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 match = (offset != 0); 1948 break; 1949 1950 case O_IN: /* "out" is "not in" */ 1951 match = (oif == NULL); 1952 break; 1953 1954 case O_LAYER2: 1955 match = (args->flags & IPFW_ARGS_ETHER); 1956 break; 1957 1958 case O_DIVERTED: 1959 if ((args->flags & IPFW_ARGS_REF) == 0) 1960 break; 1961 /* 1962 * For diverted packets, args->rule.info 1963 * contains the divert port (in host format) 1964 * reason and direction. 1965 */ 1966 match = ((args->rule.info & IPFW_IS_MASK) == 1967 IPFW_IS_DIVERT) && ( 1968 ((args->rule.info & IPFW_INFO_IN) ? 1969 1: 2) & cmd->arg1); 1970 break; 1971 1972 case O_PROTO: 1973 /* 1974 * We do not allow an arg of 0 so the 1975 * check of "proto" only suffices. 1976 */ 1977 match = (proto == cmd->arg1); 1978 break; 1979 1980 case O_IP_SRC: 1981 match = is_ipv4 && 1982 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1983 src_ip.s_addr); 1984 break; 1985 1986 case O_IP_DST_LOOKUP: 1987 { 1988 void *pkey; 1989 uint32_t vidx, key; 1990 uint16_t keylen; 1991 1992 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) { 1993 /* Determine lookup key type */ 1994 vidx = ((ipfw_insn_u32 *)cmd)->d[1]; 1995 if (vidx != 4 /* uid */ && 1996 vidx != 5 /* jail */ && 1997 is_ipv6 == 0 && is_ipv4 == 0) 1998 break; 1999 /* Determine key length */ 2000 if (vidx == 0 /* dst-ip */ || 2001 vidx == 1 /* src-ip */) 2002 keylen = is_ipv6 ? 2003 sizeof(struct in6_addr): 2004 sizeof(in_addr_t); 2005 else { 2006 keylen = sizeof(key); 2007 pkey = &key; 2008 } 2009 if (vidx == 0 /* dst-ip */) 2010 pkey = is_ipv4 ? (void *)&dst_ip: 2011 (void *)&args->f_id.dst_ip6; 2012 else if (vidx == 1 /* src-ip */) 2013 pkey = is_ipv4 ? (void *)&src_ip: 2014 (void *)&args->f_id.src_ip6; 2015 else if (vidx == 6 /* dscp */) { 2016 if (is_ipv4) 2017 key = ip->ip_tos >> 2; 2018 else { 2019 key = args->f_id.flow_id6; 2020 key = (key & 0x0f) << 2 | 2021 (key & 0xf000) >> 14; 2022 } 2023 key &= 0x3f; 2024 } else if (vidx == 2 /* dst-port */ || 2025 vidx == 3 /* src-port */) { 2026 /* Skip fragments */ 2027 if (offset != 0) 2028 break; 2029 /* Skip proto without ports */ 2030 if (proto != IPPROTO_TCP && 2031 proto != IPPROTO_UDP && 2032 proto != IPPROTO_UDPLITE && 2033 proto != IPPROTO_SCTP) 2034 break; 2035 if (vidx == 2 /* dst-port */) 2036 key = dst_port; 2037 else 2038 key = src_port; 2039 } 2040 #ifndef USERSPACE 2041 else if (vidx == 4 /* uid */ || 2042 vidx == 5 /* jail */) { 2043 check_uidgid( 2044 (ipfw_insn_u32 *)cmd, 2045 args, &ucred_lookup, 2046 #ifdef __FreeBSD__ 2047 &ucred_cache); 2048 if (vidx == 4 /* uid */) 2049 key = ucred_cache->cr_uid; 2050 else if (vidx == 5 /* jail */) 2051 key = ucred_cache->cr_prison->pr_id; 2052 #else /* !__FreeBSD__ */ 2053 (void *)&ucred_cache); 2054 if (vidx == 4 /* uid */) 2055 key = ucred_cache.uid; 2056 else if (vidx == 5 /* jail */) 2057 key = ucred_cache.xid; 2058 #endif /* !__FreeBSD__ */ 2059 } 2060 #endif /* !USERSPACE */ 2061 else 2062 break; 2063 match = ipfw_lookup_table(chain, 2064 cmd->arg1, keylen, pkey, &vidx); 2065 if (!match) 2066 break; 2067 tablearg = vidx; 2068 break; 2069 } 2070 /* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */ 2071 /* FALLTHROUGH */ 2072 } 2073 case O_IP_SRC_LOOKUP: 2074 { 2075 void *pkey; 2076 uint32_t vidx; 2077 uint16_t keylen; 2078 2079 if (is_ipv4) { 2080 keylen = sizeof(in_addr_t); 2081 if (cmd->opcode == O_IP_DST_LOOKUP) 2082 pkey = &dst_ip; 2083 else 2084 pkey = &src_ip; 2085 } else if (is_ipv6) { 2086 keylen = sizeof(struct in6_addr); 2087 if (cmd->opcode == O_IP_DST_LOOKUP) 2088 pkey = &args->f_id.dst_ip6; 2089 else 2090 pkey = &args->f_id.src_ip6; 2091 } else 2092 break; 2093 match = ipfw_lookup_table(chain, cmd->arg1, 2094 keylen, pkey, &vidx); 2095 if (!match) 2096 break; 2097 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) { 2098 match = ((ipfw_insn_u32 *)cmd)->d[0] == 2099 TARG_VAL(chain, vidx, tag); 2100 if (!match) 2101 break; 2102 } 2103 tablearg = vidx; 2104 break; 2105 } 2106 2107 case O_IP_FLOW_LOOKUP: 2108 { 2109 uint32_t v = 0; 2110 match = ipfw_lookup_table(chain, 2111 cmd->arg1, 0, &args->f_id, &v); 2112 if (!match) 2113 break; 2114 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2115 match = ((ipfw_insn_u32 *)cmd)->d[0] == 2116 TARG_VAL(chain, v, tag); 2117 if (match) 2118 tablearg = v; 2119 } 2120 break; 2121 case O_IP_SRC_MASK: 2122 case O_IP_DST_MASK: 2123 if (is_ipv4) { 2124 uint32_t a = 2125 (cmd->opcode == O_IP_DST_MASK) ? 2126 dst_ip.s_addr : src_ip.s_addr; 2127 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2128 int i = cmdlen-1; 2129 2130 for (; !match && i>0; i-= 2, p+= 2) 2131 match = (p[0] == (a & p[1])); 2132 } 2133 break; 2134 2135 case O_IP_SRC_ME: 2136 if (is_ipv4) { 2137 match = in_localip(src_ip); 2138 break; 2139 } 2140 #ifdef INET6 2141 /* FALLTHROUGH */ 2142 case O_IP6_SRC_ME: 2143 match = is_ipv6 && 2144 ipfw_localip6(&args->f_id.src_ip6); 2145 #endif 2146 break; 2147 2148 case O_IP_DST_SET: 2149 case O_IP_SRC_SET: 2150 if (is_ipv4) { 2151 u_int32_t *d = (u_int32_t *)(cmd+1); 2152 u_int32_t addr = 2153 cmd->opcode == O_IP_DST_SET ? 