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("$FreeBSD$"); 28 29 /* 30 * The FreeBSD IP packet firewall, main file 31 */ 32 33 #include "opt_ipfw.h" 34 #include "opt_ipdivert.h" 35 #include "opt_inet.h" 36 #ifndef INET 37 #error "IPFIREWALL requires INET" 38 #endif /* INET */ 39 #include "opt_inet6.h" 40 #include "opt_ipsec.h" 41 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/condvar.h> 45 #include <sys/counter.h> 46 #include <sys/eventhandler.h> 47 #include <sys/malloc.h> 48 #include <sys/mbuf.h> 49 #include <sys/kernel.h> 50 #include <sys/lock.h> 51 #include <sys/jail.h> 52 #include <sys/module.h> 53 #include <sys/priv.h> 54 #include <sys/proc.h> 55 #include <sys/rwlock.h> 56 #include <sys/rmlock.h> 57 #include <sys/socket.h> 58 #include <sys/socketvar.h> 59 #include <sys/sysctl.h> 60 #include <sys/syslog.h> 61 #include <sys/ucred.h> 62 #include <net/ethernet.h> /* for ETHERTYPE_IP */ 63 #include <net/if.h> 64 #include <net/if_var.h> 65 #include <net/route.h> 66 #include <net/pfil.h> 67 #include <net/vnet.h> 68 69 #include <netpfil/pf/pf_mtag.h> 70 71 #include <netinet/in.h> 72 #include <netinet/in_var.h> 73 #include <netinet/in_pcb.h> 74 #include <netinet/ip.h> 75 #include <netinet/ip_var.h> 76 #include <netinet/ip_icmp.h> 77 #include <netinet/ip_fw.h> 78 #include <netinet/ip_carp.h> 79 #include <netinet/pim.h> 80 #include <netinet/tcp_var.h> 81 #include <netinet/udp.h> 82 #include <netinet/udp_var.h> 83 #include <netinet/sctp.h> 84 85 #include <netinet/ip6.h> 86 #include <netinet/icmp6.h> 87 #ifdef INET6 88 #include <netinet6/in6_pcb.h> 89 #include <netinet6/scope6_var.h> 90 #include <netinet6/ip6_var.h> 91 #endif 92 93 #include <netpfil/ipfw/ip_fw_private.h> 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 static VNET_DEFINE(int, fw_deny_unknown_exthdrs); 107 #define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs) 108 109 static VNET_DEFINE(int, fw_permit_single_frag6) = 1; 110 #define V_fw_permit_single_frag6 VNET(fw_permit_single_frag6) 111 112 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 113 static int default_to_accept = 1; 114 #else 115 static int default_to_accept; 116 #endif 117 118 VNET_DEFINE(int, autoinc_step); 119 VNET_DEFINE(int, fw_one_pass) = 1; 120 121 VNET_DEFINE(unsigned int, fw_tables_max); 122 VNET_DEFINE(unsigned int, fw_tables_sets) = 0; /* Don't use set-aware tables */ 123 /* Use 128 tables by default */ 124 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT; 125 126 #ifndef LINEAR_SKIPTO 127 static int jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num, 128 int tablearg, int jump_backwards); 129 #define JUMP(ch, f, num, targ, back) jump_fast(ch, f, num, targ, back) 130 #else 131 static int jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num, 132 int tablearg, int jump_backwards); 133 #define JUMP(ch, f, num, targ, back) jump_linear(ch, f, num, targ, back) 134 #endif 135 136 /* 137 * Each rule belongs to one of 32 different sets (0..31). 138 * The variable set_disable contains one bit per set. 139 * If the bit is set, all rules in the corresponding set 140 * are disabled. Set RESVD_SET(31) is reserved for the default rule 141 * and rules that are not deleted by the flush command, 142 * and CANNOT be disabled. 143 * Rules in set RESVD_SET can only be deleted individually. 144 */ 145 VNET_DEFINE(u_int32_t, set_disable); 146 #define V_set_disable VNET(set_disable) 147 148 VNET_DEFINE(int, fw_verbose); 149 /* counter for ipfw_log(NULL...) */ 150 VNET_DEFINE(u_int64_t, norule_counter); 151 VNET_DEFINE(int, verbose_limit); 152 153 /* layer3_chain contains the list of rules for layer 3 */ 154 VNET_DEFINE(struct ip_fw_chain, layer3_chain); 155 156 /* ipfw_vnet_ready controls when we are open for business */ 157 VNET_DEFINE(int, ipfw_vnet_ready) = 0; 158 159 VNET_DEFINE(int, ipfw_nat_ready) = 0; 160 161 ipfw_nat_t *ipfw_nat_ptr = NULL; 162 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int); 163 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; 164 ipfw_nat_cfg_t *ipfw_nat_del_ptr; 165 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; 166 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; 167 168 #ifdef SYSCTL_NODE 169 uint32_t dummy_def = IPFW_DEFAULT_RULE; 170 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS); 171 static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS); 172 173 SYSBEGIN(f3) 174 175 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 176 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 177 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0, 178 "Only do a single pass through ipfw when using dummynet(4)"); 179 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, 180 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0, 181 "Rule number auto-increment step"); 182 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 183 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0, 184 "Log matches to ipfw rules"); 185 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, 186 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(verbose_limit), 0, 187 "Set upper limit of matches of ipfw rules logged"); 188 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, 189 &dummy_def, 0, 190 "The default/max possible rule number."); 191 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_max, 192 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU", 193 "Maximum number of concurrently used tables"); 194 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets, 195 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW, 196 0, 0, sysctl_ipfw_tables_sets, "IU", 197 "Use per-set namespace for tables"); 198 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN, 199 &default_to_accept, 0, 200 "Make the default rule accept all packets."); 201 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables); 202 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, 203 CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0, 204 "Number of static rules"); 205 206 #ifdef INET6 207 SYSCTL_DECL(_net_inet6_ip6); 208 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 209 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs, 210 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, 211 &VNET_NAME(fw_deny_unknown_exthdrs), 0, 212 "Deny packets with unknown IPv6 Extension Headers"); 213 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6, 214 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, 215 &VNET_NAME(fw_permit_single_frag6), 0, 216 "Permit single packet IPv6 fragments"); 217 #endif /* INET6 */ 218 219 SYSEND 220 221 #endif /* SYSCTL_NODE */ 222 223 224 /* 225 * Some macros used in the various matching options. 226 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 227 * Other macros just cast void * into the appropriate type 228 */ 229 #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 230 #define TCP(p) ((struct tcphdr *)(p)) 231 #define SCTP(p) ((struct sctphdr *)(p)) 232 #define UDP(p) ((struct udphdr *)(p)) 233 #define ICMP(p) ((struct icmphdr *)(p)) 234 #define ICMP6(p) ((struct icmp6_hdr *)(p)) 235 236 static __inline int 237 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 238 { 239 int type = icmp->icmp_type; 240 241 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 242 } 243 244 #define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 245 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 246 247 static int 248 is_icmp_query(struct icmphdr *icmp) 249 { 250 int type = icmp->icmp_type; 251 252 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 253 } 254 #undef TT 255 256 /* 257 * The following checks use two arrays of 8 or 16 bits to store the 258 * bits that we want set or clear, respectively. They are in the 259 * low and high half of cmd->arg1 or cmd->d[0]. 260 * 261 * We scan options and store the bits we find set. We succeed if 262 * 263 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 264 * 265 * The code is sometimes optimized not to store additional variables. 266 */ 267 268 static int 269 flags_match(ipfw_insn *cmd, u_int8_t bits) 270 { 271 u_char want_clear; 272 bits = ~bits; 273 274 if ( ((cmd->arg1 & 0xff) & bits) != 0) 275 return 0; /* some bits we want set were clear */ 276 want_clear = (cmd->arg1 >> 8) & 0xff; 277 if ( (want_clear & bits) != want_clear) 278 return 0; /* some bits we want clear were set */ 279 return 1; 280 } 281 282 static int 283 ipopts_match(struct ip *ip, ipfw_insn *cmd) 284 { 285 int optlen, bits = 0; 286 u_char *cp = (u_char *)(ip + 1); 287 int x = (ip->ip_hl << 2) - sizeof (struct ip); 288 289 for (; x > 0; x -= optlen, cp += optlen) { 290 int opt = cp[IPOPT_OPTVAL]; 291 292 if (opt == IPOPT_EOL) 293 break; 294 if (opt == IPOPT_NOP) 295 optlen = 1; 296 else { 297 optlen = cp[IPOPT_OLEN]; 298 if (optlen <= 0 || optlen > x) 299 return 0; /* invalid or truncated */ 300 } 301 switch (opt) { 302 303 default: 304 break; 305 306 case IPOPT_LSRR: 307 bits |= IP_FW_IPOPT_LSRR; 308 break; 309 310 case IPOPT_SSRR: 311 bits |= IP_FW_IPOPT_SSRR; 312 break; 313 314 case IPOPT_RR: 315 bits |= IP_FW_IPOPT_RR; 316 break; 317 318 case IPOPT_TS: 319 bits |= IP_FW_IPOPT_TS; 320 break; 321 } 322 } 323 return (flags_match(cmd, bits)); 324 } 325 326 static int 327 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 328 { 329 int optlen, bits = 0; 330 u_char *cp = (u_char *)(tcp + 1); 331 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 332 333 for (; x > 0; x -= optlen, cp += optlen) { 334 int opt = cp[0]; 335 if (opt == TCPOPT_EOL) 336 break; 337 if (opt == TCPOPT_NOP) 338 optlen = 1; 339 else { 340 optlen = cp[1]; 341 if (optlen <= 0) 342 break; 343 } 344 345 switch (opt) { 346 347 default: 348 break; 349 350 case TCPOPT_MAXSEG: 351 bits |= IP_FW_TCPOPT_MSS; 352 break; 353 354 case TCPOPT_WINDOW: 355 bits |= IP_FW_TCPOPT_WINDOW; 356 break; 357 358 case TCPOPT_SACK_PERMITTED: 359 case TCPOPT_SACK: 360 bits |= IP_FW_TCPOPT_SACK; 361 break; 362 363 case TCPOPT_TIMESTAMP: 364 bits |= IP_FW_TCPOPT_TS; 365 break; 366 367 } 368 } 369 return (flags_match(cmd, bits)); 370 } 371 372 static int 373 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain, 374 uint32_t *tablearg) 375 { 376 377 if (ifp == NULL) /* no iface with this packet, match fails */ 378 return (0); 379 380 /* Check by name or by IP address */ 381 if (cmd->name[0] != '\0') { /* match by name */ 382 if (cmd->name[0] == '\1') /* use tablearg to match */ 383 return ipfw_lookup_table_extended(chain, cmd->p.kidx, 0, 384 &ifp->if_index, tablearg); 385 /* Check name */ 386 if (cmd->p.glob) { 387 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 388 return(1); 389 } else { 390 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 391 return(1); 392 } 393 } else { 394 #if !defined(USERSPACE) && defined(__FreeBSD__) /* and OSX too ? */ 395 struct ifaddr *ia; 396 397 if_addr_rlock(ifp); 398 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 399 if (ia->ifa_addr->sa_family != AF_INET) 400 continue; 401 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 402 (ia->ifa_addr))->sin_addr.s_addr) { 403 if_addr_runlock(ifp); 404 return(1); /* match */ 405 } 406 } 407 if_addr_runlock(ifp); 408 #endif /* __FreeBSD__ */ 409 } 410 return(0); /* no match, fail ... */ 411 } 412 413 /* 414 * The verify_path function checks if a route to the src exists and 415 * if it is reachable via ifp (when provided). 