1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 25 * SUCH DAMAGE. 26 */ 27 28 #include <sys/cdefs.h> 29 __FBSDID("$FreeBSD$"); 30 31 /* 32 * The FreeBSD IP packet firewall, main file 33 */ 34 35 #include "opt_ipfw.h" 36 #include "opt_ipdivert.h" 37 #include "opt_inet.h" 38 #ifndef INET 39 #error "IPFIREWALL requires INET" 40 #endif /* INET */ 41 #include "opt_inet6.h" 42 43 #include <sys/param.h> 44 #include <sys/systm.h> 45 #include <sys/condvar.h> 46 #include <sys/counter.h> 47 #include <sys/eventhandler.h> 48 #include <sys/malloc.h> 49 #include <sys/mbuf.h> 50 #include <sys/kernel.h> 51 #include <sys/lock.h> 52 #include <sys/jail.h> 53 #include <sys/module.h> 54 #include <sys/priv.h> 55 #include <sys/proc.h> 56 #include <sys/rwlock.h> 57 #include <sys/rmlock.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/if_var.h> 66 #include <net/route.h> 67 #include <net/route/nhop.h> 68 #include <net/pfil.h> 69 #include <net/vnet.h> 70 71 #include <netpfil/pf/pf_mtag.h> 72 73 #include <netinet/in.h> 74 #include <netinet/in_var.h> 75 #include <netinet/in_pcb.h> 76 #include <netinet/ip.h> 77 #include <netinet/ip_var.h> 78 #include <netinet/ip_icmp.h> 79 #include <netinet/ip_fw.h> 80 #include <netinet/ip_carp.h> 81 #include <netinet/pim.h> 82 #include <netinet/tcp_var.h> 83 #include <netinet/udp.h> 84 #include <netinet/udp_var.h> 85 #include <netinet/sctp.h> 86 #include <netinet/sctp_crc32.h> 87 #include <netinet/sctp_header.h> 88 89 #include <netinet/ip6.h> 90 #include <netinet/icmp6.h> 91 #include <netinet/in_fib.h> 92 #ifdef INET6 93 #include <netinet6/in6_fib.h> 94 #include <netinet6/in6_pcb.h> 95 #include <netinet6/scope6_var.h> 96 #include <netinet6/ip6_var.h> 97 #endif 98 99 #include <net/if_gre.h> /* for struct grehdr */ 100 101 #include <netpfil/ipfw/ip_fw_private.h> 102 103 #include <machine/in_cksum.h> /* XXX for in_cksum */ 104 105 #ifdef MAC 106 #include <security/mac/mac_framework.h> 107 #endif 108 109 /* 110 * static variables followed by global ones. 111 * All ipfw global variables are here. 112 */ 113 114 VNET_DEFINE_STATIC(int, fw_deny_unknown_exthdrs); 115 #define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs) 116 117 VNET_DEFINE_STATIC(int, fw_permit_single_frag6) = 1; 118 #define V_fw_permit_single_frag6 VNET(fw_permit_single_frag6) 119 120 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 121 static int default_to_accept = 1; 122 #else 123 static int default_to_accept; 124 #endif 125 126 VNET_DEFINE(int, autoinc_step); 127 VNET_DEFINE(int, fw_one_pass) = 1; 128 129 VNET_DEFINE(unsigned int, fw_tables_max); 130 VNET_DEFINE(unsigned int, fw_tables_sets) = 0; /* Don't use set-aware tables */ 131 /* Use 128 tables by default */ 132 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT; 133 134 #ifndef LINEAR_SKIPTO 135 static int jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num, 136 int tablearg, int jump_backwards); 137 #define JUMP(ch, f, num, targ, back) jump_fast(ch, f, num, targ, back) 138 #else 139 static int jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num, 140 int tablearg, int jump_backwards); 141 #define JUMP(ch, f, num, targ, back) jump_linear(ch, f, num, targ, back) 142 #endif 143 144 /* 145 * Each rule belongs to one of 32 different sets (0..31). 146 * The variable set_disable contains one bit per set. 147 * If the bit is set, all rules in the corresponding set 148 * are disabled. Set RESVD_SET(31) is reserved for the default rule 149 * and rules that are not deleted by the flush command, 150 * and CANNOT be disabled. 151 * Rules in set RESVD_SET can only be deleted individually. 152 */ 153 VNET_DEFINE(u_int32_t, set_disable); 154 #define V_set_disable VNET(set_disable) 155 156 VNET_DEFINE(int, fw_verbose); 157 /* counter for ipfw_log(NULL...) */ 158 VNET_DEFINE(u_int64_t, norule_counter); 159 VNET_DEFINE(int, verbose_limit); 160 161 /* layer3_chain contains the list of rules for layer 3 */ 162 VNET_DEFINE(struct ip_fw_chain, layer3_chain); 163 164 /* ipfw_vnet_ready controls when we are open for business */ 165 VNET_DEFINE(int, ipfw_vnet_ready) = 0; 166 167 VNET_DEFINE(int, ipfw_nat_ready) = 0; 168 169 ipfw_nat_t *ipfw_nat_ptr = NULL; 170 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int); 171 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; 172 ipfw_nat_cfg_t *ipfw_nat_del_ptr; 173 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; 174 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; 175 176 #ifdef SYSCTL_NODE 177 uint32_t dummy_def = IPFW_DEFAULT_RULE; 178 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS); 179 static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS); 180 181 SYSBEGIN(f3) 182 183 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 184 "Firewall"); 185 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 186 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0, 187 "Only do a single pass through ipfw when using dummynet(4)"); 188 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, 189 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0, 190 "Rule number auto-increment step"); 191 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 192 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0, 193 "Log matches to ipfw rules"); 194 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, 195 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(verbose_limit), 0, 196 "Set upper limit of matches of ipfw rules logged"); 197 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, 198 &dummy_def, 0, 199 "The default/max possible rule number."); 200 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_max, 201 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, 202 0, 0, sysctl_ipfw_table_num, "IU", 203 "Maximum number of concurrently used tables"); 204 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets, 205 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, 206 0, 0, sysctl_ipfw_tables_sets, "IU", 207 "Use per-set namespace for tables"); 208 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN, 209 &default_to_accept, 0, 210 "Make the default rule accept all packets."); 211 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables); 212 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, 213 CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0, 214 "Number of static rules"); 215 216 #ifdef INET6 217 SYSCTL_DECL(_net_inet6_ip6); 218 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 219 "Firewall"); 220 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs, 221 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, 222 &VNET_NAME(fw_deny_unknown_exthdrs), 0, 223 "Deny packets with unknown IPv6 Extension Headers"); 224 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6, 225 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, 226 &VNET_NAME(fw_permit_single_frag6), 0, 227 "Permit single packet IPv6 fragments"); 228 #endif /* INET6 */ 229 230 SYSEND 231 232 #endif /* SYSCTL_NODE */ 233 234 235 /* 236 * Some macros used in the various matching options. 237 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 238 * Other macros just cast void * into the appropriate type 239 */ 240 #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 241 #define TCP(p) ((struct tcphdr *)(p)) 242 #define SCTP(p) ((struct sctphdr *)(p)) 243 #define UDP(p) ((struct udphdr *)(p)) 244 #define ICMP(p) ((struct icmphdr *)(p)) 245 #define ICMP6(p) ((struct icmp6_hdr *)(p)) 246 247 static __inline int 248 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 249 { 250 int type = icmp->icmp_type; 251 252 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 253 } 254 255 #define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 256 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 257 258 static int 259 is_icmp_query(struct icmphdr *icmp) 260 { 261 int type = icmp->icmp_type; 262 263 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 264 } 265 #undef TT 266 267 /* 268 * The following checks use two arrays of 8 or 16 bits to store the 269 * bits that we want set or clear, respectively. They are in the 270 * low and high half of cmd->arg1 or cmd->d[0]. 271 * 272 * We scan options and store the bits we find set. We succeed if 273 * 274 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 275 * 276 * The code is sometimes optimized not to store additional variables. 277 */ 278 279 static int 280 flags_match(ipfw_insn *cmd, u_int8_t bits) 281 { 282 u_char want_clear; 283 bits = ~bits; 284 285 if ( ((cmd->arg1 & 0xff) & bits) != 0) 286 return 0; /* some bits we want set were clear */ 287 want_clear = (cmd->arg1 >> 8) & 0xff; 288 if ( (want_clear & bits) != want_clear) 289 return 0; /* some bits we want clear were set */ 290 return 1; 291 } 292 293 static int 294 ipopts_match(struct ip *ip, ipfw_insn *cmd) 295 { 296 int optlen, bits = 0; 297 u_char *cp = (u_char *)(ip + 1); 298 int x = (ip->ip_hl << 2) - sizeof (struct ip); 299 300 for (; x > 0; x -= optlen, cp += optlen) { 301 int opt = cp[IPOPT_OPTVAL]; 302 303 if (opt == IPOPT_EOL) 304 break; 305 if (opt == IPOPT_NOP) 306 optlen = 1; 307 else { 308 optlen = cp[IPOPT_OLEN]; 309 if (optlen <= 0 || optlen > x) 310 return 0; /* invalid or truncated */ 311 } 312 switch (opt) { 313 314 default: 315 break; 316 317 case IPOPT_LSRR: 318 bits |= IP_FW_IPOPT_LSRR; 319 break; 320 321 case IPOPT_SSRR: 322 bits |= IP_FW_IPOPT_SSRR; 323 break; 324 325 case IPOPT_RR: 326 bits |= IP_FW_IPOPT_RR; 327 break; 328 329 case IPOPT_TS: 330 bits |= IP_FW_IPOPT_TS; 331 break; 332 } 333 } 334 return (flags_match(cmd, bits)); 335 } 336 337 /* 338 * Parse TCP options. The logic copied from tcp_dooptions(). 339 */ 340 static int 341 tcpopts_parse(const struct tcphdr *tcp, uint16_t *mss) 342 { 343 const u_char *cp = (const u_char *)(tcp + 1); 344 int optlen, bits = 0; 345 int cnt = (tcp->th_off << 2) - sizeof(struct tcphdr); 346 347 for (; cnt > 0; cnt -= optlen, cp += optlen) { 348 int opt = cp[0]; 349 if (opt == TCPOPT_EOL) 350 break; 351 if (opt == TCPOPT_NOP) 352 optlen = 1; 353 else { 354 if (cnt < 2) 355 break; 356 optlen = cp[1]; 357 if (optlen < 2 || optlen > cnt) 358 break; 359 } 360 361 switch (opt) { 362 default: 363 break; 364 365 case TCPOPT_MAXSEG: 366 if (optlen != TCPOLEN_MAXSEG) 367 break; 368 bits |= IP_FW_TCPOPT_MSS; 369 if (mss != NULL) 370 *mss = be16dec(cp + 2); 371 break; 372 373 case TCPOPT_WINDOW: 374 if (optlen == TCPOLEN_WINDOW) 375 bits |= IP_FW_TCPOPT_WINDOW; 376 break; 377 378 case TCPOPT_SACK_PERMITTED: 379 if (optlen == TCPOLEN_SACK_PERMITTED) 380 bits |= IP_FW_TCPOPT_SACK; 381 break; 382 383 case TCPOPT_SACK: 384 if (optlen > 2 && (optlen - 2) % TCPOLEN_SACK == 0) 385 bits |= IP_FW_TCPOPT_SACK; 386 break; 387 388 case TCPOPT_TIMESTAMP: 389 if (optlen == TCPOLEN_TIMESTAMP) 390 bits |= IP_FW_TCPOPT_TS; 391 break; 392 } 393 } 394 return (bits); 395 } 396 397 static int 398 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 399 { 400 401 return (flags_match(cmd, tcpopts_parse(tcp, NULL))); 402 } 403 404 static int 405 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain, 406 uint32_t *tablearg) 407 { 408 409 if (ifp == NULL) /* no iface with this packet, match fails */ 410 return (0); 411 412 /* Check by name or by IP address */ 413 if (cmd->name[0] != '\0') { /* match by name */ 414 if (cmd->name[0] == '\1') /* use tablearg to match */ 415 return ipfw_lookup_table(chain, cmd->p.kidx, 0, 416 &ifp->if_index, tablearg); 417 /* Check name */ 418 if (cmd->p.glob) { 419 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 420 return(1); 421 } else { 422 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 423 return(1); 424 } 425 } else { 426 #if !defined(USERSPACE) && defined(__FreeBSD__) /* and OSX too ? */ 427 struct ifaddr *ia; 428 429 NET_EPOCH_ASSERT(); 430 431 CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 432 if (ia->ifa_addr->sa_family != AF_INET) 433 continue; 434 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 435 (ia->ifa_addr))->sin_addr.s_addr) 436 return (1); /* match */ 437 } 438 #endif /* __FreeBSD__ */ 439 } 440 return(0); /* no match, fail ... */ 441 } 442 443 /* 444 * The verify_path function checks if a route to the src exists and 445 * if it is reachable via ifp (when provided). 