2154 args->f_id.dst_ip : 2155 args->f_id.src_ip; 2156 2157 if (addr < d[0]) 2158 break; 2159 addr -= d[0]; /* subtract base */ 2160 match = (addr < cmd->arg1) && 2161 ( d[ 1 + (addr>>5)] & 2162 (1<<(addr & 0x1f)) ); 2163 } 2164 break; 2165 2166 case O_IP_DST: 2167 match = is_ipv4 && 2168 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2169 dst_ip.s_addr); 2170 break; 2171 2172 case O_IP_DST_ME: 2173 if (is_ipv4) { 2174 match = in_localip(dst_ip); 2175 break; 2176 } 2177 #ifdef INET6 2178 /* FALLTHROUGH */ 2179 case O_IP6_DST_ME: 2180 match = is_ipv6 && 2181 ipfw_localip6(&args->f_id.dst_ip6); 2182 #endif 2183 break; 2184 2185 2186 case O_IP_SRCPORT: 2187 case O_IP_DSTPORT: 2188 /* 2189 * offset == 0 && proto != 0 is enough 2190 * to guarantee that we have a 2191 * packet with port info. 2192 */ 2193 if ((proto == IPPROTO_UDP || 2194 proto == IPPROTO_UDPLITE || 2195 proto == IPPROTO_TCP || 2196 proto == IPPROTO_SCTP) && offset == 0) { 2197 u_int16_t x = 2198 (cmd->opcode == O_IP_SRCPORT) ? 2199 src_port : dst_port ; 2200 u_int16_t *p = 2201 ((ipfw_insn_u16 *)cmd)->ports; 2202 int i; 2203 2204 for (i = cmdlen - 1; !match && i>0; 2205 i--, p += 2) 2206 match = (x>=p[0] && x<=p[1]); 2207 } 2208 break; 2209 2210 case O_ICMPTYPE: 2211 match = (offset == 0 && proto==IPPROTO_ICMP && 2212 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2213 break; 2214 2215 #ifdef INET6 2216 case O_ICMP6TYPE: 2217 match = is_ipv6 && offset == 0 && 2218 proto==IPPROTO_ICMPV6 && 2219 icmp6type_match( 2220 ICMP6(ulp)->icmp6_type, 2221 (ipfw_insn_u32 *)cmd); 2222 break; 2223 #endif /* INET6 */ 2224 2225 case O_IPOPT: 2226 match = (is_ipv4 && 2227 ipopts_match(ip, cmd) ); 2228 break; 2229 2230 case O_IPVER: 2231 match = ((is_ipv4 || is_ipv6) && 2232 cmd->arg1 == ip->ip_v); 2233 break; 2234 2235 case O_IPID: 2236 case O_IPTTL: 2237 if (!is_ipv4) 2238 break; 2239 case O_IPLEN: 2240 { /* only for IP packets */ 2241 uint16_t x; 2242 uint16_t *p; 2243 int i; 2244 2245 if (cmd->opcode == O_IPLEN) 2246 x = iplen; 2247 else if (cmd->opcode == O_IPTTL) 2248 x = ip->ip_ttl; 2249 else /* must be IPID */ 2250 x = ntohs(ip->ip_id); 2251 if (cmdlen == 1) { 2252 match = (cmd->arg1 == x); 2253 break; 2254 } 2255 /* otherwise we have ranges */ 2256 p = ((ipfw_insn_u16 *)cmd)->ports; 2257 i = cmdlen - 1; 2258 for (; !match && i>0; i--, p += 2) 2259 match = (x >= p[0] && x <= p[1]); 2260 } 2261 break; 2262 2263 case O_IPPRECEDENCE: 2264 match = (is_ipv4 && 2265 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2266 break; 2267 2268 case O_IPTOS: 2269 match = (is_ipv4 && 2270 flags_match(cmd, ip->ip_tos)); 2271 break; 2272 2273 case O_DSCP: 2274 { 2275 uint32_t *p; 2276 uint16_t x; 2277 2278 p = ((ipfw_insn_u32 *)cmd)->d; 2279 2280 if (is_ipv4) 2281 x = ip->ip_tos >> 2; 2282 else if (is_ipv6) { 2283 uint8_t *v; 2284 v = &((struct ip6_hdr *)ip)->ip6_vfc; 2285 x = (*v & 0x0F) << 2; 2286 v++; 2287 x |= *v >> 6; 2288 } else 2289 break; 2290 2291 /* DSCP bitmask is stored as low_u32 high_u32 */ 2292 if (x >= 32) 2293 match = *(p + 1) & (1 << (x - 32)); 2294 else 2295 match = *p & (1 << x); 2296 } 2297 break; 2298 2299 case O_TCPDATALEN: 2300 if (proto == IPPROTO_TCP && offset == 0) { 2301 struct tcphdr *tcp; 2302 uint16_t x; 2303 uint16_t *p; 2304 int i; 2305 #ifdef INET6 2306 if (is_ipv6) { 2307 struct ip6_hdr *ip6; 2308 2309 ip6 = (struct ip6_hdr *)ip; 2310 if (ip6->ip6_plen == 0) { 2311 /* 2312 * Jumbo payload is not 2313 * supported by this 2314 * opcode. 2315 */ 2316 break; 2317 } 2318 x = iplen - hlen; 2319 } else 2320 #endif /* INET6 */ 2321 x = iplen - (ip->ip_hl << 2); 2322 tcp = TCP(ulp); 2323 x -= tcp->th_off << 2; 2324 if (cmdlen == 1) { 2325 match = (cmd->arg1 == x); 2326 break; 2327 } 2328 /* otherwise we have ranges */ 2329 p = ((ipfw_insn_u16 *)cmd)->ports; 2330 i = cmdlen - 1; 2331 for (; !match && i>0; i--, p += 2) 2332 match = (x >= p[0] && x <= p[1]); 2333 } 2334 break; 2335 2336 case O_TCPFLAGS: 2337 match = (proto == IPPROTO_TCP && offset == 0 && 2338 flags_match(cmd, TCP(ulp)->th_flags)); 2339 break; 2340 2341 case O_TCPOPTS: 2342 if (proto == IPPROTO_TCP && offset == 0 && ulp){ 2343 PULLUP_LEN_LOCKED(hlen, ulp, 2344 (TCP(ulp)->th_off << 2)); 2345 match = tcpopts_match(TCP(ulp), cmd); 2346 } 2347 break; 2348 2349 case O_TCPSEQ: 2350 match = (proto == IPPROTO_TCP && offset == 0 && 2351 ((ipfw_insn_u32 *)cmd)->d[0] == 2352 TCP(ulp)->th_seq); 2353 break; 2354 2355 case O_TCPACK: 2356 match = (proto == IPPROTO_TCP && offset == 0 && 2357 ((ipfw_insn_u32 *)cmd)->d[0] == 2358 TCP(ulp)->th_ack); 2359 break; 2360 2361 case O_TCPMSS: 2362 if (proto == IPPROTO_TCP && 2363 (args->f_id._flags & TH_SYN) != 0 && 2364 ulp != NULL) { 2365 uint16_t mss, *p; 2366 int i; 2367 2368 PULLUP_LEN_LOCKED(hlen, ulp, 2369 (TCP(ulp)->th_off << 2)); 2370 if ((tcpopts_parse(TCP(ulp), &mss) & 2371 IP_FW_TCPOPT_MSS) == 0) 2372 break; 2373 if (cmdlen == 1) { 2374 match = (cmd->arg1 == mss); 2375 break; 2376 } 2377 /* Otherwise we have ranges. */ 2378 p = ((ipfw_insn_u16 *)cmd)->ports; 2379 i = cmdlen - 1; 2380 for (; !match && i > 0; i--, p += 2) 2381 match = (mss >= p[0] && 2382 mss <= p[1]); 2383 } 2384 break; 2385 2386 case O_TCPWIN: 2387 if (proto == IPPROTO_TCP && offset == 0) { 2388 uint16_t x; 2389 uint16_t *p; 2390 int i; 2391 2392 x = ntohs(TCP(ulp)->th_win); 2393 if (cmdlen == 1) { 2394 match = (cmd->arg1 == x); 2395 break; 2396 } 2397 /* Otherwise we have ranges. */ 2398 p = ((ipfw_insn_u16 *)cmd)->ports; 2399 i = cmdlen - 1; 2400 for (; !match && i > 0; i--, p += 2) 2401 match = (x >= p[0] && x <= p[1]); 2402 } 2403 break; 2404 2405 case O_ESTAB: 2406 /* reject packets which have SYN only */ 2407 /* XXX should i also check for TH_ACK ? */ 2408 match = (proto == IPPROTO_TCP && offset == 0 && 2409 (TCP(ulp)->th_flags & 2410 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2411 break; 2412 2413 case O_ALTQ: { 2414 struct pf_mtag *at; 2415 struct m_tag *mtag; 2416 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2417 2418 /* 2419 * ALTQ uses mbuf tags from another 2420 * packet filtering system - pf(4). 