ip_fw2.c revision 201046
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 201046 2009-12-27 10:19:10Z 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/* counter for ipfw_log(NULL...) */ 135VNET_DEFINE(u_int64_t, norule_counter); 136VNET_DEFINE(int, verbose_limit); 137 138/* layer3_chain contains the list of rules for layer 3 */ 139VNET_DEFINE(struct ip_fw_chain, layer3_chain); 140 141ipfw_nat_t *ipfw_nat_ptr = NULL; 142struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int); 143ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; 144ipfw_nat_cfg_t *ipfw_nat_del_ptr; 145ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; 146ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; 147 148#ifdef SYSCTL_NODE 149SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 150SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 151 CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0, 152 "Only do a single pass through ipfw when using dummynet(4)"); 153SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, 154 CTLFLAG_RW, &VNET_NAME(autoinc_step), 0, 155 "Rule number auto-increment step"); 156SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose, 157 CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0, 158 "Log matches to ipfw rules"); 159SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, 160 CTLFLAG_RW, &VNET_NAME(verbose_limit), 0, 161 "Set upper limit of matches of ipfw rules logged"); 162SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, 163 NULL, IPFW_DEFAULT_RULE, 164 "The default/max possible rule number."); 165SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, tables_max, CTLFLAG_RD, 166 NULL, IPFW_TABLES_MAX, 167 "The maximum number of tables."); 168SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN, 169 &default_to_accept, 0, 170 "Make the default rule accept all packets."); 171TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept); 172SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, static_count, 173 CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0, 174 "Number of static rules"); 175 176#ifdef INET6 177SYSCTL_DECL(_net_inet6_ip6); 178SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 179SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs, 180 CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0, 181 "Deny packets with unknown IPv6 Extension Headers"); 182#endif /* INET6 */ 183 184#endif /* SYSCTL_NODE */ 185 186 187/* 188 * Some macros used in the various matching options. 189 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 190 * Other macros just cast void * into the appropriate type 191 */ 192#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 193#define TCP(p) ((struct tcphdr *)(p)) 194#define SCTP(p) ((struct sctphdr *)(p)) 195#define UDP(p) ((struct udphdr *)(p)) 196#define ICMP(p) ((struct icmphdr *)(p)) 197#define ICMP6(p) ((struct icmp6_hdr *)(p)) 198 199static __inline int 200icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 201{ 202 int type = icmp->icmp_type; 203 204 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 205} 206 207#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 208 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 209 210static int 211is_icmp_query(struct icmphdr *icmp) 212{ 213 int type = icmp->icmp_type; 214 215 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 216} 217#undef TT 218 219/* 220 * The following checks use two arrays of 8 or 16 bits to store the 221 * bits that we want set or clear, respectively. They are in the 222 * low and high half of cmd->arg1 or cmd->d[0]. 223 * 224 * We scan options and store the bits we find set. We succeed if 225 * 226 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 227 * 228 * The code is sometimes optimized not to store additional variables. 229 */ 230 231static int 232flags_match(ipfw_insn *cmd, u_int8_t bits) 233{ 234 u_char want_clear; 235 bits = ~bits; 236 237 if ( ((cmd->arg1 & 0xff) & bits) != 0) 238 return 0; /* some bits we want set were clear */ 239 want_clear = (cmd->arg1 >> 8) & 0xff; 240 if ( (want_clear & bits) != want_clear) 241 return 0; /* some bits we want clear were set */ 242 return 1; 243} 244 245static int 246ipopts_match(struct ip *ip, ipfw_insn *cmd) 247{ 248 int optlen, bits = 0; 249 u_char *cp = (u_char *)(ip + 1); 250 int x = (ip->ip_hl << 2) - sizeof (struct ip); 251 252 for (; x > 0; x -= optlen, cp += optlen) { 253 int opt = cp[IPOPT_OPTVAL]; 254 255 if (opt == IPOPT_EOL) 256 break; 257 if (opt == IPOPT_NOP) 258 optlen = 1; 259 else { 260 optlen = cp[IPOPT_OLEN]; 261 if (optlen <= 0 || optlen > x) 262 return 0; /* invalid or truncated */ 263 } 264 switch (opt) { 265 266 default: 267 break; 268 269 case IPOPT_LSRR: 270 bits |= IP_FW_IPOPT_LSRR; 271 break; 272 273 case IPOPT_SSRR: 274 bits |= IP_FW_IPOPT_SSRR; 275 break; 276 277 case IPOPT_RR: 278 bits |= IP_FW_IPOPT_RR; 279 break; 280 281 case IPOPT_TS: 282 bits |= IP_FW_IPOPT_TS; 283 break; 284 } 285 } 286 return (flags_match(cmd, bits)); 287} 288 289static int 290tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 291{ 292 int optlen, bits = 0; 293 u_char *cp = (u_char *)(tcp + 1); 294 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 295 296 for (; x > 0; x -= optlen, cp += optlen) { 297 int opt = cp[0]; 298 if (opt == TCPOPT_EOL) 299 break; 300 if (opt == TCPOPT_NOP) 301 optlen = 1; 302 else { 303 optlen = cp[1]; 304 if (optlen <= 0) 305 break; 306 } 307 308 switch (opt) { 309 310 default: 311 break; 312 313 case TCPOPT_MAXSEG: 314 bits |= IP_FW_TCPOPT_MSS; 315 break; 316 317 case TCPOPT_WINDOW: 318 bits |= IP_FW_TCPOPT_WINDOW; 319 break; 320 321 case TCPOPT_SACK_PERMITTED: 322 case TCPOPT_SACK: 323 bits |= IP_FW_TCPOPT_SACK; 324 break; 325 326 case TCPOPT_TIMESTAMP: 327 bits |= IP_FW_TCPOPT_TS; 328 break; 329 330 } 331 } 332 return (flags_match(cmd, bits)); 333} 334 335static int 336iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 337{ 338 if (ifp == NULL) /* no iface with this packet, match fails */ 339 return 0; 340 /* Check by name or by IP address */ 341 if (cmd->name[0] != '\0') { /* match by name */ 342 /* Check name */ 343 if (cmd->p.glob) { 344 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 345 return(1); 346 } else { 347 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 348 return(1); 349 } 350 } else { 351 struct ifaddr *ia; 352 353 if_addr_rlock(ifp); 354 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 355 if (ia->ifa_addr->sa_family != AF_INET) 356 continue; 357 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 358 (ia->ifa_addr))->sin_addr.s_addr) { 359 if_addr_runlock(ifp); 360 return(1); /* match */ 361 } 362 } 363 if_addr_runlock(ifp); 364 } 365 return(0); /* no match, fail ... */ 366} 367 368/* 369 * The verify_path function checks if a route to the src exists and 370 * if it is reachable via ifp (when provided). 371 * 372 * The 'verrevpath' option checks that the interface that an IP packet 373 * arrives on is the same interface that traffic destined for the 374 * packet's source address would be routed out of. 375 * The 'versrcreach' option just checks that the source address is 376 * reachable via any route (except default) in the routing table. 377 * These two are a measure to block forged packets. This is also 378 * commonly known as "anti-spoofing" or Unicast Reverse Path 379 * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 380 * is purposely reminiscent of the Cisco IOS command, 381 * 382 * ip verify unicast reverse-path 383 * ip verify unicast source reachable-via any 384 * 385 * which implements the same functionality. But note that the syntax 386 * is misleading, and the check may be performed on all IP packets 387 * whether unicast, multicast, or broadcast. 388 */ 389static int 390verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) 391{ 392 struct route ro; 393 struct sockaddr_in *dst; 394 395 bzero(&ro, sizeof(ro)); 396 397 dst = (struct sockaddr_in *)&(ro.ro_dst); 398 dst->sin_family = AF_INET; 399 dst->sin_len = sizeof(*dst); 400 dst->sin_addr = src; 401 in_rtalloc_ign(&ro, 0, fib); 402 403 if (ro.ro_rt == NULL) 404 return 0; 405 406 /* 407 * If ifp is provided, check for equality with rtentry. 