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 /* 30 * The FreeBSD IP packet firewall, main file 31 */ 32 33 #include "opt_ipfw.h" 34 #include "opt_ipdivert.h" 35 #include "opt_inet.h" 36 #ifndef INET 37 #error "IPFIREWALL requires INET" 38 #endif /* INET */ 39 #include "opt_inet6.h" 40 41 #include <sys/param.h> 42 #include <sys/systm.h> 43 #include <sys/condvar.h> 44 #include <sys/counter.h> 45 #include <sys/eventhandler.h> 46 #include <sys/malloc.h> 47 #include <sys/mbuf.h> 48 #include <sys/kernel.h> 49 #include <sys/lock.h> 50 #include <sys/jail.h> 51 #include <sys/module.h> 52 #include <sys/priv.h> 53 #include <sys/proc.h> 54 #include <sys/rwlock.h> 55 #include <sys/rmlock.h> 56 #include <sys/sdt.h> 57 #include <sys/socket.h> 58 #include <sys/socketvar.h> 59 #include <sys/sysctl.h> 60 #include <sys/syslog.h> 61 #include <sys/ucred.h> 62 #include <net/ethernet.h> /* for ETHERTYPE_IP */ 63 #include <net/if.h> 64 #include <net/if_var.h> 65 #include <net/if_private.h> 66 #include <net/route.h> 67 #include <net/route/nhop.h> 68 #include <net/pfil.h> 69 #include <net/vnet.h> 70 #include <net/if_pfsync.h> 71 72 #include <netpfil/pf/pf_mtag.h> 73 74 #include <netinet/in.h> 75 #include <netinet/in_var.h> 76 #include <netinet/in_pcb.h> 77 #include <netinet/ip.h> 78 #include <netinet/ip_var.h> 79 #include <netinet/ip_icmp.h> 80 #include <netinet/ip_fw.h> 81 #include <netinet/ip_carp.h> 82 #include <netinet/pim.h> 83 #include <netinet/tcp_var.h> 84 #include <netinet/udp.h> 85 #include <netinet/udp_var.h> 86 #include <netinet/sctp.h> 87 #include <netinet/sctp_crc32.h> 88 #include <netinet/sctp_header.h> 89 90 #include <netinet/ip6.h> 91 #include <netinet/icmp6.h> 92 #include <netinet/in_fib.h> 93 #ifdef INET6 94 #include <netinet6/in6_fib.h> 95 #include <netinet6/in6_pcb.h> 96 #include <netinet6/scope6_var.h> 97 #include <netinet6/ip6_var.h> 98 #endif 99 100 #include <net/if_gre.h> /* for struct grehdr */ 101 102 #include <netpfil/ipfw/ip_fw_private.h> 103 104 #include <machine/in_cksum.h> /* XXX for in_cksum */ 105 106 #ifdef MAC 107 #include <security/mac/mac_framework.h> 108 #endif 109 110 #define IPFW_PROBE(probe, arg0, arg1, arg2, arg3, arg4, arg5) \ 111 SDT_PROBE6(ipfw, , , probe, arg0, arg1, arg2, arg3, arg4, arg5) 112 113 SDT_PROVIDER_DEFINE(ipfw); 114 SDT_PROBE_DEFINE6(ipfw, , , rule__matched, 115 "int", /* retval */ 116 "int", /* af */ 117 "void *", /* src addr */ 118 "void *", /* dst addr */ 119 "struct ip_fw_args *", /* args */ 120 "struct ip_fw *" /* rule */); 121 122 /* 123 * static variables followed by global ones. 124 * All ipfw global variables are here. 125 */ 126 127 VNET_DEFINE_STATIC(int, fw_deny_unknown_exthdrs); 128 #define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs) 129 130 VNET_DEFINE_STATIC(int, fw_permit_single_frag6) = 1; 131 #define V_fw_permit_single_frag6 VNET(fw_permit_single_frag6) 132 133 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 134 static int default_to_accept = 1; 135 #else 136 static int default_to_accept; 137 #endif 138 139 VNET_DEFINE(int, autoinc_step); 140 VNET_DEFINE(int, fw_one_pass) = 1; 141 142 VNET_DEFINE(unsigned int, fw_tables_max); 143 VNET_DEFINE(unsigned int, fw_tables_sets) = 0; /* Don't use set-aware tables */ 144 /* Use 128 tables by default */ 145 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT; 146 147 #ifndef IPFIREWALL_LINEAR_SKIPTO 148 VNET_DEFINE(int, skipto_cache) = 0; 149 #else 150 VNET_DEFINE(int, skipto_cache) = 1; 151 #endif 152 153 static uint32_t jump(struct ip_fw_chain *chain, struct ip_fw *f, 154 uint32_t num, int tablearg, bool jump_backwards); 155 156 /* 157 * Each rule belongs to one of 32 different sets (0..31). 158 * The variable set_disable contains one bit per set. 159 * If the bit is set, all rules in the corresponding set 160 * are disabled. Set RESVD_SET(31) is reserved for the default rule 161 * and rules that are not deleted by the flush command, 162 * and CANNOT be disabled. 163 * Rules in set RESVD_SET can only be deleted individually. 164 */ 165 VNET_DEFINE(u_int32_t, set_disable); 166 #define V_set_disable VNET(set_disable) 167 168 VNET_DEFINE(int, fw_verbose); 169 /* counter for ipfw_log(NULL...) */ 170 VNET_DEFINE(u_int64_t, norule_counter); 171 VNET_DEFINE(int, verbose_limit); 172 173 /* layer3_chain contains the list of rules for layer 3 */ 174 VNET_DEFINE(struct ip_fw_chain, layer3_chain); 175 176 /* ipfw_vnet_ready controls when we are open for business */ 177 VNET_DEFINE(int, ipfw_vnet_ready) = 0; 178 179 VNET_DEFINE(int, ipfw_nat_ready) = 0; 180 181 ipfw_nat_t *ipfw_nat_ptr = NULL; 182 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int); 183 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; 184 ipfw_nat_cfg_t *ipfw_nat_del_ptr; 185 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; 186 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; 187 188 #ifdef SYSCTL_NODE 189 uint32_t dummy_def = IPFW_DEFAULT_RULE; 190 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS); 191 static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS); 192 193 SYSBEGIN(f3) 194 195 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 196 "Firewall"); 197 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 198 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0, 199 "Only do a single pass through ipfw when using dummynet(4), ipfw_nat or other divert(4)-like interfaces"); 200 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, 201 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0, 202 "Rule number auto-increment step"); 203 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 204 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0, 205 "Log matches to ipfw rules"); 206 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, 207 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(verbose_limit), 0, 208 "Set upper limit of matches of ipfw rules logged"); 209 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, skipto_cache, 210 CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(skipto_cache), 0, 211 "Status of linear skipto cache: 1 - enabled, 0 - disabled."); 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 tcp_set_flags(th, TH_RST); 725 } else { 726 if (flags & TH_SYN) 727 seq++; 728 th->th_ack = htonl(seq); 729 tcp_set_flags(th, 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 tcp_set_flags(th, 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, 890 const struct ip6_hdr *ip6) 891 { 892 struct mbuf *m; 893 894 m = args->m; 895 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 896 const struct tcphdr * tcp; 897 tcp = (const struct tcphdr *)((const char *)ip6 + hlen); 898 899 if ((tcp_get_flags(tcp) & TH_RST) == 0) { 900 struct mbuf *m0; 901 m0 = ipfw_send_pkt(args->m, &(args->f_id), 902 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 903 tcp_get_flags(tcp) | TH_RST); 904 if (m0 != NULL) 905 ip6_output(m0, NULL, NULL, 0, NULL, NULL, 906 NULL); 907 } 908 FREE_PKT(m); 909 } else if (code == ICMP6_UNREACH_ABORT && 910 args->f_id.proto == IPPROTO_SCTP) { 911 struct mbuf *m0; 912 const struct sctphdr *sctp; 913 u_int32_t v_tag; 914 int reflected; 915 916 sctp = (const struct sctphdr *)((const char *)ip6 + hlen); 917 reflected = 1; 918 v_tag = ntohl(sctp->v_tag); 919 /* Investigate the first chunk header if available */ 920 if (m->m_len >= hlen + sizeof(struct sctphdr) + 921 sizeof(struct sctp_chunkhdr)) { 922 const struct sctp_chunkhdr *chunk; 923 924 chunk = (const struct sctp_chunkhdr *)(sctp + 1); 925 switch (chunk->chunk_type) { 926 case SCTP_INITIATION: 927 /* 928 * Packets containing an INIT chunk MUST have 929 * a zero v-tag. 930 */ 931 if (v_tag != 0) { 932 v_tag = 0; 933 break; 934 } 935 /* INIT chunk MUST NOT be bundled */ 936 if (m->m_pkthdr.len > 937 hlen + sizeof(struct sctphdr) + 938 ntohs(chunk->chunk_length) + 3) { 939 break; 940 } 941 /* Use the initiate tag if available */ 942 if ((m->m_len >= hlen + sizeof(struct sctphdr) + 943 sizeof(struct sctp_chunkhdr) + 944 offsetof(struct sctp_init, a_rwnd))) { 945 const struct sctp_init *init; 946 947 init = (const struct sctp_init *)(chunk + 1); 948 v_tag = ntohl(init->initiate_tag); 949 reflected = 0; 950 } 951 break; 952 case SCTP_ABORT_ASSOCIATION: 953 /* 954 * If the packet contains an ABORT chunk, don't 955 * reply. 956 * XXX: We should search through all chunks, 957 * but do not do that to avoid attacks. 