2421 * We allocate a tag in its format 2422 * and fill it in, pretending to be pf(4). 2423 */ 2424 match = 1; 2425 at = pf_find_mtag(m); 2426 if (at != NULL && at->qid != 0) 2427 break; 2428 mtag = m_tag_get(PACKET_TAG_PF, 2429 sizeof(struct pf_mtag), M_NOWAIT | M_ZERO); 2430 if (mtag == NULL) { 2431 /* 2432 * Let the packet fall back to the 2433 * default ALTQ. 2434 */ 2435 break; 2436 } 2437 m_tag_prepend(m, mtag); 2438 at = (struct pf_mtag *)(mtag + 1); 2439 at->qid = altq->qid; 2440 at->hdr = ip; 2441 break; 2442 } 2443 2444 case O_LOG: 2445 ipfw_log(chain, f, hlen, args, 2446 offset | ip6f_mf, tablearg, ip); 2447 match = 1; 2448 break; 2449 2450 case O_PROB: 2451 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2452 break; 2453 2454 case O_VERREVPATH: 2455 /* Outgoing packets automatically pass/match */ 2456 match = (args->flags & IPFW_ARGS_OUT || 2457 ( 2458 #ifdef INET6 2459 is_ipv6 ? 2460 verify_path6(&(args->f_id.src_ip6), 2461 iif, args->f_id.fib) : 2462 #endif 2463 verify_path(src_ip, iif, args->f_id.fib))); 2464 break; 2465 2466 case O_VERSRCREACH: 2467 /* Outgoing packets automatically pass/match */ 2468 match = (hlen > 0 && ((oif != NULL) || ( 2469 #ifdef INET6 2470 is_ipv6 ? 2471 verify_path6(&(args->f_id.src_ip6), 2472 NULL, args->f_id.fib) : 2473 #endif 2474 verify_path(src_ip, NULL, args->f_id.fib)))); 2475 break; 2476 2477 case O_ANTISPOOF: 2478 /* Outgoing packets automatically pass/match */ 2479 if (oif == NULL && hlen > 0 && 2480 ( (is_ipv4 && in_localaddr(src_ip)) 2481 #ifdef INET6 2482 || (is_ipv6 && 2483 in6_localaddr(&(args->f_id.src_ip6))) 2484 #endif 2485 )) 2486 match = 2487 #ifdef INET6 2488 is_ipv6 ? verify_path6( 2489 &(args->f_id.src_ip6), iif, 2490 args->f_id.fib) : 2491 #endif 2492 verify_path(src_ip, iif, 2493 args->f_id.fib); 2494 else 2495 match = 1; 2496 break; 2497 2498 case O_IPSEC: 2499 match = (m_tag_find(m, 2500 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2501 /* otherwise no match */ 2502 break; 2503 2504 #ifdef INET6 2505 case O_IP6_SRC: 2506 match = is_ipv6 && 2507 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2508 &((ipfw_insn_ip6 *)cmd)->addr6); 2509 break; 2510 2511 case O_IP6_DST: 2512 match = is_ipv6 && 2513 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2514 &((ipfw_insn_ip6 *)cmd)->addr6); 2515 break; 2516 case O_IP6_SRC_MASK: 2517 case O_IP6_DST_MASK: 2518 if (is_ipv6) { 2519 int i = cmdlen - 1; 2520 struct in6_addr p; 2521 struct in6_addr *d = 2522 &((ipfw_insn_ip6 *)cmd)->addr6; 2523 2524 for (; !match && i > 0; d += 2, 2525 i -= F_INSN_SIZE(struct in6_addr) 2526 * 2) { 2527 p = (cmd->opcode == 2528 O_IP6_SRC_MASK) ? 2529 args->f_id.src_ip6: 2530 args->f_id.dst_ip6; 2531 APPLY_MASK(&p, &d[1]); 2532 match = 2533 IN6_ARE_ADDR_EQUAL(&d[0], 2534 &p); 2535 } 2536 } 2537 break; 2538 2539 case O_FLOW6ID: 2540 match = is_ipv6 && 2541 flow6id_match(args->f_id.flow_id6, 2542 (ipfw_insn_u32 *) cmd); 2543 break; 2544 2545 case O_EXT_HDR: 2546 match = is_ipv6 && 2547 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2548 break; 2549 2550 case O_IP6: 2551 match = is_ipv6; 2552 break; 2553 #endif 2554 2555 case O_IP4: 2556 match = is_ipv4; 2557 break; 2558 2559 case O_TAG: { 2560 struct m_tag *mtag; 2561 uint32_t tag = TARG(cmd->arg1, tag); 2562 2563 /* Packet is already tagged with this tag? */ 2564 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 2565 2566 /* We have `untag' action when F_NOT flag is 2567 * present. And we must remove this mtag from 2568 * mbuf and reset `match' to zero (`match' will 2569 * be inversed later). 2570 * Otherwise we should allocate new mtag and 2571 * push it into mbuf. 2572 */ 2573 if (cmd->len & F_NOT) { /* `untag' action */ 2574 if (mtag != NULL) 2575 m_tag_delete(m, mtag); 2576 match = 0; 2577 } else { 2578 if (mtag == NULL) { 2579 mtag = m_tag_alloc( MTAG_IPFW, 2580 tag, 0, M_NOWAIT); 2581 if (mtag != NULL) 2582 m_tag_prepend(m, mtag); 2583 } 2584 match = 1; 2585 } 2586 break; 2587 } 2588 2589 case O_FIB: /* try match the specified fib */ 2590 if (args->f_id.fib == cmd->arg1) 2591 match = 1; 2592 break; 2593 2594 case O_SOCKARG: { 2595 #ifndef USERSPACE /* not supported in userspace */ 2596 struct inpcb *inp = args->inp; 2597 struct inpcbinfo *pi; 2598 2599 if (is_ipv6) /* XXX can we remove this ? */ 2600 break; 2601 2602 if (proto == IPPROTO_TCP) 2603 pi = &V_tcbinfo; 2604 else if (proto == IPPROTO_UDP) 2605 pi = &V_udbinfo; 2606 else if (proto == IPPROTO_UDPLITE) 2607 pi = &V_ulitecbinfo; 2608 else 2609 break; 2610 2611 /* 2612 * XXXRW: so_user_cookie should almost 2613 * certainly be inp_user_cookie? 