416 * 417 * The 'verrevpath' option checks that the interface that an IP packet 418 * arrives on is the same interface that traffic destined for the 419 * packet's source address would be routed out of. 420 * The 'versrcreach' option just checks that the source address is 421 * reachable via any route (except default) in the routing table. 422 * These two are a measure to block forged packets. This is also 423 * commonly known as "anti-spoofing" or Unicast Reverse Path 424 * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 425 * is purposely reminiscent of the Cisco IOS command, 426 * 427 * ip verify unicast reverse-path 428 * ip verify unicast source reachable-via any 429 * 430 * which implements the same functionality. But note that the syntax 431 * is misleading, and the check may be performed on all IP packets 432 * whether unicast, multicast, or broadcast. 433 */ 434 static int 435 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) 436 { 437 #if defined(USERSPACE) || !defined(__FreeBSD__) 438 return 0; 439 #else 440 struct route ro; 441 struct sockaddr_in *dst; 442 443 bzero(&ro, sizeof(ro)); 444 445 dst = (struct sockaddr_in *)&(ro.ro_dst); 446 dst->sin_family = AF_INET; 447 dst->sin_len = sizeof(*dst); 448 dst->sin_addr = src; 449 in_rtalloc_ign(&ro, 0, fib); 450 451 if (ro.ro_rt == NULL) 452 return 0; 453 454 /* 455 * If ifp is provided, check for equality with rtentry. 456 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 457 * in order to pass packets injected back by if_simloop(): 458 * routing entry (via lo0) for our own address 459 * may exist, so we need to handle routing assymetry. 460 */ 461 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 462 RTFREE(ro.ro_rt); 463 return 0; 464 } 465 466 /* if no ifp provided, check if rtentry is not default route */ 467 if (ifp == NULL && 468 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 469 RTFREE(ro.ro_rt); 470 return 0; 471 } 472 473 /* or if this is a blackhole/reject route */ 474 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 475 RTFREE(ro.ro_rt); 476 return 0; 477 } 478 479 /* found valid route */ 480 RTFREE(ro.ro_rt); 481 return 1; 482 #endif /* __FreeBSD__ */ 483 } 484 485 #ifdef INET6 486 /* 487 * ipv6 specific rules here... 488 */ 489 static __inline int 490 icmp6type_match (int type, ipfw_insn_u32 *cmd) 491 { 492 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 493 } 494 495 static int 496 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 497 { 498 int i; 499 for (i=0; i <= cmd->o.arg1; ++i ) 500 if (curr_flow == cmd->d[i] ) 501 return 1; 502 return 0; 503 } 504 505 /* support for IP6_*_ME opcodes */ 506 static int 507 search_ip6_addr_net (struct in6_addr * ip6_addr) 508 { 509 struct ifnet *mdc; 510 struct ifaddr *mdc2; 511 struct in6_ifaddr *fdm; 512 struct in6_addr copia; 513 514 TAILQ_FOREACH(mdc, &V_ifnet, if_link) { 515 if_addr_rlock(mdc); 516 TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) { 517 if (mdc2->ifa_addr->sa_family == AF_INET6) { 518 fdm = (struct in6_ifaddr *)mdc2; 519 copia = fdm->ia_addr.sin6_addr; 520 /* need for leaving scope_id in the sock_addr */ 521 in6_clearscope(&copia); 522 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) { 523 if_addr_runlock(mdc); 524 return 1; 525 } 526 } 527 } 528 if_addr_runlock(mdc); 529 } 530 return 0; 531 } 532 533 static int 534 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib) 535 { 536 struct route_in6 ro; 537 struct sockaddr_in6 *dst; 538 539 bzero(&ro, sizeof(ro)); 540 541 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 542 dst->sin6_family = AF_INET6; 543 dst->sin6_len = sizeof(*dst); 544 dst->sin6_addr = *src; 545 546 in6_rtalloc_ign(&ro, 0, fib); 547 if (ro.ro_rt == NULL) 548 return 0; 549 550 /* 551 * if ifp is provided, check for equality with rtentry 552 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 553 * to support the case of sending packets to an address of our own. 554 * (where the former interface is the first argument of if_simloop() 555 * (=ifp), the latter is lo0) 556 */ 557 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 558 RTFREE(ro.ro_rt); 559 return 0; 560 } 561 562 /* if no ifp provided, check if rtentry is not default route */ 563 if (ifp == NULL && 564 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 565 RTFREE(ro.ro_rt); 566 return 0; 567 } 568 569 /* or if this is a blackhole/reject route */ 570 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 571 RTFREE(ro.ro_rt); 572 return 0; 573 } 574 575 /* found valid route */ 576 RTFREE(ro.ro_rt); 577 return 1; 578 579 } 580 581 static int 582 is_icmp6_query(int icmp6_type) 583 { 584 if ((icmp6_type <= ICMP6_MAXTYPE) && 585 (icmp6_type == ICMP6_ECHO_REQUEST || 586 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 587 icmp6_type == ICMP6_WRUREQUEST || 588 icmp6_type == ICMP6_FQDN_QUERY || 589 icmp6_type == ICMP6_NI_QUERY)) 590 return (1); 591 592 return (0); 593 } 594 595 static void 596 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 597 { 598 struct mbuf *m; 599 600 m = args->m; 601 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 602 struct tcphdr *tcp; 603 tcp = (struct tcphdr *)((char *)ip6 + hlen); 604 605 if ((tcp->th_flags & TH_RST) == 0) { 606 struct mbuf *m0; 607 m0 = ipfw_send_pkt(args->m, &(args->f_id), 608 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 609 tcp->th_flags | TH_RST); 610 if (m0 != NULL) 611 ip6_output(m0, NULL, NULL, 0, NULL, NULL, 612 NULL); 613 } 614 FREE_PKT(m); 615 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 616 #if 0 617 /* 618 * Unlike above, the mbufs need to line up with the ip6 hdr, 619 * as the contents are read. We need to m_adj() the 620 * needed amount. 621 * The mbuf will however be thrown away so we can adjust it. 622 * Remember we did an m_pullup on it already so we 623 * can make some assumptions about contiguousness. 624 */ 625 if (args->L3offset) 626 m_adj(m, args->L3offset); 627 #endif 628 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 629 } else 630 FREE_PKT(m); 631 632 args->m = NULL; 633 } 634 635 #endif /* INET6 */ 636 637 638 /* 639 * sends a reject message, consuming the mbuf passed as an argument. 640 */ 641 static void 642 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip) 643 { 644 645 #if 0 646 /* XXX When ip is not guaranteed to be at mtod() we will 647 * need to account for this */ 648 * The mbuf will however be thrown away so we can adjust it. 649 * Remember we did an m_pullup on it already so we 650 * can make some assumptions about contiguousness. 651 */ 652 if (args->L3offset) 653 m_adj(m, args->L3offset); 654 #endif 655 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 656 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 657 } else if (args->f_id.proto == IPPROTO_TCP) { 658 struct tcphdr *const tcp = 659 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 660 if ( (tcp->th_flags & TH_RST) == 0) { 661 struct mbuf *m; 662 m = ipfw_send_pkt(args->m, &(args->f_id), 663 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 664 tcp->th_flags | TH_RST); 665 if (m != NULL) 666 ip_output(m, NULL, NULL, 0, NULL, NULL); 667 } 668 FREE_PKT(args->m); 669 } else 670 FREE_PKT(args->m); 671 args->m = NULL; 672 } 673 674 /* 675 * Support for uid/gid/jail lookup. These tests are expensive 676 * (because we may need to look into the list of active sockets) 677 * so we cache the results. ugid_lookupp is 0 if we have not 678 * yet done a lookup, 1 if we succeeded, and -1 if we tried 679 * and failed. The function always returns the match value. 680 * We could actually spare the variable and use *uc, setting 681 * it to '(void *)check_uidgid if we have no info, NULL if 682 * we tried and failed, or any other value if successful. 683 */ 684 static int 685 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp, 686 struct ucred **uc) 687 { 688 #if defined(USERSPACE) 689 return 0; // not supported in userspace 690 #else 691 #ifndef __FreeBSD__ 692 /* XXX */ 693 return cred_check(insn, proto, oif, 694 dst_ip, dst_port, src_ip, src_port, 695 (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb); 696 #else /* FreeBSD */ 697 struct in_addr src_ip, dst_ip; 698 struct inpcbinfo *pi; 699 struct ipfw_flow_id *id; 700 struct inpcb *pcb, *inp; 701 struct ifnet *oif; 702 int lookupflags; 703 int match; 704 705 id = &args->f_id; 706 inp = args->inp; 707 oif = args->oif; 708 709 /* 710 * Check to see if the UDP or TCP stack supplied us with 711 * the PCB. If so, rather then holding a lock and looking 712 * up the PCB, we can use the one that was supplied. 713 */ 714 if (inp && *ugid_lookupp == 0) { 715 INP_LOCK_ASSERT(inp); 716 if (inp->inp_socket != NULL) { 717 *uc = crhold(inp->inp_cred); 718 *ugid_lookupp = 1; 719 } else 720 *ugid_lookupp = -1; 721 } 722 /* 723 * If we have already been here and the packet has no 724 * PCB entry associated with it, then we can safely 725 * assume that this is a no match. 