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/pfil.h> 68 #include <net/vnet.h> 69 70 #include <netpfil/pf/pf_mtag.h> 71 72 #include <netinet/in.h> 73 #include <netinet/in_var.h> 74 #include <netinet/in_pcb.h> 75 #include <netinet/ip.h> 76 #include <netinet/ip_var.h> 77 #include <netinet/ip_icmp.h> 78 #include <netinet/ip_fw.h> 79 #include <netinet/ip_carp.h> 80 #include <netinet/pim.h> 81 #include <netinet/tcp_var.h> 82 #include <netinet/udp.h> 83 #include <netinet/udp_var.h> 84 #include <netinet/sctp.h> 85 #include <netinet/sctp_crc32.h> 86 #include <netinet/sctp_header.h> 87 88 #include <netinet/ip6.h> 89 #include <netinet/icmp6.h> 90 #include <netinet/in_fib.h> 91 #ifdef INET6 92 #include <netinet6/in6_fib.h> 93 #include <netinet6/in6_pcb.h> 94 #include <netinet6/scope6_var.h> 95 #include <netinet6/ip6_var.h> 96 #endif 97 98 #include <net/if_gre.h> /* for struct grehdr */ 99 100 #include <netpfil/ipfw/ip_fw_private.h> 101 102 #include <machine/in_cksum.h> /* XXX for in_cksum */ 103 104 #ifdef MAC 105 #include <security/mac/mac_framework.h> 106 #endif 107 108 /* 109 * static variables followed by global ones. 110 * All ipfw global variables are here. 111 */ 112 113 VNET_DEFINE_STATIC(int, fw_deny_unknown_exthdrs); 114 #define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs) 115 116 VNET_DEFINE_STATIC(int, fw_permit_single_frag6) = 1; 117 #define V_fw_permit_single_frag6 VNET(fw_permit_single_frag6) 118 119 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 120 static int default_to_accept = 1; 121 #else 122 static int default_to_accept; 123 #endif 124 125 VNET_DEFINE(int, autoinc_step); 126 VNET_DEFINE(int, fw_one_pass) = 1; 127 128 VNET_DEFINE(unsigned int, fw_tables_max); 129 VNET_DEFINE(unsigned int, fw_tables_sets) = 0; /* Don't use set-aware tables */ 130 /* Use 128 tables by default */ 131 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT; 132 133 #ifndef LINEAR_SKIPTO 134 static int jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num, 135 int tablearg, int jump_backwards); 136 #define JUMP(ch, f, num, targ, back) jump_fast(ch, f, num, targ, back) 137 #else 138 static int jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num, 139 int tablearg, int jump_backwards); 140 #define JUMP(ch, f, num, targ, back) jump_linear(ch, f, num, targ, back) 141 #endif 142 143 /* 144 * Each rule belongs to one of 32 different sets (0..31). 145 * The variable set_disable contains one bit per set. 146 * If the bit is set, all rules in the corresponding set 147 * are disabled. Set RESVD_SET(31) is reserved for the default rule 148 * and rules that are not deleted by the flush command, 149 * and CANNOT be disabled. 150 * Rules in set RESVD_SET can only be deleted individually. 151 */ 152 VNET_DEFINE(u_int32_t, set_disable); 153 #define V_set_disable VNET(set_disable) 154 155 VNET_DEFINE(int, fw_verbose); 156 /* counter for ipfw_log(NULL...) */ 157 VNET_DEFINE(u_int64_t, norule_counter); 158 VNET_DEFINE(int, verbose_limit); 159 160 /* layer3_chain contains the list of rules for layer 3 */ 161 VNET_DEFINE(struct ip_fw_chain, layer3_chain); 162 163 /* ipfw_vnet_ready controls when we are open for business */ 164 VNET_DEFINE(int, ipfw_vnet_ready) = 0; 165 166 VNET_DEFINE(int, ipfw_nat_ready) = 0; 167 168 ipfw_nat_t *ipfw_nat_ptr = NULL; 169 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int); 170 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; 171 ipfw_nat_cfg_t *ipfw_nat_del_ptr; 172 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; 173 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; 174 175 #ifdef SYSCTL_NODE 176 uint32_t dummy_def = IPFW_DEFAULT_RULE; 177 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS); 178 static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS); 179 180 SYSBEGIN(f3) 181 182 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 183 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 184 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0, 185 "Only do a single pass through ipfw when using dummynet(4)"); 186 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, 187 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0, 188 "Rule number auto-increment step"); 189 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 190 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0, 191 "Log matches to ipfw rules"); 192 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, 193 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(verbose_limit), 0, 194 "Set upper limit of matches of ipfw rules logged"); 195 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, 196 &dummy_def, 0, 197 "The default/max possible rule number."); 198 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_max, 199 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU", 200 "Maximum number of concurrently used tables"); 201 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets, 202 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW, 203 0, 0, sysctl_ipfw_tables_sets, "IU", 204 "Use per-set namespace for tables"); 205 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN, 206 &default_to_accept, 0, 207 "Make the default rule accept all packets."); 208 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables); 209 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, 210 CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0, 211 "Number of static rules"); 212 213 #ifdef INET6 214 SYSCTL_DECL(_net_inet6_ip6); 215 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 216 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs, 217 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, 218 &VNET_NAME(fw_deny_unknown_exthdrs), 0, 219 "Deny packets with unknown IPv6 Extension Headers"); 220 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6, 221 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, 222 &VNET_NAME(fw_permit_single_frag6), 0, 223 "Permit single packet IPv6 fragments"); 224 #endif /* INET6 */ 225 226 SYSEND 227 228 #endif /* SYSCTL_NODE */ 229 230 231 /* 232 * Some macros used in the various matching options. 233 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 234 * Other macros just cast void * into the appropriate type 235 */ 236 #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 237 #define TCP(p) ((struct tcphdr *)(p)) 238 #define SCTP(p) ((struct sctphdr *)(p)) 239 #define UDP(p) ((struct udphdr *)(p)) 240 #define ICMP(p) ((struct icmphdr *)(p)) 241 #define ICMP6(p) ((struct icmp6_hdr *)(p)) 242 243 static __inline int 244 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 245 { 246 int type = icmp->icmp_type; 247 248 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 249 } 250 251 #define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 252 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 253 254 static int 255 is_icmp_query(struct icmphdr *icmp) 256 { 257 int type = icmp->icmp_type; 258 259 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 260 } 261 #undef TT 262 263 /* 264 * The following checks use two arrays of 8 or 16 bits to store the 265 * bits that we want set or clear, respectively. They are in the 266 * low and high half of cmd->arg1 or cmd->d[0]. 267 * 268 * We scan options and store the bits we find set. We succeed if 269 * 270 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 271 * 272 * The code is sometimes optimized not to store additional variables. 273 */ 274 275 static int 276 flags_match(ipfw_insn *cmd, u_int8_t bits) 277 { 278 u_char want_clear; 279 bits = ~bits; 280 281 if ( ((cmd->arg1 & 0xff) & bits) != 0) 282 return 0; /* some bits we want set were clear */ 283 want_clear = (cmd->arg1 >> 8) & 0xff; 284 if ( (want_clear & bits) != want_clear) 285 return 0; /* some bits we want clear were set */ 286 return 1; 287 } 288 289 static int 290 ipopts_match(struct ip *ip, ipfw_insn *cmd) 291 { 292 int optlen, bits = 0; 293 u_char *cp = (u_char *)(ip + 1); 294 int x = (ip->ip_hl << 2) - sizeof (struct ip); 295 296 for (; x > 0; x -= optlen, cp += optlen) { 297 int opt = cp[IPOPT_OPTVAL]; 298 299 if (opt == IPOPT_EOL) 300 break; 301 if (opt == IPOPT_NOP) 302 optlen = 1; 303 else { 304 optlen = cp[IPOPT_OLEN]; 305 if (optlen <= 0 || optlen > x) 306 return 0; /* invalid or truncated */ 307 } 308 switch (opt) { 309 310 default: 311 break; 312 313 case IPOPT_LSRR: 314 bits |= IP_FW_IPOPT_LSRR; 315 break; 316 317 case IPOPT_SSRR: 318 bits |= IP_FW_IPOPT_SSRR; 319 break; 320 321 case IPOPT_RR: 322 bits |= IP_FW_IPOPT_RR; 323 break; 324 325 case IPOPT_TS: 326 bits |= IP_FW_IPOPT_TS; 327 break; 328 } 329 } 330 return (flags_match(cmd, bits)); 331 } 332 333 /* 334 * Parse TCP options. The logic copied from tcp_dooptions(). 335 */ 336 static int 337 tcpopts_parse(const struct tcphdr *tcp, uint16_t *mss) 338 { 339 const u_char *cp = (const u_char *)(tcp + 1); 340 int optlen, bits = 0; 341 int cnt = (tcp->th_off << 2) - sizeof(struct tcphdr); 342 343 for (; cnt > 0; cnt -= optlen, cp += optlen) { 344 int opt = cp[0]; 345 if (opt == TCPOPT_EOL) 346 break; 347 if (opt == TCPOPT_NOP) 348 optlen = 1; 349 else { 350 if (cnt < 2) 351 break; 352 optlen = cp[1]; 353 if (optlen < 2 || optlen > cnt) 354 break; 355 } 356 357 switch (opt) { 358 default: 359 break; 360 361 case TCPOPT_MAXSEG: 362 if (optlen != TCPOLEN_MAXSEG) 363 break; 364 bits |= IP_FW_TCPOPT_MSS; 365 if (mss != NULL) 366 *mss = be16dec(cp + 2); 367 break; 368 369 case TCPOPT_WINDOW: 370 if (optlen == TCPOLEN_WINDOW) 371 bits |= IP_FW_TCPOPT_WINDOW; 372 break; 373 374 case TCPOPT_SACK_PERMITTED: 375 if (optlen == TCPOLEN_SACK_PERMITTED) 376 bits |= IP_FW_TCPOPT_SACK; 377 break; 378 379 case TCPOPT_SACK: 380 if (optlen > 2 && (optlen - 2) % TCPOLEN_SACK == 0) 381 bits |= IP_FW_TCPOPT_SACK; 382 break; 383 384 case TCPOPT_TIMESTAMP: 385 if (optlen == TCPOLEN_TIMESTAMP) 386 bits |= IP_FW_TCPOPT_TS; 387 break; 388 } 389 } 390 return (bits); 391 } 392 393 static int 394 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 395 { 396 397 return (flags_match(cmd, tcpopts_parse(tcp, NULL))); 398 } 399 400 static int 401 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain, 402 uint32_t *tablearg) 403 { 404 405 if (ifp == NULL) /* no iface with this packet, match fails */ 406 return (0); 407 408 /* Check by name or by IP address */ 409 if (cmd->name[0] != '\0') { /* match by name */ 410 if (cmd->name[0] == '\1') /* use tablearg to match */ 411 return ipfw_lookup_table(chain, cmd->p.kidx, 0, 412 &ifp->if_index, tablearg); 413 /* Check name */ 414 if (cmd->p.glob) { 415 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 416 return(1); 417 } else { 418 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 419 return(1); 420 } 421 } else { 422 #if !defined(USERSPACE) && defined(__FreeBSD__) /* and OSX too ? */ 423 struct ifaddr *ia; 424 425 NET_EPOCH_ASSERT(); 426 427 CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 428 if (ia->ifa_addr->sa_family != AF_INET) 429 continue; 430 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 431 (ia->ifa_addr))->sin_addr.s_addr) 432 return (1); /* match */ 433 } 434 #endif /* __FreeBSD__ */ 435 } 436 return(0); /* no match, fail ... */ 437 } 438 439 /* 440 * The verify_path function checks if a route to the src exists and 441 * if it is reachable via ifp (when provided). 442 * 443 * The 'verrevpath' option checks that the interface that an IP packet 444 * arrives on is the same interface that traffic destined for the 445 * packet's source address would be routed out of. 446 * The 'versrcreach' option just checks that the source address is 447 * reachable via any route (except default) in the routing table. 448 * These two are a measure to block forged packets. This is also 449 * commonly known as "anti-spoofing" or Unicast Reverse Path 450 * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 451 * is purposely reminiscent of the Cisco IOS command, 452 * 453 * ip verify unicast reverse-path 454 * ip verify unicast source reachable-via any 455 * 456 * which implements the same functionality. But note that the syntax 457 * is misleading, and the check may be performed on all IP packets 458 * whether unicast, multicast, or broadcast. 