1 /*- 2 * SPDX-License-Identifier: BSD-2-Clause 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/sdt.h> 59 #include <sys/socket.h> 60 #include <sys/socketvar.h> 61 #include <sys/sysctl.h> 62 #include <sys/syslog.h> 63 #include <sys/ucred.h> 64 #include <net/ethernet.h> /* for ETHERTYPE_IP */ 65 #include <net/if.h> 66 #include <net/if_var.h> 67 #include <net/if_private.h> 68 #include <net/route.h> 69 #include <net/route/nhop.h> 70 #include <net/pfil.h> 71 #include <net/vnet.h> 72 #include <net/if_pfsync.h> 73 74 #include <netpfil/pf/pf_mtag.h> 75 76 #include <netinet/in.h> 77 #include <netinet/in_var.h> 78 #include <netinet/in_pcb.h> 79 #include <netinet/ip.h> 80 #include <netinet/ip_var.h> 81 #include <netinet/ip_icmp.h> 82 #include <netinet/ip_fw.h> 83 #include <netinet/ip_carp.h> 84 #include <netinet/pim.h> 85 #include <netinet/tcp_var.h> 86 #include <netinet/udp.h> 87 #include <netinet/udp_var.h> 88 #include <netinet/sctp.h> 89 #include <netinet/sctp_crc32.h> 90 #include <netinet/sctp_header.h> 91 92 #include <netinet/ip6.h> 93 #include <netinet/icmp6.h> 94 #include <netinet/in_fib.h> 95 #ifdef INET6 96 #include <netinet6/in6_fib.h> 97 #include <netinet6/in6_pcb.h> 98 #include <netinet6/scope6_var.h> 99 #include <netinet6/ip6_var.h> 100 #endif 101 102 #include <net/if_gre.h> /* for struct grehdr */ 103 104 #include <netpfil/ipfw/ip_fw_private.h> 105 106 #include <machine/in_cksum.h> /* XXX for in_cksum */ 107 108 #ifdef MAC 109 #include <security/mac/mac_framework.h> 110 #endif 111 112 #define IPFW_PROBE(probe, arg0, arg1, arg2, arg3, arg4, arg5) \ 113 SDT_PROBE6(ipfw, , , probe, arg0, arg1, arg2, arg3, arg4, arg5) 114 115 SDT_PROVIDER_DEFINE(ipfw); 116 SDT_PROBE_DEFINE6(ipfw, , , rule__matched, 117 "int", /* retval */ 118 "int", /* af */ 119 "void *", /* src addr */ 120 "void *", /* dst addr */ 121 "struct ip_fw_args *", /* args */ 122 "struct ip_fw *" /* rule */); 123 124 /* 125 * static variables followed by global ones. 126 * All ipfw global variables are here. 127 */ 128 129 VNET_DEFINE_STATIC(int, fw_deny_unknown_exthdrs); 130 #define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs) 131 132 VNET_DEFINE_STATIC(int, fw_permit_single_frag6) = 1; 133 #define V_fw_permit_single_frag6 VNET(fw_permit_single_frag6) 134 135 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 136 static int default_to_accept = 1; 137 #else 138 static int default_to_accept; 139 #endif 140 141 VNET_DEFINE(int, autoinc_step); 142 VNET_DEFINE(int, fw_one_pass) = 1; 143 144 VNET_DEFINE(unsigned int, fw_tables_max); 145 VNET_DEFINE(unsigned int, fw_tables_sets) = 0; /* Don't use set-aware tables */ 146 /* Use 128 tables by default */ 147 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT; 148 149 static int jump_lookup_pos(struct ip_fw_chain *chain, struct ip_fw *f, int num, 150 int tablearg, int jump_backwards); 151 #ifndef LINEAR_SKIPTO 152 static int jump_cached(struct ip_fw_chain *chain, struct ip_fw *f, int num, 153 int tablearg, int jump_backwards); 154 #define JUMP(ch, f, num, targ, back) jump_cached(ch, f, num, targ, back) 155 #else 156 #define JUMP(ch, f, num, targ, back) jump_lookup_pos(ch, f, num, targ, back) 157 #endif 158 159 /* 160 * Each rule belongs to one of 32 different sets (0..31). 161 * The variable set_disable contains one bit per set. 162 * If the bit is set, all rules in the corresponding set 163 * are disabled. Set RESVD_SET(31) is reserved for the default rule 164 * and rules that are not deleted by the flush command, 165 * and CANNOT be disabled. 166 * Rules in set RESVD_SET can only be deleted individually. 167 */ 168 VNET_DEFINE(u_int32_t, set_disable); 169 #define V_set_disable VNET(set_disable) 170 171 VNET_DEFINE(int, fw_verbose); 172 /* counter for ipfw_log(NULL...) */ 173 VNET_DEFINE(u_int64_t, norule_counter); 174 VNET_DEFINE(int, verbose_limit); 175 176 /* layer3_chain contains the list of rules for layer 3 */ 177 VNET_DEFINE(struct ip_fw_chain, layer3_chain); 178 179 /* ipfw_vnet_ready controls when we are open for business */ 180 VNET_DEFINE(int, ipfw_vnet_ready) = 0; 181 182 VNET_DEFINE(int, ipfw_nat_ready) = 0; 183 184 ipfw_nat_t *ipfw_nat_ptr = NULL; 185 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int); 186 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; 187 ipfw_nat_cfg_t *ipfw_nat_del_ptr; 188 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; 189 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; 190 191 #ifdef SYSCTL_NODE 192 uint32_t dummy_def = IPFW_DEFAULT_RULE; 193 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS); 194 static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS); 195 196 SYSBEGIN(f3) 197 198 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 199 "Firewall"); 200 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 201 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0, 202 "Only do a single pass through ipfw when using dummynet(4)"); 203 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, 204 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0, 205 "Rule number auto-increment step"); 206 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 207 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0, 208 "Log matches to ipfw rules"); 209 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, 210 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(verbose_limit), 0, 211 "Set upper limit of matches of ipfw rules logged"); 212 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, 213 &dummy_def, 0, 214 "The default/max possible rule number."); 215 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_max, 216 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, 217 0, 0, sysctl_ipfw_table_num, "IU", 218 "Maximum number of concurrently used tables"); 219 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets, 220 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, 221 0, 0, sysctl_ipfw_tables_sets, "IU", 222 "Use per-set namespace for tables"); 223 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN, 224 &default_to_accept, 0, 225 "Make the default rule accept all packets."); 226 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables); 227 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, 228 CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0, 229 "Number of static rules"); 230 231 #ifdef INET6 232 SYSCTL_DECL(_net_inet6_ip6); 233 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 234 "Firewall"); 235 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs, 236 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, 237 &VNET_NAME(fw_deny_unknown_exthdrs), 0, 238 "Deny packets with unknown IPv6 Extension Headers"); 239 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6, 240 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, 241 &VNET_NAME(fw_permit_single_frag6), 0, 242 "Permit single packet IPv6 fragments"); 243 #endif /* INET6 */ 244 245 SYSEND 246 247 #endif /* SYSCTL_NODE */ 248 249 /* 250 * Some macros used in the various matching options. 251 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 252 * Other macros just cast void * into the appropriate type 253 */ 254 #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 255 #define TCP(p) ((struct tcphdr *)(p)) 256 #define SCTP(p) ((struct sctphdr *)(p)) 257 #define UDP(p) ((struct udphdr *)(p)) 258 #define ICMP(p) ((struct icmphdr *)(p)) 259 #define ICMP6(p) ((struct icmp6_hdr *)(p)) 260 261 static __inline int 262 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 263 { 264 int type = icmp->icmp_type; 265 266 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 267 } 268 269 #define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 270 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 271 272 static int 273 is_icmp_query(struct icmphdr *icmp) 274 { 275 int type = icmp->icmp_type; 276 277 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 278 } 279 #undef TT 280 281 /* 282 * The following checks use two arrays of 8 or 16 bits to store the 283 * bits that we want set or clear, respectively. They are in the 284 * low and high half of cmd->arg1 or cmd->d[0]. 285 * 286 * We scan options and store the bits we find set. We succeed if 287 * 288 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 289 * 290 * The code is sometimes optimized not to store additional variables. 291 */ 292 293 static int 294 flags_match(ipfw_insn *cmd, u_int8_t bits) 295 { 296 u_char want_clear; 297 bits = ~bits; 298 299 if ( ((cmd->arg1 & 0xff) & bits) != 0) 300 return 0; /* some bits we want set were clear */ 301 want_clear = (cmd->arg1 >> 8) & 0xff; 302 if ( (want_clear & bits) != want_clear) 303 return 0; /* some bits we want clear were set */ 304 return 1; 305 } 306 307 static int 308 ipopts_match(struct ip *ip, ipfw_insn *cmd) 309 { 310 int optlen, bits = 0; 311 u_char *cp = (u_char *)(ip + 1); 312 int x = (ip->ip_hl << 2) - sizeof (struct ip); 313 314 for (; x > 0; x -= optlen, cp += optlen) { 315 int opt = cp[IPOPT_OPTVAL]; 316 317 if (opt == IPOPT_EOL) 318 break; 319 if (opt == IPOPT_NOP) 320 optlen = 1; 321 else { 322 optlen = cp[IPOPT_OLEN]; 323 if (optlen <= 0 || optlen > x) 324 return 0; /* invalid or truncated */ 325 } 326 switch (opt) { 327 default: 328 break; 329 330 case IPOPT_LSRR: 331 bits |= IP_FW_IPOPT_LSRR; 332 break; 333 334 case IPOPT_SSRR: 335 bits |= IP_FW_IPOPT_SSRR; 336 break; 337 338 case IPOPT_RR: 339 bits |= IP_FW_IPOPT_RR; 340 break; 341 342 case IPOPT_TS: 343 bits |= IP_FW_IPOPT_TS; 344 break; 345 } 346 } 347 return (flags_match(cmd, bits)); 348 } 349 350 /* 351 * Parse TCP options. The logic copied from tcp_dooptions(). 352 */ 353 static int 354 tcpopts_parse(const struct tcphdr *tcp, uint16_t *mss) 355 { 356 const u_char *cp = (const u_char *)(tcp + 1); 357 int optlen, bits = 0; 358 int cnt = (tcp->th_off << 2) - sizeof(struct tcphdr); 359 360 for (; cnt > 0; cnt -= optlen, cp += optlen) { 361 int opt = cp[0]; 362 if (opt == TCPOPT_EOL) 363 break; 364 if (opt == TCPOPT_NOP) 365 optlen = 1; 366 else { 367 if (cnt < 2) 368 break; 369 optlen = cp[1]; 370 if (optlen < 2 || optlen > cnt) 371 break; 372 } 373 374 switch (opt) { 375 default: 376 break; 377 378 case TCPOPT_MAXSEG: 379 if (optlen != TCPOLEN_MAXSEG) 380 break; 381 bits |= IP_FW_TCPOPT_MSS; 382 if (mss != NULL) 383 *mss = be16dec(cp + 2); 384 break; 385 386 case TCPOPT_WINDOW: 387 if (optlen == TCPOLEN_WINDOW) 388 bits |= IP_FW_TCPOPT_WINDOW; 389 break; 390 391 case TCPOPT_SACK_PERMITTED: 392 if (optlen == TCPOLEN_SACK_PERMITTED) 393 bits |= IP_FW_TCPOPT_SACK; 394 break; 395 396 case TCPOPT_SACK: 397 if (optlen > 2 && (optlen - 2) % TCPOLEN_SACK == 0) 398 bits |= IP_FW_TCPOPT_SACK; 399 break; 400 401 case TCPOPT_TIMESTAMP: 402 if (optlen == TCPOLEN_TIMESTAMP) 403 bits |= IP_FW_TCPOPT_TS; 404 break; 405 } 406 } 407 return (bits); 408 } 409 410 static int 411 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 412 { 413 414 return (flags_match(cmd, tcpopts_parse(tcp, NULL))); 415 } 416 417 static int 418 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain, 419 uint32_t *tablearg) 420 { 421 422 if (ifp == NULL) /* no iface with this packet, match fails */ 423 return (0); 424 425 /* Check by name or by IP address */ 426 if (cmd->name[0] != '\0') { /* match by name */ 427 if (cmd->name[0] == '\1') /* use tablearg to match */ 428 return ipfw_lookup_table(chain, cmd->p.kidx, 0, 429 &ifp->if_index, tablearg); 430 /* Check name */ 431 if (cmd->p.glob) { 432 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 433 return(1); 434 } else { 435 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 436 return(1); 437 } 438 } else { 439 #if !defined(USERSPACE) && defined(__FreeBSD__) /* and OSX too ? */ 440 struct ifaddr *ia; 441 442 NET_EPOCH_ASSERT(); 443 444 CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 445 if (ia->ifa_addr->sa_family != AF_INET) 446 continue; 447 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 448 (ia->ifa_addr))->sin_addr.s_addr) 449 return (1); /* match */ 450 } 451 #endif /* __FreeBSD__ */ 452 } 453 return(0); /* no match, fail ... */ 454 } 455 456 /* 457 * The verify_path function checks if a route to the src exists and 458 * if it is reachable via ifp (when provided). 459 * 460 * The 'verrevpath' option checks that the interface that an IP packet 461 * arrives on is the same interface that traffic destined for the 462 * packet's source address would be routed out of. 463 * The 'versrcreach' option just checks that the source address is 464 * reachable via any route (except default) in the routing table. 