446 * 447 * The 'verrevpath' option checks that the interface that an IP packet 448 * arrives on is the same interface that traffic destined for the 449 * packet's source address would be routed out of. 450 * The 'versrcreach' option just checks that the source address is 451 * reachable via any route (except default) in the routing table. 452 * These two are a measure to block forged packets. This is also 453 * commonly known as "anti-spoofing" or Unicast Reverse Path 454 * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 455 * is purposely reminiscent of the Cisco IOS command, 456 * 457 * ip verify unicast reverse-path 458 * ip verify unicast source reachable-via any 459 * 460 * which implements the same functionality. But note that the syntax 461 * is misleading, and the check may be performed on all IP packets 462 * whether unicast, multicast, or broadcast. 463 */ 464 static int 465 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) 466 { 467 #if defined(USERSPACE) || !defined(__FreeBSD__) 468 return 0; 469 #else 470 struct nhop_object *nh; 471 472 nh = fib4_lookup(fib, src, 0, NHR_NONE, 0); 473 if (nh == NULL) 474 return (0); 475 476 /* 477 * If ifp is provided, check for equality with rtentry. 478 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 479 * in order to pass packets injected back by if_simloop(): 480 * routing entry (via lo0) for our own address 481 * may exist, so we need to handle routing assymetry. 482 */ 483 if (ifp != NULL && ifp != nh->nh_aifp) 484 return (0); 485 486 /* if no ifp provided, check if rtentry is not default route */ 487 if (ifp == NULL && (nh->nh_flags & NHF_DEFAULT) != 0) 488 return (0); 489 490 /* or if this is a blackhole/reject route */ 491 if (ifp == NULL && (nh->nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0) 492 return (0); 493 494 /* found valid route */ 495 return 1; 496 #endif /* __FreeBSD__ */ 497 } 498 499 /* 500 * Generate an SCTP packet containing an ABORT chunk. The verification tag 501 * is given by vtag. The T-bit is set in the ABORT chunk if and only if 502 * reflected is not 0. 503 */ 504 505 static struct mbuf * 506 ipfw_send_abort(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t vtag, 507 int reflected) 508 { 509 struct mbuf *m; 510 struct ip *ip; 511 #ifdef INET6 512 struct ip6_hdr *ip6; 513 #endif 514 struct sctphdr *sctp; 515 struct sctp_chunkhdr *chunk; 516 u_int16_t hlen, plen, tlen; 517 518 MGETHDR(m, M_NOWAIT, MT_DATA); 519 if (m == NULL) 520 return (NULL); 521 522 M_SETFIB(m, id->fib); 523 #ifdef MAC 524 if (replyto != NULL) 525 mac_netinet_firewall_reply(replyto, m); 526 else 527 mac_netinet_firewall_send(m); 528 #else 529 (void)replyto; /* don't warn about unused arg */ 530 #endif 531 532 switch (id->addr_type) { 533 case 4: 534 hlen = sizeof(struct ip); 535 break; 536 #ifdef INET6 537 case 6: 538 hlen = sizeof(struct ip6_hdr); 539 break; 540 #endif 541 default: 542 /* XXX: log me?!? */ 543 FREE_PKT(m); 544 return (NULL); 545 } 546 plen = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); 547 tlen = hlen + plen; 548 m->m_data += max_linkhdr; 549 m->m_flags |= M_SKIP_FIREWALL; 550 m->m_pkthdr.len = m->m_len = tlen; 551 m->m_pkthdr.rcvif = NULL; 552 bzero(m->m_data, tlen); 553 554 switch (id->addr_type) { 555 case 4: 556 ip = mtod(m, struct ip *); 557 558 ip->ip_v = 4; 559 ip->ip_hl = sizeof(struct ip) >> 2; 560 ip->ip_tos = IPTOS_LOWDELAY; 561 ip->ip_len = htons(tlen); 562 ip->ip_id = htons(0); 563 ip->ip_off = htons(0); 564 ip->ip_ttl = V_ip_defttl; 565 ip->ip_p = IPPROTO_SCTP; 566 ip->ip_sum = 0; 567 ip->ip_src.s_addr = htonl(id->dst_ip); 568 ip->ip_dst.s_addr = htonl(id->src_ip); 569 570 sctp = (struct sctphdr *)(ip + 1); 571 break; 572 #ifdef INET6 573 case 6: 574 ip6 = mtod(m, struct ip6_hdr *); 575 576 ip6->ip6_vfc = IPV6_VERSION; 577 ip6->ip6_plen = htons(plen); 578 ip6->ip6_nxt = IPPROTO_SCTP; 579 ip6->ip6_hlim = IPV6_DEFHLIM; 580 ip6->ip6_src = id->dst_ip6; 581 ip6->ip6_dst = id->src_ip6; 582 583 sctp = (struct sctphdr *)(ip6 + 1); 584 break; 585 #endif 586 } 587 588 sctp->src_port = htons(id->dst_port); 589 sctp->dest_port = htons(id->src_port); 590 sctp->v_tag = htonl(vtag); 591 sctp->checksum = htonl(0); 592 593 chunk = (struct sctp_chunkhdr *)(sctp + 1); 594 chunk->chunk_type = SCTP_ABORT_ASSOCIATION; 595 chunk->chunk_flags = 0; 596 if (reflected != 0) { 597 chunk->chunk_flags |= SCTP_HAD_NO_TCB; 598 } 599 chunk->chunk_length = htons(sizeof(struct sctp_chunkhdr)); 600 601 sctp->checksum = sctp_calculate_cksum(m, hlen); 602 603 return (m); 604 } 605 606 /* 607 * Generate a TCP packet, containing either a RST or a keepalive. 608 * When flags & TH_RST, we are sending a RST packet, because of a 609 * "reset" action matched the packet. 610 * Otherwise we are sending a keepalive, and flags & TH_ 611 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 612 * so that MAC can label the reply appropriately. 613 */ 614 struct mbuf * 615 ipfw_send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 616 u_int32_t ack, int flags) 617 { 618 struct mbuf *m = NULL; /* stupid compiler */ 619 struct ip *h = NULL; /* stupid compiler */ 620 #ifdef INET6 621 struct ip6_hdr *h6 = NULL; 622 #endif 623 struct tcphdr *th = NULL; 624 int len, dir; 625 626 MGETHDR(m, M_NOWAIT, MT_DATA); 627 if (m == NULL) 628 return (NULL); 629 630 M_SETFIB(m, id->fib); 631 #ifdef MAC 632 if (replyto != NULL) 633 mac_netinet_firewall_reply(replyto, m); 634 else 635 mac_netinet_firewall_send(m); 636 #else 637 (void)replyto; /* don't warn about unused arg */ 638 #endif 639 640 switch (id->addr_type) { 641 case 4: 642 len = sizeof(struct ip) + sizeof(struct tcphdr); 643 break; 644 #ifdef INET6 645 case 6: 646 len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); 647 break; 648 #endif 649 default: 650 /* XXX: log me?!? */ 651 FREE_PKT(m); 652 return (NULL); 653 } 654 dir = ((flags & (TH_SYN | TH_RST)) == TH_SYN); 655 656 m->m_data += max_linkhdr; 657 m->m_flags |= M_SKIP_FIREWALL; 658 m->m_pkthdr.len = m->m_len = len; 659 m->m_pkthdr.rcvif = NULL; 660 bzero(m->m_data, len); 661 662 switch (id->addr_type) { 663 case 4: 664 h = mtod(m, struct ip *); 665 666 /* prepare for checksum */ 667 h->ip_p = IPPROTO_TCP; 668 h->ip_len = htons(sizeof(struct tcphdr)); 669 if (dir) { 670 h->ip_src.s_addr = htonl(id->src_ip); 671 h->ip_dst.s_addr = htonl(id->dst_ip); 672 } else { 673 h->ip_src.s_addr = htonl(id->dst_ip); 674 h->ip_dst.s_addr = htonl(id->src_ip); 675 } 676 677 th = (struct tcphdr *)(h + 1); 678 break; 679 #ifdef INET6 680 case 6: 681 h6 = mtod(m, struct ip6_hdr *); 682 683 /* prepare for checksum */ 684 h6->ip6_nxt = IPPROTO_TCP; 685 h6->ip6_plen = htons(sizeof(struct tcphdr)); 686 if (dir) { 687 h6->ip6_src = id->src_ip6; 688 h6->ip6_dst = id->dst_ip6; 689 } else { 690 h6->ip6_src = id->dst_ip6; 691 h6->ip6_dst = id->src_ip6; 692 } 693 694 th = (struct tcphdr *)(h6 + 1); 695 break; 696 #endif 697 } 698 699 if (dir) { 700 th->th_sport = htons(id->src_port); 701 th->th_dport = htons(id->dst_port); 702 } else { 703 th->th_sport = htons(id->dst_port); 704 th->th_dport = htons(id->src_port); 705 } 706 th->th_off = sizeof(struct tcphdr) >> 2; 707 708 if (flags & TH_RST) { 709 if (flags & TH_ACK) { 710 th->th_seq = htonl(ack); 711 th->th_flags = TH_RST; 712 } else { 713 if (flags & TH_SYN) 714 seq++; 715 th->th_ack = htonl(seq); 716 th->th_flags = TH_RST | TH_ACK; 717 } 718 } else { 719 /* 720 * Keepalive - use caller provided sequence numbers 721 */ 722 th->th_seq = htonl(seq); 723 th->th_ack = htonl(ack); 724 th->th_flags = TH_ACK; 725 } 726 727 switch (id->addr_type) { 728 case 4: 729 th->th_sum = in_cksum(m, len); 730 731 /* finish the ip header */ 732 h->ip_v = 4; 733 h->ip_hl = sizeof(*h) >> 2; 734 h->ip_tos = IPTOS_LOWDELAY; 735 h->ip_off = htons(0); 736 h->ip_len = htons(len); 737 h->ip_ttl = V_ip_defttl; 738 h->ip_sum = 0; 739 break; 740 #ifdef INET6 741 case 6: 742 th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(*h6), 743 sizeof(struct tcphdr)); 744 745 /* finish the ip6 header */ 746 h6->ip6_vfc |= IPV6_VERSION; 747 h6->ip6_hlim = IPV6_DEFHLIM; 748 break; 749 #endif 750 } 751 752 return (m); 753 } 754 755 #ifdef INET6 756 /* 757 * ipv6 specific rules here... 758 */ 759 static __inline int 760 icmp6type_match(int type, ipfw_insn_u32 *cmd) 761 { 762 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 763 } 764 765 static int 766 flow6id_match(int curr_flow, ipfw_insn_u32 *cmd) 767 { 768 int i; 769 for (i=0; i <= cmd->o.arg1; ++i) 770 if (curr_flow == cmd->d[i]) 771 return 1; 772 return 0; 773 } 774 775 /* support for IP6_*_ME opcodes */ 776 static const struct in6_addr lla_mask = {{{ 777 0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff, 778 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff 779 }}}; 780 781 static int 782 ipfw_localip6(struct in6_addr *in6) 783 { 784 struct rm_priotracker in6_ifa_tracker; 785 struct in6_ifaddr *ia; 786 787 if (IN6_IS_ADDR_MULTICAST(in6)) 788 return (0); 789 790 if (!IN6_IS_ADDR_LINKLOCAL(in6)) 791 return (in6_localip(in6)); 792 793 IN6_IFADDR_RLOCK(&in6_ifa_tracker); 794 CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) { 795 if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) 796 continue; 797 if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr, 798 in6, &lla_mask)) { 799 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); 800 return (1); 801 } 802 } 803 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); 804 return (0); 805 } 806 807 static int 808 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib) 809 { 810 struct nhop_object *nh; 811 812 if (IN6_IS_SCOPE_LINKLOCAL(src)) 813 return (1); 814 815 nh = fib6_lookup(fib, src, 0, NHR_NONE, 0); 816 if (nh == NULL) 817 return (0); 818 819 /* If ifp is provided, check for equality with route table. */ 820 if (ifp != NULL && ifp != nh->nh_aifp) 821 return (0); 822 823 /* if no ifp provided, check if rtentry is not default route */ 824 if (ifp == NULL && (nh->nh_flags & NHF_DEFAULT) != 0) 825 return (0); 826 827 /* or if this is a blackhole/reject route */ 828 if (ifp == NULL && (nh->nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0) 829 return (0); 830 831 /* found valid route */ 832 return 1; 833 } 834 835 static int 836 is_icmp6_query(int icmp6_type) 837 { 838 if ((icmp6_type <= ICMP6_MAXTYPE) && 839 (icmp6_type == ICMP6_ECHO_REQUEST || 840 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 841 icmp6_type == ICMP6_WRUREQUEST || 842 icmp6_type == ICMP6_FQDN_QUERY || 843 icmp6_type == ICMP6_NI_QUERY)) 844 return (1); 845 846 return (0); 847 } 848 849 static int 850 map_icmp_unreach(int code) 851 { 852 853 /* RFC 7915 p4.2 */ 854 switch (code) { 855 case ICMP_UNREACH_NET: 856 case ICMP_UNREACH_HOST: 857 case ICMP_UNREACH_SRCFAIL: 858 case ICMP_UNREACH_NET_UNKNOWN: 859 case ICMP_UNREACH_HOST_UNKNOWN: 860 case ICMP_UNREACH_TOSNET: 861 case ICMP_UNREACH_TOSHOST: 862 return (ICMP6_DST_UNREACH_NOROUTE); 863 case ICMP_UNREACH_PORT: 864 return (ICMP6_DST_UNREACH_NOPORT); 865 default: 866 /* 867 * Map the rest of codes into admit prohibited. 868 * XXX: unreach proto should be mapped into ICMPv6 869 * parameter problem, but we use only unreach type. 870 */ 871 return (ICMP6_DST_UNREACH_ADMIN); 872 } 873 } 874 875 static void 876 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 877 { 878 struct mbuf *m; 879 880 m = args->m; 881 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 882 struct tcphdr *tcp; 883 tcp = (struct tcphdr *)((char *)ip6 + hlen); 884 885 if ((tcp->th_flags & TH_RST) == 0) { 886 struct mbuf *m0; 887 m0 = ipfw_send_pkt(args->m, &(args->f_id), 888 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 889 tcp->th_flags | TH_RST); 890 if (m0 != NULL) 891 ip6_output(m0, NULL, NULL, 0, NULL, NULL, 892 NULL); 893 } 894 FREE_PKT(m); 895 } else if (code == ICMP6_UNREACH_ABORT && 896 args->f_id.proto == IPPROTO_SCTP) { 897 struct mbuf *m0; 898 struct sctphdr *sctp; 899 u_int32_t v_tag; 900 int reflected; 901 902 sctp = (struct sctphdr *)((char *)ip6 + hlen); 903 reflected = 1; 904 v_tag = ntohl(sctp->v_tag); 905 /* Investigate the first chunk header if available */ 906 if (m->m_len >= hlen + sizeof(struct sctphdr) + 907 sizeof(struct sctp_chunkhdr)) { 908 struct sctp_chunkhdr *chunk; 909 910 chunk = (struct sctp_chunkhdr *)(sctp + 1); 911 switch (chunk->chunk_type) { 912 case SCTP_INITIATION: 913 /* 914 * Packets containing an INIT chunk MUST have 915 * a zero v-tag. 916 */ 917 if (v_tag != 0) { 918 v_tag = 0; 919 break; 920 } 921 /* INIT chunk MUST NOT be bundled */ 922 if (m->m_pkthdr.len > 923 hlen + sizeof(struct sctphdr) + 924 ntohs(chunk->chunk_length) + 3) { 925 break; 926 } 927 /* Use the initiate tag if available */ 928 if ((m->m_len >= hlen + sizeof(struct sctphdr) + 929 sizeof(struct sctp_chunkhdr) + 930 offsetof(struct sctp_init, a_rwnd))) { 931 struct sctp_init *init; 932 933 init = (struct sctp_init *)(chunk + 1); 934 v_tag = ntohl(init->initiate_tag); 935 reflected = 0; 936 } 937 break; 938 case SCTP_ABORT_ASSOCIATION: 939 /* 940 * If the packet contains an ABORT chunk, don't 941 * reply. 