459 */ 460 static int 461 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) 462 { 463 #if defined(USERSPACE) || !defined(__FreeBSD__) 464 return 0; 465 #else 466 struct nhop4_basic nh4; 467 468 if (fib4_lookup_nh_basic(fib, src, NHR_IFAIF, 0, &nh4) != 0) 469 return (0); 470 471 /* 472 * If ifp is provided, check for equality with rtentry. 473 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 474 * in order to pass packets injected back by if_simloop(): 475 * routing entry (via lo0) for our own address 476 * may exist, so we need to handle routing assymetry. 477 */ 478 if (ifp != NULL && ifp != nh4.nh_ifp) 479 return (0); 480 481 /* if no ifp provided, check if rtentry is not default route */ 482 if (ifp == NULL && (nh4.nh_flags & NHF_DEFAULT) != 0) 483 return (0); 484 485 /* or if this is a blackhole/reject route */ 486 if (ifp == NULL && (nh4.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0) 487 return (0); 488 489 /* found valid route */ 490 return 1; 491 #endif /* __FreeBSD__ */ 492 } 493 494 /* 495 * Generate an SCTP packet containing an ABORT chunk. The verification tag 496 * is given by vtag. The T-bit is set in the ABORT chunk if and only if 497 * reflected is not 0. 498 */ 499 500 static struct mbuf * 501 ipfw_send_abort(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t vtag, 502 int reflected) 503 { 504 struct mbuf *m; 505 struct ip *ip; 506 #ifdef INET6 507 struct ip6_hdr *ip6; 508 #endif 509 struct sctphdr *sctp; 510 struct sctp_chunkhdr *chunk; 511 u_int16_t hlen, plen, tlen; 512 513 MGETHDR(m, M_NOWAIT, MT_DATA); 514 if (m == NULL) 515 return (NULL); 516 517 M_SETFIB(m, id->fib); 518 #ifdef MAC 519 if (replyto != NULL) 520 mac_netinet_firewall_reply(replyto, m); 521 else 522 mac_netinet_firewall_send(m); 523 #else 524 (void)replyto; /* don't warn about unused arg */ 525 #endif 526 527 switch (id->addr_type) { 528 case 4: 529 hlen = sizeof(struct ip); 530 break; 531 #ifdef INET6 532 case 6: 533 hlen = sizeof(struct ip6_hdr); 534 break; 535 #endif 536 default: 537 /* XXX: log me?!? */ 538 FREE_PKT(m); 539 return (NULL); 540 } 541 plen = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); 542 tlen = hlen + plen; 543 m->m_data += max_linkhdr; 544 m->m_flags |= M_SKIP_FIREWALL; 545 m->m_pkthdr.len = m->m_len = tlen; 546 m->m_pkthdr.rcvif = NULL; 547 bzero(m->m_data, tlen); 548 549 switch (id->addr_type) { 550 case 4: 551 ip = mtod(m, struct ip *); 552 553 ip->ip_v = 4; 554 ip->ip_hl = sizeof(struct ip) >> 2; 555 ip->ip_tos = IPTOS_LOWDELAY; 556 ip->ip_len = htons(tlen); 557 ip->ip_id = htons(0); 558 ip->ip_off = htons(0); 559 ip->ip_ttl = V_ip_defttl; 560 ip->ip_p = IPPROTO_SCTP; 561 ip->ip_sum = 0; 562 ip->ip_src.s_addr = htonl(id->dst_ip); 563 ip->ip_dst.s_addr = htonl(id->src_ip); 564 565 sctp = (struct sctphdr *)(ip + 1); 566 break; 567 #ifdef INET6 568 case 6: 569 ip6 = mtod(m, struct ip6_hdr *); 570 571 ip6->ip6_vfc = IPV6_VERSION; 572 ip6->ip6_plen = htons(plen); 573 ip6->ip6_nxt = IPPROTO_SCTP; 574 ip6->ip6_hlim = IPV6_DEFHLIM; 575 ip6->ip6_src = id->dst_ip6; 576 ip6->ip6_dst = id->src_ip6; 577 578 sctp = (struct sctphdr *)(ip6 + 1); 579 break; 580 #endif 581 } 582 583 sctp->src_port = htons(id->dst_port); 584 sctp->dest_port = htons(id->src_port); 585 sctp->v_tag = htonl(vtag); 586 sctp->checksum = htonl(0); 587 588 chunk = (struct sctp_chunkhdr *)(sctp + 1); 589 chunk->chunk_type = SCTP_ABORT_ASSOCIATION; 590 chunk->chunk_flags = 0; 591 if (reflected != 0) { 592 chunk->chunk_flags |= SCTP_HAD_NO_TCB; 593 } 594 chunk->chunk_length = htons(sizeof(struct sctp_chunkhdr)); 595 596 sctp->checksum = sctp_calculate_cksum(m, hlen); 597 598 return (m); 599 } 600 601 /* 602 * Generate a TCP packet, containing either a RST or a keepalive. 603 * When flags & TH_RST, we are sending a RST packet, because of a 604 * "reset" action matched the packet. 605 * Otherwise we are sending a keepalive, and flags & TH_ 606 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 607 * so that MAC can label the reply appropriately. 608 */ 609 struct mbuf * 610 ipfw_send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 611 u_int32_t ack, int flags) 612 { 613 struct mbuf *m = NULL; /* stupid compiler */ 614 struct ip *h = NULL; /* stupid compiler */ 615 #ifdef INET6 616 struct ip6_hdr *h6 = NULL; 617 #endif 618 struct tcphdr *th = NULL; 619 int len, dir; 620 621 MGETHDR(m, M_NOWAIT, MT_DATA); 622 if (m == NULL) 623 return (NULL); 624 625 M_SETFIB(m, id->fib); 626 #ifdef MAC 627 if (replyto != NULL) 628 mac_netinet_firewall_reply(replyto, m); 629 else 630 mac_netinet_firewall_send(m); 631 #else 632 (void)replyto; /* don't warn about unused arg */ 633 #endif 634 635 switch (id->addr_type) { 636 case 4: 637 len = sizeof(struct ip) + sizeof(struct tcphdr); 638 break; 639 #ifdef INET6 640 case 6: 641 len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); 642 break; 643 #endif 644 default: 645 /* XXX: log me?!? */ 646 FREE_PKT(m); 647 return (NULL); 648 } 649 dir = ((flags & (TH_SYN | TH_RST)) == TH_SYN); 650 651 m->m_data += max_linkhdr; 652 m->m_flags |= M_SKIP_FIREWALL; 653 m->m_pkthdr.len = m->m_len = len; 654 m->m_pkthdr.rcvif = NULL; 655 bzero(m->m_data, len); 656 657 switch (id->addr_type) { 658 case 4: 659 h = mtod(m, struct ip *); 660 661 /* prepare for checksum */ 662 h->ip_p = IPPROTO_TCP; 663 h->ip_len = htons(sizeof(struct tcphdr)); 664 if (dir) { 665 h->ip_src.s_addr = htonl(id->src_ip); 666 h->ip_dst.s_addr = htonl(id->dst_ip); 667 } else { 668 h->ip_src.s_addr = htonl(id->dst_ip); 669 h->ip_dst.s_addr = htonl(id->src_ip); 670 } 671 672 th = (struct tcphdr *)(h + 1); 673 break; 674 #ifdef INET6 675 case 6: 676 h6 = mtod(m, struct ip6_hdr *); 677 678 /* prepare for checksum */ 679 h6->ip6_nxt = IPPROTO_TCP; 680 h6->ip6_plen = htons(sizeof(struct tcphdr)); 681 if (dir) { 682 h6->ip6_src = id->src_ip6; 683 h6->ip6_dst = id->dst_ip6; 684 } else { 685 h6->ip6_src = id->dst_ip6; 686 h6->ip6_dst = id->src_ip6; 687 } 688 689 th = (struct tcphdr *)(h6 + 1); 690 break; 691 #endif 692 } 693 694 if (dir) { 695 th->th_sport = htons(id->src_port); 696 th->th_dport = htons(id->dst_port); 697 } else { 698 th->th_sport = htons(id->dst_port); 699 th->th_dport = htons(id->src_port); 700 } 701 th->th_off = sizeof(struct tcphdr) >> 2; 702 703 if (flags & TH_RST) { 704 if (flags & TH_ACK) { 705 th->th_seq = htonl(ack); 706 th->th_flags = TH_RST; 707 } else { 708 if (flags & TH_SYN) 709 seq++; 710 th->th_ack = htonl(seq); 711 th->th_flags = TH_RST | TH_ACK; 712 } 713 } else { 714 /* 715 * Keepalive - use caller provided sequence numbers 716 */ 717 th->th_seq = htonl(seq); 718 th->th_ack = htonl(ack); 719 th->th_flags = TH_ACK; 720 } 721 722 switch (id->addr_type) { 723 case 4: 724 th->th_sum = in_cksum(m, len); 725 726 /* finish the ip header */ 727 h->ip_v = 4; 728 h->ip_hl = sizeof(*h) >> 2; 729 h->ip_tos = IPTOS_LOWDELAY; 730 h->ip_off = htons(0); 731 h->ip_len = htons(len); 732 h->ip_ttl = V_ip_defttl; 733 h->ip_sum = 0; 734 break; 735 #ifdef INET6 736 case 6: 737 th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(*h6), 738 sizeof(struct tcphdr)); 739 740 /* finish the ip6 header */ 741 h6->ip6_vfc |= IPV6_VERSION; 742 h6->ip6_hlim = IPV6_DEFHLIM; 743 break; 744 #endif 745 } 746 747 return (m); 748 } 749 750 #ifdef INET6 751 /* 752 * ipv6 specific rules here... 753 */ 754 static __inline int 755 icmp6type_match (int type, ipfw_insn_u32 *cmd) 756 { 757 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 758 } 759 760 static int 761 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 762 { 763 int i; 764 for (i=0; i <= cmd->o.arg1; ++i ) 765 if (curr_flow == cmd->d[i] ) 766 return 1; 767 return 0; 768 } 769 770 /* support for IP6_*_ME opcodes */ 771 static const struct in6_addr lla_mask = {{{ 772 0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff, 773 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff 774 }}}; 775 776 static int 777 ipfw_localip6(struct in6_addr *in6) 778 { 779 struct rm_priotracker in6_ifa_tracker; 780 struct in6_ifaddr *ia; 781 782 if (IN6_IS_ADDR_MULTICAST(in6)) 783 return (0); 784 785 if (!IN6_IS_ADDR_LINKLOCAL(in6)) 786 return (in6_localip(in6)); 787 788 IN6_IFADDR_RLOCK(&in6_ifa_tracker); 789 CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) { 790 if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) 791 continue; 792 if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr, 793 in6, &lla_mask)) { 794 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); 795 return (1); 796 } 797 } 798 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); 799 return (0); 800 } 801 802 static int 803 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib) 804 { 805 struct nhop6_basic nh6; 806 807 if (IN6_IS_SCOPE_LINKLOCAL(src)) 808 return (1); 809 810 if (fib6_lookup_nh_basic(fib, src, 0, NHR_IFAIF, 0, &nh6) != 0) 811 return (0); 812 813 /* If ifp is provided, check for equality with route table. */ 814 if (ifp != NULL && ifp != nh6.nh_ifp) 815 return (0); 816 817 /* if no ifp provided, check if rtentry is not default route */ 818 if (ifp == NULL && (nh6.nh_flags & NHF_DEFAULT) != 0) 819 return (0); 820 821 /* or if this is a blackhole/reject route */ 822 if (ifp == NULL && (nh6.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0) 823 return (0); 824 825 /* found valid route */ 826 return 1; 827 } 828 829 static int 830 is_icmp6_query(int icmp6_type) 831 { 832 if ((icmp6_type <= ICMP6_MAXTYPE) && 833 (icmp6_type == ICMP6_ECHO_REQUEST || 834 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 835 icmp6_type == ICMP6_WRUREQUEST || 836 icmp6_type == ICMP6_FQDN_QUERY || 837 icmp6_type == ICMP6_NI_QUERY)) 838 return (1); 839 840 return (0); 841 } 842 843 static int 844 map_icmp_unreach(int code) 845 { 846 847 /* RFC 7915 p4.2 */ 848 switch (code) { 849 case ICMP_UNREACH_NET: 850 case ICMP_UNREACH_HOST: 851 case ICMP_UNREACH_SRCFAIL: 852 case ICMP_UNREACH_NET_UNKNOWN: 853 case ICMP_UNREACH_HOST_UNKNOWN: 854 case ICMP_UNREACH_TOSNET: 855 case ICMP_UNREACH_TOSHOST: 856 return (ICMP6_DST_UNREACH_NOROUTE); 857 case ICMP_UNREACH_PORT: 858 return (ICMP6_DST_UNREACH_NOPORT); 859 default: 860 /* 861 * Map the rest of codes into admit prohibited. 862 * XXX: unreach proto should be mapped into ICMPv6 863 * parameter problem, but we use only unreach type. 864 */ 865 return (ICMP6_DST_UNREACH_ADMIN); 866 } 867 } 868 869 static void 870 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 871 { 872 struct mbuf *m; 873 874 m = args->m; 875 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 876 struct tcphdr *tcp; 877 tcp = (struct tcphdr *)((char *)ip6 + hlen); 878 879 if ((tcp->th_flags & TH_RST) == 0) { 880 struct mbuf *m0; 881 m0 = ipfw_send_pkt(args->m, &(args->f_id), 882 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 883 tcp->th_flags | TH_RST); 884 if (m0 != NULL) 885 ip6_output(m0, NULL, NULL, 0, NULL, NULL, 886 NULL); 887 } 888 FREE_PKT(m); 889 } else if (code == ICMP6_UNREACH_ABORT && 890 args->f_id.proto == IPPROTO_SCTP) { 891 struct mbuf *m0; 892 struct sctphdr *sctp; 893 u_int32_t v_tag; 894 int reflected; 895 896 sctp = (struct sctphdr *)((char *)ip6 + hlen); 897 reflected = 1; 898 v_tag = ntohl(sctp->v_tag); 899 /* Investigate the first chunk header if available */ 900 if (m->m_len >= hlen + sizeof(struct sctphdr) + 901 sizeof(struct sctp_chunkhdr)) { 902 struct sctp_chunkhdr *chunk; 903 904 chunk = (struct sctp_chunkhdr *)(sctp + 1); 905 switch (chunk->chunk_type) { 906 case SCTP_INITIATION: 907 /* 908 * Packets containing an INIT chunk MUST have 909 * a zero v-tag. 910 */ 911 if (v_tag != 0) { 912 v_tag = 0; 913 break; 914 } 915 /* INIT chunk MUST NOT be bundled */ 916 if (m->m_pkthdr.len > 917 hlen + sizeof(struct sctphdr) + 918 ntohs(chunk->chunk_length) + 3) { 919 break; 920 } 921 /* Use the initiate tag if available */ 922 if ((m->m_len >= hlen + sizeof(struct sctphdr) + 923 sizeof(struct sctp_chunkhdr) + 924 offsetof(struct sctp_init, a_rwnd))) { 925 struct sctp_init *init; 926 927 init = (struct sctp_init *)(chunk + 1); 928 v_tag = ntohl(init->initiate_tag); 929 reflected = 0; 930 } 931 break; 932 case SCTP_ABORT_ASSOCIATION: 933 /* 934 * If the packet contains an ABORT chunk, don't 935 * reply. 