2614 */ 2615 2616 /* For incoming packet, lookup up the 2617 inpcb using the src/dest ip/port tuple */ 2618 if (inp == NULL) { 2619 inp = in_pcblookup(pi, 2620 src_ip, htons(src_port), 2621 dst_ip, htons(dst_port), 2622 INPLOOKUP_RLOCKPCB, NULL); 2623 if (inp != NULL) { 2624 tablearg = 2625 inp->inp_socket->so_user_cookie; 2626 if (tablearg) 2627 match = 1; 2628 INP_RUNLOCK(inp); 2629 } 2630 } else { 2631 if (inp->inp_socket) { 2632 tablearg = 2633 inp->inp_socket->so_user_cookie; 2634 if (tablearg) 2635 match = 1; 2636 } 2637 } 2638 #endif /* !USERSPACE */ 2639 break; 2640 } 2641 2642 case O_TAGGED: { 2643 struct m_tag *mtag; 2644 uint32_t tag = TARG(cmd->arg1, tag); 2645 2646 if (cmdlen == 1) { 2647 match = m_tag_locate(m, MTAG_IPFW, 2648 tag, NULL) != NULL; 2649 break; 2650 } 2651 2652 /* we have ranges */ 2653 for (mtag = m_tag_first(m); 2654 mtag != NULL && !match; 2655 mtag = m_tag_next(m, mtag)) { 2656 uint16_t *p; 2657 int i; 2658 2659 if (mtag->m_tag_cookie != MTAG_IPFW) 2660 continue; 2661 2662 p = ((ipfw_insn_u16 *)cmd)->ports; 2663 i = cmdlen - 1; 2664 for(; !match && i > 0; i--, p += 2) 2665 match = 2666 mtag->m_tag_id >= p[0] && 2667 mtag->m_tag_id <= p[1]; 2668 } 2669 break; 2670 } 2671 2672 /* 2673 * The second set of opcodes represents 'actions', 2674 * i.e. the terminal part of a rule once the packet 2675 * matches all previous patterns. 2676 * Typically there is only one action for each rule, 2677 * and the opcode is stored at the end of the rule 2678 * (but there are exceptions -- see below). 2679 * 2680 * In general, here we set retval and terminate the 2681 * outer loop (would be a 'break 3' in some language, 2682 * but we need to set l=0, done=1) 2683 * 2684 * Exceptions: 2685 * O_COUNT and O_SKIPTO actions: 2686 * instead of terminating, we jump to the next rule 2687 * (setting l=0), or to the SKIPTO target (setting 2688 * f/f_len, cmd and l as needed), respectively. 2689 * 2690 * O_TAG, O_LOG and O_ALTQ action parameters: 2691 * perform some action and set match = 1; 2692 * 2693 * O_LIMIT and O_KEEP_STATE: these opcodes are 2694 * not real 'actions', and are stored right 2695 * before the 'action' part of the rule (one 2696 * exception is O_SKIP_ACTION which could be 2697 * between these opcodes and 'action' one). 2698 * These opcodes try to install an entry in the 2699 * state tables; if successful, we continue with 2700 * the next opcode (match=1; break;), otherwise 2701 * the packet must be dropped (set retval, 2702 * break loops with l=0, done=1) 2703 * 2704 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2705 * cause a lookup of the state table, and a jump 2706 * to the 'action' part of the parent rule 2707 * if an entry is found, or 2708 * (CHECK_STATE only) a jump to the next rule if 2709 * the entry is not found. 2710 * The result of the lookup is cached so that 2711 * further instances of these opcodes become NOPs. 2712 * The jump to the next rule is done by setting 2713 * l=0, cmdlen=0. 2714 * 2715 * O_SKIP_ACTION: this opcode is not a real 'action' 2716 * either, and is stored right before the 'action' 2717 * part of the rule, right after the O_KEEP_STATE 2718 * opcode. It causes match failure so the real 2719 * 'action' could be executed only if the rule 2720 * is checked via dynamic rule from the state 2721 * table, as in such case execution starts 2722 * from the true 'action' opcode directly. 2723 * 2724 */ 2725 case O_LIMIT: 2726 case O_KEEP_STATE: 2727 if (ipfw_dyn_install_state(chain, f, 2728 (ipfw_insn_limit *)cmd, args, ulp, 2729 pktlen, &dyn_info, tablearg)) { 2730 /* error or limit violation */ 2731 retval = IP_FW_DENY; 2732 l = 0; /* exit inner loop */ 2733 done = 1; /* exit outer loop */ 2734 } 2735 match = 1; 2736 break; 2737 2738 case O_PROBE_STATE: 2739 case O_CHECK_STATE: 2740 /* 2741 * dynamic rules are checked at the first 2742 * keep-state or check-state occurrence, 2743 * with the result being stored in dyn_info. 2744 * The compiler introduces a PROBE_STATE 2745 * instruction for us when we have a 2746 * KEEP_STATE (because PROBE_STATE needs 2747 * to be run first). 2748 */ 2749 if (DYN_LOOKUP_NEEDED(&dyn_info, cmd) && 2750 (q = ipfw_dyn_lookup_state(args, ulp, 2751 pktlen, cmd, &dyn_info)) != NULL) { 2752 /* 2753 * Found dynamic entry, jump to the 2754 * 'action' part of the parent rule 2755 * by setting f, cmd, l and clearing 2756 * cmdlen. 2757 */ 2758 f = q; 2759 f_pos = dyn_info.f_pos; 2760 cmd = ACTION_PTR(f); 2761 l = f->cmd_len - f->act_ofs; 2762 cmdlen = 0; 2763 match = 1; 2764 break; 2765 } 2766 /* 2767 * Dynamic entry not found. If CHECK_STATE, 2768 * skip to next rule, if PROBE_STATE just 2769 * ignore and continue with next opcode. 