465 * These two are a measure to block forged packets. This is also 466 * commonly known as "anti-spoofing" or Unicast Reverse Path 467 * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 468 * is purposely reminiscent of the Cisco IOS command, 469 * 470 * ip verify unicast reverse-path 471 * ip verify unicast source reachable-via any 472 * 473 * which implements the same functionality. But note that the syntax 474 * is misleading, and the check may be performed on all IP packets 475 * whether unicast, multicast, or broadcast. 476 */ 477 static int 478 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) 479 { 480 #if defined(USERSPACE) || !defined(__FreeBSD__) 481 return 0; 482 #else 483 struct nhop_object *nh; 484 485 nh = fib4_lookup(fib, src, 0, NHR_NONE, 0); 486 if (nh == NULL) 487 return (0); 488 489 /* 490 * If ifp is provided, check for equality with rtentry. 491 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 492 * in order to pass packets injected back by if_simloop(): 493 * routing entry (via lo0) for our own address 494 * may exist, so we need to handle routing assymetry. 495 */ 496 if (ifp != NULL && ifp != nh->nh_aifp) 497 return (0); 498 499 /* if no ifp provided, check if rtentry is not default route */ 500 if (ifp == NULL && (nh->nh_flags & NHF_DEFAULT) != 0) 501 return (0); 502 503 /* or if this is a blackhole/reject route */ 504 if (ifp == NULL && (nh->nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0) 505 return (0); 506 507 /* found valid route */ 508 return 1; 509 #endif /* __FreeBSD__ */ 510 } 511 512 /* 513 * Generate an SCTP packet containing an ABORT chunk. The verification tag 514 * is given by vtag. The T-bit is set in the ABORT chunk if and only if 515 * reflected is not 0. 516 */ 517 518 static struct mbuf * 519 ipfw_send_abort(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t vtag, 520 int reflected) 521 { 522 struct mbuf *m; 523 struct ip *ip; 524 #ifdef INET6 525 struct ip6_hdr *ip6; 526 #endif 527 struct sctphdr *sctp; 528 struct sctp_chunkhdr *chunk; 529 u_int16_t hlen, plen, tlen; 530 531 MGETHDR(m, M_NOWAIT, MT_DATA); 532 if (m == NULL) 533 return (NULL); 534 535 M_SETFIB(m, id->fib); 536 #ifdef MAC 537 if (replyto != NULL) 538 mac_netinet_firewall_reply(replyto, m); 539 else 540 mac_netinet_firewall_send(m); 541 #else 542 (void)replyto; /* don't warn about unused arg */ 543 #endif 544 545 switch (id->addr_type) { 546 case 4: 547 hlen = sizeof(struct ip); 548 break; 549 #ifdef INET6 550 case 6: 551 hlen = sizeof(struct ip6_hdr); 552 break; 553 #endif 554 default: 555 /* XXX: log me?!? */ 556 FREE_PKT(m); 557 return (NULL); 558 } 559 plen = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); 560 tlen = hlen + plen; 561 m->m_data += max_linkhdr; 562 m->m_flags |= M_SKIP_FIREWALL; 563 m->m_pkthdr.len = m->m_len = tlen; 564 m->m_pkthdr.rcvif = NULL; 565 bzero(m->m_data, tlen); 566 567 switch (id->addr_type) { 568 case 4: 569 ip = mtod(m, struct ip *); 570 571 ip->ip_v = 4; 572 ip->ip_hl = sizeof(struct ip) >> 2; 573 ip->ip_tos = IPTOS_LOWDELAY; 574 ip->ip_len = htons(tlen); 575 ip->ip_id = htons(0); 576 ip->ip_off = htons(0); 577 ip->ip_ttl = V_ip_defttl; 578 ip->ip_p = IPPROTO_SCTP; 579 ip->ip_sum = 0; 580 ip->ip_src.s_addr = htonl(id->dst_ip); 581 ip->ip_dst.s_addr = htonl(id->src_ip); 582 583 sctp = (struct sctphdr *)(ip + 1); 584 break; 585 #ifdef INET6 586 case 6: 587 ip6 = mtod(m, struct ip6_hdr *); 588 589 ip6->ip6_vfc = IPV6_VERSION; 590 ip6->ip6_plen = htons(plen); 591 ip6->ip6_nxt = IPPROTO_SCTP; 592 ip6->ip6_hlim = IPV6_DEFHLIM; 593 ip6->ip6_src = id->dst_ip6; 594 ip6->ip6_dst = id->src_ip6; 595 596 sctp = (struct sctphdr *)(ip6 + 1); 597 break; 598 #endif 599 } 600 601 sctp->src_port = htons(id->dst_port); 602 sctp->dest_port = htons(id->src_port); 603 sctp->v_tag = htonl(vtag); 604 sctp->checksum = htonl(0); 605 606 chunk = (struct sctp_chunkhdr *)(sctp + 1); 607 chunk->chunk_type = SCTP_ABORT_ASSOCIATION; 608 chunk->chunk_flags = 0; 609 if (reflected != 0) { 610 chunk->chunk_flags |= SCTP_HAD_NO_TCB; 611 } 612 chunk->chunk_length = htons(sizeof(struct sctp_chunkhdr)); 613 614 sctp->checksum = sctp_calculate_cksum(m, hlen); 615 616 return (m); 617 } 618 619 /* 620 * Generate a TCP packet, containing either a RST or a keepalive. 621 * When flags & TH_RST, we are sending a RST packet, because of a 622 * "reset" action matched the packet. 623 * Otherwise we are sending a keepalive, and flags & TH_ 624 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 625 * so that MAC can label the reply appropriately. 626 */ 627 struct mbuf * 628 ipfw_send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 629 u_int32_t ack, int flags) 630 { 631 struct mbuf *m = NULL; /* stupid compiler */ 632 struct ip *h = NULL; /* stupid compiler */ 633 #ifdef INET6 634 struct ip6_hdr *h6 = NULL; 635 #endif 636 struct tcphdr *th = NULL; 637 int len, dir; 638 639 MGETHDR(m, M_NOWAIT, MT_DATA); 640 if (m == NULL) 641 return (NULL); 642 643 M_SETFIB(m, id->fib); 644 #ifdef MAC 645 if (replyto != NULL) 646 mac_netinet_firewall_reply(replyto, m); 647 else 648 mac_netinet_firewall_send(m); 649 #else 650 (void)replyto; /* don't warn about unused arg */ 651 #endif 652 653 switch (id->addr_type) { 654 case 4: 655 len = sizeof(struct ip) + sizeof(struct tcphdr); 656 break; 657 #ifdef INET6 658 case 6: 659 len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); 660 break; 661 #endif 662 default: 663 /* XXX: log me?!? */ 664 FREE_PKT(m); 665 return (NULL); 666 } 667 dir = ((flags & (TH_SYN | TH_RST)) == TH_SYN); 668 669 m->m_data += max_linkhdr; 670 m->m_flags |= M_SKIP_FIREWALL; 671 m->m_pkthdr.len = m->m_len = len; 672 m->m_pkthdr.rcvif = NULL; 673 bzero(m->m_data, len); 674 675 switch (id->addr_type) { 676 case 4: 677 h = mtod(m, struct ip *); 678 679 /* prepare for checksum */ 680 h->ip_p = IPPROTO_TCP; 681 h->ip_len = htons(sizeof(struct tcphdr)); 682 if (dir) { 683 h->ip_src.s_addr = htonl(id->src_ip); 684 h->ip_dst.s_addr = htonl(id->dst_ip); 685 } else { 686 h->ip_src.s_addr = htonl(id->dst_ip); 687 h->ip_dst.s_addr = htonl(id->src_ip); 688 } 689 690 th = (struct tcphdr *)(h + 1); 691 break; 692 #ifdef INET6 693 case 6: 694 h6 = mtod(m, struct ip6_hdr *); 695 696 /* prepare for checksum */ 697 h6->ip6_nxt = IPPROTO_TCP; 698 h6->ip6_plen = htons(sizeof(struct tcphdr)); 699 if (dir) { 700 h6->ip6_src = id->src_ip6; 701 h6->ip6_dst = id->dst_ip6; 702 } else { 703 h6->ip6_src = id->dst_ip6; 704 h6->ip6_dst = id->src_ip6; 705 } 706 707 th = (struct tcphdr *)(h6 + 1); 708 break; 709 #endif 710 } 711 712 if (dir) { 713 th->th_sport = htons(id->src_port); 714 th->th_dport = htons(id->dst_port); 715 } else { 716 th->th_sport = htons(id->dst_port); 717 th->th_dport = htons(id->src_port); 718 } 719 th->th_off = sizeof(struct tcphdr) >> 2; 720 721 if (flags & TH_RST) { 722 if (flags & TH_ACK) { 723 th->th_seq = htonl(ack); 724 th->th_flags = TH_RST; 725 } else { 726 if (flags & TH_SYN) 727 seq++; 728 th->th_ack = htonl(seq); 729 th->th_flags = TH_RST | TH_ACK; 730 } 731 } else { 732 /* 733 * Keepalive - use caller provided sequence numbers 734 */ 735 th->th_seq = htonl(seq); 736 th->th_ack = htonl(ack); 737 th->th_flags = TH_ACK; 738 } 739 740 switch (id->addr_type) { 741 case 4: 742 th->th_sum = in_cksum(m, len); 743 744 /* finish the ip header */ 745 h->ip_v = 4; 746 h->ip_hl = sizeof(*h) >> 2; 747 h->ip_tos = IPTOS_LOWDELAY; 748 h->ip_off = htons(0); 749 h->ip_len = htons(len); 750 h->ip_ttl = V_ip_defttl; 751 h->ip_sum = 0; 752 break; 753 #ifdef INET6 754 case 6: 755 th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(*h6), 756 sizeof(struct tcphdr)); 757 758 /* finish the ip6 header */ 759 h6->ip6_vfc |= IPV6_VERSION; 760 h6->ip6_hlim = IPV6_DEFHLIM; 761 break; 762 #endif 763 } 764 765 return (m); 766 } 767 768 #ifdef INET6 769 /* 770 * ipv6 specific rules here... 771 */ 772 static __inline int 773 icmp6type_match(int type, ipfw_insn_u32 *cmd) 774 { 775 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 776 } 777 778 static int 779 flow6id_match(int curr_flow, ipfw_insn_u32 *cmd) 780 { 781 int i; 782 for (i=0; i <= cmd->o.arg1; ++i) 783 if (curr_flow == cmd->d[i]) 784 return 1; 785 return 0; 786 } 787 788 /* support for IP6_*_ME opcodes */ 789 static const struct in6_addr lla_mask = {{{ 790 0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff, 791 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff 792 }}}; 793 794 static int 795 ipfw_localip6(struct in6_addr *in6) 796 { 797 struct rm_priotracker in6_ifa_tracker; 798 struct in6_ifaddr *ia; 799 800 if (IN6_IS_ADDR_MULTICAST(in6)) 801 return (0); 802 803 if (!IN6_IS_ADDR_LINKLOCAL(in6)) 804 return (in6_localip(in6)); 805 806 IN6_IFADDR_RLOCK(&in6_ifa_tracker); 807 CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) { 808 if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) 809 continue; 810 if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr, 811 in6, &lla_mask)) { 812 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); 813 return (1); 814 } 815 } 816 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); 817 return (0); 818 } 819 820 static int 821 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib) 822 { 823 struct nhop_object *nh; 824 825 if (IN6_IS_SCOPE_LINKLOCAL(src)) 826 return (1); 827 828 nh = fib6_lookup(fib, src, 0, NHR_NONE, 0); 829 if (nh == NULL) 830 return (0); 831 832 /* If ifp is provided, check for equality with route table. */ 833 if (ifp != NULL && ifp != nh->nh_aifp) 834 return (0); 835 836 /* if no ifp provided, check if rtentry is not default route */ 837 if (ifp == NULL && (nh->nh_flags & NHF_DEFAULT) != 0) 838 return (0); 839 840 /* or if this is a blackhole/reject route */ 841 if (ifp == NULL && (nh->nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0) 842 return (0); 843 844 /* found valid route */ 845 return 1; 846 } 847 848 static int 849 is_icmp6_query(int icmp6_type) 850 { 851 if ((icmp6_type <= ICMP6_MAXTYPE) && 852 (icmp6_type == ICMP6_ECHO_REQUEST || 853 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 854 icmp6_type == ICMP6_WRUREQUEST || 855 icmp6_type == ICMP6_FQDN_QUERY || 856 icmp6_type == ICMP6_NI_QUERY)) 857 return (1); 858 859 return (0); 860 } 861 862 static int 863 map_icmp_unreach(int code) 864 { 865 866 /* RFC 7915 p4.2 */ 867 switch (code) { 868 case ICMP_UNREACH_NET: 869 case ICMP_UNREACH_HOST: 870 case ICMP_UNREACH_SRCFAIL: 871 case ICMP_UNREACH_NET_UNKNOWN: 872 case ICMP_UNREACH_HOST_UNKNOWN: 873 case ICMP_UNREACH_TOSNET: 874 case ICMP_UNREACH_TOSHOST: 875 return (ICMP6_DST_UNREACH_NOROUTE); 876 case ICMP_UNREACH_PORT: 877 return (ICMP6_DST_UNREACH_NOPORT); 878 default: 879 /* 880 * Map the rest of codes into admit prohibited. 881 * XXX: unreach proto should be mapped into ICMPv6 882 * parameter problem, but we use only unreach type. 883 */ 884 return (ICMP6_DST_UNREACH_ADMIN); 885 } 886 } 887 888 static void 889 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 890 { 891 struct mbuf *m; 892 893 m = args->m; 894 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 895 struct tcphdr *tcp; 896 tcp = (struct tcphdr *)((char *)ip6 + hlen); 897 898 if ((tcp->th_flags & TH_RST) == 0) { 899 struct mbuf *m0; 900 m0 = ipfw_send_pkt(args->m, &(args->f_id), 901 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 902 tcp->th_flags | TH_RST); 903 if (m0 != NULL) 904 ip6_output(m0, NULL, NULL, 0, NULL, NULL, 905 NULL); 906 } 907 FREE_PKT(m); 908 } else if (code == ICMP6_UNREACH_ABORT && 909 args->f_id.proto == IPPROTO_SCTP) { 910 struct mbuf *m0; 911 struct sctphdr *sctp; 912 u_int32_t v_tag; 913 int reflected; 914 915 sctp = (struct sctphdr *)((char *)ip6 + hlen); 916 reflected = 1; 917 v_tag = ntohl(sctp->v_tag); 918 /* Investigate the first chunk header if available */ 919 if (m->m_len >= hlen + sizeof(struct sctphdr) + 920 sizeof(struct sctp_chunkhdr)) { 921 struct sctp_chunkhdr *chunk; 922 923 chunk = (struct sctp_chunkhdr *)(sctp + 1); 924 switch (chunk->chunk_type) { 925 case SCTP_INITIATION: 926 /* 927 * Packets containing an INIT chunk MUST have 928 * a zero v-tag. 929 */ 930 if (v_tag != 0) { 931 v_tag = 0; 932 break; 933 } 934 /* INIT chunk MUST NOT be bundled */ 935 if (m->m_pkthdr.len > 936 hlen + sizeof(struct sctphdr) + 937 ntohs(chunk->chunk_length) + 3) { 938 break; 939 } 940 /* Use the initiate tag if available */ 941 if ((m->m_len >= hlen + sizeof(struct sctphdr) + 942 sizeof(struct sctp_chunkhdr) + 943 offsetof(struct sctp_init, a_rwnd))) { 944 struct sctp_init *init; 945 946 init = (struct sctp_init *)(chunk + 1); 947 v_tag = ntohl(init->initiate_tag); 948 reflected = 0; 949 } 950 break; 951 case SCTP_ABORT_ASSOCIATION: 952 /* 953 * If the packet contains an ABORT chunk, don't 954 * reply. 