936 * XXX: We should search through all chunks, 937 * but don't do to avoid attacks. 938 */ 939 v_tag = 0; 940 break; 941 } 942 } 943 if (v_tag == 0) { 944 m0 = NULL; 945 } else { 946 m0 = ipfw_send_abort(args->m, &(args->f_id), v_tag, 947 reflected); 948 } 949 if (m0 != NULL) 950 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL); 951 FREE_PKT(m); 952 } else if (code != ICMP6_UNREACH_RST && code != ICMP6_UNREACH_ABORT) { 953 /* Send an ICMPv6 unreach. */ 954 #if 0 955 /* 956 * Unlike above, the mbufs need to line up with the ip6 hdr, 957 * as the contents are read. We need to m_adj() the 958 * needed amount. 959 * The mbuf will however be thrown away so we can adjust it. 960 * Remember we did an m_pullup on it already so we 961 * can make some assumptions about contiguousness. 962 */ 963 if (args->L3offset) 964 m_adj(m, args->L3offset); 965 #endif 966 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 967 } else 968 FREE_PKT(m); 969 970 args->m = NULL; 971 } 972 973 #endif /* INET6 */ 974 975 976 /* 977 * sends a reject message, consuming the mbuf passed as an argument. 978 */ 979 static void 980 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip) 981 { 982 983 #if 0 984 /* XXX When ip is not guaranteed to be at mtod() we will 985 * need to account for this */ 986 * The mbuf will however be thrown away so we can adjust it. 987 * Remember we did an m_pullup on it already so we 988 * can make some assumptions about contiguousness. 989 */ 990 if (args->L3offset) 991 m_adj(m, args->L3offset); 992 #endif 993 if (code != ICMP_REJECT_RST && code != ICMP_REJECT_ABORT) { 994 /* Send an ICMP unreach */ 995 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 996 } else if (code == ICMP_REJECT_RST && args->f_id.proto == IPPROTO_TCP) { 997 struct tcphdr *const tcp = 998 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 999 if ( (tcp->th_flags & TH_RST) == 0) { 1000 struct mbuf *m; 1001 m = ipfw_send_pkt(args->m, &(args->f_id), 1002 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1003 tcp->th_flags | TH_RST); 1004 if (m != NULL) 1005 ip_output(m, NULL, NULL, 0, NULL, NULL); 1006 } 1007 FREE_PKT(args->m); 1008 } else if (code == ICMP_REJECT_ABORT && 1009 args->f_id.proto == IPPROTO_SCTP) { 1010 struct mbuf *m; 1011 struct sctphdr *sctp; 1012 struct sctp_chunkhdr *chunk; 1013 struct sctp_init *init; 1014 u_int32_t v_tag; 1015 int reflected; 1016 1017 sctp = L3HDR(struct sctphdr, mtod(args->m, struct ip *)); 1018 reflected = 1; 1019 v_tag = ntohl(sctp->v_tag); 1020 if (iplen >= (ip->ip_hl << 2) + sizeof(struct sctphdr) + 1021 sizeof(struct sctp_chunkhdr)) { 1022 /* Look at the first chunk header if available */ 1023 chunk = (struct sctp_chunkhdr *)(sctp + 1); 1024 switch (chunk->chunk_type) { 1025 case SCTP_INITIATION: 1026 /* 1027 * Packets containing an INIT chunk MUST have 1028 * a zero v-tag. 1029 */ 1030 if (v_tag != 0) { 1031 v_tag = 0; 1032 break; 1033 } 1034 /* INIT chunk MUST NOT be bundled */ 1035 if (iplen > 1036 (ip->ip_hl << 2) + sizeof(struct sctphdr) + 1037 ntohs(chunk->chunk_length) + 3) { 1038 break; 1039 } 1040 /* Use the initiate tag if available */ 1041 if ((iplen >= (ip->ip_hl << 2) + 1042 sizeof(struct sctphdr) + 1043 sizeof(struct sctp_chunkhdr) + 1044 offsetof(struct sctp_init, a_rwnd))) { 1045 init = (struct sctp_init *)(chunk + 1); 1046 v_tag = ntohl(init->initiate_tag); 1047 reflected = 0; 1048 } 1049 break; 1050 case SCTP_ABORT_ASSOCIATION: 1051 /* 1052 * If the packet contains an ABORT chunk, don't 1053 * reply. 1054 * XXX: We should search through all chunks, 1055 * but don't do to avoid attacks. 1056 */ 1057 v_tag = 0; 1058 break; 1059 } 1060 } 1061 if (v_tag == 0) { 1062 m = NULL; 1063 } else { 1064 m = ipfw_send_abort(args->m, &(args->f_id), v_tag, 1065 reflected); 1066 } 1067 if (m != NULL) 1068 ip_output(m, NULL, NULL, 0, NULL, NULL); 1069 FREE_PKT(args->m); 1070 } else 1071 FREE_PKT(args->m); 1072 args->m = NULL; 1073 } 1074 1075 /* 1076 * Support for uid/gid/jail lookup. These tests are expensive 1077 * (because we may need to look into the list of active sockets) 1078 * so we cache the results. ugid_lookupp is 0 if we have not 1079 * yet done a lookup, 1 if we succeeded, and -1 if we tried 1080 * and failed. The function always returns the match value. 1081 * We could actually spare the variable and use *uc, setting 1082 * it to '(void *)check_uidgid if we have no info, NULL if 1083 * we tried and failed, or any other value if successful. 1084 */ 1085 static int 1086 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp, 1087 struct ucred **uc) 1088 { 1089 #if defined(USERSPACE) 1090 return 0; // not supported in userspace 1091 #else 1092 #ifndef __FreeBSD__ 1093 /* XXX */ 1094 return cred_check(insn, proto, oif, 1095 dst_ip, dst_port, src_ip, src_port, 1096 (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb); 1097 #else /* FreeBSD */ 1098 struct in_addr src_ip, dst_ip; 1099 struct inpcbinfo *pi; 1100 struct ipfw_flow_id *id; 1101 struct inpcb *pcb, *inp; 1102 int lookupflags; 1103 int match; 1104 1105 id = &args->f_id; 1106 inp = args->inp; 1107 1108 /* 1109 * Check to see if the UDP or TCP stack supplied us with 1110 * the PCB. If so, rather then holding a lock and looking 1111 * up the PCB, we can use the one that was supplied. 1112 */ 1113 if (inp && *ugid_lookupp == 0) { 1114 INP_LOCK_ASSERT(inp); 1115 if (inp->inp_socket != NULL) { 1116 *uc = crhold(inp->inp_cred); 1117 *ugid_lookupp = 1; 1118 } else 1119 *ugid_lookupp = -1; 1120 } 1121 /* 1122 * If we have already been here and the packet has no 1123 * PCB entry associated with it, then we can safely 1124 * assume that this is a no match. 1125 */ 1126 if (*ugid_lookupp == -1) 1127 return (0); 1128 if (id->proto == IPPROTO_TCP) { 1129 lookupflags = 0; 1130 pi = &V_tcbinfo; 1131 } else if (id->proto == IPPROTO_UDP) { 1132 lookupflags = INPLOOKUP_WILDCARD; 1133 pi = &V_udbinfo; 1134 } else if (id->proto == IPPROTO_UDPLITE) { 1135 lookupflags = INPLOOKUP_WILDCARD; 1136 pi = &V_ulitecbinfo; 1137 } else 1138 return 0; 1139 lookupflags |= INPLOOKUP_RLOCKPCB; 1140 match = 0; 1141 if (*ugid_lookupp == 0) { 1142 if (id->addr_type == 6) { 1143 #ifdef INET6 1144 if (args->flags & IPFW_ARGS_IN) 1145 pcb = in6_pcblookup_mbuf(pi, 1146 &id->src_ip6, htons(id->src_port), 1147 &id->dst_ip6, htons(id->dst_port), 1148 lookupflags, NULL, args->m); 1149 else 1150 pcb = in6_pcblookup_mbuf(pi, 1151 &id->dst_ip6, htons(id->dst_port), 1152 &id->src_ip6, htons(id->src_port), 1153 lookupflags, args->ifp, args->m); 1154 #else 1155 *ugid_lookupp = -1; 1156 return (0); 1157 #endif 1158 } else { 1159 src_ip.s_addr = htonl(id->src_ip); 1160 dst_ip.s_addr = htonl(id->dst_ip); 1161 if (args->flags & IPFW_ARGS_IN) 1162 pcb = in_pcblookup_mbuf(pi, 1163 src_ip, htons(id->src_port), 1164 dst_ip, htons(id->dst_port), 1165 lookupflags, NULL, args->m); 1166 else 1167 pcb = in_pcblookup_mbuf(pi, 1168 dst_ip, htons(id->dst_port), 1169 src_ip, htons(id->src_port), 1170 lookupflags, args->ifp, args->m); 1171 } 1172 if (pcb != NULL) { 1173 INP_RLOCK_ASSERT(pcb); 1174 *uc = crhold(pcb->inp_cred); 1175 *ugid_lookupp = 1; 1176 INP_RUNLOCK(pcb); 1177 } 1178 if (*ugid_lookupp == 0) { 1179 /* 1180 * We tried and failed, set the variable to -1 1181 * so we will not try again on this packet. 1182 */ 1183 *ugid_lookupp = -1; 1184 return (0); 1185 } 1186 } 1187 if (insn->o.opcode == O_UID) 1188 match = ((*uc)->cr_uid == (uid_t)insn->d[0]); 1189 else if (insn->o.opcode == O_GID) 1190 match = groupmember((gid_t)insn->d[0], *uc); 1191 else if (insn->o.opcode == O_JAIL) 1192 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]); 1193 return (match); 1194 #endif /* __FreeBSD__ */ 1195 #endif /* not supported in userspace */ 1196 } 1197 1198 /* 1199 * Helper function to set args with info on the rule after the matching 1200 * one. slot is precise, whereas we guess rule_id as they are 1201 * assigned sequentially. 1202 */ 1203 static inline void 1204 set_match(struct ip_fw_args *args, int slot, 1205 struct ip_fw_chain *chain) 1206 { 1207 args->rule.chain_id = chain->id; 1208 args->rule.slot = slot + 1; /* we use 0 as a marker */ 1209 args->rule.rule_id = 1 + chain->map[slot]->id; 1210 args->rule.rulenum = chain->map[slot]->rulenum; 1211 args->flags |= IPFW_ARGS_REF; 1212 } 1213 1214 #ifndef LINEAR_SKIPTO 1215 /* 1216 * Helper function to enable cached rule lookups using 1217 * cached_id and cached_pos fields in ipfw rule. 1218 */ 1219 static int 1220 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num, 1221 int tablearg, int jump_backwards) 1222 { 1223 int f_pos; 1224 1225 /* If possible use cached f_pos (in f->cached_pos), 1226 * whose version is written in f->cached_id 1227 * (horrible hacks to avoid changing the ABI). 1228 */ 1229 if (num != IP_FW_TARG && f->cached_id == chain->id) 1230 f_pos = f->cached_pos; 1231 else { 1232 int i = IP_FW_ARG_TABLEARG(chain, num, skipto); 1233 /* make sure we do not jump backward */ 1234 if (jump_backwards == 0 && i <= f->rulenum) 1235 i = f->rulenum + 1; 1236 if (chain->idxmap != NULL) 1237 f_pos = chain->idxmap[i]; 1238 else 1239 f_pos = ipfw_find_rule(chain, i, 0); 1240 /* update the cache */ 1241 if (num != IP_FW_TARG) { 1242 f->cached_id = chain->id; 1243 f->cached_pos = f_pos; 1244 } 1245 } 1246 1247 return (f_pos); 1248 } 1249 #else 1250 /* 1251 * Helper function to enable real fast rule lookups. 1252 */ 1253 static int 1254 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num, 1255 int tablearg, int jump_backwards) 1256 { 1257 int f_pos; 1258 1259 num = IP_FW_ARG_TABLEARG(chain, num, skipto); 1260 /* make sure we do not jump backward */ 1261 if (jump_backwards == 0 && num <= f->rulenum) 1262 num = f->rulenum + 1; 1263 f_pos = chain->idxmap[num]; 1264 1265 return (f_pos); 1266 } 1267 #endif 1268 1269 #define TARG(k, f) IP_FW_ARG_TABLEARG(chain, k, f) 1270 /* 1271 * The main check routine for the firewall. 1272 * 1273 * All arguments are in args so we can modify them and return them 1274 * back to the caller. 1275 * 1276 * Parameters: 1277 * 1278 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1279 * Starts with the IP header. 1280 * args->L3offset Number of bytes bypassed if we came from L2. 1281 * e.g. often sizeof(eh) ** NOTYET ** 1282 * args->ifp Incoming or outgoing interface. 1283 * args->divert_rule (in/out) 1284 * Skip up to the first rule past this rule number; 1285 * upon return, non-zero port number for divert or tee. 1286 * 1287 * args->rule Pointer to the last matching rule (in/out) 1288 * args->next_hop Socket we are forwarding to (out). 1289 * args->next_hop6 IPv6 next hop we are forwarding to (out). 1290 * args->f_id Addresses grabbed from the packet (out) 1291 * args->rule.info a cookie depending on rule action 1292 * 1293 * Return value: 1294 * 1295 * IP_FW_PASS the packet must be accepted 1296 * IP_FW_DENY the packet must be dropped 1297 * IP_FW_DIVERT divert packet, port in m_tag 1298 * IP_FW_TEE tee packet, port in m_tag 1299 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 1300 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 1301 * args->rule contains the matching rule, 1302 * args->rule.info has additional information. 1303 * 1304 */ 1305 int 1306 ipfw_chk(struct ip_fw_args *args) 1307 { 1308 1309 /* 1310 * Local variables holding state while processing a packet: 1311 * 1312 * IMPORTANT NOTE: to speed up the processing of rules, there 1313 * are some assumption on the values of the variables, which 1314 * are documented here. Should you change them, please check 1315 * the implementation of the various instructions to make sure 1316 * that they still work. 1317 * 1318 * m | args->m Pointer to the mbuf, as received from the caller. 1319 * It may change if ipfw_chk() does an m_pullup, or if it 1320 * consumes the packet because it calls send_reject(). 1321 * XXX This has to change, so that ipfw_chk() never modifies 1322 * or consumes the buffer. 