942 * XXX: We should search through all chunks, 943 * but do not do that to avoid attacks. 944 */ 945 v_tag = 0; 946 break; 947 } 948 } 949 if (v_tag == 0) { 950 m0 = NULL; 951 } else { 952 m0 = ipfw_send_abort(args->m, &(args->f_id), v_tag, 953 reflected); 954 } 955 if (m0 != NULL) 956 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL); 957 FREE_PKT(m); 958 } else if (code != ICMP6_UNREACH_RST && code != ICMP6_UNREACH_ABORT) { 959 /* Send an ICMPv6 unreach. */ 960 #if 0 961 /* 962 * Unlike above, the mbufs need to line up with the ip6 hdr, 963 * as the contents are read. We need to m_adj() the 964 * needed amount. 965 * The mbuf will however be thrown away so we can adjust it. 966 * Remember we did an m_pullup on it already so we 967 * can make some assumptions about contiguousness. 968 */ 969 if (args->L3offset) 970 m_adj(m, args->L3offset); 971 #endif 972 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 973 } else 974 FREE_PKT(m); 975 976 args->m = NULL; 977 } 978 979 #endif /* INET6 */ 980 981 982 /* 983 * sends a reject message, consuming the mbuf passed as an argument. 984 */ 985 static void 986 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip) 987 { 988 989 #if 0 990 /* XXX When ip is not guaranteed to be at mtod() we will 991 * need to account for this */ 992 * The mbuf will however be thrown away so we can adjust it. 993 * Remember we did an m_pullup on it already so we 994 * can make some assumptions about contiguousness. 995 */ 996 if (args->L3offset) 997 m_adj(m, args->L3offset); 998 #endif 999 if (code != ICMP_REJECT_RST && code != ICMP_REJECT_ABORT) { 1000 /* Send an ICMP unreach */ 1001 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1002 } else if (code == ICMP_REJECT_RST && args->f_id.proto == IPPROTO_TCP) { 1003 struct tcphdr *const tcp = 1004 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1005 if ( (tcp->th_flags & TH_RST) == 0) { 1006 struct mbuf *m; 1007 m = ipfw_send_pkt(args->m, &(args->f_id), 1008 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1009 tcp->th_flags | TH_RST); 1010 if (m != NULL) 1011 ip_output(m, NULL, NULL, 0, NULL, NULL); 1012 } 1013 FREE_PKT(args->m); 1014 } else if (code == ICMP_REJECT_ABORT && 1015 args->f_id.proto == IPPROTO_SCTP) { 1016 struct mbuf *m; 1017 struct sctphdr *sctp; 1018 struct sctp_chunkhdr *chunk; 1019 struct sctp_init *init; 1020 u_int32_t v_tag; 1021 int reflected; 1022 1023 sctp = L3HDR(struct sctphdr, mtod(args->m, struct ip *)); 1024 reflected = 1; 1025 v_tag = ntohl(sctp->v_tag); 1026 if (iplen >= (ip->ip_hl << 2) + sizeof(struct sctphdr) + 1027 sizeof(struct sctp_chunkhdr)) { 1028 /* Look at the first chunk header if available */ 1029 chunk = (struct sctp_chunkhdr *)(sctp + 1); 1030 switch (chunk->chunk_type) { 1031 case SCTP_INITIATION: 1032 /* 1033 * Packets containing an INIT chunk MUST have 1034 * a zero v-tag. 1035 */ 1036 if (v_tag != 0) { 1037 v_tag = 0; 1038 break; 1039 } 1040 /* INIT chunk MUST NOT be bundled */ 1041 if (iplen > 1042 (ip->ip_hl << 2) + sizeof(struct sctphdr) + 1043 ntohs(chunk->chunk_length) + 3) { 1044 break; 1045 } 1046 /* Use the initiate tag if available */ 1047 if ((iplen >= (ip->ip_hl << 2) + 1048 sizeof(struct sctphdr) + 1049 sizeof(struct sctp_chunkhdr) + 1050 offsetof(struct sctp_init, a_rwnd))) { 1051 init = (struct sctp_init *)(chunk + 1); 1052 v_tag = ntohl(init->initiate_tag); 1053 reflected = 0; 1054 } 1055 break; 1056 case SCTP_ABORT_ASSOCIATION: 1057 /* 1058 * If the packet contains an ABORT chunk, don't 1059 * reply. 1060 * XXX: We should search through all chunks, 1061 * but do not do that to avoid attacks. 1062 */ 1063 v_tag = 0; 1064 break; 1065 } 1066 } 1067 if (v_tag == 0) { 1068 m = NULL; 1069 } else { 1070 m = ipfw_send_abort(args->m, &(args->f_id), v_tag, 1071 reflected); 1072 } 1073 if (m != NULL) 1074 ip_output(m, NULL, NULL, 0, NULL, NULL); 1075 FREE_PKT(args->m); 1076 } else 1077 FREE_PKT(args->m); 1078 args->m = NULL; 1079 } 1080 1081 /* 1082 * Support for uid/gid/jail lookup. These tests are expensive 1083 * (because we may need to look into the list of active sockets) 1084 * so we cache the results. ugid_lookupp is 0 if we have not 1085 * yet done a lookup, 1 if we succeeded, and -1 if we tried 1086 * and failed. The function always returns the match value. 1087 * We could actually spare the variable and use *uc, setting 1088 * it to '(void *)check_uidgid if we have no info, NULL if 1089 * we tried and failed, or any other value if successful. 1090 */ 1091 static int 1092 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp, 1093 struct ucred **uc) 1094 { 1095 #if defined(USERSPACE) 1096 return 0; // not supported in userspace 1097 #else 1098 #ifndef __FreeBSD__ 1099 /* XXX */ 1100 return cred_check(insn, proto, oif, 1101 dst_ip, dst_port, src_ip, src_port, 1102 (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb); 1103 #else /* FreeBSD */ 1104 struct in_addr src_ip, dst_ip; 1105 struct inpcbinfo *pi; 1106 struct ipfw_flow_id *id; 1107 struct inpcb *pcb, *inp; 1108 int lookupflags; 1109 int match; 1110 1111 id = &args->f_id; 1112 inp = args->inp; 1113 1114 /* 1115 * Check to see if the UDP or TCP stack supplied us with 1116 * the PCB. If so, rather then holding a lock and looking 1117 * up the PCB, we can use the one that was supplied. 1118 */ 1119 if (inp && *ugid_lookupp == 0) { 1120 INP_LOCK_ASSERT(inp); 1121 if (inp->inp_socket != NULL) { 1122 *uc = crhold(inp->inp_cred); 1123 *ugid_lookupp = 1; 1124 } else 1125 *ugid_lookupp = -1; 1126 } 1127 /* 1128 * If we have already been here and the packet has no 1129 * PCB entry associated with it, then we can safely 1130 * assume that this is a no match. 1131 */ 1132 if (*ugid_lookupp == -1) 1133 return (0); 1134 if (id->proto == IPPROTO_TCP) { 1135 lookupflags = 0; 1136 pi = &V_tcbinfo; 1137 } else if (id->proto == IPPROTO_UDP) { 1138 lookupflags = INPLOOKUP_WILDCARD; 1139 pi = &V_udbinfo; 1140 } else if (id->proto == IPPROTO_UDPLITE) { 1141 lookupflags = INPLOOKUP_WILDCARD; 1142 pi = &V_ulitecbinfo; 1143 } else 1144 return 0; 1145 lookupflags |= INPLOOKUP_RLOCKPCB; 1146 match = 0; 1147 if (*ugid_lookupp == 0) { 1148 if (id->addr_type == 6) { 1149 #ifdef INET6 1150 if (args->flags & IPFW_ARGS_IN) 1151 pcb = in6_pcblookup_mbuf(pi, 1152 &id->src_ip6, htons(id->src_port), 1153 &id->dst_ip6, htons(id->dst_port), 1154 lookupflags, NULL, args->m); 1155 else 1156 pcb = in6_pcblookup_mbuf(pi, 1157 &id->dst_ip6, htons(id->dst_port), 1158 &id->src_ip6, htons(id->src_port), 1159 lookupflags, args->ifp, args->m); 1160 #else 1161 *ugid_lookupp = -1; 1162 return (0); 1163 #endif 1164 } else { 1165 src_ip.s_addr = htonl(id->src_ip); 1166 dst_ip.s_addr = htonl(id->dst_ip); 1167 if (args->flags & IPFW_ARGS_IN) 1168 pcb = in_pcblookup_mbuf(pi, 1169 src_ip, htons(id->src_port), 1170 dst_ip, htons(id->dst_port), 1171 lookupflags, NULL, args->m); 1172 else 1173 pcb = in_pcblookup_mbuf(pi, 1174 dst_ip, htons(id->dst_port), 1175 src_ip, htons(id->src_port), 1176 lookupflags, args->ifp, args->m); 1177 } 1178 if (pcb != NULL) { 1179 INP_RLOCK_ASSERT(pcb); 1180 *uc = crhold(pcb->inp_cred); 1181 *ugid_lookupp = 1; 1182 INP_RUNLOCK(pcb); 1183 } 1184 if (*ugid_lookupp == 0) { 1185 /* 1186 * We tried and failed, set the variable to -1 1187 * so we will not try again on this packet. 1188 */ 1189 *ugid_lookupp = -1; 1190 return (0); 1191 } 1192 } 1193 if (insn->o.opcode == O_UID) 1194 match = ((*uc)->cr_uid == (uid_t)insn->d[0]); 1195 else if (insn->o.opcode == O_GID) 1196 match = groupmember((gid_t)insn->d[0], *uc); 1197 else if (insn->o.opcode == O_JAIL) 1198 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]); 1199 return (match); 1200 #endif /* __FreeBSD__ */ 1201 #endif /* not supported in userspace */ 1202 } 1203 1204 /* 1205 * Helper function to set args with info on the rule after the matching 1206 * one. slot is precise, whereas we guess rule_id as they are 1207 * assigned sequentially. 1208 */ 1209 static inline void 1210 set_match(struct ip_fw_args *args, int slot, 1211 struct ip_fw_chain *chain) 1212 { 1213 args->rule.chain_id = chain->id; 1214 args->rule.slot = slot + 1; /* we use 0 as a marker */ 1215 args->rule.rule_id = 1 + chain->map[slot]->id; 1216 args->rule.rulenum = chain->map[slot]->rulenum; 1217 args->flags |= IPFW_ARGS_REF; 1218 } 1219 1220 #ifndef LINEAR_SKIPTO 1221 /* 1222 * Helper function to enable cached rule lookups using 1223 * cached_id and cached_pos fields in ipfw rule. 1224 */ 1225 static int 1226 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num, 1227 int tablearg, int jump_backwards) 1228 { 1229 int f_pos; 1230 1231 /* If possible use cached f_pos (in f->cached_pos), 1232 * whose version is written in f->cached_id 1233 * (horrible hacks to avoid changing the ABI). 1234 */ 1235 if (num != IP_FW_TARG && f->cached_id == chain->id) 1236 f_pos = f->cached_pos; 1237 else { 1238 int i = IP_FW_ARG_TABLEARG(chain, num, skipto); 1239 /* make sure we do not jump backward */ 1240 if (jump_backwards == 0 && i <= f->rulenum) 1241 i = f->rulenum + 1; 1242 if (chain->idxmap != NULL) 1243 f_pos = chain->idxmap[i]; 1244 else 1245 f_pos = ipfw_find_rule(chain, i, 0); 1246 /* update the cache */ 1247 if (num != IP_FW_TARG) { 1248 f->cached_id = chain->id; 1249 f->cached_pos = f_pos; 1250 } 1251 } 1252 1253 return (f_pos); 1254 } 1255 #else 1256 /* 1257 * Helper function to enable real fast rule lookups. 1258 */ 1259 static int 1260 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num, 1261 int tablearg, int jump_backwards) 1262 { 1263 int f_pos; 1264 1265 num = IP_FW_ARG_TABLEARG(chain, num, skipto); 1266 /* make sure we do not jump backward */ 1267 if (jump_backwards == 0 && num <= f->rulenum) 1268 num = f->rulenum + 1; 1269 f_pos = chain->idxmap[num]; 1270 1271 return (f_pos); 1272 } 1273 #endif 1274 1275 #define TARG(k, f) IP_FW_ARG_TABLEARG(chain, k, f) 1276 /* 1277 * The main check routine for the firewall. 1278 * 1279 * All arguments are in args so we can modify them and return them 1280 * back to the caller. 1281 * 1282 * Parameters: 1283 * 1284 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1285 * Starts with the IP header. 1286 * args->L3offset Number of bytes bypassed if we came from L2. 1287 * e.g. often sizeof(eh) ** NOTYET ** 1288 * args->ifp Incoming or outgoing interface. 1289 * args->divert_rule (in/out) 1290 * Skip up to the first rule past this rule number; 1291 * upon return, non-zero port number for divert or tee. 1292 * 1293 * args->rule Pointer to the last matching rule (in/out) 1294 * args->next_hop Socket we are forwarding to (out). 1295 * args->next_hop6 IPv6 next hop we are forwarding to (out). 1296 * args->f_id Addresses grabbed from the packet (out) 1297 * args->rule.info a cookie depending on rule action 1298 * 1299 * Return value: 1300 * 1301 * IP_FW_PASS the packet must be accepted 1302 * IP_FW_DENY the packet must be dropped 1303 * IP_FW_DIVERT divert packet, port in m_tag 1304 * IP_FW_TEE tee packet, port in m_tag 1305 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 1306 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 1307 * args->rule contains the matching rule, 1308 * args->rule.info has additional information. 1309 * 1310 */ 1311 int 1312 ipfw_chk(struct ip_fw_args *args) 1313 { 1314 1315 /* 1316 * Local variables holding state while processing a packet: 1317 * 1318 * IMPORTANT NOTE: to speed up the processing of rules, there 1319 * are some assumption on the values of the variables, which 1320 * are documented here. Should you change them, please check 1321 * the implementation of the various instructions to make sure 1322 * that they still work. 1323 * 1324 * m | args->m Pointer to the mbuf, as received from the caller. 