955 * XXX: We should search through all chunks, 956 * but do not do that to avoid attacks. 957 */ 958 v_tag = 0; 959 break; 960 } 961 } 962 if (v_tag == 0) { 963 m0 = NULL; 964 } else { 965 m0 = ipfw_send_abort(args->m, &(args->f_id), v_tag, 966 reflected); 967 } 968 if (m0 != NULL) 969 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL); 970 FREE_PKT(m); 971 } else if (code != ICMP6_UNREACH_RST && code != ICMP6_UNREACH_ABORT) { 972 /* Send an ICMPv6 unreach. */ 973 #if 0 974 /* 975 * Unlike above, the mbufs need to line up with the ip6 hdr, 976 * as the contents are read. We need to m_adj() the 977 * needed amount. 978 * The mbuf will however be thrown away so we can adjust it. 979 * Remember we did an m_pullup on it already so we 980 * can make some assumptions about contiguousness. 981 */ 982 if (args->L3offset) 983 m_adj(m, args->L3offset); 984 #endif 985 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 986 } else 987 FREE_PKT(m); 988 989 args->m = NULL; 990 } 991 992 #endif /* INET6 */ 993 994 /* 995 * sends a reject message, consuming the mbuf passed as an argument. 996 */ 997 static void 998 send_reject(struct ip_fw_args *args, const ipfw_insn *cmd, int iplen, 999 struct ip *ip) 1000 { 1001 int code, mtu; 1002 1003 code = cmd->arg1; 1004 if (code == ICMP_UNREACH_NEEDFRAG && 1005 cmd->len == F_INSN_SIZE(ipfw_insn_u16)) 1006 mtu = ((const ipfw_insn_u16 *)cmd)->ports[0]; 1007 else 1008 mtu = 0; 1009 1010 #if 0 1011 /* XXX When ip is not guaranteed to be at mtod() we will 1012 * need to account for this */ 1013 * The mbuf will however be thrown away so we can adjust it. 1014 * Remember we did an m_pullup on it already so we 1015 * can make some assumptions about contiguousness. 1016 */ 1017 if (args->L3offset) 1018 m_adj(m, args->L3offset); 1019 #endif 1020 if (code != ICMP_REJECT_RST && code != ICMP_REJECT_ABORT) { 1021 /* Send an ICMP unreach */ 1022 icmp_error(args->m, ICMP_UNREACH, code, 0L, mtu); 1023 } else if (code == ICMP_REJECT_RST && args->f_id.proto == IPPROTO_TCP) { 1024 struct tcphdr *const tcp = 1025 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1026 if ( (tcp->th_flags & TH_RST) == 0) { 1027 struct mbuf *m; 1028 m = ipfw_send_pkt(args->m, &(args->f_id), 1029 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1030 tcp->th_flags | TH_RST); 1031 if (m != NULL) 1032 ip_output(m, NULL, NULL, 0, NULL, NULL); 1033 } 1034 FREE_PKT(args->m); 1035 } else if (code == ICMP_REJECT_ABORT && 1036 args->f_id.proto == IPPROTO_SCTP) { 1037 struct mbuf *m; 1038 struct sctphdr *sctp; 1039 struct sctp_chunkhdr *chunk; 1040 struct sctp_init *init; 1041 u_int32_t v_tag; 1042 int reflected; 1043 1044 sctp = L3HDR(struct sctphdr, mtod(args->m, struct ip *)); 1045 reflected = 1; 1046 v_tag = ntohl(sctp->v_tag); 1047 if (iplen >= (ip->ip_hl << 2) + sizeof(struct sctphdr) + 1048 sizeof(struct sctp_chunkhdr)) { 1049 /* Look at the first chunk header if available */ 1050 chunk = (struct sctp_chunkhdr *)(sctp + 1); 1051 switch (chunk->chunk_type) { 1052 case SCTP_INITIATION: 1053 /* 1054 * Packets containing an INIT chunk MUST have 1055 * a zero v-tag. 1056 */ 1057 if (v_tag != 0) { 1058 v_tag = 0; 1059 break; 1060 } 1061 /* INIT chunk MUST NOT be bundled */ 1062 if (iplen > 1063 (ip->ip_hl << 2) + sizeof(struct sctphdr) + 1064 ntohs(chunk->chunk_length) + 3) { 1065 break; 1066 } 1067 /* Use the initiate tag if available */ 1068 if ((iplen >= (ip->ip_hl << 2) + 1069 sizeof(struct sctphdr) + 1070 sizeof(struct sctp_chunkhdr) + 1071 offsetof(struct sctp_init, a_rwnd))) { 1072 init = (struct sctp_init *)(chunk + 1); 1073 v_tag = ntohl(init->initiate_tag); 1074 reflected = 0; 1075 } 1076 break; 1077 case SCTP_ABORT_ASSOCIATION: 1078 /* 1079 * If the packet contains an ABORT chunk, don't 1080 * reply. 1081 * XXX: We should search through all chunks, 1082 * but do not do that to avoid attacks. 1083 */ 1084 v_tag = 0; 1085 break; 1086 } 1087 } 1088 if (v_tag == 0) { 1089 m = NULL; 1090 } else { 1091 m = ipfw_send_abort(args->m, &(args->f_id), v_tag, 1092 reflected); 1093 } 1094 if (m != NULL) 1095 ip_output(m, NULL, NULL, 0, NULL, NULL); 1096 FREE_PKT(args->m); 1097 } else 1098 FREE_PKT(args->m); 1099 args->m = NULL; 1100 } 1101 1102 /* 1103 * Support for uid/gid/jail lookup. These tests are expensive 1104 * (because we may need to look into the list of active sockets) 1105 * so we cache the results. ugid_lookupp is 0 if we have not 1106 * yet done a lookup, 1 if we succeeded, and -1 if we tried 1107 * and failed. The function always returns the match value. 1108 * We could actually spare the variable and use *uc, setting 1109 * it to '(void *)check_uidgid if we have no info, NULL if 1110 * we tried and failed, or any other value if successful. 1111 */ 1112 static int 1113 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp, 1114 struct ucred **uc) 1115 { 1116 #if defined(USERSPACE) 1117 return 0; // not supported in userspace 1118 #else 1119 #ifndef __FreeBSD__ 1120 /* XXX */ 1121 return cred_check(insn, proto, oif, 1122 dst_ip, dst_port, src_ip, src_port, 1123 (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb); 1124 #else /* FreeBSD */ 1125 struct in_addr src_ip, dst_ip; 1126 struct inpcbinfo *pi; 1127 struct ipfw_flow_id *id; 1128 struct inpcb *pcb, *inp; 1129 int lookupflags; 1130 int match; 1131 1132 id = &args->f_id; 1133 inp = args->inp; 1134 1135 /* 1136 * Check to see if the UDP or TCP stack supplied us with 1137 * the PCB. If so, rather then holding a lock and looking 1138 * up the PCB, we can use the one that was supplied. 1139 */ 1140 if (inp && *ugid_lookupp == 0) { 1141 INP_LOCK_ASSERT(inp); 1142 if (inp->inp_socket != NULL) { 1143 *uc = crhold(inp->inp_cred); 1144 *ugid_lookupp = 1; 1145 } else 1146 *ugid_lookupp = -1; 1147 } 1148 /* 1149 * If we have already been here and the packet has no 1150 * PCB entry associated with it, then we can safely 1151 * assume that this is a no match. 1152 */ 1153 if (*ugid_lookupp == -1) 1154 return (0); 1155 if (id->proto == IPPROTO_TCP) { 1156 lookupflags = 0; 1157 pi = &V_tcbinfo; 1158 } else if (id->proto == IPPROTO_UDP) { 1159 lookupflags = INPLOOKUP_WILDCARD; 1160 pi = &V_udbinfo; 1161 } else if (id->proto == IPPROTO_UDPLITE) { 1162 lookupflags = INPLOOKUP_WILDCARD; 1163 pi = &V_ulitecbinfo; 1164 } else 1165 return 0; 1166 lookupflags |= INPLOOKUP_RLOCKPCB; 1167 match = 0; 1168 if (*ugid_lookupp == 0) { 1169 if (id->addr_type == 6) { 1170 #ifdef INET6 1171 if (args->flags & IPFW_ARGS_IN) 1172 pcb = in6_pcblookup_mbuf(pi, 1173 &id->src_ip6, htons(id->src_port), 1174 &id->dst_ip6, htons(id->dst_port), 1175 lookupflags, NULL, args->m); 1176 else 1177 pcb = in6_pcblookup_mbuf(pi, 1178 &id->dst_ip6, htons(id->dst_port), 1179 &id->src_ip6, htons(id->src_port), 1180 lookupflags, args->ifp, args->m); 1181 #else 1182 *ugid_lookupp = -1; 1183 return (0); 1184 #endif 1185 } else { 1186 src_ip.s_addr = htonl(id->src_ip); 1187 dst_ip.s_addr = htonl(id->dst_ip); 1188 if (args->flags & IPFW_ARGS_IN) 1189 pcb = in_pcblookup_mbuf(pi, 1190 src_ip, htons(id->src_port), 1191 dst_ip, htons(id->dst_port), 1192 lookupflags, NULL, args->m); 1193 else 1194 pcb = in_pcblookup_mbuf(pi, 1195 dst_ip, htons(id->dst_port), 1196 src_ip, htons(id->src_port), 1197 lookupflags, args->ifp, args->m); 1198 } 1199 if (pcb != NULL) { 1200 INP_RLOCK_ASSERT(pcb); 1201 *uc = crhold(pcb->inp_cred); 1202 *ugid_lookupp = 1; 1203 INP_RUNLOCK(pcb); 1204 } 1205 if (*ugid_lookupp == 0) { 1206 /* 1207 * We tried and failed, set the variable to -1 1208 * so we will not try again on this packet. 1209 */ 1210 *ugid_lookupp = -1; 1211 return (0); 1212 } 1213 } 1214 if (insn->o.opcode == O_UID) 1215 match = ((*uc)->cr_uid == (uid_t)insn->d[0]); 1216 else if (insn->o.opcode == O_GID) 1217 match = groupmember((gid_t)insn->d[0], *uc); 1218 else if (insn->o.opcode == O_JAIL) 1219 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]); 1220 return (match); 1221 #endif /* __FreeBSD__ */ 1222 #endif /* not supported in userspace */ 1223 } 1224 1225 /* 1226 * Helper function to set args with info on the rule after the matching 1227 * one. slot is precise, whereas we guess rule_id as they are 1228 * assigned sequentially. 1229 */ 1230 static inline void 1231 set_match(struct ip_fw_args *args, int slot, 1232 struct ip_fw_chain *chain) 1233 { 1234 args->rule.chain_id = chain->id; 1235 args->rule.slot = slot + 1; /* we use 0 as a marker */ 1236 args->rule.rule_id = 1 + chain->map[slot]->id; 1237 args->rule.rulenum = chain->map[slot]->rulenum; 1238 args->flags |= IPFW_ARGS_REF; 1239 } 1240 1241 static int 1242 jump_lookup_pos(struct ip_fw_chain *chain, struct ip_fw *f, int num, 1243 int tablearg, int jump_backwards) 1244 { 1245 int f_pos, i; 1246 1247 i = IP_FW_ARG_TABLEARG(chain, num, skipto); 1248 /* make sure we do not jump backward */ 1249 if (jump_backwards == 0 && i <= f->rulenum) 1250 i = f->rulenum + 1; 1251 1252 #ifndef LINEAR_SKIPTO 1253 if (chain->idxmap != NULL) 1254 f_pos = chain->idxmap[i]; 1255 else 1256 f_pos = ipfw_find_rule(chain, i, 0); 1257 #else 1258 f_pos = chain->idxmap[i]; 1259 #endif /* LINEAR_SKIPTO */ 1260 1261 return (f_pos); 1262 } 1263 1264 1265 #ifndef LINEAR_SKIPTO 1266 /* 1267 * Helper function to enable cached rule lookups using 1268 * cache.id and cache.pos fields in ipfw rule. 1269 */ 1270 static int 1271 jump_cached(struct ip_fw_chain *chain, struct ip_fw *f, int num, 1272 int tablearg, int jump_backwards) 1273 { 1274 int f_pos; 1275 1276 /* Can't use cache with IP_FW_TARG */ 1277 if (num == IP_FW_TARG) 1278 return jump_lookup_pos(chain, f, num, tablearg, jump_backwards); 1279 1280 /* 1281 * If possible use cached f_pos (in f->cache.pos), 1282 * whose version is written in f->cache.id (horrible hacks 1283 * to avoid changing the ABI). 1284 * 1285 * Multiple threads can execute the same rule simultaneously, 1286 * we need to ensure that cache.pos is updated before cache.id. 1287 */ 1288 1289 #ifdef __LP64__ 1290 struct ip_fw_jump_cache cache; 1291 1292 cache.raw_value = f->cache.raw_value; 1293 if (cache.id == chain->id) 1294 return (cache.pos); 1295 1296 f_pos = jump_lookup_pos(chain, f, num, tablearg, jump_backwards); 1297 1298 cache.pos = f_pos; 1299 cache.id = chain->id; 1300 f->cache.raw_value = cache.raw_value; 1301 #else 1302 if (f->cache.id == chain->id) { 1303 /* Load pos after id */ 1304 atomic_thread_fence_acq(); 1305 return (f->cache.pos); 1306 } 1307 1308 f_pos = jump_lookup_pos(chain, f, num, tablearg, jump_backwards); 1309 1310 f->cache.pos = f_pos; 1311 /* Store id after pos */ 1312 atomic_thread_fence_rel(); 1313 f->cache.id = chain->id; 1314 #endif /* !__LP64__ */ 1315 return (f_pos); 1316 } 1317 #endif /* !LINEAR_SKIPTO */ 1318 1319 #define TARG(k, f) IP_FW_ARG_TABLEARG(chain, k, f) 1320 /* 1321 * The main check routine for the firewall. 1322 * 1323 * All arguments are in args so we can modify them and return them 1324 * back to the caller. 1325 * 1326 * Parameters: 1327 * 1328 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1329 * Starts with the IP header. 1330 * args->L3offset Number of bytes bypassed if we came from L2. 1331 * e.g. often sizeof(eh) ** NOTYET ** 1332 * args->ifp Incoming or outgoing interface. 1333 * args->divert_rule (in/out) 1334 * Skip up to the first rule past this rule number; 1335 * upon return, non-zero port number for divert or tee. 1336 * 1337 * args->rule Pointer to the last matching rule (in/out) 1338 * args->next_hop Socket we are forwarding to (out). 1339 * args->next_hop6 IPv6 next hop we are forwarding to (out). 1340 * args->f_id Addresses grabbed from the packet (out) 1341 * args->rule.info a cookie depending on rule action 1342 * 1343 * Return value: 1344 * 1345 * IP_FW_PASS the packet must be accepted 1346 * IP_FW_DENY the packet must be dropped 1347 * IP_FW_DIVERT divert packet, port in m_tag 1348 * IP_FW_TEE tee packet, port in m_tag 1349 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 1350 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 1351 * args->rule contains the matching rule, 1352 * args->rule.info has additional information. 1353 * 1354 */ 1355 int 1356 ipfw_chk(struct ip_fw_args *args) 1357 { 1358 1359 /* 1360 * Local variables holding state while processing a packet: 1361 * 1362 * IMPORTANT NOTE: to speed up the processing of rules, there 1363 * are some assumption on the values of the variables, which 1364 * are documented here. Should you change them, please check 1365 * the implementation of the various instructions to make sure 1366 * that they still work. 1367 * 1368 * m | args->m Pointer to the mbuf, as received from the caller. 