1323 * OR 1324 * args->mem Pointer to contigous memory chunk. 1325 * ip Is the beginning of the ip(4 or 6) header. 1326 * eh Ethernet header in case if input is Layer2. 1327 */ 1328 struct mbuf *m; 1329 struct ip *ip; 1330 struct ether_header *eh; 1331 1332 /* 1333 * For rules which contain uid/gid or jail constraints, cache 1334 * a copy of the users credentials after the pcb lookup has been 1335 * executed. This will speed up the processing of rules with 1336 * these types of constraints, as well as decrease contention 1337 * on pcb related locks. 1338 */ 1339 #ifndef __FreeBSD__ 1340 struct bsd_ucred ucred_cache; 1341 #else 1342 struct ucred *ucred_cache = NULL; 1343 #endif 1344 int ucred_lookup = 0; 1345 int f_pos = 0; /* index of current rule in the array */ 1346 int retval = 0; 1347 struct ifnet *oif, *iif; 1348 1349 /* 1350 * hlen The length of the IP header. 1351 */ 1352 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 1353 1354 /* 1355 * offset The offset of a fragment. offset != 0 means that 1356 * we have a fragment at this offset of an IPv4 packet. 1357 * offset == 0 means that (if this is an IPv4 packet) 1358 * this is the first or only fragment. 1359 * For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header 1360 * or there is a single packet fragment (fragment header added 1361 * without needed). We will treat a single packet fragment as if 1362 * there was no fragment header (or log/block depending on the 1363 * V_fw_permit_single_frag6 sysctl setting). 1364 */ 1365 u_short offset = 0; 1366 u_short ip6f_mf = 0; 1367 1368 /* 1369 * Local copies of addresses. They are only valid if we have 1370 * an IP packet. 1371 * 1372 * proto The protocol. Set to 0 for non-ip packets, 1373 * or to the protocol read from the packet otherwise. 1374 * proto != 0 means that we have an IPv4 packet. 1375 * 1376 * src_port, dst_port port numbers, in HOST format. Only 1377 * valid for TCP and UDP packets. 1378 * 1379 * src_ip, dst_ip ip addresses, in NETWORK format. 1380 * Only valid for IPv4 packets. 1381 */ 1382 uint8_t proto; 1383 uint16_t src_port, dst_port; /* NOTE: host format */ 1384 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 1385 int iplen = 0; 1386 int pktlen; 1387 1388 struct ipfw_dyn_info dyn_info; 1389 struct ip_fw *q = NULL; 1390 struct ip_fw_chain *chain = &V_layer3_chain; 1391 1392 /* 1393 * We store in ulp a pointer to the upper layer protocol header. 1394 * In the ipv4 case this is easy to determine from the header, 1395 * but for ipv6 we might have some additional headers in the middle. 1396 * ulp is NULL if not found. 1397 */ 1398 void *ulp = NULL; /* upper layer protocol pointer. */ 1399 1400 /* XXX ipv6 variables */ 1401 int is_ipv6 = 0; 1402 uint8_t icmp6_type = 0; 1403 uint16_t ext_hd = 0; /* bits vector for extension header filtering */ 1404 /* end of ipv6 variables */ 1405 1406 int is_ipv4 = 0; 1407 1408 int done = 0; /* flag to exit the outer loop */ 1409 IPFW_RLOCK_TRACKER; 1410 bool mem; 1411 1412 if ((mem = (args->flags & IPFW_ARGS_LENMASK))) { 1413 if (args->flags & IPFW_ARGS_ETHER) { 1414 eh = (struct ether_header *)args->mem; 1415 if (eh->ether_type == htons(ETHERTYPE_VLAN)) 1416 ip = (struct ip *) 1417 ((struct ether_vlan_header *)eh + 1); 1418 else 1419 ip = (struct ip *)(eh + 1); 1420 } else { 1421 eh = NULL; 1422 ip = (struct ip *)args->mem; 1423 } 1424 pktlen = IPFW_ARGS_LENGTH(args->flags); 1425 args->f_id.fib = args->ifp->if_fib; /* best guess */ 1426 } else { 1427 m = args->m; 1428 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready)) 1429 return (IP_FW_PASS); /* accept */ 1430 if (args->flags & IPFW_ARGS_ETHER) { 1431 /* We need some amount of data to be contiguous. */ 1432 if (m->m_len < min(m->m_pkthdr.len, max_protohdr) && 1433 (args->m = m = m_pullup(m, min(m->m_pkthdr.len, 1434 max_protohdr))) == NULL) 1435 goto pullup_failed; 1436 eh = mtod(m, struct ether_header *); 1437 ip = (struct ip *)(eh + 1); 1438 } else { 1439 eh = NULL; 1440 ip = mtod(m, struct ip *); 1441 } 1442 pktlen = m->m_pkthdr.len; 1443 args->f_id.fib = M_GETFIB(m); /* mbuf not altered */ 1444 } 1445 1446 dst_ip.s_addr = 0; /* make sure it is initialized */ 1447 src_ip.s_addr = 0; /* make sure it is initialized */ 1448 src_port = dst_port = 0; 1449 1450 DYN_INFO_INIT(&dyn_info); 1451 /* 1452 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 1453 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 1454 * pointer might become stale after other pullups (but we never use it 1455 * this way). 1456 */ 1457 #define PULLUP_TO(_len, p, T) PULLUP_LEN(_len, p, sizeof(T)) 1458 #define EHLEN (eh != NULL ? ((char *)ip - (char *)eh) : 0) 1459 #define _PULLUP_LOCKED(_len, p, T, unlock) \ 1460 do { \ 1461 int x = (_len) + T + EHLEN; \ 1462 if (mem) { \ 1463 if (__predict_false(pktlen < x)) { \ 1464 unlock; \ 1465 goto pullup_failed; \ 1466 } \ 1467 p = (char *)args->mem + (_len) + EHLEN; \ 1468 } else { \ 1469 if (__predict_false((m)->m_len < x)) { \ 1470 args->m = m = m_pullup(m, x); \ 1471 if (m == NULL) { \ 1472 unlock; \ 1473 goto pullup_failed; \ 1474 } \ 1475 } \ 1476 p = mtod(m, char *) + (_len) + EHLEN; \ 1477 } \ 1478 } while (0) 1479 1480 #define PULLUP_LEN(_len, p, T) _PULLUP_LOCKED(_len, p, T, ) 1481 #define PULLUP_LEN_LOCKED(_len, p, T) \ 1482 _PULLUP_LOCKED(_len, p, T, IPFW_PF_RUNLOCK(chain)); \ 1483 UPDATE_POINTERS() 1484 /* 1485 * In case pointers got stale after pullups, update them. 1486 */ 1487 #define UPDATE_POINTERS() \ 1488 do { \ 1489 if (!mem) { \ 1490 if (eh != NULL) { \ 1491 eh = mtod(m, struct ether_header *); \ 1492 ip = (struct ip *)(eh + 1); \ 1493 } else \ 1494 ip = mtod(m, struct ip *); \ 1495 args->m = m; \ 1496 } \ 1497 } while (0) 1498 1499 /* Identify IP packets and fill up variables. */ 1500 if (pktlen >= sizeof(struct ip6_hdr) && 1501 (eh == NULL || eh->ether_type == htons(ETHERTYPE_IPV6)) && 1502 ip->ip_v == 6) { 1503 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 1504 1505 is_ipv6 = 1; 1506 args->flags |= IPFW_ARGS_IP6; 1507 hlen = sizeof(struct ip6_hdr); 1508 proto = ip6->ip6_nxt; 1509 /* Search extension headers to find upper layer protocols */ 1510 while (ulp == NULL && offset == 0) { 1511 switch (proto) { 1512 case IPPROTO_ICMPV6: 1513 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 1514 icmp6_type = ICMP6(ulp)->icmp6_type; 1515 break; 1516 1517 case IPPROTO_TCP: 1518 PULLUP_TO(hlen, ulp, struct tcphdr); 1519 dst_port = TCP(ulp)->th_dport; 1520 src_port = TCP(ulp)->th_sport; 1521 /* save flags for dynamic rules */ 1522 args->f_id._flags = TCP(ulp)->th_flags; 1523 break; 1524 1525 case IPPROTO_SCTP: 1526 if (pktlen >= hlen + sizeof(struct sctphdr) + 1527 sizeof(struct sctp_chunkhdr) + 1528 offsetof(struct sctp_init, a_rwnd)) 1529 PULLUP_LEN(hlen, ulp, 1530 sizeof(struct sctphdr) + 1531 sizeof(struct sctp_chunkhdr) + 1532 offsetof(struct sctp_init, a_rwnd)); 1533 else if (pktlen >= hlen + sizeof(struct sctphdr)) 1534 PULLUP_LEN(hlen, ulp, pktlen - hlen); 1535 else 1536 PULLUP_LEN(hlen, ulp, 1537 sizeof(struct sctphdr)); 1538 src_port = SCTP(ulp)->src_port; 1539 dst_port = SCTP(ulp)->dest_port; 1540 break; 1541 1542 case IPPROTO_UDP: 1543 case IPPROTO_UDPLITE: 1544 PULLUP_TO(hlen, ulp, struct udphdr); 1545 dst_port = UDP(ulp)->uh_dport; 1546 src_port = UDP(ulp)->uh_sport; 1547 break; 1548 1549 case IPPROTO_HOPOPTS: /* RFC 2460 */ 1550 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1551 ext_hd |= EXT_HOPOPTS; 1552 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1553 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1554 ulp = NULL; 1555 break; 1556 1557 case IPPROTO_ROUTING: /* RFC 2460 */ 1558 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 1559 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 1560 case 0: 1561 ext_hd |= EXT_RTHDR0; 1562 break; 1563 case 2: 1564 ext_hd |= EXT_RTHDR2; 1565 break; 1566 default: 1567 if (V_fw_verbose) 1568 printf("IPFW2: IPV6 - Unknown " 1569 "Routing Header type(%d)\n", 1570 ((struct ip6_rthdr *) 1571 ulp)->ip6r_type); 1572 if (V_fw_deny_unknown_exthdrs) 1573 return (IP_FW_DENY); 1574 break; 1575 } 1576 ext_hd |= EXT_ROUTING; 1577 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 1578 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 1579 ulp = NULL; 1580 break; 1581 1582 case IPPROTO_FRAGMENT: /* RFC 2460 */ 1583 PULLUP_TO(hlen, ulp, struct ip6_frag); 1584 ext_hd |= EXT_FRAGMENT; 1585 hlen += sizeof (struct ip6_frag); 1586 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 1587 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 1588 IP6F_OFF_MASK; 1589 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg & 1590 IP6F_MORE_FRAG; 1591 if (V_fw_permit_single_frag6 == 0 && 1592 offset == 0 && ip6f_mf == 0) { 1593 if (V_fw_verbose) 1594 printf("IPFW2: IPV6 - Invalid " 1595 "Fragment Header\n"); 1596 if (V_fw_deny_unknown_exthdrs) 1597 return (IP_FW_DENY); 1598 break; 1599 } 1600 args->f_id.extra = 1601 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 1602 ulp = NULL; 1603 break; 1604 1605 case IPPROTO_DSTOPTS: /* RFC 2460 */ 1606 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1607 ext_hd |= EXT_DSTOPTS; 1608 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1609 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1610 ulp = NULL; 1611 break; 1612 1613 case IPPROTO_AH: /* RFC 2402 */ 1614 PULLUP_TO(hlen, ulp, struct ip6_ext); 1615 ext_hd |= EXT_AH; 1616 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 1617 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 1618 ulp = NULL; 1619 break; 1620 1621 case IPPROTO_ESP: /* RFC 2406 */ 1622 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 1623 /* Anything past Seq# is variable length and 1624 * data past this ext. header is encrypted. */ 1625 ext_hd |= EXT_ESP; 1626 break; 1627 1628 case IPPROTO_NONE: /* RFC 2460 */ 1629 /* 1630 * Packet ends here, and IPv6 header has 1631 * already been pulled up. If ip6e_len!=0 1632 * then octets must be ignored. 1633 */ 1634 ulp = ip; /* non-NULL to get out of loop. */ 1635 break; 1636 1637 case IPPROTO_OSPFIGP: 1638 /* XXX OSPF header check? */ 1639 PULLUP_TO(hlen, ulp, struct ip6_ext); 1640 break; 1641 1642 case IPPROTO_PIM: 1643 /* XXX PIM header check? */ 1644 PULLUP_TO(hlen, ulp, struct pim); 1645 break; 1646 1647 case IPPROTO_GRE: /* RFC 1701 */ 1648 /* XXX GRE header check? */ 1649 PULLUP_TO(hlen, ulp, struct grehdr); 1650 break; 1651 1652 case IPPROTO_CARP: 1653 PULLUP_TO(hlen, ulp, offsetof( 1654 struct carp_header, carp_counter)); 1655 if (CARP_ADVERTISEMENT != 1656 ((struct carp_header *)ulp)->carp_type) 1657 return (IP_FW_DENY); 1658 break; 1659 1660 case IPPROTO_IPV6: /* RFC 2893 */ 1661 PULLUP_TO(hlen, ulp, struct ip6_hdr); 1662 break; 1663 1664 case IPPROTO_IPV4: /* RFC 2893 */ 1665 PULLUP_TO(hlen, ulp, struct ip); 1666 break; 1667 1668 default: 1669 if (V_fw_verbose) 1670 printf("IPFW2: IPV6 - Unknown " 1671 "Extension Header(%d), ext_hd=%x\n", 1672 proto, ext_hd); 1673 if (V_fw_deny_unknown_exthdrs) 1674 return (IP_FW_DENY); 1675 PULLUP_TO(hlen, ulp, struct ip6_ext); 1676 break; 1677 } /*switch */ 1678 } 1679 UPDATE_POINTERS(); 1680 ip6 = (struct ip6_hdr *)ip; 1681 args->f_id.addr_type = 6; 1682 args->f_id.src_ip6 = ip6->ip6_src; 1683 args->f_id.dst_ip6 = ip6->ip6_dst; 1684 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 1685 iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6); 1686 } else if (pktlen >= sizeof(struct ip) && 1687 (eh == NULL || eh->ether_type == htons(ETHERTYPE_IP)) && 1688 ip->ip_v == 4) { 1689 is_ipv4 = 1; 1690 args->flags |= IPFW_ARGS_IP4; 1691 hlen = ip->ip_hl << 2; 1692 /* 1693 * Collect parameters into local variables for faster 1694 * matching. 