408 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 409 * in order to pass packets injected back by if_simloop(): 410 * if useloopback == 1 routing entry (via lo0) for our own address 411 * may exist, so we need to handle routing assymetry. 412 */ 413 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 414 RTFREE(ro.ro_rt); 415 return 0; 416 } 417 418 /* if no ifp provided, check if rtentry is not default route */ 419 if (ifp == NULL && 420 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 421 RTFREE(ro.ro_rt); 422 return 0; 423 } 424 425 /* or if this is a blackhole/reject route */ 426 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 427 RTFREE(ro.ro_rt); 428 return 0; 429 } 430 431 /* found valid route */ 432 RTFREE(ro.ro_rt); 433 return 1; 434} 435 436#ifdef INET6 437/* 438 * ipv6 specific rules here... 439 */ 440static __inline int 441icmp6type_match (int type, ipfw_insn_u32 *cmd) 442{ 443 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 444} 445 446static int 447flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 448{ 449 int i; 450 for (i=0; i <= cmd->o.arg1; ++i ) 451 if (curr_flow == cmd->d[i] ) 452 return 1; 453 return 0; 454} 455 456/* support for IP6_*_ME opcodes */ 457static int 458search_ip6_addr_net (struct in6_addr * ip6_addr) 459{ 460 struct ifnet *mdc; 461 struct ifaddr *mdc2; 462 struct in6_ifaddr *fdm; 463 struct in6_addr copia; 464 465 TAILQ_FOREACH(mdc, &V_ifnet, if_link) { 466 if_addr_rlock(mdc); 467 TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) { 468 if (mdc2->ifa_addr->sa_family == AF_INET6) { 469 fdm = (struct in6_ifaddr *)mdc2; 470 copia = fdm->ia_addr.sin6_addr; 471 /* need for leaving scope_id in the sock_addr */ 472 in6_clearscope(&copia); 473 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) { 474 if_addr_runlock(mdc); 475 return 1; 476 } 477 } 478 } 479 if_addr_runlock(mdc); 480 } 481 return 0; 482} 483 484static int 485verify_path6(struct in6_addr *src, struct ifnet *ifp) 486{ 487 struct route_in6 ro; 488 struct sockaddr_in6 *dst; 489 490 bzero(&ro, sizeof(ro)); 491 492 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 493 dst->sin6_family = AF_INET6; 494 dst->sin6_len = sizeof(*dst); 495 dst->sin6_addr = *src; 496 /* XXX MRT 0 for ipv6 at this time */ 497 rtalloc_ign((struct route *)&ro, 0); 498 499 if (ro.ro_rt == NULL) 500 return 0; 501 502 /* 503 * if ifp is provided, check for equality with rtentry 504 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 505 * to support the case of sending packets to an address of our own. 506 * (where the former interface is the first argument of if_simloop() 507 * (=ifp), the latter is lo0) 508 */ 509 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 510 RTFREE(ro.ro_rt); 511 return 0; 512 } 513 514 /* if no ifp provided, check if rtentry is not default route */ 515 if (ifp == NULL && 516 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 517 RTFREE(ro.ro_rt); 518 return 0; 519 } 520 521 /* or if this is a blackhole/reject route */ 522 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 523 RTFREE(ro.ro_rt); 524 return 0; 525 } 526 527 /* found valid route */ 528 RTFREE(ro.ro_rt); 529 return 1; 530 531} 532 533static int 534is_icmp6_query(int icmp6_type) 535{ 536 if ((icmp6_type <= ICMP6_MAXTYPE) && 537 (icmp6_type == ICMP6_ECHO_REQUEST || 538 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 539 icmp6_type == ICMP6_WRUREQUEST || 540 icmp6_type == ICMP6_FQDN_QUERY || 541 icmp6_type == ICMP6_NI_QUERY)) 542 return (1); 543 544 return (0); 545} 546 547static void 548send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 549{ 550 struct mbuf *m; 551 552 m = args->m; 553 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 554 struct tcphdr *tcp; 555 tcp = (struct tcphdr *)((char *)ip6 + hlen); 556 557 if ((tcp->th_flags & TH_RST) == 0) { 558 struct mbuf *m0; 559 m0 = ipfw_send_pkt(args->m, &(args->f_id), 560 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 561 tcp->th_flags | TH_RST); 562 if (m0 != NULL) 563 ip6_output(m0, NULL, NULL, 0, NULL, NULL, 564 NULL); 565 } 566 m_freem(m); 567 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 568#if 0 569 /* 570 * Unlike above, the mbufs need to line up with the ip6 hdr, 571 * as the contents are read. We need to m_adj() the 572 * needed amount. 573 * The mbuf will however be thrown away so we can adjust it. 574 * Remember we did an m_pullup on it already so we 575 * can make some assumptions about contiguousness. 576 */ 577 if (args->L3offset) 578 m_adj(m, args->L3offset); 579#endif 580 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 581 } else 582 m_freem(m); 583 584 args->m = NULL; 585} 586 587#endif /* INET6 */ 588 589 590/* 591 * sends a reject message, consuming the mbuf passed as an argument. 592 */ 593static void 594send_reject(struct ip_fw_args *args, int code, int ip_len, struct ip *ip) 595{ 596 597#if 0 598 /* XXX When ip is not guaranteed to be at mtod() we will 599 * need to account for this */ 600 * The mbuf will however be thrown away so we can adjust it. 601 * Remember we did an m_pullup on it already so we 602 * can make some assumptions about contiguousness. 603 */ 604 if (args->L3offset) 605 m_adj(m, args->L3offset); 606#endif 607 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 608 /* We need the IP header in host order for icmp_error(). */ 609 if (args->eh != NULL) { 610 ip->ip_len = ntohs(ip->ip_len); 611 ip->ip_off = ntohs(ip->ip_off); 612 } 613 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 614 } else if (args->f_id.proto == IPPROTO_TCP) { 615 struct tcphdr *const tcp = 616 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 617 if ( (tcp->th_flags & TH_RST) == 0) { 618 struct mbuf *m; 619 m = ipfw_send_pkt(args->m, &(args->f_id), 620 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 621 tcp->th_flags | TH_RST); 622 if (m != NULL) 623 ip_output(m, NULL, NULL, 0, NULL, NULL); 624 } 625 m_freem(args->m); 626 } else 627 m_freem(args->m); 628 args->m = NULL; 629} 630 631/* 632 * Support for uid/gid/jail lookup. These tests are expensive 633 * (because we may need to look into the list of active sockets) 634 * so we cache the results. ugid_lookupp is 0 if we have not 635 * yet done a lookup, 1 if we succeeded, and -1 if we tried 636 * and failed. The function always returns the match value. 637 * We could actually spare the variable and use *uc, setting 638 * it to '(void *)check_uidgid if we have no info, NULL if 639 * we tried and failed, or any other value if successful. 640 */ 641static int 642check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif, 643 struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip, 644 u_int16_t src_port, struct ucred **uc, int *ugid_lookupp, 645 struct inpcb *inp) 646{ 647 struct inpcbinfo *pi; 648 int wildcard; 649 struct inpcb *pcb; 650 int match; 651 652 /* 653 * Check to see if the UDP or TCP stack supplied us with 654 * the PCB. If so, rather then holding a lock and looking 655 * up the PCB, we can use the one that was supplied. 656 */ 657 if (inp && *ugid_lookupp == 0) { 658 INP_LOCK_ASSERT(inp); 659 if (inp->inp_socket != NULL) { 660 *uc = crhold(inp->inp_cred); 661 *ugid_lookupp = 1; 662 } else 663 *ugid_lookupp = -1; 664 } 665 /* 666 * If we have already been here and the packet has no 667 * PCB entry associated with it, then we can safely 668 * assume that this is a no match. 669 */ 670 if (*ugid_lookupp == -1) 671 return (0); 672 if (proto == IPPROTO_TCP) { 673 wildcard = 0; 674 pi = &V_tcbinfo; 675 } else if (proto == IPPROTO_UDP) { 676 wildcard = INPLOOKUP_WILDCARD; 677 pi = &V_udbinfo; 678 } else 679 return 0; 680 match = 0; 681 if (*ugid_lookupp == 0) { 682 INP_INFO_RLOCK(pi); 683 pcb = (oif) ? 684 in_pcblookup_hash(pi, 685 dst_ip, htons(dst_port), 686 src_ip, htons(src_port), 687 wildcard, oif) : 688 in_pcblookup_hash(pi, 689 src_ip, htons(src_port), 690 dst_ip, htons(dst_port), 691 wildcard, NULL); 692 if (pcb != NULL) { 693 *uc = crhold(pcb->inp_cred); 694 *ugid_lookupp = 1; 695 } 696 INP_INFO_RUNLOCK(pi); 697 if (*ugid_lookupp == 0) { 698 /* 699 * We tried and failed, set the variable to -1 700 * so we will not try again on this packet. 