1369 * It may change if ipfw_chk() does an m_pullup, or if it 1370 * consumes the packet because it calls send_reject(). 1371 * XXX This has to change, so that ipfw_chk() never modifies 1372 * or consumes the buffer. 1373 * OR 1374 * args->mem Pointer to contigous memory chunk. 1375 * ip Is the beginning of the ip(4 or 6) header. 1376 * eh Ethernet header in case if input is Layer2. 1377 */ 1378 struct mbuf *m; 1379 struct ip *ip; 1380 struct ether_header *eh; 1381 1382 /* 1383 * For rules which contain uid/gid or jail constraints, cache 1384 * a copy of the users credentials after the pcb lookup has been 1385 * executed. This will speed up the processing of rules with 1386 * these types of constraints, as well as decrease contention 1387 * on pcb related locks. 1388 */ 1389 #ifndef __FreeBSD__ 1390 struct bsd_ucred ucred_cache; 1391 #else 1392 struct ucred *ucred_cache = NULL; 1393 #endif 1394 int ucred_lookup = 0; 1395 int f_pos = 0; /* index of current rule in the array */ 1396 int retval = 0; 1397 struct ifnet *oif, *iif; 1398 1399 /* 1400 * hlen The length of the IP header. 1401 */ 1402 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 1403 1404 /* 1405 * offset The offset of a fragment. offset != 0 means that 1406 * we have a fragment at this offset of an IPv4 packet. 1407 * offset == 0 means that (if this is an IPv4 packet) 1408 * this is the first or only fragment. 1409 * For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header 1410 * or there is a single packet fragment (fragment header added 1411 * without needed). We will treat a single packet fragment as if 1412 * there was no fragment header (or log/block depending on the 1413 * V_fw_permit_single_frag6 sysctl setting). 1414 */ 1415 u_short offset = 0; 1416 u_short ip6f_mf = 0; 1417 1418 /* 1419 * Local copies of addresses. They are only valid if we have 1420 * an IP packet. 1421 * 1422 * proto The protocol. Set to 0 for non-ip packets, 1423 * or to the protocol read from the packet otherwise. 1424 * proto != 0 means that we have an IPv4 packet. 1425 * 1426 * src_port, dst_port port numbers, in HOST format. Only 1427 * valid for TCP and UDP packets. 1428 * 1429 * src_ip, dst_ip ip addresses, in NETWORK format. 1430 * Only valid for IPv4 packets. 1431 */ 1432 uint8_t proto; 1433 uint16_t src_port, dst_port; /* NOTE: host format */ 1434 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 1435 int iplen = 0; 1436 int pktlen; 1437 1438 struct ipfw_dyn_info dyn_info; 1439 struct ip_fw *q = NULL; 1440 struct ip_fw_chain *chain = &V_layer3_chain; 1441 1442 /* 1443 * We store in ulp a pointer to the upper layer protocol header. 1444 * In the ipv4 case this is easy to determine from the header, 1445 * but for ipv6 we might have some additional headers in the middle. 1446 * ulp is NULL if not found. 1447 */ 1448 void *ulp = NULL; /* upper layer protocol pointer. */ 1449 1450 /* XXX ipv6 variables */ 1451 int is_ipv6 = 0; 1452 #ifdef INET6 1453 uint8_t icmp6_type = 0; 1454 #endif 1455 uint16_t ext_hd = 0; /* bits vector for extension header filtering */ 1456 /* end of ipv6 variables */ 1457 1458 int is_ipv4 = 0; 1459 1460 int done = 0; /* flag to exit the outer loop */ 1461 IPFW_RLOCK_TRACKER; 1462 bool mem; 1463 1464 if ((mem = (args->flags & IPFW_ARGS_LENMASK))) { 1465 if (args->flags & IPFW_ARGS_ETHER) { 1466 eh = (struct ether_header *)args->mem; 1467 if (eh->ether_type == htons(ETHERTYPE_VLAN)) 1468 ip = (struct ip *) 1469 ((struct ether_vlan_header *)eh + 1); 1470 else 1471 ip = (struct ip *)(eh + 1); 1472 } else { 1473 eh = NULL; 1474 ip = (struct ip *)args->mem; 1475 } 1476 pktlen = IPFW_ARGS_LENGTH(args->flags); 1477 args->f_id.fib = args->ifp->if_fib; /* best guess */ 1478 } else { 1479 m = args->m; 1480 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready)) 1481 return (IP_FW_PASS); /* accept */ 1482 if (args->flags & IPFW_ARGS_ETHER) { 1483 /* We need some amount of data to be contiguous. */ 1484 if (m->m_len < min(m->m_pkthdr.len, max_protohdr) && 1485 (args->m = m = m_pullup(m, min(m->m_pkthdr.len, 1486 max_protohdr))) == NULL) 1487 goto pullup_failed; 1488 eh = mtod(m, struct ether_header *); 1489 ip = (struct ip *)(eh + 1); 1490 } else { 1491 eh = NULL; 1492 ip = mtod(m, struct ip *); 1493 } 1494 pktlen = m->m_pkthdr.len; 1495 args->f_id.fib = M_GETFIB(m); /* mbuf not altered */ 1496 } 1497 1498 dst_ip.s_addr = 0; /* make sure it is initialized */ 1499 src_ip.s_addr = 0; /* make sure it is initialized */ 1500 src_port = dst_port = 0; 1501 1502 DYN_INFO_INIT(&dyn_info); 1503 /* 1504 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 1505 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 1506 * pointer might become stale after other pullups (but we never use it 1507 * this way). 1508 */ 1509 #define PULLUP_TO(_len, p, T) PULLUP_LEN(_len, p, sizeof(T)) 1510 #define EHLEN (eh != NULL ? ((char *)ip - (char *)eh) : 0) 1511 #define _PULLUP_LOCKED(_len, p, T, unlock) \ 1512 do { \ 1513 int x = (_len) + T + EHLEN; \ 1514 if (mem) { \ 1515 if (__predict_false(pktlen < x)) { \ 1516 unlock; \ 1517 goto pullup_failed; \ 1518 } \ 1519 p = (char *)args->mem + (_len) + EHLEN; \ 1520 } else { \ 1521 if (__predict_false((m)->m_len < x)) { \ 1522 args->m = m = m_pullup(m, x); \ 1523 if (m == NULL) { \ 1524 unlock; \ 1525 goto pullup_failed; \ 1526 } \ 1527 } \ 1528 p = mtod(m, char *) + (_len) + EHLEN; \ 1529 } \ 1530 } while (0) 1531 1532 #define PULLUP_LEN(_len, p, T) _PULLUP_LOCKED(_len, p, T, ) 1533 #define PULLUP_LEN_LOCKED(_len, p, T) \ 1534 _PULLUP_LOCKED(_len, p, T, IPFW_PF_RUNLOCK(chain)); \ 1535 UPDATE_POINTERS() 1536 /* 1537 * In case pointers got stale after pullups, update them. 1538 */ 1539 #define UPDATE_POINTERS() \ 1540 do { \ 1541 if (!mem) { \ 1542 if (eh != NULL) { \ 1543 eh = mtod(m, struct ether_header *); \ 1544 ip = (struct ip *)(eh + 1); \ 1545 } else \ 1546 ip = mtod(m, struct ip *); \ 1547 args->m = m; \ 1548 } \ 1549 } while (0) 1550 1551 /* Identify IP packets and fill up variables. */ 1552 if (pktlen >= sizeof(struct ip6_hdr) && 1553 (eh == NULL || eh->ether_type == htons(ETHERTYPE_IPV6)) && 1554 ip->ip_v == 6) { 1555 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 1556 1557 is_ipv6 = 1; 1558 args->flags |= IPFW_ARGS_IP6; 1559 hlen = sizeof(struct ip6_hdr); 1560 proto = ip6->ip6_nxt; 1561 /* Search extension headers to find upper layer protocols */ 1562 while (ulp == NULL && offset == 0) { 1563 switch (proto) { 1564 case IPPROTO_ICMPV6: 1565 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 1566 #ifdef INET6 1567 icmp6_type = ICMP6(ulp)->icmp6_type; 1568 #endif 1569 break; 1570 1571 case IPPROTO_TCP: 1572 PULLUP_TO(hlen, ulp, struct tcphdr); 1573 dst_port = TCP(ulp)->th_dport; 1574 src_port = TCP(ulp)->th_sport; 1575 /* save flags for dynamic rules */ 1576 args->f_id._flags = TCP(ulp)->th_flags; 1577 break; 1578 1579 case IPPROTO_SCTP: 1580 if (pktlen >= hlen + sizeof(struct sctphdr) + 1581 sizeof(struct sctp_chunkhdr) + 1582 offsetof(struct sctp_init, a_rwnd)) 1583 PULLUP_LEN(hlen, ulp, 1584 sizeof(struct sctphdr) + 1585 sizeof(struct sctp_chunkhdr) + 1586 offsetof(struct sctp_init, a_rwnd)); 1587 else if (pktlen >= hlen + sizeof(struct sctphdr)) 1588 PULLUP_LEN(hlen, ulp, pktlen - hlen); 1589 else 1590 PULLUP_LEN(hlen, ulp, 1591 sizeof(struct sctphdr)); 1592 src_port = SCTP(ulp)->src_port; 1593 dst_port = SCTP(ulp)->dest_port; 1594 break; 1595 1596 case IPPROTO_UDP: 1597 case IPPROTO_UDPLITE: 1598 PULLUP_TO(hlen, ulp, struct udphdr); 1599 dst_port = UDP(ulp)->uh_dport; 1600 src_port = UDP(ulp)->uh_sport; 1601 break; 1602 1603 case IPPROTO_HOPOPTS: /* RFC 2460 */ 1604 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1605 ext_hd |= EXT_HOPOPTS; 1606 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1607 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1608 ulp = NULL; 1609 break; 1610 1611 case IPPROTO_ROUTING: /* RFC 2460 */ 1612 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 1613 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 1614 case 0: 1615 ext_hd |= EXT_RTHDR0; 1616 break; 1617 case 2: 1618 ext_hd |= EXT_RTHDR2; 1619 break; 1620 default: 1621 if (V_fw_verbose) 1622 printf("IPFW2: IPV6 - Unknown " 1623 "Routing Header type(%d)\n", 1624 ((struct ip6_rthdr *) 1625 ulp)->ip6r_type); 1626 if (V_fw_deny_unknown_exthdrs) 1627 return (IP_FW_DENY); 1628 break; 1629 } 1630 ext_hd |= EXT_ROUTING; 1631 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 1632 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 1633 ulp = NULL; 1634 break; 1635 1636 case IPPROTO_FRAGMENT: /* RFC 2460 */ 1637 PULLUP_TO(hlen, ulp, struct ip6_frag); 1638 ext_hd |= EXT_FRAGMENT; 1639 hlen += sizeof (struct ip6_frag); 1640 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 1641 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 1642 IP6F_OFF_MASK; 1643 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg & 1644 IP6F_MORE_FRAG; 1645 if (V_fw_permit_single_frag6 == 0 && 1646 offset == 0 && ip6f_mf == 0) { 1647 if (V_fw_verbose) 1648 printf("IPFW2: IPV6 - Invalid " 1649 "Fragment Header\n"); 1650 if (V_fw_deny_unknown_exthdrs) 1651 return (IP_FW_DENY); 1652 break; 1653 } 1654 args->f_id.extra = 1655 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 1656 ulp = NULL; 1657 break; 1658 1659 case IPPROTO_DSTOPTS: /* RFC 2460 */ 1660 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1661 ext_hd |= EXT_DSTOPTS; 1662 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1663 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1664 ulp = NULL; 1665 break; 1666 1667 case IPPROTO_AH: /* RFC 2402 */ 1668 PULLUP_TO(hlen, ulp, struct ip6_ext); 1669 ext_hd |= EXT_AH; 1670 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 1671 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 1672 ulp = NULL; 1673 break; 1674 1675 case IPPROTO_ESP: /* RFC 2406 */ 1676 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 1677 /* Anything past Seq# is variable length and 1678 * data past this ext. header is encrypted. */ 1679 ext_hd |= EXT_ESP; 1680 break; 1681 1682 case IPPROTO_NONE: /* RFC 2460 */ 1683 /* 1684 * Packet ends here, and IPv6 header has 1685 * already been pulled up. If ip6e_len!=0 1686 * then octets must be ignored. 1687 */ 1688 ulp = ip; /* non-NULL to get out of loop. */ 1689 break; 1690 1691 case IPPROTO_OSPFIGP: 1692 /* XXX OSPF header check? */ 1693 PULLUP_TO(hlen, ulp, struct ip6_ext); 1694 break; 1695 1696 case IPPROTO_PIM: 1697 /* XXX PIM header check? */ 1698 PULLUP_TO(hlen, ulp, struct pim); 1699 break; 1700 1701 case IPPROTO_GRE: /* RFC 1701 */ 1702 /* XXX GRE header check? */ 1703 PULLUP_TO(hlen, ulp, struct grehdr); 1704 break; 1705 1706 case IPPROTO_CARP: 1707 PULLUP_TO(hlen, ulp, offsetof( 1708 struct carp_header, carp_counter)); 1709 if (CARP_ADVERTISEMENT != 1710 ((struct carp_header *)ulp)->carp_type) 1711 return (IP_FW_DENY); 1712 break; 1713 1714 case IPPROTO_IPV6: /* RFC 2893 */ 1715 PULLUP_TO(hlen, ulp, struct ip6_hdr); 1716 break; 1717 1718 case IPPROTO_IPV4: /* RFC 2893 */ 1719 PULLUP_TO(hlen, ulp, struct ip); 1720 break; 1721 1722 case IPPROTO_PFSYNC: 1723 PULLUP_TO(hlen, ulp, struct pfsync_header); 1724 break; 1725 1726 default: 1727 if (V_fw_verbose) 1728 printf("IPFW2: IPV6 - Unknown " 1729 "Extension Header(%d), ext_hd=%x\n", 1730 proto, ext_hd); 1731 if (V_fw_deny_unknown_exthdrs) 1732 return (IP_FW_DENY); 1733 PULLUP_TO(hlen, ulp, struct ip6_ext); 1734 break; 1735 } /*switch */ 1736 } 1737 UPDATE_POINTERS(); 1738 ip6 = (struct ip6_hdr *)ip; 1739 args->f_id.addr_type = 6; 1740 args->f_id.src_ip6 = ip6->ip6_src; 1741 args->f_id.dst_ip6 = ip6->ip6_dst; 1742 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 1743 iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6); 1744 } else if (pktlen >= sizeof(struct ip) && 1745 (eh == NULL || eh->ether_type == htons(ETHERTYPE_IP)) && 1746 ip->ip_v == 4) { 1747 is_ipv4 = 1; 1748 args->flags |= IPFW_ARGS_IP4; 1749 hlen = ip->ip_hl << 2; 1750 /* 1751 * Collect parameters into local variables for faster 1752 * matching. 