1325 * It may change if ipfw_chk() does an m_pullup, or if it 1326 * consumes the packet because it calls send_reject(). 1327 * XXX This has to change, so that ipfw_chk() never modifies 1328 * or consumes the buffer. 1329 * OR 1330 * args->mem Pointer to contigous memory chunk. 1331 * ip Is the beginning of the ip(4 or 6) header. 1332 * eh Ethernet header in case if input is Layer2. 1333 */ 1334 struct mbuf *m; 1335 struct ip *ip; 1336 struct ether_header *eh; 1337 1338 /* 1339 * For rules which contain uid/gid or jail constraints, cache 1340 * a copy of the users credentials after the pcb lookup has been 1341 * executed. This will speed up the processing of rules with 1342 * these types of constraints, as well as decrease contention 1343 * on pcb related locks. 1344 */ 1345 #ifndef __FreeBSD__ 1346 struct bsd_ucred ucred_cache; 1347 #else 1348 struct ucred *ucred_cache = NULL; 1349 #endif 1350 int ucred_lookup = 0; 1351 int f_pos = 0; /* index of current rule in the array */ 1352 int retval = 0; 1353 struct ifnet *oif, *iif; 1354 1355 /* 1356 * hlen The length of the IP header. 1357 */ 1358 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 1359 1360 /* 1361 * offset The offset of a fragment. offset != 0 means that 1362 * we have a fragment at this offset of an IPv4 packet. 1363 * offset == 0 means that (if this is an IPv4 packet) 1364 * this is the first or only fragment. 1365 * For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header 1366 * or there is a single packet fragment (fragment header added 1367 * without needed). We will treat a single packet fragment as if 1368 * there was no fragment header (or log/block depending on the 1369 * V_fw_permit_single_frag6 sysctl setting). 1370 */ 1371 u_short offset = 0; 1372 u_short ip6f_mf = 0; 1373 1374 /* 1375 * Local copies of addresses. They are only valid if we have 1376 * an IP packet. 1377 * 1378 * proto The protocol. Set to 0 for non-ip packets, 1379 * or to the protocol read from the packet otherwise. 1380 * proto != 0 means that we have an IPv4 packet. 1381 * 1382 * src_port, dst_port port numbers, in HOST format. Only 1383 * valid for TCP and UDP packets. 1384 * 1385 * src_ip, dst_ip ip addresses, in NETWORK format. 1386 * Only valid for IPv4 packets. 1387 */ 1388 uint8_t proto; 1389 uint16_t src_port, dst_port; /* NOTE: host format */ 1390 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 1391 int iplen = 0; 1392 int pktlen; 1393 1394 struct ipfw_dyn_info dyn_info; 1395 struct ip_fw *q = NULL; 1396 struct ip_fw_chain *chain = &V_layer3_chain; 1397 1398 /* 1399 * We store in ulp a pointer to the upper layer protocol header. 1400 * In the ipv4 case this is easy to determine from the header, 1401 * but for ipv6 we might have some additional headers in the middle. 1402 * ulp is NULL if not found. 1403 */ 1404 void *ulp = NULL; /* upper layer protocol pointer. */ 1405 1406 /* XXX ipv6 variables */ 1407 int is_ipv6 = 0; 1408 uint8_t icmp6_type = 0; 1409 uint16_t ext_hd = 0; /* bits vector for extension header filtering */ 1410 /* end of ipv6 variables */ 1411 1412 int is_ipv4 = 0; 1413 1414 int done = 0; /* flag to exit the outer loop */ 1415 IPFW_RLOCK_TRACKER; 1416 bool mem; 1417 1418 if ((mem = (args->flags & IPFW_ARGS_LENMASK))) { 1419 if (args->flags & IPFW_ARGS_ETHER) { 1420 eh = (struct ether_header *)args->mem; 1421 if (eh->ether_type == htons(ETHERTYPE_VLAN)) 1422 ip = (struct ip *) 1423 ((struct ether_vlan_header *)eh + 1); 1424 else 1425 ip = (struct ip *)(eh + 1); 1426 } else { 1427 eh = NULL; 1428 ip = (struct ip *)args->mem; 1429 } 1430 pktlen = IPFW_ARGS_LENGTH(args->flags); 1431 args->f_id.fib = args->ifp->if_fib; /* best guess */ 1432 } else { 1433 m = args->m; 1434 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready)) 1435 return (IP_FW_PASS); /* accept */ 1436 if (args->flags & IPFW_ARGS_ETHER) { 1437 /* We need some amount of data to be contiguous. */ 1438 if (m->m_len < min(m->m_pkthdr.len, max_protohdr) && 1439 (args->m = m = m_pullup(m, min(m->m_pkthdr.len, 1440 max_protohdr))) == NULL) 1441 goto pullup_failed; 1442 eh = mtod(m, struct ether_header *); 1443 ip = (struct ip *)(eh + 1); 1444 } else { 1445 eh = NULL; 1446 ip = mtod(m, struct ip *); 1447 } 1448 pktlen = m->m_pkthdr.len; 1449 args->f_id.fib = M_GETFIB(m); /* mbuf not altered */ 1450 } 1451 1452 dst_ip.s_addr = 0; /* make sure it is initialized */ 1453 src_ip.s_addr = 0; /* make sure it is initialized */ 1454 src_port = dst_port = 0; 1455 1456 DYN_INFO_INIT(&dyn_info); 1457 /* 1458 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 1459 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 1460 * pointer might become stale after other pullups (but we never use it 1461 * this way). 1462 */ 1463 #define PULLUP_TO(_len, p, T) PULLUP_LEN(_len, p, sizeof(T)) 1464 #define EHLEN (eh != NULL ? ((char *)ip - (char *)eh) : 0) 1465 #define _PULLUP_LOCKED(_len, p, T, unlock) \ 1466 do { \ 1467 int x = (_len) + T + EHLEN; \ 1468 if (mem) { \ 1469 if (__predict_false(pktlen < x)) { \ 1470 unlock; \ 1471 goto pullup_failed; \ 1472 } \ 1473 p = (char *)args->mem + (_len) + EHLEN; \ 1474 } else { \ 1475 if (__predict_false((m)->m_len < x)) { \ 1476 args->m = m = m_pullup(m, x); \ 1477 if (m == NULL) { \ 1478 unlock; \ 1479 goto pullup_failed; \ 1480 } \ 1481 } \ 1482 p = mtod(m, char *) + (_len) + EHLEN; \ 1483 } \ 1484 } while (0) 1485 1486 #define PULLUP_LEN(_len, p, T) _PULLUP_LOCKED(_len, p, T, ) 1487 #define PULLUP_LEN_LOCKED(_len, p, T) \ 1488 _PULLUP_LOCKED(_len, p, T, IPFW_PF_RUNLOCK(chain)); \ 1489 UPDATE_POINTERS() 1490 /* 1491 * In case pointers got stale after pullups, update them. 1492 */ 1493 #define UPDATE_POINTERS() \ 1494 do { \ 1495 if (!mem) { \ 1496 if (eh != NULL) { \ 1497 eh = mtod(m, struct ether_header *); \ 1498 ip = (struct ip *)(eh + 1); \ 1499 } else \ 1500 ip = mtod(m, struct ip *); \ 1501 args->m = m; \ 1502 } \ 1503 } while (0) 1504 1505 /* Identify IP packets and fill up variables. */ 1506 if (pktlen >= sizeof(struct ip6_hdr) && 1507 (eh == NULL || eh->ether_type == htons(ETHERTYPE_IPV6)) && 1508 ip->ip_v == 6) { 1509 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 1510 1511 is_ipv6 = 1; 1512 args->flags |= IPFW_ARGS_IP6; 1513 hlen = sizeof(struct ip6_hdr); 1514 proto = ip6->ip6_nxt; 1515 /* Search extension headers to find upper layer protocols */ 1516 while (ulp == NULL && offset == 0) { 1517 switch (proto) { 1518 case IPPROTO_ICMPV6: 1519 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 1520 icmp6_type = ICMP6(ulp)->icmp6_type; 1521 break; 1522 1523 case IPPROTO_TCP: 1524 PULLUP_TO(hlen, ulp, struct tcphdr); 1525 dst_port = TCP(ulp)->th_dport; 1526 src_port = TCP(ulp)->th_sport; 1527 /* save flags for dynamic rules */ 1528 args->f_id._flags = TCP(ulp)->th_flags; 1529 break; 1530 1531 case IPPROTO_SCTP: 1532 if (pktlen >= hlen + sizeof(struct sctphdr) + 1533 sizeof(struct sctp_chunkhdr) + 1534 offsetof(struct sctp_init, a_rwnd)) 1535 PULLUP_LEN(hlen, ulp, 1536 sizeof(struct sctphdr) + 1537 sizeof(struct sctp_chunkhdr) + 1538 offsetof(struct sctp_init, a_rwnd)); 1539 else if (pktlen >= hlen + sizeof(struct sctphdr)) 1540 PULLUP_LEN(hlen, ulp, pktlen - hlen); 1541 else 1542 PULLUP_LEN(hlen, ulp, 1543 sizeof(struct sctphdr)); 1544 src_port = SCTP(ulp)->src_port; 1545 dst_port = SCTP(ulp)->dest_port; 1546 break; 1547 1548 case IPPROTO_UDP: 1549 case IPPROTO_UDPLITE: 1550 PULLUP_TO(hlen, ulp, struct udphdr); 1551 dst_port = UDP(ulp)->uh_dport; 1552 src_port = UDP(ulp)->uh_sport; 1553 break; 1554 1555 case IPPROTO_HOPOPTS: /* RFC 2460 */ 1556 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1557 ext_hd |= EXT_HOPOPTS; 1558 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1559 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1560 ulp = NULL; 1561 break; 1562 1563 case IPPROTO_ROUTING: /* RFC 2460 */ 1564 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 1565 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 1566 case 0: 1567 ext_hd |= EXT_RTHDR0; 1568 break; 1569 case 2: 1570 ext_hd |= EXT_RTHDR2; 1571 break; 1572 default: 1573 if (V_fw_verbose) 1574 printf("IPFW2: IPV6 - Unknown " 1575 "Routing Header type(%d)\n", 1576 ((struct ip6_rthdr *) 1577 ulp)->ip6r_type); 1578 if (V_fw_deny_unknown_exthdrs) 1579 return (IP_FW_DENY); 1580 break; 1581 } 1582 ext_hd |= EXT_ROUTING; 1583 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 1584 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 1585 ulp = NULL; 1586 break; 1587 1588 case IPPROTO_FRAGMENT: /* RFC 2460 */ 1589 PULLUP_TO(hlen, ulp, struct ip6_frag); 1590 ext_hd |= EXT_FRAGMENT; 1591 hlen += sizeof (struct ip6_frag); 1592 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 1593 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 1594 IP6F_OFF_MASK; 1595 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg & 1596 IP6F_MORE_FRAG; 1597 if (V_fw_permit_single_frag6 == 0 && 1598 offset == 0 && ip6f_mf == 0) { 1599 if (V_fw_verbose) 1600 printf("IPFW2: IPV6 - Invalid " 1601 "Fragment Header\n"); 1602 if (V_fw_deny_unknown_exthdrs) 1603 return (IP_FW_DENY); 1604 break; 1605 } 1606 args->f_id.extra = 1607 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 1608 ulp = NULL; 1609 break; 1610 1611 case IPPROTO_DSTOPTS: /* RFC 2460 */ 1612 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1613 ext_hd |= EXT_DSTOPTS; 1614 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1615 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1616 ulp = NULL; 1617 break; 1618 1619 case IPPROTO_AH: /* RFC 2402 */ 1620 PULLUP_TO(hlen, ulp, struct ip6_ext); 1621 ext_hd |= EXT_AH; 1622 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 1623 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 1624 ulp = NULL; 1625 break; 1626 1627 case IPPROTO_ESP: /* RFC 2406 */ 1628 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 1629 /* Anything past Seq# is variable length and 1630 * data past this ext. header is encrypted. */ 1631 ext_hd |= EXT_ESP; 1632 break; 1633 1634 case IPPROTO_NONE: /* RFC 2460 */ 1635 /* 1636 * Packet ends here, and IPv6 header has 1637 * already been pulled up. If ip6e_len!=0 1638 * then octets must be ignored. 1639 */ 1640 ulp = ip; /* non-NULL to get out of loop. */ 1641 break; 1642 1643 case IPPROTO_OSPFIGP: 1644 /* XXX OSPF header check? */ 1645 PULLUP_TO(hlen, ulp, struct ip6_ext); 1646 break; 1647 1648 case IPPROTO_PIM: 1649 /* XXX PIM header check? */ 1650 PULLUP_TO(hlen, ulp, struct pim); 1651 break; 1652 1653 case IPPROTO_GRE: /* RFC 1701 */ 1654 /* XXX GRE header check? */ 1655 PULLUP_TO(hlen, ulp, struct grehdr); 1656 break; 1657 1658 case IPPROTO_CARP: 1659 PULLUP_TO(hlen, ulp, offsetof( 1660 struct carp_header, carp_counter)); 1661 if (CARP_ADVERTISEMENT != 1662 ((struct carp_header *)ulp)->carp_type) 1663 return (IP_FW_DENY); 1664 break; 1665 1666 case IPPROTO_IPV6: /* RFC 2893 */ 1667 PULLUP_TO(hlen, ulp, struct ip6_hdr); 1668 break; 1669 1670 case IPPROTO_IPV4: /* RFC 2893 */ 1671 PULLUP_TO(hlen, ulp, struct ip); 1672 break; 1673 1674 default: 1675 if (V_fw_verbose) 1676 printf("IPFW2: IPV6 - Unknown " 1677 "Extension Header(%d), ext_hd=%x\n", 1678 proto, ext_hd); 1679 if (V_fw_deny_unknown_exthdrs) 1680 return (IP_FW_DENY); 1681 PULLUP_TO(hlen, ulp, struct ip6_ext); 1682 break; 1683 } /*switch */ 1684 } 1685 UPDATE_POINTERS(); 1686 ip6 = (struct ip6_hdr *)ip; 1687 args->f_id.addr_type = 6; 1688 args->f_id.src_ip6 = ip6->ip6_src; 1689 args->f_id.dst_ip6 = ip6->ip6_dst; 1690 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 1691 iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6); 1692 } else if (pktlen >= sizeof(struct ip) && 1693 (eh == NULL || eh->ether_type == htons(ETHERTYPE_IP)) && 1694 ip->ip_v == 4) { 1695 is_ipv4 = 1; 1696 args->flags |= IPFW_ARGS_IP4; 1697 hlen = ip->ip_hl << 2; 1698 /* 1699 * Collect parameters into local variables for faster 1700 * matching. 