1695 */ 1696 proto = ip->ip_p; 1697 src_ip = ip->ip_src; 1698 dst_ip = ip->ip_dst; 1699 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1700 iplen = ntohs(ip->ip_len); 1701 1702 if (offset == 0) { 1703 switch (proto) { 1704 case IPPROTO_TCP: 1705 PULLUP_TO(hlen, ulp, struct tcphdr); 1706 dst_port = TCP(ulp)->th_dport; 1707 src_port = TCP(ulp)->th_sport; 1708 /* save flags for dynamic rules */ 1709 args->f_id._flags = TCP(ulp)->th_flags; 1710 break; 1711 1712 case IPPROTO_SCTP: 1713 if (pktlen >= hlen + sizeof(struct sctphdr) + 1714 sizeof(struct sctp_chunkhdr) + 1715 offsetof(struct sctp_init, a_rwnd)) 1716 PULLUP_LEN(hlen, ulp, 1717 sizeof(struct sctphdr) + 1718 sizeof(struct sctp_chunkhdr) + 1719 offsetof(struct sctp_init, a_rwnd)); 1720 else if (pktlen >= hlen + sizeof(struct sctphdr)) 1721 PULLUP_LEN(hlen, ulp, pktlen - hlen); 1722 else 1723 PULLUP_LEN(hlen, ulp, 1724 sizeof(struct sctphdr)); 1725 src_port = SCTP(ulp)->src_port; 1726 dst_port = SCTP(ulp)->dest_port; 1727 break; 1728 1729 case IPPROTO_UDP: 1730 case IPPROTO_UDPLITE: 1731 PULLUP_TO(hlen, ulp, struct udphdr); 1732 dst_port = UDP(ulp)->uh_dport; 1733 src_port = UDP(ulp)->uh_sport; 1734 break; 1735 1736 case IPPROTO_ICMP: 1737 PULLUP_TO(hlen, ulp, struct icmphdr); 1738 //args->f_id.flags = ICMP(ulp)->icmp_type; 1739 break; 1740 1741 default: 1742 break; 1743 } 1744 } else { 1745 if (offset == 1 && proto == IPPROTO_TCP) { 1746 /* RFC 3128 */ 1747 goto pullup_failed; 1748 } 1749 } 1750 1751 UPDATE_POINTERS(); 1752 args->f_id.addr_type = 4; 1753 args->f_id.src_ip = ntohl(src_ip.s_addr); 1754 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1755 } else { 1756 proto = 0; 1757 dst_ip.s_addr = src_ip.s_addr = 0; 1758 1759 args->f_id.addr_type = 1; /* XXX */ 1760 } 1761 #undef PULLUP_TO 1762 pktlen = iplen < pktlen ? iplen: pktlen; 1763 1764 /* Properly initialize the rest of f_id */ 1765 args->f_id.proto = proto; 1766 args->f_id.src_port = src_port = ntohs(src_port); 1767 args->f_id.dst_port = dst_port = ntohs(dst_port); 1768 1769 IPFW_PF_RLOCK(chain); 1770 if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */ 1771 IPFW_PF_RUNLOCK(chain); 1772 return (IP_FW_PASS); /* accept */ 1773 } 1774 if (args->flags & IPFW_ARGS_REF) { 1775 /* 1776 * Packet has already been tagged as a result of a previous 1777 * match on rule args->rule aka args->rule_id (PIPE, QUEUE, 1778 * REASS, NETGRAPH, DIVERT/TEE...) 1779 * Validate the slot and continue from the next one 1780 * if still present, otherwise do a lookup. 1781 */ 1782 f_pos = (args->rule.chain_id == chain->id) ? 1783 args->rule.slot : 1784 ipfw_find_rule(chain, args->rule.rulenum, 1785 args->rule.rule_id); 1786 } else { 1787 f_pos = 0; 1788 } 1789 1790 if (args->flags & IPFW_ARGS_IN) { 1791 iif = args->ifp; 1792 oif = NULL; 1793 } else { 1794 MPASS(args->flags & IPFW_ARGS_OUT); 1795 iif = mem ? NULL : m_rcvif(m); 1796 oif = args->ifp; 1797 } 1798 1799 /* 1800 * Now scan the rules, and parse microinstructions for each rule. 1801 * We have two nested loops and an inner switch. Sometimes we 1802 * need to break out of one or both loops, or re-enter one of 1803 * the loops with updated variables. Loop variables are: 1804 * 1805 * f_pos (outer loop) points to the current rule. 1806 * On output it points to the matching rule. 1807 * done (outer loop) is used as a flag to break the loop. 1808 * l (inner loop) residual length of current rule. 1809 * cmd points to the current microinstruction. 1810 * 1811 * We break the inner loop by setting l=0 and possibly 1812 * cmdlen=0 if we don't want to advance cmd. 1813 * We break the outer loop by setting done=1 1814 * We can restart the inner loop by setting l>0 and f_pos, f, cmd 1815 * as needed. 1816 */ 1817 for (; f_pos < chain->n_rules; f_pos++) { 1818 ipfw_insn *cmd; 1819 uint32_t tablearg = 0; 1820 int l, cmdlen, skip_or; /* skip rest of OR block */ 1821 struct ip_fw *f; 1822 1823 f = chain->map[f_pos]; 1824 if (V_set_disable & (1 << f->set) ) 1825 continue; 1826 1827 skip_or = 0; 1828 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1829 l -= cmdlen, cmd += cmdlen) { 1830 int match; 1831 1832 /* 1833 * check_body is a jump target used when we find a 1834 * CHECK_STATE, and need to jump to the body of 1835 * the target rule. 1836 */ 1837 1838 /* check_body: */ 1839 cmdlen = F_LEN(cmd); 1840 /* 1841 * An OR block (insn_1 || .. || insn_n) has the 1842 * F_OR bit set in all but the last instruction. 1843 * The first match will set "skip_or", and cause 1844 * the following instructions to be skipped until 1845 * past the one with the F_OR bit clear. 1846 */ 1847 if (skip_or) { /* skip this instruction */ 1848 if ((cmd->len & F_OR) == 0) 1849 skip_or = 0; /* next one is good */ 1850 continue; 1851 } 1852 match = 0; /* set to 1 if we succeed */ 1853 1854 switch (cmd->opcode) { 1855 /* 1856 * The first set of opcodes compares the packet's 1857 * fields with some pattern, setting 'match' if a 1858 * match is found. At the end of the loop there is 1859 * logic to deal with F_NOT and F_OR flags associated 1860 * with the opcode. 1861 */ 1862 case O_NOP: 1863 match = 1; 1864 break; 1865 1866 case O_FORWARD_MAC: 1867 printf("ipfw: opcode %d unimplemented\n", 1868 cmd->opcode); 1869 break; 1870 1871 case O_GID: 1872 case O_UID: 1873 case O_JAIL: 1874 /* 1875 * We only check offset == 0 && proto != 0, 1876 * as this ensures that we have a 1877 * packet with the ports info. 1878 */ 1879 if (offset != 0) 1880 break; 1881 if (proto == IPPROTO_TCP || 1882 proto == IPPROTO_UDP || 1883 proto == IPPROTO_UDPLITE) 1884 match = check_uidgid( 1885 (ipfw_insn_u32 *)cmd, 1886 args, &ucred_lookup, 1887 #ifdef __FreeBSD__ 1888 &ucred_cache); 1889 #else 1890 (void *)&ucred_cache); 1891 #endif 1892 break; 1893 1894 case O_RECV: 1895 match = iface_match(iif, (ipfw_insn_if *)cmd, 1896 chain, &tablearg); 1897 break; 1898 1899 case O_XMIT: 1900 match = iface_match(oif, (ipfw_insn_if *)cmd, 1901 chain, &tablearg); 1902 break; 1903 1904 case O_VIA: 1905 match = iface_match(args->ifp, 1906 (ipfw_insn_if *)cmd, chain, &tablearg); 1907 break; 1908 1909 case O_MACADDR2: 1910 if (args->flags & IPFW_ARGS_ETHER) { 1911 u_int32_t *want = (u_int32_t *) 1912 ((ipfw_insn_mac *)cmd)->addr; 1913 u_int32_t *mask = (u_int32_t *) 1914 ((ipfw_insn_mac *)cmd)->mask; 1915 u_int32_t *hdr = (u_int32_t *)eh; 1916 1917 match = 1918 ( want[0] == (hdr[0] & mask[0]) && 1919 want[1] == (hdr[1] & mask[1]) && 1920 want[2] == (hdr[2] & mask[2]) ); 1921 } 1922 break; 1923 1924 case O_MAC_TYPE: 1925 if (args->flags & IPFW_ARGS_ETHER) { 1926 u_int16_t *p = 1927 ((ipfw_insn_u16 *)cmd)->ports; 1928 int i; 1929 1930 for (i = cmdlen - 1; !match && i>0; 1931 i--, p += 2) 1932 match = 1933 (ntohs(eh->ether_type) >= 1934 p[0] && 1935 ntohs(eh->ether_type) <= 1936 p[1]); 1937 } 1938 break; 1939 1940 case O_FRAG: 1941 match = (offset != 0); 1942 break; 1943 1944 case O_IN: /* "out" is "not in" */ 1945 match = (oif == NULL); 1946 break; 1947 1948 case O_LAYER2: 1949 match = (args->flags & IPFW_ARGS_ETHER); 1950 break; 1951 1952 case O_DIVERTED: 1953 if ((args->flags & IPFW_ARGS_REF) == 0) 1954 break; 1955 /* 1956 * For diverted packets, args->rule.info 1957 * contains the divert port (in host format) 1958 * reason and direction. 1959 */ 1960 match = ((args->rule.info & IPFW_IS_MASK) == 1961 IPFW_IS_DIVERT) && ( 1962 ((args->rule.info & IPFW_INFO_IN) ? 1963 1: 2) & cmd->arg1); 1964 break; 1965 1966 case O_PROTO: 1967 /* 1968 * We do not allow an arg of 0 so the 1969 * check of "proto" only suffices. 1970 */ 1971 match = (proto == cmd->arg1); 1972 break; 1973 1974 case O_IP_SRC: 1975 match = is_ipv4 && 1976 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1977 src_ip.s_addr); 1978 break; 1979 1980 case O_IP_DST_LOOKUP: 1981 { 1982 void *pkey; 1983 uint32_t vidx, key; 1984 uint16_t keylen; 1985 1986 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) { 1987 /* Determine lookup key type */ 1988 vidx = ((ipfw_insn_u32 *)cmd)->d[1]; 1989 if (vidx != 4 /* uid */ && 1990 vidx != 5 /* jail */ && 1991 is_ipv6 == 0 && is_ipv4 == 0) 1992 break; 1993 /* Determine key length */ 1994 if (vidx == 0 /* dst-ip */ || 1995 vidx == 1 /* src-ip */) 1996 keylen = is_ipv6 ? 1997 sizeof(struct in6_addr): 1998 sizeof(in_addr_t); 1999 else { 2000 keylen = sizeof(key); 2001 pkey = &key; 2002 } 2003 if (vidx == 0 /* dst-ip */) 2004 pkey = is_ipv4 ? (void *)&dst_ip: 2005 (void *)&args->f_id.dst_ip6; 2006 else if (vidx == 1 /* src-ip */) 2007 pkey = is_ipv4 ? (void *)&src_ip: 2008 (void *)&args->f_id.src_ip6; 2009 else if (vidx == 6 /* dscp */) { 2010 if (is_ipv4) 2011 key = ip->ip_tos >> 2; 2012 else { 2013 key = args->f_id.flow_id6; 2014 key = (key & 0x0f) << 2 | 2015 (key & 0xf000) >> 14; 2016 } 2017 key &= 0x3f; 2018 } else if (vidx == 2 /* dst-port */ || 2019 vidx == 3 /* src-port */) { 2020 /* Skip fragments */ 2021 if (offset != 0) 2022 break; 2023 /* Skip proto without ports */ 2024 if (proto != IPPROTO_TCP && 2025 proto != IPPROTO_UDP && 2026 proto != IPPROTO_UDPLITE && 2027 proto != IPPROTO_SCTP) 2028 break; 2029 if (vidx == 2 /* dst-port */) 2030 key = dst_port; 2031 else 2032 key = src_port; 2033 } 2034 #ifndef USERSPACE 2035 else if (vidx == 4 /* uid */ || 2036 vidx == 5 /* jail */) { 2037 check_uidgid( 2038 (ipfw_insn_u32 *)cmd, 2039 args, &ucred_lookup, 2040 #ifdef __FreeBSD__ 2041 &ucred_cache); 2042 if (vidx == 4 /* uid */) 2043 key = ucred_cache->cr_uid; 2044 else if (vidx == 5 /* jail */) 2045 key = ucred_cache->cr_prison->pr_id; 2046 #else /* !__FreeBSD__ */ 2047 (void *)&ucred_cache); 2048 if (vidx == 4 /* uid */) 2049 key = ucred_cache.uid; 2050 else if (vidx == 5 /* jail */) 2051 key = ucred_cache.xid; 2052 #endif /* !__FreeBSD__ */ 2053 } 2054 #endif /* !USERSPACE */ 2055 else 2056 break; 2057 match = ipfw_lookup_table(chain, 2058 cmd->arg1, keylen, pkey, &vidx); 2059 if (!match) 2060 break; 2061 tablearg = vidx; 2062 break; 2063 } 2064 /* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */ 2065 /* FALLTHROUGH */ 2066 } 2067 case O_IP_SRC_LOOKUP: 2068 { 2069 void *pkey; 2070 uint32_t vidx; 2071 uint16_t keylen; 2072 2073 if (is_ipv4) { 2074 keylen = sizeof(in_addr_t); 2075 if (cmd->opcode == O_IP_DST_LOOKUP) 2076 pkey = &dst_ip; 2077 else 2078 pkey = &src_ip; 2079 } else if (is_ipv6) { 2080 keylen = sizeof(struct in6_addr); 2081 if (cmd->opcode == O_IP_DST_LOOKUP) 2082 pkey = &args->f_id.dst_ip6; 2083 else 2084 pkey = &args->f_id.src_ip6; 2085 } else 2086 break; 2087 match = ipfw_lookup_table(chain, cmd->arg1, 2088 keylen, pkey, &vidx); 2089 if (!match) 2090 break; 2091 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) { 2092 match = ((ipfw_insn_u32 *)cmd)->d[0] == 2093 TARG_VAL(chain, vidx, tag); 2094 if (!match) 2095 break; 2096 } 2097 tablearg = vidx; 2098 break; 2099 } 2100 2101 case O_IP_FLOW_LOOKUP: 2102 { 2103 uint32_t v = 0; 2104 match = ipfw_lookup_table(chain, 2105 cmd->arg1, 0, &args->f_id, &v); 2106 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2107 match = ((ipfw_insn_u32 *)cmd)->d[0] == 2108 TARG_VAL(chain, v, tag); 2109 if (match) 2110 tablearg = v; 2111 } 2112 break; 2113 case O_IP_SRC_MASK: 2114 case O_IP_DST_MASK: 2115 if (is_ipv4) { 2116 uint32_t a = 2117 (cmd->opcode == O_IP_DST_MASK) ? 2118 dst_ip.s_addr : src_ip.s_addr; 2119 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2120 int i = cmdlen-1; 2121 2122 for (; !match && i>0; i-= 2, p+= 2) 2123 match = (p[0] == (a & p[1])); 2124 } 2125 break; 2126 2127 case O_IP_SRC_ME: 2128 if (is_ipv4) { 2129 match = in_localip(src_ip); 2130 break; 2131 } 2132 #ifdef INET6 2133 /* FALLTHROUGH */ 2134 case O_IP6_SRC_ME: 2135 match = is_ipv6 && 2136 ipfw_localip6(&args->f_id.src_ip6); 2137 #endif 2138 break; 2139 2140 case O_IP_DST_SET: 2141 case O_IP_SRC_SET: 2142 if (is_ipv4) { 2143 u_int32_t *d = (u_int32_t *)(cmd+1); 2144 u_int32_t addr = 2145 cmd->opcode == O_IP_DST_SET ? 2146 args->f_id.dst_ip : 2147 args->f_id.src_ip; 2148 2149 if (addr < d[0]) 2150 break; 2151 addr -= d[0]; /* subtract base */ 2152 match = (addr < cmd->arg1) && 2153 ( d[ 1 + (addr>>5)] & 2154 (1<<(addr & 0x1f)) ); 2155 } 2156 break; 2157 2158 case O_IP_DST: 2159 match = is_ipv4 && 2160 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2161 dst_ip.s_addr); 2162 break; 2163 2164 case O_IP_DST_ME: 2165 if (is_ipv4) { 2166 match = in_localip(dst_ip); 2167 break; 2168 } 2169 #ifdef INET6 2170 /* FALLTHROUGH */ 2171 case O_IP6_DST_ME: 2172 match = is_ipv6 && 2173 ipfw_localip6(&args->f_id.dst_ip6); 2174 #endif 2175 break; 2176 2177 2178 case O_IP_SRCPORT: 2179 case O_IP_DSTPORT: 2180 /* 2181 * offset == 0 && proto != 0 is enough 2182 * to guarantee that we have a 2183 * packet with port info. 2184 */ 2185 if ((proto == IPPROTO_UDP || 2186 proto == IPPROTO_UDPLITE || 2187 proto == IPPROTO_TCP || 2188 proto == IPPROTO_SCTP) && offset == 0) { 2189 u_int16_t x = 2190 (cmd->opcode == O_IP_SRCPORT) ? 