1753 */ 1754 proto = ip->ip_p; 1755 src_ip = ip->ip_src; 1756 dst_ip = ip->ip_dst; 1757 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1758 iplen = ntohs(ip->ip_len); 1759 1760 if (offset == 0) { 1761 switch (proto) { 1762 case IPPROTO_TCP: 1763 PULLUP_TO(hlen, ulp, struct tcphdr); 1764 dst_port = TCP(ulp)->th_dport; 1765 src_port = TCP(ulp)->th_sport; 1766 /* save flags for dynamic rules */ 1767 args->f_id._flags = TCP(ulp)->th_flags; 1768 break; 1769 1770 case IPPROTO_SCTP: 1771 if (pktlen >= hlen + sizeof(struct sctphdr) + 1772 sizeof(struct sctp_chunkhdr) + 1773 offsetof(struct sctp_init, a_rwnd)) 1774 PULLUP_LEN(hlen, ulp, 1775 sizeof(struct sctphdr) + 1776 sizeof(struct sctp_chunkhdr) + 1777 offsetof(struct sctp_init, a_rwnd)); 1778 else if (pktlen >= hlen + sizeof(struct sctphdr)) 1779 PULLUP_LEN(hlen, ulp, pktlen - hlen); 1780 else 1781 PULLUP_LEN(hlen, ulp, 1782 sizeof(struct sctphdr)); 1783 src_port = SCTP(ulp)->src_port; 1784 dst_port = SCTP(ulp)->dest_port; 1785 break; 1786 1787 case IPPROTO_UDP: 1788 case IPPROTO_UDPLITE: 1789 PULLUP_TO(hlen, ulp, struct udphdr); 1790 dst_port = UDP(ulp)->uh_dport; 1791 src_port = UDP(ulp)->uh_sport; 1792 break; 1793 1794 case IPPROTO_ICMP: 1795 PULLUP_TO(hlen, ulp, struct icmphdr); 1796 //args->f_id.flags = ICMP(ulp)->icmp_type; 1797 break; 1798 1799 default: 1800 break; 1801 } 1802 } else { 1803 if (offset == 1 && proto == IPPROTO_TCP) { 1804 /* RFC 3128 */ 1805 goto pullup_failed; 1806 } 1807 } 1808 1809 UPDATE_POINTERS(); 1810 args->f_id.addr_type = 4; 1811 args->f_id.src_ip = ntohl(src_ip.s_addr); 1812 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1813 } else { 1814 proto = 0; 1815 dst_ip.s_addr = src_ip.s_addr = 0; 1816 1817 args->f_id.addr_type = 1; /* XXX */ 1818 } 1819 #undef PULLUP_TO 1820 pktlen = iplen < pktlen ? iplen: pktlen; 1821 1822 /* Properly initialize the rest of f_id */ 1823 args->f_id.proto = proto; 1824 args->f_id.src_port = src_port = ntohs(src_port); 1825 args->f_id.dst_port = dst_port = ntohs(dst_port); 1826 1827 IPFW_PF_RLOCK(chain); 1828 if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */ 1829 IPFW_PF_RUNLOCK(chain); 1830 return (IP_FW_PASS); /* accept */ 1831 } 1832 if (args->flags & IPFW_ARGS_REF) { 1833 /* 1834 * Packet has already been tagged as a result of a previous 1835 * match on rule args->rule aka args->rule_id (PIPE, QUEUE, 1836 * REASS, NETGRAPH, DIVERT/TEE...) 1837 * Validate the slot and continue from the next one 1838 * if still present, otherwise do a lookup. 1839 */ 1840 f_pos = (args->rule.chain_id == chain->id) ? 1841 args->rule.slot : 1842 ipfw_find_rule(chain, args->rule.rulenum, 1843 args->rule.rule_id); 1844 } else { 1845 f_pos = 0; 1846 } 1847 1848 if (args->flags & IPFW_ARGS_IN) { 1849 iif = args->ifp; 1850 oif = NULL; 1851 } else { 1852 MPASS(args->flags & IPFW_ARGS_OUT); 1853 iif = mem ? NULL : m_rcvif(m); 1854 oif = args->ifp; 1855 } 1856 1857 /* 1858 * Now scan the rules, and parse microinstructions for each rule. 1859 * We have two nested loops and an inner switch. Sometimes we 1860 * need to break out of one or both loops, or re-enter one of 1861 * the loops with updated variables. Loop variables are: 1862 * 1863 * f_pos (outer loop) points to the current rule. 1864 * On output it points to the matching rule. 1865 * done (outer loop) is used as a flag to break the loop. 1866 * l (inner loop) residual length of current rule. 1867 * cmd points to the current microinstruction. 1868 * 1869 * We break the inner loop by setting l=0 and possibly 1870 * cmdlen=0 if we don't want to advance cmd. 1871 * We break the outer loop by setting done=1 1872 * We can restart the inner loop by setting l>0 and f_pos, f, cmd 1873 * as needed. 1874 */ 1875 for (; f_pos < chain->n_rules; f_pos++) { 1876 ipfw_insn *cmd; 1877 uint32_t tablearg = 0; 1878 int l, cmdlen, skip_or; /* skip rest of OR block */ 1879 struct ip_fw *f; 1880 1881 f = chain->map[f_pos]; 1882 if (V_set_disable & (1 << f->set) ) 1883 continue; 1884 1885 skip_or = 0; 1886 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1887 l -= cmdlen, cmd += cmdlen) { 1888 int match; 1889 1890 /* 1891 * check_body is a jump target used when we find a 1892 * CHECK_STATE, and need to jump to the body of 1893 * the target rule. 1894 */ 1895 1896 /* check_body: */ 1897 cmdlen = F_LEN(cmd); 1898 /* 1899 * An OR block (insn_1 || .. || insn_n) has the 1900 * F_OR bit set in all but the last instruction. 1901 * The first match will set "skip_or", and cause 1902 * the following instructions to be skipped until 1903 * past the one with the F_OR bit clear. 1904 */ 1905 if (skip_or) { /* skip this instruction */ 1906 if ((cmd->len & F_OR) == 0) 1907 skip_or = 0; /* next one is good */ 1908 continue; 1909 } 1910 match = 0; /* set to 1 if we succeed */ 1911 1912 switch (cmd->opcode) { 1913 /* 1914 * The first set of opcodes compares the packet's 1915 * fields with some pattern, setting 'match' if a 1916 * match is found. At the end of the loop there is 1917 * logic to deal with F_NOT and F_OR flags associated 1918 * with the opcode. 1919 */ 1920 case O_NOP: 1921 match = 1; 1922 break; 1923 1924 case O_FORWARD_MAC: 1925 printf("ipfw: opcode %d unimplemented\n", 1926 cmd->opcode); 1927 break; 1928 1929 case O_GID: 1930 case O_UID: 1931 case O_JAIL: 1932 /* 1933 * We only check offset == 0 && proto != 0, 1934 * as this ensures that we have a 1935 * packet with the ports info. 1936 */ 1937 if (offset != 0) 1938 break; 1939 if (proto == IPPROTO_TCP || 1940 proto == IPPROTO_UDP || 1941 proto == IPPROTO_UDPLITE) 1942 match = check_uidgid( 1943 (ipfw_insn_u32 *)cmd, 1944 args, &ucred_lookup, 1945 #ifdef __FreeBSD__ 1946 &ucred_cache); 1947 #else 1948 (void *)&ucred_cache); 1949 #endif 1950 break; 1951 1952 case O_RECV: 1953 match = iface_match(iif, (ipfw_insn_if *)cmd, 1954 chain, &tablearg); 1955 break; 1956 1957 case O_XMIT: 1958 match = iface_match(oif, (ipfw_insn_if *)cmd, 1959 chain, &tablearg); 1960 break; 1961 1962 case O_VIA: 1963 match = iface_match(args->ifp, 1964 (ipfw_insn_if *)cmd, chain, &tablearg); 1965 break; 1966 1967 case O_MACADDR2: 1968 if (args->flags & IPFW_ARGS_ETHER) { 1969 u_int32_t *want = (u_int32_t *) 1970 ((ipfw_insn_mac *)cmd)->addr; 1971 u_int32_t *mask = (u_int32_t *) 1972 ((ipfw_insn_mac *)cmd)->mask; 1973 u_int32_t *hdr = (u_int32_t *)eh; 1974 1975 match = 1976 ( want[0] == (hdr[0] & mask[0]) && 1977 want[1] == (hdr[1] & mask[1]) && 1978 want[2] == (hdr[2] & mask[2]) ); 1979 } 1980 break; 1981 1982 case O_MAC_TYPE: 1983 if (args->flags & IPFW_ARGS_ETHER) { 1984 u_int16_t *p = 1985 ((ipfw_insn_u16 *)cmd)->ports; 1986 int i; 1987 1988 for (i = cmdlen - 1; !match && i>0; 1989 i--, p += 2) 1990 match = 1991 (ntohs(eh->ether_type) >= 1992 p[0] && 1993 ntohs(eh->ether_type) <= 1994 p[1]); 1995 } 1996 break; 1997 1998 case O_FRAG: 1999 if (is_ipv4) { 2000 /* 2001 * Since flags_match() works with 2002 * uint8_t we pack ip_off into 8 bits. 2003 * For this match offset is a boolean. 2004 */ 2005 match = flags_match(cmd, 2006 ((ntohs(ip->ip_off) & ~IP_OFFMASK) 2007 >> 8) | (offset != 0)); 2008 } else { 2009 /* 2010 * Compatibility: historically bare 2011 * "frag" would match IPv6 fragments. 2012 */ 2013 match = (cmd->arg1 == 0x1 && 2014 (offset != 0)); 2015 } 2016 break; 2017 2018 case O_IN: /* "out" is "not in" */ 2019 match = (oif == NULL); 2020 break; 2021 2022 case O_LAYER2: 2023 match = (args->flags & IPFW_ARGS_ETHER); 2024 break; 2025 2026 case O_DIVERTED: 2027 if ((args->flags & IPFW_ARGS_REF) == 0) 2028 break; 2029 /* 2030 * For diverted packets, args->rule.info 2031 * contains the divert port (in host format) 2032 * reason and direction. 2033 */ 2034 match = ((args->rule.info & IPFW_IS_MASK) == 2035 IPFW_IS_DIVERT) && ( 2036 ((args->rule.info & IPFW_INFO_IN) ? 2037 1: 2) & cmd->arg1); 2038 break; 2039 2040 case O_PROTO: 2041 /* 2042 * We do not allow an arg of 0 so the 2043 * check of "proto" only suffices. 2044 */ 2045 match = (proto == cmd->arg1); 2046 break; 2047 2048 case O_IP_SRC: 2049 match = is_ipv4 && 2050 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2051 src_ip.s_addr); 2052 break; 2053 2054 case O_IP_DST_LOOKUP: 2055 { 2056 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) { 2057 void *pkey; 2058 uint32_t vidx, key; 2059 uint16_t keylen = 0; /* zero if can't match the packet */ 2060 2061 /* Determine lookup key type */ 2062 vidx = ((ipfw_insn_u32 *)cmd)->d[1]; 2063 switch (vidx) { 2064 case LOOKUP_DST_IP: 2065 case LOOKUP_SRC_IP: 2066 /* Need IP frame */ 2067 if (is_ipv6 == 0 && is_ipv4 == 0) 2068 break; 2069 if (vidx == LOOKUP_DST_IP) 2070 pkey = is_ipv6 ? 2071 (void *)&args->f_id.dst_ip6: 2072 (void *)&dst_ip; 2073 else 2074 pkey = is_ipv6 ? 2075 (void *)&args->f_id.src_ip6: 2076 (void *)&src_ip; 2077 keylen = is_ipv6 ? 2078 sizeof(struct in6_addr): 2079 sizeof(in_addr_t); 2080 break; 2081 case LOOKUP_DST_PORT: 2082 case LOOKUP_SRC_PORT: 2083 /* Need IP frame */ 2084 if (is_ipv6 == 0 && is_ipv4 == 0) 2085 break; 2086 /* Skip fragments */ 2087 if (offset != 0) 2088 break; 2089 /* Skip proto without ports */ 2090 if (proto != IPPROTO_TCP && 2091 proto != IPPROTO_UDP && 2092 proto != IPPROTO_UDPLITE && 2093 proto != IPPROTO_SCTP) 2094 break; 2095 key = vidx == LOOKUP_DST_PORT ? 2096 dst_port: 2097 src_port; 2098 pkey = &key; 2099 keylen = sizeof(key); 2100 break; 2101 case LOOKUP_UID: 2102 case LOOKUP_JAIL: 2103 check_uidgid( 2104 (ipfw_insn_u32 *)cmd, 2105 args, &ucred_lookup, 2106 &ucred_cache); 2107 key = vidx == LOOKUP_UID ? 2108 ucred_cache->cr_uid: 2109 ucred_cache->cr_prison->pr_id; 2110 pkey = &key; 2111 keylen = sizeof(key); 2112 break; 2113 case LOOKUP_DSCP: 2114 /* Need IP frame */ 2115 if (is_ipv6 == 0 && is_ipv4 == 0) 2116 break; 2117 if (is_ipv6) 2118 key = IPV6_DSCP( 2119 (struct ip6_hdr *)ip) >> 2; 2120 else 2121 key = ip->ip_tos >> 2; 2122 pkey = &key; 2123 keylen = sizeof(key); 2124 break; 2125 case LOOKUP_DST_MAC: 2126 case LOOKUP_SRC_MAC: 2127 /* Need ether frame */ 2128 if ((args->flags & IPFW_ARGS_ETHER) == 0) 2129 break; 2130 pkey = vidx == LOOKUP_DST_MAC ? 2131 eh->ether_dhost: 2132 eh->ether_shost; 2133 keylen = ETHER_ADDR_LEN; 2134 break; 2135 case LOOKUP_MARK: 2136 key = args->rule.pkt_mark; 2137 pkey = &key; 2138 keylen = sizeof(key); 2139 break; 2140 } 2141 if (keylen == 0) 2142 break; 2143 match = ipfw_lookup_table(chain, 2144 cmd->arg1, keylen, pkey, &vidx); 2145 if (!match) 2146 break; 2147 tablearg = vidx; 2148 break; 2149 } 2150 /* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */ 2151 /* FALLTHROUGH */ 2152 } 2153 case O_IP_SRC_LOOKUP: 2154 { 2155 void *pkey; 2156 uint32_t vidx; 2157 uint16_t keylen; 2158 2159 if (is_ipv4) { 2160 keylen = sizeof(in_addr_t); 2161 if (cmd->opcode == O_IP_DST_LOOKUP) 2162 pkey = &dst_ip; 2163 else 2164 pkey = &src_ip; 2165 } else if (is_ipv6) { 2166 keylen = sizeof(struct in6_addr); 2167 if (cmd->opcode == O_IP_DST_LOOKUP) 2168 pkey = &args->f_id.dst_ip6; 2169 else 2170 pkey = &args->f_id.src_ip6; 2171 } else 2172 break; 2173 match = ipfw_lookup_table(chain, cmd->arg1, 2174 keylen, pkey, &vidx); 2175 if (!match) 2176 break; 2177 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) { 2178 match = ((ipfw_insn_u32 *)cmd)->d[0] == 2179 TARG_VAL(chain, vidx, tag); 2180 if (!match) 2181 break; 2182 } 2183 tablearg = vidx; 2184 break; 2185 } 2186 2187 case O_MAC_SRC_LOOKUP: 2188 case O_MAC_DST_LOOKUP: 2189 { 2190 void *pkey; 2191 uint32_t vidx; 2192 uint16_t keylen = ETHER_ADDR_LEN; 2193 2194 /* Need ether frame */ 2195 if ((args->flags & IPFW_ARGS_ETHER) == 0) 2196 break; 2197 2198 if (cmd->opcode == O_MAC_DST_LOOKUP) 2199 pkey = eh->ether_dhost; 2200 else 2201 pkey = eh->ether_shost; 2202 2203 match = ipfw_lookup_table(chain, cmd->arg1, 2204 keylen, pkey, &vidx); 2205 if (!match) 2206 break; 2207 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) { 2208 match = ((ipfw_insn_u32 *)cmd)->d[0] == 2209 TARG_VAL(chain, vidx, tag); 2210 if (!match) 2211 break; 2212 } 2213 tablearg = vidx; 2214 break; 2215 } 2216 2217 case O_IP_FLOW_LOOKUP: 2218 { 2219 uint32_t v = 0; 2220 match = ipfw_lookup_table(chain, 2221 cmd->arg1, 0, &args->f_id, &v); 2222 if (!match) 2223 break; 2224 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2225 match = ((ipfw_insn_u32 *)cmd)->d[0] == 2226 TARG_VAL(chain, v, tag); 2227 if (match) 2228 tablearg = v; 2229 } 2230 break; 2231 case O_IP_SRC_MASK: 2232 case O_IP_DST_MASK: 2233 if (is_ipv4) { 2234 uint32_t a = 2235 (cmd->opcode == O_IP_DST_MASK) ? 2236 dst_ip.s_addr : src_ip.s_addr; 2237 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2238 int i = cmdlen-1; 2239 2240 for (; !match && i>0; i-= 2, p+= 2) 2241 match = (p[0] == (a & p[1])); 2242 } 2243 break; 2244 2245 case O_IP_SRC_ME: 2246 if (is_ipv4) { 2247 match = in_localip(src_ip); 2248 break; 2249 } 2250 #ifdef INET6 2251 /* FALLTHROUGH */ 2252 case O_IP6_SRC_ME: 2253 match = is_ipv6 && 2254 ipfw_localip6(&args->f_id.src_ip6); 2255 #endif 2256 break; 2257 2258 case O_IP_DST_SET: 2259 case O_IP_SRC_SET: 2260 if (is_ipv4) { 2261 u_int32_t *d = (u_int32_t *)(cmd+1); 2262 u_int32_t addr = 2263 cmd->opcode == O_IP_DST_SET ? 