726 */ 727 if (*ugid_lookupp == -1) 728 return (0); 729 if (id->proto == IPPROTO_TCP) { 730 lookupflags = 0; 731 pi = &V_tcbinfo; 732 } else if (id->proto == IPPROTO_UDP) { 733 lookupflags = INPLOOKUP_WILDCARD; 734 pi = &V_udbinfo; 735 } else 736 return 0; 737 lookupflags |= INPLOOKUP_RLOCKPCB; 738 match = 0; 739 if (*ugid_lookupp == 0) { 740 if (id->addr_type == 6) { 741 #ifdef INET6 742 if (oif == NULL) 743 pcb = in6_pcblookup_mbuf(pi, 744 &id->src_ip6, htons(id->src_port), 745 &id->dst_ip6, htons(id->dst_port), 746 lookupflags, oif, args->m); 747 else 748 pcb = in6_pcblookup_mbuf(pi, 749 &id->dst_ip6, htons(id->dst_port), 750 &id->src_ip6, htons(id->src_port), 751 lookupflags, oif, args->m); 752 #else 753 *ugid_lookupp = -1; 754 return (0); 755 #endif 756 } else { 757 src_ip.s_addr = htonl(id->src_ip); 758 dst_ip.s_addr = htonl(id->dst_ip); 759 if (oif == NULL) 760 pcb = in_pcblookup_mbuf(pi, 761 src_ip, htons(id->src_port), 762 dst_ip, htons(id->dst_port), 763 lookupflags, oif, args->m); 764 else 765 pcb = in_pcblookup_mbuf(pi, 766 dst_ip, htons(id->dst_port), 767 src_ip, htons(id->src_port), 768 lookupflags, oif, args->m); 769 } 770 if (pcb != NULL) { 771 INP_RLOCK_ASSERT(pcb); 772 *uc = crhold(pcb->inp_cred); 773 *ugid_lookupp = 1; 774 INP_RUNLOCK(pcb); 775 } 776 if (*ugid_lookupp == 0) { 777 /* 778 * We tried and failed, set the variable to -1 779 * so we will not try again on this packet. 780 */ 781 *ugid_lookupp = -1; 782 return (0); 783 } 784 } 785 if (insn->o.opcode == O_UID) 786 match = ((*uc)->cr_uid == (uid_t)insn->d[0]); 787 else if (insn->o.opcode == O_GID) 788 match = groupmember((gid_t)insn->d[0], *uc); 789 else if (insn->o.opcode == O_JAIL) 790 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]); 791 return (match); 792 #endif /* __FreeBSD__ */ 793 #endif /* not supported in userspace */ 794 } 795 796 /* 797 * Helper function to set args with info on the rule after the matching 798 * one. slot is precise, whereas we guess rule_id as they are 799 * assigned sequentially. 800 */ 801 static inline void 802 set_match(struct ip_fw_args *args, int slot, 803 struct ip_fw_chain *chain) 804 { 805 args->rule.chain_id = chain->id; 806 args->rule.slot = slot + 1; /* we use 0 as a marker */ 807 args->rule.rule_id = 1 + chain->map[slot]->id; 808 args->rule.rulenum = chain->map[slot]->rulenum; 809 } 810 811 #ifndef LINEAR_SKIPTO 812 /* 813 * Helper function to enable cached rule lookups using 814 * cached_id and cached_pos fields in ipfw rule. 815 */ 816 static int 817 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num, 818 int tablearg, int jump_backwards) 819 { 820 int f_pos; 821 822 /* If possible use cached f_pos (in f->cached_pos), 823 * whose version is written in f->cached_id 824 * (horrible hacks to avoid changing the ABI). 825 */ 826 if (num != IP_FW_TARG && f->cached_id == chain->id) 827 f_pos = f->cached_pos; 828 else { 829 int i = IP_FW_ARG_TABLEARG(chain, num, skipto); 830 /* make sure we do not jump backward */ 831 if (jump_backwards == 0 && i <= f->rulenum) 832 i = f->rulenum + 1; 833 if (chain->idxmap != NULL) 834 f_pos = chain->idxmap[i]; 835 else 836 f_pos = ipfw_find_rule(chain, i, 0); 837 /* update the cache */ 838 if (num != IP_FW_TARG) { 839 f->cached_id = chain->id; 840 f->cached_pos = f_pos; 841 } 842 } 843 844 return (f_pos); 845 } 846 #else 847 /* 848 * Helper function to enable real fast rule lookups. 849 */ 850 static int 851 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num, 852 int tablearg, int jump_backwards) 853 { 854 int f_pos; 855 856 num = IP_FW_ARG_TABLEARG(chain, num, skipto); 857 /* make sure we do not jump backward */ 858 if (jump_backwards == 0 && num <= f->rulenum) 859 num = f->rulenum + 1; 860 f_pos = chain->idxmap[num]; 861 862 return (f_pos); 863 } 864 #endif 865 866 #define TARG(k, f) IP_FW_ARG_TABLEARG(chain, k, f) 867 /* 868 * The main check routine for the firewall. 869 * 870 * All arguments are in args so we can modify them and return them 871 * back to the caller. 872 * 873 * Parameters: 874 * 875 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 876 * Starts with the IP header. 877 * args->eh (in) Mac header if present, NULL for layer3 packet. 878 * args->L3offset Number of bytes bypassed if we came from L2. 879 * e.g. often sizeof(eh) ** NOTYET ** 880 * args->oif Outgoing interface, NULL if packet is incoming. 881 * The incoming interface is in the mbuf. (in) 882 * args->divert_rule (in/out) 883 * Skip up to the first rule past this rule number; 884 * upon return, non-zero port number for divert or tee. 885 * 886 * args->rule Pointer to the last matching rule (in/out) 887 * args->next_hop Socket we are forwarding to (out). 888 * args->next_hop6 IPv6 next hop we are forwarding to (out). 889 * args->f_id Addresses grabbed from the packet (out) 890 * args->rule.info a cookie depending on rule action 891 * 892 * Return value: 893 * 894 * IP_FW_PASS the packet must be accepted 895 * IP_FW_DENY the packet must be dropped 896 * IP_FW_DIVERT divert packet, port in m_tag 897 * IP_FW_TEE tee packet, port in m_tag 898 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 899 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 900 * args->rule contains the matching rule, 901 * args->rule.info has additional information. 902 * 903 */ 904 int 905 ipfw_chk(struct ip_fw_args *args) 906 { 907 908 /* 909 * Local variables holding state while processing a packet: 910 * 911 * IMPORTANT NOTE: to speed up the processing of rules, there 912 * are some assumption on the values of the variables, which 913 * are documented here. Should you change them, please check 914 * the implementation of the various instructions to make sure 915 * that they still work. 916 * 917 * args->eh The MAC header. It is non-null for a layer2 918 * packet, it is NULL for a layer-3 packet. 919 * **notyet** 920 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. 921 * 922 * m | args->m Pointer to the mbuf, as received from the caller. 923 * It may change if ipfw_chk() does an m_pullup, or if it 924 * consumes the packet because it calls send_reject(). 925 * XXX This has to change, so that ipfw_chk() never modifies 926 * or consumes the buffer. 927 * ip is the beginning of the ip(4 or 6) header. 928 * Calculated by adding the L3offset to the start of data. 929 * (Until we start using L3offset, the packet is 930 * supposed to start with the ip header). 931 */ 932 struct mbuf *m = args->m; 933 struct ip *ip = mtod(m, struct ip *); 934 935 /* 936 * For rules which contain uid/gid or jail constraints, cache 937 * a copy of the users credentials after the pcb lookup has been 938 * executed. This will speed up the processing of rules with 939 * these types of constraints, as well as decrease contention 940 * on pcb related locks. 941 */ 942 #ifndef __FreeBSD__ 943 struct bsd_ucred ucred_cache; 944 #else 945 struct ucred *ucred_cache = NULL; 946 #endif 947 int ucred_lookup = 0; 948 949 /* 950 * oif | args->oif If NULL, ipfw_chk has been called on the 951 * inbound path (ether_input, ip_input). 952 * If non-NULL, ipfw_chk has been called on the outbound path 953 * (ether_output, ip_output). 954 */ 955 struct ifnet *oif = args->oif; 956 957 int f_pos = 0; /* index of current rule in the array */ 958 int retval = 0; 959 960 /* 961 * hlen The length of the IP header. 962 */ 963 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 964 965 /* 966 * offset The offset of a fragment. offset != 0 means that 967 * we have a fragment at this offset of an IPv4 packet. 968 * offset == 0 means that (if this is an IPv4 packet) 969 * this is the first or only fragment. 970 * For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header 971 * or there is a single packet fragement (fragement header added 972 * without needed). We will treat a single packet fragment as if 973 * there was no fragment header (or log/block depending on the 974 * V_fw_permit_single_frag6 sysctl setting). 975 */ 976 u_short offset = 0; 977 u_short ip6f_mf = 0; 978 979 /* 980 * Local copies of addresses. They are only valid if we have 981 * an IP packet. 982 * 983 * proto The protocol. Set to 0 for non-ip packets, 984 * or to the protocol read from the packet otherwise. 985 * proto != 0 means that we have an IPv4 packet. 986 * 987 * src_port, dst_port port numbers, in HOST format. Only 988 * valid for TCP and UDP packets. 989 * 990 * src_ip, dst_ip ip addresses, in NETWORK format. 991 * Only valid for IPv4 packets. 992 */ 993 uint8_t proto; 994 uint16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 995 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 996 uint16_t iplen=0; 997 int pktlen; 998 uint16_t etype = 0; /* Host order stored ether type */ 999 1000 /* 1001 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 1002 * MATCH_NONE when checked and not matched (q = NULL), 1003 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 1004 */ 1005 int dyn_dir = MATCH_UNKNOWN; 1006 ipfw_dyn_rule *q = NULL; 1007 struct ip_fw_chain *chain = &V_layer3_chain; 1008 1009 /* 1010 * We store in ulp a pointer to the upper layer protocol header. 1011 * In the ipv4 case this is easy to determine from the header, 1012 * but for ipv6 we might have some additional headers in the middle. 1013 * ulp is NULL if not found. 1014 */ 1015 void *ulp = NULL; /* upper layer protocol pointer. */ 1016 1017 /* XXX ipv6 variables */ 1018 int is_ipv6 = 0; 1019 uint8_t icmp6_type = 0; 1020 uint16_t ext_hd = 0; /* bits vector for extension header filtering */ 1021 /* end of ipv6 variables */ 1022 1023 int is_ipv4 = 0; 1024 1025 int done = 0; /* flag to exit the outer loop */ 1026 IPFW_RLOCK_TRACKER; 1027 1028 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready)) 1029 return (IP_FW_PASS); /* accept */ 1030 1031 dst_ip.s_addr = 0; /* make sure it is initialized */ 1032 src_ip.s_addr = 0; /* make sure it is initialized */ 1033 pktlen = m->m_pkthdr.len; 1034 args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */ 1035 proto = args->f_id.proto = 0; /* mark f_id invalid */ 1036 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 1037 1038 /* 1039 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 1040 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 1041 * pointer might become stale after other pullups (but we never use it 1042 * this way). 