1701 */ 1702 proto = ip->ip_p; 1703 src_ip = ip->ip_src; 1704 dst_ip = ip->ip_dst; 1705 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1706 iplen = ntohs(ip->ip_len); 1707 1708 if (offset == 0) { 1709 switch (proto) { 1710 case IPPROTO_TCP: 1711 PULLUP_TO(hlen, ulp, struct tcphdr); 1712 dst_port = TCP(ulp)->th_dport; 1713 src_port = TCP(ulp)->th_sport; 1714 /* save flags for dynamic rules */ 1715 args->f_id._flags = TCP(ulp)->th_flags; 1716 break; 1717 1718 case IPPROTO_SCTP: 1719 if (pktlen >= hlen + sizeof(struct sctphdr) + 1720 sizeof(struct sctp_chunkhdr) + 1721 offsetof(struct sctp_init, a_rwnd)) 1722 PULLUP_LEN(hlen, ulp, 1723 sizeof(struct sctphdr) + 1724 sizeof(struct sctp_chunkhdr) + 1725 offsetof(struct sctp_init, a_rwnd)); 1726 else if (pktlen >= hlen + sizeof(struct sctphdr)) 1727 PULLUP_LEN(hlen, ulp, pktlen - hlen); 1728 else 1729 PULLUP_LEN(hlen, ulp, 1730 sizeof(struct sctphdr)); 1731 src_port = SCTP(ulp)->src_port; 1732 dst_port = SCTP(ulp)->dest_port; 1733 break; 1734 1735 case IPPROTO_UDP: 1736 case IPPROTO_UDPLITE: 1737 PULLUP_TO(hlen, ulp, struct udphdr); 1738 dst_port = UDP(ulp)->uh_dport; 1739 src_port = UDP(ulp)->uh_sport; 1740 break; 1741 1742 case IPPROTO_ICMP: 1743 PULLUP_TO(hlen, ulp, struct icmphdr); 1744 //args->f_id.flags = ICMP(ulp)->icmp_type; 1745 break; 1746 1747 default: 1748 break; 1749 } 1750 } else { 1751 if (offset == 1 && proto == IPPROTO_TCP) { 1752 /* RFC 3128 */ 1753 goto pullup_failed; 1754 } 1755 } 1756 1757 UPDATE_POINTERS(); 1758 args->f_id.addr_type = 4; 1759 args->f_id.src_ip = ntohl(src_ip.s_addr); 1760 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1761 } else { 1762 proto = 0; 1763 dst_ip.s_addr = src_ip.s_addr = 0; 1764 1765 args->f_id.addr_type = 1; /* XXX */ 1766 } 1767 #undef PULLUP_TO 1768 pktlen = iplen < pktlen ? iplen: pktlen; 1769 1770 /* Properly initialize the rest of f_id */ 1771 args->f_id.proto = proto; 1772 args->f_id.src_port = src_port = ntohs(src_port); 1773 args->f_id.dst_port = dst_port = ntohs(dst_port); 1774 1775 IPFW_PF_RLOCK(chain); 1776 if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */ 1777 IPFW_PF_RUNLOCK(chain); 1778 return (IP_FW_PASS); /* accept */ 1779 } 1780 if (args->flags & IPFW_ARGS_REF) { 1781 /* 1782 * Packet has already been tagged as a result of a previous 1783 * match on rule args->rule aka args->rule_id (PIPE, QUEUE, 1784 * REASS, NETGRAPH, DIVERT/TEE...) 1785 * Validate the slot and continue from the next one 1786 * if still present, otherwise do a lookup. 1787 */ 1788 f_pos = (args->rule.chain_id == chain->id) ? 1789 args->rule.slot : 1790 ipfw_find_rule(chain, args->rule.rulenum, 1791 args->rule.rule_id); 1792 } else { 1793 f_pos = 0; 1794 } 1795 1796 if (args->flags & IPFW_ARGS_IN) { 1797 iif = args->ifp; 1798 oif = NULL; 1799 } else { 1800 MPASS(args->flags & IPFW_ARGS_OUT); 1801 iif = mem ? NULL : m_rcvif(m); 1802 oif = args->ifp; 1803 } 1804 1805 /* 1806 * Now scan the rules, and parse microinstructions for each rule. 1807 * We have two nested loops and an inner switch. Sometimes we 1808 * need to break out of one or both loops, or re-enter one of 1809 * the loops with updated variables. Loop variables are: 1810 * 1811 * f_pos (outer loop) points to the current rule. 1812 * On output it points to the matching rule. 1813 * done (outer loop) is used as a flag to break the loop. 1814 * l (inner loop) residual length of current rule. 1815 * cmd points to the current microinstruction. 1816 * 1817 * We break the inner loop by setting l=0 and possibly 1818 * cmdlen=0 if we don't want to advance cmd. 1819 * We break the outer loop by setting done=1 1820 * We can restart the inner loop by setting l>0 and f_pos, f, cmd 1821 * as needed. 1822 */ 1823 for (; f_pos < chain->n_rules; f_pos++) { 1824 ipfw_insn *cmd; 1825 uint32_t tablearg = 0; 1826 int l, cmdlen, skip_or; /* skip rest of OR block */ 1827 struct ip_fw *f; 1828 1829 f = chain->map[f_pos]; 1830 if (V_set_disable & (1 << f->set) ) 1831 continue; 1832 1833 skip_or = 0; 1834 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1835 l -= cmdlen, cmd += cmdlen) { 1836 int match; 1837 1838 /* 1839 * check_body is a jump target used when we find a 1840 * CHECK_STATE, and need to jump to the body of 1841 * the target rule. 1842 */ 1843 1844 /* check_body: */ 1845 cmdlen = F_LEN(cmd); 1846 /* 1847 * An OR block (insn_1 || .. || insn_n) has the 1848 * F_OR bit set in all but the last instruction. 1849 * The first match will set "skip_or", and cause 1850 * the following instructions to be skipped until 1851 * past the one with the F_OR bit clear. 1852 */ 1853 if (skip_or) { /* skip this instruction */ 1854 if ((cmd->len & F_OR) == 0) 1855 skip_or = 0; /* next one is good */ 1856 continue; 1857 } 1858 match = 0; /* set to 1 if we succeed */ 1859 1860 switch (cmd->opcode) { 1861 /* 1862 * The first set of opcodes compares the packet's 1863 * fields with some pattern, setting 'match' if a 1864 * match is found. At the end of the loop there is 1865 * logic to deal with F_NOT and F_OR flags associated 1866 * with the opcode. 1867 */ 1868 case O_NOP: 1869 match = 1; 1870 break; 1871 1872 case O_FORWARD_MAC: 1873 printf("ipfw: opcode %d unimplemented\n", 1874 cmd->opcode); 1875 break; 1876 1877 case O_GID: 1878 case O_UID: 1879 case O_JAIL: 1880 /* 1881 * We only check offset == 0 && proto != 0, 1882 * as this ensures that we have a 1883 * packet with the ports info. 1884 */ 1885 if (offset != 0) 1886 break; 1887 if (proto == IPPROTO_TCP || 1888 proto == IPPROTO_UDP || 1889 proto == IPPROTO_UDPLITE) 1890 match = check_uidgid( 1891 (ipfw_insn_u32 *)cmd, 1892 args, &ucred_lookup, 1893 #ifdef __FreeBSD__ 1894 &ucred_cache); 1895 #else 1896 (void *)&ucred_cache); 1897 #endif 1898 break; 1899 1900 case O_RECV: 1901 match = iface_match(iif, (ipfw_insn_if *)cmd, 1902 chain, &tablearg); 1903 break; 1904 1905 case O_XMIT: 1906 match = iface_match(oif, (ipfw_insn_if *)cmd, 1907 chain, &tablearg); 1908 break; 1909 1910 case O_VIA: 1911 match = iface_match(args->ifp, 1912 (ipfw_insn_if *)cmd, chain, &tablearg); 1913 break; 1914 1915 case O_MACADDR2: 1916 if (args->flags & IPFW_ARGS_ETHER) { 1917 u_int32_t *want = (u_int32_t *) 1918 ((ipfw_insn_mac *)cmd)->addr; 1919 u_int32_t *mask = (u_int32_t *) 1920 ((ipfw_insn_mac *)cmd)->mask; 1921 u_int32_t *hdr = (u_int32_t *)eh; 1922 1923 match = 1924 ( want[0] == (hdr[0] & mask[0]) && 1925 want[1] == (hdr[1] & mask[1]) && 1926 want[2] == (hdr[2] & mask[2]) ); 1927 } 1928 break; 1929 1930 case O_MAC_TYPE: 1931 if (args->flags & IPFW_ARGS_ETHER) { 1932 u_int16_t *p = 1933 ((ipfw_insn_u16 *)cmd)->ports; 1934 int i; 1935 1936 for (i = cmdlen - 1; !match && i>0; 1937 i--, p += 2) 1938 match = 1939 (ntohs(eh->ether_type) >= 1940 p[0] && 1941 ntohs(eh->ether_type) <= 1942 p[1]); 1943 } 1944 break; 1945 1946 case O_FRAG: 1947 if (is_ipv4) { 1948 /* 1949 * Since flags_match() works with 1950 * uint8_t we pack ip_off into 8 bits. 1951 * For this match offset is a boolean. 1952 */ 1953 match = flags_match(cmd, 1954 ((ntohs(ip->ip_off) & ~IP_OFFMASK) 1955 >> 8) | (offset != 0)); 1956 } else { 1957 /* 1958 * Compatiblity: historically bare 1959 * "frag" would match IPv6 fragments. 1960 */ 1961 match = (cmd->arg1 == 0x1 && 1962 (offset != 0)); 1963 } 1964 break; 1965 1966 case O_IN: /* "out" is "not in" */ 1967 match = (oif == NULL); 1968 break; 1969 1970 case O_LAYER2: 1971 match = (args->flags & IPFW_ARGS_ETHER); 1972 break; 1973 1974 case O_DIVERTED: 1975 if ((args->flags & IPFW_ARGS_REF) == 0) 1976 break; 1977 /* 1978 * For diverted packets, args->rule.info 1979 * contains the divert port (in host format) 1980 * reason and direction. 1981 */ 1982 match = ((args->rule.info & IPFW_IS_MASK) == 1983 IPFW_IS_DIVERT) && ( 1984 ((args->rule.info & IPFW_INFO_IN) ? 1985 1: 2) & cmd->arg1); 1986 break; 1987 1988 case O_PROTO: 1989 /* 1990 * We do not allow an arg of 0 so the 1991 * check of "proto" only suffices. 1992 */ 1993 match = (proto == cmd->arg1); 1994 break; 1995 1996 case O_IP_SRC: 1997 match = is_ipv4 && 1998 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1999 src_ip.s_addr); 2000 break; 2001 2002 case O_IP_DST_LOOKUP: 2003 { 2004 void *pkey; 2005 uint32_t vidx, key; 2006 uint16_t keylen; 2007 2008 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) { 2009 /* Determine lookup key type */ 2010 vidx = ((ipfw_insn_u32 *)cmd)->d[1]; 2011 if (vidx != 4 /* uid */ && 2012 vidx != 5 /* jail */ && 2013 is_ipv6 == 0 && is_ipv4 == 0) 2014 break; 2015 /* Determine key length */ 2016 if (vidx == 0 /* dst-ip */ || 2017 vidx == 1 /* src-ip */) 2018 keylen = is_ipv6 ? 2019 sizeof(struct in6_addr): 2020 sizeof(in_addr_t); 2021 else { 2022 keylen = sizeof(key); 2023 pkey = &key; 2024 } 2025 if (vidx == 0 /* dst-ip */) 2026 pkey = is_ipv4 ? (void *)&dst_ip: 2027 (void *)&args->f_id.dst_ip6; 2028 else if (vidx == 1 /* src-ip */) 2029 pkey = is_ipv4 ? (void *)&src_ip: 2030 (void *)&args->f_id.src_ip6; 2031 else if (vidx == 6 /* dscp */) { 2032 if (is_ipv4) 2033 key = ip->ip_tos >> 2; 2034 else { 2035 key = args->f_id.flow_id6; 2036 key = (key & 0x0f) << 2 | 2037 (key & 0xf000) >> 14; 2038 } 2039 key &= 0x3f; 2040 } else if (vidx == 2 /* dst-port */ || 2041 vidx == 3 /* src-port */) { 2042 /* Skip fragments */ 2043 if (offset != 0) 2044 break; 2045 /* Skip proto without ports */ 2046 if (proto != IPPROTO_TCP && 2047 proto != IPPROTO_UDP && 2048 proto != IPPROTO_UDPLITE && 2049 proto != IPPROTO_SCTP) 2050 break; 2051 if (vidx == 2 /* dst-port */) 2052 key = dst_port; 2053 else 2054 key = src_port; 2055 } 2056 #ifndef USERSPACE 2057 else if (vidx == 4 /* uid */ || 2058 vidx == 5 /* jail */) { 2059 check_uidgid( 2060 (ipfw_insn_u32 *)cmd, 2061 args, &ucred_lookup, 2062 #ifdef __FreeBSD__ 2063 &ucred_cache); 2064 if (vidx == 4 /* uid */) 2065 key = ucred_cache->cr_uid; 2066 else if (vidx == 5 /* jail */) 2067 key = ucred_cache->cr_prison->pr_id; 2068 #else /* !__FreeBSD__ */ 2069 (void *)&ucred_cache); 2070 if (vidx == 4 /* uid */) 2071 key = ucred_cache.uid; 2072 else if (vidx == 5 /* jail */) 2073 key = ucred_cache.xid; 2074 #endif /* !__FreeBSD__ */ 2075 } 2076 #endif /* !USERSPACE */ 2077 else 2078 break; 2079 match = ipfw_lookup_table(chain, 2080 cmd->arg1, keylen, pkey, &vidx); 2081 if (!match) 2082 break; 2083 tablearg = vidx; 2084 break; 2085 } 2086 /* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */ 2087 /* FALLTHROUGH */ 2088 } 2089 case O_IP_SRC_LOOKUP: 2090 { 2091 void *pkey; 2092 uint32_t vidx; 2093 uint16_t keylen; 2094 2095 if (is_ipv4) { 2096 keylen = sizeof(in_addr_t); 2097 if (cmd->opcode == O_IP_DST_LOOKUP) 2098 pkey = &dst_ip; 2099 else 2100 pkey = &src_ip; 2101 } else if (is_ipv6) { 2102 keylen = sizeof(struct in6_addr); 2103 if (cmd->opcode == O_IP_DST_LOOKUP) 2104 pkey = &args->f_id.dst_ip6; 2105 else 2106 pkey = &args->f_id.src_ip6; 2107 } else 2108 break; 2109 match = ipfw_lookup_table(chain, cmd->arg1, 2110 keylen, pkey, &vidx); 2111 if (!match) 2112 break; 2113 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) { 2114 match = ((ipfw_insn_u32 *)cmd)->d[0] == 2115 TARG_VAL(chain, vidx, tag); 2116 if (!match) 2117 break; 2118 } 2119 tablearg = vidx; 2120 break; 2121 } 2122 2123 case O_IP_FLOW_LOOKUP: 2124 { 2125 uint32_t v = 0; 2126 match = ipfw_lookup_table(chain, 2127 cmd->arg1, 0, &args->f_id, &v); 2128 if (!match) 2129 break; 2130 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2131 match = ((ipfw_insn_u32 *)cmd)->d[0] == 2132 TARG_VAL(chain, v, tag); 2133 if (match) 2134 tablearg = v; 2135 } 2136 break; 2137 case O_IP_SRC_MASK: 2138 case O_IP_DST_MASK: 2139 if (is_ipv4) { 2140 uint32_t a = 2141 (cmd->opcode == O_IP_DST_MASK) ? 2142 dst_ip.s_addr : src_ip.s_addr; 2143 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2144 int i = cmdlen-1; 2145 2146 for (; !match && i>0; i-= 2, p+= 2) 2147 match = (p[0] == (a & p[1])); 2148 } 2149 break; 2150 2151 case O_IP_SRC_ME: 2152 if (is_ipv4) { 2153 match = in_localip(src_ip); 2154 break; 2155 } 2156 #ifdef INET6 2157 /* FALLTHROUGH */ 2158 case O_IP6_SRC_ME: 2159 match = is_ipv6 && 2160 ipfw_localip6(&args->f_id.src_ip6); 2161 #endif 2162 break; 2163 2164 case O_IP_DST_SET: 2165 case O_IP_SRC_SET: 2166 if (is_ipv4) { 2167 u_int32_t *d = (u_int32_t *)(cmd+1); 2168 u_int32_t addr = 2169 cmd->opcode == O_IP_DST_SET ? 