, htons(src_port), 1841 dst_ip, htons(dst_port), 1842 INPLOOKUP_RLOCKPCB, NULL); 1843 if (inp != NULL) { 1844 tablearg = 1845 inp->inp_socket->so_user_cookie; 1846 if (tablearg) 1847 match = 1; 1848 INP_RUNLOCK(inp); 1849 } 1850 } else { 1851 if (inp->inp_socket) { 1852 tablearg = 1853 inp->inp_socket->so_user_cookie; 1854 if (tablearg) 1855 match = 1; 1856 } 1857 } 1858 break; 1859 } 1860 1861 case O_TAGGED: { 1862 struct m_tag *mtag; 1863 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 1864 tablearg : cmd->arg1; 1865 1866 if (cmdlen == 1) { 1867 match = m_tag_locate(m, MTAG_IPFW, 1868 tag, NULL) != NULL; 1869 break; 1870 } 1871 1872 /* we have ranges */ 1873 for (mtag = m_tag_first(m); 1874 mtag != NULL && !match; 1875 mtag = m_tag_next(m, mtag)) { 1876 uint16_t *p; 1877 int i; 1878 1879 if (mtag->m_tag_cookie != MTAG_IPFW) 1880 continue; 1881 1882 p = ((ipfw_insn_u16 *)cmd)->ports; 1883 i = cmdlen - 1; 1884 for(; !match && i > 0; i--, p += 2) 1885 match = 1886 mtag->m_tag_id >= p[0] && 1887 mtag->m_tag_id <= p[1]; 1888 } 1889 break; 1890 } 1891 1892 /* 1893 * The second set of opcodes represents 'actions', 1894 * i.e. the terminal part of a rule once the packet 1895 * matches all previous patterns. 1896 * Typically there is only one action for each rule, 1897 * and the opcode is stored at the end of the rule 1898 * (but there are exceptions -- see below). 1899 * 1900 * In general, here we set retval and terminate the 1901 * outer loop (would be a 'break 3' in some language, 1902 * but we need to set l=0, done=1) 1903 * 1904 * Exceptions: 1905 * O_COUNT and O_SKIPTO actions: 1906 * instead of terminating, we jump to the next rule 1907 * (setting l=0), or to the SKIPTO target (setting 1908 * f/f_len, cmd and l as needed), respectively. 1909 * 1910 * O_TAG, O_LOG and O_ALTQ action parameters: 1911 * perform some action and set match = 1; 1912 * 1913 * O_LIMIT and O_KEEP_STATE: these opcodes are 1914 * not real 'actions', and are stored right 1915 * before the 'action' part of the rule. 1916 * These opcodes try to install an entry in the 1917 * state tables; if successful, we continue with 1918 * the next opcode (match=1; break;), otherwise 1919 * the packet must be dropped (set retval, 1920 * break loops with l=0, done=1) 1921 * 1922 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 1923 * cause a lookup of the state table, and a jump 1924 * to the 'action' part of the parent rule 1925 * if an entry is found, or 1926 * (CHECK_STATE only) a jump to the next rule if 1927 * the entry is not found. 1928 * The result of the lookup is cached so that 1929 * further instances of these opcodes become NOPs. 1930 * The jump to the next rule is done by setting 1931 * l=0, cmdlen=0. 1932 */ 1933 case O_LIMIT: 1934 case O_KEEP_STATE: 1935 if (ipfw_install_state(f, 1936 (ipfw_insn_limit *)cmd, args, tablearg)) { 1937 /* error or limit violation */ 1938 retval = IP_FW_DENY; 1939 l = 0; /* exit inner loop */ 1940 done = 1; /* exit outer loop */ 1941 } 1942 match = 1; 1943 break; 1944 1945 case O_PROBE_STATE: 1946 case O_CHECK_STATE: 1947 /* 1948 * dynamic rules are checked at the first 1949 * keep-state or check-state occurrence, 1950 * with the result being stored in dyn_dir. 1951 * The compiler introduces a PROBE_STATE 1952 * instruction for us when we have a 1953 * KEEP_STATE (because PROBE_STATE needs 1954 * to be run first). 