1043 */ 1044 #define PULLUP_TO(_len, p, T) PULLUP_LEN(_len, p, sizeof(T)) 1045 #define PULLUP_LEN(_len, p, T) \ 1046 do { \ 1047 int x = (_len) + T; \ 1048 if ((m)->m_len < x) { \ 1049 args->m = m = m_pullup(m, x); \ 1050 if (m == NULL) \ 1051 goto pullup_failed; \ 1052 } \ 1053 p = (mtod(m, char *) + (_len)); \ 1054 } while (0) 1055 1056 /* 1057 * if we have an ether header, 1058 */ 1059 if (args->eh) 1060 etype = ntohs(args->eh->ether_type); 1061 1062 /* Identify IP packets and fill up variables. */ 1063 if (pktlen >= sizeof(struct ip6_hdr) && 1064 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { 1065 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 1066 is_ipv6 = 1; 1067 args->f_id.addr_type = 6; 1068 hlen = sizeof(struct ip6_hdr); 1069 proto = ip6->ip6_nxt; 1070 1071 /* Search extension headers to find upper layer protocols */ 1072 while (ulp == NULL && offset == 0) { 1073 switch (proto) { 1074 case IPPROTO_ICMPV6: 1075 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 1076 icmp6_type = ICMP6(ulp)->icmp6_type; 1077 break; 1078 1079 case IPPROTO_TCP: 1080 PULLUP_TO(hlen, ulp, struct tcphdr); 1081 dst_port = TCP(ulp)->th_dport; 1082 src_port = TCP(ulp)->th_sport; 1083 /* save flags for dynamic rules */ 1084 args->f_id._flags = TCP(ulp)->th_flags; 1085 break; 1086 1087 case IPPROTO_SCTP: 1088 PULLUP_TO(hlen, ulp, struct sctphdr); 1089 src_port = SCTP(ulp)->src_port; 1090 dst_port = SCTP(ulp)->dest_port; 1091 break; 1092 1093 case IPPROTO_UDP: 1094 PULLUP_TO(hlen, ulp, struct udphdr); 1095 dst_port = UDP(ulp)->uh_dport; 1096 src_port = UDP(ulp)->uh_sport; 1097 break; 1098 1099 case IPPROTO_HOPOPTS: /* RFC 2460 */ 1100 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1101 ext_hd |= EXT_HOPOPTS; 1102 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1103 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1104 ulp = NULL; 1105 break; 1106 1107 case IPPROTO_ROUTING: /* RFC 2460 */ 1108 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 1109 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 1110 case 0: 1111 ext_hd |= EXT_RTHDR0; 1112 break; 1113 case 2: 1114 ext_hd |= EXT_RTHDR2; 1115 break; 1116 default: 1117 if (V_fw_verbose) 1118 printf("IPFW2: IPV6 - Unknown " 1119 "Routing Header type(%d)\n", 1120 ((struct ip6_rthdr *) 1121 ulp)->ip6r_type); 1122 if (V_fw_deny_unknown_exthdrs) 1123 return (IP_FW_DENY); 1124 break; 1125 } 1126 ext_hd |= EXT_ROUTING; 1127 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 1128 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 1129 ulp = NULL; 1130 break; 1131 1132 case IPPROTO_FRAGMENT: /* RFC 2460 */ 1133 PULLUP_TO(hlen, ulp, struct ip6_frag); 1134 ext_hd |= EXT_FRAGMENT; 1135 hlen += sizeof (struct ip6_frag); 1136 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 1137 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 1138 IP6F_OFF_MASK; 1139 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg & 1140 IP6F_MORE_FRAG; 1141 if (V_fw_permit_single_frag6 == 0 && 1142 offset == 0 && ip6f_mf == 0) { 1143 if (V_fw_verbose) 1144 printf("IPFW2: IPV6 - Invalid " 1145 "Fragment Header\n"); 1146 if (V_fw_deny_unknown_exthdrs) 1147 return (IP_FW_DENY); 1148 break; 1149 } 1150 args->f_id.extra = 1151 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 1152 ulp = NULL; 1153 break; 1154 1155 case IPPROTO_DSTOPTS: /* RFC 2460 */ 1156 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1157 ext_hd |= EXT_DSTOPTS; 1158 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1159 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1160 ulp = NULL; 1161 break; 1162 1163 case IPPROTO_AH: /* RFC 2402 */ 1164 PULLUP_TO(hlen, ulp, struct ip6_ext); 1165 ext_hd |= EXT_AH; 1166 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 1167 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 1168 ulp = NULL; 1169 break; 1170 1171 case IPPROTO_ESP: /* RFC 2406 */ 1172 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 1173 /* Anything past Seq# is variable length and 1174 * data past this ext. header is encrypted. */ 1175 ext_hd |= EXT_ESP; 1176 break; 1177 1178 case IPPROTO_NONE: /* RFC 2460 */ 1179 /* 1180 * Packet ends here, and IPv6 header has 1181 * already been pulled up. If ip6e_len!=0 1182 * then octets must be ignored. 1183 */ 1184 ulp = ip; /* non-NULL to get out of loop. */ 1185 break; 1186 1187 case IPPROTO_OSPFIGP: 1188 /* XXX OSPF header check? */ 1189 PULLUP_TO(hlen, ulp, struct ip6_ext); 1190 break; 1191 1192 case IPPROTO_PIM: 1193 /* XXX PIM header check? */ 1194 PULLUP_TO(hlen, ulp, struct pim); 1195 break; 1196 1197 case IPPROTO_CARP: 1198 PULLUP_TO(hlen, ulp, struct carp_header); 1199 if (((struct carp_header *)ulp)->carp_version != 1200 CARP_VERSION) 1201 return (IP_FW_DENY); 1202 if (((struct carp_header *)ulp)->carp_type != 1203 CARP_ADVERTISEMENT) 1204 return (IP_FW_DENY); 1205 break; 1206 1207 case IPPROTO_IPV6: /* RFC 2893 */ 1208 PULLUP_TO(hlen, ulp, struct ip6_hdr); 1209 break; 1210 1211 case IPPROTO_IPV4: /* RFC 2893 */ 1212 PULLUP_TO(hlen, ulp, struct ip); 1213 break; 1214 1215 default: 1216 if (V_fw_verbose) 1217 printf("IPFW2: IPV6 - Unknown " 1218 "Extension Header(%d), ext_hd=%x\n", 1219 proto, ext_hd); 1220 if (V_fw_deny_unknown_exthdrs) 1221 return (IP_FW_DENY); 1222 PULLUP_TO(hlen, ulp, struct ip6_ext); 1223 break; 1224 } /*switch */ 1225 } 1226 ip = mtod(m, struct ip *); 1227 ip6 = (struct ip6_hdr *)ip; 1228 args->f_id.src_ip6 = ip6->ip6_src; 1229 args->f_id.dst_ip6 = ip6->ip6_dst; 1230 args->f_id.src_ip = 0; 1231 args->f_id.dst_ip = 0; 1232 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 1233 } else if (pktlen >= sizeof(struct ip) && 1234 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { 1235 is_ipv4 = 1; 1236 hlen = ip->ip_hl << 2; 1237 args->f_id.addr_type = 4; 1238 1239 /* 1240 * Collect parameters into local variables for faster matching. 1241 */ 1242 proto = ip->ip_p; 1243 src_ip = ip->ip_src; 1244 dst_ip = ip->ip_dst; 1245 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1246 iplen = ntohs(ip->ip_len); 1247 pktlen = iplen < pktlen ? iplen : pktlen; 1248 1249 if (offset == 0) { 1250 switch (proto) { 1251 case IPPROTO_TCP: 1252 PULLUP_TO(hlen, ulp, struct tcphdr); 1253 dst_port = TCP(ulp)->th_dport; 1254 src_port = TCP(ulp)->th_sport; 1255 /* save flags for dynamic rules */ 1256 args->f_id._flags = TCP(ulp)->th_flags; 1257 break; 1258 1259 case IPPROTO_SCTP: 1260 PULLUP_TO(hlen, ulp, struct sctphdr); 1261 src_port = SCTP(ulp)->src_port; 1262 dst_port = SCTP(ulp)->dest_port; 1263 break; 1264 1265 case IPPROTO_UDP: 1266 PULLUP_TO(hlen, ulp, struct udphdr); 1267 dst_port = UDP(ulp)->uh_dport; 1268 src_port = UDP(ulp)->uh_sport; 1269 break; 1270 1271 case IPPROTO_ICMP: 1272 PULLUP_TO(hlen, ulp, struct icmphdr); 1273 //args->f_id.flags = ICMP(ulp)->icmp_type; 1274 break; 1275 1276 default: 1277 break; 1278 } 1279 } 1280 1281 ip = mtod(m, struct ip *); 1282 args->f_id.src_ip = ntohl(src_ip.s_addr); 1283 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1284 } 1285 #undef PULLUP_TO 1286 if (proto) { /* we may have port numbers, store them */ 1287 args->f_id.proto = proto; 1288 args->f_id.src_port = src_port = ntohs(src_port); 1289 args->f_id.dst_port = dst_port = ntohs(dst_port); 1290 } 1291 1292 IPFW_PF_RLOCK(chain); 1293 if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */ 1294 IPFW_PF_RUNLOCK(chain); 1295 return (IP_FW_PASS); /* accept */ 1296 } 1297 if (args->rule.slot) { 1298 /* 1299 * Packet has already been tagged as a result of a previous 1300 * match on rule args->rule aka args->rule_id (PIPE, QUEUE, 1301 * REASS, NETGRAPH, DIVERT/TEE...) 1302 * Validate the slot and continue from the next one 1303 * if still present, otherwise do a lookup. 1304 */ 1305 f_pos = (args->rule.chain_id == chain->id) ? 1306 args->rule.slot : 1307 ipfw_find_rule(chain, args->rule.rulenum, 1308 args->rule.rule_id); 1309 } else { 1310 f_pos = 0; 1311 } 1312 1313 /* 1314 * Now scan the rules, and parse microinstructions for each rule. 1315 * We have two nested loops and an inner switch. Sometimes we 1316 * need to break out of one or both loops, or re-enter one of 1317 * the loops with updated variables. Loop variables are: 1318 * 1319 * f_pos (outer loop) points to the current rule. 1320 * On output it points to the matching rule. 1321 * done (outer loop) is used as a flag to break the loop. 1322 * l (inner loop) residual length of current rule. 1323 * cmd points to the current microinstruction. 1324 * 1325 * We break the inner loop by setting l=0 and possibly 1326 * cmdlen=0 if we don't want to advance cmd. 1327 * We break the outer loop by setting done=1 1328 * We can restart the inner loop by setting l>0 and f_pos, f, cmd 1329 * as needed. 1330 */ 1331 for (; f_pos < chain->n_rules; f_pos++) { 1332 ipfw_insn *cmd; 1333 uint32_t tablearg = 0; 1334 int l, cmdlen, skip_or; /* skip rest of OR block */ 1335 struct ip_fw *f; 1336 1337 f = chain->map[f_pos]; 1338 if (V_set_disable & (1 << f->set) ) 1339 continue; 1340 1341 skip_or = 0; 1342 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1343 l -= cmdlen, cmd += cmdlen) { 1344 int match; 1345 1346 /* 1347 * check_body is a jump target used when we find a 1348 * CHECK_STATE, and need to jump to the body of 1349 * the target rule. 1350 */ 1351 1352 /* check_body: */ 1353 cmdlen = F_LEN(cmd); 1354 /* 1355 * An OR block (insn_1 || .. || insn_n) has the 1356 * F_OR bit set in all but the last instruction. 1357 * The first match will set "skip_or", and cause 1358 * the following instructions to be skipped until 1359 * past the one with the F_OR bit clear. 1360 */ 1361 if (skip_or) { /* skip this instruction */ 1362 if ((cmd->len & F_OR) == 0) 1363 skip_or = 0; /* next one is good */ 1364 continue; 1365 } 1366 match = 0; /* set to 1 if we succeed */ 1367 1368 switch (cmd->opcode) { 1369 /* 1370 * The first set of opcodes compares the packet's 1371 * fields with some pattern, setting 'match' if a 1372 * match is found. At the end of the loop there is 1373 * logic to deal with F_NOT and F_OR flags associated 1374 * with the opcode. 