2170 args->f_id.dst_ip : 2171 args->f_id.src_ip; 2172 2173 if (addr < d[0]) 2174 break; 2175 addr -= d[0]; /* subtract base */ 2176 match = (addr < cmd->arg1) && 2177 ( d[ 1 + (addr>>5)] & 2178 (1<<(addr & 0x1f)) ); 2179 } 2180 break; 2181 2182 case O_IP_DST: 2183 match = is_ipv4 && 2184 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2185 dst_ip.s_addr); 2186 break; 2187 2188 case O_IP_DST_ME: 2189 if (is_ipv4) { 2190 match = in_localip(dst_ip); 2191 break; 2192 } 2193 #ifdef INET6 2194 /* FALLTHROUGH */ 2195 case O_IP6_DST_ME: 2196 match = is_ipv6 && 2197 ipfw_localip6(&args->f_id.dst_ip6); 2198 #endif 2199 break; 2200 2201 2202 case O_IP_SRCPORT: 2203 case O_IP_DSTPORT: 2204 /* 2205 * offset == 0 && proto != 0 is enough 2206 * to guarantee that we have a 2207 * packet with port info. 2208 */ 2209 if ((proto == IPPROTO_UDP || 2210 proto == IPPROTO_UDPLITE || 2211 proto == IPPROTO_TCP || 2212 proto == IPPROTO_SCTP) && offset == 0) { 2213 u_int16_t x = 2214 (cmd->opcode == O_IP_SRCPORT) ? 2215 src_port : dst_port ; 2216 u_int16_t *p = 2217 ((ipfw_insn_u16 *)cmd)->ports; 2218 int i; 2219 2220 for (i = cmdlen - 1; !match && i>0; 2221 i--, p += 2) 2222 match = (x>=p[0] && x<=p[1]); 2223 } 2224 break; 2225 2226 case O_ICMPTYPE: 2227 match = (offset == 0 && proto==IPPROTO_ICMP && 2228 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2229 break; 2230 2231 #ifdef INET6 2232 case O_ICMP6TYPE: 2233 match = is_ipv6 && offset == 0 && 2234 proto==IPPROTO_ICMPV6 && 2235 icmp6type_match( 2236 ICMP6(ulp)->icmp6_type, 2237 (ipfw_insn_u32 *)cmd); 2238 break; 2239 #endif /* INET6 */ 2240 2241 case O_IPOPT: 2242 match = (is_ipv4 && 2243 ipopts_match(ip, cmd) ); 2244 break; 2245 2246 case O_IPVER: 2247 match = ((is_ipv4 || is_ipv6) && 2248 cmd->arg1 == ip->ip_v); 2249 break; 2250 2251 case O_IPID: 2252 case O_IPTTL: 2253 if (!is_ipv4) 2254 break; 2255 case O_IPLEN: 2256 { /* only for IP packets */ 2257 uint16_t x; 2258 uint16_t *p; 2259 int i; 2260 2261 if (cmd->opcode == O_IPLEN) 2262 x = iplen; 2263 else if (cmd->opcode == O_IPTTL) 2264 x = ip->ip_ttl; 2265 else /* must be IPID */ 2266 x = ntohs(ip->ip_id); 2267 if (cmdlen == 1) { 2268 match = (cmd->arg1 == x); 2269 break; 2270 } 2271 /* otherwise we have ranges */ 2272 p = ((ipfw_insn_u16 *)cmd)->ports; 2273 i = cmdlen - 1; 2274 for (; !match && i>0; i--, p += 2) 2275 match = (x >= p[0] && x <= p[1]); 2276 } 2277 break; 2278 2279 case O_IPPRECEDENCE: 2280 match = (is_ipv4 && 2281 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2282 break; 2283 2284 case O_IPTOS: 2285 match = (is_ipv4 && 2286 flags_match(cmd, ip->ip_tos)); 2287 break; 2288 2289 case O_DSCP: 2290 { 2291 uint32_t *p; 2292 uint16_t x; 2293 2294 p = ((ipfw_insn_u32 *)cmd)->d; 2295 2296 if (is_ipv4) 2297 x = ip->ip_tos >> 2; 2298 else if (is_ipv6) { 2299 uint8_t *v; 2300 v = &((struct ip6_hdr *)ip)->ip6_vfc; 2301 x = (*v & 0x0F) << 2; 2302 v++; 2303 x |= *v >> 6; 2304 } else 2305 break; 2306 2307 /* DSCP bitmask is stored as low_u32 high_u32 */ 2308 if (x >= 32) 2309 match = *(p + 1) & (1 << (x - 32)); 2310 else 2311 match = *p & (1 << x); 2312 } 2313 break; 2314 2315 case O_TCPDATALEN: 2316 if (proto == IPPROTO_TCP && offset == 0) { 2317 struct tcphdr *tcp; 2318 uint16_t x; 2319 uint16_t *p; 2320 int i; 2321 #ifdef INET6 2322 if (is_ipv6) { 2323 struct ip6_hdr *ip6; 2324 2325 ip6 = (struct ip6_hdr *)ip; 2326 if (ip6->ip6_plen == 0) { 2327 /* 2328 * Jumbo payload is not 2329 * supported by this 2330 * opcode. 2331 */ 2332 break; 2333 } 2334 x = iplen - hlen; 2335 } else 2336 #endif /* INET6 */ 2337 x = iplen - (ip->ip_hl << 2); 2338 tcp = TCP(ulp); 2339 x -= tcp->th_off << 2; 2340 if (cmdlen == 1) { 2341 match = (cmd->arg1 == x); 2342 break; 2343 } 2344 /* otherwise we have ranges */ 2345 p = ((ipfw_insn_u16 *)cmd)->ports; 2346 i = cmdlen - 1; 2347 for (; !match && i>0; i--, p += 2) 2348 match = (x >= p[0] && x <= p[1]); 2349 } 2350 break; 2351 2352 case O_TCPFLAGS: 2353 match = (proto == IPPROTO_TCP && offset == 0 && 2354 flags_match(cmd, TCP(ulp)->th_flags)); 2355 break; 2356 2357 case O_TCPOPTS: 2358 if (proto == IPPROTO_TCP && offset == 0 && ulp){ 2359 PULLUP_LEN_LOCKED(hlen, ulp, 2360 (TCP(ulp)->th_off << 2)); 2361 match = tcpopts_match(TCP(ulp), cmd); 2362 } 2363 break; 2364 2365 case O_TCPSEQ: 2366 match = (proto == IPPROTO_TCP && offset == 0 && 2367 ((ipfw_insn_u32 *)cmd)->d[0] == 2368 TCP(ulp)->th_seq); 2369 break; 2370 2371 case O_TCPACK: 2372 match = (proto == IPPROTO_TCP && offset == 0 && 2373 ((ipfw_insn_u32 *)cmd)->d[0] == 2374 TCP(ulp)->th_ack); 2375 break; 2376 2377 case O_TCPMSS: 2378 if (proto == IPPROTO_TCP && 2379 (args->f_id._flags & TH_SYN) != 0 && 2380 ulp != NULL) { 2381 uint16_t mss, *p; 2382 int i; 2383 2384 PULLUP_LEN_LOCKED(hlen, ulp, 2385 (TCP(ulp)->th_off << 2)); 2386 if ((tcpopts_parse(TCP(ulp), &mss) & 2387 IP_FW_TCPOPT_MSS) == 0) 2388 break; 2389 if (cmdlen == 1) { 2390 match = (cmd->arg1 == mss); 2391 break; 2392 } 2393 /* Otherwise we have ranges. */ 2394 p = ((ipfw_insn_u16 *)cmd)->ports; 2395 i = cmdlen - 1; 2396 for (; !match && i > 0; i--, p += 2) 2397 match = (mss >= p[0] && 2398 mss <= p[1]); 2399 } 2400 break; 2401 2402 case O_TCPWIN: 2403 if (proto == IPPROTO_TCP && offset == 0) { 2404 uint16_t x; 2405 uint16_t *p; 2406 int i; 2407 2408 x = ntohs(TCP(ulp)->th_win); 2409 if (cmdlen == 1) { 2410 match = (cmd->arg1 == x); 2411 break; 2412 } 2413 /* Otherwise we have ranges. */ 2414 p = ((ipfw_insn_u16 *)cmd)->ports; 2415 i = cmdlen - 1; 2416 for (; !match && i > 0; i--, p += 2) 2417 match = (x >= p[0] && x <= p[1]); 2418 } 2419 break; 2420 2421 case O_ESTAB: 2422 /* reject packets which have SYN only */ 2423 /* XXX should i also check for TH_ACK ? */ 2424 match = (proto == IPPROTO_TCP && offset == 0 && 2425 (TCP(ulp)->th_flags & 2426 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2427 break; 2428 2429 case O_ALTQ: { 2430 struct pf_mtag *at; 2431 struct m_tag *mtag; 2432 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2433 2434 /* 2435 * ALTQ uses mbuf tags from another 2436 * packet filtering system - pf(4). 2437 * We allocate a tag in its format 2438 * and fill it in, pretending to be pf(4). 2439 */ 2440 match = 1; 2441 at = pf_find_mtag(m); 2442 if (at != NULL && at->qid != 0) 2443 break; 2444 mtag = m_tag_get(PACKET_TAG_PF, 2445 sizeof(struct pf_mtag), M_NOWAIT | M_ZERO); 2446 if (mtag == NULL) { 2447 /* 2448 * Let the packet fall back to the 2449 * default ALTQ. 2450 */ 2451 break; 2452 } 2453 m_tag_prepend(m, mtag); 2454 at = (struct pf_mtag *)(mtag + 1); 2455 at->qid = altq->qid; 2456 at->hdr = ip; 2457 break; 2458 } 2459 2460 case O_LOG: 2461 ipfw_log(chain, f, hlen, args, 2462 offset | ip6f_mf, tablearg, ip); 2463 match = 1; 2464 break; 2465 2466 case O_PROB: 2467 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2468 break; 2469 2470 case O_VERREVPATH: 2471 /* Outgoing packets automatically pass/match */ 2472 match = (args->flags & IPFW_ARGS_OUT || 2473 ( 2474 #ifdef INET6 2475 is_ipv6 ? 2476 verify_path6(&(args->f_id.src_ip6), 2477 iif, args->f_id.fib) : 2478 #endif 2479 verify_path(src_ip, iif, args->f_id.fib))); 2480 break; 2481 2482 case O_VERSRCREACH: 2483 /* Outgoing packets automatically pass/match */ 2484 match = (hlen > 0 && ((oif != NULL) || ( 2485 #ifdef INET6 2486 is_ipv6 ? 2487 verify_path6(&(args->f_id.src_ip6), 2488 NULL, args->f_id.fib) : 2489 #endif 2490 verify_path(src_ip, NULL, args->f_id.fib)))); 2491 break; 2492 2493 case O_ANTISPOOF: 2494 /* Outgoing packets automatically pass/match */ 2495 if (oif == NULL && hlen > 0 && 2496 ( (is_ipv4 && in_localaddr(src_ip)) 2497 #ifdef INET6 2498 || (is_ipv6 && 2499 in6_localaddr(&(args->f_id.src_ip6))) 2500 #endif 2501 )) 2502 match = 2503 #ifdef INET6 2504 is_ipv6 ? verify_path6( 2505 &(args->f_id.src_ip6), iif, 2506 args->f_id.fib) : 2507 #endif 2508 verify_path(src_ip, iif, 2509 args->f_id.fib); 2510 else 2511 match = 1; 2512 break; 2513 2514 case O_IPSEC: 2515 match = (m_tag_find(m, 2516 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2517 /* otherwise no match */ 2518 break; 2519 2520 #ifdef INET6 2521 case O_IP6_SRC: 2522 match = is_ipv6 && 2523 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2524 &((ipfw_insn_ip6 *)cmd)->addr6); 2525 break; 2526 2527 case O_IP6_DST: 2528 match = is_ipv6 && 2529 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2530 &((ipfw_insn_ip6 *)cmd)->addr6); 2531 break; 2532 case O_IP6_SRC_MASK: 2533 case O_IP6_DST_MASK: 2534 if (is_ipv6) { 2535 int i = cmdlen - 1; 2536 struct in6_addr p; 2537 struct in6_addr *d = 2538 &((ipfw_insn_ip6 *)cmd)->addr6; 2539 2540 for (; !match && i > 0; d += 2, 2541 i -= F_INSN_SIZE(struct in6_addr) 2542 * 2) { 2543 p = (cmd->opcode == 2544 O_IP6_SRC_MASK) ? 2545 args->f_id.src_ip6: 2546 args->f_id.dst_ip6; 2547 APPLY_MASK(&p, &d[1]); 2548 match = 2549 IN6_ARE_ADDR_EQUAL(&d[0], 2550 &p); 2551 } 2552 } 2553 break; 2554 2555 case O_FLOW6ID: 2556 match = is_ipv6 && 2557 flow6id_match(args->f_id.flow_id6, 2558 (ipfw_insn_u32 *) cmd); 2559 break; 2560 2561 case O_EXT_HDR: 2562 match = is_ipv6 && 2563 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2564 break; 2565 2566 case O_IP6: 2567 match = is_ipv6; 2568 break; 2569 #endif 2570 2571 case O_IP4: 2572 match = is_ipv4; 2573 break; 2574 2575 case O_TAG: { 2576 struct m_tag *mtag; 2577 uint32_t tag = TARG(cmd->arg1, tag); 2578 2579 /* Packet is already tagged with this tag? */ 2580 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 2581 2582 /* We have `untag' action when F_NOT flag is 2583 * present. And we must remove this mtag from 2584 * mbuf and reset `match' to zero (`match' will 2585 * be inversed later). 2586 * Otherwise we should allocate new mtag and 2587 * push it into mbuf. 2588 */ 2589 if (cmd->len & F_NOT) { /* `untag' action */ 2590 if (mtag != NULL) 2591 m_tag_delete(m, mtag); 2592 match = 0; 2593 } else { 2594 if (mtag == NULL) { 2595 mtag = m_tag_alloc( MTAG_IPFW, 2596 tag, 0, M_NOWAIT); 2597 if (mtag != NULL) 2598 m_tag_prepend(m, mtag); 2599 } 2600 match = 1; 2601 } 2602 break; 2603 } 2604 2605 case O_FIB: /* try match the specified fib */ 2606 if (args->f_id.fib == cmd->arg1) 2607 match = 1; 2608 break; 2609 2610 case O_SOCKARG: { 2611 #ifndef USERSPACE /* not supported in userspace */ 2612 struct inpcb *inp = args->inp; 2613 struct inpcbinfo *pi; 2614 2615 if (is_ipv6) /* XXX can we remove this ? */ 2616 break; 2617 2618 if (proto == IPPROTO_TCP) 2619 pi = &V_tcbinfo; 2620 else if (proto == IPPROTO_UDP) 2621 pi = &V_udbinfo; 2622 else if (proto == IPPROTO_UDPLITE) 2623 pi = &V_ulitecbinfo; 2624 else 2625 break; 2626 2627 /* 2628 * XXXRW: so_user_cookie should almost 2629 * certainly be inp_user_cookie? 2630 */ 2631 2632 /* For incoming packet, lookup up the 2633 inpcb using the src/dest ip/port tuple */ 2634 if (inp == NULL) { 2635 inp = in_pcblookup(pi, 2636 src_ip, htons(src_port), 2637 dst_ip, htons(dst_port), 2638 INPLOOKUP_RLOCKPCB, NULL); 2639 if (inp != NULL) { 2640 tablearg = 2641 inp->inp_socket->so_user_cookie; 2642 if (tablearg) 2643 match = 1; 2644 INP_RUNLOCK(inp); 2645 } 2646 } else { 2647 if (inp->inp_socket) { 2648 tablearg = 2649 inp->inp_socket->so_user_cookie; 2650 if (tablearg) 2651 match = 1; 2652 } 2653 } 2654 #endif /* !USERSPACE */ 2655 break; 2656 } 2657 2658 case O_TAGGED: { 2659 struct m_tag *mtag; 2660 uint32_t tag = TARG(cmd->arg1, tag); 2661 2662 if (cmdlen == 1) { 2663 match = m_tag_locate(m, MTAG_IPFW, 2664 tag, NULL) != NULL; 2665 break; 2666 } 2667 2668 /* we have ranges */ 2669 for (mtag = m_tag_first(m); 2670 mtag != NULL && !match; 2671 mtag = m_tag_next(m, mtag)) { 2672 uint16_t *p; 2673 int i; 2674 2675 if (mtag->m_tag_cookie != MTAG_IPFW) 2676 continue; 2677 2678 p = ((ipfw_insn_u16 *)cmd)->ports; 2679 i = cmdlen - 1; 2680 for(; !match && i > 0; i--, p += 2) 2681 match = 2682 mtag->m_tag_id >= p[0] && 2683 mtag->m_tag_id <= p[1]; 2684 } 2685 break; 2686 } 2687 2688 /* 2689 * The second set of opcodes represents 'actions', 2690 * i.e. the terminal part of a rule once the packet 2691 * matches all previous patterns. 2692 * Typically there is only one action for each rule, 2693 * and the opcode is stored at the end of the rule 2694 * (but there are exceptions -- see below). 