2191 src_port : dst_port ; 2192 u_int16_t *p = 2193 ((ipfw_insn_u16 *)cmd)->ports; 2194 int i; 2195 2196 for (i = cmdlen - 1; !match && i>0; 2197 i--, p += 2) 2198 match = (x>=p[0] && x<=p[1]); 2199 } 2200 break; 2201 2202 case O_ICMPTYPE: 2203 match = (offset == 0 && proto==IPPROTO_ICMP && 2204 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2205 break; 2206 2207 #ifdef INET6 2208 case O_ICMP6TYPE: 2209 match = is_ipv6 && offset == 0 && 2210 proto==IPPROTO_ICMPV6 && 2211 icmp6type_match( 2212 ICMP6(ulp)->icmp6_type, 2213 (ipfw_insn_u32 *)cmd); 2214 break; 2215 #endif /* INET6 */ 2216 2217 case O_IPOPT: 2218 match = (is_ipv4 && 2219 ipopts_match(ip, cmd) ); 2220 break; 2221 2222 case O_IPVER: 2223 match = (is_ipv4 && 2224 cmd->arg1 == ip->ip_v); 2225 break; 2226 2227 case O_IPID: 2228 case O_IPTTL: 2229 if (!is_ipv4) 2230 break; 2231 case O_IPLEN: 2232 { /* only for IP packets */ 2233 uint16_t x; 2234 uint16_t *p; 2235 int i; 2236 2237 if (cmd->opcode == O_IPLEN) 2238 x = iplen; 2239 else if (cmd->opcode == O_IPTTL) 2240 x = ip->ip_ttl; 2241 else /* must be IPID */ 2242 x = ntohs(ip->ip_id); 2243 if (cmdlen == 1) { 2244 match = (cmd->arg1 == x); 2245 break; 2246 } 2247 /* otherwise we have ranges */ 2248 p = ((ipfw_insn_u16 *)cmd)->ports; 2249 i = cmdlen - 1; 2250 for (; !match && i>0; i--, p += 2) 2251 match = (x >= p[0] && x <= p[1]); 2252 } 2253 break; 2254 2255 case O_IPPRECEDENCE: 2256 match = (is_ipv4 && 2257 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2258 break; 2259 2260 case O_IPTOS: 2261 match = (is_ipv4 && 2262 flags_match(cmd, ip->ip_tos)); 2263 break; 2264 2265 case O_DSCP: 2266 { 2267 uint32_t *p; 2268 uint16_t x; 2269 2270 p = ((ipfw_insn_u32 *)cmd)->d; 2271 2272 if (is_ipv4) 2273 x = ip->ip_tos >> 2; 2274 else if (is_ipv6) { 2275 uint8_t *v; 2276 v = &((struct ip6_hdr *)ip)->ip6_vfc; 2277 x = (*v & 0x0F) << 2; 2278 v++; 2279 x |= *v >> 6; 2280 } else 2281 break; 2282 2283 /* DSCP bitmask is stored as low_u32 high_u32 */ 2284 if (x >= 32) 2285 match = *(p + 1) & (1 << (x - 32)); 2286 else 2287 match = *p & (1 << x); 2288 } 2289 break; 2290 2291 case O_TCPDATALEN: 2292 if (proto == IPPROTO_TCP && offset == 0) { 2293 struct tcphdr *tcp; 2294 uint16_t x; 2295 uint16_t *p; 2296 int i; 2297 #ifdef INET6 2298 if (is_ipv6) { 2299 struct ip6_hdr *ip6; 2300 2301 ip6 = (struct ip6_hdr *)ip; 2302 if (ip6->ip6_plen == 0) { 2303 /* 2304 * Jumbo payload is not 2305 * supported by this 2306 * opcode. 2307 */ 2308 break; 2309 } 2310 x = iplen - hlen; 2311 } else 2312 #endif /* INET6 */ 2313 x = iplen - (ip->ip_hl << 2); 2314 tcp = TCP(ulp); 2315 x -= tcp->th_off << 2; 2316 if (cmdlen == 1) { 2317 match = (cmd->arg1 == x); 2318 break; 2319 } 2320 /* otherwise we have ranges */ 2321 p = ((ipfw_insn_u16 *)cmd)->ports; 2322 i = cmdlen - 1; 2323 for (; !match && i>0; i--, p += 2) 2324 match = (x >= p[0] && x <= p[1]); 2325 } 2326 break; 2327 2328 case O_TCPFLAGS: 2329 match = (proto == IPPROTO_TCP && offset == 0 && 2330 flags_match(cmd, TCP(ulp)->th_flags)); 2331 break; 2332 2333 case O_TCPOPTS: 2334 if (proto == IPPROTO_TCP && offset == 0 && ulp){ 2335 PULLUP_LEN_LOCKED(hlen, ulp, 2336 (TCP(ulp)->th_off << 2)); 2337 match = tcpopts_match(TCP(ulp), cmd); 2338 } 2339 break; 2340 2341 case O_TCPSEQ: 2342 match = (proto == IPPROTO_TCP && offset == 0 && 2343 ((ipfw_insn_u32 *)cmd)->d[0] == 2344 TCP(ulp)->th_seq); 2345 break; 2346 2347 case O_TCPACK: 2348 match = (proto == IPPROTO_TCP && offset == 0 && 2349 ((ipfw_insn_u32 *)cmd)->d[0] == 2350 TCP(ulp)->th_ack); 2351 break; 2352 2353 case O_TCPMSS: 2354 if (proto == IPPROTO_TCP && 2355 (args->f_id._flags & TH_SYN) != 0 && 2356 ulp != NULL) { 2357 uint16_t mss, *p; 2358 int i; 2359 2360 PULLUP_LEN_LOCKED(hlen, ulp, 2361 (TCP(ulp)->th_off << 2)); 2362 if ((tcpopts_parse(TCP(ulp), &mss) & 2363 IP_FW_TCPOPT_MSS) == 0) 2364 break; 2365 if (cmdlen == 1) { 2366 match = (cmd->arg1 == mss); 2367 break; 2368 } 2369 /* Otherwise we have ranges. */ 2370 p = ((ipfw_insn_u16 *)cmd)->ports; 2371 i = cmdlen - 1; 2372 for (; !match && i > 0; i--, p += 2) 2373 match = (mss >= p[0] && 2374 mss <= p[1]); 2375 } 2376 break; 2377 2378 case O_TCPWIN: 2379 if (proto == IPPROTO_TCP && offset == 0) { 2380 uint16_t x; 2381 uint16_t *p; 2382 int i; 2383 2384 x = ntohs(TCP(ulp)->th_win); 2385 if (cmdlen == 1) { 2386 match = (cmd->arg1 == x); 2387 break; 2388 } 2389 /* Otherwise we have ranges. */ 2390 p = ((ipfw_insn_u16 *)cmd)->ports; 2391 i = cmdlen - 1; 2392 for (; !match && i > 0; i--, p += 2) 2393 match = (x >= p[0] && x <= p[1]); 2394 } 2395 break; 2396 2397 case O_ESTAB: 2398 /* reject packets which have SYN only */ 2399 /* XXX should i also check for TH_ACK ? */ 2400 match = (proto == IPPROTO_TCP && offset == 0 && 2401 (TCP(ulp)->th_flags & 2402 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2403 break; 2404 2405 case O_ALTQ: { 2406 struct pf_mtag *at; 2407 struct m_tag *mtag; 2408 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2409 2410 /* 2411 * ALTQ uses mbuf tags from another 2412 * packet filtering system - pf(4). 2413 * We allocate a tag in its format 2414 * and fill it in, pretending to be pf(4). 2415 */ 2416 match = 1; 2417 at = pf_find_mtag(m); 2418 if (at != NULL && at->qid != 0) 2419 break; 2420 mtag = m_tag_get(PACKET_TAG_PF, 2421 sizeof(struct pf_mtag), M_NOWAIT | M_ZERO); 2422 if (mtag == NULL) { 2423 /* 2424 * Let the packet fall back to the 2425 * default ALTQ. 2426 */ 2427 break; 2428 } 2429 m_tag_prepend(m, mtag); 2430 at = (struct pf_mtag *)(mtag + 1); 2431 at->qid = altq->qid; 2432 at->hdr = ip; 2433 break; 2434 } 2435 2436 case O_LOG: 2437 ipfw_log(chain, f, hlen, args, 2438 offset | ip6f_mf, tablearg, ip); 2439 match = 1; 2440 break; 2441 2442 case O_PROB: 2443 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2444 break; 2445 2446 case O_VERREVPATH: 2447 /* Outgoing packets automatically pass/match */ 2448 match = (args->flags & IPFW_ARGS_OUT || 2449 ( 2450 #ifdef INET6 2451 is_ipv6 ? 2452 verify_path6(&(args->f_id.src_ip6), 2453 iif, args->f_id.fib) : 2454 #endif 2455 verify_path(src_ip, iif, args->f_id.fib))); 2456 break; 2457 2458 case O_VERSRCREACH: 2459 /* Outgoing packets automatically pass/match */ 2460 match = (hlen > 0 && ((oif != NULL) || ( 2461 #ifdef INET6 2462 is_ipv6 ? 2463 verify_path6(&(args->f_id.src_ip6), 2464 NULL, args->f_id.fib) : 2465 #endif 2466 verify_path(src_ip, NULL, args->f_id.fib)))); 2467 break; 2468 2469 case O_ANTISPOOF: 2470 /* Outgoing packets automatically pass/match */ 2471 if (oif == NULL && hlen > 0 && 2472 ( (is_ipv4 && in_localaddr(src_ip)) 2473 #ifdef INET6 2474 || (is_ipv6 && 2475 in6_localaddr(&(args->f_id.src_ip6))) 2476 #endif 2477 )) 2478 match = 2479 #ifdef INET6 2480 is_ipv6 ? verify_path6( 2481 &(args->f_id.src_ip6), iif, 2482 args->f_id.fib) : 2483 #endif 2484 verify_path(src_ip, iif, 2485 args->f_id.fib); 2486 else 2487 match = 1; 2488 break; 2489 2490 case O_IPSEC: 2491 match = (m_tag_find(m, 2492 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2493 /* otherwise no match */ 2494 break; 2495 2496 #ifdef INET6 2497 case O_IP6_SRC: 2498 match = is_ipv6 && 2499 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2500 &((ipfw_insn_ip6 *)cmd)->addr6); 2501 break; 2502 2503 case O_IP6_DST: 2504 match = is_ipv6 && 2505 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2506 &((ipfw_insn_ip6 *)cmd)->addr6); 2507 break; 2508 case O_IP6_SRC_MASK: 2509 case O_IP6_DST_MASK: 2510 if (is_ipv6) { 2511 int i = cmdlen - 1; 2512 struct in6_addr p; 2513 struct in6_addr *d = 2514 &((ipfw_insn_ip6 *)cmd)->addr6; 2515 2516 for (; !match && i > 0; d += 2, 2517 i -= F_INSN_SIZE(struct in6_addr) 2518 * 2) { 2519 p = (cmd->opcode == 2520 O_IP6_SRC_MASK) ? 2521 args->f_id.src_ip6: 2522 args->f_id.dst_ip6; 2523 APPLY_MASK(&p, &d[1]); 2524 match = 2525 IN6_ARE_ADDR_EQUAL(&d[0], 2526 &p); 2527 } 2528 } 2529 break; 2530 2531 case O_FLOW6ID: 2532 match = is_ipv6 && 2533 flow6id_match(args->f_id.flow_id6, 2534 (ipfw_insn_u32 *) cmd); 2535 break; 2536 2537 case O_EXT_HDR: 2538 match = is_ipv6 && 2539 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2540 break; 2541 2542 case O_IP6: 2543 match = is_ipv6; 2544 break; 2545 #endif 2546 2547 case O_IP4: 2548 match = is_ipv4; 2549 break; 2550 2551 case O_TAG: { 2552 struct m_tag *mtag; 2553 uint32_t tag = TARG(cmd->arg1, tag); 2554 2555 /* Packet is already tagged with this tag? */ 2556 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 2557 2558 /* We have `untag' action when F_NOT flag is 2559 * present. And we must remove this mtag from 2560 * mbuf and reset `match' to zero (`match' will 2561 * be inversed later). 2562 * Otherwise we should allocate new mtag and 2563 * push it into mbuf. 2564 */ 2565 if (cmd->len & F_NOT) { /* `untag' action */ 2566 if (mtag != NULL) 2567 m_tag_delete(m, mtag); 2568 match = 0; 2569 } else { 2570 if (mtag == NULL) { 2571 mtag = m_tag_alloc( MTAG_IPFW, 2572 tag, 0, M_NOWAIT); 2573 if (mtag != NULL) 2574 m_tag_prepend(m, mtag); 2575 } 2576 match = 1; 2577 } 2578 break; 2579 } 2580 2581 case O_FIB: /* try match the specified fib */ 2582 if (args->f_id.fib == cmd->arg1) 2583 match = 1; 2584 break; 2585 2586 case O_SOCKARG: { 2587 #ifndef USERSPACE /* not supported in userspace */ 2588 struct inpcb *inp = args->inp; 2589 struct inpcbinfo *pi; 2590 2591 if (is_ipv6) /* XXX can we remove this ? */ 2592 break; 2593 2594 if (proto == IPPROTO_TCP) 2595 pi = &V_tcbinfo; 2596 else if (proto == IPPROTO_UDP) 2597 pi = &V_udbinfo; 2598 else if (proto == IPPROTO_UDPLITE) 2599 pi = &V_ulitecbinfo; 2600 else 2601 break; 2602 2603 /* 2604 * XXXRW: so_user_cookie should almost 2605 * certainly be inp_user_cookie? 2606 */ 2607 2608 /* For incoming packet, lookup up the 2609 inpcb using the src/dest ip/port tuple */ 2610 if (inp == NULL) { 2611 inp = in_pcblookup(pi, 2612 src_ip, htons(src_port), 2613 dst_ip, htons(dst_port), 2614 INPLOOKUP_RLOCKPCB, NULL); 2615 if (inp != NULL) { 2616 tablearg = 2617 inp->inp_socket->so_user_cookie; 2618 if (tablearg) 2619 match = 1; 2620 INP_RUNLOCK(inp); 2621 } 2622 } else { 2623 if (inp->inp_socket) { 2624 tablearg = 2625 inp->inp_socket->so_user_cookie; 2626 if (tablearg) 2627 match = 1; 2628 } 2629 } 2630 #endif /* !USERSPACE */ 2631 break; 2632 } 2633 2634 case O_TAGGED: { 2635 struct m_tag *mtag; 2636 uint32_t tag = TARG(cmd->arg1, tag); 2637 2638 if (cmdlen == 1) { 2639 match = m_tag_locate(m, MTAG_IPFW, 2640 tag, NULL) != NULL; 2641 break; 2642 } 2643 2644 /* we have ranges */ 2645 for (mtag = m_tag_first(m); 2646 mtag != NULL && !match; 2647 mtag = m_tag_next(m, mtag)) { 2648 uint16_t *p; 2649 int i; 2650 2651 if (mtag->m_tag_cookie != MTAG_IPFW) 2652 continue; 2653 2654 p = ((ipfw_insn_u16 *)cmd)->ports; 2655 i = cmdlen - 1; 2656 for(; !match && i > 0; i--, p += 2) 2657 match = 2658 mtag->m_tag_id >= p[0] && 2659 mtag->m_tag_id <= p[1]; 2660 } 2661 break; 2662 } 2663 2664 /* 2665 * The second set of opcodes represents 'actions', 2666 * i.e. the terminal part of a rule once the packet 2667 * matches all previous patterns. 2668 * Typically there is only one action for each rule, 2669 * and the opcode is stored at the end of the rule 2670 * (but there are exceptions -- see below). 2671 * 2672 * In general, here we set retval and terminate the 2673 * outer loop (would be a 'break 3' in some language, 2674 * but we need to set l=0, done=1) 2675 * 2676 * Exceptions: 2677 * O_COUNT and O_SKIPTO actions: 2678 * instead of terminating, we jump to the next rule 2679 * (setting l=0), or to the SKIPTO target (setting 2680 * f/f_len, cmd and l as needed), respectively. 