701 */ 702 *ugid_lookupp = -1; 703 return (0); 704 } 705 } 706 if (insn->o.opcode == O_UID) 707 match = ((*uc)->cr_uid == (uid_t)insn->d[0]); 708 else if (insn->o.opcode == O_GID) 709 match = groupmember((gid_t)insn->d[0], *uc); 710 else if (insn->o.opcode == O_JAIL) 711 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]); 712 return match; 713} 714 715/* 716 * Helper function to write the matching rule into args 717 */ 718static inline void 719set_match(struct ip_fw_args *args, int slot, 720 struct ip_fw_chain *chain) 721{ 722 args->chain_id = chain->id; 723 args->slot = slot + 1; /* we use 0 as a marker */ 724 args->rule_id = chain->map[slot]->id; 725 args->rulenum = chain->map[slot]->rulenum; 726} 727 728/* 729 * The main check routine for the firewall. 730 * 731 * All arguments are in args so we can modify them and return them 732 * back to the caller. 733 * 734 * Parameters: 735 * 736 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 737 * Starts with the IP header. 738 * args->eh (in) Mac header if present, NULL for layer3 packet. 739 * args->L3offset Number of bytes bypassed if we came from L2. 740 * e.g. often sizeof(eh) ** NOTYET ** 741 * args->oif Outgoing interface, NULL if packet is incoming. 742 * The incoming interface is in the mbuf. (in) 743 * args->divert_rule (in/out) 744 * Skip up to the first rule past this rule number; 745 * upon return, non-zero port number for divert or tee. 746 * 747 * args->rule Pointer to the last matching rule (in/out) 748 * args->next_hop Socket we are forwarding to (out). 749 * args->f_id Addresses grabbed from the packet (out) 750 * args->cookie a cookie depending on rule action 751 * 752 * Return value: 753 * 754 * IP_FW_PASS the packet must be accepted 755 * IP_FW_DENY the packet must be dropped 756 * IP_FW_DIVERT divert packet, port in m_tag 757 * IP_FW_TEE tee packet, port in m_tag 758 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 759 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 760 * 761 */ 762int 763ipfw_chk(struct ip_fw_args *args) 764{ 765 766 /* 767 * Local variables holding state while processing a packet: 768 * 769 * IMPORTANT NOTE: to speed up the processing of rules, there 770 * are some assumption on the values of the variables, which 771 * are documented here. Should you change them, please check 772 * the implementation of the various instructions to make sure 773 * that they still work. 774 * 775 * args->eh The MAC header. It is non-null for a layer2 776 * packet, it is NULL for a layer-3 packet. 777 * **notyet** 778 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. 779 * 780 * m | args->m Pointer to the mbuf, as received from the caller. 781 * It may change if ipfw_chk() does an m_pullup, or if it 782 * consumes the packet because it calls send_reject(). 783 * XXX This has to change, so that ipfw_chk() never modifies 784 * or consumes the buffer. 785 * ip is the beginning of the ip(4 or 6) header. 786 * Calculated by adding the L3offset to the start of data. 787 * (Until we start using L3offset, the packet is 788 * supposed to start with the ip header). 789 */ 790 struct mbuf *m = args->m; 791 struct ip *ip = mtod(m, struct ip *); 792 793 /* 794 * For rules which contain uid/gid or jail constraints, cache 795 * a copy of the users credentials after the pcb lookup has been 796 * executed. This will speed up the processing of rules with 797 * these types of constraints, as well as decrease contention 798 * on pcb related locks. 799 */ 800 struct ucred *ucred_cache = NULL; 801 int ucred_lookup = 0; 802 803 /* 804 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 805 * associated with a packet input on a divert socket. This 806 * will allow to distinguish traffic and its direction when 807 * it originates from a divert socket. 808 */ 809 u_int divinput_flags = 0; 810 811 /* 812 * oif | args->oif If NULL, ipfw_chk has been called on the 813 * inbound path (ether_input, ip_input). 814 * If non-NULL, ipfw_chk has been called on the outbound path 815 * (ether_output, ip_output). 816 */ 817 struct ifnet *oif = args->oif; 818 819 int f_pos = 0; /* index of current rule in the array */ 820 int retval = 0; 821 822 /* 823 * hlen The length of the IP header. 824 */ 825 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 826 827 /* 828 * offset The offset of a fragment. offset != 0 means that 829 * we have a fragment at this offset of an IPv4 packet. 830 * offset == 0 means that (if this is an IPv4 packet) 831 * this is the first or only fragment. 832 * For IPv6 offset == 0 means there is no Fragment Header. 833 * If offset != 0 for IPv6 always use correct mask to 834 * get the correct offset because we add IP6F_MORE_FRAG 835 * to be able to dectect the first fragment which would 836 * otherwise have offset = 0. 837 */ 838 u_short offset = 0; 839 840 /* 841 * Local copies of addresses. They are only valid if we have 842 * an IP packet. 843 * 844 * proto The protocol. Set to 0 for non-ip packets, 845 * or to the protocol read from the packet otherwise. 846 * proto != 0 means that we have an IPv4 packet. 847 * 848 * src_port, dst_port port numbers, in HOST format. Only 849 * valid for TCP and UDP packets. 850 * 851 * src_ip, dst_ip ip addresses, in NETWORK format. 852 * Only valid for IPv4 packets. 853 */ 854 u_int8_t proto; 855 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 856 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 857 u_int16_t ip_len=0; 858 int pktlen; 859 u_int16_t etype = 0; /* Host order stored ether type */ 860 861 /* 862 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 863 * MATCH_NONE when checked and not matched (q = NULL), 864 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 865 */ 866 int dyn_dir = MATCH_UNKNOWN; 867 ipfw_dyn_rule *q = NULL; 868 struct ip_fw_chain *chain = &V_layer3_chain; 869 struct m_tag *mtag; 870 871 /* 872 * We store in ulp a pointer to the upper layer protocol header. 873 * In the ipv4 case this is easy to determine from the header, 874 * but for ipv6 we might have some additional headers in the middle. 875 * ulp is NULL if not found. 876 */ 877 void *ulp = NULL; /* upper layer protocol pointer. */ 878 /* XXX ipv6 variables */ 879 int is_ipv6 = 0; 880 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 881 /* end of ipv6 variables */ 882 int is_ipv4 = 0; 883 884 int done = 0; /* flag to exit the outer loop */ 885 886 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready)) 887 return (IP_FW_PASS); /* accept */ 888 889 dst_ip.s_addr = 0; /* make sure it is initialized */ 890 src_ip.s_addr = 0; /* make sure it is initialized */ 891 pktlen = m->m_pkthdr.len; 892 args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */ 893 proto = args->f_id.proto = 0; /* mark f_id invalid */ 894 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 895 896/* 897 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 898 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 899 * pointer might become stale after other pullups (but we never use it 900 * this way). 901 */ 902#define PULLUP_TO(_len, p, T) \ 903do { \ 904 int x = (_len) + sizeof(T); \ 905 if ((m)->m_len < x) { \ 906 args->m = m = m_pullup(m, x); \ 907 if (m == NULL) \ 908 goto pullup_failed; \ 909 } \ 910 p = (mtod(m, char *) + (_len)); \ 911} while (0) 912 913 /* 914 * if we have an ether header, 915 */ 916 if (args->eh) 917 etype = ntohs(args->eh->ether_type); 918 919 /* Identify IP packets and fill up variables. */ 920 if (pktlen >= sizeof(struct ip6_hdr) && 921 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { 922 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 923 is_ipv6 = 1; 924 args->f_id.addr_type = 6; 925 hlen = sizeof(struct ip6_hdr); 926 proto = ip6->ip6_nxt; 927 928 /* Search extension headers to find upper layer protocols */ 929 while (ulp == NULL) { 930 switch (proto) { 931 case IPPROTO_ICMPV6: 932 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 933 args->f_id.flags = ICMP6(ulp)->icmp6_type; 934 break; 935 936 case IPPROTO_TCP: 937 PULLUP_TO(hlen, ulp, struct tcphdr); 938 dst_port = TCP(ulp)->th_dport; 939 src_port = TCP(ulp)->th_sport; 940 args->f_id.flags = TCP(ulp)->th_flags; 941 break; 942 943 case IPPROTO_SCTP: 944 PULLUP_TO(hlen, ulp, struct sctphdr); 945 src_port = SCTP(ulp)->src_port; 946 dst_port = SCTP(ulp)->dest_port; 947 break; 948 949 case IPPROTO_UDP: 950 PULLUP_TO(hlen, ulp, struct udphdr); 951 dst_port = UDP(ulp)->uh_dport; 952 src_port = UDP(ulp)->uh_sport; 953 break; 954 955 case IPPROTO_HOPOPTS: /* RFC 2460 */ 956 PULLUP_TO(hlen, ulp, struct ip6_hbh); 957 ext_hd |= EXT_HOPOPTS; 958 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 959 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 960 ulp = NULL; 961 break; 962 963 case IPPROTO_ROUTING: /* RFC 2460 */ 964 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 965 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 966 case 0: 967 ext_hd |= EXT_RTHDR0; 968 break; 969 case 2: 970 ext_hd |= EXT_RTHDR2; 971 break; 972 default: 973 printf("IPFW2: IPV6 - Unknown Routing " 974 "Header type(%d)\n", 975 ((struct ip6_rthdr *)ulp)->ip6r_type); 976 if (V_fw_deny_unknown_exthdrs) 977 return (IP_FW_DENY); 978 break; 979 } 980 ext_hd |= EXT_ROUTING; 981 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 982 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 983 ulp = NULL; 984 break; 985 986 case IPPROTO_FRAGMENT: /* RFC 2460 */ 987 PULLUP_TO(hlen, ulp, struct ip6_frag); 988 ext_hd |= EXT_FRAGMENT; 989 hlen += sizeof (struct ip6_frag); 990 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 991 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 992 IP6F_OFF_MASK; 993 /* Add IP6F_MORE_FRAG for offset of first 994 * fragment to be != 0. */ 995 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 996 IP6F_MORE_FRAG; 997 if (offset == 0) { 998 printf("IPFW2: IPV6 - Invalid Fragment " 999 "Header\n"); 1000 if (V_fw_deny_unknown_exthdrs) 1001 return (IP_FW_DENY); 1002 break; 1003 } 1004 args->f_id.frag_id6 = 1005 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 1006 ulp = NULL; 1007 break; 1008 1009 case IPPROTO_DSTOPTS: /* RFC 2460 */ 1010 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1011 ext_hd |= EXT_DSTOPTS; 1012 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1013 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1014 ulp = NULL; 1015 break; 1016 1017 case IPPROTO_AH: /* RFC 2402 */ 1018 PULLUP_TO(hlen, ulp, struct ip6_ext); 1019 ext_hd |= EXT_AH; 1020 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 1021 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 1022 ulp = NULL; 1023 break; 1024 1025 case IPPROTO_ESP: /* RFC 2406 */ 1026 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 1027 /* Anything past Seq# is variable length and 1028 * data past this ext. header is encrypted. */ 1029 ext_hd |= EXT_ESP; 1030 break; 1031 1032 case IPPROTO_NONE: /* RFC 2460 */ 1033 /* 1034 * Packet ends here, and IPv6 header has 1035 * already been pulled up. If ip6e_len!=0 1036 * then octets must be ignored. 1037 */ 1038 ulp = ip; /* non-NULL to get out of loop. */ 1039 break; 1040 1041 case IPPROTO_OSPFIGP: 1042 /* XXX OSPF header check? */ 1043 PULLUP_TO(hlen, ulp, struct ip6_ext); 1044 break; 1045 1046 case IPPROTO_PIM: 1047 /* XXX PIM header check? */ 1048 PULLUP_TO(hlen, ulp, struct pim); 1049 break; 1050 1051 case IPPROTO_CARP: 1052 PULLUP_TO(hlen, ulp, struct carp_header); 1053 if (((struct carp_header *)ulp)->carp_version != 1054 CARP_VERSION) 1055 return (IP_FW_DENY); 1056 if (((struct carp_header *)ulp)->carp_type != 1057 CARP_ADVERTISEMENT) 1058 return (IP_FW_DENY); 1059 break; 1060 1061 case IPPROTO_IPV6: /* RFC 2893 */ 1062 PULLUP_TO(hlen, ulp, struct ip6_hdr); 1063 break; 1064 1065 case IPPROTO_IPV4: /* RFC 2893 */ 1066 PULLUP_TO(hlen, ulp, struct ip); 1067 break; 1068 1069 default: 1070 printf("IPFW2: IPV6 - Unknown Extension " 1071 "Header(%d), ext_hd=%x\n", proto, ext_hd); 1072 if (V_fw_deny_unknown_exthdrs) 1073 return (IP_FW_DENY); 1074 PULLUP_TO(hlen, ulp, struct ip6_ext); 1075 break; 1076 } /*switch */ 1077 } 1078 ip = mtod(m, struct ip *); 1079 ip6 = (struct ip6_hdr *)ip; 1080 args->f_id.src_ip6 = ip6->ip6_src; 1081 args->f_id.dst_ip6 = ip6->ip6_dst; 1082 args->f_id.src_ip = 0; 1083 args->f_id.dst_ip = 0; 1084 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 1085 } else if (pktlen >= sizeof(struct ip) && 1086 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { 1087 is_ipv4 = 1; 1088 hlen = ip->ip_hl << 2; 1089 args->f_id.addr_type = 4; 1090 1091 /* 1092 * Collect parameters into local variables for faster matching. 1093 */ 1094 proto = ip->ip_p; 1095 src_ip = ip->ip_src; 1096 dst_ip = ip->ip_dst; 1097 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 1098 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1099 ip_len = ntohs(ip->ip_len); 1100 } else { 1101 offset = ip->ip_off & IP_OFFMASK; 1102 ip_len = ip->ip_len; 1103 } 1104 pktlen = ip_len < pktlen ? ip_len : pktlen; 1105 1106 if (offset == 0) { 1107 switch (proto) { 1108 case IPPROTO_TCP: 1109 PULLUP_TO(hlen, ulp, struct tcphdr); 1110 dst_port = TCP(ulp)->th_dport; 1111 src_port = TCP(ulp)->th_sport; 1112 args->f_id.flags = TCP(ulp)->th_flags; 1113 break; 1114 1115 case IPPROTO_UDP: 1116 PULLUP_TO(hlen, ulp, struct udphdr); 1117 dst_port = UDP(ulp)->uh_dport; 1118 src_port = UDP(ulp)->uh_sport; 1119 break; 1120 1121 case IPPROTO_ICMP: 1122 PULLUP_TO(hlen, ulp, struct icmphdr); 1123 args->f_id.flags = ICMP(ulp)->icmp_type; 1124 break; 1125 1126 default: 1127 break; 1128 } 1129 } 1130 1131 ip = mtod(m, struct ip *); 1132 args->f_id.src_ip = ntohl(src_ip.s_addr); 1133 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1134 } 1135#undef PULLUP_TO 1136 if (proto) { /* we may have port numbers, store them */ 1137 args->f_id.proto = proto; 1138 args->f_id.src_port = src_port = ntohs(src_port); 1139 args->f_id.dst_port = dst_port = ntohs(dst_port); 1140 } 1141 1142 IPFW_RLOCK(chain); 1143 if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */ 1144 IPFW_RUNLOCK(chain); 1145 return (IP_FW_PASS); /* accept */ 1146 } 1147 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 1148 if (args->slot) { 1149 /* 1150 * Packet has already been tagged as a result of a previous 1151 * match on rule args->rule aka args->rule_id (PIPE, QUEUE, 1152 * REASS, NETGRAPH and similar, never a skipto). 1153 * Validate the slot and continue from the next one 1154 * if still present, otherwise do a lookup. 1155 */ 1156 if (V_fw_one_pass) { 1157 IPFW_RUNLOCK(chain); 1158 return (IP_FW_PASS); 1159 } 1160 f_pos = (args->chain_id == chain->id) ? 1161 args->slot /* already incremented */ : 1162 ipfw_find_rule(chain, args->rulenum, args->rule_id+1); 1163 } else { 1164 /* 1165 * Find the starting rule. It can be either the first 1166 * one, or the one after divert_rule if asked so. 1167 */ 1168 int skipto = mtag ? divert_cookie(mtag) : 0; 1169 1170 f_pos = 0; 1171 if (args->eh == NULL && skipto != 0) { 1172 if (skipto >= IPFW_DEFAULT_RULE) { 1173 IPFW_RUNLOCK(chain); 1174 return (IP_FW_DENY); /* invalid */ 1175 } 1176 f_pos = ipfw_find_rule(chain, skipto+1, 0); 1177 } 1178 } 1179 /* reset divert rule to avoid confusion later */ 1180 if (mtag) { 1181 divinput_flags = divert_info(mtag) & 1182 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 1183 m_tag_delete(m, mtag); 1184 } 1185 1186 /* 1187 * Now scan the rules, and parse microinstructions for each rule. 1188 * We have two nested loops and an inner switch. Sometimes we 1189 * need to break out of one or both loops, or re-enter one of 1190 * the loops with updated variables. Loop variables are: 1191 * 1192 * f_pos (outer loop) points to the current rule. 1193 * On output it points to the matching rule. 1194 * done (outer loop) is used as a flag to break the loop. 1195 * l (inner loop) residual length of current rule. 1196 * cmd points to the current microinstruction. 1197 * 1198 * We break the inner loop by setting l=0 and possibly 1199 * cmdlen=0 if we don't want to advance cmd. 1200 * We break the outer loop by setting done=1 1201 * We can restart the inner loop by setting l>0 and f_pos, f, cmd 1202 * as needed. 1203 */ 1204 for (; f_pos < chain->n_rules; f_pos++) { 1205 ipfw_insn *cmd; 1206 uint32_t tablearg = 0; 1207 int l, cmdlen, skip_or; /* skip rest of OR block */ 1208 struct ip_fw *f; 1209 1210 f = chain->map[f_pos]; 1211 if (V_set_disable & (1 << f->set) ) 1212 continue; 1213 1214 skip_or = 0; 1215 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1216 l -= cmdlen, cmd += cmdlen) { 1217 int match; 1218 1219 /* 1220 * check_body is a jump target used when we find a 1221 * CHECK_STATE, and need to jump to the body of 1222 * the target rule. 1223 */ 1224 1225/* check_body: */ 1226 cmdlen = F_LEN(cmd); 1227 /* 1228 * An OR block (insn_1 || .. || insn_n) has the 1229 * F_OR bit set in all but the last instruction. 