958 */ 959 v_tag = 0; 960 break; 961 } 962 } 963 if (v_tag == 0) { 964 m0 = NULL; 965 } else { 966 m0 = ipfw_send_abort(args->m, &(args->f_id), v_tag, 967 reflected); 968 } 969 if (m0 != NULL) 970 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL); 971 FREE_PKT(m); 972 } else if (code != ICMP6_UNREACH_RST && code != ICMP6_UNREACH_ABORT) { 973 /* Send an ICMPv6 unreach. */ 974 #if 0 975 /* 976 * Unlike above, the mbufs need to line up with the ip6 hdr, 977 * as the contents are read. We need to m_adj() the 978 * needed amount. 979 * The mbuf will however be thrown away so we can adjust it. 980 * Remember we did an m_pullup on it already so we 981 * can make some assumptions about contiguousness. 982 */ 983 if (args->L3offset) 984 m_adj(m, args->L3offset); 985 #endif 986 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 987 } else 988 FREE_PKT(m); 989 990 args->m = NULL; 991 } 992 993 #endif /* INET6 */ 994 995 /* 996 * sends a reject message, consuming the mbuf passed as an argument. 997 */ 998 static void 999 send_reject(struct ip_fw_args *args, int code, uint16_t mtu, int iplen, 1000 const struct ip *ip) 1001 { 1002 #if 0 1003 /* XXX When ip is not guaranteed to be at mtod() we will 1004 * need to account for this */ 1005 * The mbuf will however be thrown away so we can adjust it. 1006 * Remember we did an m_pullup on it already so we 1007 * can make some assumptions about contiguousness. 1008 */ 1009 if (args->L3offset) 1010 m_adj(m, args->L3offset); 1011 #endif 1012 if (code != ICMP_REJECT_RST && code != ICMP_REJECT_ABORT) { 1013 /* Send an ICMP unreach */ 1014 icmp_error(args->m, ICMP_UNREACH, code, 0L, mtu); 1015 } else if (code == ICMP_REJECT_RST && args->f_id.proto == IPPROTO_TCP) { 1016 struct tcphdr *const tcp = 1017 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1018 if ( (tcp_get_flags(tcp) & TH_RST) == 0) { 1019 struct mbuf *m; 1020 m = ipfw_send_pkt(args->m, &(args->f_id), 1021 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1022 tcp_get_flags(tcp) | TH_RST); 1023 if (m != NULL) 1024 ip_output(m, NULL, NULL, 0, NULL, NULL); 1025 } 1026 FREE_PKT(args->m); 1027 } else if (code == ICMP_REJECT_ABORT && 1028 args->f_id.proto == IPPROTO_SCTP) { 1029 struct mbuf *m; 1030 struct sctphdr *sctp; 1031 struct sctp_chunkhdr *chunk; 1032 struct sctp_init *init; 1033 u_int32_t v_tag; 1034 int reflected; 1035 1036 sctp = L3HDR(struct sctphdr, mtod(args->m, struct ip *)); 1037 reflected = 1; 1038 v_tag = ntohl(sctp->v_tag); 1039 if (iplen >= (ip->ip_hl << 2) + sizeof(struct sctphdr) + 1040 sizeof(struct sctp_chunkhdr)) { 1041 /* Look at the first chunk header if available */ 1042 chunk = (struct sctp_chunkhdr *)(sctp + 1); 1043 switch (chunk->chunk_type) { 1044 case SCTP_INITIATION: 1045 /* 1046 * Packets containing an INIT chunk MUST have 1047 * a zero v-tag. 1048 */ 1049 if (v_tag != 0) { 1050 v_tag = 0; 1051 break; 1052 } 1053 /* INIT chunk MUST NOT be bundled */ 1054 if (iplen > 1055 (ip->ip_hl << 2) + sizeof(struct sctphdr) + 1056 ntohs(chunk->chunk_length) + 3) { 1057 break; 1058 } 1059 /* Use the initiate tag if available */ 1060 if ((iplen >= (ip->ip_hl << 2) + 1061 sizeof(struct sctphdr) + 1062 sizeof(struct sctp_chunkhdr) + 1063 offsetof(struct sctp_init, a_rwnd))) { 1064 init = (struct sctp_init *)(chunk + 1); 1065 v_tag = ntohl(init->initiate_tag); 1066 reflected = 0; 1067 } 1068 break; 1069 case SCTP_ABORT_ASSOCIATION: 1070 /* 1071 * If the packet contains an ABORT chunk, don't 1072 * reply. 1073 * XXX: We should search through all chunks, 1074 * but do not do that to avoid attacks. 1075 */ 1076 v_tag = 0; 1077 break; 1078 } 1079 } 1080 if (v_tag == 0) { 1081 m = NULL; 1082 } else { 1083 m = ipfw_send_abort(args->m, &(args->f_id), v_tag, 1084 reflected); 1085 } 1086 if (m != NULL) 1087 ip_output(m, NULL, NULL, 0, NULL, NULL); 1088 FREE_PKT(args->m); 1089 } else 1090 FREE_PKT(args->m); 1091 args->m = NULL; 1092 } 1093 1094 /* 1095 * Support for uid/gid/jail lookup. These tests are expensive 1096 * (because we may need to look into the list of active sockets) 1097 * so we cache the results. ugid_lookupp is 0 if we have not 1098 * yet done a lookup, 1 if we succeeded, and -1 if we tried 1099 * and failed. The function always returns the match value. 1100 * We could actually spare the variable and use *uc, setting 1101 * it to '(void *)check_uidgid if we have no info, NULL if 1102 * we tried and failed, or any other value if successful. 1103 */ 1104 static int 1105 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp, 1106 struct ucred **uc) 1107 { 1108 #if defined(USERSPACE) 1109 return 0; // not supported in userspace 1110 #else 1111 #ifndef __FreeBSD__ 1112 /* XXX */ 1113 return cred_check(insn, proto, oif, 1114 dst_ip, dst_port, src_ip, src_port, 1115 (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb); 1116 #else /* FreeBSD */ 1117 struct in_addr src_ip, dst_ip; 1118 struct inpcbinfo *pi; 1119 struct ipfw_flow_id *id; 1120 struct inpcb *pcb, *inp; 1121 int lookupflags; 1122 int match; 1123 1124 id = &args->f_id; 1125 inp = args->inp; 1126 1127 /* 1128 * Check to see if the UDP or TCP stack supplied us with 1129 * the PCB. If so, rather then holding a lock and looking 1130 * up the PCB, we can use the one that was supplied. 1131 */ 1132 if (inp && *ugid_lookupp == 0) { 1133 INP_LOCK_ASSERT(inp); 1134 if (inp->inp_socket != NULL) { 1135 *uc = crhold(inp->inp_cred); 1136 *ugid_lookupp = 1; 1137 } else 1138 *ugid_lookupp = -1; 1139 } 1140 /* 1141 * If we have already been here and the packet has no 1142 * PCB entry associated with it, then we can safely 1143 * assume that this is a no match. 1144 */ 1145 if (*ugid_lookupp == -1) 1146 return (0); 1147 if (id->proto == IPPROTO_TCP) { 1148 lookupflags = 0; 1149 pi = &V_tcbinfo; 1150 } else if (id->proto == IPPROTO_UDP) { 1151 lookupflags = INPLOOKUP_WILDCARD; 1152 pi = &V_udbinfo; 1153 } else if (id->proto == IPPROTO_UDPLITE) { 1154 lookupflags = INPLOOKUP_WILDCARD; 1155 pi = &V_ulitecbinfo; 1156 } else 1157 return 0; 1158 lookupflags |= INPLOOKUP_RLOCKPCB; 1159 match = 0; 1160 if (*ugid_lookupp == 0) { 1161 if (id->addr_type == 6) { 1162 #ifdef INET6 1163 if (args->flags & IPFW_ARGS_IN) 1164 pcb = in6_pcblookup_mbuf(pi, 1165 &id->src_ip6, htons(id->src_port), 1166 &id->dst_ip6, htons(id->dst_port), 1167 lookupflags, NULL, args->m); 1168 else 1169 pcb = in6_pcblookup_mbuf(pi, 1170 &id->dst_ip6, htons(id->dst_port), 1171 &id->src_ip6, htons(id->src_port), 1172 lookupflags, args->ifp, args->m); 1173 #else 1174 *ugid_lookupp = -1; 1175 return (0); 1176 #endif 1177 } else { 1178 src_ip.s_addr = htonl(id->src_ip); 1179 dst_ip.s_addr = htonl(id->dst_ip); 1180 if (args->flags & IPFW_ARGS_IN) 1181 pcb = in_pcblookup_mbuf(pi, 1182 src_ip, htons(id->src_port), 1183 dst_ip, htons(id->dst_port), 1184 lookupflags, NULL, args->m); 1185 else 1186 pcb = in_pcblookup_mbuf(pi, 1187 dst_ip, htons(id->dst_port), 1188 src_ip, htons(id->src_port), 1189 lookupflags, args->ifp, args->m); 1190 } 1191 if (pcb != NULL) { 1192 INP_RLOCK_ASSERT(pcb); 1193 *uc = crhold(pcb->inp_cred); 1194 *ugid_lookupp = 1; 1195 INP_RUNLOCK(pcb); 1196 } 1197 if (*ugid_lookupp == 0) { 1198 /* 1199 * We tried and failed, set the variable to -1 1200 * so we will not try again on this packet. 1201 */ 1202 *ugid_lookupp = -1; 1203 return (0); 1204 } 1205 } 1206 if (insn->o.opcode == O_UID) 1207 match = ((*uc)->cr_uid == (uid_t)insn->d[0]); 1208 else if (insn->o.opcode == O_GID) 1209 match = groupmember((gid_t)insn->d[0], *uc); 1210 else if (insn->o.opcode == O_JAIL) 1211 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]); 1212 return (match); 1213 #endif /* __FreeBSD__ */ 1214 #endif /* not supported in userspace */ 1215 } 1216 1217 /* 1218 * Helper function to set args with info on the rule after the matching 1219 * one. slot is precise, whereas we guess rule_id as they are 1220 * assigned sequentially. 1221 */ 1222 static inline void 1223 set_match(struct ip_fw_args *args, int slot, 1224 struct ip_fw_chain *chain) 1225 { 1226 args->rule.chain_id = chain->id; 1227 args->rule.slot = slot + 1; /* we use 0 as a marker */ 1228 args->rule.rule_id = 1 + chain->map[slot]->id; 1229 args->rule.rulenum = chain->map[slot]->rulenum; 1230 args->flags |= IPFW_ARGS_REF; 1231 } 1232 1233 static uint32_t 1234 jump_lookup_pos(struct ip_fw_chain *chain, struct ip_fw *f, uint32_t num, 1235 int tablearg, bool jump_backwards) 1236 { 1237 int f_pos, i; 1238 1239 /* 1240 * Make sure we do not jump backward. 1241 */ 1242 i = IP_FW_ARG_TABLEARG(chain, num, skipto); 1243 if (!jump_backwards && i <= f->rulenum) 1244 i = f->rulenum + 1; 1245 1246 if (V_skipto_cache == 0) 1247 f_pos = ipfw_find_rule(chain, i, 0); 1248 else { 1249 /* 1250 * Make sure we do not do out of bounds access. 