2695 * 2696 * In general, here we set retval and terminate the 2697 * outer loop (would be a 'break 3' in some language, 2698 * but we need to set l=0, done=1) 2699 * 2700 * Exceptions: 2701 * O_COUNT and O_SKIPTO actions: 2702 * instead of terminating, we jump to the next rule 2703 * (setting l=0), or to the SKIPTO target (setting 2704 * f/f_len, cmd and l as needed), respectively. 2705 * 2706 * O_TAG, O_LOG and O_ALTQ action parameters: 2707 * perform some action and set match = 1; 2708 * 2709 * O_LIMIT and O_KEEP_STATE: these opcodes are 2710 * not real 'actions', and are stored right 2711 * before the 'action' part of the rule (one 2712 * exception is O_SKIP_ACTION which could be 2713 * between these opcodes and 'action' one). 2714 * These opcodes try to install an entry in the 2715 * state tables; if successful, we continue with 2716 * the next opcode (match=1; break;), otherwise 2717 * the packet must be dropped (set retval, 2718 * break loops with l=0, done=1) 2719 * 2720 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2721 * cause a lookup of the state table, and a jump 2722 * to the 'action' part of the parent rule 2723 * if an entry is found, or 2724 * (CHECK_STATE only) a jump to the next rule if 2725 * the entry is not found. 2726 * The result of the lookup is cached so that 2727 * further instances of these opcodes become NOPs. 2728 * The jump to the next rule is done by setting 2729 * l=0, cmdlen=0. 2730 * 2731 * O_SKIP_ACTION: this opcode is not a real 'action' 2732 * either, and is stored right before the 'action' 2733 * part of the rule, right after the O_KEEP_STATE 2734 * opcode. It causes match failure so the real 2735 * 'action' could be executed only if the rule 2736 * is checked via dynamic rule from the state 2737 * table, as in such case execution starts 2738 * from the true 'action' opcode directly. 2739 * 2740 */ 2741 case O_LIMIT: 2742 case O_KEEP_STATE: 2743 if (ipfw_dyn_install_state(chain, f, 2744 (ipfw_insn_limit *)cmd, args, ulp, 2745 pktlen, &dyn_info, tablearg)) { 2746 /* error or limit violation */ 2747 retval = IP_FW_DENY; 2748 l = 0; /* exit inner loop */ 2749 done = 1; /* exit outer loop */ 2750 } 2751 match = 1; 2752 break; 2753 2754 case O_PROBE_STATE: 2755 case O_CHECK_STATE: 2756 /* 2757 * dynamic rules are checked at the first 2758 * keep-state or check-state occurrence, 2759 * with the result being stored in dyn_info. 2760 * The compiler introduces a PROBE_STATE 2761 * instruction for us when we have a 2762 * KEEP_STATE (because PROBE_STATE needs 2763 * to be run first). 2764 */ 2765 if (DYN_LOOKUP_NEEDED(&dyn_info, cmd) && 2766 (q = ipfw_dyn_lookup_state(args, ulp, 2767 pktlen, cmd, &dyn_info)) != NULL) { 2768 /* 2769 * Found dynamic entry, jump to the 2770 * 'action' part of the parent rule 2771 * by setting f, cmd, l and clearing 2772 * cmdlen. 2773 */ 2774 f = q; 2775 f_pos = dyn_info.f_pos; 2776 cmd = ACTION_PTR(f); 2777 l = f->cmd_len - f->act_ofs; 2778 cmdlen = 0; 2779 match = 1; 2780 break; 2781 } 2782 /* 2783 * Dynamic entry not found. If CHECK_STATE, 2784 * skip to next rule, if PROBE_STATE just 2785 * ignore and continue with next opcode. 2786 */ 2787 if (cmd->opcode == O_CHECK_STATE) 2788 l = 0; /* exit inner loop */ 2789 match = 1; 2790 break; 2791 2792 case O_SKIP_ACTION: 2793 match = 0; /* skip to the next rule */ 2794 l = 0; /* exit inner loop */ 2795 break; 2796 2797 case O_ACCEPT: 2798 retval = 0; /* accept */ 2799 l = 0; /* exit inner loop */ 2800 done = 1; /* exit outer loop */ 2801 break; 2802 2803 case O_PIPE: 2804 case O_QUEUE: 2805 set_match(args, f_pos, chain); 2806 args->rule.info = TARG(cmd->arg1, pipe); 2807 if (cmd->opcode == O_PIPE) 2808 args->rule.info |= IPFW_IS_PIPE; 2809 if (V_fw_one_pass) 2810 args->rule.info |= IPFW_ONEPASS; 2811 retval = IP_FW_DUMMYNET; 2812 l = 0; /* exit inner loop */ 2813 done = 1; /* exit outer loop */ 2814 break; 2815 2816 case O_DIVERT: 2817 case O_TEE: 2818 if (args->flags & IPFW_ARGS_ETHER) 2819 break; /* not on layer 2 */ 2820 /* otherwise this is terminal */ 2821 l = 0; /* exit inner loop */ 2822 done = 1; /* exit outer loop */ 2823 retval = (cmd->opcode == O_DIVERT) ? 2824 IP_FW_DIVERT : IP_FW_TEE; 2825 set_match(args, f_pos, chain); 2826 args->rule.info = TARG(cmd->arg1, divert); 2827 break; 2828 2829 case O_COUNT: 2830 IPFW_INC_RULE_COUNTER(f, pktlen); 2831 l = 0; /* exit inner loop */ 2832 break; 2833 2834 case O_SKIPTO: 2835 IPFW_INC_RULE_COUNTER(f, pktlen); 2836 f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0); 2837 /* 2838 * Skip disabled rules, and re-enter 2839 * the inner loop with the correct 2840 * f_pos, f, l and cmd. 2841 * Also clear cmdlen and skip_or 2842 */ 2843 for (; f_pos < chain->n_rules - 1 && 2844 (V_set_disable & 2845 (1 << chain->map[f_pos]->set)); 2846 f_pos++) 2847 ; 2848 /* Re-enter the inner loop at the skipto rule. */ 2849 f = chain->map[f_pos]; 2850 l = f->cmd_len; 2851 cmd = f->cmd; 2852 match = 1; 2853 cmdlen = 0; 2854 skip_or = 0; 2855 continue; 2856 break; /* not reached */ 2857 2858 case O_CALLRETURN: { 2859 /* 2860 * Implementation of `subroutine' call/return, 2861 * in the stack carried in an mbuf tag. This 2862 * is different from `skipto' in that any call 2863 * address is possible (`skipto' must prevent 2864 * backward jumps to avoid endless loops). 2865 * We have `return' action when F_NOT flag is 2866 * present. The `m_tag_id' field is used as 2867 * stack pointer. 2868 */ 2869 struct m_tag *mtag; 2870 uint16_t jmpto, *stack; 2871 2872 #define IS_CALL ((cmd->len & F_NOT) == 0) 2873 #define IS_RETURN ((cmd->len & F_NOT) != 0) 2874 /* 2875 * Hand-rolled version of m_tag_locate() with 2876 * wildcard `type'. 2877 * If not already tagged, allocate new tag. 2878 */ 2879 mtag = m_tag_first(m); 2880 while (mtag != NULL) { 2881 if (mtag->m_tag_cookie == 2882 MTAG_IPFW_CALL) 2883 break; 2884 mtag = m_tag_next(m, mtag); 2885 } 2886 if (mtag == NULL && IS_CALL) { 2887 mtag = m_tag_alloc(MTAG_IPFW_CALL, 0, 2888 IPFW_CALLSTACK_SIZE * 2889 sizeof(uint16_t), M_NOWAIT); 2890 if (mtag != NULL) 2891 m_tag_prepend(m, mtag); 2892 } 2893 2894 /* 2895 * On error both `call' and `return' just 2896 * continue with next rule. 2897 */ 2898 if (IS_RETURN && (mtag == NULL || 2899 mtag->m_tag_id == 0)) { 2900 l = 0; /* exit inner loop */ 2901 break; 2902 } 2903 if (IS_CALL && (mtag == NULL || 2904 mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) { 2905 printf("ipfw: call stack error, " 2906 "go to next rule\n"); 2907 l = 0; /* exit inner loop */ 2908 break; 2909 } 2910 2911 IPFW_INC_RULE_COUNTER(f, pktlen); 2912 stack = (uint16_t *)(mtag + 1); 2913 2914 /* 2915 * The `call' action may use cached f_pos 2916 * (in f->next_rule), whose version is written 2917 * in f->next_rule. 2918 * The `return' action, however, doesn't have 2919 * fixed jump address in cmd->arg1 and can't use 2920 * cache. 2921 */ 2922 if (IS_CALL) { 2923 stack[mtag->m_tag_id] = f->rulenum; 2924 mtag->m_tag_id++; 2925 f_pos = JUMP(chain, f, cmd->arg1, 2926 tablearg, 1); 2927 } else { /* `return' action */ 2928 mtag->m_tag_id--; 2929 jmpto = stack[mtag->m_tag_id] + 1; 2930 f_pos = ipfw_find_rule(chain, jmpto, 0); 2931 } 2932 2933 /* 2934 * Skip disabled rules, and re-enter 2935 * the inner loop with the correct 2936 * f_pos, f, l and cmd. 2937 * Also clear cmdlen and skip_or 2938 */ 2939 for (; f_pos < chain->n_rules - 1 && 2940 (V_set_disable & 2941 (1 << chain->map[f_pos]->set)); f_pos++) 2942 ; 2943 /* Re-enter the inner loop at the dest rule. */ 2944 f = chain->map[f_pos]; 2945 l = f->cmd_len; 2946 cmd = f->cmd; 2947 cmdlen = 0; 2948 skip_or = 0; 2949 continue; 2950 break; /* NOTREACHED */ 2951 } 2952 #undef IS_CALL 2953 #undef IS_RETURN 2954 2955 case O_REJECT: 2956 /* 2957 * Drop the packet and send a reject notice 2958 * if the packet is not ICMP (or is an ICMP 2959 * query), and it is not multicast/broadcast. 2960 */ 2961 if (hlen > 0 && is_ipv4 && offset == 0 && 2962 (proto != IPPROTO_ICMP || 2963 is_icmp_query(ICMP(ulp))) && 2964 !(m->m_flags & (M_BCAST|M_MCAST)) && 2965 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 2966 send_reject(args, cmd->arg1, iplen, ip); 2967 m = args->m; 2968 } 2969 /* FALLTHROUGH */ 2970 #ifdef INET6 2971 case O_UNREACH6: 2972 if (hlen > 0 && is_ipv6 && 2973 ((offset & IP6F_OFF_MASK) == 0) && 2974 (proto != IPPROTO_ICMPV6 || 2975 (is_icmp6_query(icmp6_type) == 1)) && 2976 !(m->m_flags & (M_BCAST|M_MCAST)) && 2977 !IN6_IS_ADDR_MULTICAST( 2978 &args->f_id.dst_ip6)) { 2979 send_reject6(args, 2980 cmd->opcode == O_REJECT ? 2981 map_icmp_unreach(cmd->arg1): 2982 cmd->arg1, hlen, 2983 (struct ip6_hdr *)ip); 2984 m = args->m; 2985 } 2986 /* FALLTHROUGH */ 2987 #endif 2988 case O_DENY: 2989 retval = IP_FW_DENY; 2990 l = 0; /* exit inner loop */ 2991 done = 1; /* exit outer loop */ 2992 break; 2993 2994 case O_FORWARD_IP: 2995 if (args->flags & IPFW_ARGS_ETHER) 2996 break; /* not valid on layer2 pkts */ 2997 if (q != f || 2998 dyn_info.direction == MATCH_FORWARD) { 2999 struct sockaddr_in *sa; 3000 3001 sa = &(((ipfw_insn_sa *)cmd)->sa); 3002 if (sa->sin_addr.s_addr == INADDR_ANY) { 3003 #ifdef INET6 3004 /* 3005 * We use O_FORWARD_IP opcode for 3006 * fwd rule with tablearg, but tables 3007 * now support IPv6 addresses. And 3008 * when we are inspecting IPv6 packet, 3009 * we can use nh6 field from 3010 * table_value as next_hop6 address. 3011 */ 3012 if (is_ipv6) { 3013 struct ip_fw_nh6 *nh6; 3014 3015 args->flags |= IPFW_ARGS_NH6; 3016 nh6 = &args->hopstore6; 3017 nh6->sin6_addr = TARG_VAL( 3018 chain, tablearg, nh6); 3019 nh6->sin6_port = sa->sin_port; 3020 nh6->sin6_scope_id = TARG_VAL( 3021 chain, tablearg, zoneid); 3022 } else 3023 #endif 3024 { 3025 args->flags |= IPFW_ARGS_NH4; 3026 args->hopstore.sin_port = 3027 sa->sin_port; 3028 sa = &args->hopstore; 3029 sa->sin_family = AF_INET; 3030 sa->sin_len = sizeof(*sa); 3031 sa->sin_addr.s_addr = htonl( 3032 TARG_VAL(chain, tablearg, 3033 nh4)); 3034 } 3035 } else { 3036 args->flags |= IPFW_ARGS_NH4PTR; 3037 args->next_hop = sa; 3038 } 3039 } 3040 retval = IP_FW_PASS; 3041 l = 0; /* exit inner loop */ 3042 done = 1; /* exit outer loop */ 3043 break; 3044 3045 #ifdef INET6 3046 case O_FORWARD_IP6: 3047 if (args->flags & IPFW_ARGS_ETHER) 3048 break; /* not valid on layer2 pkts */ 3049 if (q != f || 3050 dyn_info.direction == MATCH_FORWARD) { 3051 struct sockaddr_in6 *sin6; 3052 3053 sin6 = &(((ipfw_insn_sa6 *)cmd)->sa); 3054 args->flags |= IPFW_ARGS_NH6PTR; 3055 args->next_hop6 = sin6; 3056 } 3057 retval = IP_FW_PASS; 3058 l = 0; /* exit inner loop */ 3059 done = 1; /* exit outer loop */ 3060 break; 3061 #endif 3062 3063 case O_NETGRAPH: 3064 case O_NGTEE: 3065 set_match(args, f_pos, chain); 3066 args->rule.info = TARG(cmd->arg1, netgraph); 3067 if (V_fw_one_pass) 3068 args->rule.info |= IPFW_ONEPASS; 3069 retval = (cmd->opcode == O_NETGRAPH) ? 3070 IP_FW_NETGRAPH : IP_FW_NGTEE; 3071 l = 0; /* exit inner loop */ 3072 done = 1; /* exit outer loop */ 3073 break; 3074 3075 case O_SETFIB: { 3076 uint32_t fib; 3077 3078 IPFW_INC_RULE_COUNTER(f, pktlen); 3079 fib = TARG(cmd->arg1, fib) & 0x7FFF; 3080 if (fib >= rt_numfibs) 3081 fib = 0; 3082 M_SETFIB(m, fib); 3083 args->f_id.fib = fib; /* XXX */ 3084 l = 0; /* exit inner loop */ 3085 break; 3086 } 3087 3088 case O_SETDSCP: { 3089 uint16_t code; 3090 3091 code = TARG(cmd->arg1, dscp) & 0x3F; 3092 l = 0; /* exit inner loop */ 3093 if (is_ipv4) { 3094 uint16_t old; 3095 3096 old = *(uint16_t *)ip; 3097 ip->ip_tos = (code << 2) | 3098 (ip->ip_tos & 0x03); 3099 ip->ip_sum = cksum_adjust(ip->ip_sum, 3100 old, *(uint16_t *)ip); 3101 } else if (is_ipv6) { 3102 uint8_t *v; 3103 3104 v = &((struct ip6_hdr *)ip)->ip6_vfc; 3105 *v = (*v & 0xF0) | (code >> 2); 3106 v++; 3107 *v = (*v & 0x3F) | ((code & 0x03) << 6); 3108 } else 3109 break; 3110 3111 IPFW_INC_RULE_COUNTER(f, pktlen); 3112 break; 3113 } 3114 3115 case O_NAT: 3116 l = 0; /* exit inner loop */ 3117 done = 1; /* exit outer loop */ 3118 /* 3119 * Ensure that we do not invoke NAT handler for 3120 * non IPv4 packets. Libalias expects only IPv4. 