2264 args->f_id.dst_ip : 2265 args->f_id.src_ip; 2266 2267 if (addr < d[0]) 2268 break; 2269 addr -= d[0]; /* subtract base */ 2270 match = (addr < cmd->arg1) && 2271 ( d[ 1 + (addr>>5)] & 2272 (1<<(addr & 0x1f)) ); 2273 } 2274 break; 2275 2276 case O_IP_DST: 2277 match = is_ipv4 && 2278 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2279 dst_ip.s_addr); 2280 break; 2281 2282 case O_IP_DST_ME: 2283 if (is_ipv4) { 2284 match = in_localip(dst_ip); 2285 break; 2286 } 2287 #ifdef INET6 2288 /* FALLTHROUGH */ 2289 case O_IP6_DST_ME: 2290 match = is_ipv6 && 2291 ipfw_localip6(&args->f_id.dst_ip6); 2292 #endif 2293 break; 2294 2295 case O_IP_SRCPORT: 2296 case O_IP_DSTPORT: 2297 /* 2298 * offset == 0 && proto != 0 is enough 2299 * to guarantee that we have a 2300 * packet with port info. 2301 */ 2302 if ((proto == IPPROTO_UDP || 2303 proto == IPPROTO_UDPLITE || 2304 proto == IPPROTO_TCP || 2305 proto == IPPROTO_SCTP) && offset == 0) { 2306 u_int16_t x = 2307 (cmd->opcode == O_IP_SRCPORT) ? 2308 src_port : dst_port ; 2309 u_int16_t *p = 2310 ((ipfw_insn_u16 *)cmd)->ports; 2311 int i; 2312 2313 for (i = cmdlen - 1; !match && i>0; 2314 i--, p += 2) 2315 match = (x>=p[0] && x<=p[1]); 2316 } 2317 break; 2318 2319 case O_ICMPTYPE: 2320 match = (offset == 0 && proto==IPPROTO_ICMP && 2321 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2322 break; 2323 2324 #ifdef INET6 2325 case O_ICMP6TYPE: 2326 match = is_ipv6 && offset == 0 && 2327 proto==IPPROTO_ICMPV6 && 2328 icmp6type_match( 2329 ICMP6(ulp)->icmp6_type, 2330 (ipfw_insn_u32 *)cmd); 2331 break; 2332 #endif /* INET6 */ 2333 2334 case O_IPOPT: 2335 match = (is_ipv4 && 2336 ipopts_match(ip, cmd) ); 2337 break; 2338 2339 case O_IPVER: 2340 match = ((is_ipv4 || is_ipv6) && 2341 cmd->arg1 == ip->ip_v); 2342 break; 2343 2344 case O_IPID: 2345 case O_IPTTL: 2346 if (!is_ipv4) 2347 break; 2348 case O_IPLEN: 2349 { /* only for IP packets */ 2350 uint16_t x; 2351 uint16_t *p; 2352 int i; 2353 2354 if (cmd->opcode == O_IPLEN) 2355 x = iplen; 2356 else if (cmd->opcode == O_IPTTL) 2357 x = ip->ip_ttl; 2358 else /* must be IPID */ 2359 x = ntohs(ip->ip_id); 2360 if (cmdlen == 1) { 2361 match = (cmd->arg1 == x); 2362 break; 2363 } 2364 /* otherwise we have ranges */ 2365 p = ((ipfw_insn_u16 *)cmd)->ports; 2366 i = cmdlen - 1; 2367 for (; !match && i>0; i--, p += 2) 2368 match = (x >= p[0] && x <= p[1]); 2369 } 2370 break; 2371 2372 case O_IPPRECEDENCE: 2373 match = (is_ipv4 && 2374 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2375 break; 2376 2377 case O_IPTOS: 2378 match = (is_ipv4 && 2379 flags_match(cmd, ip->ip_tos)); 2380 break; 2381 2382 case O_DSCP: 2383 { 2384 uint32_t *p; 2385 uint16_t x; 2386 2387 p = ((ipfw_insn_u32 *)cmd)->d; 2388 2389 if (is_ipv4) 2390 x = ip->ip_tos >> 2; 2391 else if (is_ipv6) { 2392 x = IPV6_DSCP( 2393 (struct ip6_hdr *)ip) >> 2; 2394 x &= 0x3f; 2395 } else 2396 break; 2397 2398 /* DSCP bitmask is stored as low_u32 high_u32 */ 2399 if (x >= 32) 2400 match = *(p + 1) & (1 << (x - 32)); 2401 else 2402 match = *p & (1 << x); 2403 } 2404 break; 2405 2406 case O_TCPDATALEN: 2407 if (proto == IPPROTO_TCP && offset == 0) { 2408 struct tcphdr *tcp; 2409 uint16_t x; 2410 uint16_t *p; 2411 int i; 2412 #ifdef INET6 2413 if (is_ipv6) { 2414 struct ip6_hdr *ip6; 2415 2416 ip6 = (struct ip6_hdr *)ip; 2417 if (ip6->ip6_plen == 0) { 2418 /* 2419 * Jumbo payload is not 2420 * supported by this 2421 * opcode. 2422 */ 2423 break; 2424 } 2425 x = iplen - hlen; 2426 } else 2427 #endif /* INET6 */ 2428 x = iplen - (ip->ip_hl << 2); 2429 tcp = TCP(ulp); 2430 x -= tcp->th_off << 2; 2431 if (cmdlen == 1) { 2432 match = (cmd->arg1 == x); 2433 break; 2434 } 2435 /* otherwise we have ranges */ 2436 p = ((ipfw_insn_u16 *)cmd)->ports; 2437 i = cmdlen - 1; 2438 for (; !match && i>0; i--, p += 2) 2439 match = (x >= p[0] && x <= p[1]); 2440 } 2441 break; 2442 2443 case O_TCPFLAGS: 2444 match = (proto == IPPROTO_TCP && offset == 0 && 2445 flags_match(cmd, TCP(ulp)->th_flags)); 2446 break; 2447 2448 case O_TCPOPTS: 2449 if (proto == IPPROTO_TCP && offset == 0 && ulp){ 2450 PULLUP_LEN_LOCKED(hlen, ulp, 2451 (TCP(ulp)->th_off << 2)); 2452 match = tcpopts_match(TCP(ulp), cmd); 2453 } 2454 break; 2455 2456 case O_TCPSEQ: 2457 match = (proto == IPPROTO_TCP && offset == 0 && 2458 ((ipfw_insn_u32 *)cmd)->d[0] == 2459 TCP(ulp)->th_seq); 2460 break; 2461 2462 case O_TCPACK: 2463 match = (proto == IPPROTO_TCP && offset == 0 && 2464 ((ipfw_insn_u32 *)cmd)->d[0] == 2465 TCP(ulp)->th_ack); 2466 break; 2467 2468 case O_TCPMSS: 2469 if (proto == IPPROTO_TCP && 2470 (args->f_id._flags & TH_SYN) != 0 && 2471 ulp != NULL) { 2472 uint16_t mss, *p; 2473 int i; 2474 2475 PULLUP_LEN_LOCKED(hlen, ulp, 2476 (TCP(ulp)->th_off << 2)); 2477 if ((tcpopts_parse(TCP(ulp), &mss) & 2478 IP_FW_TCPOPT_MSS) == 0) 2479 break; 2480 if (cmdlen == 1) { 2481 match = (cmd->arg1 == mss); 2482 break; 2483 } 2484 /* Otherwise we have ranges. */ 2485 p = ((ipfw_insn_u16 *)cmd)->ports; 2486 i = cmdlen - 1; 2487 for (; !match && i > 0; i--, p += 2) 2488 match = (mss >= p[0] && 2489 mss <= p[1]); 2490 } 2491 break; 2492 2493 case O_TCPWIN: 2494 if (proto == IPPROTO_TCP && offset == 0) { 2495 uint16_t x; 2496 uint16_t *p; 2497 int i; 2498 2499 x = ntohs(TCP(ulp)->th_win); 2500 if (cmdlen == 1) { 2501 match = (cmd->arg1 == x); 2502 break; 2503 } 2504 /* Otherwise we have ranges. */ 2505 p = ((ipfw_insn_u16 *)cmd)->ports; 2506 i = cmdlen - 1; 2507 for (; !match && i > 0; i--, p += 2) 2508 match = (x >= p[0] && x <= p[1]); 2509 } 2510 break; 2511 2512 case O_ESTAB: 2513 /* reject packets which have SYN only */ 2514 /* XXX should i also check for TH_ACK ? */ 2515 match = (proto == IPPROTO_TCP && offset == 0 && 2516 (TCP(ulp)->th_flags & 2517 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2518 break; 2519 2520 case O_ALTQ: { 2521 struct pf_mtag *at; 2522 struct m_tag *mtag; 2523 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2524 2525 /* 2526 * ALTQ uses mbuf tags from another 2527 * packet filtering system - pf(4). 2528 * We allocate a tag in its format 2529 * and fill it in, pretending to be pf(4). 2530 */ 2531 match = 1; 2532 at = pf_find_mtag(m); 2533 if (at != NULL && at->qid != 0) 2534 break; 2535 mtag = m_tag_get(PACKET_TAG_PF, 2536 sizeof(struct pf_mtag), M_NOWAIT | M_ZERO); 2537 if (mtag == NULL) { 2538 /* 2539 * Let the packet fall back to the 2540 * default ALTQ. 2541 */ 2542 break; 2543 } 2544 m_tag_prepend(m, mtag); 2545 at = (struct pf_mtag *)(mtag + 1); 2546 at->qid = altq->qid; 2547 at->hdr = ip; 2548 break; 2549 } 2550 2551 case O_LOG: 2552 ipfw_log(chain, f, hlen, args, 2553 offset | ip6f_mf, tablearg, ip); 2554 match = 1; 2555 break; 2556 2557 case O_PROB: 2558 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2559 break; 2560 2561 case O_VERREVPATH: 2562 /* Outgoing packets automatically pass/match */ 2563 match = (args->flags & IPFW_ARGS_OUT || 2564 ( 2565 #ifdef INET6 2566 is_ipv6 ? 2567 verify_path6(&(args->f_id.src_ip6), 2568 iif, args->f_id.fib) : 2569 #endif 2570 verify_path(src_ip, iif, args->f_id.fib))); 2571 break; 2572 2573 case O_VERSRCREACH: 2574 /* Outgoing packets automatically pass/match */ 2575 match = (hlen > 0 && ((oif != NULL) || ( 2576 #ifdef INET6 2577 is_ipv6 ? 2578 verify_path6(&(args->f_id.src_ip6), 2579 NULL, args->f_id.fib) : 2580 #endif 2581 verify_path(src_ip, NULL, args->f_id.fib)))); 2582 break; 2583 2584 case O_ANTISPOOF: 2585 /* Outgoing packets automatically pass/match */ 2586 if (oif == NULL && hlen > 0 && 2587 ( (is_ipv4 && in_localaddr(src_ip)) 2588 #ifdef INET6 2589 || (is_ipv6 && 2590 in6_localaddr(&(args->f_id.src_ip6))) 2591 #endif 2592 )) 2593 match = 2594 #ifdef INET6 2595 is_ipv6 ? verify_path6( 2596 &(args->f_id.src_ip6), iif, 2597 args->f_id.fib) : 2598 #endif 2599 verify_path(src_ip, iif, 2600 args->f_id.fib); 2601 else 2602 match = 1; 2603 break; 2604 2605 case O_IPSEC: 2606 match = (m_tag_find(m, 2607 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2608 /* otherwise no match */ 2609 break; 2610 2611 #ifdef INET6 2612 case O_IP6_SRC: 2613 match = is_ipv6 && 2614 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2615 &((ipfw_insn_ip6 *)cmd)->addr6); 2616 break; 2617 2618 case O_IP6_DST: 2619 match = is_ipv6 && 2620 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2621 &((ipfw_insn_ip6 *)cmd)->addr6); 2622 break; 2623 case O_IP6_SRC_MASK: 2624 case O_IP6_DST_MASK: 2625 if (is_ipv6) { 2626 int i = cmdlen - 1; 2627 struct in6_addr p; 2628 struct in6_addr *d = 2629 &((ipfw_insn_ip6 *)cmd)->addr6; 2630 2631 for (; !match && i > 0; d += 2, 2632 i -= F_INSN_SIZE(struct in6_addr) 2633 * 2) { 2634 p = (cmd->opcode == 2635 O_IP6_SRC_MASK) ? 2636 args->f_id.src_ip6: 2637 args->f_id.dst_ip6; 2638 APPLY_MASK(&p, &d[1]); 2639 match = 2640 IN6_ARE_ADDR_EQUAL(&d[0], 2641 &p); 2642 } 2643 } 2644 break; 2645 2646 case O_FLOW6ID: 2647 match = is_ipv6 && 2648 flow6id_match(args->f_id.flow_id6, 2649 (ipfw_insn_u32 *) cmd); 2650 break; 2651 2652 case O_EXT_HDR: 2653 match = is_ipv6 && 2654 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2655 break; 2656 2657 case O_IP6: 2658 match = is_ipv6; 2659 break; 2660 #endif 2661 2662 case O_IP4: 2663 match = is_ipv4; 2664 break; 2665 2666 case O_TAG: { 2667 struct m_tag *mtag; 2668 uint32_t tag = TARG(cmd->arg1, tag); 2669 2670 /* Packet is already tagged with this tag? */ 2671 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 2672 2673 /* We have `untag' action when F_NOT flag is 2674 * present. And we must remove this mtag from 2675 * mbuf and reset `match' to zero (`match' will 2676 * be inversed later). 2677 * Otherwise we should allocate new mtag and 2678 * push it into mbuf. 2679 */ 2680 if (cmd->len & F_NOT) { /* `untag' action */ 2681 if (mtag != NULL) 2682 m_tag_delete(m, mtag); 2683 match = 0; 2684 } else { 2685 if (mtag == NULL) { 2686 mtag = m_tag_alloc( MTAG_IPFW, 2687 tag, 0, M_NOWAIT); 2688 if (mtag != NULL) 2689 m_tag_prepend(m, mtag); 2690 } 2691 match = 1; 2692 } 2693 break; 2694 } 2695 2696 case O_FIB: /* try match the specified fib */ 2697 if (args->f_id.fib == cmd->arg1) 2698 match = 1; 2699 break; 2700 2701 case O_SOCKARG: { 2702 #ifndef USERSPACE /* not supported in userspace */ 2703 struct inpcb *inp = args->inp; 2704 struct inpcbinfo *pi; 2705 bool inp_locked = false; 2706 2707 if (proto == IPPROTO_TCP) 2708 pi = &V_tcbinfo; 2709 else if (proto == IPPROTO_UDP) 2710 pi = &V_udbinfo; 2711 else if (proto == IPPROTO_UDPLITE) 2712 pi = &V_ulitecbinfo; 2713 else 2714 break; 2715 2716 /* 2717 * XXXRW: so_user_cookie should almost 2718 * certainly be inp_user_cookie? 2719 */ 2720 2721 /* 2722 * For incoming packet lookup the inpcb 2723 * using the src/dest ip/port tuple. 2724 */ 2725 if (is_ipv4 && inp == NULL) { 2726 inp = in_pcblookup(pi, 2727 src_ip, htons(src_port), 2728 dst_ip, htons(dst_port), 2729 INPLOOKUP_RLOCKPCB, NULL); 2730 inp_locked = true; 2731 } 2732 #ifdef INET6 2733 if (is_ipv6 && inp == NULL) { 2734 inp = in6_pcblookup(pi, 2735 &args->f_id.src_ip6, 2736 htons(src_port), 2737 &args->f_id.dst_ip6, 2738 htons(dst_port), 2739 INPLOOKUP_RLOCKPCB, NULL); 2740 inp_locked = true; 2741 } 2742 #endif /* INET6 */ 2743 if (inp != NULL) { 2744 if (inp->inp_socket) { 2745 tablearg = 2746 inp->inp_socket->so_user_cookie; 2747 if (tablearg) 2748 match = 1; 2749 } 2750 if (inp_locked) 2751 INP_RUNLOCK(inp); 2752 } 2753 #endif /* !USERSPACE */ 2754 break; 2755 } 2756 2757 case O_TAGGED: { 2758 struct m_tag *mtag; 2759 uint32_t tag = TARG(cmd->arg1, tag); 2760 2761 if (cmdlen == 1) { 2762 match = m_tag_locate(m, MTAG_IPFW, 2763 tag, NULL) != NULL; 2764 break; 2765 } 2766 2767 /* we have ranges */ 2768 for (mtag = m_tag_first(m); 2769 mtag != NULL && !match; 2770 mtag = m_tag_next(m, mtag)) { 2771 uint16_t *p; 2772 int i; 2773 2774 if (mtag->m_tag_cookie != MTAG_IPFW) 2775 continue; 2776 2777 p = ((ipfw_insn_u16 *)cmd)->ports; 2778 i = cmdlen - 1; 2779 for(; !match && i > 0; i--, p += 2) 2780 match = 2781 mtag->m_tag_id >= p[0] && 2782 mtag->m_tag_id <= p[1]; 2783 } 2784 break; 2785 } 2786 2787 case O_MARK: { 2788 uint32_t mark; 2789 if (cmd->arg1 == IP_FW_TARG) 2790 mark = TARG_VAL(chain, tablearg, mark); 2791 else 2792 mark = ((ipfw_insn_u32 *)cmd)->d[0]; 2793 match = 2794 (args->rule.pkt_mark & 2795 ((ipfw_insn_u32 *)cmd)->d[1]) == 2796 (mark & ((ipfw_insn_u32 *)cmd)->d[1]); 2797 break; 2798 } 2799 2800 /* 2801 * The second set of opcodes represents 'actions', 2802 * i.e. the terminal part of a rule once the packet 2803 * matches all previous patterns. 2804 * Typically there is only one action for each rule, 2805 * and the opcode is stored at the end of the rule 2806 * (but there are exceptions -- see below). 