2770 */ 2771 if (cmd->opcode == O_CHECK_STATE) 2772 l = 0; /* exit inner loop */ 2773 match = 1; 2774 break; 2775 2776 case O_SKIP_ACTION: 2777 match = 0; /* skip to the next rule */ 2778 l = 0; /* exit inner loop */ 2779 break; 2780 2781 case O_ACCEPT: 2782 retval = 0; /* accept */ 2783 l = 0; /* exit inner loop */ 2784 done = 1; /* exit outer loop */ 2785 break; 2786 2787 case O_PIPE: 2788 case O_QUEUE: 2789 set_match(args, f_pos, chain); 2790 args->rule.info = TARG(cmd->arg1, pipe); 2791 if (cmd->opcode == O_PIPE) 2792 args->rule.info |= IPFW_IS_PIPE; 2793 if (V_fw_one_pass) 2794 args->rule.info |= IPFW_ONEPASS; 2795 retval = IP_FW_DUMMYNET; 2796 l = 0; /* exit inner loop */ 2797 done = 1; /* exit outer loop */ 2798 break; 2799 2800 case O_DIVERT: 2801 case O_TEE: 2802 if (args->flags & IPFW_ARGS_ETHER) 2803 break; /* not on layer 2 */ 2804 /* otherwise this is terminal */ 2805 l = 0; /* exit inner loop */ 2806 done = 1; /* exit outer loop */ 2807 retval = (cmd->opcode == O_DIVERT) ? 2808 IP_FW_DIVERT : IP_FW_TEE; 2809 set_match(args, f_pos, chain); 2810 args->rule.info = TARG(cmd->arg1, divert); 2811 break; 2812 2813 case O_COUNT: 2814 IPFW_INC_RULE_COUNTER(f, pktlen); 2815 l = 0; /* exit inner loop */ 2816 break; 2817 2818 case O_SKIPTO: 2819 IPFW_INC_RULE_COUNTER(f, pktlen); 2820 f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0); 2821 /* 2822 * Skip disabled rules, and re-enter 2823 * the inner loop with the correct 2824 * f_pos, f, l and cmd. 2825 * Also clear cmdlen and skip_or 2826 */ 2827 for (; f_pos < chain->n_rules - 1 && 2828 (V_set_disable & 2829 (1 << chain->map[f_pos]->set)); 2830 f_pos++) 2831 ; 2832 /* Re-enter the inner loop at the skipto rule. */ 2833 f = chain->map[f_pos]; 2834 l = f->cmd_len; 2835 cmd = f->cmd; 2836 match = 1; 2837 cmdlen = 0; 2838 skip_or = 0; 2839 continue; 2840 break; /* not reached */ 2841 2842 case O_CALLRETURN: { 2843 /* 2844 * Implementation of `subroutine' call/return, 2845 * in the stack carried in an mbuf tag. This 2846 * is different from `skipto' in that any call 2847 * address is possible (`skipto' must prevent 2848 * backward jumps to avoid endless loops). 2849 * We have `return' action when F_NOT flag is 2850 * present. The `m_tag_id' field is used as 2851 * stack pointer. 2852 */ 2853 struct m_tag *mtag; 2854 uint16_t jmpto, *stack; 2855 2856 #define IS_CALL ((cmd->len & F_NOT) == 0) 2857 #define IS_RETURN ((cmd->len & F_NOT) != 0) 2858 /* 2859 * Hand-rolled version of m_tag_locate() with 2860 * wildcard `type'. 2861 * If not already tagged, allocate new tag. 2862 */ 2863 mtag = m_tag_first(m); 2864 while (mtag != NULL) { 2865 if (mtag->m_tag_cookie == 2866 MTAG_IPFW_CALL) 2867 break; 2868 mtag = m_tag_next(m, mtag); 2869 } 2870 if (mtag == NULL && IS_CALL) { 2871 mtag = m_tag_alloc(MTAG_IPFW_CALL, 0, 2872 IPFW_CALLSTACK_SIZE * 2873 sizeof(uint16_t), M_NOWAIT); 2874 if (mtag != NULL) 2875 m_tag_prepend(m, mtag); 2876 } 2877 2878 /* 2879 * On error both `call' and `return' just 2880 * continue with next rule. 2881 */ 2882 if (IS_RETURN && (mtag == NULL || 2883 mtag->m_tag_id == 0)) { 2884 l = 0; /* exit inner loop */ 2885 break; 2886 } 2887 if (IS_CALL && (mtag == NULL || 2888 mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) { 2889 printf("ipfw: call stack error, " 2890 "go to next rule\n"); 2891 l = 0; /* exit inner loop */ 2892 break; 2893 } 2894 2895 IPFW_INC_RULE_COUNTER(f, pktlen); 2896 stack = (uint16_t *)(mtag + 1); 2897 2898 /* 2899 * The `call' action may use cached f_pos 2900 * (in f->next_rule), whose version is written 2901 * in f->next_rule. 2902 * The `return' action, however, doesn't have 2903 * fixed jump address in cmd->arg1 and can't use 2904 * cache. 2905 */ 2906 if (IS_CALL) { 2907 stack[mtag->m_tag_id] = f->rulenum; 2908 mtag->m_tag_id++; 2909 f_pos = JUMP(chain, f, cmd->arg1, 2910 tablearg, 1); 2911 } else { /* `return' action */ 2912 mtag->m_tag_id--; 2913 jmpto = stack[mtag->m_tag_id] + 1; 2914 f_pos = ipfw_find_rule(chain, jmpto, 0); 2915 } 2916 2917 /* 2918 * Skip disabled rules, and re-enter 2919 * the inner loop with the correct 2920 * f_pos, f, l and cmd. 2921 * Also clear cmdlen and skip_or 2922 */ 2923 for (; f_pos < chain->n_rules - 1 && 2924 (V_set_disable & 2925 (1 << chain->map[f_pos]->set)); f_pos++) 2926 ; 2927 /* Re-enter the inner loop at the dest rule. */ 2928 f = chain->map[f_pos]; 2929 l = f->cmd_len; 2930 cmd = f->cmd; 2931 cmdlen = 0; 2932 skip_or = 0; 2933 continue; 2934 break; /* NOTREACHED */ 2935 } 2936 #undef IS_CALL 2937 #undef IS_RETURN 2938 2939 case O_REJECT: 2940 /* 2941 * Drop the packet and send a reject notice 2942 * if the packet is not ICMP (or is an ICMP 2943 * query), and it is not multicast/broadcast. 2944 */ 2945 if (hlen > 0 && is_ipv4 && offset == 0 && 2946 (proto != IPPROTO_ICMP || 2947 is_icmp_query(ICMP(ulp))) && 2948 !(m->m_flags & (M_BCAST|M_MCAST)) && 2949 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 2950 send_reject(args, cmd->arg1, iplen, ip); 2951 m = args->m; 2952 } 2953 /* FALLTHROUGH */ 2954 #ifdef INET6 2955 case O_UNREACH6: 2956 if (hlen > 0 && is_ipv6 && 2957 ((offset & IP6F_OFF_MASK) == 0) && 2958 (proto != IPPROTO_ICMPV6 || 2959 (is_icmp6_query(icmp6_type) == 1)) && 2960 !(m->m_flags & (M_BCAST|M_MCAST)) && 2961 !IN6_IS_ADDR_MULTICAST( 2962 &args->f_id.dst_ip6)) { 2963 send_reject6(args, 2964 cmd->opcode == O_REJECT ? 