1375 */ 1376 case O_NOP: 1377 match = 1; 1378 break; 1379 1380 case O_FORWARD_MAC: 1381 printf("ipfw: opcode %d unimplemented\n", 1382 cmd->opcode); 1383 break; 1384 1385 case O_GID: 1386 case O_UID: 1387 case O_JAIL: 1388 /* 1389 * We only check offset == 0 && proto != 0, 1390 * as this ensures that we have a 1391 * packet with the ports info. 1392 */ 1393 if (offset != 0) 1394 break; 1395 if (proto == IPPROTO_TCP || 1396 proto == IPPROTO_UDP) 1397 match = check_uidgid( 1398 (ipfw_insn_u32 *)cmd, 1399 args, &ucred_lookup, 1400 #ifdef __FreeBSD__ 1401 &ucred_cache); 1402 #else 1403 (void *)&ucred_cache); 1404 #endif 1405 break; 1406 1407 case O_RECV: 1408 match = iface_match(m->m_pkthdr.rcvif, 1409 (ipfw_insn_if *)cmd, chain, &tablearg); 1410 break; 1411 1412 case O_XMIT: 1413 match = iface_match(oif, (ipfw_insn_if *)cmd, 1414 chain, &tablearg); 1415 break; 1416 1417 case O_VIA: 1418 match = iface_match(oif ? oif : 1419 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd, 1420 chain, &tablearg); 1421 break; 1422 1423 case O_MACADDR2: 1424 if (args->eh != NULL) { /* have MAC header */ 1425 u_int32_t *want = (u_int32_t *) 1426 ((ipfw_insn_mac *)cmd)->addr; 1427 u_int32_t *mask = (u_int32_t *) 1428 ((ipfw_insn_mac *)cmd)->mask; 1429 u_int32_t *hdr = (u_int32_t *)args->eh; 1430 1431 match = 1432 ( want[0] == (hdr[0] & mask[0]) && 1433 want[1] == (hdr[1] & mask[1]) && 1434 want[2] == (hdr[2] & mask[2]) ); 1435 } 1436 break; 1437 1438 case O_MAC_TYPE: 1439 if (args->eh != NULL) { 1440 u_int16_t *p = 1441 ((ipfw_insn_u16 *)cmd)->ports; 1442 int i; 1443 1444 for (i = cmdlen - 1; !match && i>0; 1445 i--, p += 2) 1446 match = (etype >= p[0] && 1447 etype <= p[1]); 1448 } 1449 break; 1450 1451 case O_FRAG: 1452 match = (offset != 0); 1453 break; 1454 1455 case O_IN: /* "out" is "not in" */ 1456 match = (oif == NULL); 1457 break; 1458 1459 case O_LAYER2: 1460 match = (args->eh != NULL); 1461 break; 1462 1463 case O_DIVERTED: 1464 { 1465 /* For diverted packets, args->rule.info 1466 * contains the divert port (in host format) 1467 * reason and direction. 1468 */ 1469 uint32_t i = args->rule.info; 1470 match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT && 1471 cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2); 1472 } 1473 break; 1474 1475 case O_PROTO: 1476 /* 1477 * We do not allow an arg of 0 so the 1478 * check of "proto" only suffices. 1479 */ 1480 match = (proto == cmd->arg1); 1481 break; 1482 1483 case O_IP_SRC: 1484 match = is_ipv4 && 1485 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1486 src_ip.s_addr); 1487 break; 1488 1489 case O_IP_SRC_LOOKUP: 1490 case O_IP_DST_LOOKUP: 1491 if (is_ipv4) { 1492 uint32_t key = 1493 (cmd->opcode == O_IP_DST_LOOKUP) ? 1494 dst_ip.s_addr : src_ip.s_addr; 1495 uint32_t v = 0; 1496 1497 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) { 1498 /* generic lookup. The key must be 1499 * in 32bit big-endian format. 1500 */ 1501 v = ((ipfw_insn_u32 *)cmd)->d[1]; 1502 if (v == 0) 1503 key = dst_ip.s_addr; 1504 else if (v == 1) 1505 key = src_ip.s_addr; 1506 else if (v == 6) /* dscp */ 1507 key = (ip->ip_tos >> 2) & 0x3f; 1508 else if (offset != 0) 1509 break; 1510 else if (proto != IPPROTO_TCP && 1511 proto != IPPROTO_UDP) 1512 break; 1513 else if (v == 2) 1514 key = dst_port; 1515 else if (v == 3) 1516 key = src_port; 1517 #ifndef USERSPACE 1518 else if (v == 4 || v == 5) { 1519 check_uidgid( 1520 (ipfw_insn_u32 *)cmd, 1521 args, &ucred_lookup, 1522 #ifdef __FreeBSD__ 1523 &ucred_cache); 1524 if (v == 4 /* O_UID */) 1525 key = ucred_cache->cr_uid; 1526 else if (v == 5 /* O_JAIL */) 1527 key = ucred_cache->cr_prison->pr_id; 1528 #else /* !__FreeBSD__ */ 1529 (void *)&ucred_cache); 1530 if (v ==4 /* O_UID */) 1531 key = ucred_cache.uid; 1532 else if (v == 5 /* O_JAIL */) 1533 key = ucred_cache.xid; 1534 #endif /* !__FreeBSD__ */ 1535 } 1536 #endif /* !USERSPACE */ 1537 else 1538 break; 1539 } 1540 match = ipfw_lookup_table(chain, 1541 cmd->arg1, key, &v); 1542 if (!match) 1543 break; 1544 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 1545 match = 1546 ((ipfw_insn_u32 *)cmd)->d[0] == v; 1547 else 1548 tablearg = v; 1549 } else if (is_ipv6) { 1550 uint32_t v = 0; 1551 void *pkey = (cmd->opcode == O_IP_DST_LOOKUP) ? 1552 &args->f_id.dst_ip6: &args->f_id.src_ip6; 1553 match = ipfw_lookup_table_extended(chain, 1554 cmd->arg1, 1555 sizeof(struct in6_addr), 1556 pkey, &v); 1557 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 1558 match = ((ipfw_insn_u32 *)cmd)->d[0] == v; 1559 if (match) 1560 tablearg = v; 1561 } 1562 break; 1563 1564 case O_IP_FLOW_LOOKUP: 1565 { 1566 uint32_t v = 0; 1567 match = ipfw_lookup_table_extended(chain, 1568 cmd->arg1, 0, &args->f_id, &v); 1569 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 1570 match = ((ipfw_insn_u32 *)cmd)->d[0] == v; 1571 if (match) 1572 tablearg = v; 1573 } 1574 break; 1575 case O_IP_SRC_MASK: 1576 case O_IP_DST_MASK: 1577 if (is_ipv4) { 1578 uint32_t a = 1579 (cmd->opcode == O_IP_DST_MASK) ? 1580 dst_ip.s_addr : src_ip.s_addr; 1581 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 1582 int i = cmdlen-1; 1583 1584 for (; !match && i>0; i-= 2, p+= 2) 1585 match = (p[0] == (a & p[1])); 1586 } 1587 break; 1588 1589 case O_IP_SRC_ME: 1590 if (is_ipv4) { 1591 struct ifnet *tif; 1592 1593 INADDR_TO_IFP(src_ip, tif); 1594 match = (tif != NULL); 1595 break; 1596 } 1597 #ifdef INET6 1598 /* FALLTHROUGH */ 1599 case O_IP6_SRC_ME: 1600 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 1601 #endif 1602 break; 1603 1604 case O_IP_DST_SET: 1605 case O_IP_SRC_SET: 1606 if (is_ipv4) { 1607 u_int32_t *d = (u_int32_t *)(cmd+1); 1608 u_int32_t addr = 1609 cmd->opcode == O_IP_DST_SET ? 1610 args->f_id.dst_ip : 1611 args->f_id.src_ip; 1612 1613 if (addr < d[0]) 1614 break; 1615 addr -= d[0]; /* subtract base */ 1616 match = (addr < cmd->arg1) && 1617 ( d[ 1 + (addr>>5)] & 1618 (1<<(addr & 0x1f)) ); 1619 } 1620 break; 1621 1622 case O_IP_DST: 1623 match = is_ipv4 && 1624 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1625 dst_ip.s_addr); 1626 break; 1627 1628 case O_IP_DST_ME: 1629 if (is_ipv4) { 1630 struct ifnet *tif; 1631 1632 INADDR_TO_IFP(dst_ip, tif); 1633 match = (tif != NULL); 1634 break; 1635 } 1636 #ifdef INET6 1637 /* FALLTHROUGH */ 1638 case O_IP6_DST_ME: 1639 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 1640 #endif 1641 break; 1642 1643 1644 case O_IP_SRCPORT: 1645 case O_IP_DSTPORT: 1646 /* 1647 * offset == 0 && proto != 0 is enough 1648 * to guarantee that we have a 1649 * packet with port info. 1650 */ 1651 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 1652 && offset == 0) { 1653 u_int16_t x = 1654 (cmd->opcode == O_IP_SRCPORT) ? 1655 src_port : dst_port ; 1656 u_int16_t *p = 1657 ((ipfw_insn_u16 *)cmd)->ports; 1658 int i; 1659 1660 for (i = cmdlen - 1; !match && i>0; 1661 i--, p += 2) 1662 match = (x>=p[0] && x<=p[1]); 1663 } 1664 break; 1665 1666 case O_ICMPTYPE: 1667 match = (offset == 0 && proto==IPPROTO_ICMP && 1668 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 1669 break; 1670 1671 #ifdef INET6 1672 case O_ICMP6TYPE: 1673 match = is_ipv6 && offset == 0 && 1674 proto==IPPROTO_ICMPV6 && 1675 icmp6type_match( 1676 ICMP6(ulp)->icmp6_type, 1677 (ipfw_insn_u32 *)cmd); 1678 break; 1679 #endif /* INET6 */ 1680 1681 case O_IPOPT: 1682 match = (is_ipv4 && 1683 ipopts_match(ip, cmd) ); 1684 break; 1685 1686 case O_IPVER: 1687 match = (is_ipv4 && 1688 cmd->arg1 == ip->ip_v); 1689 break; 1690 1691 case O_IPID: 1692 case O_IPLEN: 1693 case O_IPTTL: 1694 if (is_ipv4) { /* only for IP packets */ 1695 uint16_t x; 1696 uint16_t *p; 1697 int i; 1698 1699 if (cmd->opcode == O_IPLEN) 1700 x = iplen; 1701 else if (cmd->opcode == O_IPTTL) 1702 x = ip->ip_ttl; 1703 else /* must be IPID */ 1704 x = ntohs(ip->ip_id); 1705 if (cmdlen == 1) { 1706 match = (cmd->arg1 == x); 1707 break; 1708 } 1709 /* otherwise we have ranges */ 1710 p = ((ipfw_insn_u16 *)cmd)->ports; 1711 i = cmdlen - 1; 1712 for (; !match && i>0; i--, p += 2) 1713 match = (x >= p[0] && x <= p[1]); 1714 } 1715 break; 1716 1717 case O_IPPRECEDENCE: 1718 match = (is_ipv4 && 1719 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 1720 break; 1721 1722 case O_IPTOS: 1723 match = (is_ipv4 && 1724 flags_match(cmd, ip->ip_tos)); 1725 break; 1726 1727 case O_DSCP: 1728 { 1729 uint32_t *p; 1730 uint16_t x; 1731 1732 p = ((ipfw_insn_u32 *)cmd)->d; 1733 1734 if (is_ipv4) 1735 x = ip->ip_tos >> 2; 1736 else if (is_ipv6) { 1737 uint8_t *v; 1738 v = &((struct ip6_hdr *)ip)->ip6_vfc; 1739 x = (*v & 0x0F) << 2; 1740 v++; 1741 x |= *v >> 6; 1742 } else 1743 break; 1744 1745 /* DSCP bitmask is stored as low_u32 high_u32 */ 1746 if (x > 32) 1747 match = *(p + 1) & (1 << (x - 32)); 1748 else 1749 match = *p & (1 << x); 1750 } 1751 break; 1752 1753 case O_TCPDATALEN: 1754 if (proto == IPPROTO_TCP && offset == 0) { 1755 struct tcphdr *tcp; 1756 uint16_t x; 1757 uint16_t *p; 1758 int i; 1759 1760 tcp = TCP(ulp); 1761 x = iplen - 1762 ((ip->ip_hl + tcp->th_off) << 2); 1763 if (cmdlen == 1) { 1764 match = (cmd->arg1 == x); 1765 break; 1766 } 1767 /* otherwise we have ranges */ 1768 p = ((ipfw_insn_u16 *)cmd)->ports; 1769 i = cmdlen - 1; 1770 for (; !match && i>0; i--, p += 2) 1771 match = (x >= p[0] && x <= p[1]); 1772 } 1773 break; 1774 1775 case O_TCPFLAGS: 1776 match = (proto == IPPROTO_TCP && offset == 0 && 1777 flags_match(cmd, TCP(ulp)->th_flags)); 1778 break; 1779 1780 case O_TCPOPTS: 1781 if (proto == IPPROTO_TCP && offset == 0 && ulp){ 1782 PULLUP_LEN(hlen, ulp, 1783 (TCP(ulp)->th_off << 2)); 1784 match = tcpopts_match(TCP(ulp), cmd); 1785 } 1786 break; 1787 1788 case O_TCPSEQ: 1789 match = (proto == IPPROTO_TCP && offset == 0 && 1790 ((ipfw_insn_u32 *)cmd)->d[0] == 1791 TCP(ulp)->th_seq); 1792 break; 1793 1794 case O_TCPACK: 1795 match = (proto == IPPROTO_TCP && offset == 0 && 1796 ((ipfw_insn_u32 *)cmd)->d[0] == 1797 TCP(ulp)->th_ack); 1798 break; 1799 1800 case O_TCPWIN: 1801 if (proto == IPPROTO_TCP && offset == 0) { 1802 uint16_t x; 1803 uint16_t *p; 1804 int i; 1805 1806 x = ntohs(TCP(ulp)->th_win); 1807 if (cmdlen == 1) { 1808 match = (cmd->arg1 == x); 1809 break; 1810 } 1811 /* Otherwise we have ranges. */ 1812 p = ((ipfw_insn_u16 *)cmd)->ports; 1813 i = cmdlen - 1; 1814 for (; !match && i > 0; i--, p += 2) 1815 match = (x >= p[0] && x <= p[1]); 1816 } 1817 break; 1818 1819 case O_ESTAB: 1820 /* reject packets which have SYN only */ 1821 /* XXX should i also check for TH_ACK ? */ 1822 match = (proto == IPPROTO_TCP && offset == 0 && 1823 (TCP(ulp)->th_flags & 1824 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 1825 break; 1826 1827 case O_ALTQ: { 1828 struct pf_mtag *at; 1829 struct m_tag *mtag; 1830 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 1831 1832 /* 1833 * ALTQ uses mbuf tags from another 1834 * packet filtering system - pf(4). 