1230 * The first match will set "skip_or", and cause 1231 * the following instructions to be skipped until 1232 * past the one with the F_OR bit clear. 1233 */ 1234 if (skip_or) { /* skip this instruction */ 1235 if ((cmd->len & F_OR) == 0) 1236 skip_or = 0; /* next one is good */ 1237 continue; 1238 } 1239 match = 0; /* set to 1 if we succeed */ 1240 1241 switch (cmd->opcode) { 1242 /* 1243 * The first set of opcodes compares the packet's 1244 * fields with some pattern, setting 'match' if a 1245 * match is found. At the end of the loop there is 1246 * logic to deal with F_NOT and F_OR flags associated 1247 * with the opcode. 1248 */ 1249 case O_NOP: 1250 match = 1; 1251 break; 1252 1253 case O_FORWARD_MAC: 1254 printf("ipfw: opcode %d unimplemented\n", 1255 cmd->opcode); 1256 break; 1257 1258 case O_GID: 1259 case O_UID: 1260 case O_JAIL: 1261 /* 1262 * We only check offset == 0 && proto != 0, 1263 * as this ensures that we have a 1264 * packet with the ports info. 1265 */ 1266 if (offset!=0) 1267 break; 1268 if (is_ipv6) /* XXX to be fixed later */ 1269 break; 1270 if (proto == IPPROTO_TCP || 1271 proto == IPPROTO_UDP) 1272 match = check_uidgid( 1273 (ipfw_insn_u32 *)cmd, 1274 proto, oif, 1275 dst_ip, dst_port, 1276 src_ip, src_port, &ucred_cache, 1277 &ucred_lookup, args->inp); 1278 break; 1279 1280 case O_RECV: 1281 match = iface_match(m->m_pkthdr.rcvif, 1282 (ipfw_insn_if *)cmd); 1283 break; 1284 1285 case O_XMIT: 1286 match = iface_match(oif, (ipfw_insn_if *)cmd); 1287 break; 1288 1289 case O_VIA: 1290 match = iface_match(oif ? oif : 1291 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 1292 break; 1293 1294 case O_MACADDR2: 1295 if (args->eh != NULL) { /* have MAC header */ 1296 u_int32_t *want = (u_int32_t *) 1297 ((ipfw_insn_mac *)cmd)->addr; 1298 u_int32_t *mask = (u_int32_t *) 1299 ((ipfw_insn_mac *)cmd)->mask; 1300 u_int32_t *hdr = (u_int32_t *)args->eh; 1301 1302 match = 1303 ( want[0] == (hdr[0] & mask[0]) && 1304 want[1] == (hdr[1] & mask[1]) && 1305 want[2] == (hdr[2] & mask[2]) ); 1306 } 1307 break; 1308 1309 case O_MAC_TYPE: 1310 if (args->eh != NULL) { 1311 u_int16_t *p = 1312 ((ipfw_insn_u16 *)cmd)->ports; 1313 int i; 1314 1315 for (i = cmdlen - 1; !match && i>0; 1316 i--, p += 2) 1317 match = (etype >= p[0] && 1318 etype <= p[1]); 1319 } 1320 break; 1321 1322 case O_FRAG: 1323 match = (offset != 0); 1324 break; 1325 1326 case O_IN: /* "out" is "not in" */ 1327 match = (oif == NULL); 1328 break; 1329 1330 case O_LAYER2: 1331 match = (args->eh != NULL); 1332 break; 1333 1334 case O_DIVERTED: 1335 match = (cmd->arg1 & 1 && divinput_flags & 1336 IP_FW_DIVERT_LOOPBACK_FLAG) || 1337 (cmd->arg1 & 2 && divinput_flags & 1338 IP_FW_DIVERT_OUTPUT_FLAG); 1339 break; 1340 1341 case O_PROTO: 1342 /* 1343 * We do not allow an arg of 0 so the 1344 * check of "proto" only suffices. 1345 */ 1346 match = (proto == cmd->arg1); 1347 break; 1348 1349 case O_IP_SRC: 1350 match = is_ipv4 && 1351 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1352 src_ip.s_addr); 1353 break; 1354 1355 case O_IP_SRC_LOOKUP: 1356 case O_IP_DST_LOOKUP: 1357 if (is_ipv4) { 1358 uint32_t a = 1359 (cmd->opcode == O_IP_DST_LOOKUP) ? 1360 dst_ip.s_addr : src_ip.s_addr; 1361 uint32_t v = 0; 1362 1363 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) { 1364 /* generic lookup */ 1365 v = ((ipfw_insn_u32 *)cmd)->d[1]; 1366 if (v == 0) 1367 a = dst_ip.s_addr; 1368 else if (v == 1) 1369 a = src_ip.s_addr; 1370 else if (offset != 0) 1371 break; 1372 else if (proto != IPPROTO_TCP && 1373 proto != IPPROTO_UDP) 1374 break; 1375 else if (v == 2) 1376 a = dst_port; 1377 else if (v == 3) 1378 a = src_port; 1379 else if (v == 4 || v == 5) { 1380 check_uidgid( 1381 (ipfw_insn_u32 *)cmd, 1382 proto, oif, 1383 dst_ip, dst_port, 1384 src_ip, src_port, &ucred_cache, 1385 &ucred_lookup, args->inp); 1386 if (v == 4 /* O_UID */) 1387 a = ucred_cache->cr_uid; 1388 else if (v == 5 /* O_JAIL */) 1389 a = ucred_cache->cr_prison->pr_id; 1390 } else 1391 break; 1392 } 1393 match = ipfw_lookup_table(chain, cmd->arg1, a, 1394 &v); 1395 if (!match) 1396 break; 1397 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 1398 match = 1399 ((ipfw_insn_u32 *)cmd)->d[0] == v; 1400 else 1401 tablearg = v; 1402 } 1403 break; 1404 1405 case O_IP_SRC_MASK: 1406 case O_IP_DST_MASK: 1407 if (is_ipv4) { 1408 uint32_t a = 1409 (cmd->opcode == O_IP_DST_MASK) ? 1410 dst_ip.s_addr : src_ip.s_addr; 1411 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 1412 int i = cmdlen-1; 1413 1414 for (; !match && i>0; i-= 2, p+= 2) 1415 match = (p[0] == (a & p[1])); 1416 } 1417 break; 1418 1419 case O_IP_SRC_ME: 1420 if (is_ipv4) { 1421 struct ifnet *tif; 1422 1423 INADDR_TO_IFP(src_ip, tif); 1424 match = (tif != NULL); 1425 } 1426 break; 1427 1428 case O_IP_DST_SET: 1429 case O_IP_SRC_SET: 1430 if (is_ipv4) { 1431 u_int32_t *d = (u_int32_t *)(cmd+1); 1432 u_int32_t addr = 1433 cmd->opcode == O_IP_DST_SET ? 1434 args->f_id.dst_ip : 1435 args->f_id.src_ip; 1436 1437 if (addr < d[0]) 1438 break; 1439 addr -= d[0]; /* subtract base */ 1440 match = (addr < cmd->arg1) && 1441 ( d[ 1 + (addr>>5)] & 1442 (1<<(addr & 0x1f)) ); 1443 } 1444 break; 1445 1446 case O_IP_DST: 1447 match = is_ipv4 && 1448 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1449 dst_ip.s_addr); 1450 break; 1451 1452 case O_IP_DST_ME: 1453 if (is_ipv4) { 1454 struct ifnet *tif; 1455 1456 INADDR_TO_IFP(dst_ip, tif); 1457 match = (tif != NULL); 1458 } 1459 break; 1460 1461 case O_IP_SRCPORT: 1462 case O_IP_DSTPORT: 1463 /* 1464 * offset == 0 && proto != 0 is enough 1465 * to guarantee that we have a 1466 * packet with port info. 1467 */ 1468 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 1469 && offset == 0) { 1470 u_int16_t x = 1471 (cmd->opcode == O_IP_SRCPORT) ? 1472 src_port : dst_port ; 1473 u_int16_t *p = 1474 ((ipfw_insn_u16 *)cmd)->ports; 1475 int i; 1476 1477 for (i = cmdlen - 1; !match && i>0; 1478 i--, p += 2) 1479 match = (x>=p[0] && x<=p[1]); 1480 } 1481 break; 1482 1483 case O_ICMPTYPE: 1484 match = (offset == 0 && proto==IPPROTO_ICMP && 1485 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 1486 break; 1487 1488#ifdef INET6 1489 case O_ICMP6TYPE: 1490 match = is_ipv6 && offset == 0 && 1491 proto==IPPROTO_ICMPV6 && 1492 icmp6type_match( 1493 ICMP6(ulp)->icmp6_type, 1494 (ipfw_insn_u32 *)cmd); 1495 break; 1496#endif /* INET6 */ 1497 1498 case O_IPOPT: 1499 match = (is_ipv4 && 1500 ipopts_match(ip, cmd) ); 1501 break; 1502 1503 case O_IPVER: 1504 match = (is_ipv4 && 1505 cmd->arg1 == ip->ip_v); 1506 break; 1507 1508 case O_IPID: 1509 case O_IPLEN: 1510 case O_IPTTL: 1511 if (is_ipv4) { /* only for IP packets */ 1512 uint16_t x; 1513 uint16_t *p; 1514 int i; 1515 1516 if (cmd->opcode == O_IPLEN) 1517 x = ip_len; 1518 else if (cmd->opcode == O_IPTTL) 1519 x = ip->ip_ttl; 1520 else /* must be IPID */ 1521 x = ntohs(ip->ip_id); 1522 if (cmdlen == 1) { 1523 match = (cmd->arg1 == x); 1524 break; 1525 } 1526 /* otherwise we have ranges */ 1527 p = ((ipfw_insn_u16 *)cmd)->ports; 1528 i = cmdlen - 1; 1529 for (; !match && i>0; i--, p += 2) 1530 match = (x >= p[0] && x <= p[1]); 1531 } 1532 break; 1533 1534 case O_IPPRECEDENCE: 1535 match = (is_ipv4 && 1536 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 1537 break; 1538 1539 case O_IPTOS: 1540 match = (is_ipv4 && 1541 flags_match(cmd, ip->ip_tos)); 1542 break; 1543 1544 case O_TCPDATALEN: 1545 if (proto == IPPROTO_TCP && offset == 0) { 1546 struct tcphdr *tcp; 1547 uint16_t x; 1548 uint16_t *p; 1549 int i; 1550 1551 tcp = TCP(ulp); 1552 x = ip_len - 1553 ((ip->ip_hl + tcp->th_off) << 2); 1554 if (cmdlen == 1) { 1555 match = (cmd->arg1 == x); 1556 break; 1557 } 1558 /* otherwise we have ranges */ 1559 p = ((ipfw_insn_u16 *)cmd)->ports; 1560 i = cmdlen - 1; 1561 for (; !match && i>0; i--, p += 2) 1562 match = (x >= p[0] && x <= p[1]); 1563 } 1564 break; 1565 1566 case O_TCPFLAGS: 1567 match = (proto == IPPROTO_TCP && offset == 0 && 1568 flags_match(cmd, TCP(ulp)->th_flags)); 1569 break; 1570 1571 case O_TCPOPTS: 1572 match = (proto == IPPROTO_TCP && offset == 0 && 1573 tcpopts_match(TCP(ulp), cmd)); 1574 break; 1575 1576 case O_TCPSEQ: 1577 match = (proto == IPPROTO_TCP && offset == 0 && 1578 ((ipfw_insn_u32 *)cmd)->d[0] == 1579 TCP(ulp)->th_seq); 1580 break; 1581 1582 case O_TCPACK: 1583 match = (proto == IPPROTO_TCP && offset == 0 && 1584 ((ipfw_insn_u32 *)cmd)->d[0] == 1585 TCP(ulp)->th_ack); 1586 break; 1587 1588 case O_TCPWIN: 1589 match = (proto == IPPROTO_TCP && offset == 0 && 1590 cmd->arg1 == TCP(ulp)->th_win); 1591 break; 1592 1593 case O_ESTAB: 1594 /* reject packets which have SYN only */ 1595 /* XXX should i also check for TH_ACK ? */ 1596 match = (proto == IPPROTO_TCP && offset == 0 && 1597 (TCP(ulp)->th_flags & 1598 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 1599 break; 1600 1601 case O_ALTQ: { 1602 struct pf_mtag *at; 1603 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 1604 1605 match = 1; 1606 at = pf_find_mtag(m); 1607 if (at != NULL && at->qid != 0) 1608 break; 1609 at = pf_get_mtag(m); 1610 if (at == NULL) { 1611 /* 1612 * Let the packet fall back to the 1613 * default ALTQ. 1614 */ 1615 break; 1616 } 1617 at->qid = altq->qid; 1618 if (is_ipv4) 1619 at->af = AF_INET; 1620 else 1621 at->af = AF_LINK; 1622 at->hdr = ip; 1623 break; 1624 } 1625 1626 case O_LOG: 1627 ipfw_log(f, hlen, args, m, 1628 oif, offset, tablearg, ip); 1629 match = 1; 1630 break; 1631 1632 case O_PROB: 1633 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 1634 break; 1635 1636 case O_VERREVPATH: 1637 /* Outgoing packets automatically pass/match */ 1638 match = ((oif != NULL) || 1639 (m->m_pkthdr.rcvif == NULL) || 1640 ( 1641#ifdef INET6 1642 is_ipv6 ? 1643 verify_path6(&(args->f_id.src_ip6), 1644 m->m_pkthdr.rcvif) : 1645#endif 1646 verify_path(src_ip, m->m_pkthdr.rcvif, 1647 args->f_id.fib))); 1648 break; 1649 1650 case O_VERSRCREACH: 1651 /* Outgoing packets automatically pass/match */ 1652 match = (hlen > 0 && ((oif != NULL) || 1653#ifdef INET6 1654 is_ipv6 ? 1655 verify_path6(&(args->f_id.src_ip6), 1656 NULL) : 1657#endif 1658 verify_path(src_ip, NULL, args->f_id.fib))); 1659 break; 1660 1661 case O_ANTISPOOF: 1662 /* Outgoing packets automatically pass/match */ 1663 if (oif == NULL && hlen > 0 && 1664 ( (is_ipv4 && in_localaddr(src_ip)) 1665#ifdef INET6 1666 || (is_ipv6 && 1667 in6_localaddr(&(args->f_id.src_ip6))) 1668#endif 1669 )) 1670 match = 1671#ifdef INET6 1672 is_ipv6 ? verify_path6( 1673 &(args->f_id.src_ip6), 1674 m->m_pkthdr.rcvif) : 1675#endif 1676 verify_path(src_ip, 1677 m->m_pkthdr.rcvif, 1678 args->f_id.fib); 1679 else 1680 match = 1; 1681 break; 1682 1683 case O_IPSEC: 1684#ifdef IPSEC 1685 match = (m_tag_find(m, 1686 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 1687#endif 1688 /* otherwise no match */ 1689 break; 1690 1691#ifdef INET6 1692 case O_IP6_SRC: 1693 match = is_ipv6 && 1694 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 1695 &((ipfw_insn_ip6 *)cmd)->addr6); 1696 break; 1697 1698 case O_IP6_DST: 1699 match = is_ipv6 && 1700 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 1701 &((ipfw_insn_ip6 *)cmd)->addr6); 1702 break; 1703 case O_IP6_SRC_MASK: 1704 case O_IP6_DST_MASK: 1705 if (is_ipv6) { 1706 int i = cmdlen - 1; 1707 struct in6_addr p; 1708 struct in6_addr *d = 1709 &((ipfw_insn_ip6 *)cmd)->addr6; 1710 1711 for (; !match && i > 0; d += 2, 1712 i -= F_INSN_SIZE(struct in6_addr) 1713 * 2) { 1714 p = (cmd->opcode == 1715 O_IP6_SRC_MASK) ? 1716 args->f_id.src_ip6: 1717 args->f_id.dst_ip6; 1718 APPLY_MASK(&p, &d[1]); 1719 match = 1720 IN6_ARE_ADDR_EQUAL(&d[0], 1721 &p); 1722 } 1723 } 1724 break; 1725 1726 case O_IP6_SRC_ME: 1727 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 1728 break; 1729 1730 case O_IP6_DST_ME: 1731 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 1732 break; 1733 1734 case O_FLOW6ID: 1735 match = is_ipv6 && 1736 flow6id_match(args->f_id.flow_id6, 1737 (ipfw_insn_u32 *) cmd); 1738 break; 1739 1740 case O_EXT_HDR: 1741 match = is_ipv6 && 1742 (ext_hd & ((ipfw_insn *) cmd)->arg1); 1743 break; 1744 1745 case O_IP6: 1746 match = is_ipv6; 1747 break; 1748#endif 1749 1750 case O_IP4: 1751 match = is_ipv4; 1752 break; 1753 1754 case O_TAG: { 1755 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 1756 tablearg : cmd->arg1; 1757 1758 /* Packet is already tagged with this tag? */ 1759 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 1760 1761 /* We have `untag' action when F_NOT flag is 1762 * present. And we must remove this mtag from 1763 * mbuf and reset `match' to zero (`match' will 1764 * be inversed later). 1765 * Otherwise we should allocate new mtag and 1766 * push it into mbuf. 1767 */ 1768 if (cmd->len & F_NOT) { /* `untag' action */ 1769 if (mtag != NULL) 1770 m_tag_delete(m, mtag); 1771 } else if (mtag == NULL) { 1772 if ((mtag = m_tag_alloc(MTAG_IPFW, 1773 tag, 0, M_NOWAIT)) != NULL) 1774 m_tag_prepend(m, mtag); 1775 } 1776 match = (cmd->len & F_NOT) ? 0: 1; 1777 break; 1778 } 1779 1780 case O_FIB: /* try match the specified fib */ 1781 if (args->f_id.fib == cmd->arg1) 1782 match = 1; 1783 break; 1784 1785 case O_TAGGED: { 1786 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 1787 tablearg : cmd->arg1; 1788 1789 if (cmdlen == 1) { 1790 match = m_tag_locate(m, MTAG_IPFW, 1791 tag, NULL) != NULL; 1792 break; 1793 } 1794 1795 /* we have ranges */ 1796 for (mtag = m_tag_first(m); 1797 mtag != NULL && !match; 1798 mtag = m_tag_next(m, mtag)) { 1799 uint16_t *p; 1800 int i; 1801 1802 if (mtag->m_tag_cookie != MTAG_IPFW) 1803 continue; 1804 1805 p = ((ipfw_insn_u16 *)cmd)->ports; 1806 i = cmdlen - 1; 1807 for(; !match && i > 0; i--, p += 2) 1808 match = 1809 mtag->m_tag_id >= p[0] && 1810 mtag->m_tag_id <= p[1]; 1811 } 1812 break; 1813 } 1814 1815 /* 1816 * The second set of opcodes represents 'actions', 1817 * i.e. the terminal part of a rule once the packet 1818 * matches all previous patterns. 1819 * Typically there is only one action for each rule, 1820 * and the opcode is stored at the end of the rule 1821 * (but there are exceptions -- see below). 1822 * 1823 * In general, here we set retval and terminate the 1824 * outer loop (would be a 'break 3' in some language, 1825 * but we need to set l=0, done=1) 1826 * 1827 * Exceptions: 1828 * O_COUNT and O_SKIPTO actions: 1829 * instead of terminating, we jump to the next rule 1830 * (setting l=0), or to the SKIPTO target (setting 1831 * f/f_len, cmd and l as needed), respectively. 1832 * 1833 * O_TAG, O_LOG and O_ALTQ action parameters: 1834 * perform some action and set match = 1; 1835 * 1836 * O_LIMIT and O_KEEP_STATE: these opcodes are 1837 * not real 'actions', and are stored right 1838 * before the 'action' part of the rule. 1839 * These opcodes try to install an entry in the 1840 * state tables; if successful, we continue with 1841 * the next opcode (match=1; break;), otherwise 1842 * the packet must be dropped (set retval, 1843 * break loops with l=0, done=1) 1844 * 1845 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 1846 * cause a lookup of the state table, and a jump 1847 * to the 'action' part of the parent rule 1848 * if an entry is found, or 1849 * (CHECK_STATE only) a jump to the next rule if 1850 * the entry is not found. 1851 * The result of the lookup is cached so that 1852 * further instances of these opcodes become NOPs. 1853 * The jump to the next rule is done by setting 1854 * l=0, cmdlen=0. 1855 */ 1856 case O_LIMIT: 1857 case O_KEEP_STATE: 1858 if (ipfw_install_state(f, 1859 (ipfw_insn_limit *)cmd, args, tablearg)) { 1860 /* error or limit violation */ 1861 retval = IP_FW_DENY; 1862 l = 0; /* exit inner loop */ 1863 done = 1; /* exit outer loop */ 1864 } 1865 match = 1; 1866 break; 1867 1868 case O_PROBE_STATE: 1869 case O_CHECK_STATE: 1870 /* 1871 * dynamic rules are checked at the first 1872 * keep-state or check-state occurrence, 1873 * with the result being stored in dyn_dir. 1874 * The compiler introduces a PROBE_STATE 1875 * instruction for us when we have a 1876 * KEEP_STATE (because PROBE_STATE needs 1877 * to be run first). 1878 */ 1879 if (dyn_dir == MATCH_UNKNOWN && 1880 (q = ipfw_lookup_dyn_rule(&args->f_id, 1881 &dyn_dir, proto == IPPROTO_TCP ? 