1955 */ 1956 if (dyn_dir == MATCH_UNKNOWN && 1957 (q = ipfw_lookup_dyn_rule(&args->f_id, 1958 &dyn_dir, proto == IPPROTO_TCP ? 1959 TCP(ulp) : NULL)) 1960 != NULL) { 1961 /* 1962 * Found dynamic entry, update stats 1963 * and jump to the 'action' part of 1964 * the parent rule by setting 1965 * f, cmd, l and clearing cmdlen. 1966 */ 1967 q->pcnt++; 1968 q->bcnt += pktlen; 1969 /* XXX we would like to have f_pos 1970 * readily accessible in the dynamic 1971 * rule, instead of having to 1972 * lookup q->rule. 1973 */ 1974 f = q->rule; 1975 f_pos = ipfw_find_rule(chain, 1976 f->rulenum, f->id); 1977 cmd = ACTION_PTR(f); 1978 l = f->cmd_len - f->act_ofs; 1979 ipfw_dyn_unlock(); 1980 cmdlen = 0; 1981 match = 1; 1982 break; 1983 } 1984 /* 1985 * Dynamic entry not found. If CHECK_STATE, 1986 * skip to next rule, if PROBE_STATE just 1987 * ignore and continue with next opcode. 1988 */ 1989 if (cmd->opcode == O_CHECK_STATE) 1990 l = 0; /* exit inner loop */ 1991 match = 1; 1992 break; 1993 1994 case O_ACCEPT: 1995 retval = 0; /* accept */ 1996 l = 0; /* exit inner loop */ 1997 done = 1; /* exit outer loop */ 1998 break; 1999 2000 case O_PIPE: 2001 case O_QUEUE: 2002 set_match(args, f_pos, chain); 2003 args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ? 2004 tablearg : cmd->arg1; 2005 if (cmd->opcode == O_PIPE) 2006 args->rule.info |= IPFW_IS_PIPE; 2007 if (V_fw_one_pass) 2008 args->rule.info |= IPFW_ONEPASS; 2009 retval = IP_FW_DUMMYNET; 2010 l = 0; /* exit inner loop */ 2011 done = 1; /* exit outer loop */ 2012 break; 2013 2014 case O_DIVERT: 2015 case O_TEE: 2016 if (args->eh) /* not on layer 2 */ 2017 break; 2018 /* otherwise this is terminal */ 2019 l = 0; /* exit inner loop */ 2020 done = 1; /* exit outer loop */ 2021 retval = (cmd->opcode == O_DIVERT) ? 2022 IP_FW_DIVERT : IP_FW_TEE; 2023 set_match(args, f_pos, chain); 2024 args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ? 2025 tablearg : cmd->arg1; 2026 break; 2027 2028 case O_COUNT: 2029 f->pcnt++; /* update stats */ 2030 f->bcnt += pktlen; 2031 f->timestamp = time_uptime; 2032 l = 0; /* exit inner loop */ 2033 break; 2034 2035 case O_SKIPTO: 2036 f->pcnt++; /* update stats */ 2037 f->bcnt += pktlen; 2038 f->timestamp = time_uptime; 2039 /* If possible use cached f_pos (in f->next_rule), 2040 * whose version is written in f->next_rule 2041 * (horrible hacks to avoid changing the ABI). 2042 */ 2043 if (cmd->arg1 != IP_FW_TABLEARG && 2044 (uintptr_t)f->x_next == chain->id) { 2045 f_pos = (uintptr_t)f->next_rule; 2046 } else { 2047 int i = (cmd->arg1 == IP_FW_TABLEARG) ? 2048 tablearg : cmd->arg1; 2049 /* make sure we do not jump backward */ 2050 if (i <= f->rulenum) 2051 i = f->rulenum + 1; 2052 f_pos = ipfw_find_rule(chain, i, 0); 2053 /* update the cache */ 2054 if (cmd->arg1 != IP_FW_TABLEARG) { 2055 f->next_rule = 2056 (void *)(uintptr_t)f_pos; 2057 f->x_next = 2058 (void *)(uintptr_t)chain->id; 2059 } 2060 } 2061 /* 2062 * Skip disabled rules, and re-enter 2063 * the inner loop with the correct 2064 * f_pos, f, l and cmd. 2065 * Also clear cmdlen and skip_or 2066 */ 2067 for (; f_pos < chain->n_rules - 1 && 2068 (V_set_disable & 2069 (1 << chain->map[f_pos]->set)); 2070 f_pos++) 2071 ; 2072 /* Re-enter the inner loop at the skipto rule. */ 2073 f = chain->map[f_pos]; 2074 l = f->cmd_len; 2075 cmd = f->cmd; 2076 match = 1; 2077 cmdlen = 0; 2078 skip_or = 0; 2079 continue; 2080 break; /* not reached */ 2081 2082 case O_REJECT: 2083 /* 2084 * Drop the packet and send a reject notice 2085 * if the packet is not ICMP (or is an ICMP 2086 * query), and it is not multicast/broadcast. 