1835 * We allocate a tag in its format 1836 * and fill it in, pretending to be pf(4). 1837 */ 1838 match = 1; 1839 at = pf_find_mtag(m); 1840 if (at != NULL && at->qid != 0) 1841 break; 1842 mtag = m_tag_get(PACKET_TAG_PF, 1843 sizeof(struct pf_mtag), M_NOWAIT | M_ZERO); 1844 if (mtag == NULL) { 1845 /* 1846 * Let the packet fall back to the 1847 * default ALTQ. 1848 */ 1849 break; 1850 } 1851 m_tag_prepend(m, mtag); 1852 at = (struct pf_mtag *)(mtag + 1); 1853 at->qid = altq->qid; 1854 at->hdr = ip; 1855 break; 1856 } 1857 1858 case O_LOG: 1859 ipfw_log(chain, f, hlen, args, m, 1860 oif, offset | ip6f_mf, tablearg, ip); 1861 match = 1; 1862 break; 1863 1864 case O_PROB: 1865 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 1866 break; 1867 1868 case O_VERREVPATH: 1869 /* Outgoing packets automatically pass/match */ 1870 match = ((oif != NULL) || 1871 (m->m_pkthdr.rcvif == NULL) || 1872 ( 1873 #ifdef INET6 1874 is_ipv6 ? 1875 verify_path6(&(args->f_id.src_ip6), 1876 m->m_pkthdr.rcvif, args->f_id.fib) : 1877 #endif 1878 verify_path(src_ip, m->m_pkthdr.rcvif, 1879 args->f_id.fib))); 1880 break; 1881 1882 case O_VERSRCREACH: 1883 /* Outgoing packets automatically pass/match */ 1884 match = (hlen > 0 && ((oif != NULL) || 1885 #ifdef INET6 1886 is_ipv6 ? 1887 verify_path6(&(args->f_id.src_ip6), 1888 NULL, args->f_id.fib) : 1889 #endif 1890 verify_path(src_ip, NULL, args->f_id.fib))); 1891 break; 1892 1893 case O_ANTISPOOF: 1894 /* Outgoing packets automatically pass/match */ 1895 if (oif == NULL && hlen > 0 && 1896 ( (is_ipv4 && in_localaddr(src_ip)) 1897 #ifdef INET6 1898 || (is_ipv6 && 1899 in6_localaddr(&(args->f_id.src_ip6))) 1900 #endif 1901 )) 1902 match = 1903 #ifdef INET6 1904 is_ipv6 ? verify_path6( 1905 &(args->f_id.src_ip6), 1906 m->m_pkthdr.rcvif, 1907 args->f_id.fib) : 1908 #endif 1909 verify_path(src_ip, 1910 m->m_pkthdr.rcvif, 1911 args->f_id.fib); 1912 else 1913 match = 1; 1914 break; 1915 1916 case O_IPSEC: 1917 #ifdef IPSEC 1918 match = (m_tag_find(m, 1919 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 1920 #endif 1921 /* otherwise no match */ 1922 break; 1923 1924 #ifdef INET6 1925 case O_IP6_SRC: 1926 match = is_ipv6 && 1927 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 1928 &((ipfw_insn_ip6 *)cmd)->addr6); 1929 break; 1930 1931 case O_IP6_DST: 1932 match = is_ipv6 && 1933 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 1934 &((ipfw_insn_ip6 *)cmd)->addr6); 1935 break; 1936 case O_IP6_SRC_MASK: 1937 case O_IP6_DST_MASK: 1938 if (is_ipv6) { 1939 int i = cmdlen - 1; 1940 struct in6_addr p; 1941 struct in6_addr *d = 1942 &((ipfw_insn_ip6 *)cmd)->addr6; 1943 1944 for (; !match && i > 0; d += 2, 1945 i -= F_INSN_SIZE(struct in6_addr) 1946 * 2) { 1947 p = (cmd->opcode == 1948 O_IP6_SRC_MASK) ? 1949 args->f_id.src_ip6: 1950 args->f_id.dst_ip6; 1951 APPLY_MASK(&p, &d[1]); 1952 match = 1953 IN6_ARE_ADDR_EQUAL(&d[0], 1954 &p); 1955 } 1956 } 1957 break; 1958 1959 case O_FLOW6ID: 1960 match = is_ipv6 && 1961 flow6id_match(args->f_id.flow_id6, 1962 (ipfw_insn_u32 *) cmd); 1963 break; 1964 1965 case O_EXT_HDR: 1966 match = is_ipv6 && 1967 (ext_hd & ((ipfw_insn *) cmd)->arg1); 1968 break; 1969 1970 case O_IP6: 1971 match = is_ipv6; 1972 break; 1973 #endif 1974 1975 case O_IP4: 1976 match = is_ipv4; 1977 break; 1978 1979 case O_TAG: { 1980 struct m_tag *mtag; 1981 uint32_t tag = TARG(cmd->arg1, tag); 1982 1983 /* Packet is already tagged with this tag? */ 1984 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 1985 1986 /* We have `untag' action when F_NOT flag is 1987 * present. And we must remove this mtag from 1988 * mbuf and reset `match' to zero (`match' will 1989 * be inversed later). 1990 * Otherwise we should allocate new mtag and 1991 * push it into mbuf. 1992 */ 1993 if (cmd->len & F_NOT) { /* `untag' action */ 1994 if (mtag != NULL) 1995 m_tag_delete(m, mtag); 1996 match = 0; 1997 } else { 1998 if (mtag == NULL) { 1999 mtag = m_tag_alloc( MTAG_IPFW, 2000 tag, 0, M_NOWAIT); 2001 if (mtag != NULL) 2002 m_tag_prepend(m, mtag); 2003 } 2004 match = 1; 2005 } 2006 break; 2007 } 2008 2009 case O_FIB: /* try match the specified fib */ 2010 if (args->f_id.fib == cmd->arg1) 2011 match = 1; 2012 break; 2013 2014 case O_SOCKARG: { 2015 #ifndef USERSPACE /* not supported in userspace */ 2016 struct inpcb *inp = args->inp; 2017 struct inpcbinfo *pi; 2018 2019 if (is_ipv6) /* XXX can we remove this ? */ 2020 break; 2021 2022 if (proto == IPPROTO_TCP) 2023 pi = &V_tcbinfo; 2024 else if (proto == IPPROTO_UDP) 2025 pi = &V_udbinfo; 2026 else 2027 break; 2028 2029 /* 2030 * XXXRW: so_user_cookie should almost 2031 * certainly be inp_user_cookie? 2032 */ 2033 2034 /* For incomming packet, lookup up the 2035 inpcb using the src/dest ip/port tuple */ 2036 if (inp == NULL) { 2037 inp = in_pcblookup(pi, 2038 src_ip, htons(src_port), 2039 dst_ip, htons(dst_port), 2040 INPLOOKUP_RLOCKPCB, NULL); 2041 if (inp != NULL) { 2042 tablearg = 2043 inp->inp_socket->so_user_cookie; 2044 if (tablearg) 2045 match = 1; 2046 INP_RUNLOCK(inp); 2047 } 2048 } else { 2049 if (inp->inp_socket) { 2050 tablearg = 2051 inp->inp_socket->so_user_cookie; 2052 if (tablearg) 2053 match = 1; 2054 } 2055 } 2056 #endif /* !USERSPACE */ 2057 break; 2058 } 2059 2060 case O_TAGGED: { 2061 struct m_tag *mtag; 2062 uint32_t tag = TARG(cmd->arg1, tag); 2063 2064 if (cmdlen == 1) { 2065 match = m_tag_locate(m, MTAG_IPFW, 2066 tag, NULL) != NULL; 2067 break; 2068 } 2069 2070 /* we have ranges */ 2071 for (mtag = m_tag_first(m); 2072 mtag != NULL && !match; 2073 mtag = m_tag_next(m, mtag)) { 2074 uint16_t *p; 2075 int i; 2076 2077 if (mtag->m_tag_cookie != MTAG_IPFW) 2078 continue; 2079 2080 p = ((ipfw_insn_u16 *)cmd)->ports; 2081 i = cmdlen - 1; 2082 for(; !match && i > 0; i--, p += 2) 2083 match = 2084 mtag->m_tag_id >= p[0] && 2085 mtag->m_tag_id <= p[1]; 2086 } 2087 break; 2088 } 2089 2090 /* 2091 * The second set of opcodes represents 'actions', 2092 * i.e. the terminal part of a rule once the packet 2093 * matches all previous patterns. 2094 * Typically there is only one action for each rule, 2095 * and the opcode is stored at the end of the rule 2096 * (but there are exceptions -- see below). 2097 * 2098 * In general, here we set retval and terminate the 2099 * outer loop (would be a 'break 3' in some language, 2100 * but we need to set l=0, done=1) 2101 * 2102 * Exceptions: 2103 * O_COUNT and O_SKIPTO actions: 2104 * instead of terminating, we jump to the next rule 2105 * (setting l=0), or to the SKIPTO target (setting 2106 * f/f_len, cmd and l as needed), respectively. 2107 * 2108 * O_TAG, O_LOG and O_ALTQ action parameters: 2109 * perform some action and set match = 1; 2110 * 2111 * O_LIMIT and O_KEEP_STATE: these opcodes are 2112 * not real 'actions', and are stored right 2113 * before the 'action' part of the rule. 2114 * These opcodes try to install an entry in the 2115 * state tables; if successful, we continue with 2116 * the next opcode (match=1; break;), otherwise 2117 * the packet must be dropped (set retval, 2118 * break loops with l=0, done=1) 2119 * 2120 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2121 * cause a lookup of the state table, and a jump 2122 * to the 'action' part of the parent rule 2123 * if an entry is found, or 2124 * (CHECK_STATE only) a jump to the next rule if 2125 * the entry is not found. 2126 * The result of the lookup is cached so that 2127 * further instances of these opcodes become NOPs. 2128 * The jump to the next rule is done by setting 2129 * l=0, cmdlen=0. 2130 */ 2131 case O_LIMIT: 2132 case O_KEEP_STATE: 2133 if (ipfw_install_state(chain, f, 2134 (ipfw_insn_limit *)cmd, args, tablearg)) { 2135 /* error or limit violation */ 2136 retval = IP_FW_DENY; 2137 l = 0; /* exit inner loop */ 2138 done = 1; /* exit outer loop */ 2139 } 2140 match = 1; 2141 break; 2142 2143 case O_PROBE_STATE: 2144 case O_CHECK_STATE: 2145 /* 2146 * dynamic rules are checked at the first 2147 * keep-state or check-state occurrence, 2148 * with the result being stored in dyn_dir. 2149 * The compiler introduces a PROBE_STATE 2150 * instruction for us when we have a 2151 * KEEP_STATE (because PROBE_STATE needs 2152 * to be run first). 2153 */ 2154 if (dyn_dir == MATCH_UNKNOWN && 2155 (q = ipfw_lookup_dyn_rule(&args->f_id, 2156 &dyn_dir, proto == IPPROTO_TCP ? 2157 TCP(ulp) : NULL)) 2158 != NULL) { 2159 /* 2160 * Found dynamic entry, update stats 2161 * and jump to the 'action' part of 2162 * the parent rule by setting 2163 * f, cmd, l and clearing cmdlen. 2164 */ 2165 IPFW_INC_DYN_COUNTER(q, pktlen); 2166 /* XXX we would like to have f_pos 2167 * readily accessible in the dynamic 2168 * rule, instead of having to 2169 * lookup q->rule. 2170 */ 2171 f = q->rule; 2172 f_pos = ipfw_find_rule(chain, 2173 f->rulenum, f->id); 2174 cmd = ACTION_PTR(f); 2175 l = f->cmd_len - f->act_ofs; 2176 ipfw_dyn_unlock(q); 2177 cmdlen = 0; 2178 match = 1; 2179 break; 2180 } 2181 /* 2182 * Dynamic entry not found. If CHECK_STATE, 2183 * skip to next rule, if PROBE_STATE just 2184 * ignore and continue with next opcode. 