1251 */ 1252 if (i >= IPFW_DEFAULT_RULE) 1253 i = IPFW_DEFAULT_RULE - 1; 1254 f_pos = chain->idxmap[i]; 1255 } 1256 1257 return (f_pos); 1258 } 1259 1260 static uint32_t 1261 jump(struct ip_fw_chain *chain, struct ip_fw *f, uint32_t num, 1262 int tablearg, bool jump_backwards) 1263 { 1264 int f_pos; 1265 1266 /* Can't use cache with IP_FW_TARG */ 1267 if (num == IP_FW_TARG) 1268 return jump_lookup_pos(chain, f, num, tablearg, jump_backwards); 1269 1270 /* 1271 * If possible use cached f_pos (in f->cache.pos), 1272 * whose version is written in f->cache.id (horrible hacks 1273 * to avoid changing the ABI). 1274 * 1275 * Multiple threads can execute the same rule simultaneously, 1276 * we need to ensure that cache.pos is updated before cache.id. 1277 */ 1278 1279 #ifdef __LP64__ 1280 struct ip_fw_jump_cache cache; 1281 1282 cache.raw_value = f->cache.raw_value; 1283 if (cache.id == chain->id) 1284 return (cache.pos); 1285 1286 f_pos = jump_lookup_pos(chain, f, num, tablearg, jump_backwards); 1287 1288 cache.pos = f_pos; 1289 cache.id = chain->id; 1290 f->cache.raw_value = cache.raw_value; 1291 #else 1292 if (f->cache.id == chain->id) { 1293 /* Load pos after id */ 1294 atomic_thread_fence_acq(); 1295 return (f->cache.pos); 1296 } 1297 1298 f_pos = jump_lookup_pos(chain, f, num, tablearg, jump_backwards); 1299 1300 f->cache.pos = f_pos; 1301 /* Store id after pos */ 1302 atomic_thread_fence_rel(); 1303 f->cache.id = chain->id; 1304 #endif /* !__LP64__ */ 1305 return (f_pos); 1306 } 1307 1308 #define TARG(k, f) IP_FW_ARG_TABLEARG(chain, k, f) 1309 1310 static inline int 1311 tvalue_match(struct ip_fw_chain *ch, const ipfw_insn_table *cmd, 1312 uint32_t tablearg) 1313 { 1314 uint32_t tvalue; 1315 1316 switch (IPFW_TVALUE_TYPE(&cmd->o)) { 1317 case TVALUE_PIPE: 1318 tvalue = TARG_VAL(ch, tablearg, pipe); 1319 break; 1320 case TVALUE_DIVERT: 1321 tvalue = TARG_VAL(ch, tablearg, divert); 1322 break; 1323 case TVALUE_SKIPTO: 1324 tvalue = TARG_VAL(ch, tablearg, skipto); 1325 break; 1326 case TVALUE_NETGRAPH: 1327 tvalue = TARG_VAL(ch, tablearg, netgraph); 1328 break; 1329 case TVALUE_FIB: 1330 tvalue = TARG_VAL(ch, tablearg, fib); 1331 break; 1332 case TVALUE_NAT: 1333 tvalue = TARG_VAL(ch, tablearg, nat); 1334 break; 1335 case TVALUE_NH4: 1336 tvalue = TARG_VAL(ch, tablearg, nh4); 1337 break; 1338 case TVALUE_DSCP: 1339 tvalue = TARG_VAL(ch, tablearg, dscp); 1340 break; 1341 case TVALUE_LIMIT: 1342 tvalue = TARG_VAL(ch, tablearg, limit); 1343 break; 1344 case TVALUE_MARK: 1345 tvalue = TARG_VAL(ch, tablearg, mark); 1346 break; 1347 case TVALUE_TAG: 1348 default: 1349 tvalue = TARG_VAL(ch, tablearg, tag); 1350 break; 1351 } 1352 return (tvalue == cmd->value); 1353 } 1354 1355 /* 1356 * The main check routine for the firewall. 1357 * 1358 * All arguments are in args so we can modify them and return them 1359 * back to the caller. 1360 * 1361 * Parameters: 1362 * 1363 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1364 * Starts with the IP header. 1365 * args->L3offset Number of bytes bypassed if we came from L2. 1366 * e.g. often sizeof(eh) ** NOTYET ** 1367 * args->ifp Incoming or outgoing interface. 1368 * args->divert_rule (in/out) 1369 * Skip up to the first rule past this rule number; 1370 * upon return, non-zero port number for divert or tee. 1371 * 1372 * args->rule Pointer to the last matching rule (in/out) 1373 * args->next_hop Socket we are forwarding to (out). 1374 * args->next_hop6 IPv6 next hop we are forwarding to (out). 1375 * args->f_id Addresses grabbed from the packet (out) 1376 * args->rule.info a cookie depending on rule action 1377 * 1378 * Return value: 1379 * 1380 * IP_FW_PASS the packet must be accepted 1381 * IP_FW_DENY the packet must be dropped 1382 * IP_FW_DIVERT divert packet, port in m_tag 1383 * IP_FW_TEE tee packet, port in m_tag 1384 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 1385 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 1386 * args->rule contains the matching rule, 1387 * args->rule.info has additional information. 1388 * 1389 */ 1390 int 1391 ipfw_chk(struct ip_fw_args *args) 1392 { 1393 1394 /* 1395 * Local variables holding state while processing a packet: 1396 * 1397 * IMPORTANT NOTE: to speed up the processing of rules, there 1398 * are some assumption on the values of the variables, which 1399 * are documented here. Should you change them, please check 1400 * the implementation of the various instructions to make sure 1401 * that they still work. 1402 * 1403 * m | args->m Pointer to the mbuf, as received from the caller. 1404 * It may change if ipfw_chk() does an m_pullup, or if it 1405 * consumes the packet because it calls send_reject(). 1406 * XXX This has to change, so that ipfw_chk() never modifies 1407 * or consumes the buffer. 1408 * OR 1409 * args->mem Pointer to contigous memory chunk. 1410 * ip Is the beginning of the ip(4 or 6) header. 1411 * eh Ethernet header in case if input is Layer2. 1412 */ 1413 struct mbuf *m; 1414 struct ip *ip; 1415 struct ether_header *eh; 1416 1417 /* 1418 * For rules which contain uid/gid or jail constraints, cache 1419 * a copy of the users credentials after the pcb lookup has been 1420 * executed. This will speed up the processing of rules with 1421 * these types of constraints, as well as decrease contention 1422 * on pcb related locks. 1423 */ 1424 #ifndef __FreeBSD__ 1425 struct bsd_ucred ucred_cache; 1426 #else 1427 struct ucred *ucred_cache = NULL; 1428 #endif 1429 uint32_t f_pos = 0; /* index of current rule in the array */ 1430 int ucred_lookup = 0; 1431 int retval = 0; 1432 struct ifnet *oif, *iif; 1433 1434 /* 1435 * hlen The length of the IP header. 1436 */ 1437 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 1438 1439 /* 1440 * offset The offset of a fragment. offset != 0 means that 1441 * we have a fragment at this offset of an IPv4 packet. 1442 * offset == 0 means that (if this is an IPv4 packet) 1443 * this is the first or only fragment. 1444 * For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header 1445 * or there is a single packet fragment (fragment header added 1446 * without needed). We will treat a single packet fragment as if 1447 * there was no fragment header (or log/block depending on the 1448 * V_fw_permit_single_frag6 sysctl setting). 1449 */ 1450 u_short offset = 0; 1451 u_short ip6f_mf = 0; 1452 1453 /* 1454 * Local copies of addresses. They are only valid if we have 1455 * an IP packet. 1456 * 1457 * proto The protocol. Set to 0 for non-ip packets, 1458 * or to the protocol read from the packet otherwise. 1459 * proto != 0 means that we have an IPv4 packet. 1460 * 1461 * src_port, dst_port port numbers, in HOST format. Only 1462 * valid for TCP and UDP packets. 1463 * 1464 * src_ip, dst_ip ip addresses, in NETWORK format. 1465 * Only valid for IPv4 packets. 1466 */ 1467 uint8_t proto; 1468 uint16_t src_port, dst_port; /* NOTE: host format */ 1469 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 1470 int iplen = 0; 1471 int pktlen; 1472 1473 struct ipfw_dyn_info dyn_info; 1474 struct ip_fw *q = NULL; 1475 struct ip_fw_chain *chain = &V_layer3_chain; 1476 1477 /* 1478 * We store in ulp a pointer to the upper layer protocol header. 1479 * In the ipv4 case this is easy to determine from the header, 1480 * but for ipv6 we might have some additional headers in the middle. 1481 * ulp is NULL if not found. 1482 */ 1483 void *ulp = NULL; /* upper layer protocol pointer. */ 1484 1485 /* XXX ipv6 variables */ 1486 int is_ipv6 = 0; 1487 #ifdef INET6 1488 uint8_t icmp6_type = 0; 1489 #endif 1490 uint16_t ext_hd = 0; /* bits vector for extension header filtering */ 1491 /* end of ipv6 variables */ 1492 1493 int is_ipv4 = 0; 1494 1495 int done = 0; /* flag to exit the outer loop */ 1496 IPFW_RLOCK_TRACKER; 1497 bool mem; 1498 bool need_send_reject = false; 1499 int reject_code; 1500 uint16_t reject_mtu; 1501 1502 if ((mem = (args->flags & IPFW_ARGS_LENMASK))) { 1503 if (args->flags & IPFW_ARGS_ETHER) { 1504 eh = (struct ether_header *)args->mem; 1505 if (eh->ether_type == htons(ETHERTYPE_VLAN)) 1506 ip = (struct ip *) 1507 ((struct ether_vlan_header *)eh + 1); 1508 else 1509 ip = (struct ip *)(eh + 1); 1510 } else { 1511 eh = NULL; 1512 ip = (struct ip *)args->mem; 1513 } 1514 pktlen = IPFW_ARGS_LENGTH(args->flags); 1515 args->f_id.fib = args->ifp->if_fib; /* best guess */ 1516 } else { 1517 m = args->m; 1518 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready)) 1519 return (IP_FW_PASS); /* accept */ 1520 if (args->flags & IPFW_ARGS_ETHER) { 1521 /* We need some amount of data to be contiguous. */ 1522 if (m->m_len < min(m->m_pkthdr.len, max_protohdr) && 1523 (args->m = m = m_pullup(m, min(m->m_pkthdr.len, 1524 max_protohdr))) == NULL) 1525 goto pullup_failed; 1526 eh = mtod(m, struct ether_header *); 1527 ip = (struct ip *)(eh + 1); 1528 } else { 1529 eh = NULL; 1530 ip = mtod(m, struct ip *); 1531 } 1532 pktlen = m->m_pkthdr.len; 1533 args->f_id.fib = M_GETFIB(m); /* mbuf not altered */ 1534 } 1535 1536 dst_ip.s_addr = 0; /* make sure it is initialized */ 1537 src_ip.s_addr = 0; /* make sure it is initialized */ 1538 src_port = dst_port = 0; 1539 1540 DYN_INFO_INIT(&dyn_info); 1541 /* 1542 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 1543 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 1544 * pointer might become stale after other pullups (but we never use it 1545 * this way). 