1882 TCP(ulp) : NULL)) 1883 != NULL) { 1884 /* 1885 * Found dynamic entry, update stats 1886 * and jump to the 'action' part of 1887 * the parent rule by setting 1888 * f, cmd, l and clearing cmdlen. 1889 */ 1890 q->pcnt++; 1891 q->bcnt += pktlen; 1892 /* XXX we would like to have f_pos 1893 * readily accessible in the dynamic 1894 * rule, instead of having to 1895 * lookup q->rule. 1896 */ 1897 f = q->rule; 1898 f_pos = ipfw_find_rule(chain, 1899 f->rulenum, f->id); 1900 cmd = ACTION_PTR(f); 1901 l = f->cmd_len - f->act_ofs; 1902 ipfw_dyn_unlock(); 1903 cmdlen = 0; 1904 match = 1; 1905 break; 1906 } 1907 /* 1908 * Dynamic entry not found. If CHECK_STATE, 1909 * skip to next rule, if PROBE_STATE just 1910 * ignore and continue with next opcode. 1911 */ 1912 if (cmd->opcode == O_CHECK_STATE) 1913 l = 0; /* exit inner loop */ 1914 match = 1; 1915 break; 1916 1917 case O_ACCEPT: 1918 retval = 0; /* accept */ 1919 l = 0; /* exit inner loop */ 1920 done = 1; /* exit outer loop */ 1921 break; 1922 1923 case O_PIPE: 1924 case O_QUEUE: 1925 set_match(args, f_pos, chain); 1926 args->cookie = (cmd->arg1 == IP_FW_TABLEARG) ? 1927 tablearg : cmd->arg1; 1928 if (cmd->opcode == O_QUEUE) 1929 args->cookie |= 0x80000000; 1930 retval = IP_FW_DUMMYNET; 1931 l = 0; /* exit inner loop */ 1932 done = 1; /* exit outer loop */ 1933 break; 1934 1935 case O_DIVERT: 1936 case O_TEE: 1937 if (args->eh) /* not on layer 2 */ 1938 break; 1939 /* otherwise this is terminal */ 1940 l = 0; /* exit inner loop */ 1941 done = 1; /* exit outer loop */ 1942 mtag = m_tag_get(PACKET_TAG_DIVERT, 1943 sizeof(struct divert_tag), 1944 M_NOWAIT); 1945 if (mtag == NULL) { 1946 retval = IP_FW_DENY; 1947 } else { 1948 struct divert_tag *dt; 1949 dt = (struct divert_tag *)(mtag+1); 1950 dt->cookie = f->rulenum; 1951 if (cmd->arg1 == IP_FW_TABLEARG) 1952 dt->info = tablearg; 1953 else 1954 dt->info = cmd->arg1; 1955 m_tag_prepend(m, mtag); 1956 retval = (cmd->opcode == O_DIVERT) ? 1957 IP_FW_DIVERT : IP_FW_TEE; 1958 } 1959 break; 1960 1961 case O_COUNT: 1962 f->pcnt++; /* update stats */ 1963 f->bcnt += pktlen; 1964 f->timestamp = time_uptime; 1965 l = 0; /* exit inner loop */ 1966 break; 1967 1968 case O_SKIPTO: 1969 f->pcnt++; /* update stats */ 1970 f->bcnt += pktlen; 1971 f->timestamp = time_uptime; 1972 /* If possible use cached f_pos (in f->next_rule), 1973 * whose version is written in f->next_rule 1974 * (horrible hacks to avoid changing the ABI). 1975 */ 1976 if (cmd->arg1 != IP_FW_TABLEARG && 1977 (uintptr_t)f->x_next == chain->id) { 1978 f_pos = (uintptr_t)f->next_rule; 1979 } else { 1980 int i = (cmd->arg1 == IP_FW_TABLEARG) ? 1981 tablearg : cmd->arg1; 1982 /* make sure we do not jump backward */ 1983 if (i <= f->rulenum) 1984 i = f->rulenum + 1; 1985 f_pos = ipfw_find_rule(chain, i, 0); 1986 /* update the cache */ 1987 if (cmd->arg1 != IP_FW_TABLEARG) { 1988 f->next_rule = 1989 (void *)(uintptr_t)f_pos; 1990 f->x_next = 1991 (void *)(uintptr_t)chain->id; 1992 } 1993 } 1994 /* 1995 * Skip disabled rules, and re-enter 1996 * the inner loop with the correct 1997 * f_pos, f, l and cmd. 1998 * Also clear cmdlen and skip_or 1999 */ 2000 for (; f_pos < chain->n_rules - 1 && 2001 (V_set_disable & 2002 (1 << chain->map[f_pos]->set)); 2003 f_pos++) 2004 ; 2005 /* prepare to enter the inner loop */ 2006 f = chain->map[f_pos]; 2007 l = f->cmd_len; 2008 cmd = f->cmd; 2009 match = 1; 2010 cmdlen = 0; 2011 skip_or = 0; 2012 break; 2013 2014 case O_REJECT: 2015 /* 2016 * Drop the packet and send a reject notice 2017 * if the packet is not ICMP (or is an ICMP 2018 * query), and it is not multicast/broadcast. 2019 */ 2020 if (hlen > 0 && is_ipv4 && offset == 0 && 2021 (proto != IPPROTO_ICMP || 2022 is_icmp_query(ICMP(ulp))) && 2023 !(m->m_flags & (M_BCAST|M_MCAST)) && 2024 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 2025 send_reject(args, cmd->arg1, ip_len, ip); 2026 m = args->m; 2027 } 2028 /* FALLTHROUGH */ 2029#ifdef INET6 2030 case O_UNREACH6: 2031 if (hlen > 0 && is_ipv6 && 2032 ((offset & IP6F_OFF_MASK) == 0) && 2033 (proto != IPPROTO_ICMPV6 || 2034 (is_icmp6_query(args->f_id.flags) == 1)) && 2035 !(m->m_flags & (M_BCAST|M_MCAST)) && 2036 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 2037 send_reject6( 2038 args, cmd->arg1, hlen, 2039 (struct ip6_hdr *)ip); 2040 m = args->m; 2041 } 2042 /* FALLTHROUGH */ 2043#endif 2044 case O_DENY: 2045 retval = IP_FW_DENY; 2046 l = 0; /* exit inner loop */ 2047 done = 1; /* exit outer loop */ 2048 break; 2049 2050 case O_FORWARD_IP: 2051 if (args->eh) /* not valid on layer2 pkts */ 2052 break; 2053 if (!q || dyn_dir == MATCH_FORWARD) { 2054 struct sockaddr_in *sa; 2055 sa = &(((ipfw_insn_sa *)cmd)->sa); 2056 if (sa->sin_addr.s_addr == INADDR_ANY) { 2057 bcopy(sa, &args->hopstore, 2058 sizeof(*sa)); 2059 args->hopstore.sin_addr.s_addr = 2060 htonl(tablearg); 2061 args->next_hop = &args->hopstore; 2062 } else { 2063 args->next_hop = sa; 2064 } 2065 } 2066 retval = IP_FW_PASS; 2067 l = 0; /* exit inner loop */ 2068 done = 1; /* exit outer loop */ 2069 break; 2070 2071 case O_NETGRAPH: 2072 case O_NGTEE: 2073 set_match(args, f_pos, chain); 2074 args->cookie = (cmd->arg1 == IP_FW_TABLEARG) ? 2075 tablearg : cmd->arg1; 2076 retval = (cmd->opcode == O_NETGRAPH) ? 2077 IP_FW_NETGRAPH : IP_FW_NGTEE; 2078 l = 0; /* exit inner loop */ 2079 done = 1; /* exit outer loop */ 2080 break; 2081 2082 case O_SETFIB: 2083 f->pcnt++; /* update stats */ 2084 f->bcnt += pktlen; 2085 f->timestamp = time_uptime; 2086 M_SETFIB(m, cmd->arg1); 2087 args->f_id.fib = cmd->arg1; 2088 l = 0; /* exit inner loop */ 2089 break; 2090 2091 case O_NAT: 2092 if (!IPFW_NAT_LOADED) { 2093 retval = IP_FW_DENY; 2094 } else { 2095 struct cfg_nat *t; 2096 int nat_id; 2097 2098 set_match(args, f_pos, chain); 2099 t = ((ipfw_insn_nat *)cmd)->nat; 2100 if (t == NULL) { 2101 nat_id = (cmd->arg1 == IP_FW_TABLEARG) ? 2102 tablearg : cmd->arg1; 2103 t = (*lookup_nat_ptr)(&chain->nat, nat_id); 2104 2105 if (t == NULL) { 2106 retval = IP_FW_DENY; 2107 l = 0; /* exit inner loop */ 2108 done = 1; /* exit outer loop */ 2109 break; 2110 } 2111 if (cmd->arg1 != IP_FW_TABLEARG) 2112 ((ipfw_insn_nat *)cmd)->nat = t; 2113 } 2114 retval = ipfw_nat_ptr(args, t, m); 2115 } 2116 l = 0; /* exit inner loop */ 2117 done = 1; /* exit outer loop */ 2118 break; 2119 2120 case O_REASS: { 2121 int ip_off; 2122 2123 f->pcnt++; 2124 f->bcnt += pktlen; 2125 l = 0; /* in any case exit inner loop */ 2126 2127 ip_off = (args->eh != NULL) ? 2128 ntohs(ip->ip_off) : ip->ip_off; 2129 /* if not fragmented, go to next rule */ 2130 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0) 2131 break; 2132 /* 2133 * ip_reass() expects len & off in host 2134 * byte order: fix them in case we come 2135 * from layer2. 2136 */ 2137 if (args->eh != NULL) { 2138 ip->ip_len = ntohs(ip->ip_len); 2139 ip->ip_off = ntohs(ip->ip_off); 2140 } 2141 2142 args->m = m = ip_reass(m); 2143 2144 /* 2145 * IP header checksum fixup after 2146 * reassembly and leave header 2147 * in network byte order. 2148 */ 2149 if (m == NULL) { /* fragment got swallowed */ 2150 retval = IP_FW_DENY; 2151 } else { /* good, packet complete */ 2152 int hlen; 2153 2154 ip = mtod(m, struct ip *); 2155 hlen = ip->ip_hl << 2; 2156 /* revert len & off for layer2 pkts */ 2157 if (args->eh != NULL) 2158 ip->ip_len = htons(ip->ip_len); 2159 ip->ip_sum = 0; 2160 if (hlen == sizeof(struct ip)) 2161 ip->ip_sum = in_cksum_hdr(ip); 2162 else 2163 ip->ip_sum = in_cksum(m, hlen); 2164 retval = IP_FW_REASS; 2165 set_match(args, f_pos, chain); 2166 } 2167 done = 1; /* exit outer loop */ 2168 break; 2169 } 2170 2171 default: 2172 panic("-- unknown opcode %d\n", cmd->opcode); 2173 } /* end of switch() on opcodes */ 2174 /* 2175 * if we get here with l=0, then match is irrelevant. 2176 */ 2177 2178 if (cmd->len & F_NOT) 2179 match = !match; 2180 2181 if (match) { 2182 if (cmd->len & F_OR) 2183 skip_or = 1; 2184 } else { 2185 if (!