2185 */ 2186 if (cmd->opcode == O_CHECK_STATE) 2187 l = 0; /* exit inner loop */ 2188 match = 1; 2189 break; 2190 2191 case O_ACCEPT: 2192 retval = 0; /* accept */ 2193 l = 0; /* exit inner loop */ 2194 done = 1; /* exit outer loop */ 2195 break; 2196 2197 case O_PIPE: 2198 case O_QUEUE: 2199 set_match(args, f_pos, chain); 2200 args->rule.info = TARG(cmd->arg1, pipe); 2201 if (cmd->opcode == O_PIPE) 2202 args->rule.info |= IPFW_IS_PIPE; 2203 if (V_fw_one_pass) 2204 args->rule.info |= IPFW_ONEPASS; 2205 retval = IP_FW_DUMMYNET; 2206 l = 0; /* exit inner loop */ 2207 done = 1; /* exit outer loop */ 2208 break; 2209 2210 case O_DIVERT: 2211 case O_TEE: 2212 if (args->eh) /* not on layer 2 */ 2213 break; 2214 /* otherwise this is terminal */ 2215 l = 0; /* exit inner loop */ 2216 done = 1; /* exit outer loop */ 2217 retval = (cmd->opcode == O_DIVERT) ? 2218 IP_FW_DIVERT : IP_FW_TEE; 2219 set_match(args, f_pos, chain); 2220 args->rule.info = TARG(cmd->arg1, divert); 2221 break; 2222 2223 case O_COUNT: 2224 IPFW_INC_RULE_COUNTER(f, pktlen); 2225 l = 0; /* exit inner loop */ 2226 break; 2227 2228 case O_SKIPTO: 2229 IPFW_INC_RULE_COUNTER(f, pktlen); 2230 f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0); 2231 /* 2232 * Skip disabled rules, and re-enter 2233 * the inner loop with the correct 2234 * f_pos, f, l and cmd. 2235 * Also clear cmdlen and skip_or 2236 */ 2237 for (; f_pos < chain->n_rules - 1 && 2238 (V_set_disable & 2239 (1 << chain->map[f_pos]->set)); 2240 f_pos++) 2241 ; 2242 /* Re-enter the inner loop at the skipto rule. */ 2243 f = chain->map[f_pos]; 2244 l = f->cmd_len; 2245 cmd = f->cmd; 2246 match = 1; 2247 cmdlen = 0; 2248 skip_or = 0; 2249 continue; 2250 break; /* not reached */ 2251 2252 case O_CALLRETURN: { 2253 /* 2254 * Implementation of `subroutine' call/return, 2255 * in the stack carried in an mbuf tag. This 2256 * is different from `skipto' in that any call 2257 * address is possible (`skipto' must prevent 2258 * backward jumps to avoid endless loops). 2259 * We have `return' action when F_NOT flag is 2260 * present. The `m_tag_id' field is used as 2261 * stack pointer. 2262 */ 2263 struct m_tag *mtag; 2264 uint16_t jmpto, *stack; 2265 2266 #define IS_CALL ((cmd->len & F_NOT) == 0) 2267 #define IS_RETURN ((cmd->len & F_NOT) != 0) 2268 /* 2269 * Hand-rolled version of m_tag_locate() with 2270 * wildcard `type'. 2271 * If not already tagged, allocate new tag. 2272 */ 2273 mtag = m_tag_first(m); 2274 while (mtag != NULL) { 2275 if (mtag->m_tag_cookie == 2276 MTAG_IPFW_CALL) 2277 break; 2278 mtag = m_tag_next(m, mtag); 2279 } 2280 if (mtag == NULL && IS_CALL) { 2281 mtag = m_tag_alloc(MTAG_IPFW_CALL, 0, 2282 IPFW_CALLSTACK_SIZE * 2283 sizeof(uint16_t), M_NOWAIT); 2284 if (mtag != NULL) 2285 m_tag_prepend(m, mtag); 2286 } 2287 2288 /* 2289 * On error both `call' and `return' just 2290 * continue with next rule. 2291 */ 2292 if (IS_RETURN && (mtag == NULL || 2293 mtag->m_tag_id == 0)) { 2294 l = 0; /* exit inner loop */ 2295 break; 2296 } 2297 if (IS_CALL && (mtag == NULL || 2298 mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) { 2299 printf("ipfw: call stack error, " 2300 "go to next rule\n"); 2301 l = 0; /* exit inner loop */ 2302 break; 2303 } 2304 2305 IPFW_INC_RULE_COUNTER(f, pktlen); 2306 stack = (uint16_t *)(mtag + 1); 2307 2308 /* 2309 * The `call' action may use cached f_pos 2310 * (in f->next_rule), whose version is written 2311 * in f->next_rule. 2312 * The `return' action, however, doesn't have 2313 * fixed jump address in cmd->arg1 and can't use 2314 * cache. 2315 */ 2316 if (IS_CALL) { 2317 stack[mtag->m_tag_id] = f->rulenum; 2318 mtag->m_tag_id++; 2319 f_pos = JUMP(chain, f, cmd->arg1, 2320 tablearg, 1); 2321 } else { /* `return' action */ 2322 mtag->m_tag_id--; 2323 jmpto = stack[mtag->m_tag_id] + 1; 2324 f_pos = ipfw_find_rule(chain, jmpto, 0); 2325 } 2326 2327 /* 2328 * Skip disabled rules, and re-enter 2329 * the inner loop with the correct 2330 * f_pos, f, l and cmd. 2331 * Also clear cmdlen and skip_or 2332 */ 2333 for (; f_pos < chain->n_rules - 1 && 2334 (V_set_disable & 2335 (1 << chain->map[f_pos]->set)); f_pos++) 2336 ; 2337 /* Re-enter the inner loop at the dest rule. */ 2338 f = chain->map[f_pos]; 2339 l = f->cmd_len; 2340 cmd = f->cmd; 2341 cmdlen = 0; 2342 skip_or = 0; 2343 continue; 2344 break; /* NOTREACHED */ 2345 } 2346 #undef IS_CALL 2347 #undef IS_RETURN 2348 2349 case O_REJECT: 2350 /* 2351 * Drop the packet and send a reject notice 2352 * if the packet is not ICMP (or is an ICMP 2353 * query), and it is not multicast/broadcast. 2354 */ 2355 if (hlen > 0 && is_ipv4 && offset == 0 && 2356 (proto != IPPROTO_ICMP || 2357 is_icmp_query(ICMP(ulp))) && 2358 !(m->m_flags & (M_BCAST|M_MCAST)) && 2359 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 2360 send_reject(args, cmd->arg1, iplen, ip); 2361 m = args->m; 2362 } 2363 /* FALLTHROUGH */ 2364 #ifdef INET6 2365 case O_UNREACH6: 2366 if (hlen > 0 && is_ipv6 && 2367 ((offset & IP6F_OFF_MASK) == 0) && 2368 (proto != IPPROTO_ICMPV6 || 2369 (is_icmp6_query(icmp6_type) == 1)) && 2370 !(m->m_flags & (M_BCAST|M_MCAST)) && 2371 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 2372 send_reject6( 2373 args, cmd->arg1, hlen, 2374 (struct ip6_hdr *)ip); 2375 m = args->m; 2376 } 2377 /* FALLTHROUGH */ 2378 #endif 2379 case O_DENY: 2380 retval = IP_FW_DENY; 2381 l = 0; /* exit inner loop */ 2382 done = 1; /* exit outer loop */ 2383 break; 2384 2385 case O_FORWARD_IP: 2386 if (args->eh) /* not valid on layer2 pkts */ 2387 break; 2388 if (q == NULL || q->rule != f || 2389 dyn_dir == MATCH_FORWARD) { 2390 struct sockaddr_in *sa; 2391 2392 sa = &(((ipfw_insn_sa *)cmd)->sa); 2393 if (sa->sin_addr.s_addr == INADDR_ANY) { 2394 #ifdef INET6 2395 /* 2396 * We use O_FORWARD_IP opcode for 2397 * fwd rule with tablearg, but tables 2398 * now support IPv6 addresses. And 2399 * when we are inspecting IPv6 packet, 2400 * we can use nh6 field from 2401 * table_value as next_hop6 address. 2402 */ 2403 if (is_ipv6) { 2404 struct sockaddr_in6 *sa6; 2405 2406 sa6 = args->next_hop6 = 2407 &args->hopstore6; 2408 sa6->sin6_family = AF_INET6; 2409 sa6->sin6_len = sizeof(*sa6); 2410 sa6->sin6_addr = TARG_VAL( 2411 chain, tablearg, nh6); 2412 /* 2413 * Set sin6_scope_id only for 2414 * link-local unicast addresses. 2415 */ 2416 if (IN6_IS_ADDR_LINKLOCAL( 2417 &sa6->sin6_addr)) 2418 sa6->sin6_scope_id = 2419 TARG_VAL(chain, 2420 tablearg, 2421 zoneid); 2422 } else 2423 #endif 2424 { 2425 sa = args->next_hop = 2426 &args->hopstore; 2427 sa->sin_family = AF_INET; 2428 sa->sin_len = sizeof(*sa); 2429 sa->sin_addr.s_addr = htonl( 2430 TARG_VAL(chain, tablearg, 2431 nh4)); 2432 } 2433 } else { 2434 args->next_hop = sa; 2435 } 2436 } 2437 retval = IP_FW_PASS; 2438 l = 0; /* exit inner loop */ 2439 done = 1; /* exit outer loop */ 2440 break; 2441 2442 #ifdef INET6 2443 case O_FORWARD_IP6: 2444 if (args->eh) /* not valid on layer2 pkts */ 2445 break; 2446 if (q == NULL || q->rule != f || 2447 dyn_dir == MATCH_FORWARD) { 2448 struct sockaddr_in6 *sin6; 2449 2450 sin6 = &(((ipfw_insn_sa6 *)cmd)->sa); 2451 args->next_hop6 = sin6; 2452 } 2453 retval = IP_FW_PASS; 2454 l = 0; /* exit inner loop */ 2455 done = 1; /* exit outer loop */ 2456 break; 2457 #endif 2458 2459 case O_NETGRAPH: 2460 case O_NGTEE: 2461 set_match(args, f_pos, chain); 2462 args->rule.info = TARG(cmd->arg1, netgraph); 2463 if (V_fw_one_pass) 2464 args->rule.info |= IPFW_ONEPASS; 2465 retval = (cmd->opcode == O_NETGRAPH) ? 2466 IP_FW_NETGRAPH : IP_FW_NGTEE; 2467 l = 0; /* exit inner loop */ 2468 done = 1; /* exit outer loop */ 2469 break; 2470 2471 case O_SETFIB: { 2472 uint32_t fib; 2473 2474 IPFW_INC_RULE_COUNTER(f, pktlen); 2475 fib = TARG(cmd->arg1, fib) & 0x7FFFF; 2476 if (fib >= rt_numfibs) 2477 fib = 0; 2478 M_SETFIB(m, fib); 2479 args->f_id.fib = fib; 2480 l = 0; /* exit inner loop */ 2481 break; 2482 } 2483 2484 case O_SETDSCP: { 2485 uint16_t code; 2486 2487 code = TARG(cmd->arg1, dscp) & 0x3F; 2488 l = 0; /* exit inner loop */ 2489 if (is_ipv4) { 2490 uint16_t old; 2491 2492 old = *(uint16_t *)ip; 2493 ip->ip_tos = (code << 2) | 2494 (ip->ip_tos & 0x03); 2495 ip->ip_sum = cksum_adjust(ip->ip_sum, 2496 old, *(uint16_t *)ip); 2497 } else if (is_ipv6) { 2498 uint8_t *v; 2499 2500 v = &((struct ip6_hdr *)ip)->ip6_vfc; 2501 *v = (*v & 0xF0) | (code >> 2); 2502 v++; 2503 *v = (*v & 0x3F) | ((code & 0x03) << 6); 2504 } else 2505 break; 2506 2507 IPFW_INC_RULE_COUNTER(f, pktlen); 2508 break; 2509 } 2510 2511 case O_NAT: 2512 l = 0; /* exit inner loop */ 2513 done = 1; /* exit outer loop */ 2514 if (!IPFW_NAT_LOADED) { 2515 retval = IP_FW_DENY; 2516 break; 2517 } 2518 2519 struct cfg_nat *t; 2520 int nat_id; 2521 2522 set_match(args, f_pos, chain); 2523 /* Check if this is 'global' nat rule */ 2524 if (cmd->arg1 == 0) { 2525 retval = ipfw_nat_ptr(args, NULL, m); 2526 break; 2527 } 2528 t = ((ipfw_insn_nat *)cmd)->nat; 2529 if (t == NULL) { 2530 nat_id = TARG(cmd->arg1, nat); 2531 t = (*lookup_nat_ptr)(&chain->nat, nat_id); 2532 2533 if (t == NULL) { 2534 retval = IP_FW_DENY; 2535 break; 2536 } 2537 if (cmd->arg1 != IP_FW_TARG) 2538 ((ipfw_insn_nat *)cmd)->nat = t; 2539 } 2540 retval = ipfw_nat_ptr(args, t, m); 2541 break; 2542 2543 case O_REASS: { 2544 int ip_off; 2545 2546 IPFW_INC_RULE_COUNTER(f, pktlen); 2547 l = 0; /* in any case exit inner loop */ 2548 ip_off = ntohs(ip->ip_off); 2549 2550 /* if not fragmented, go to next rule */ 2551 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0) 2552 break; 2553 2554 args->m = m = ip_reass(m); 2555 2556 /* 2557 * do IP header checksum fixup. 2558 */ 2559 if (m == NULL) { /* fragment got swallowed */ 2560 retval = IP_FW_DENY; 2561 } else { /* good, packet complete */ 2562 int hlen; 2563 2564 ip = mtod(m, struct ip *); 2565 hlen = ip->ip_hl << 2; 2566 ip->ip_sum = 0; 2567 if (hlen == sizeof(struct ip)) 2568 ip->ip_sum = in_cksum_hdr(ip); 2569 else 2570 ip->ip_sum = in_cksum(m, hlen); 2571 retval = IP_FW_REASS; 2572 set_match(args, f_pos, chain); 2573 } 2574 done = 1; /* exit outer loop */ 2575 break; 2576 } 2577 2578 default: 2579 panic("-- unknown opcode %d\n", cmd->opcode); 2580 } /* end of switch() on opcodes */ 2581 /* 2582 * if we get here with l=0, then match is irrelevant. 2583 */ 2584 2585 if (cmd->len & F_NOT) 2586 match = !match; 2587 2588 if (match) { 2589 if (cmd->len & F_OR) 2590 skip_or = 1; 2591 } else { 2592 if (!