1546 */ 1547 #define PULLUP_TO(_len, p, T) PULLUP_LEN(_len, p, sizeof(T)) 1548 #define EHLEN (eh != NULL ? ((char *)ip - (char *)eh) : 0) 1549 #define _PULLUP_LOCKED(_len, p, T, unlock) \ 1550 do { \ 1551 int x = (_len) + T + EHLEN; \ 1552 if (mem) { \ 1553 if (__predict_false(pktlen < x)) { \ 1554 unlock; \ 1555 goto pullup_failed; \ 1556 } \ 1557 p = (char *)args->mem + (_len) + EHLEN; \ 1558 } else { \ 1559 if (__predict_false((m)->m_len < x)) { \ 1560 args->m = m = m_pullup(m, x); \ 1561 if (m == NULL) { \ 1562 unlock; \ 1563 goto pullup_failed; \ 1564 } \ 1565 } \ 1566 p = mtod(m, char *) + (_len) + EHLEN; \ 1567 } \ 1568 } while (0) 1569 1570 #define PULLUP_LEN(_len, p, T) _PULLUP_LOCKED(_len, p, T, ) 1571 #define PULLUP_LEN_LOCKED(_len, p, T) \ 1572 _PULLUP_LOCKED(_len, p, T, IPFW_PF_RUNLOCK(chain)); \ 1573 UPDATE_POINTERS() 1574 /* 1575 * In case pointers got stale after pullups, update them. 1576 */ 1577 #define UPDATE_POINTERS() \ 1578 do { \ 1579 if (!mem) { \ 1580 if (eh != NULL) { \ 1581 eh = mtod(m, struct ether_header *); \ 1582 ip = (struct ip *)(eh + 1); \ 1583 } else \ 1584 ip = mtod(m, struct ip *); \ 1585 args->m = m; \ 1586 } \ 1587 } while (0) 1588 1589 /* Identify IP packets and fill up variables. */ 1590 if (pktlen >= sizeof(struct ip6_hdr) && 1591 (eh == NULL || eh->ether_type == htons(ETHERTYPE_IPV6)) && 1592 ip->ip_v == 6) { 1593 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 1594 1595 is_ipv6 = 1; 1596 args->flags |= IPFW_ARGS_IP6; 1597 hlen = sizeof(struct ip6_hdr); 1598 proto = ip6->ip6_nxt; 1599 /* Search extension headers to find upper layer protocols */ 1600 while (ulp == NULL && offset == 0) { 1601 switch (proto) { 1602 case IPPROTO_ICMPV6: 1603 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 1604 #ifdef INET6 1605 icmp6_type = ICMP6(ulp)->icmp6_type; 1606 #endif 1607 break; 1608 1609 case IPPROTO_TCP: 1610 PULLUP_TO(hlen, ulp, struct tcphdr); 1611 dst_port = TCP(ulp)->th_dport; 1612 src_port = TCP(ulp)->th_sport; 1613 /* save flags for dynamic rules */ 1614 args->f_id._flags = tcp_get_flags(TCP(ulp)); 1615 break; 1616 1617 case IPPROTO_SCTP: 1618 if (pktlen >= hlen + sizeof(struct sctphdr) + 1619 sizeof(struct sctp_chunkhdr) + 1620 offsetof(struct sctp_init, a_rwnd)) 1621 PULLUP_LEN(hlen, ulp, 1622 sizeof(struct sctphdr) + 1623 sizeof(struct sctp_chunkhdr) + 1624 offsetof(struct sctp_init, a_rwnd)); 1625 else if (pktlen >= hlen + sizeof(struct sctphdr)) 1626 PULLUP_LEN(hlen, ulp, pktlen - hlen); 1627 else 1628 PULLUP_LEN(hlen, ulp, 1629 sizeof(struct sctphdr)); 1630 src_port = SCTP(ulp)->src_port; 1631 dst_port = SCTP(ulp)->dest_port; 1632 break; 1633 1634 case IPPROTO_UDP: 1635 case IPPROTO_UDPLITE: 1636 PULLUP_TO(hlen, ulp, struct udphdr); 1637 dst_port = UDP(ulp)->uh_dport; 1638 src_port = UDP(ulp)->uh_sport; 1639 break; 1640 1641 case IPPROTO_HOPOPTS: /* RFC 2460 */ 1642 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1643 ext_hd |= EXT_HOPOPTS; 1644 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1645 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1646 ulp = NULL; 1647 break; 1648 1649 case IPPROTO_ROUTING: /* RFC 2460 */ 1650 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 1651 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 1652 case 0: 1653 ext_hd |= EXT_RTHDR0; 1654 break; 1655 case 2: 1656 ext_hd |= EXT_RTHDR2; 1657 break; 1658 default: 1659 if (V_fw_verbose) 1660 printf("IPFW2: IPV6 - Unknown " 1661 "Routing Header type(%d)\n", 1662 ((struct ip6_rthdr *) 1663 ulp)->ip6r_type); 1664 if (V_fw_deny_unknown_exthdrs) 1665 return (IP_FW_DENY); 1666 break; 1667 } 1668 ext_hd |= EXT_ROUTING; 1669 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 1670 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 1671 ulp = NULL; 1672 break; 1673 1674 case IPPROTO_FRAGMENT: /* RFC 2460 */ 1675 PULLUP_TO(hlen, ulp, struct ip6_frag); 1676 ext_hd |= EXT_FRAGMENT; 1677 hlen += sizeof (struct ip6_frag); 1678 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 1679 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 1680 IP6F_OFF_MASK; 1681 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg & 1682 IP6F_MORE_FRAG; 1683 if (V_fw_permit_single_frag6 == 0 && 1684 offset == 0 && ip6f_mf == 0) { 1685 if (V_fw_verbose) 1686 printf("IPFW2: IPV6 - Invalid " 1687 "Fragment Header\n"); 1688 if (V_fw_deny_unknown_exthdrs) 1689 return (IP_FW_DENY); 1690 break; 1691 } 1692 args->f_id.extra = 1693 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 1694 ulp = NULL; 1695 break; 1696 1697 case IPPROTO_DSTOPTS: /* RFC 2460 */ 1698 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1699 ext_hd |= EXT_DSTOPTS; 1700 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1701 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1702 ulp = NULL; 1703 break; 1704 1705 case IPPROTO_AH: /* RFC 2402 */ 1706 PULLUP_TO(hlen, ulp, struct ip6_ext); 1707 ext_hd |= EXT_AH; 1708 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 1709 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 1710 ulp = NULL; 1711 break; 1712 1713 case IPPROTO_ESP: /* RFC 2406 */ 1714 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 1715 /* Anything past Seq# is variable length and 1716 * data past this ext. header is encrypted. */ 1717 ext_hd |= EXT_ESP; 1718 break; 1719 1720 case IPPROTO_NONE: /* RFC 2460 */ 1721 /* 1722 * Packet ends here, and IPv6 header has 1723 * already been pulled up. If ip6e_len!=0 1724 * then octets must be ignored. 1725 */ 1726 ulp = ip; /* non-NULL to get out of loop. */ 1727 break; 1728 1729 case IPPROTO_OSPFIGP: 1730 /* XXX OSPF header check? */ 1731 PULLUP_TO(hlen, ulp, struct ip6_ext); 1732 break; 1733 1734 case IPPROTO_PIM: 1735 /* XXX PIM header check? */ 1736 PULLUP_TO(hlen, ulp, struct pim); 1737 break; 1738 1739 case IPPROTO_GRE: /* RFC 1701 */ 1740 /* XXX GRE header check? */ 1741 PULLUP_TO(hlen, ulp, struct grehdr); 1742 break; 1743 1744 case IPPROTO_CARP: 1745 PULLUP_TO(hlen, ulp, offsetof( 1746 struct carp_header, carp_counter)); 1747 if (CARP_ADVERTISEMENT != 1748 ((struct carp_header *)ulp)->carp_type) 1749 return (IP_FW_DENY); 1750 break; 1751 1752 case IPPROTO_IPV6: /* RFC 2893 */ 1753 PULLUP_TO(hlen, ulp, struct ip6_hdr); 1754 break; 1755 1756 case IPPROTO_IPV4: /* RFC 2893 */ 1757 PULLUP_TO(hlen, ulp, struct ip); 1758 break; 1759 1760 case IPPROTO_PFSYNC: 1761 PULLUP_TO(hlen, ulp, struct pfsync_header); 1762 break; 1763 1764 default: 1765 if (V_fw_verbose) 1766 printf("IPFW2: IPV6 - Unknown " 1767 "Extension Header(%d), ext_hd=%x\n", 1768 proto, ext_hd); 1769 if (V_fw_deny_unknown_exthdrs) 1770 return (IP_FW_DENY); 1771 PULLUP_TO(hlen, ulp, struct ip6_ext); 1772 break; 1773 } /*switch */ 1774 } 1775 UPDATE_POINTERS(); 1776 ip6 = (struct ip6_hdr *)ip; 1777 args->f_id.addr_type = 6; 1778 args->f_id.src_ip6 = ip6->ip6_src; 1779 args->f_id.dst_ip6 = ip6->ip6_dst; 1780 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 1781 iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6); 1782 } else if (pktlen >= sizeof(struct ip) && 1783 (eh == NULL || eh->ether_type == htons(ETHERTYPE_IP)) && 1784 ip->ip_v == 4) { 1785 is_ipv4 = 1; 1786 args->flags |= IPFW_ARGS_IP4; 1787 hlen = ip->ip_hl << 2; 1788 /* 1789 * Collect parameters into local variables for faster 1790 * matching. 