2807 * 2808 * In general, here we set retval and terminate the 2809 * outer loop (would be a 'break 3' in some language, 2810 * but we need to set l=0, done=1) 2811 * 2812 * Exceptions: 2813 * O_COUNT and O_SKIPTO actions: 2814 * instead of terminating, we jump to the next rule 2815 * (setting l=0), or to the SKIPTO target (setting 2816 * f/f_len, cmd and l as needed), respectively. 2817 * 2818 * O_TAG, O_LOG and O_ALTQ action parameters: 2819 * perform some action and set match = 1; 2820 * 2821 * O_LIMIT and O_KEEP_STATE: these opcodes are 2822 * not real 'actions', and are stored right 2823 * before the 'action' part of the rule (one 2824 * exception is O_SKIP_ACTION which could be 2825 * between these opcodes and 'action' one). 2826 * These opcodes try to install an entry in the 2827 * state tables; if successful, we continue with 2828 * the next opcode (match=1; break;), otherwise 2829 * the packet must be dropped (set retval, 2830 * break loops with l=0, done=1) 2831 * 2832 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2833 * cause a lookup of the state table, and a jump 2834 * to the 'action' part of the parent rule 2835 * if an entry is found, or 2836 * (CHECK_STATE only) a jump to the next rule if 2837 * the entry is not found. 2838 * The result of the lookup is cached so that 2839 * further instances of these opcodes become NOPs. 2840 * The jump to the next rule is done by setting 2841 * l=0, cmdlen=0. 2842 * 2843 * O_SKIP_ACTION: this opcode is not a real 'action' 2844 * either, and is stored right before the 'action' 2845 * part of the rule, right after the O_KEEP_STATE 2846 * opcode. It causes match failure so the real 2847 * 'action' could be executed only if the rule 2848 * is checked via dynamic rule from the state 2849 * table, as in such case execution starts 2850 * from the true 'action' opcode directly. 2851 * 2852 */ 2853 case O_LIMIT: 2854 case O_KEEP_STATE: 2855 if (ipfw_dyn_install_state(chain, f, 2856 (ipfw_insn_limit *)cmd, args, ulp, 2857 pktlen, &dyn_info, tablearg)) { 2858 /* error or limit violation */ 2859 retval = IP_FW_DENY; 2860 l = 0; /* exit inner loop */ 2861 done = 1; /* exit outer loop */ 2862 } 2863 match = 1; 2864 break; 2865 2866 case O_PROBE_STATE: 2867 case O_CHECK_STATE: 2868 /* 2869 * dynamic rules are checked at the first 2870 * keep-state or check-state occurrence, 2871 * with the result being stored in dyn_info. 2872 * The compiler introduces a PROBE_STATE 2873 * instruction for us when we have a 2874 * KEEP_STATE (because PROBE_STATE needs 2875 * to be run first). 2876 */ 2877 if (DYN_LOOKUP_NEEDED(&dyn_info, cmd) && 2878 (q = ipfw_dyn_lookup_state(args, ulp, 2879 pktlen, cmd, &dyn_info)) != NULL) { 2880 /* 2881 * Found dynamic entry, jump to the 2882 * 'action' part of the parent rule 2883 * by setting f, cmd, l and clearing 2884 * cmdlen. 2885 */ 2886 f = q; 2887 f_pos = dyn_info.f_pos; 2888 cmd = ACTION_PTR(f); 2889 l = f->cmd_len - f->act_ofs; 2890 cmdlen = 0; 2891 match = 1; 2892 break; 2893 } 2894 /* 2895 * Dynamic entry not found. If CHECK_STATE, 2896 * skip to next rule, if PROBE_STATE just 2897 * ignore and continue with next opcode. 2898 */ 2899 if (cmd->opcode == O_CHECK_STATE) 2900 l = 0; /* exit inner loop */ 2901 match = 1; 2902 break; 2903 2904 case O_SKIP_ACTION: 2905 match = 0; /* skip to the next rule */ 2906 l = 0; /* exit inner loop */ 2907 break; 2908 2909 case O_ACCEPT: 2910 retval = 0; /* accept */ 2911 l = 0; /* exit inner loop */ 2912 done = 1; /* exit outer loop */ 2913 break; 2914 2915 case O_PIPE: 2916 case O_QUEUE: 2917 set_match(args, f_pos, chain); 2918 args->rule.info = TARG(cmd->arg1, pipe); 2919 if (cmd->opcode == O_PIPE) 2920 args->rule.info |= IPFW_IS_PIPE; 2921 if (V_fw_one_pass) 2922 args->rule.info |= IPFW_ONEPASS; 2923 retval = IP_FW_DUMMYNET; 2924 l = 0; /* exit inner loop */ 2925 done = 1; /* exit outer loop */ 2926 break; 2927 2928 case O_DIVERT: 2929 case O_TEE: 2930 if (args->flags & IPFW_ARGS_ETHER) 2931 break; /* not on layer 2 */ 2932 /* otherwise this is terminal */ 2933 l = 0; /* exit inner loop */ 2934 done = 1; /* exit outer loop */ 2935 retval = (cmd->opcode == O_DIVERT) ? 2936 IP_FW_DIVERT : IP_FW_TEE; 2937 set_match(args, f_pos, chain); 2938 args->rule.info = TARG(cmd->arg1, divert); 2939 break; 2940 2941 case O_COUNT: 2942 IPFW_INC_RULE_COUNTER(f, pktlen); 2943 l = 0; /* exit inner loop */ 2944 break; 2945 2946 case O_SKIPTO: 2947 IPFW_INC_RULE_COUNTER(f, pktlen); 2948 f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0); 2949 /* 2950 * Skip disabled rules, and re-enter 2951 * the inner loop with the correct 2952 * f_pos, f, l and cmd. 2953 * Also clear cmdlen and skip_or 2954 */ 2955 for (; f_pos < chain->n_rules - 1 && 2956 (V_set_disable & 2957 (1 << chain->map[f_pos]->set)); 2958 f_pos++) 2959 ; 2960 /* Re-enter the inner loop at the skipto rule. */ 2961 f = chain->map[f_pos]; 2962 l = f->cmd_len; 2963 cmd = f->cmd; 2964 match = 1; 2965 cmdlen = 0; 2966 skip_or = 0; 2967 continue; 2968 break; /* not reached */ 2969 2970 case O_CALLRETURN: { 2971 /* 2972 * Implementation of `subroutine' call/return, 2973 * in the stack carried in an mbuf tag. This 2974 * is different from `skipto' in that any call 2975 * address is possible (`skipto' must prevent 2976 * backward jumps to avoid endless loops). 2977 * We have `return' action when F_NOT flag is 2978 * present. The `m_tag_id' field is used as 2979 * stack pointer. 2980 */ 2981 struct m_tag *mtag; 2982 uint16_t jmpto, *stack; 2983 2984 #define IS_CALL ((cmd->len & F_NOT) == 0) 2985 #define IS_RETURN ((cmd->len & F_NOT) != 0) 2986 /* 2987 * Hand-rolled version of m_tag_locate() with 2988 * wildcard `type'. 2989 * If not already tagged, allocate new tag. 2990 */ 2991 mtag = m_tag_first(m); 2992 while (mtag != NULL) { 2993 if (mtag->m_tag_cookie == 2994 MTAG_IPFW_CALL) 2995 break; 2996 mtag = m_tag_next(m, mtag); 2997 } 2998 if (mtag == NULL && IS_CALL) { 2999 mtag = m_tag_alloc(MTAG_IPFW_CALL, 0, 3000 IPFW_CALLSTACK_SIZE * 3001 sizeof(uint16_t), M_NOWAIT); 3002 if (mtag != NULL) 3003 m_tag_prepend(m, mtag); 3004 } 3005 3006 /* 3007 * On error both `call' and `return' just 3008 * continue with next rule. 3009 */ 3010 if (IS_RETURN && (mtag == NULL || 3011 mtag->m_tag_id == 0)) { 3012 l = 0; /* exit inner loop */ 3013 break; 3014 } 3015 if (IS_CALL && (mtag == NULL || 3016 mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) { 3017 printf("ipfw: call stack error, " 3018 "go to next rule\n"); 3019 l = 0; /* exit inner loop */ 3020 break; 3021 } 3022 3023 IPFW_INC_RULE_COUNTER(f, pktlen); 3024 stack = (uint16_t *)(mtag + 1); 3025 3026 /* 3027 * The `call' action may use cached f_pos 3028 * (in f->next_rule), whose version is written 3029 * in f->next_rule. 3030 * The `return' action, however, doesn't have 3031 * fixed jump address in cmd->arg1 and can't use 3032 * cache. 3033 */ 3034 if (IS_CALL) { 3035 stack[mtag->m_tag_id] = f->rulenum; 3036 mtag->m_tag_id++; 3037 f_pos = JUMP(chain, f, cmd->arg1, 3038 tablearg, 1); 3039 } else { /* `return' action */ 3040 mtag->m_tag_id--; 3041 jmpto = stack[mtag->m_tag_id] + 1; 3042 f_pos = ipfw_find_rule(chain, jmpto, 0); 3043 } 3044 3045 /* 3046 * Skip disabled rules, and re-enter 3047 * the inner loop with the correct 3048 * f_pos, f, l and cmd. 3049 * Also clear cmdlen and skip_or 3050 */ 3051 for (; f_pos < chain->n_rules - 1 && 3052 (V_set_disable & 3053 (1 << chain->map[f_pos]->set)); f_pos++) 3054 ; 3055 /* Re-enter the inner loop at the dest rule. */ 3056 f = chain->map[f_pos]; 3057 l = f->cmd_len; 3058 cmd = f->cmd; 3059 cmdlen = 0; 3060 skip_or = 0; 3061 continue; 3062 break; /* NOTREACHED */ 3063 } 3064 #undef IS_CALL 3065 #undef IS_RETURN 3066 3067 case O_REJECT: 3068 /* 3069 * Drop the packet and send a reject notice 3070 * if the packet is not ICMP (or is an ICMP 3071 * query), and it is not multicast/broadcast. 3072 */ 3073 if (hlen > 0 && is_ipv4 && offset == 0 && 3074 (proto != IPPROTO_ICMP || 3075 is_icmp_query(ICMP(ulp))) && 3076 !(m->m_flags & (M_BCAST|M_MCAST)) && 3077 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3078 send_reject(args, cmd, iplen, ip); 3079 m = args->m; 3080 } 3081 /* FALLTHROUGH */ 3082 #ifdef INET6 3083 case O_UNREACH6: 3084 if (hlen > 0 && is_ipv6 && 3085 ((offset & IP6F_OFF_MASK) == 0) && 3086 (proto != IPPROTO_ICMPV6 || 3087 (is_icmp6_query(icmp6_type) == 1)) && 3088 !(m->m_flags & (M_BCAST|M_MCAST)) && 3089 !IN6_IS_ADDR_MULTICAST( 3090 &args->f_id.dst_ip6)) { 3091 send_reject6(args, 3092 cmd->opcode == O_REJECT ? 3093 map_icmp_unreach(cmd->arg1): 3094 cmd->arg1, hlen, 3095 (struct ip6_hdr *)ip); 3096 m = args->m; 3097 } 3098 /* FALLTHROUGH */ 3099 #endif 3100 case O_DENY: 3101 retval = IP_FW_DENY; 3102 l = 0; /* exit inner loop */ 3103 done = 1; /* exit outer loop */ 3104 break; 3105 3106 case O_FORWARD_IP: 3107 if (args->flags & IPFW_ARGS_ETHER) 3108 break; /* not valid on layer2 pkts */ 3109 if (q != f || 3110 dyn_info.direction == MATCH_FORWARD) { 3111 struct sockaddr_in *sa; 3112 3113 sa = &(((ipfw_insn_sa *)cmd)->sa); 3114 if (sa->sin_addr.s_addr == INADDR_ANY) { 3115 #ifdef INET6 3116 /* 3117 * We use O_FORWARD_IP opcode for 3118 * fwd rule with tablearg, but tables 3119 * now support IPv6 addresses. And 3120 * when we are inspecting IPv6 packet, 3121 * we can use nh6 field from 3122 * table_value as next_hop6 address. 3123 */ 3124 if (is_ipv6) { 3125 struct ip_fw_nh6 *nh6; 3126 3127 args->flags |= IPFW_ARGS_NH6; 3128 nh6 = &args->hopstore6; 3129 nh6->sin6_addr = TARG_VAL( 3130 chain, tablearg, nh6); 3131 nh6->sin6_port = sa->sin_port; 3132 nh6->sin6_scope_id = TARG_VAL( 3133 chain, tablearg, zoneid); 3134 } else 3135 #endif 3136 { 3137 args->flags |= IPFW_ARGS_NH4; 3138 args->hopstore.sin_port = 3139 sa->sin_port; 3140 sa = &args->hopstore; 3141 sa->sin_family = AF_INET; 3142 sa->sin_len = sizeof(*sa); 3143 sa->sin_addr.s_addr = htonl( 3144 TARG_VAL(chain, tablearg, 3145 nh4)); 3146 } 3147 } else { 3148 args->flags |= IPFW_ARGS_NH4PTR; 3149 args->next_hop = sa; 3150 } 3151 } 3152 retval = IP_FW_PASS; 3153 l = 0; /* exit inner loop */ 3154 done = 1; /* exit outer loop */ 3155 break; 3156 3157 #ifdef INET6 3158 case O_FORWARD_IP6: 3159 if (args->flags & IPFW_ARGS_ETHER) 3160 break; /* not valid on layer2 pkts */ 3161 if (q != f || 3162 dyn_info.direction == MATCH_FORWARD) { 3163 struct sockaddr_in6 *sin6; 3164 3165 sin6 = &(((ipfw_insn_sa6 *)cmd)->sa); 3166 args->flags |= IPFW_ARGS_NH6PTR; 3167 args->next_hop6 = sin6; 3168 } 3169 retval = IP_FW_PASS; 3170 l = 0; /* exit inner loop */ 3171 done = 1; /* exit outer loop */ 3172 break; 3173 #endif 3174 3175 case O_NETGRAPH: 3176 case O_NGTEE: 3177 set_match(args, f_pos, chain); 3178 args->rule.info = TARG(cmd->arg1, netgraph); 3179 if (V_fw_one_pass) 3180 args->rule.info |= IPFW_ONEPASS; 3181 retval = (cmd->opcode == O_NETGRAPH) ? 3182 IP_FW_NETGRAPH : IP_FW_NGTEE; 3183 l = 0; /* exit inner loop */ 3184 done = 1; /* exit outer loop */ 3185 break; 3186 3187 case O_SETFIB: { 3188 uint32_t fib; 3189 3190 IPFW_INC_RULE_COUNTER(f, pktlen); 3191 fib = TARG(cmd->arg1, fib) & 0x7FFF; 3192 if (fib >= rt_numfibs) 3193 fib = 0; 3194 M_SETFIB(m, fib); 3195 args->f_id.fib = fib; /* XXX */ 3196 l = 0; /* exit inner loop */ 3197 break; 3198 } 3199 3200 case O_SETDSCP: { 3201 uint16_t code; 3202 3203 code = TARG(cmd->arg1, dscp) & 0x3F; 3204 l = 0; /* exit inner loop */ 3205 if (is_ipv4) { 3206 uint16_t old; 3207 3208 old = *(uint16_t *)ip; 3209 ip->ip_tos = (code << 2) | 3210 (ip->ip_tos & 0x03); 3211 ip->ip_sum = cksum_adjust(ip->ip_sum, 3212 old, *(uint16_t *)ip); 3213 } else if (is_ipv6) { 3214 /* update cached value */ 3215 args->f_id.flow_id6 = 3216 ntohl(*(uint32_t *)ip) & ~0x0FC00000; 3217 args->f_id.flow_id6 |= code << 22; 3218 3219 *((uint32_t *)ip) = 3220 htonl(args->f_id.flow_id6); 3221 } else 3222 break; 3223 3224 IPFW_INC_RULE_COUNTER(f, pktlen); 3225 break; 3226 } 3227 3228 case O_NAT: 3229 l = 0; /* exit inner loop */ 3230 done = 1; /* exit outer loop */ 3231 /* 3232 * Ensure that we do not invoke NAT handler for 3233 * non IPv4 packets. Libalias expects only IPv4. 