2965 map_icmp_unreach(cmd->arg1): 2966 cmd->arg1, hlen, 2967 (struct ip6_hdr *)ip); 2968 m = args->m; 2969 } 2970 /* FALLTHROUGH */ 2971 #endif 2972 case O_DENY: 2973 retval = IP_FW_DENY; 2974 l = 0; /* exit inner loop */ 2975 done = 1; /* exit outer loop */ 2976 break; 2977 2978 case O_FORWARD_IP: 2979 if (args->flags & IPFW_ARGS_ETHER) 2980 break; /* not valid on layer2 pkts */ 2981 if (q != f || 2982 dyn_info.direction == MATCH_FORWARD) { 2983 struct sockaddr_in *sa; 2984 2985 sa = &(((ipfw_insn_sa *)cmd)->sa); 2986 if (sa->sin_addr.s_addr == INADDR_ANY) { 2987 #ifdef INET6 2988 /* 2989 * We use O_FORWARD_IP opcode for 2990 * fwd rule with tablearg, but tables 2991 * now support IPv6 addresses. And 2992 * when we are inspecting IPv6 packet, 2993 * we can use nh6 field from 2994 * table_value as next_hop6 address. 2995 */ 2996 if (is_ipv6) { 2997 struct ip_fw_nh6 *nh6; 2998 2999 args->flags |= IPFW_ARGS_NH6; 3000 nh6 = &args->hopstore6; 3001 nh6->sin6_addr = TARG_VAL( 3002 chain, tablearg, nh6); 3003 nh6->sin6_port = sa->sin_port; 3004 nh6->sin6_scope_id = TARG_VAL( 3005 chain, tablearg, zoneid); 3006 } else 3007 #endif 3008 { 3009 args->flags |= IPFW_ARGS_NH4; 3010 args->hopstore.sin_port = 3011 sa->sin_port; 3012 sa = &args->hopstore; 3013 sa->sin_family = AF_INET; 3014 sa->sin_len = sizeof(*sa); 3015 sa->sin_addr.s_addr = htonl( 3016 TARG_VAL(chain, tablearg, 3017 nh4)); 3018 } 3019 } else { 3020 args->flags |= IPFW_ARGS_NH4PTR; 3021 args->next_hop = sa; 3022 } 3023 } 3024 retval = IP_FW_PASS; 3025 l = 0; /* exit inner loop */ 3026 done = 1; /* exit outer loop */ 3027 break; 3028 3029 #ifdef INET6 3030 case O_FORWARD_IP6: 3031 if (args->flags & IPFW_ARGS_ETHER) 3032 break; /* not valid on layer2 pkts */ 3033 if (q != f || 3034 dyn_info.direction == MATCH_FORWARD) { 3035 struct sockaddr_in6 *sin6; 3036 3037 sin6 = &(((ipfw_insn_sa6 *)cmd)->sa); 3038 args->flags |= IPFW_ARGS_NH6PTR; 3039 args->next_hop6 = sin6; 3040 } 3041 retval = IP_FW_PASS; 3042 l = 0; /* exit inner loop */ 3043 done = 1; /* exit outer loop */ 3044 break; 3045 #endif 3046 3047 case O_NETGRAPH: 3048 case O_NGTEE: 3049 set_match(args, f_pos, chain); 3050 args->rule.info = TARG(cmd->arg1, netgraph); 3051 if (V_fw_one_pass) 3052 args->rule.info |= IPFW_ONEPASS; 3053 retval = (cmd->opcode == O_NETGRAPH) ? 3054 IP_FW_NETGRAPH : IP_FW_NGTEE; 3055 l = 0; /* exit inner loop */ 3056 done = 1; /* exit outer loop */ 3057 break; 3058 3059 case O_SETFIB: { 3060 uint32_t fib; 3061 3062 IPFW_INC_RULE_COUNTER(f, pktlen); 3063 fib = TARG(cmd->arg1, fib) & 0x7FFF; 3064 if (fib >= rt_numfibs) 3065 fib = 0; 3066 M_SETFIB(m, fib); 3067 args->f_id.fib = fib; /* XXX */ 3068 l = 0; /* exit inner loop */ 3069 break; 3070 } 3071 3072 case O_SETDSCP: { 3073 uint16_t code; 3074 3075 code = TARG(cmd->arg1, dscp) & 0x3F; 3076 l = 0; /* exit inner loop */ 3077 if (is_ipv4) { 3078 uint16_t old; 3079 3080 old = *(uint16_t *)ip; 3081 ip->ip_tos = (code << 2) | 3082 (ip->ip_tos & 0x03); 3083 ip->ip_sum = cksum_adjust(ip->ip_sum, 3084 old, *(uint16_t *)ip); 3085 } else if (is_ipv6) { 3086 uint8_t *v; 3087 3088 v = &((struct ip6_hdr *)ip)->ip6_vfc; 3089 *v = (*v & 0xF0) | (code >> 2); 3090 v++; 3091 *v = (*v & 0x3F) | ((code & 0x03) << 6); 3092 } else 3093 break; 3094 3095 IPFW_INC_RULE_COUNTER(f, pktlen); 3096 break; 3097 } 3098 3099 case O_NAT: 3100 l = 0; /* exit inner loop */ 3101 done = 1; /* exit outer loop */ 3102 /* 3103 * Ensure that we do not invoke NAT handler for 3104 * non IPv4 packets. Libalias expects only IPv4. 3105 */ 3106 if (!is_ipv4 || !IPFW_NAT_LOADED) { 3107 retval = IP_FW_DENY; 3108 break; 3109 } 3110 3111 struct cfg_nat *t; 3112 int nat_id; 3113 3114 args->rule.info = 0; 3115 set_match(args, f_pos, chain); 3116 /* Check if this is 'global' nat rule */ 3117 if (cmd->arg1 == IP_FW_NAT44_GLOBAL) { 3118 retval = ipfw_nat_ptr(args, NULL, m); 3119 break; 3120 } 3121 t = ((ipfw_insn_nat *)cmd)->nat; 3122 if (t == NULL) { 3123 nat_id = TARG(cmd->arg1, nat); 3124 t = (*lookup_nat_ptr)(&chain->nat, nat_id); 3125 3126 if (t == NULL) { 3127 retval = IP_FW_DENY; 3128 break; 3129 } 3130 if (cmd->arg1 != IP_FW_TARG) 3131 ((ipfw_insn_nat *)cmd)->nat = t; 3132 } 3133 retval = ipfw_nat_ptr(args, t, m); 3134 break; 3135 3136 case O_REASS: { 3137 int ip_off; 3138 3139 l = 0; /* in any case exit inner loop */ 3140 if (is_ipv6) /* IPv6 is not supported yet */ 3141 break; 3142 IPFW_INC_RULE_COUNTER(f, pktlen); 3143 ip_off = ntohs(ip->ip_off); 3144 3145 /* if not fragmented, go to next rule */ 3146 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0) 3147 break; 3148 3149 args->m = m = ip_reass(m); 3150 3151 /* 3152 * do IP header checksum fixup. 3153 */ 3154 if (m == NULL) { /* fragment got swallowed */ 3155 retval = IP_FW_DENY; 3156 } else { /* good, packet complete */ 3157 int hlen; 3158 3159 ip = mtod(m, struct ip *); 3160 hlen = ip->ip_hl << 2; 3161 ip->ip_sum = 0; 3162 if (hlen == sizeof(struct ip)) 3163 ip->ip_sum = in_cksum_hdr(ip); 3164 else 3165 ip->ip_sum = in_cksum(m, hlen); 3166 retval = IP_FW_REASS; 3167 args->rule.info = 0; 3168 set_match(args, f_pos, chain); 3169 } 3170 done = 1; /* exit outer loop */ 3171 break; 3172 } 3173 case O_EXTERNAL_ACTION: 3174 l = 0; /* in any case exit inner loop */ 3175 retval = ipfw_run_eaction(chain, args, 3176 cmd, &done); 3177 /* 3178 * If both @retval and @done are zero, 3179 * consider this as rule matching and 3180 * update counters. 3181 */ 3182 if (retval == 0 && done == 0) { 3183 IPFW_INC_RULE_COUNTER(f, pktlen); 3184 /* 3185 * Reset the result of the last 3186 * dynamic state lookup. 