2681 * 2682 * O_TAG, O_LOG and O_ALTQ action parameters: 2683 * perform some action and set match = 1; 2684 * 2685 * O_LIMIT and O_KEEP_STATE: these opcodes are 2686 * not real 'actions', and are stored right 2687 * before the 'action' part of the rule (one 2688 * exception is O_SKIP_ACTION which could be 2689 * between these opcodes and 'action' one). 2690 * These opcodes try to install an entry in the 2691 * state tables; if successful, we continue with 2692 * the next opcode (match=1; break;), otherwise 2693 * the packet must be dropped (set retval, 2694 * break loops with l=0, done=1) 2695 * 2696 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2697 * cause a lookup of the state table, and a jump 2698 * to the 'action' part of the parent rule 2699 * if an entry is found, or 2700 * (CHECK_STATE only) a jump to the next rule if 2701 * the entry is not found. 2702 * The result of the lookup is cached so that 2703 * further instances of these opcodes become NOPs. 2704 * The jump to the next rule is done by setting 2705 * l=0, cmdlen=0. 2706 * 2707 * O_SKIP_ACTION: this opcode is not a real 'action' 2708 * either, and is stored right before the 'action' 2709 * part of the rule, right after the O_KEEP_STATE 2710 * opcode. It causes match failure so the real 2711 * 'action' could be executed only if the rule 2712 * is checked via dynamic rule from the state 2713 * table, as in such case execution starts 2714 * from the true 'action' opcode directly. 2715 * 2716 */ 2717 case O_LIMIT: 2718 case O_KEEP_STATE: 2719 if (ipfw_dyn_install_state(chain, f, 2720 (ipfw_insn_limit *)cmd, args, ulp, 2721 pktlen, &dyn_info, tablearg)) { 2722 /* error or limit violation */ 2723 retval = IP_FW_DENY; 2724 l = 0; /* exit inner loop */ 2725 done = 1; /* exit outer loop */ 2726 } 2727 match = 1; 2728 break; 2729 2730 case O_PROBE_STATE: 2731 case O_CHECK_STATE: 2732 /* 2733 * dynamic rules are checked at the first 2734 * keep-state or check-state occurrence, 2735 * with the result being stored in dyn_info. 2736 * The compiler introduces a PROBE_STATE 2737 * instruction for us when we have a 2738 * KEEP_STATE (because PROBE_STATE needs 2739 * to be run first). 2740 */ 2741 if (DYN_LOOKUP_NEEDED(&dyn_info, cmd) && 2742 (q = ipfw_dyn_lookup_state(args, ulp, 2743 pktlen, cmd, &dyn_info)) != NULL) { 2744 /* 2745 * Found dynamic entry, jump to the 2746 * 'action' part of the parent rule 2747 * by setting f, cmd, l and clearing 2748 * cmdlen. 2749 */ 2750 f = q; 2751 f_pos = dyn_info.f_pos; 2752 cmd = ACTION_PTR(f); 2753 l = f->cmd_len - f->act_ofs; 2754 cmdlen = 0; 2755 match = 1; 2756 break; 2757 } 2758 /* 2759 * Dynamic entry not found. If CHECK_STATE, 2760 * skip to next rule, if PROBE_STATE just 2761 * ignore and continue with next opcode. 2762 */ 2763 if (cmd->opcode == O_CHECK_STATE) 2764 l = 0; /* exit inner loop */ 2765 match = 1; 2766 break; 2767 2768 case O_SKIP_ACTION: 2769 match = 0; /* skip to the next rule */ 2770 l = 0; /* exit inner loop */ 2771 break; 2772 2773 case O_ACCEPT: 2774 retval = 0; /* accept */ 2775 l = 0; /* exit inner loop */ 2776 done = 1; /* exit outer loop */ 2777 break; 2778 2779 case O_PIPE: 2780 case O_QUEUE: 2781 set_match(args, f_pos, chain); 2782 args->rule.info = TARG(cmd->arg1, pipe); 2783 if (cmd->opcode == O_PIPE) 2784 args->rule.info |= IPFW_IS_PIPE; 2785 if (V_fw_one_pass) 2786 args->rule.info |= IPFW_ONEPASS; 2787 retval = IP_FW_DUMMYNET; 2788 l = 0; /* exit inner loop */ 2789 done = 1; /* exit outer loop */ 2790 break; 2791 2792 case O_DIVERT: 2793 case O_TEE: 2794 if (args->flags & IPFW_ARGS_ETHER) 2795 break; /* not on layer 2 */ 2796 /* otherwise this is terminal */ 2797 l = 0; /* exit inner loop */ 2798 done = 1; /* exit outer loop */ 2799 retval = (cmd->opcode == O_DIVERT) ? 2800 IP_FW_DIVERT : IP_FW_TEE; 2801 set_match(args, f_pos, chain); 2802 args->rule.info = TARG(cmd->arg1, divert); 2803 break; 2804 2805 case O_COUNT: 2806 IPFW_INC_RULE_COUNTER(f, pktlen); 2807 l = 0; /* exit inner loop */ 2808 break; 2809 2810 case O_SKIPTO: 2811 IPFW_INC_RULE_COUNTER(f, pktlen); 2812 f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0); 2813 /* 2814 * Skip disabled rules, and re-enter 2815 * the inner loop with the correct 2816 * f_pos, f, l and cmd. 2817 * Also clear cmdlen and skip_or 2818 */ 2819 for (; f_pos < chain->n_rules - 1 && 2820 (V_set_disable & 2821 (1 << chain->map[f_pos]->set)); 2822 f_pos++) 2823 ; 2824 /* Re-enter the inner loop at the skipto rule. */ 2825 f = chain->map[f_pos]; 2826 l = f->cmd_len; 2827 cmd = f->cmd; 2828 match = 1; 2829 cmdlen = 0; 2830 skip_or = 0; 2831 continue; 2832 break; /* not reached */ 2833 2834 case O_CALLRETURN: { 2835 /* 2836 * Implementation of `subroutine' call/return, 2837 * in the stack carried in an mbuf tag. This 2838 * is different from `skipto' in that any call 2839 * address is possible (`skipto' must prevent 2840 * backward jumps to avoid endless loops). 2841 * We have `return' action when F_NOT flag is 2842 * present. The `m_tag_id' field is used as 2843 * stack pointer. 2844 */ 2845 struct m_tag *mtag; 2846 uint16_t jmpto, *stack; 2847 2848 #define IS_CALL ((cmd->len & F_NOT) == 0) 2849 #define IS_RETURN ((cmd->len & F_NOT) != 0) 2850 /* 2851 * Hand-rolled version of m_tag_locate() with 2852 * wildcard `type'. 2853 * If not already tagged, allocate new tag. 2854 */ 2855 mtag = m_tag_first(m); 2856 while (mtag != NULL) { 2857 if (mtag->m_tag_cookie == 2858 MTAG_IPFW_CALL) 2859 break; 2860 mtag = m_tag_next(m, mtag); 2861 } 2862 if (mtag == NULL && IS_CALL) { 2863 mtag = m_tag_alloc(MTAG_IPFW_CALL, 0, 2864 IPFW_CALLSTACK_SIZE * 2865 sizeof(uint16_t), M_NOWAIT); 2866 if (mtag != NULL) 2867 m_tag_prepend(m, mtag); 2868 } 2869 2870 /* 2871 * On error both `call' and `return' just 2872 * continue with next rule. 2873 */ 2874 if (IS_RETURN && (mtag == NULL || 2875 mtag->m_tag_id == 0)) { 2876 l = 0; /* exit inner loop */ 2877 break; 2878 } 2879 if (IS_CALL && (mtag == NULL || 2880 mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) { 2881 printf("ipfw: call stack error, " 2882 "go to next rule\n"); 2883 l = 0; /* exit inner loop */ 2884 break; 2885 } 2886 2887 IPFW_INC_RULE_COUNTER(f, pktlen); 2888 stack = (uint16_t *)(mtag + 1); 2889 2890 /* 2891 * The `call' action may use cached f_pos 2892 * (in f->next_rule), whose version is written 2893 * in f->next_rule. 2894 * The `return' action, however, doesn't have 2895 * fixed jump address in cmd->arg1 and can't use 2896 * cache. 2897 */ 2898 if (IS_CALL) { 2899 stack[mtag->m_tag_id] = f->rulenum; 2900 mtag->m_tag_id++; 2901 f_pos = JUMP(chain, f, cmd->arg1, 2902 tablearg, 1); 2903 } else { /* `return' action */ 2904 mtag->m_tag_id--; 2905 jmpto = stack[mtag->m_tag_id] + 1; 2906 f_pos = ipfw_find_rule(chain, jmpto, 0); 2907 } 2908 2909 /* 2910 * Skip disabled rules, and re-enter 2911 * the inner loop with the correct 2912 * f_pos, f, l and cmd. 2913 * Also clear cmdlen and skip_or 2914 */ 2915 for (; f_pos < chain->n_rules - 1 && 2916 (V_set_disable & 2917 (1 << chain->map[f_pos]->set)); f_pos++) 2918 ; 2919 /* Re-enter the inner loop at the dest rule. */ 2920 f = chain->map[f_pos]; 2921 l = f->cmd_len; 2922 cmd = f->cmd; 2923 cmdlen = 0; 2924 skip_or = 0; 2925 continue; 2926 break; /* NOTREACHED */ 2927 } 2928 #undef IS_CALL 2929 #undef IS_RETURN 2930 2931 case O_REJECT: 2932 /* 2933 * Drop the packet and send a reject notice 2934 * if the packet is not ICMP (or is an ICMP 2935 * query), and it is not multicast/broadcast. 2936 */ 2937 if (hlen > 0 && is_ipv4 && offset == 0 && 2938 (proto != IPPROTO_ICMP || 2939 is_icmp_query(ICMP(ulp))) && 2940 !(m->m_flags & (M_BCAST|M_MCAST)) && 2941 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 2942 send_reject(args, cmd->arg1, iplen, ip); 2943 m = args->m; 2944 } 2945 /* FALLTHROUGH */ 2946 #ifdef INET6 2947 case O_UNREACH6: 2948 if (hlen > 0 && is_ipv6 && 2949 ((offset & IP6F_OFF_MASK) == 0) && 2950 (proto != IPPROTO_ICMPV6 || 2951 (is_icmp6_query(icmp6_type) == 1)) && 2952 !(m->m_flags & (M_BCAST|M_MCAST)) && 2953 !IN6_IS_ADDR_MULTICAST( 2954 &args->f_id.dst_ip6)) { 2955 send_reject6(args, 2956 cmd->opcode == O_REJECT ? 2957 map_icmp_unreach(cmd->arg1): 2958 cmd->arg1, hlen, 2959 (struct ip6_hdr *)ip); 2960 m = args->m; 2961 } 2962 /* FALLTHROUGH */ 2963 #endif 2964 case O_DENY: 2965 retval = IP_FW_DENY; 2966 l = 0; /* exit inner loop */ 2967 done = 1; /* exit outer loop */ 2968 break; 2969 2970 case O_FORWARD_IP: 2971 if (args->flags & IPFW_ARGS_ETHER) 2972 break; /* not valid on layer2 pkts */ 2973 if (q != f || 2974 dyn_info.direction == MATCH_FORWARD) { 2975 struct sockaddr_in *sa; 2976 2977 sa = &(((ipfw_insn_sa *)cmd)->sa); 2978 if (sa->sin_addr.s_addr == INADDR_ANY) { 2979 #ifdef INET6 2980 /* 2981 * We use O_FORWARD_IP opcode for 2982 * fwd rule with tablearg, but tables 2983 * now support IPv6 addresses. And 2984 * when we are inspecting IPv6 packet, 2985 * we can use nh6 field from 2986 * table_value as next_hop6 address. 2987 */ 2988 if (is_ipv6) { 2989 struct ip_fw_nh6 *nh6; 2990 2991 args->flags |= IPFW_ARGS_NH6; 2992 nh6 = &args->hopstore6; 2993 nh6->sin6_addr = TARG_VAL( 2994 chain, tablearg, nh6); 2995 nh6->sin6_port = sa->sin_port; 2996 nh6->sin6_scope_id = TARG_VAL( 2997 chain, tablearg, zoneid); 2998 } else 2999 #endif 3000 { 3001 args->flags |= IPFW_ARGS_NH4; 3002 args->hopstore.sin_port = 3003 sa->sin_port; 3004 sa = &args->hopstore; 3005 sa->sin_family = AF_INET; 3006 sa->sin_len = sizeof(*sa); 3007 sa->sin_addr.s_addr = htonl( 3008 TARG_VAL(chain, tablearg, 3009 nh4)); 3010 } 3011 } else { 3012 args->flags |= IPFW_ARGS_NH4PTR; 3013 args->next_hop = sa; 3014 } 3015 } 3016 retval = IP_FW_PASS; 3017 l = 0; /* exit inner loop */ 3018 done = 1; /* exit outer loop */ 3019 break; 3020 3021 #ifdef INET6 3022 case O_FORWARD_IP6: 3023 if (args->flags & IPFW_ARGS_ETHER) 3024 break; /* not valid on layer2 pkts */ 3025 if (q != f || 3026 dyn_info.direction == MATCH_FORWARD) { 3027 struct sockaddr_in6 *sin6; 3028 3029 sin6 = &(((ipfw_insn_sa6 *)cmd)->sa); 3030 args->flags |= IPFW_ARGS_NH6PTR; 3031 args->next_hop6 = sin6; 3032 } 3033 retval = IP_FW_PASS; 3034 l = 0; /* exit inner loop */ 3035 done = 1; /* exit outer loop */ 3036 break; 3037 #endif 3038 3039 case O_NETGRAPH: 3040 case O_NGTEE: 3041 set_match(args, f_pos, chain); 3042 args->rule.info = TARG(cmd->arg1, netgraph); 3043 if (V_fw_one_pass) 3044 args->rule.info |= IPFW_ONEPASS; 3045 retval = (cmd->opcode == O_NETGRAPH) ? 3046 IP_FW_NETGRAPH : IP_FW_NGTEE; 3047 l = 0; /* exit inner loop */ 3048 done = 1; /* exit outer loop */ 3049 break; 3050 3051 case O_SETFIB: { 3052 uint32_t fib; 3053 3054 IPFW_INC_RULE_COUNTER(f, pktlen); 3055 fib = TARG(cmd->arg1, fib) & 0x7FFF; 3056 if (fib >= rt_numfibs) 3057 fib = 0; 3058 M_SETFIB(m, fib); 3059 args->f_id.fib = fib; /* XXX */ 3060 l = 0; /* exit inner loop */ 3061 break; 3062 } 3063 3064 case O_SETDSCP: { 3065 uint16_t code; 3066 3067 code = TARG(cmd->arg1, dscp) & 0x3F; 3068 l = 0; /* exit inner loop */ 3069 if (is_ipv4) { 3070 uint16_t old; 3071 3072 old = *(uint16_t *)ip; 3073 ip->ip_tos = (code << 2) | 3074 (ip->ip_tos & 0x03); 3075 ip->ip_sum = cksum_adjust(ip->ip_sum, 3076 old, *(uint16_t *)ip); 3077 } else if (is_ipv6) { 3078 uint8_t *v; 3079 3080 v = &((struct ip6_hdr *)ip)->ip6_vfc; 3081 *v = (*v & 0xF0) | (code >> 2); 3082 v++; 3083 *v = (*v & 0x3F) | ((code & 0x03) << 6); 3084 } else 3085 break; 3086 3087 IPFW_INC_RULE_COUNTER(f, pktlen); 3088 break; 3089 } 3090 3091 case O_NAT: 3092 l = 0; /* exit inner loop */ 3093 done = 1; /* exit outer loop */ 3094 /* 3095 * Ensure that we do not invoke NAT handler for 3096 * non IPv4 packets. Libalias expects only IPv4. 