(cmd->len & F_OR)) /* not an OR block, */ 2593 break; /* try next rule */ 2594 } 2595 2596 } /* end of inner loop, scan opcodes */ 2597 #undef PULLUP_LEN 2598 2599 if (done) 2600 break; 2601 2602 /* next_rule:; */ /* try next rule */ 2603 2604 } /* end of outer for, scan rules */ 2605 2606 if (done) { 2607 struct ip_fw *rule = chain->map[f_pos]; 2608 /* Update statistics */ 2609 IPFW_INC_RULE_COUNTER(rule, pktlen); 2610 } else { 2611 retval = IP_FW_DENY; 2612 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 2613 } 2614 IPFW_PF_RUNLOCK(chain); 2615 #ifdef __FreeBSD__ 2616 if (ucred_cache != NULL) 2617 crfree(ucred_cache); 2618 #endif 2619 return (retval); 2620 2621 pullup_failed: 2622 if (V_fw_verbose) 2623 printf("ipfw: pullup failed\n"); 2624 return (IP_FW_DENY); 2625 } 2626 2627 /* 2628 * Set maximum number of tables that can be used in given VNET ipfw instance. 2629 */ 2630 #ifdef SYSCTL_NODE 2631 static int 2632 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS) 2633 { 2634 int error; 2635 unsigned int ntables; 2636 2637 ntables = V_fw_tables_max; 2638 2639 error = sysctl_handle_int(oidp, &ntables, 0, req); 2640 /* Read operation or some error */ 2641 if ((error != 0) || (req->newptr == NULL)) 2642 return (error); 2643 2644 return (ipfw_resize_tables(&V_layer3_chain, ntables)); 2645 } 2646 2647 /* 2648 * Switches table namespace between global and per-set. 2649 */ 2650 static int 2651 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS) 2652 { 2653 int error; 2654 unsigned int sets; 2655 2656 sets = V_fw_tables_sets; 2657 2658 error = sysctl_handle_int(oidp, &sets, 0, req); 2659 /* Read operation or some error */ 2660 if ((error != 0) || (req->newptr == NULL)) 2661 return (error); 2662 2663 return (ipfw_switch_tables_namespace(&V_layer3_chain, sets)); 2664 } 2665 #endif 2666 2667 /* 2668 * Module and VNET glue 2669 */ 2670 2671 /* 2672 * Stuff that must be initialised only on boot or module load 2673 */ 2674 static int 2675 ipfw_init(void) 2676 { 2677 int error = 0; 2678 2679 /* 2680 * Only print out this stuff the first time around, 2681 * when called from the sysinit code. 2682 */ 2683 printf("ipfw2 " 2684 #ifdef INET6 2685 "(+ipv6) " 2686 #endif 2687 "initialized, divert %s, nat %s, " 2688 "default to %s, logging ", 2689 #ifdef IPDIVERT 2690 "enabled", 2691 #else 2692 "loadable", 2693 #endif 2694 #ifdef IPFIREWALL_NAT 2695 "enabled", 2696 #else 2697 "loadable", 2698 #endif 2699 default_to_accept ? "accept" : "deny"); 2700 2701 /* 2702 * Note: V_xxx variables can be accessed here but the vnet specific 2703 * initializer may not have been called yet for the VIMAGE case. 2704 * Tuneables will have been processed. We will print out values for 2705 * the default vnet. 2706 * XXX This should all be rationalized AFTER 8.0 2707 */ 2708 if (V_fw_verbose == 0) 2709 printf("disabled\n"); 2710 else if (V_verbose_limit == 0) 2711 printf("unlimited\n"); 2712 else 2713 printf("limited to %d packets/entry by default\n", 2714 V_verbose_limit); 2715 2716 /* Check user-supplied table count for validness */ 2717 if (default_fw_tables > IPFW_TABLES_MAX) 2718 default_fw_tables = IPFW_TABLES_MAX; 2719 2720 ipfw_init_sopt_handler(); 2721 ipfw_log_bpf(1); /* init */ 2722 ipfw_iface_init(); 2723 return (error); 2724 } 2725 2726 /* 2727 * Called for the removal of the last instance only on module unload. 2728 */ 2729 static void 2730 ipfw_destroy(void) 2731 { 2732 2733 ipfw_iface_destroy(); 2734 ipfw_log_bpf(0); /* uninit */ 2735 ipfw_destroy_sopt_handler(); 2736 printf("IP firewall unloaded\n"); 2737 } 2738 2739 /* 2740 * Stuff that must be initialized for every instance 2741 * (including the first of course). 2742 */ 2743 static int 2744 vnet_ipfw_init(const void *unused) 2745 { 2746 int error, first; 2747 struct ip_fw *rule = NULL; 2748 struct ip_fw_chain *chain; 2749 2750 chain = &V_layer3_chain; 2751 2752 first = IS_DEFAULT_VNET(curvnet) ? 1 : 0; 2753 2754 /* First set up some values that are compile time options */ 2755 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 2756 V_fw_deny_unknown_exthdrs = 1; 2757 #ifdef IPFIREWALL_VERBOSE 2758 V_fw_verbose = 1; 2759 #endif 2760 #ifdef IPFIREWALL_VERBOSE_LIMIT 2761 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 2762 #endif 2763 #ifdef IPFIREWALL_NAT 2764 LIST_INIT(&chain->nat); 2765 #endif 2766 2767 /* Init shared services hash table */ 2768 ipfw_init_srv(chain); 2769 2770 ipfw_init_obj_rewriter(); 2771 ipfw_init_counters(); 2772 /* insert the default rule and create the initial map */ 2773 chain->n_rules = 1; 2774 chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_WAITOK | M_ZERO); 2775 rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw)); 2776 2777 /* Set initial number of tables */ 2778 V_fw_tables_max = default_fw_tables; 2779 error = ipfw_init_tables(chain, first); 2780 if (error) { 2781 printf("ipfw2: setting up tables failed\n"); 2782 free(chain->map, M_IPFW); 2783 free(rule, M_IPFW); 2784 return (ENOSPC); 2785 } 2786 2787 /* fill and insert the default rule */ 2788 rule->act_ofs = 0; 2789 rule->rulenum = IPFW_DEFAULT_RULE; 2790 rule->cmd_len = 1; 2791 rule->set = RESVD_SET; 2792 rule->cmd[0].len = 1; 2793 rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; 2794 chain->default_rule = chain->map[0] = rule; 2795 chain->id = rule->id = 1; 2796 /* Pre-calculate rules length for legacy dump format */ 2797 chain->static_len = sizeof(struct ip_fw_rule0); 2798 2799 IPFW_LOCK_INIT(chain); 2800 ipfw_dyn_init(chain); 2801 #ifdef LINEAR_SKIPTO 2802 ipfw_init_skipto_cache(chain); 2803 #endif 2804 2805 /* First set up some values that are compile time options */ 2806 V_ipfw_vnet_ready = 1; /* Open for business */ 2807 2808 /* 2809 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6. 2810 * Even if the latter two fail we still keep the module alive 2811 * because the sockopt and layer2 paths are still useful. 2812 * ipfw[6]_hook return 0 on success, ENOENT on failure, 2813 * so we can ignore the exact return value and just set a flag. 2814 * 2815 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so 2816 * changes in the underlying (per-vnet) variables trigger 2817 * immediate hook()/unhook() calls. 2818 * In layer2 we have the same behaviour, except that V_ether_ipfw 2819 * is checked on each packet because there are no pfil hooks. 2820 */ 2821 V_ip_fw_ctl_ptr = ipfw_ctl3; 2822 error = ipfw_attach_hooks(1); 2823 return (error); 2824 } 2825 2826 /* 2827 * Called for the removal of each instance. 2828 */ 2829 static int 2830 vnet_ipfw_uninit(const void *unused) 2831 { 2832 struct ip_fw *reap; 2833 struct ip_fw_chain *chain = &V_layer3_chain; 2834 int i, last; 2835 2836 V_ipfw_vnet_ready = 0; /* tell new callers to go away */ 2837 /* 2838 * disconnect from ipv4, ipv6, layer2 and sockopt. 2839 * Then grab, release and grab again the WLOCK so we make 2840 * sure the update is propagated and nobody will be in. 2841 */ 2842 (void)ipfw_attach_hooks(0 /* detach */); 2843 V_ip_fw_ctl_ptr = NULL; 2844 2845 last = IS_DEFAULT_VNET(curvnet) ? 1 : 0; 2846 2847 IPFW_UH_WLOCK(chain); 2848 IPFW_UH_WUNLOCK(chain); 2849 IPFW_UH_WLOCK(chain); 2850 2851 IPFW_WLOCK(chain); 2852 ipfw_dyn_uninit(0); /* run the callout_drain */ 2853 IPFW_WUNLOCK(chain); 2854 2855 reap = NULL; 2856 IPFW_WLOCK(chain); 2857 for (i = 0; i < chain->n_rules; i++) 2858 ipfw_reap_add(chain, &reap, chain->map[i]); 2859 free(chain->map, M_IPFW); 2860 #ifdef LINEAR_SKIPTO 2861 ipfw_destroy_skipto_cache(chain); 2862 #endif 2863 IPFW_WUNLOCK(chain); 2864 IPFW_UH_WUNLOCK(chain); 2865 ipfw_destroy_tables(chain, last); 2866 if (reap != NULL) 2867 ipfw_reap_rules(reap); 2868 vnet_ipfw_iface_destroy(chain); 2869 ipfw_destroy_srv(chain); 2870 IPFW_LOCK_DESTROY(chain); 2871 ipfw_dyn_uninit(1); /* free the remaining parts */ 2872 ipfw_destroy_counters(); 2873 ipfw_destroy_obj_rewriter(); 2874 return (0); 2875 } 2876 2877 /* 2878 * Module event handler. 2879 * In general we have the choice of handling most of these events by the 2880 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to 2881 * use the SYSINIT handlers as they are more capable of expressing the 2882 * flow of control during module and vnet operations, so this is just 2883 * a skeleton. Note there is no SYSINIT equivalent of the module 2884 * SHUTDOWN handler, but we don't have anything to do in that case anyhow. 2885 */ 2886 static int 2887 ipfw_modevent(module_t mod, int type, void *unused) 2888 { 2889 int err = 0; 2890 2891 switch (type) { 2892 case MOD_LOAD: 2893 /* Called once at module load or 2894 * system boot if compiled in. */ 2895 break; 2896 case MOD_QUIESCE: 2897 /* Called before unload. May veto unloading. */ 2898 break; 2899 case MOD_UNLOAD: 2900 /* Called during unload. */ 2901 break; 2902 case MOD_SHUTDOWN: 2903 /* Called during system shutdown. */ 2904 break; 2905 default: 2906 err = EOPNOTSUPP; 2907 break; 2908 } 2909 return err; 2910 } 2911 2912 static moduledata_t ipfwmod = { 2913 "ipfw", 2914 ipfw_modevent, 2915 0 2916 }; 2917 2918 /* Define startup order. */ 2919 #define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_IFATTACHDOMAIN 2920 #define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */ 2921 #define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */ 2922 #define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */ 2923 2924 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER); 2925 FEATURE(ipfw_ctl3, "ipfw new sockopt calls"); 2926 MODULE_VERSION(ipfw, 3); 2927 /* should declare some dependencies here */ 2928 2929 /* 2930 * Starting up. Done in order after ipfwmod() has been called. 2931 * VNET_SYSINIT is also called for each existing vnet and each new vnet. 2932 */ 2933 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 2934 ipfw_init, NULL); 2935 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 2936 vnet_ipfw_init, NULL); 2937 2938 /* 2939 * Closing up shop. These are done in REVERSE ORDER, but still 2940 * after ipfwmod() has been called. Not called on reboot. 2941 * VNET_SYSUNINIT is also called for each exiting vnet as it exits. 2942 * or when the module is unloaded. 2943 */ 2944 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 2945 ipfw_destroy, NULL); 2946 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 2947 vnet_ipfw_uninit, NULL); 2948 /* end of file */ 2949