3187 * External action can change 3188 * @args content, and it may be 3189 * used for new state lookup later. 3190 */ 3191 DYN_INFO_INIT(&dyn_info); 3192 } 3193 break; 3194 3195 default: 3196 panic("-- unknown opcode %d\n", cmd->opcode); 3197 } /* end of switch() on opcodes */ 3198 /* 3199 * if we get here with l=0, then match is irrelevant. 3200 */ 3201 3202 if (cmd->len & F_NOT) 3203 match = !match; 3204 3205 if (match) { 3206 if (cmd->len & F_OR) 3207 skip_or = 1; 3208 } else { 3209 if (!(cmd->len & F_OR)) /* not an OR block, */ 3210 break; /* try next rule */ 3211 } 3212 3213 } /* end of inner loop, scan opcodes */ 3214 #undef PULLUP_LEN 3215 #undef PULLUP_LEN_LOCKED 3216 3217 if (done) 3218 break; 3219 3220 /* next_rule:; */ /* try next rule */ 3221 3222 } /* end of outer for, scan rules */ 3223 3224 if (done) { 3225 struct ip_fw *rule = chain->map[f_pos]; 3226 /* Update statistics */ 3227 IPFW_INC_RULE_COUNTER(rule, pktlen); 3228 } else { 3229 retval = IP_FW_DENY; 3230 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3231 } 3232 IPFW_PF_RUNLOCK(chain); 3233 #ifdef __FreeBSD__ 3234 if (ucred_cache != NULL) 3235 crfree(ucred_cache); 3236 #endif 3237 return (retval); 3238 3239 pullup_failed: 3240 if (V_fw_verbose) 3241 printf("ipfw: pullup failed\n"); 3242 return (IP_FW_DENY); 3243 } 3244 3245 /* 3246 * Set maximum number of tables that can be used in given VNET ipfw instance. 3247 */ 3248 #ifdef SYSCTL_NODE 3249 static int 3250 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS) 3251 { 3252 int error; 3253 unsigned int ntables; 3254 3255 ntables = V_fw_tables_max; 3256 3257 error = sysctl_handle_int(oidp, &ntables, 0, req); 3258 /* Read operation or some error */ 3259 if ((error != 0) || (req->newptr == NULL)) 3260 return (error); 3261 3262 return (ipfw_resize_tables(&V_layer3_chain, ntables)); 3263 } 3264 3265 /* 3266 * Switches table namespace between global and per-set. 3267 */ 3268 static int 3269 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS) 3270 { 3271 int error; 3272 unsigned int sets; 3273 3274 sets = V_fw_tables_sets; 3275 3276 error = sysctl_handle_int(oidp, &sets, 0, req); 3277 /* Read operation or some error */ 3278 if ((error != 0) || (req->newptr == NULL)) 3279 return (error); 3280 3281 return (ipfw_switch_tables_namespace(&V_layer3_chain, sets)); 3282 } 3283 #endif 3284 3285 /* 3286 * Module and VNET glue 3287 */ 3288 3289 /* 3290 * Stuff that must be initialised only on boot or module load 3291 */ 3292 static int 3293 ipfw_init(void) 3294 { 3295 int error = 0; 3296 3297 /* 3298 * Only print out this stuff the first time around, 3299 * when called from the sysinit code. 3300 */ 3301 printf("ipfw2 " 3302 #ifdef INET6 3303 "(+ipv6) " 3304 #endif 3305 "initialized, divert %s, nat %s, " 3306 "default to %s, logging ", 3307 #ifdef IPDIVERT 3308 "enabled", 3309 #else 3310 "loadable", 3311 #endif 3312 #ifdef IPFIREWALL_NAT 3313 "enabled", 3314 #else 3315 "loadable", 3316 #endif 3317 default_to_accept ? "accept" : "deny"); 3318 3319 /* 3320 * Note: V_xxx variables can be accessed here but the vnet specific 3321 * initializer may not have been called yet for the VIMAGE case. 3322 * Tuneables will have been processed. We will print out values for 3323 * the default vnet. 3324 * XXX This should all be rationalized AFTER 8.0 3325 */ 3326 if (V_fw_verbose == 0) 3327 printf("disabled\n"); 3328 else if (V_verbose_limit == 0) 3329 printf("unlimited\n"); 3330 else 3331 printf("limited to %d packets/entry by default\n", 3332 V_verbose_limit); 3333 3334 /* Check user-supplied table count for validness */ 3335 if (default_fw_tables > IPFW_TABLES_MAX) 3336 default_fw_tables = IPFW_TABLES_MAX; 3337 3338 ipfw_init_sopt_handler(); 3339 ipfw_init_obj_rewriter(); 3340 ipfw_iface_init(); 3341 return (error); 3342 } 3343 3344 /* 3345 * Called for the removal of the last instance only on module unload. 3346 */ 3347 static void 3348 ipfw_destroy(void) 3349 { 3350 3351 ipfw_iface_destroy(); 3352 ipfw_destroy_sopt_handler(); 3353 ipfw_destroy_obj_rewriter(); 3354 printf("IP firewall unloaded\n"); 3355 } 3356 3357 /* 3358 * Stuff that must be initialized for every instance 3359 * (including the first of course). 3360 */ 3361 static int 3362 vnet_ipfw_init(const void *unused) 3363 { 3364 int error, first; 3365 struct ip_fw *rule = NULL; 3366 struct ip_fw_chain *chain; 3367 3368 chain = &V_layer3_chain; 3369 3370 first = IS_DEFAULT_VNET(curvnet) ? 1 : 0; 3371 3372 /* First set up some values that are compile time options */ 3373 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 3374 V_fw_deny_unknown_exthdrs = 1; 3375 #ifdef IPFIREWALL_VERBOSE 3376 V_fw_verbose = 1; 3377 #endif 3378 #ifdef IPFIREWALL_VERBOSE_LIMIT 3379 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 3380 #endif 3381 #ifdef IPFIREWALL_NAT 3382 LIST_INIT(&chain->nat); 3383 #endif 3384 3385 /* Init shared services hash table */ 3386 ipfw_init_srv(chain); 3387 3388 ipfw_init_counters(); 3389 /* Set initial number of tables */ 3390 V_fw_tables_max = default_fw_tables; 3391 error = ipfw_init_tables(chain, first); 3392 if (error) { 3393 printf("ipfw2: setting up tables failed\n"); 3394 free(chain->map, M_IPFW); 3395 free(rule, M_IPFW); 3396 return (ENOSPC); 3397 } 3398 3399 IPFW_LOCK_INIT(chain); 3400 3401 /* fill and insert the default rule */ 3402 rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw)); 3403 rule->flags |= IPFW_RULE_NOOPT; 3404 rule->cmd_len = 1; 3405 rule->cmd[0].len = 1; 3406 rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; 3407 chain->default_rule = rule; 3408 ipfw_add_protected_rule(chain, rule, 0); 3409 3410 ipfw_dyn_init(chain); 3411 ipfw_eaction_init(chain, first); 3412 #ifdef LINEAR_SKIPTO 3413 ipfw_init_skipto_cache(chain); 3414 #endif 3415 ipfw_bpf_init(first); 3416 3417 /* First set up some values that are compile time options */ 3418 V_ipfw_vnet_ready = 1; /* Open for business */ 3419 3420 /* 3421 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6. 