3097 */ 3098 if (!is_ipv4 || !IPFW_NAT_LOADED) { 3099 retval = IP_FW_DENY; 3100 break; 3101 } 3102 3103 struct cfg_nat *t; 3104 int nat_id; 3105 3106 args->rule.info = 0; 3107 set_match(args, f_pos, chain); 3108 /* Check if this is 'global' nat rule */ 3109 if (cmd->arg1 == IP_FW_NAT44_GLOBAL) { 3110 retval = ipfw_nat_ptr(args, NULL, m); 3111 break; 3112 } 3113 t = ((ipfw_insn_nat *)cmd)->nat; 3114 if (t == NULL) { 3115 nat_id = TARG(cmd->arg1, nat); 3116 t = (*lookup_nat_ptr)(&chain->nat, nat_id); 3117 3118 if (t == NULL) { 3119 retval = IP_FW_DENY; 3120 break; 3121 } 3122 if (cmd->arg1 != IP_FW_TARG) 3123 ((ipfw_insn_nat *)cmd)->nat = t; 3124 } 3125 retval = ipfw_nat_ptr(args, t, m); 3126 break; 3127 3128 case O_REASS: { 3129 int ip_off; 3130 3131 l = 0; /* in any case exit inner loop */ 3132 if (is_ipv6) /* IPv6 is not supported yet */ 3133 break; 3134 IPFW_INC_RULE_COUNTER(f, pktlen); 3135 ip_off = ntohs(ip->ip_off); 3136 3137 /* if not fragmented, go to next rule */ 3138 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0) 3139 break; 3140 3141 args->m = m = ip_reass(m); 3142 3143 /* 3144 * do IP header checksum fixup. 3145 */ 3146 if (m == NULL) { /* fragment got swallowed */ 3147 retval = IP_FW_DENY; 3148 } else { /* good, packet complete */ 3149 int hlen; 3150 3151 ip = mtod(m, struct ip *); 3152 hlen = ip->ip_hl << 2; 3153 ip->ip_sum = 0; 3154 if (hlen == sizeof(struct ip)) 3155 ip->ip_sum = in_cksum_hdr(ip); 3156 else 3157 ip->ip_sum = in_cksum(m, hlen); 3158 retval = IP_FW_REASS; 3159 args->rule.info = 0; 3160 set_match(args, f_pos, chain); 3161 } 3162 done = 1; /* exit outer loop */ 3163 break; 3164 } 3165 case O_EXTERNAL_ACTION: 3166 l = 0; /* in any case exit inner loop */ 3167 retval = ipfw_run_eaction(chain, args, 3168 cmd, &done); 3169 /* 3170 * If both @retval and @done are zero, 3171 * consider this as rule matching and 3172 * update counters. 3173 */ 3174 if (retval == 0 && done == 0) { 3175 IPFW_INC_RULE_COUNTER(f, pktlen); 3176 /* 3177 * Reset the result of the last 3178 * dynamic state lookup. 3179 * External action can change 3180 * @args content, and it may be 3181 * used for new state lookup later. 3182 */ 3183 DYN_INFO_INIT(&dyn_info); 3184 } 3185 break; 3186 3187 default: 3188 panic("-- unknown opcode %d\n", cmd->opcode); 3189 } /* end of switch() on opcodes */ 3190 /* 3191 * if we get here with l=0, then match is irrelevant. 3192 */ 3193 3194 if (cmd->len & F_NOT) 3195 match = !match; 3196 3197 if (match) { 3198 if (cmd->len & F_OR) 3199 skip_or = 1; 3200 } else { 3201 if (!(cmd->len & F_OR)) /* not an OR block, */ 3202 break; /* try next rule */ 3203 } 3204 3205 } /* end of inner loop, scan opcodes */ 3206 #undef PULLUP_LEN 3207 #undef PULLUP_LEN_LOCKED 3208 3209 if (done) 3210 break; 3211 3212 /* next_rule:; */ /* try next rule */ 3213 3214 } /* end of outer for, scan rules */ 3215 3216 if (done) { 3217 struct ip_fw *rule = chain->map[f_pos]; 3218 /* Update statistics */ 3219 IPFW_INC_RULE_COUNTER(rule, pktlen); 3220 } else { 3221 retval = IP_FW_DENY; 3222 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3223 } 3224 IPFW_PF_RUNLOCK(chain); 3225 #ifdef __FreeBSD__ 3226 if (ucred_cache != NULL) 3227 crfree(ucred_cache); 3228 #endif 3229 return (retval); 3230 3231 pullup_failed: 3232 if (V_fw_verbose) 3233 printf("ipfw: pullup failed\n"); 3234 return (IP_FW_DENY); 3235 } 3236 3237 /* 3238 * Set maximum number of tables that can be used in given VNET ipfw instance. 3239 */ 3240 #ifdef SYSCTL_NODE 3241 static int 3242 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS) 3243 { 3244 int error; 3245 unsigned int ntables; 3246 3247 ntables = V_fw_tables_max; 3248 3249 error = sysctl_handle_int(oidp, &ntables, 0, req); 3250 /* Read operation or some error */ 3251 if ((error != 0) || (req->newptr == NULL)) 3252 return (error); 3253 3254 return (ipfw_resize_tables(&V_layer3_chain, ntables)); 3255 } 3256 3257 /* 3258 * Switches table namespace between global and per-set. 3259 */ 3260 static int 3261 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS) 3262 { 3263 int error; 3264 unsigned int sets; 3265 3266 sets = V_fw_tables_sets; 3267 3268 error = sysctl_handle_int(oidp, &sets, 0, req); 3269 /* Read operation or some error */ 3270 if ((error != 0) || (req->newptr == NULL)) 3271 return (error); 3272 3273 return (ipfw_switch_tables_namespace(&V_layer3_chain, sets)); 3274 } 3275 #endif 3276 3277 /* 3278 * Module and VNET glue 3279 */ 3280 3281 /* 3282 * Stuff that must be initialised only on boot or module load 3283 */ 3284 static int 3285 ipfw_init(void) 3286 { 3287 int error = 0; 3288 3289 /* 3290 * Only print out this stuff the first time around, 3291 * when called from the sysinit code. 3292 */ 3293 printf("ipfw2 " 3294 #ifdef INET6 3295 "(+ipv6) " 3296 #endif 3297 "initialized, divert %s, nat %s, " 3298 "default to %s, logging ", 3299 #ifdef IPDIVERT 3300 "enabled", 3301 #else 3302 "loadable", 3303 #endif 3304 #ifdef IPFIREWALL_NAT 3305 "enabled", 3306 #else 3307 "loadable", 3308 #endif 3309 default_to_accept ? "accept" : "deny"); 3310 3311 /* 3312 * Note: V_xxx variables can be accessed here but the vnet specific 3313 * initializer may not have been called yet for the VIMAGE case. 3314 * Tuneables will have been processed. We will print out values for 3315 * the default vnet. 3316 * XXX This should all be rationalized AFTER 8.0 3317 */ 3318 if (V_fw_verbose == 0) 3319 printf("disabled\n"); 3320 else if (V_verbose_limit == 0) 3321 printf("unlimited\n"); 3322 else 3323 printf("limited to %d packets/entry by default\n", 3324 V_verbose_limit); 3325 3326 /* Check user-supplied table count for validness */ 3327 if (default_fw_tables > IPFW_TABLES_MAX) 3328 default_fw_tables = IPFW_TABLES_MAX; 3329 3330 ipfw_init_sopt_handler(); 3331 ipfw_init_obj_rewriter(); 3332 ipfw_iface_init(); 3333 return (error); 3334 } 3335 3336 /* 3337 * Called for the removal of the last instance only on module unload. 3338 */ 3339 static void 3340 ipfw_destroy(void) 3341 { 3342 3343 ipfw_iface_destroy(); 3344 ipfw_destroy_sopt_handler(); 3345 ipfw_destroy_obj_rewriter(); 3346 printf("IP firewall unloaded\n"); 3347 } 3348 3349 /* 3350 * Stuff that must be initialized for every instance 3351 * (including the first of course). 3352 */ 3353 static int 3354 vnet_ipfw_init(const void *unused) 3355 { 3356 int error, first; 3357 struct ip_fw *rule = NULL; 3358 struct ip_fw_chain *chain; 3359 3360 chain = &V_layer3_chain; 3361 3362 first = IS_DEFAULT_VNET(curvnet) ? 1 : 0; 3363 3364 /* First set up some values that are compile time options */ 3365 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 3366 V_fw_deny_unknown_exthdrs = 1; 3367 #ifdef IPFIREWALL_VERBOSE 3368 V_fw_verbose = 1; 3369 #endif 3370 #ifdef IPFIREWALL_VERBOSE_LIMIT 3371 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 3372 #endif 3373 #ifdef IPFIREWALL_NAT 3374 LIST_INIT(&chain->nat); 3375 #endif 3376 3377 /* Init shared services hash table */ 3378 ipfw_init_srv(chain); 3379 3380 ipfw_init_counters(); 3381 /* Set initial number of tables */ 3382 V_fw_tables_max = default_fw_tables; 3383 error = ipfw_init_tables(chain, first); 3384 if (error) { 3385 printf("ipfw2: setting up tables failed\n"); 3386 free(chain->map, M_IPFW); 3387 free(rule, M_IPFW); 3388 return (ENOSPC); 3389 } 3390 3391 IPFW_LOCK_INIT(chain); 3392 3393 /* fill and insert the default rule */ 3394 rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw)); 3395 rule->flags |= IPFW_RULE_NOOPT; 3396 rule->cmd_len = 1; 3397 rule->cmd[0].len = 1; 3398 rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; 3399 chain->default_rule = rule; 3400 ipfw_add_protected_rule(chain, rule, 0); 3401 3402 ipfw_dyn_init(chain); 3403 ipfw_eaction_init(chain, first); 3404 #ifdef LINEAR_SKIPTO 3405 ipfw_init_skipto_cache(chain); 3406 #endif 3407 ipfw_bpf_init(first); 3408 3409 /* First set up some values that are compile time options */ 3410 V_ipfw_vnet_ready = 1; /* Open for business */ 3411 3412 /* 3413 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6. 3414 * Even if the latter two fail we still keep the module alive 3415 * because the sockopt and layer2 paths are still useful. 3416 * ipfw[6]_hook return 0 on success, ENOENT on failure, 3417 * so we can ignore the exact return value and just set a flag. 3418 * 3419 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so 3420 * changes in the underlying (per-vnet) variables trigger 3421 * immediate hook()/unhook() calls. 3422 * In layer2 we have the same behaviour, except that V_ether_ipfw 3423 * is checked on each packet because there are no pfil hooks. 3424 */ 3425 V_ip_fw_ctl_ptr = ipfw_ctl3; 3426 error = ipfw_attach_hooks(); 3427 return (error); 3428 } 3429 3430 /* 3431 * Called for the removal of each instance. 3432 */ 3433 static int 3434 vnet_ipfw_uninit(const void *unused) 3435 { 3436 struct ip_fw *reap; 3437 struct ip_fw_chain *chain = &V_layer3_chain; 3438 int i, last; 3439 3440 V_ipfw_vnet_ready = 0; /* tell new callers to go away */ 3441 /* 3442 * disconnect from ipv4, ipv6, layer2 and sockopt. 3443 * Then grab, release and grab again the WLOCK so we make 3444 * sure the update is propagated and nobody will be in. 3445 */ 3446 ipfw_detach_hooks(); 3447 V_ip_fw_ctl_ptr = NULL; 3448 3449 last = IS_DEFAULT_VNET(curvnet) ? 1 : 0; 3450 3451 IPFW_UH_WLOCK(chain); 3452 IPFW_UH_WUNLOCK(chain); 3453 3454 ipfw_dyn_uninit(0); /* run the callout_drain */ 3455 3456 IPFW_UH_WLOCK(chain); 3457 3458 reap = NULL; 3459 IPFW_WLOCK(chain); 3460 for (i = 0; i < chain->n_rules; i++) 3461 ipfw_reap_add(chain, &reap, chain->map[i]); 3462 free(chain->map, M_IPFW); 3463 #ifdef LINEAR_SKIPTO 3464 ipfw_destroy_skipto_cache(chain); 3465 #endif 3466 IPFW_WUNLOCK(chain); 3467 IPFW_UH_WUNLOCK(chain); 3468 ipfw_destroy_tables(chain, last); 3469 ipfw_eaction_uninit(chain, last); 3470 if (reap != NULL) 3471 ipfw_reap_rules(reap); 3472 vnet_ipfw_iface_destroy(chain); 3473 ipfw_destroy_srv(chain); 3474 IPFW_LOCK_DESTROY(chain); 3475 ipfw_dyn_uninit(1); /* free the remaining parts */ 3476 ipfw_destroy_counters(); 3477 ipfw_bpf_uninit(last); 3478 return (0); 3479 } 3480 3481 /* 3482 * Module event handler. 3483 * In general we have the choice of handling most of these events by the 3484 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to 3485 * use the SYSINIT handlers as they are more capable of expressing the 3486 * flow of control during module and vnet operations, so this is just 3487 * a skeleton. Note there is no SYSINIT equivalent of the module 3488 * SHUTDOWN handler, but we don't have anything to do in that case anyhow. 3489 */ 3490 static int 3491 ipfw_modevent(module_t mod, int type, void *unused) 3492 { 3493 int err = 0; 3494 3495 switch (type) { 3496 case MOD_LOAD: 3497 /* Called once at module load or 3498 * system boot if compiled in. */ 3499 break; 3500 case MOD_QUIESCE: 3501 /* Called before unload. May veto unloading. */ 3502 break; 3503 case MOD_UNLOAD: 3504 /* Called during unload. */ 3505 break; 3506 case MOD_SHUTDOWN: 3507 /* Called during system shutdown. */ 3508 break; 3509 default: 3510 err = EOPNOTSUPP; 3511 break; 3512 } 3513 return err; 3514 } 3515 3516 static moduledata_t ipfwmod = { 3517 "ipfw", 3518 ipfw_modevent, 3519 0 3520 }; 3521 3522 /* Define startup order. */ 3523 #define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_FIREWALL 3524 #define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */ 3525 #define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */ 3526 #define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */ 3527 3528 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER); 3529 FEATURE(ipfw_ctl3, "ipfw new sockopt calls"); 3530 MODULE_VERSION(ipfw, 3); 3531 /* should declare some dependencies here */ 3532 3533 /* 3534 * Starting up. Done in order after ipfwmod() has been called. 3535 * VNET_SYSINIT is also called for each existing vnet and each new vnet. 3536 */ 3537 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 3538 ipfw_init, NULL); 3539 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 3540 vnet_ipfw_init, NULL); 3541 3542 /* 3543 * Closing up shop. These are done in REVERSE ORDER, but still 3544 * after ipfwmod() has been called. Not called on reboot. 3545 * VNET_SYSUNINIT is also called for each exiting vnet as it exits. 3546 * or when the module is unloaded. 3547 */ 3548 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 3549 ipfw_destroy, NULL); 3550 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 3551 vnet_ipfw_uninit, NULL); 3552 /* end of file */ 3553