1791 */ 1792 proto = ip->ip_p; 1793 src_ip = ip->ip_src; 1794 dst_ip = ip->ip_dst; 1795 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1796 iplen = ntohs(ip->ip_len); 1797 1798 if (offset == 0) { 1799 switch (proto) { 1800 case IPPROTO_TCP: 1801 PULLUP_TO(hlen, ulp, struct tcphdr); 1802 dst_port = TCP(ulp)->th_dport; 1803 src_port = TCP(ulp)->th_sport; 1804 /* save flags for dynamic rules */ 1805 args->f_id._flags = tcp_get_flags(TCP(ulp)); 1806 break; 1807 1808 case IPPROTO_SCTP: 1809 if (pktlen >= hlen + sizeof(struct sctphdr) + 1810 sizeof(struct sctp_chunkhdr) + 1811 offsetof(struct sctp_init, a_rwnd)) 1812 PULLUP_LEN(hlen, ulp, 1813 sizeof(struct sctphdr) + 1814 sizeof(struct sctp_chunkhdr) + 1815 offsetof(struct sctp_init, a_rwnd)); 1816 else if (pktlen >= hlen + sizeof(struct sctphdr)) 1817 PULLUP_LEN(hlen, ulp, pktlen - hlen); 1818 else 1819 PULLUP_LEN(hlen, ulp, 1820 sizeof(struct sctphdr)); 1821 src_port = SCTP(ulp)->src_port; 1822 dst_port = SCTP(ulp)->dest_port; 1823 break; 1824 1825 case IPPROTO_UDP: 1826 case IPPROTO_UDPLITE: 1827 PULLUP_TO(hlen, ulp, struct udphdr); 1828 dst_port = UDP(ulp)->uh_dport; 1829 src_port = UDP(ulp)->uh_sport; 1830 break; 1831 1832 case IPPROTO_ICMP: 1833 PULLUP_TO(hlen, ulp, struct icmphdr); 1834 //args->f_id.flags = ICMP(ulp)->icmp_type; 1835 break; 1836 1837 default: 1838 break; 1839 } 1840 } else { 1841 if (offset == 1 && proto == IPPROTO_TCP) { 1842 /* RFC 3128 */ 1843 goto pullup_failed; 1844 } 1845 } 1846 1847 UPDATE_POINTERS(); 1848 args->f_id.addr_type = 4; 1849 args->f_id.src_ip = ntohl(src_ip.s_addr); 1850 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1851 } else { 1852 proto = 0; 1853 dst_ip.s_addr = src_ip.s_addr = 0; 1854 1855 args->f_id.addr_type = 1; /* XXX */ 1856 } 1857 #undef PULLUP_TO 1858 pktlen = iplen < pktlen ? iplen: pktlen; 1859 1860 /* Properly initialize the rest of f_id */ 1861 args->f_id.proto = proto; 1862 args->f_id.src_port = src_port = ntohs(src_port); 1863 args->f_id.dst_port = dst_port = ntohs(dst_port); 1864 1865 IPFW_PF_RLOCK(chain); 1866 if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */ 1867 IPFW_PF_RUNLOCK(chain); 1868 return (IP_FW_PASS); /* accept */ 1869 } 1870 if (args->flags & IPFW_ARGS_REF) { 1871 /* 1872 * Packet has already been tagged as a result of a previous 1873 * match on rule args->rule aka args->rule_id (PIPE, QUEUE, 1874 * REASS, NETGRAPH, DIVERT/TEE...) 1875 * Validate the slot and continue from the next one 1876 * if still present, otherwise do a lookup. 1877 */ 1878 f_pos = (args->rule.chain_id == chain->id) ? 1879 args->rule.slot : 1880 ipfw_find_rule(chain, args->rule.rulenum, 1881 args->rule.rule_id); 1882 } else { 1883 f_pos = 0; 1884 } 1885 1886 if (args->flags & IPFW_ARGS_IN) { 1887 iif = args->ifp; 1888 oif = NULL; 1889 } else { 1890 MPASS(args->flags & IPFW_ARGS_OUT); 1891 iif = mem ? NULL : m_rcvif(m); 1892 oif = args->ifp; 1893 } 1894 1895 /* 1896 * Now scan the rules, and parse microinstructions for each rule. 1897 * We have two nested loops and an inner switch. Sometimes we 1898 * need to break out of one or both loops, or re-enter one of 1899 * the loops with updated variables. Loop variables are: 1900 * 1901 * f_pos (outer loop) points to the current rule. 1902 * On output it points to the matching rule. 1903 * done (outer loop) is used as a flag to break the loop. 1904 * l (inner loop) residual length of current rule. 1905 * cmd points to the current microinstruction. 1906 * 1907 * We break the inner loop by setting l=0 and possibly 1908 * cmdlen=0 if we don't want to advance cmd. 1909 * We break the outer loop by setting done=1 1910 * We can restart the inner loop by setting l>0 and f_pos, f, cmd 1911 * as needed. 1912 */ 1913 for (; f_pos < chain->n_rules; f_pos++) { 1914 ipfw_insn *cmd; 1915 uint32_t tablearg = 0; 1916 int l, cmdlen, skip_or; /* skip rest of OR block */ 1917 struct ip_fw *f; 1918 1919 f = chain->map[f_pos]; 1920 if (V_set_disable & (1 << f->set) ) 1921 continue; 1922 1923 skip_or = 0; 1924 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1925 l -= cmdlen, cmd += cmdlen) { 1926 int match; 1927 1928 /* 1929 * check_body is a jump target used when we find a 1930 * CHECK_STATE, and need to jump to the body of 1931 * the target rule. 1932 */ 1933 1934 /* check_body: */ 1935 cmdlen = F_LEN(cmd); 1936 /* 1937 * An OR block (insn_1 || .. || insn_n) has the 1938 * F_OR bit set in all but the last instruction. 1939 * The first match will set "skip_or", and cause 1940 * the following instructions to be skipped until 1941 * past the one with the F_OR bit clear. 1942 */ 1943 if (skip_or) { /* skip this instruction */ 1944 if ((cmd->len & F_OR) == 0) 1945 skip_or = 0; /* next one is good */ 1946 continue; 1947 } 1948 match = 0; /* set to 1 if we succeed */ 1949 1950 switch (cmd->opcode) { 1951 /* 1952 * The first set of opcodes compares the packet's 1953 * fields with some pattern, setting 'match' if a 1954 * match is found. At the end of the loop there is 1955 * logic to deal with F_NOT and F_OR flags associated 1956 * with the opcode. 1957 */ 1958 case O_NOP: 1959 match = 1; 1960 break; 1961 1962 case O_FORWARD_MAC: 1963 printf("ipfw: opcode %d unimplemented\n", 1964 cmd->opcode); 1965 break; 1966 1967 case O_GID: 1968 case O_UID: 1969 case O_JAIL: 1970 /* 1971 * We only check offset == 0 && proto != 0, 1972 * as this ensures that we have a 1973 * packet with the ports info. 1974 */ 1975 if (offset != 0) 1976 break; 1977 if (proto == IPPROTO_TCP || 1978 proto == IPPROTO_UDP || 1979 proto == IPPROTO_UDPLITE) 1980 match = check_uidgid( 1981 (ipfw_insn_u32 *)cmd, 1982 args, &ucred_lookup, 1983 #ifdef __FreeBSD__ 1984 &ucred_cache); 1985 #else 1986 (void *)&ucred_cache); 1987 #endif 1988 break; 1989 1990 case O_RECV: 1991 match = iface_match(iif, (ipfw_insn_if *)cmd, 1992 chain, &tablearg); 1993 break; 1994 1995 case O_XMIT: 1996 match = iface_match(oif, (ipfw_insn_if *)cmd, 1997 chain, &tablearg); 1998 break; 1999 2000 case O_VIA: 2001 match = iface_match(args->ifp, 2002 (ipfw_insn_if *)cmd, chain, &tablearg); 2003 break; 2004 2005 case O_MACADDR2: 2006 if (args->flags & IPFW_ARGS_ETHER) { 2007 u_int32_t *want = (u_int32_t *) 2008 ((ipfw_insn_mac *)cmd)->addr; 2009 u_int32_t *mask = (u_int32_t *) 2010 ((ipfw_insn_mac *)cmd)->mask; 2011 u_int32_t *hdr = (u_int32_t *)eh; 2012 2013 match = 2014 ( want[0] == (hdr[0] & mask[0]) && 2015 want[1] == (hdr[1] & mask[1]) && 2016 want[2] == (hdr[2] & mask[2]) ); 2017 } 2018 break; 2019 2020 case O_MAC_TYPE: 2021 if (args->flags & IPFW_ARGS_ETHER) { 2022 u_int16_t *p = 2023 ((ipfw_insn_u16 *)cmd)->ports; 2024 int i; 2025 2026 for (i = cmdlen - 1; !match && i>0; 2027 i--, p += 2) 2028 match = 2029 (ntohs(eh->ether_type) >= 2030 p[0] && 2031 ntohs(eh->ether_type) <= 2032 p[1]); 2033 } 2034 break; 2035 2036 case O_FRAG: 2037 if (is_ipv4) { 2038 /* 2039 * Since flags_match() works with 2040 * uint8_t we pack ip_off into 8 bits. 2041 * For this match offset is a boolean. 2042 */ 2043 match = flags_match(cmd, 2044 ((ntohs(ip->ip_off) & ~IP_OFFMASK) 2045 >> 8) | (offset != 0)); 2046 } else { 2047 /* 2048 * Compatibility: historically bare 2049 * "frag" would match IPv6 fragments. 2050 */ 2051 match = (cmd->arg1 == 0x1 && 2052 (offset != 0)); 2053 } 2054 break; 2055 2056 case O_IN: /* "out" is "not in" */ 2057 match = (oif == NULL); 2058 break; 2059 2060 case O_LAYER2: 2061 match = (args->flags & IPFW_ARGS_ETHER); 2062 break; 2063 2064 case O_DIVERTED: 2065 if ((args->flags & IPFW_ARGS_REF) == 0) 2066 break; 2067 /* 2068 * For diverted packets, args->rule.info 2069 * contains the divert port (in host format) 2070 * reason and direction. 2071 */ 2072 match = ((args->rule.info & IPFW_IS_MASK) == 2073 IPFW_IS_DIVERT) && ( 2074 ((args->rule.info & IPFW_INFO_IN) ? 2075 1: 2) & cmd->arg1); 2076 break; 2077 2078 case O_PROTO: 2079 /* 2080 * We do not allow an arg of 0 so the 2081 * check of "proto" only suffices. 2082 */ 2083 match = (proto == cmd->arg1); 2084 break; 2085 2086 case O_IP_SRC: 2087 match = is_ipv4 && 2088 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2089 src_ip.s_addr); 2090 break; 2091 2092 case O_IP_DST_LOOKUP: 2093 if (IPFW_LOOKUP_TYPE(cmd) != LOOKUP_NONE) { 2094 void *pkey = NULL; 2095 uint32_t key, vidx; 2096 uint16_t keylen = 0; /* zero if can't match the packet */ 2097 uint8_t lookup_type; 2098 2099 lookup_type = IPFW_LOOKUP_TYPE(cmd); 2100 2101 switch (lookup_type) { 2102 case LOOKUP_DST_IP: 2103 case LOOKUP_SRC_IP: 2104 if (is_ipv4) { 2105 keylen = sizeof(in_addr_t); 2106 if (lookup_type == LOOKUP_DST_IP) 2107 pkey = &dst_ip; 2108 else 2109 pkey = &src_ip; 2110 } else if (is_ipv6) { 2111 keylen = sizeof(struct in6_addr); 2112 if (lookup_type == LOOKUP_DST_IP) 2113 pkey = &args->f_id.dst_ip6; 2114 else 2115 pkey = &args->f_id.src_ip6; 2116 } else /* only for L3 */ 2117 break; 2118 case LOOKUP_DSCP: 2119 if (is_ipv4) 2120 key = ip->ip_tos >> 2; 2121 else if (is_ipv6) 2122 key = IPV6_DSCP( 2123 (struct ip6_hdr *)ip) >> 2; 2124 else 2125 break; /* only for L3 */ 2126 2127 key &= 0x3f; 2128 if (cmdlen == F_INSN_SIZE(ipfw_insn_table)) 2129 key &= insntod(cmd, table)->value; 2130 pkey = &key; 2131 keylen = sizeof(key); 2132 break; 2133 case LOOKUP_DST_PORT: 2134 case LOOKUP_SRC_PORT: 2135 /* only for L3 */ 2136 if (is_ipv6 == 0 && is_ipv4 == 0) { 2137 break; 2138 } 2139 /* Skip fragments */ 2140 if (offset != 0) { 2141 break; 2142 } 2143 /* Skip proto without ports */ 2144 if (proto != IPPROTO_TCP && 2145 proto != IPPROTO_UDP && 2146 proto != IPPROTO_UDPLITE && 2147 proto != IPPROTO_SCTP) 2148 break; 2149 if (lookup_type == LOOKUP_DST_PORT) 2150 key = dst_port; 2151 else 2152 key = src_port; 2153 pkey = &key; 2154 if (cmdlen == F_INSN_SIZE(ipfw_insn_table)) 2155 key &= insntod(cmd, table)->value; 2156 keylen = sizeof(key); 2157 break; 2158 case LOOKUP_DST_MAC: 2159 case LOOKUP_SRC_MAC: 2160 /* only for L2 */ 2161 if ((args->flags & IPFW_ARGS_ETHER) == 0) 2162 break; 2163 2164 pkey = lookup_type == LOOKUP_DST_MAC ? 