3422 * Even if the latter two fail we still keep the module alive 3423 * because the sockopt and layer2 paths are still useful. 3424 * ipfw[6]_hook return 0 on success, ENOENT on failure, 3425 * so we can ignore the exact return value and just set a flag. 3426 * 3427 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so 3428 * changes in the underlying (per-vnet) variables trigger 3429 * immediate hook()/unhook() calls. 3430 * In layer2 we have the same behaviour, except that V_ether_ipfw 3431 * is checked on each packet because there are no pfil hooks. 3432 */ 3433 V_ip_fw_ctl_ptr = ipfw_ctl3; 3434 error = ipfw_attach_hooks(); 3435 return (error); 3436 } 3437 3438 /* 3439 * Called for the removal of each instance. 3440 */ 3441 static int 3442 vnet_ipfw_uninit(const void *unused) 3443 { 3444 struct ip_fw *reap; 3445 struct ip_fw_chain *chain = &V_layer3_chain; 3446 int i, last; 3447 3448 V_ipfw_vnet_ready = 0; /* tell new callers to go away */ 3449 /* 3450 * disconnect from ipv4, ipv6, layer2 and sockopt. 3451 * Then grab, release and grab again the WLOCK so we make 3452 * sure the update is propagated and nobody will be in. 3453 */ 3454 ipfw_detach_hooks(); 3455 V_ip_fw_ctl_ptr = NULL; 3456 3457 last = IS_DEFAULT_VNET(curvnet) ? 1 : 0; 3458 3459 IPFW_UH_WLOCK(chain); 3460 IPFW_UH_WUNLOCK(chain); 3461 3462 ipfw_dyn_uninit(0); /* run the callout_drain */ 3463 3464 IPFW_UH_WLOCK(chain); 3465 3466 reap = NULL; 3467 IPFW_WLOCK(chain); 3468 for (i = 0; i < chain->n_rules; i++) 3469 ipfw_reap_add(chain, &reap, chain->map[i]); 3470 free(chain->map, M_IPFW); 3471 #ifdef LINEAR_SKIPTO 3472 ipfw_destroy_skipto_cache(chain); 3473 #endif 3474 IPFW_WUNLOCK(chain); 3475 IPFW_UH_WUNLOCK(chain); 3476 ipfw_destroy_tables(chain, last); 3477 ipfw_eaction_uninit(chain, last); 3478 if (reap != NULL) 3479 ipfw_reap_rules(reap); 3480 vnet_ipfw_iface_destroy(chain); 3481 ipfw_destroy_srv(chain); 3482 IPFW_LOCK_DESTROY(chain); 3483 ipfw_dyn_uninit(1); /* free the remaining parts */ 3484 ipfw_destroy_counters(); 3485 ipfw_bpf_uninit(last); 3486 return (0); 3487 } 3488 3489 /* 3490 * Module event handler. 3491 * In general we have the choice of handling most of these events by the 3492 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to 3493 * use the SYSINIT handlers as they are more capable of expressing the 3494 * flow of control during module and vnet operations, so this is just 3495 * a skeleton. Note there is no SYSINIT equivalent of the module 3496 * SHUTDOWN handler, but we don't have anything to do in that case anyhow. 3497 */ 3498 static int 3499 ipfw_modevent(module_t mod, int type, void *unused) 3500 { 3501 int err = 0; 3502 3503 switch (type) { 3504 case MOD_LOAD: 3505 /* Called once at module load or 3506 * system boot if compiled in. */ 3507 break; 3508 case MOD_QUIESCE: 3509 /* Called before unload. May veto unloading. */ 3510 break; 3511 case MOD_UNLOAD: 3512 /* Called during unload. */ 3513 break; 3514 case MOD_SHUTDOWN: 3515 /* Called during system shutdown. */ 3516 break; 3517 default: 3518 err = EOPNOTSUPP; 3519 break; 3520 } 3521 return err; 3522 } 3523 3524 static moduledata_t ipfwmod = { 3525 "ipfw", 3526 ipfw_modevent, 3527 0 3528 }; 3529 3530 /* Define startup order. */ 3531 #define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_FIREWALL 3532 #define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */ 3533 #define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */ 3534 #define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */ 3535 3536 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER); 3537 FEATURE(ipfw_ctl3, "ipfw new sockopt calls"); 3538 MODULE_VERSION(ipfw, 3); 3539 /* should declare some dependencies here */ 3540 3541 /* 3542 * Starting up. Done in order after ipfwmod() has been called. 3543 * VNET_SYSINIT is also called for each existing vnet and each new vnet. 3544 */ 3545 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 3546 ipfw_init, NULL); 3547 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 3548 vnet_ipfw_init, NULL); 3549 3550 /* 3551 * Closing up shop. These are done in REVERSE ORDER, but still 3552 * after ipfwmod() has been called. Not called on reboot. 3553 * VNET_SYSUNINIT is also called for each exiting vnet as it exits. 3554 * or when the module is unloaded. 3555 */ 3556 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 3557 ipfw_destroy, NULL); 3558 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 3559 vnet_ipfw_uninit, NULL); 3560 /* end of file */ 3561