2087 */ 2088 if (hlen > 0 && is_ipv4 && offset == 0 && 2089 (proto != IPPROTO_ICMP || 2090 is_icmp_query(ICMP(ulp))) && 2091 !(m->m_flags & (M_BCAST|M_MCAST)) && 2092 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 2093 send_reject(args, cmd->arg1, iplen, ip); 2094 m = args->m; 2095 } 2096 /* FALLTHROUGH */ 2097#ifdef INET6 2098 case O_UNREACH6: 2099 if (hlen > 0 && is_ipv6 && 2100 ((offset & IP6F_OFF_MASK) == 0) && 2101 (proto != IPPROTO_ICMPV6 || 2102 (is_icmp6_query(icmp6_type) == 1)) && 2103 !(m->m_flags & (M_BCAST|M_MCAST)) && 2104 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 2105 send_reject6( 2106 args, cmd->arg1, hlen, 2107 (struct ip6_hdr *)ip); 2108 m = args->m; 2109 } 2110 /* FALLTHROUGH */ 2111#endif 2112 case O_DENY: 2113 retval = IP_FW_DENY; 2114 l = 0; /* exit inner loop */ 2115 done = 1; /* exit outer loop */ 2116 break; 2117 2118 case O_FORWARD_IP: 2119 if (args->eh) /* not valid on layer2 pkts */ 2120 break; 2121 if (q == NULL || q->rule != f || 2122 dyn_dir == MATCH_FORWARD) { 2123 struct sockaddr_in *sa; 2124 sa = &(((ipfw_insn_sa *)cmd)->sa); 2125 if (sa->sin_addr.s_addr == INADDR_ANY) { 2126 bcopy(sa, &args->hopstore, 2127 sizeof(*sa)); 2128 args->hopstore.sin_addr.s_addr = 2129 htonl(tablearg); 2130 args->next_hop = &args->hopstore; 2131 } else { 2132 args->next_hop = sa; 2133 } 2134 } 2135 retval = IP_FW_PASS; 2136 l = 0; /* exit inner loop */ 2137 done = 1; /* exit outer loop */ 2138 break; 2139 2140 case O_NETGRAPH: 2141 case O_NGTEE: 2142 set_match(args, f_pos, chain); 2143 args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ? 2144 tablearg : cmd->arg1; 2145 if (V_fw_one_pass) 2146 args->rule.info |= IPFW_ONEPASS; 2147 retval = (cmd->opcode == O_NETGRAPH) ? 2148 IP_FW_NETGRAPH : IP_FW_NGTEE; 2149 l = 0; /* exit inner loop */ 2150 done = 1; /* exit outer loop */ 2151 break; 2152 2153 case O_SETFIB: { 2154 uint32_t fib; 2155 2156 f->pcnt++; /* update stats */ 2157 f->bcnt += pktlen; 2158 f->timestamp = time_uptime; 2159 fib = (cmd->arg1 == IP_FW_TABLEARG) ? tablearg: 2160 cmd->arg1; 2161 if (fib >= rt_numfibs) 2162 fib = 0; 2163 M_SETFIB(m, fib); 2164 args->f_id.fib = fib; 2165 l = 0; /* exit inner loop */ 2166 break; 2167 } 2168 2169 case O_NAT: 2170 if (!IPFW_NAT_LOADED) { 2171 retval = IP_FW_DENY; 2172 } else { 2173 struct cfg_nat *t; 2174 int nat_id; 2175 2176 set_match(args, f_pos, chain); 2177 t = ((ipfw_insn_nat *)cmd)->nat; 2178 if (t == NULL) { 2179 nat_id = (cmd->arg1 == IP_FW_TABLEARG) ? 2180 tablearg : cmd->arg1; 2181 t = (*lookup_nat_ptr)(&chain->nat, nat_id); 2182 2183 if (t == NULL) { 2184 retval = IP_FW_DENY; 2185 l = 0; /* exit inner loop */ 2186 done = 1; /* exit outer loop */ 2187 break; 2188 } 2189 if (cmd->arg1 != IP_FW_TABLEARG) 2190 ((ipfw_insn_nat *)cmd)->nat = t; 2191 } 2192 retval = ipfw_nat_ptr(args, t, m); 2193 } 2194 l = 0; /* exit inner loop */ 2195 done = 1; /* exit outer loop */ 2196 break; 2197 2198 case O_REASS: { 2199 int ip_off; 2200 2201 f->pcnt++; 2202 f->bcnt += pktlen; 2203 l = 0; /* in any case exit inner loop */ 2204 ip_off = ntohs(ip->ip_off); 2205 2206 /* if not fragmented, go to next rule */ 2207 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0) 2208 break; 2209 /* 2210 * ip_reass() expects len & off in host 2211 * byte order. 