2165 eh->ether_dhost : eh->ether_shost; 2166 keylen = ETHER_ADDR_LEN; 2167 break; 2168 #ifndef USERSPACE 2169 case LOOKUP_UID: 2170 case LOOKUP_JAIL: 2171 check_uidgid(insntod(cmd, u32), 2172 args, &ucred_lookup, 2173 #ifdef __FreeBSD__ 2174 &ucred_cache); 2175 if (lookup_type == LOOKUP_UID) 2176 key = ucred_cache->cr_uid; 2177 else if (lookup_type == LOOKUP_JAIL) 2178 key = ucred_cache->cr_prison->pr_id; 2179 #else /* !__FreeBSD__ */ 2180 (void *)&ucred_cache); 2181 if (lookup_type == LOOKUP_UID) 2182 key = ucred_cache.uid; 2183 else if (lookup_type == LOOKUP_JAIL) 2184 key = ucred_cache.xid; 2185 #endif /* !__FreeBSD__ */ 2186 pkey = &key; 2187 if (cmdlen == F_INSN_SIZE(ipfw_insn_table)) 2188 key &= insntod(cmd, table)->value; 2189 keylen = sizeof(key); 2190 break; 2191 #endif /* !USERSPACE */ 2192 case LOOKUP_MARK: 2193 key = args->rule.pkt_mark; 2194 if (cmdlen == F_INSN_SIZE(ipfw_insn_table)) 2195 key &= insntod(cmd, table)->value; 2196 pkey = &key; 2197 keylen = sizeof(key); 2198 break; 2199 case LOOKUP_RULENUM: 2200 key = f->rulenum; 2201 if (cmdlen == F_INSN_SIZE(ipfw_insn_table)) 2202 key &= insntod(cmd, table)->value; 2203 pkey = &key; 2204 keylen = sizeof(key); 2205 break; 2206 } 2207 /* unknown key type */ 2208 if (keylen == 0) 2209 break; 2210 match = ipfw_lookup_table(chain, 2211 insntod(cmd, kidx)->kidx, keylen, 2212 pkey, &vidx); 2213 2214 if (match) 2215 tablearg = vidx; 2216 break; 2217 } 2218 /* LOOKUP_NONE */ 2219 /* FALLTHROUGH */ 2220 case O_IP_SRC_LOOKUP: 2221 { 2222 void *pkey; 2223 uint32_t vidx; 2224 uint16_t keylen; 2225 2226 if (is_ipv4) { 2227 keylen = sizeof(in_addr_t); 2228 if (cmd->opcode == O_IP_DST_LOOKUP) 2229 pkey = &dst_ip; 2230 else 2231 pkey = &src_ip; 2232 } else if (is_ipv6) { 2233 keylen = sizeof(struct in6_addr); 2234 if (cmd->opcode == O_IP_DST_LOOKUP) 2235 pkey = &args->f_id.dst_ip6; 2236 else 2237 pkey = &args->f_id.src_ip6; 2238 } else 2239 break; 2240 match = ipfw_lookup_table(chain, 2241 insntod(cmd, kidx)->kidx, 2242 keylen, pkey, &vidx); 2243 if (!match) 2244 break; 2245 if (cmdlen == F_INSN_SIZE(ipfw_insn_table)) { 2246 match = tvalue_match(chain, 2247 insntod(cmd, table), vidx); 2248 if (!match) 2249 break; 2250 } 2251 tablearg = vidx; 2252 break; 2253 } 2254 2255 case O_MAC_SRC_LOOKUP: 2256 case O_MAC_DST_LOOKUP: 2257 { 2258 void *pkey; 2259 uint32_t vidx; 2260 uint16_t keylen = ETHER_ADDR_LEN; 2261 2262 /* Need ether frame */ 2263 if ((args->flags & IPFW_ARGS_ETHER) == 0) 2264 break; 2265 2266 if (cmd->opcode == O_MAC_DST_LOOKUP) 2267 pkey = eh->ether_dhost; 2268 else 2269 pkey = eh->ether_shost; 2270 2271 match = ipfw_lookup_table(chain, 2272 insntod(cmd, kidx)->kidx, 2273 keylen, pkey, &vidx); 2274 if (!match) 2275 break; 2276 if (cmdlen == F_INSN_SIZE(ipfw_insn_table)) { 2277 match = tvalue_match(chain, 2278 insntod(cmd, table), vidx); 2279 if (!match) 2280 break; 2281 } 2282 tablearg = vidx; 2283 break; 2284 } 2285 2286 case O_IP_FLOW_LOOKUP: 2287 { 2288 uint32_t vidx = 0; 2289 2290 match = ipfw_lookup_table(chain, 2291 insntod(cmd, kidx)->kidx, 0, 2292 &args->f_id, &vidx); 2293 if (!match) 2294 break; 2295 if (cmdlen == F_INSN_SIZE(ipfw_insn_table)) 2296 match = tvalue_match(chain, 2297 insntod(cmd, table), vidx); 2298 if (match) 2299 tablearg = vidx; 2300 break; 2301 } 2302 2303 case O_IP_SRC_MASK: 2304 case O_IP_DST_MASK: 2305 if (is_ipv4) { 2306 uint32_t a = 2307 (cmd->opcode == O_IP_DST_MASK) ? 2308 dst_ip.s_addr : src_ip.s_addr; 2309 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2310 int i = cmdlen-1; 2311 2312 for (; !match && i>0; i-= 2, p+= 2) 2313 match = (p[0] == (a & p[1])); 2314 } 2315 break; 2316 2317 case O_IP_SRC_ME: 2318 if (is_ipv4) { 2319 match = in_localip(src_ip); 2320 break; 2321 } 2322 #ifdef INET6 2323 /* FALLTHROUGH */ 2324 case O_IP6_SRC_ME: 2325 match = is_ipv6 && 2326 ipfw_localip6(&args->f_id.src_ip6); 2327 #endif 2328 break; 2329 2330 case O_IP_DST_SET: 2331 case O_IP_SRC_SET: 2332 if (is_ipv4) { 2333 u_int32_t *d = (u_int32_t *)(cmd+1); 2334 u_int32_t addr = 2335 cmd->opcode == O_IP_DST_SET ? 2336 args->f_id.dst_ip : 2337 args->f_id.src_ip; 2338 2339 if (addr < d[0]) 2340 break; 2341 addr -= d[0]; /* subtract base */ 2342 match = (addr < cmd->arg1) && 2343 ( d[ 1 + (addr>>5)] & 2344 (1<<(addr & 0x1f)) ); 2345 } 2346 break; 2347 2348 case O_IP_DST: 2349 match = is_ipv4 && 2350 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2351 dst_ip.s_addr); 2352 break; 2353 2354 case O_IP_DST_ME: 2355 if (is_ipv4) { 2356 match = in_localip(dst_ip); 2357 break; 2358 } 2359 #ifdef INET6 2360 /* FALLTHROUGH */ 2361 case O_IP6_DST_ME: 2362 match = is_ipv6 && 2363 ipfw_localip6(&args->f_id.dst_ip6); 2364 #endif 2365 break; 2366 2367 case O_IP_SRCPORT: 2368 case O_IP_DSTPORT: 2369 /* 2370 * offset == 0 && proto != 0 is enough 2371 * to guarantee that we have a 2372 * packet with port info. 2373 */ 2374 if ((proto == IPPROTO_UDP || 2375 proto == IPPROTO_UDPLITE || 2376 proto == IPPROTO_TCP || 2377 proto == IPPROTO_SCTP) && offset == 0) { 2378 u_int16_t x = 2379 (cmd->opcode == O_IP_SRCPORT) ? 2380 src_port : dst_port ; 2381 u_int16_t *p = 2382 ((ipfw_insn_u16 *)cmd)->ports; 2383 int i; 2384 2385 for (i = cmdlen - 1; !match && i>0; 2386 i--, p += 2) 2387 match = (x>=p[0] && x<=p[1]); 2388 } 2389 break; 2390 2391 case O_ICMPTYPE: 2392 match = (offset == 0 && proto==IPPROTO_ICMP && 2393 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2394 break; 2395 2396 #ifdef INET6 2397 case O_ICMP6TYPE: 2398 match = is_ipv6 && offset == 0 && 2399 proto==IPPROTO_ICMPV6 && 2400 icmp6type_match( 2401 ICMP6(ulp)->icmp6_type, 2402 (ipfw_insn_u32 *)cmd); 2403 break; 2404 #endif /* INET6 */ 2405 2406 case O_IPOPT: 2407 match = (is_ipv4 && 2408 ipopts_match(ip, cmd) ); 2409 break; 2410 2411 case O_IPVER: 2412 match = ((is_ipv4 || is_ipv6) && 2413 cmd->arg1 == ip->ip_v); 2414 break; 2415 2416 case O_IPID: 2417 case O_IPTTL: 2418 if (!is_ipv4) 2419 break; 2420 case O_IPLEN: 2421 { /* only for IP packets */ 2422 uint16_t x; 2423 uint16_t *p; 2424 int i; 2425 2426 if (cmd->opcode == O_IPLEN) 2427 x = iplen; 2428 else if (cmd->opcode == O_IPTTL) 2429 x = ip->ip_ttl; 2430 else /* must be IPID */ 2431 x = ntohs(ip->ip_id); 2432 if (cmdlen == 1) { 2433 match = (cmd->arg1 == x); 2434 break; 2435 } 2436 /* otherwise we have ranges */ 2437 p = ((ipfw_insn_u16 *)cmd)->ports; 2438 i = cmdlen - 1; 2439 for (; !match && i>0; i--, p += 2) 2440 match = (x >= p[0] && x <= p[1]); 2441 } 2442 break; 2443 2444 case O_IPPRECEDENCE: 2445 match = (is_ipv4 && 2446 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2447 break; 2448 2449 case O_IPTOS: 2450 match = (is_ipv4 && 2451 flags_match(cmd, ip->ip_tos)); 2452 break; 2453 2454 case O_DSCP: 2455 { 2456 uint32_t *p; 2457 uint16_t x; 2458 2459 p = ((ipfw_insn_u32 *)cmd)->d; 2460 2461 if (is_ipv4) 2462 x = ip->ip_tos >> 2; 2463 else if (is_ipv6) { 2464 x = IPV6_DSCP( 2465 (struct ip6_hdr *)ip) >> 2; 2466 x &= 0x3f; 2467 } else 2468 break; 2469 2470 /* DSCP bitmask is stored as low_u32 high_u32 */ 2471 if (x >= 32) 2472 match = *(p + 1) & (1 << (x - 32)); 2473 else 2474 match = *p & (1 << x); 2475 } 2476 break; 2477 2478 case O_TCPDATALEN: 2479 if (proto == IPPROTO_TCP && offset == 0) { 2480 struct tcphdr *tcp; 2481 uint16_t x; 2482 uint16_t *p; 2483 int i; 2484 #ifdef INET6 2485 if (is_ipv6) { 2486 struct ip6_hdr *ip6; 2487 2488 ip6 = (struct ip6_hdr *)ip; 2489 if (ip6->ip6_plen == 0) { 2490 /* 2491 * Jumbo payload is not 2492 * supported by this 2493 * opcode. 