3234 */ 3235 if (!is_ipv4 || !IPFW_NAT_LOADED) { 3236 retval = IP_FW_DENY; 3237 break; 3238 } 3239 3240 struct cfg_nat *t; 3241 int nat_id; 3242 3243 args->rule.info = 0; 3244 set_match(args, f_pos, chain); 3245 /* Check if this is 'global' nat rule */ 3246 if (cmd->arg1 == IP_FW_NAT44_GLOBAL) { 3247 retval = ipfw_nat_ptr(args, NULL, m); 3248 break; 3249 } 3250 t = ((ipfw_insn_nat *)cmd)->nat; 3251 if (t == NULL) { 3252 nat_id = TARG(cmd->arg1, nat); 3253 t = (*lookup_nat_ptr)(&chain->nat, nat_id); 3254 3255 if (t == NULL) { 3256 retval = IP_FW_DENY; 3257 break; 3258 } 3259 if (cmd->arg1 != IP_FW_TARG) 3260 ((ipfw_insn_nat *)cmd)->nat = t; 3261 } 3262 retval = ipfw_nat_ptr(args, t, m); 3263 break; 3264 3265 case O_REASS: { 3266 int ip_off; 3267 3268 l = 0; /* in any case exit inner loop */ 3269 if (is_ipv6) /* IPv6 is not supported yet */ 3270 break; 3271 IPFW_INC_RULE_COUNTER(f, pktlen); 3272 ip_off = ntohs(ip->ip_off); 3273 3274 /* if not fragmented, go to next rule */ 3275 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0) 3276 break; 3277 3278 args->m = m = ip_reass(m); 3279 3280 /* 3281 * do IP header checksum fixup. 3282 */ 3283 if (m == NULL) { /* fragment got swallowed */ 3284 retval = IP_FW_DENY; 3285 } else { /* good, packet complete */ 3286 int hlen; 3287 3288 ip = mtod(m, struct ip *); 3289 hlen = ip->ip_hl << 2; 3290 ip->ip_sum = 0; 3291 if (hlen == sizeof(struct ip)) 3292 ip->ip_sum = in_cksum_hdr(ip); 3293 else 3294 ip->ip_sum = in_cksum(m, hlen); 3295 retval = IP_FW_REASS; 3296 args->rule.info = 0; 3297 set_match(args, f_pos, chain); 3298 } 3299 done = 1; /* exit outer loop */ 3300 break; 3301 } 3302 3303 case O_SETMARK: { 3304 l = 0; /* exit inner loop */ 3305 args->rule.pkt_mark = ( 3306 (cmd->arg1 == IP_FW_TARG) ? 3307 TARG_VAL(chain, tablearg, mark) : 3308 ((ipfw_insn_u32 *)cmd)->d[0]); 3309 3310 IPFW_INC_RULE_COUNTER(f, pktlen); 3311 break; 3312 } 3313 3314 case O_EXTERNAL_ACTION: 3315 l = 0; /* in any case exit inner loop */ 3316 retval = ipfw_run_eaction(chain, args, 3317 cmd, &done); 3318 /* 3319 * If both @retval and @done are zero, 3320 * consider this as rule matching and 3321 * update counters. 3322 */ 3323 if (retval == 0 && done == 0) { 3324 IPFW_INC_RULE_COUNTER(f, pktlen); 3325 /* 3326 * Reset the result of the last 3327 * dynamic state lookup. 3328 * External action can change 3329 * @args content, and it may be 3330 * used for new state lookup later. 3331 */ 3332 DYN_INFO_INIT(&dyn_info); 3333 } 3334 break; 3335 3336 default: 3337 panic("-- unknown opcode %d\n", cmd->opcode); 3338 } /* end of switch() on opcodes */ 3339 /* 3340 * if we get here with l=0, then match is irrelevant. 3341 */ 3342 3343 if (cmd->len & F_NOT) 3344 match = !match; 3345 3346 if (match) { 3347 if (cmd->len & F_OR) 3348 skip_or = 1; 3349 } else { 3350 if (!(cmd->len & F_OR)) /* not an OR block, */ 3351 break; /* try next rule */ 3352 } 3353 3354 } /* end of inner loop, scan opcodes */ 3355 #undef PULLUP_LEN 3356 #undef PULLUP_LEN_LOCKED 3357 3358 if (done) 3359 break; 3360 3361 /* next_rule:; */ /* try next rule */ 3362 3363 } /* end of outer for, scan rules */ 3364 3365 if (done) { 3366 struct ip_fw *rule = chain->map[f_pos]; 3367 /* Update statistics */ 3368 IPFW_INC_RULE_COUNTER(rule, pktlen); 3369 IPFW_PROBE(rule__matched, retval, 3370 is_ipv4 ? AF_INET : AF_INET6, 3371 is_ipv4 ? (uintptr_t)&src_ip : 3372 (uintptr_t)&args->f_id.src_ip6, 3373 is_ipv4 ? (uintptr_t)&dst_ip : 3374 (uintptr_t)&args->f_id.dst_ip6, 3375 args, rule); 3376 } else { 3377 retval = IP_FW_DENY; 3378 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3379 } 3380 IPFW_PF_RUNLOCK(chain); 3381 #ifdef __FreeBSD__ 3382 if (ucred_cache != NULL) 3383 crfree(ucred_cache); 3384 #endif 3385 return (retval); 3386 3387 pullup_failed: 3388 if (V_fw_verbose) 3389 printf("ipfw: pullup failed\n"); 3390 return (IP_FW_DENY); 3391 } 3392 3393 /* 3394 * Set maximum number of tables that can be used in given VNET ipfw instance. 3395 */ 3396 #ifdef SYSCTL_NODE 3397 static int 3398 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS) 3399 { 3400 int error; 3401 unsigned int ntables; 3402 3403 ntables = V_fw_tables_max; 3404 3405 error = sysctl_handle_int(oidp, &ntables, 0, req); 3406 /* Read operation or some error */ 3407 if ((error != 0) || (req->newptr == NULL)) 3408 return (error); 3409 3410 return (ipfw_resize_tables(&V_layer3_chain, ntables)); 3411 } 3412 3413 /* 3414 * Switches table namespace between global and per-set. 3415 */ 3416 static int 3417 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS) 3418 { 3419 int error; 3420 unsigned int sets; 3421 3422 sets = V_fw_tables_sets; 3423 3424 error = sysctl_handle_int(oidp, &sets, 0, req); 3425 /* Read operation or some error */ 3426 if ((error != 0) || (req->newptr == NULL)) 3427 return (error); 3428 3429 return (ipfw_switch_tables_namespace(&V_layer3_chain, sets)); 3430 } 3431 #endif 3432 3433 /* 3434 * Module and VNET glue 3435 */ 3436 3437 /* 3438 * Stuff that must be initialised only on boot or module load 3439 */ 3440 static int 3441 ipfw_init(void) 3442 { 3443 int error = 0; 3444 3445 /* 3446 * Only print out this stuff the first time around, 3447 * when called from the sysinit code. 3448 */ 3449 printf("ipfw2 " 3450 #ifdef INET6 3451 "(+ipv6) " 3452 #endif 3453 "initialized, divert %s, nat %s, " 3454 "default to %s, logging ", 3455 #ifdef IPDIVERT 3456 "enabled", 3457 #else 3458 "loadable", 3459 #endif 3460 #ifdef IPFIREWALL_NAT 3461 "enabled", 3462 #else 3463 "loadable", 3464 #endif 3465 default_to_accept ? "accept" : "deny"); 3466 3467 /* 3468 * Note: V_xxx variables can be accessed here but the vnet specific 3469 * initializer may not have been called yet for the VIMAGE case. 3470 * Tuneables will have been processed. We will print out values for 3471 * the default vnet. 3472 * XXX This should all be rationalized AFTER 8.0 3473 */ 3474 if (V_fw_verbose == 0) 3475 printf("disabled\n"); 3476 else if (V_verbose_limit == 0) 3477 printf("unlimited\n"); 3478 else 3479 printf("limited to %d packets/entry by default\n", 3480 V_verbose_limit); 3481 3482 /* Check user-supplied table count for validness */ 3483 if (default_fw_tables > IPFW_TABLES_MAX) 3484 default_fw_tables = IPFW_TABLES_MAX; 3485 3486 ipfw_init_sopt_handler(); 3487 ipfw_init_obj_rewriter(); 3488 ipfw_iface_init(); 3489 return (error); 3490 } 3491 3492 /* 3493 * Called for the removal of the last instance only on module unload. 3494 */ 3495 static void 3496 ipfw_destroy(void) 3497 { 3498 3499 ipfw_iface_destroy(); 3500 ipfw_destroy_sopt_handler(); 3501 ipfw_destroy_obj_rewriter(); 3502 printf("IP firewall unloaded\n"); 3503 } 3504 3505 /* 3506 * Stuff that must be initialized for every instance 3507 * (including the first of course). 3508 */ 3509 static int 3510 vnet_ipfw_init(const void *unused) 3511 { 3512 int error, first; 3513 struct ip_fw *rule = NULL; 3514 struct ip_fw_chain *chain; 3515 3516 chain = &V_layer3_chain; 3517 3518 first = IS_DEFAULT_VNET(curvnet) ? 1 : 0; 3519 3520 /* First set up some values that are compile time options */ 3521 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 3522 V_fw_deny_unknown_exthdrs = 1; 3523 #ifdef IPFIREWALL_VERBOSE 3524 V_fw_verbose = 1; 3525 #endif 3526 #ifdef IPFIREWALL_VERBOSE_LIMIT 3527 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 3528 #endif 3529 #ifdef IPFIREWALL_NAT 3530 LIST_INIT(&chain->nat); 3531 #endif 3532 3533 /* Init shared services hash table */ 3534 ipfw_init_srv(chain); 3535 3536 ipfw_init_counters(); 3537 /* Set initial number of tables */ 3538 V_fw_tables_max = default_fw_tables; 3539 error = ipfw_init_tables(chain, first); 3540 if (error) { 3541 printf("ipfw2: setting up tables failed\n"); 3542 free(chain->map, M_IPFW); 3543 free(rule, M_IPFW); 3544 return (ENOSPC); 3545 } 3546 3547 IPFW_LOCK_INIT(chain); 3548 3549 /* fill and insert the default rule */ 3550 rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw)); 3551 rule->flags |= IPFW_RULE_NOOPT; 3552 rule->cmd_len = 1; 3553 rule->cmd[0].len = 1; 3554 rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; 3555 chain->default_rule = rule; 3556 ipfw_add_protected_rule(chain, rule, 0); 3557 3558 ipfw_dyn_init(chain); 3559 ipfw_eaction_init(chain, first); 3560 #ifdef LINEAR_SKIPTO 3561 ipfw_init_skipto_cache(chain); 3562 #endif 3563 ipfw_bpf_init(first); 3564 3565 /* First set up some values that are compile time options */ 3566 V_ipfw_vnet_ready = 1; /* Open for business */ 3567 3568 /* 3569 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6. 3570 * Even if the latter two fail we still keep the module alive 3571 * because the sockopt and layer2 paths are still useful. 3572 * ipfw[6]_hook return 0 on success, ENOENT on failure, 3573 * so we can ignore the exact return value and just set a flag. 3574 * 3575 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so 3576 * changes in the underlying (per-vnet) variables trigger 3577 * immediate hook()/unhook() calls. 3578 * In layer2 we have the same behaviour, except that V_ether_ipfw 3579 * is checked on each packet because there are no pfil hooks. 3580 */ 3581 V_ip_fw_ctl_ptr = ipfw_ctl3; 3582 error = ipfw_attach_hooks(); 3583 return (error); 3584 } 3585 3586 /* 3587 * Called for the removal of each instance. 3588 */ 3589 static int 3590 vnet_ipfw_uninit(const void *unused) 3591 { 3592 struct ip_fw *reap; 3593 struct ip_fw_chain *chain = &V_layer3_chain; 3594 int i, last; 3595 3596 V_ipfw_vnet_ready = 0; /* tell new callers to go away */ 3597 /* 3598 * disconnect from ipv4, ipv6, layer2 and sockopt. 3599 * Then grab, release and grab again the WLOCK so we make 3600 * sure the update is propagated and nobody will be in. 3601 */ 3602 ipfw_detach_hooks(); 3603 V_ip_fw_ctl_ptr = NULL; 3604 3605 last = IS_DEFAULT_VNET(curvnet) ? 1 : 0; 3606 3607 IPFW_UH_WLOCK(chain); 3608 IPFW_UH_WUNLOCK(chain); 3609 3610 ipfw_dyn_uninit(0); /* run the callout_drain */ 3611 3612 IPFW_UH_WLOCK(chain); 3613 3614 reap = NULL; 3615 IPFW_WLOCK(chain); 3616 for (i = 0; i < chain->n_rules; i++) 3617 ipfw_reap_add(chain, &reap, chain->map[i]); 3618 free(chain->map, M_IPFW); 3619 #ifdef LINEAR_SKIPTO 3620 ipfw_destroy_skipto_cache(chain); 3621 #endif 3622 IPFW_WUNLOCK(chain); 3623 IPFW_UH_WUNLOCK(chain); 3624 ipfw_destroy_tables(chain, last); 3625 ipfw_eaction_uninit(chain, last); 3626 if (reap != NULL) 3627 ipfw_reap_rules(reap); 3628 vnet_ipfw_iface_destroy(chain); 3629 ipfw_destroy_srv(chain); 3630 IPFW_LOCK_DESTROY(chain); 3631 ipfw_dyn_uninit(1); /* free the remaining parts */ 3632 ipfw_destroy_counters(); 3633 ipfw_bpf_uninit(last); 3634 return (0); 3635 } 3636 3637 /* 3638 * Module event handler. 3639 * In general we have the choice of handling most of these events by the 3640 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to 3641 * use the SYSINIT handlers as they are more capable of expressing the 3642 * flow of control during module and vnet operations, so this is just 3643 * a skeleton. Note there is no SYSINIT equivalent of the module 3644 * SHUTDOWN handler, but we don't have anything to do in that case anyhow. 3645 */ 3646 static int 3647 ipfw_modevent(module_t mod, int type, void *unused) 3648 { 3649 int err = 0; 3650 3651 switch (type) { 3652 case MOD_LOAD: 3653 /* Called once at module load or 3654 * system boot if compiled in. */ 3655 break; 3656 case MOD_QUIESCE: 3657 /* Called before unload. May veto unloading. */ 3658 break; 3659 case MOD_UNLOAD: 3660 /* Called during unload. */ 3661 break; 3662 case MOD_SHUTDOWN: 3663 /* Called during system shutdown. */ 3664 break; 3665 default: 3666 err = EOPNOTSUPP; 3667 break; 3668 } 3669 return err; 3670 } 3671 3672 static moduledata_t ipfwmod = { 3673 "ipfw", 3674 ipfw_modevent, 3675 0 3676 }; 3677 3678 /* Define startup order. */ 3679 #define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_FIREWALL 3680 #define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */ 3681 #define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */ 3682 #define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */ 3683 3684 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER); 3685 FEATURE(ipfw_ctl3, "ipfw new sockopt calls"); 3686 MODULE_VERSION(ipfw, 3); 3687 /* should declare some dependencies here */ 3688 3689 /* 3690 * Starting up. Done in order after ipfwmod() has been called. 3691 * VNET_SYSINIT is also called for each existing vnet and each new vnet. 3692 */ 3693 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 3694 ipfw_init, NULL); 3695 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 3696 vnet_ipfw_init, NULL); 3697 3698 /* 3699 * Closing up shop. These are done in REVERSE ORDER, but still 3700 * after ipfwmod() has been called. Not called on reboot. 3701 * VNET_SYSUNINIT is also called for each exiting vnet as it exits. 3702 * or when the module is unloaded. 3703 */ 3704 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 3705 ipfw_destroy, NULL); 3706 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 3707 vnet_ipfw_uninit, NULL); 3708 /* end of file */ 3709