2212 */ 2213 SET_HOST_IPLEN(ip); 2214 2215 args->m = m = ip_reass(m); 2216 2217 /* 2218 * do IP header checksum fixup. 2219 */ 2220 if (m == NULL) { /* fragment got swallowed */ 2221 retval = IP_FW_DENY; 2222 } else { /* good, packet complete */ 2223 int hlen; 2224 2225 ip = mtod(m, struct ip *); 2226 hlen = ip->ip_hl << 2; 2227 SET_NET_IPLEN(ip); 2228 ip->ip_sum = 0; 2229 if (hlen == sizeof(struct ip)) 2230 ip->ip_sum = in_cksum_hdr(ip); 2231 else 2232 ip->ip_sum = in_cksum(m, hlen); 2233 retval = IP_FW_REASS; 2234 set_match(args, f_pos, chain); 2235 } 2236 done = 1; /* exit outer loop */ 2237 break; 2238 } 2239 2240 default: 2241 panic("-- unknown opcode %d\n", cmd->opcode); 2242 } /* end of switch() on opcodes */ 2243 /* 2244 * if we get here with l=0, then match is irrelevant. 2245 */ 2246 2247 if (cmd->len & F_NOT) 2248 match = !match; 2249 2250 if (match) { 2251 if (cmd->len & F_OR) 2252 skip_or = 1; 2253 } else { 2254 if (!(cmd->len & F_OR)) /* not an OR block, */ 2255 break; /* try next rule */ 2256 } 2257 2258 } /* end of inner loop, scan opcodes */ 2259#undef PULLUP_LEN 2260 2261 if (done) 2262 break; 2263 2264/* next_rule:; */ /* try next rule */ 2265 2266 } /* end of outer for, scan rules */ 2267 2268 if (done) { 2269 struct ip_fw *rule = chain->map[f_pos]; 2270 /* Update statistics */ 2271 rule->pcnt++; 2272 rule->bcnt += pktlen; 2273 rule->timestamp = time_uptime; 2274 } else { 2275 retval = IP_FW_DENY; 2276 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 2277 } 2278 IPFW_RUNLOCK(chain); 2279#ifdef __FreeBSD__ 2280 if (ucred_cache != NULL) 2281 crfree(ucred_cache); 2282#endif 2283 return (retval); 2284 2285pullup_failed: 2286 if (V_fw_verbose) 2287 printf("ipfw: pullup failed\n"); 2288 return (IP_FW_DENY); 2289} 2290 2291/* 2292 * Module and VNET glue 2293 */ 2294 2295/* 2296 * Stuff that must be initialised only on boot or module load 2297 */ 2298static int 2299ipfw_init(void) 2300{ 2301 int error = 0; 2302 2303 ipfw_dyn_attach(); 2304 /* 2305 * Only print out this stuff the first time around, 2306 * when called from the sysinit code. 2307 */ 2308 printf("ipfw2 " 2309#ifdef INET6 2310 "(+ipv6) " 2311#endif 2312 "initialized, divert %s, nat %s, " 2313 "rule-based forwarding " 2314#ifdef IPFIREWALL_FORWARD 2315 "enabled, " 2316#else 2317 "disabled, " 2318#endif 2319 "default to %s, logging ", 2320#ifdef IPDIVERT 2321 "enabled", 2322#else 2323 "loadable", 2324#endif 2325#ifdef IPFIREWALL_NAT 2326 "enabled", 2327#else 2328 "loadable", 2329#endif 2330 default_to_accept ? "accept" : "deny"); 2331 2332 /* 2333 * Note: V_xxx variables can be accessed here but the vnet specific 2334 * initializer may not have been called yet for the VIMAGE case. 2335 * Tuneables will have been processed. We will print out values for 2336 * the default vnet. 2337 * XXX This should all be rationalized AFTER 8.0 2338 */ 2339 if (V_fw_verbose == 0) 2340 printf("disabled\n"); 2341 else if (V_verbose_limit == 0) 2342 printf("unlimited\n"); 2343 else 2344 printf("limited to %d packets/entry by default\n", 2345 V_verbose_limit); 2346 2347 ipfw_log_bpf(1); /* init */ 2348 return (error); 2349} 2350 2351/* 2352 * Called for the removal of the last instance only on module unload. 