2494 */ 2495 break; 2496 } 2497 x = iplen - hlen; 2498 } else 2499 #endif /* INET6 */ 2500 x = iplen - (ip->ip_hl << 2); 2501 tcp = TCP(ulp); 2502 x -= tcp->th_off << 2; 2503 if (cmdlen == 1) { 2504 match = (cmd->arg1 == x); 2505 break; 2506 } 2507 /* otherwise we have ranges */ 2508 p = ((ipfw_insn_u16 *)cmd)->ports; 2509 i = cmdlen - 1; 2510 for (; !match && i>0; i--, p += 2) 2511 match = (x >= p[0] && x <= p[1]); 2512 } 2513 break; 2514 2515 case O_TCPFLAGS: 2516 /* 2517 * Note that this is currently only set up to 2518 * match the lower 8 TCP header flag bits, not 2519 * the full compliment of all 12 flags. 2520 */ 2521 match = (proto == IPPROTO_TCP && offset == 0 && 2522 flags_match(cmd, tcp_get_flags(TCP(ulp)))); 2523 break; 2524 2525 case O_TCPOPTS: 2526 if (proto == IPPROTO_TCP && offset == 0 && ulp){ 2527 PULLUP_LEN_LOCKED(hlen, ulp, 2528 (TCP(ulp)->th_off << 2)); 2529 match = tcpopts_match(TCP(ulp), cmd); 2530 } 2531 break; 2532 2533 case O_TCPSEQ: 2534 match = (proto == IPPROTO_TCP && offset == 0 && 2535 ((ipfw_insn_u32 *)cmd)->d[0] == 2536 TCP(ulp)->th_seq); 2537 break; 2538 2539 case O_TCPACK: 2540 match = (proto == IPPROTO_TCP && offset == 0 && 2541 ((ipfw_insn_u32 *)cmd)->d[0] == 2542 TCP(ulp)->th_ack); 2543 break; 2544 2545 case O_TCPMSS: 2546 if (proto == IPPROTO_TCP && 2547 (args->f_id._flags & TH_SYN) != 0 && 2548 ulp != NULL) { 2549 uint16_t mss, *p; 2550 int i; 2551 2552 PULLUP_LEN_LOCKED(hlen, ulp, 2553 (TCP(ulp)->th_off << 2)); 2554 if ((tcpopts_parse(TCP(ulp), &mss) & 2555 IP_FW_TCPOPT_MSS) == 0) 2556 break; 2557 if (cmdlen == 1) { 2558 match = (cmd->arg1 == mss); 2559 break; 2560 } 2561 /* Otherwise we have ranges. */ 2562 p = ((ipfw_insn_u16 *)cmd)->ports; 2563 i = cmdlen - 1; 2564 for (; !match && i > 0; i--, p += 2) 2565 match = (mss >= p[0] && 2566 mss <= p[1]); 2567 } 2568 break; 2569 2570 case O_TCPWIN: 2571 if (proto == IPPROTO_TCP && offset == 0) { 2572 uint16_t x; 2573 uint16_t *p; 2574 int i; 2575 2576 x = ntohs(TCP(ulp)->th_win); 2577 if (cmdlen == 1) { 2578 match = (cmd->arg1 == x); 2579 break; 2580 } 2581 /* Otherwise we have ranges. */ 2582 p = ((ipfw_insn_u16 *)cmd)->ports; 2583 i = cmdlen - 1; 2584 for (; !match && i > 0; i--, p += 2) 2585 match = (x >= p[0] && x <= p[1]); 2586 } 2587 break; 2588 2589 case O_ESTAB: 2590 /* reject packets which have SYN only */ 2591 /* XXX should i also check for TH_ACK ? */ 2592 match = (proto == IPPROTO_TCP && offset == 0 && 2593 (tcp_get_flags(TCP(ulp)) & 2594 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2595 break; 2596 2597 case O_ALTQ: { 2598 struct pf_mtag *at; 2599 struct m_tag *mtag; 2600 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2601 2602 /* 2603 * ALTQ uses mbuf tags from another 2604 * packet filtering system - pf(4). 2605 * We allocate a tag in its format 2606 * and fill it in, pretending to be pf(4). 2607 */ 2608 match = 1; 2609 at = pf_find_mtag(m); 2610 if (at != NULL && at->qid != 0) 2611 break; 2612 mtag = m_tag_get(PACKET_TAG_PF, 2613 sizeof(struct pf_mtag), M_NOWAIT | M_ZERO); 2614 if (mtag == NULL) { 2615 /* 2616 * Let the packet fall back to the 2617 * default ALTQ. 2618 */ 2619 break; 2620 } 2621 m_tag_prepend(m, mtag); 2622 at = (struct pf_mtag *)(mtag + 1); 2623 at->qid = altq->qid; 2624 at->hdr = ip; 2625 break; 2626 } 2627 2628 case O_LOG: 2629 ipfw_log(chain, f, hlen, args, 2630 offset | ip6f_mf, tablearg, ip, eh); 2631 match = 1; 2632 break; 2633 2634 case O_PROB: 2635 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2636 break; 2637 2638 case O_VERREVPATH: 2639 /* Outgoing packets automatically pass/match */ 2640 match = (args->flags & IPFW_ARGS_OUT || 2641 ( 2642 #ifdef INET6 2643 is_ipv6 ? 2644 verify_path6(&(args->f_id.src_ip6), 2645 iif, args->f_id.fib) : 2646 #endif 2647 verify_path(src_ip, iif, args->f_id.fib))); 2648 break; 2649 2650 case O_VERSRCREACH: 2651 /* Outgoing packets automatically pass/match */ 2652 match = (hlen > 0 && ((oif != NULL) || ( 2653 #ifdef INET6 2654 is_ipv6 ? 2655 verify_path6(&(args->f_id.src_ip6), 2656 NULL, args->f_id.fib) : 2657 #endif 2658 verify_path(src_ip, NULL, args->f_id.fib)))); 2659 break; 2660 2661 case O_ANTISPOOF: 2662 /* Outgoing packets automatically pass/match */ 2663 if (oif == NULL && hlen > 0 && 2664 ( (is_ipv4 && in_localaddr(src_ip)) 2665 #ifdef INET6 2666 || (is_ipv6 && 2667 in6_localaddr(&(args->f_id.src_ip6))) 2668 #endif 2669 )) 2670 match = 2671 #ifdef INET6 2672 is_ipv6 ? verify_path6( 2673 &(args->f_id.src_ip6), iif, 2674 args->f_id.fib) : 2675 #endif 2676 verify_path(src_ip, iif, 2677 args->f_id.fib); 2678 else 2679 match = 1; 2680 break; 2681 2682 case O_IPSEC: 2683 match = (m_tag_find(m, 2684 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2685 /* otherwise no match */ 2686 break; 2687 2688 #ifdef INET6 2689 case O_IP6_SRC: 2690 match = is_ipv6 && 2691 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2692 &((ipfw_insn_ip6 *)cmd)->addr6); 2693 break; 2694 2695 case O_IP6_DST: 2696 match = is_ipv6 && 2697 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2698 &((ipfw_insn_ip6 *)cmd)->addr6); 2699 break; 2700 case O_IP6_SRC_MASK: 2701 case O_IP6_DST_MASK: 2702 if (is_ipv6) { 2703 int i = cmdlen - 1; 2704 struct in6_addr p; 2705 struct in6_addr *d = 2706 &((ipfw_insn_ip6 *)cmd)->addr6; 2707 2708 for (; !match && i > 0; d += 2, 2709 i -= F_INSN_SIZE(struct in6_addr) 2710 * 2) { 2711 p = (cmd->opcode == 2712 O_IP6_SRC_MASK) ? 2713 args->f_id.src_ip6: 2714 args->f_id.dst_ip6; 2715 APPLY_MASK(&p, &d[1]); 2716 match = 2717 IN6_ARE_ADDR_EQUAL(&d[0], 2718 &p); 2719 } 2720 } 2721 break; 2722 2723 case O_FLOW6ID: 2724 match = is_ipv6 && 2725 flow6id_match(args->f_id.flow_id6, 2726 (ipfw_insn_u32 *) cmd); 2727 break; 2728 2729 case O_EXT_HDR: 2730 match = is_ipv6 && 2731 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2732 break; 2733 2734 case O_IP6: 2735 match = is_ipv6; 2736 break; 2737 #endif 2738 2739 case O_IP4: 2740 match = is_ipv4; 2741 break; 2742 2743 case O_TAG: { 2744 struct m_tag *mtag; 2745 uint32_t tag = TARG(cmd->arg1, tag); 2746 2747 /* Packet is already tagged with this tag? */ 2748 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 2749 2750 /* We have `untag' action when F_NOT flag is 2751 * present. And we must remove this mtag from 2752 * mbuf and reset `match' to zero (`match' will 2753 * be inversed later). 2754 * Otherwise we should allocate new mtag and 2755 * push it into mbuf. 2756 */ 2757 if (cmd->len & F_NOT) { /* `untag' action */ 2758 if (mtag != NULL) 2759 m_tag_delete(m, mtag); 2760 match = 0; 2761 } else { 2762 if (mtag == NULL) { 2763 mtag = m_tag_alloc( MTAG_IPFW, 2764 tag, 0, M_NOWAIT); 2765 if (mtag != NULL) 2766 m_tag_prepend(m, mtag); 2767 } 2768 match = 1; 2769 } 2770 break; 2771 } 2772 2773 case O_FIB: /* try match the specified fib */ 2774 if (args->f_id.fib == cmd->arg1) 2775 match = 1; 2776 break; 2777 2778 case O_SOCKARG: { 2779 #ifndef USERSPACE /* not supported in userspace */ 2780 struct inpcb *inp = args->inp; 2781 struct inpcbinfo *pi; 2782 bool inp_locked = false; 2783 2784 if (proto == IPPROTO_TCP) 2785 pi = &V_tcbinfo; 2786 else if (proto == IPPROTO_UDP) 2787 pi = &V_udbinfo; 2788 else if (proto == IPPROTO_UDPLITE) 2789 pi = &V_ulitecbinfo; 2790 else 2791 break; 2792 2793 /* 2794 * XXXRW: so_user_cookie should almost 2795 * certainly be inp_user_cookie? 2796 */ 2797 2798 /* 2799 * For incoming packet lookup the inpcb 2800 * using the src/dest ip/port tuple. 2801 */ 2802 if (is_ipv4 && inp == NULL) { 2803 inp = in_pcblookup(pi, 2804 src_ip, htons(src_port), 2805 dst_ip, htons(dst_port), 2806 INPLOOKUP_RLOCKPCB, NULL); 2807 inp_locked = true; 2808 } 2809 #ifdef INET6 2810 if (is_ipv6 && inp == NULL) { 2811 inp = in6_pcblookup(pi, 2812 &args->f_id.src_ip6, 2813 htons(src_port), 2814 &args->f_id.dst_ip6, 2815 htons(dst_port), 2816 INPLOOKUP_RLOCKPCB, NULL); 2817 inp_locked = true; 2818 } 2819 #endif /* INET6 */ 2820 if (inp != NULL) { 2821 if (inp->inp_socket) { 2822 tablearg = 2823 inp->inp_socket->so_user_cookie; 2824 if (tablearg) 2825 match = 1; 2826 } 2827 if (inp_locked) 2828 INP_RUNLOCK(inp); 2829 } 2830 #endif /* !USERSPACE */ 2831 break; 2832 } 2833 2834 case O_TAGGED: { 2835 struct m_tag *mtag; 2836 uint32_t tag = TARG(cmd->arg1, tag); 2837 2838 if (cmdlen == 1) { 2839 match = m_tag_locate(m, MTAG_IPFW, 2840 tag, NULL) != NULL; 2841 break; 2842 } 2843 2844 /* we have ranges */ 2845 for (mtag = m_tag_first(m); 2846 mtag != NULL && !match; 2847 mtag = m_tag_next(m, mtag)) { 2848 uint16_t *p; 2849 int i; 2850 2851 if (mtag->m_tag_cookie != MTAG_IPFW) 2852 continue; 2853 2854 p = ((ipfw_insn_u16 *)cmd)->ports; 2855 i = cmdlen - 1; 2856 for(; !match && i > 0; i--, p += 2) 2857 match = 2858 mtag->m_tag_id >= p[0] && 2859 mtag->m_tag_id <= p[1]; 2860 } 2861 break; 2862 } 2863 2864 case O_MARK: { 2865 uint32_t mark; 2866 if (cmd->arg1 == IP_FW_TARG) 2867 mark = TARG_VAL(chain, tablearg, mark); 2868 else 2869 mark = insntoc(cmd, u32)->d[0]; 2870 match = 2871 (args->rule.pkt_mark & 2872 insntoc(cmd, u32)->d[1]) == 2873 (mark & insntoc(cmd, u32)->d[1]); 2874 break; 2875 } 2876 2877 /* 2878 * The second set of opcodes represents 'actions', 2879 * i.e. the terminal part of a rule once the packet 2880 * matches all previous patterns. 