3121 */ 3122 if (!is_ipv4 || !IPFW_NAT_LOADED) { 3123 retval = IP_FW_DENY; 3124 break; 3125 } 3126 3127 struct cfg_nat *t; 3128 int nat_id; 3129 3130 args->rule.info = 0; 3131 set_match(args, f_pos, chain); 3132 /* Check if this is 'global' nat rule */ 3133 if (cmd->arg1 == IP_FW_NAT44_GLOBAL) { 3134 retval = ipfw_nat_ptr(args, NULL, m); 3135 break; 3136 } 3137 t = ((ipfw_insn_nat *)cmd)->nat; 3138 if (t == NULL) { 3139 nat_id = TARG(cmd->arg1, nat); 3140 t = (*lookup_nat_ptr)(&chain->nat, nat_id); 3141 3142 if (t == NULL) { 3143 retval = IP_FW_DENY; 3144 break; 3145 } 3146 if (cmd->arg1 != IP_FW_TARG) 3147 ((ipfw_insn_nat *)cmd)->nat = t; 3148 } 3149 retval = ipfw_nat_ptr(args, t, m); 3150 break; 3151 3152 case O_REASS: { 3153 int ip_off; 3154 3155 l = 0; /* in any case exit inner loop */ 3156 if (is_ipv6) /* IPv6 is not supported yet */ 3157 break; 3158 IPFW_INC_RULE_COUNTER(f, pktlen); 3159 ip_off = ntohs(ip->ip_off); 3160 3161 /* if not fragmented, go to next rule */ 3162 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0) 3163 break; 3164 3165 args->m = m = ip_reass(m); 3166 3167 /* 3168 * do IP header checksum fixup. 3169 */ 3170 if (m == NULL) { /* fragment got swallowed */ 3171 retval = IP_FW_DENY; 3172 } else { /* good, packet complete */ 3173 int hlen; 3174 3175 ip = mtod(m, struct ip *); 3176 hlen = ip->ip_hl << 2; 3177 ip->ip_sum = 0; 3178 if (hlen == sizeof(struct ip)) 3179 ip->ip_sum = in_cksum_hdr(ip); 3180 else 3181 ip->ip_sum = in_cksum(m, hlen); 3182 retval = IP_FW_REASS; 3183 args->rule.info = 0; 3184 set_match(args, f_pos, chain); 3185 } 3186 done = 1; /* exit outer loop */ 3187 break; 3188 } 3189 case O_EXTERNAL_ACTION: 3190 l = 0; /* in any case exit inner loop */ 3191 retval = ipfw_run_eaction(chain, args, 3192 cmd, &done); 3193 /* 3194 * If both @retval and @done are zero, 3195 * consider this as rule matching and 3196 * update counters. 3197 */ 3198 if (retval == 0 && done == 0) { 3199 IPFW_INC_RULE_COUNTER(f, pktlen); 3200 /* 3201 * Reset the result of the last 3202 * dynamic state lookup. 3203 * External action can change 3204 * @args content, and it may be 3205 * used for new state lookup later. 3206 */ 3207 DYN_INFO_INIT(&dyn_info); 3208 } 3209 break; 3210 3211 default: 3212 panic("-- unknown opcode %d\n", cmd->opcode); 3213 } /* end of switch() on opcodes */ 3214 /* 3215 * if we get here with l=0, then match is irrelevant. 3216 */ 3217 3218 if (cmd->len & F_NOT) 3219 match = !match; 3220 3221 if (match) { 3222 if (cmd->len & F_OR) 3223 skip_or = 1; 3224 } else { 3225 if (!(cmd->len & F_OR)) /* not an OR block, */ 3226 break; /* try next rule */ 3227 } 3228 3229 } /* end of inner loop, scan opcodes */ 3230 #undef PULLUP_LEN 3231 #undef PULLUP_LEN_LOCKED 3232 3233 if (done) 3234 break; 3235 3236 /* next_rule:; */ /* try next rule */ 3237 3238 } /* end of outer for, scan rules */ 3239 3240 if (done) { 3241 struct ip_fw *rule = chain->map[f_pos]; 3242 /* Update statistics */ 3243 IPFW_INC_RULE_COUNTER(rule, pktlen); 3244 } else { 3245 retval = IP_FW_DENY; 3246 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3247 } 3248 IPFW_PF_RUNLOCK(chain); 3249 #ifdef __FreeBSD__ 3250 if (ucred_cache != NULL) 3251 crfree(ucred_cache); 3252 #endif 3253 return (retval); 3254 3255 pullup_failed: 3256 if (V_fw_verbose) 3257 printf("ipfw: pullup failed\n"); 3258 return (IP_FW_DENY); 3259 } 3260 3261 /* 3262 * Set maximum number of tables that can be used in given VNET ipfw instance. 3263 */ 3264 #ifdef SYSCTL_NODE 3265 static int 3266 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS) 3267 { 3268 int error; 3269 unsigned int ntables; 3270 3271 ntables = V_fw_tables_max; 3272 3273 error = sysctl_handle_int(oidp, &ntables, 0, req); 3274 /* Read operation or some error */ 3275 if ((error != 0) || (req->newptr == NULL)) 3276 return (error); 3277 3278 return (ipfw_resize_tables(&V_layer3_chain, ntables)); 3279 } 3280 3281 /* 3282 * Switches table namespace between global and per-set. 3283 */ 3284 static int 3285 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS) 3286 { 3287 int error; 3288 unsigned int sets; 3289 3290 sets = V_fw_tables_sets; 3291 3292 error = sysctl_handle_int(oidp, &sets, 0, req); 3293 /* Read operation or some error */ 3294 if ((error != 0) || (req->newptr == NULL)) 3295 return (error); 3296 3297 return (ipfw_switch_tables_namespace(&V_layer3_chain, sets)); 3298 } 3299 #endif 3300 3301 /* 3302 * Module and VNET glue 3303 */ 3304 3305 /* 3306 * Stuff that must be initialised only on boot or module load 3307 */ 3308 static int 3309 ipfw_init(void) 3310 { 3311 int error = 0; 3312 3313 /* 3314 * Only print out this stuff the first time around, 3315 * when called from the sysinit code. 3316 */ 3317 printf("ipfw2 " 3318 #ifdef INET6 3319 "(+ipv6) " 3320 #endif 3321 "initialized, divert %s, nat %s, " 3322 "default to %s, logging ", 3323 #ifdef IPDIVERT 3324 "enabled", 3325 #else 3326 "loadable", 3327 #endif 3328 #ifdef IPFIREWALL_NAT 3329 "enabled", 3330 #else 3331 "loadable", 3332 #endif 3333 default_to_accept ? "accept" : "deny"); 3334 3335 /* 3336 * Note: V_xxx variables can be accessed here but the vnet specific 3337 * initializer may not have been called yet for the VIMAGE case. 3338 * Tuneables will have been processed. We will print out values for 3339 * the default vnet. 3340 * XXX This should all be rationalized AFTER 8.0 3341 */ 3342 if (V_fw_verbose == 0) 3343 printf("disabled\n"); 3344 else if (V_verbose_limit == 0) 3345 printf("unlimited\n"); 3346 else 3347 printf("limited to %d packets/entry by default\n", 3348 V_verbose_limit); 3349 3350 /* Check user-supplied table count for validness */ 3351 if (default_fw_tables > IPFW_TABLES_MAX) 3352 default_fw_tables = IPFW_TABLES_MAX; 3353 3354 ipfw_init_sopt_handler(); 3355 ipfw_init_obj_rewriter(); 3356 ipfw_iface_init(); 3357 return (error); 3358 } 3359 3360 /* 3361 * Called for the removal of the last instance only on module unload. 3362 */ 3363 static void 3364 ipfw_destroy(void) 3365 { 3366 3367 ipfw_iface_destroy(); 3368 ipfw_destroy_sopt_handler(); 3369 ipfw_destroy_obj_rewriter(); 3370 printf("IP firewall unloaded\n"); 3371 } 3372 3373 /* 3374 * Stuff that must be initialized for every instance 3375 * (including the first of course). 3376 */ 3377 static int 3378 vnet_ipfw_init(const void *unused) 3379 { 3380 int error, first; 3381 struct ip_fw *rule = NULL; 3382 struct ip_fw_chain *chain; 3383 3384 chain = &V_layer3_chain; 3385 3386 first = IS_DEFAULT_VNET(curvnet) ? 1 : 0; 3387 3388 /* First set up some values that are compile time options */ 3389 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 3390 V_fw_deny_unknown_exthdrs = 1; 3391 #ifdef IPFIREWALL_VERBOSE 3392 V_fw_verbose = 1; 3393 #endif 3394 #ifdef IPFIREWALL_VERBOSE_LIMIT 3395 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 3396 #endif 3397 #ifdef IPFIREWALL_NAT 3398 LIST_INIT(&chain->nat); 3399 #endif 3400 3401 /* Init shared services hash table */ 3402 ipfw_init_srv(chain); 3403 3404 ipfw_init_counters(); 3405 /* Set initial number of tables */ 3406 V_fw_tables_max = default_fw_tables; 3407 error = ipfw_init_tables(chain, first); 3408 if (error) { 3409 printf("ipfw2: setting up tables failed\n"); 3410 free(chain->map, M_IPFW); 3411 free(rule, M_IPFW); 3412 return (ENOSPC); 3413 } 3414 3415 IPFW_LOCK_INIT(chain); 3416 3417 /* fill and insert the default rule */ 3418 rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw)); 3419 rule->flags |= IPFW_RULE_NOOPT; 3420 rule->cmd_len = 1; 3421 rule->cmd[0].len = 1; 3422 rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; 3423 chain->default_rule = rule; 3424 ipfw_add_protected_rule(chain, rule, 0); 3425 3426 ipfw_dyn_init(chain); 3427 ipfw_eaction_init(chain, first); 3428 #ifdef LINEAR_SKIPTO 3429 ipfw_init_skipto_cache(chain); 3430 #endif 3431 ipfw_bpf_init(first); 3432 3433 /* First set up some values that are compile time options */ 3434 V_ipfw_vnet_ready = 1; /* Open for business */ 3435 3436 /* 3437 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6. 3438 * Even if the latter two fail we still keep the module alive 3439 * because the sockopt and layer2 paths are still useful. 3440 * ipfw[6]_hook return 0 on success, ENOENT on failure, 3441 * so we can ignore the exact return value and just set a flag. 3442 * 3443 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so 3444 * changes in the underlying (per-vnet) variables trigger 3445 * immediate hook()/unhook() calls. 3446 * In layer2 we have the same behaviour, except that V_ether_ipfw 3447 * is checked on each packet because there are no pfil hooks. 3448 */ 3449 V_ip_fw_ctl_ptr = ipfw_ctl3; 3450 error = ipfw_attach_hooks(); 3451 return (error); 3452 } 3453 3454 /* 3455 * Called for the removal of each instance. 3456 */ 3457 static int 3458 vnet_ipfw_uninit(const void *unused) 3459 { 3460 struct ip_fw *reap; 3461 struct ip_fw_chain *chain = &V_layer3_chain; 3462 int i, last; 3463 3464 V_ipfw_vnet_ready = 0; /* tell new callers to go away */ 3465 /* 3466 * disconnect from ipv4, ipv6, layer2 and sockopt. 3467 * Then grab, release and grab again the WLOCK so we make 3468 * sure the update is propagated and nobody will be in. 3469 */ 3470 ipfw_detach_hooks(); 3471 V_ip_fw_ctl_ptr = NULL; 3472 3473 last = IS_DEFAULT_VNET(curvnet) ? 1 : 0; 3474 3475 IPFW_UH_WLOCK(chain); 3476 IPFW_UH_WUNLOCK(chain); 3477 3478 ipfw_dyn_uninit(0); /* run the callout_drain */ 3479 3480 IPFW_UH_WLOCK(chain); 3481 3482 reap = NULL; 3483 IPFW_WLOCK(chain); 3484 for (i = 0; i < chain->n_rules; i++) 3485 ipfw_reap_add(chain, &reap, chain->map[i]); 3486 free(chain->map, M_IPFW); 3487 #ifdef LINEAR_SKIPTO 3488 ipfw_destroy_skipto_cache(chain); 3489 #endif 3490 IPFW_WUNLOCK(chain); 3491 IPFW_UH_WUNLOCK(chain); 3492 ipfw_destroy_tables(chain, last); 3493 ipfw_eaction_uninit(chain, last); 3494 if (reap != NULL) 3495 ipfw_reap_rules(reap); 3496 vnet_ipfw_iface_destroy(chain); 3497 ipfw_destroy_srv(chain); 3498 IPFW_LOCK_DESTROY(chain); 3499 ipfw_dyn_uninit(1); /* free the remaining parts */ 3500 ipfw_destroy_counters(); 3501 ipfw_bpf_uninit(last); 3502 return (0); 3503 } 3504 3505 /* 3506 * Module event handler. 3507 * In general we have the choice of handling most of these events by the 3508 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to 3509 * use the SYSINIT handlers as they are more capable of expressing the 3510 * flow of control during module and vnet operations, so this is just 3511 * a skeleton. Note there is no SYSINIT equivalent of the module 3512 * SHUTDOWN handler, but we don't have anything to do in that case anyhow. 3513 */ 3514 static int 3515 ipfw_modevent(module_t mod, int type, void *unused) 3516 { 3517 int err = 0; 3518 3519 switch (type) { 3520 case MOD_LOAD: 3521 /* Called once at module load or 3522 * system boot if compiled in. */ 3523 break; 3524 case MOD_QUIESCE: 3525 /* Called before unload. May veto unloading. */ 3526 break; 3527 case MOD_UNLOAD: 3528 /* Called during unload. */ 3529 break; 3530 case MOD_SHUTDOWN: 3531 /* Called during system shutdown. */ 3532 break; 3533 default: 3534 err = EOPNOTSUPP; 3535 break; 3536 } 3537 return err; 3538 } 3539 3540 static moduledata_t ipfwmod = { 3541 "ipfw", 3542 ipfw_modevent, 3543 0 3544 }; 3545 3546 /* Define startup order. */ 3547 #define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_FIREWALL 3548 #define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */ 3549 #define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */ 3550 #define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */ 3551 3552 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER); 3553 FEATURE(ipfw_ctl3, "ipfw new sockopt calls"); 3554 MODULE_VERSION(ipfw, 3); 3555 /* should declare some dependencies here */ 3556 3557 /* 3558 * Starting up. Done in order after ipfwmod() has been called. 3559 * VNET_SYSINIT is also called for each existing vnet and each new vnet. 3560 */ 3561 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 3562 ipfw_init, NULL); 3563 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 3564 vnet_ipfw_init, NULL); 3565 3566 /* 3567 * Closing up shop. These are done in REVERSE ORDER, but still 3568 * after ipfwmod() has been called. Not called on reboot. 3569 * VNET_SYSUNINIT is also called for each exiting vnet as it exits. 3570 * or when the module is unloaded. 3571 */ 3572 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 3573 ipfw_destroy, NULL); 3574 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 3575 vnet_ipfw_uninit, NULL); 3576 /* end of file */ 3577