2353 */ 2354static void 2355ipfw_destroy(void) 2356{ 2357 2358 ipfw_log_bpf(0); /* uninit */ 2359 ipfw_dyn_detach(); 2360 printf("IP firewall unloaded\n"); 2361} 2362 2363/* 2364 * Stuff that must be initialized for every instance 2365 * (including the first of course). 2366 */ 2367static int 2368vnet_ipfw_init(const void *unused) 2369{ 2370 int error; 2371 struct ip_fw *rule = NULL; 2372 struct ip_fw_chain *chain; 2373 2374 chain = &V_layer3_chain; 2375 2376 /* First set up some values that are compile time options */ 2377 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 2378 V_fw_deny_unknown_exthdrs = 1; 2379#ifdef IPFIREWALL_VERBOSE 2380 V_fw_verbose = 1; 2381#endif 2382#ifdef IPFIREWALL_VERBOSE_LIMIT 2383 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 2384#endif 2385#ifdef IPFIREWALL_NAT 2386 LIST_INIT(&chain->nat); 2387#endif 2388 2389 /* insert the default rule and create the initial map */ 2390 chain->n_rules = 1; 2391 chain->static_len = sizeof(struct ip_fw); 2392 chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_NOWAIT | M_ZERO); 2393 if (chain->map) 2394 rule = malloc(chain->static_len, M_IPFW, M_NOWAIT | M_ZERO); 2395 if (rule == NULL) { 2396 if (chain->map) 2397 free(chain->map, M_IPFW); 2398 printf("ipfw2: ENOSPC initializing default rule " 2399 "(support disabled)\n"); 2400 return (ENOSPC); 2401 } 2402 error = ipfw_init_tables(chain); 2403 if (error) { 2404 panic("init_tables"); /* XXX Marko fix this ! */ 2405 } 2406 2407 /* fill and insert the default rule */ 2408 rule->act_ofs = 0; 2409 rule->rulenum = IPFW_DEFAULT_RULE; 2410 rule->cmd_len = 1; 2411 rule->set = RESVD_SET; 2412 rule->cmd[0].len = 1; 2413 rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; 2414 chain->rules = chain->default_rule = chain->map[0] = rule; 2415 chain->id = rule->id = 1; 2416 2417 IPFW_LOCK_INIT(chain); 2418 ipfw_dyn_init(); 2419 2420 /* First set up some values that are compile time options */ 2421 V_ipfw_vnet_ready = 1; /* Open for business */ 2422 2423 /* 2424 * Hook the sockopt handler, and the layer2 (V_ip_fw_chk_ptr) 2425 * and pfil hooks for ipv4 and ipv6. Even if the latter two fail 2426 * we still keep the module alive because the sockopt and 2427 * layer2 paths are still useful. 2428 * ipfw[6]_hook return 0 on success, ENOENT on failure, 2429 * so we can ignore the exact return value and just set a flag. 2430 * 2431 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so 2432 * changes in the underlying (per-vnet) variables trigger 2433 * immediate hook()/unhook() calls. 2434 * In layer2 we have the same behaviour, except that V_ether_ipfw 2435 * is checked on each packet because there are no pfil hooks. 2436 */ 2437 V_ip_fw_ctl_ptr = ipfw_ctl; 2438 V_ip_fw_chk_ptr = ipfw_chk; 2439 error = ipfw_attach_hooks(1); 2440 return (error); 2441} 2442 2443/* 2444 * Called for the removal of each instance. 2445 */ 2446static int 2447vnet_ipfw_uninit(const void *unused) 2448{ 2449 struct ip_fw *reap, *rule; 2450 struct ip_fw_chain *chain = &V_layer3_chain; 2451 int i; 2452 2453 V_ipfw_vnet_ready = 0; /* tell new callers to go away */ 2454 /* 2455 * disconnect from ipv4, ipv6, layer2 and sockopt. 