(cmd->len & F_OR)) /* not an OR block, */ 2186 break; /* try next rule */ 2187 } 2188 2189 } /* end of inner loop, scan opcodes */ 2190 2191 if (done) 2192 break; 2193 2194/* next_rule:; */ /* try next rule */ 2195 2196 } /* end of outer for, scan rules */ 2197 2198 if (done) { 2199 struct ip_fw *rule = chain->map[f_pos]; 2200 /* Update statistics */ 2201 rule->pcnt++; 2202 rule->bcnt += pktlen; 2203 rule->timestamp = time_uptime; 2204 } else { 2205 retval = IP_FW_DENY; 2206 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 2207 } 2208 IPFW_RUNLOCK(chain); 2209 if (ucred_cache != NULL) 2210 crfree(ucred_cache); 2211 return (retval); 2212 2213pullup_failed: 2214 if (V_fw_verbose) 2215 printf("ipfw: pullup failed\n"); 2216 return (IP_FW_DENY); 2217} 2218 2219/* 2220 * Module and VNET glue 2221 */ 2222 2223/* 2224 * Stuff that must be initialised only on boot or module load 2225 */ 2226static int 2227ipfw_init(void) 2228{ 2229 int error = 0; 2230 2231 ipfw_dyn_attach(); 2232 /* 2233 * Only print out this stuff the first time around, 2234 * when called from the sysinit code. 2235 */ 2236 printf("ipfw2 " 2237#ifdef INET6 2238 "(+ipv6) " 2239#endif 2240 "initialized, divert %s, nat %s, " 2241 "rule-based forwarding " 2242#ifdef IPFIREWALL_FORWARD 2243 "enabled, " 2244#else 2245 "disabled, " 2246#endif 2247 "default to %s, logging ", 2248#ifdef IPDIVERT 2249 "enabled", 2250#else 2251 "loadable", 2252#endif 2253#ifdef IPFIREWALL_NAT 2254 "enabled", 2255#else 2256 "loadable", 2257#endif 2258 default_to_accept ? "accept" : "deny"); 2259 2260 /* 2261 * Note: V_xxx variables can be accessed here but the vnet specific 2262 * initializer may not have been called yet for the VIMAGE case. 2263 * Tuneables will have been processed. We will print out values for 2264 * the default vnet. 2265 * XXX This should all be rationalized AFTER 8.0 2266 */ 2267 if (V_fw_verbose == 0) 2268 printf("disabled\n"); 2269 else if (V_verbose_limit == 0) 2270 printf("unlimited\n"); 2271 else 2272 printf("limited to %d packets/entry by default\n", 2273 V_verbose_limit); 2274 2275 ipfw_log_bpf(1); /* init */ 2276 return (error); 2277} 2278 2279/* 2280 * Called for the removal of the last instance only on module unload. 2281 */ 2282static void 2283ipfw_destroy(void) 2284{ 2285 2286 ipfw_log_bpf(0); /* uninit */ 2287 ipfw_dyn_detach(); 2288 printf("IP firewall unloaded\n"); 2289} 2290 2291/* 2292 * Stuff that must be initialized for every instance 2293 * (including the first of course). 2294 */ 2295static int 2296vnet_ipfw_init(const void *unused) 2297{ 2298 int error; 2299 struct ip_fw *rule = NULL; 2300 struct ip_fw_chain *chain; 2301 2302 chain = &V_layer3_chain; 2303 2304 /* First set up some values that are compile time options */ 2305 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 2306 V_fw_deny_unknown_exthdrs = 1; 2307#ifdef IPFIREWALL_VERBOSE 2308 V_fw_verbose = 1; 2309#endif 2310#ifdef IPFIREWALL_VERBOSE_LIMIT 2311 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 2312#endif 2313#ifdef IPFIREWALL_NAT 2314 LIST_INIT(&chain->nat); 2315#endif 2316 2317 /* insert the default rule and create the initial map */ 2318 chain->n_rules = 1; 2319 chain->static_len = sizeof(struct ip_fw); 2320 chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_NOWAIT | M_ZERO); 2321 if (chain->map) 2322 rule = malloc(chain->static_len, M_IPFW, M_NOWAIT | M_ZERO); 2323 if (rule == NULL) { 2324 if (chain->map) 2325 free(chain->map, M_IPFW); 2326 printf("ipfw2: ENOSPC initializing default rule " 2327 "(support disabled)\n"); 2328 return (ENOSPC); 2329 } 2330 error = ipfw_init_tables(chain); 2331 if (error) { 2332 panic("init_tables"); /* XXX Marko fix this ! */ 2333 } 2334 2335 /* fill and insert the default rule */ 2336 rule->act_ofs = 0; 2337 rule->rulenum = IPFW_DEFAULT_RULE; 2338 rule->cmd_len = 1; 2339 rule->set = RESVD_SET; 2340 rule->cmd[0].len = 1; 2341 rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; 2342 chain->rules = chain->default_rule = chain->map[0] = rule; 2343 chain->id = rule->id = 1; 2344 2345 IPFW_LOCK_INIT(chain); 2346 ipfw_dyn_init(); 2347 2348 /* First set up some values that are compile time options */ 2349 V_ipfw_vnet_ready = 1; /* Open for business */ 2350 2351 /* 2352 * Hook the sockopt handler, and the layer2 (V_ip_fw_chk_ptr) 2353 * and pfil hooks for ipv4 and ipv6. Even if the latter two fail 2354 * we still keep the module alive because the sockopt and 2355 * layer2 paths are still useful. 2356 * ipfw[6]_hook return 0 on success, ENOENT on failure, 2357 * so we can ignore the exact return value and just set a flag. 2358 * 2359 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so 2360 * changes in the underlying (per-vnet) variables trigger 2361 * immediate hook()/unhook() calls. 2362 * In layer2 we have the same behaviour, except that V_ether_ipfw 2363 * is checked on each packet because there are no pfil hooks. 2364 */ 2365 V_ip_fw_ctl_ptr = ipfw_ctl; 2366 V_ip_fw_chk_ptr = ipfw_chk; 2367 error = ipfw_attach_hooks(); 2368 return (error); 2369} 2370 2371/* 2372 * Called for the removal of each instance. 2373 */ 2374static int 2375vnet_ipfw_uninit(const void *unused) 2376{ 2377 struct ip_fw *reap, *rule; 2378 struct ip_fw_chain *chain = &V_layer3_chain; 2379 int i; 2380 2381 V_ipfw_vnet_ready = 0; /* tell new callers to go away */ 2382 /* 2383 * disconnect from ipv4, ipv6, layer2 and sockopt. 2384 * Then grab, release and grab again the WLOCK so we make 2385 * sure the update is propagated and nobody will be in. 2386 */ 2387 ipfw_unhook(); 2388#ifdef INET6 2389 ipfw6_unhook(); 2390#endif 2391 V_ip_fw_chk_ptr = NULL; 2392 V_ip_fw_ctl_ptr = NULL; 2393 IPFW_UH_WLOCK(chain); 2394 IPFW_UH_WUNLOCK(chain); 2395 IPFW_UH_WLOCK(chain); 2396 2397 IPFW_WLOCK(chain); 2398 IPFW_WUNLOCK(chain); 2399 IPFW_WLOCK(chain); 2400 2401 ipfw_dyn_uninit(0); /* run the callout_drain */ 2402 ipfw_flush_tables(chain); 2403 reap = NULL; 2404 for (i = 0; i < chain->n_rules; i++) { 2405 rule = chain->map[i]; 2406 rule->x_next = reap; 2407 reap = rule; 2408 } 2409 if (chain->map) 2410 free(chain->map, M_IPFW); 2411 IPFW_WUNLOCK(chain); 2412 IPFW_UH_WUNLOCK(chain); 2413 if (reap != NULL) 2414 ipfw_reap_rules(reap); 2415 IPFW_LOCK_DESTROY(chain); 2416 ipfw_dyn_uninit(1); /* free the remaining parts */ 2417 return 0; 2418} 2419 2420/* 2421 * Module event handler. 2422 * In general we have the choice of handling most of these events by the 2423 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to 2424 * use the SYSINIT handlers as they are more capable of expressing the 2425 * flow of control during module and vnet operations, so this is just 2426 * a skeleton. Note there is no SYSINIT equivalent of the module 2427 * SHUTDOWN handler, but we don't have anything to do in that case anyhow. 2428 */ 2429static int 2430ipfw_modevent(module_t mod, int type, void *unused) 2431{ 2432 int err = 0; 2433 2434 switch (type) { 2435 case MOD_LOAD: 2436 /* Called once at module load or 2437 * system boot if compiled in. */ 2438 break; 2439 case MOD_QUIESCE: 2440 /* Called before unload. May veto unloading. */ 2441 break; 2442 case MOD_UNLOAD: 2443 /* Called during unload. */ 2444 break; 2445 case MOD_SHUTDOWN: 2446 /* Called during system shutdown. */ 2447 break; 2448 default: 2449 err = EOPNOTSUPP; 2450 break; 2451 } 2452 return err; 2453} 2454 2455static moduledata_t ipfwmod = { 2456 "ipfw", 2457 ipfw_modevent, 2458 0 2459}; 2460 2461/* Define startup order. */ 2462#define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_IFATTACHDOMAIN 2463#define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */ 2464#define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */ 2465#define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */ 2466 2467DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER); 2468MODULE_VERSION(ipfw, 2); 2469/* should declare some dependencies here */ 2470 2471/* 2472 * Starting up. Done in order after ipfwmod() has been called. 2473 * VNET_SYSINIT is also called for each existing vnet and each new vnet. 2474 */ 2475SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 2476 ipfw_init, NULL); 2477VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 2478 vnet_ipfw_init, NULL); 2479 2480/* 2481 * Closing up shop. These are done in REVERSE ORDER, but still 2482 * after ipfwmod() has been called. Not called on reboot. 2483 * VNET_SYSUNINIT is also called for each exiting vnet as it exits. 2484 * or when the module is unloaded. 2485 */ 2486SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 2487 ipfw_destroy, NULL); 2488VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 2489 vnet_ipfw_uninit, NULL); 2490/* end of file */ 2491