2456 * Then grab, release and grab again the WLOCK so we make 2457 * sure the update is propagated and nobody will be in. 2458 */ 2459 (void)ipfw_attach_hooks(0 /* detach */); 2460 V_ip_fw_chk_ptr = NULL; 2461 V_ip_fw_ctl_ptr = NULL; 2462 IPFW_UH_WLOCK(chain); 2463 IPFW_UH_WUNLOCK(chain); 2464 IPFW_UH_WLOCK(chain); 2465 2466 IPFW_WLOCK(chain); 2467 IPFW_WUNLOCK(chain); 2468 IPFW_WLOCK(chain); 2469 2470 ipfw_dyn_uninit(0); /* run the callout_drain */ 2471 ipfw_destroy_tables(chain); 2472 reap = NULL; 2473 for (i = 0; i < chain->n_rules; i++) { 2474 rule = chain->map[i]; 2475 rule->x_next = reap; 2476 reap = rule; 2477 } 2478 if (chain->map) 2479 free(chain->map, M_IPFW); 2480 IPFW_WUNLOCK(chain); 2481 IPFW_UH_WUNLOCK(chain); 2482 if (reap != NULL) 2483 ipfw_reap_rules(reap); 2484 IPFW_LOCK_DESTROY(chain); 2485 ipfw_dyn_uninit(1); /* free the remaining parts */ 2486 return 0; 2487} 2488 2489/* 2490 * Module event handler. 2491 * In general we have the choice of handling most of these events by the 2492 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to 2493 * use the SYSINIT handlers as they are more capable of expressing the 2494 * flow of control during module and vnet operations, so this is just 2495 * a skeleton. Note there is no SYSINIT equivalent of the module 2496 * SHUTDOWN handler, but we don't have anything to do in that case anyhow. 2497 */ 2498static int 2499ipfw_modevent(module_t mod, int type, void *unused) 2500{ 2501 int err = 0; 2502 2503 switch (type) { 2504 case MOD_LOAD: 2505 /* Called once at module load or 2506 * system boot if compiled in. */ 2507 break; 2508 case MOD_QUIESCE: 2509 /* Called before unload. May veto unloading. */ 2510 break; 2511 case MOD_UNLOAD: 2512 /* Called during unload. */ 2513 break; 2514 case MOD_SHUTDOWN: 2515 /* Called during system shutdown. */ 2516 break; 2517 default: 2518 err = EOPNOTSUPP; 2519 break; 2520 } 2521 return err; 2522} 2523 2524static moduledata_t ipfwmod = { 2525 "ipfw", 2526 ipfw_modevent, 2527 0 2528}; 2529 2530/* Define startup order. */ 2531#define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_IFATTACHDOMAIN 2532#define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */ 2533#define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */ 2534#define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */ 2535 2536DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER); 2537MODULE_VERSION(ipfw, 2); 2538/* should declare some dependencies here */ 2539 2540/* 2541 * Starting up. Done in order after ipfwmod() has been called. 2542 * VNET_SYSINIT is also called for each existing vnet and each new vnet. 2543 */ 2544SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 2545 ipfw_init, NULL); 2546VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 2547 vnet_ipfw_init, NULL); 2548 2549/* 2550 * Closing up shop. These are done in REVERSE ORDER, but still 2551 * after ipfwmod() has been called. Not called on reboot. 2552 * VNET_SYSUNINIT is also called for each exiting vnet as it exits. 2553 * or when the module is unloaded. 2554 */ 2555SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 2556 ipfw_destroy, NULL); 2557VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 2558 vnet_ipfw_uninit, NULL); 2559/* end of file */ 2560