2881 * Typically there is only one action for each rule, 2882 * and the opcode is stored at the end of the rule 2883 * (but there are exceptions -- see below). 2884 * 2885 * In general, here we set retval and terminate the 2886 * outer loop (would be a 'break 3' in some language, 2887 * but we need to set l=0, done=1) 2888 * 2889 * Exceptions: 2890 * O_COUNT and O_SKIPTO actions: 2891 * instead of terminating, we jump to the next rule 2892 * (setting l=0), or to the SKIPTO target (setting 2893 * f/f_len, cmd and l as needed), respectively. 2894 * 2895 * O_TAG, O_LOG and O_ALTQ action parameters: 2896 * perform some action and set match = 1; 2897 * 2898 * O_LIMIT and O_KEEP_STATE: these opcodes are 2899 * not real 'actions', and are stored right 2900 * before the 'action' part of the rule (one 2901 * exception is O_SKIP_ACTION which could be 2902 * between these opcodes and 'action' one). 2903 * These opcodes try to install an entry in the 2904 * state tables; if successful, we continue with 2905 * the next opcode (match=1; break;), otherwise 2906 * the packet must be dropped (set retval, 2907 * break loops with l=0, done=1) 2908 * 2909 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2910 * cause a lookup of the state table, and a jump 2911 * to the 'action' part of the parent rule 2912 * if an entry is found, or 2913 * (CHECK_STATE only) a jump to the next rule if 2914 * the entry is not found. 2915 * The result of the lookup is cached so that 2916 * further instances of these opcodes become NOPs. 2917 * The jump to the next rule is done by setting 2918 * l=0, cmdlen=0. 2919 * 2920 * O_SKIP_ACTION: this opcode is not a real 'action' 2921 * either, and is stored right before the 'action' 2922 * part of the rule, right after the O_KEEP_STATE 2923 * opcode. It causes match failure so the real 2924 * 'action' could be executed only if the rule 2925 * is checked via dynamic rule from the state 2926 * table, as in such case execution starts 2927 * from the true 'action' opcode directly. 2928 * 2929 */ 2930 case O_LIMIT: 2931 case O_KEEP_STATE: 2932 if (ipfw_dyn_install_state(chain, f, 2933 (ipfw_insn_limit *)cmd, args, ulp, 2934 pktlen, &dyn_info, tablearg)) { 2935 /* error or limit violation */ 2936 retval = IP_FW_DENY; 2937 l = 0; /* exit inner loop */ 2938 done = 1; /* exit outer loop */ 2939 } 2940 match = 1; 2941 break; 2942 2943 case O_PROBE_STATE: 2944 case O_CHECK_STATE: 2945 /* 2946 * dynamic rules are checked at the first 2947 * keep-state or check-state occurrence, 2948 * with the result being stored in dyn_info. 2949 * The compiler introduces a PROBE_STATE 2950 * instruction for us when we have a 2951 * KEEP_STATE (because PROBE_STATE needs 2952 * to be run first). 2953 */ 2954 if (DYN_LOOKUP_NEEDED(&dyn_info, cmd) && 2955 (q = ipfw_dyn_lookup_state(args, ulp, 2956 pktlen, cmd, &dyn_info)) != NULL) { 2957 /* 2958 * Found dynamic entry, jump to the 2959 * 'action' part of the parent rule 2960 * by setting f, cmd, l and clearing 2961 * cmdlen. 2962 */ 2963 f = q; 2964 f_pos = dyn_info.f_pos; 2965 cmd = ACTION_PTR(f); 2966 l = f->cmd_len - f->act_ofs; 2967 cmdlen = 0; 2968 continue; 2969 } 2970 /* 2971 * Dynamic entry not found. If CHECK_STATE, 2972 * skip to next rule, if PROBE_STATE just 2973 * ignore and continue with next opcode. 2974 */ 2975 if (cmd->opcode == O_CHECK_STATE) 2976 l = 0; /* exit inner loop */ 2977 match = 1; 2978 break; 2979 2980 case O_SKIP_ACTION: 2981 match = 0; /* skip to the next rule */ 2982 l = 0; /* exit inner loop */ 2983 break; 2984 2985 case O_ACCEPT: 2986 retval = 0; /* accept */ 2987 l = 0; /* exit inner loop */ 2988 done = 1; /* exit outer loop */ 2989 break; 2990 2991 case O_PIPE: 2992 case O_QUEUE: 2993 set_match(args, f_pos, chain); 2994 args->rule.info = TARG(cmd->arg1, pipe); 2995 if (cmd->opcode == O_PIPE) 2996 args->rule.info |= IPFW_IS_PIPE; 2997 if (V_fw_one_pass) 2998 args->rule.info |= IPFW_ONEPASS; 2999 retval = IP_FW_DUMMYNET; 3000 l = 0; /* exit inner loop */ 3001 done = 1; /* exit outer loop */ 3002 break; 3003 3004 case O_DIVERT: 3005 case O_TEE: 3006 if (args->flags & IPFW_ARGS_ETHER) 3007 break; /* not on layer 2 */ 3008 /* otherwise this is terminal */ 3009 l = 0; /* exit inner loop */ 3010 done = 1; /* exit outer loop */ 3011 retval = (cmd->opcode == O_DIVERT) ? 3012 IP_FW_DIVERT : IP_FW_TEE; 3013 set_match(args, f_pos, chain); 3014 args->rule.info = TARG(cmd->arg1, divert); 3015 break; 3016 3017 case O_COUNT: 3018 IPFW_INC_RULE_COUNTER(f, pktlen); 3019 l = 0; /* exit inner loop */ 3020 break; 3021 3022 case O_SKIPTO: 3023 IPFW_INC_RULE_COUNTER(f, pktlen); 3024 f_pos = jump(chain, f, 3025 insntod(cmd, u32)->d[0], tablearg, false); 3026 /* 3027 * Skip disabled rules, and re-enter 3028 * the inner loop with the correct 3029 * f_pos, f, l and cmd. 3030 * Also clear cmdlen and skip_or 3031 */ 3032 for (; f_pos < chain->n_rules - 1 && 3033 (V_set_disable & 3034 (1 << chain->map[f_pos]->set)); 3035 f_pos++) 3036 ; 3037 /* Re-enter the inner loop at the skipto rule. */ 3038 f = chain->map[f_pos]; 3039 l = f->cmd_len; 3040 cmd = f->cmd; 3041 match = 1; 3042 cmdlen = 0; 3043 skip_or = 0; 3044 continue; 3045 break; /* not reached */ 3046 3047 case O_CALLRETURN: { 3048 /* 3049 * Implementation of `subroutine' call/return, 3050 * in the stack carried in an mbuf tag. This 3051 * is different from `skipto' in that any call 3052 * address is possible (`skipto' must prevent 3053 * backward jumps to avoid endless loops). 3054 * We have `return' action when F_NOT flag is 3055 * present. The `m_tag_id' field is used as 3056 * stack pointer. 3057 */ 3058 struct m_tag *mtag; 3059 uint32_t jmpto, *stack; 3060 3061 #define IS_CALL ((cmd->len & F_NOT) == 0) 3062 #define IS_RETURN ((cmd->len & F_NOT) != 0) 3063 /* 3064 * Hand-rolled version of m_tag_locate() with 3065 * wildcard `type'. 3066 * If not already tagged, allocate new tag. 3067 */ 3068 mtag = m_tag_first(m); 3069 while (mtag != NULL) { 3070 if (mtag->m_tag_cookie == 3071 MTAG_IPFW_CALL) 3072 break; 3073 mtag = m_tag_next(m, mtag); 3074 } 3075 3076 /* 3077 * We keep ruleset id in the first element 3078 * of stack. If it doesn't match chain->id, 3079 * then we can't trust information in the 3080 * stack, since rules were changed. 3081 * We reset stack pointer to be able reuse 3082 * tag if it will be needed. 3083 */ 3084 if (mtag != NULL) { 3085 stack = (uint32_t *)(mtag + 1); 3086 if (stack[0] != chain->id) { 3087 stack[0] = chain->id; 3088 mtag->m_tag_id = 0; 3089 } 3090 } 3091 3092 /* 3093 * If there is no mtag or stack is empty, 3094 * `return` continues with next rule. 3095 */ 3096 if (IS_RETURN && (mtag == NULL || 3097 mtag->m_tag_id == 0)) { 3098 l = 0; /* exit inner loop */ 3099 break; 3100 } 3101 3102 if (mtag == NULL) { 3103 MPASS(IS_CALL); 3104 mtag = m_tag_alloc(MTAG_IPFW_CALL, 0, 3105 IPFW_CALLSTACK_SIZE * 3106 sizeof(uint32_t), M_NOWAIT); 3107 if (mtag != NULL) { 3108 m_tag_prepend(m, mtag); 3109 stack = (uint32_t *)(mtag + 1); 3110 stack[0] = chain->id; 3111 } 3112 } 3113 3114 if (mtag == NULL) { 3115 printf("ipfw: rule %u: failed to " 3116 "allocate call stack. " 3117 "Denying packet.\n", 3118 f->rulenum); 3119 l = 0; /* exit inner loop */ 3120 done = 1; /* exit outer loop */ 3121 retval = IP_FW_DENY; /* drop packet */ 3122 break; 3123 } 3124 3125 if (IS_CALL && mtag->m_tag_id >= 3126 IPFW_CALLSTACK_SIZE - 1) { 3127 printf("ipfw: rule %u: call stack " 3128 "overflow. Denying packet.\n", 3129 f->rulenum); 3130 l = 0; /* exit inner loop */ 3131 done = 1; /* exit outer loop */ 3132 retval = IP_FW_DENY; /* drop packet */ 3133 break; 3134 } 3135 3136 MPASS(stack == (uint32_t *)(mtag + 1)); 3137 IPFW_INC_RULE_COUNTER(f, pktlen); 3138 3139 if (IS_CALL) { 3140 stack[++mtag->m_tag_id] = f_pos; 3141 f_pos = jump(chain, f, 3142 insntod(cmd, u32)->d[0], 3143 tablearg, true); 3144 } else { /* `return' action */ 3145 jmpto = stack[mtag->m_tag_id--]; 3146 if (cmd->arg1 == RETURN_NEXT_RULE) 3147 f_pos = jmpto + 1; 3148 else /* RETURN_NEXT_RULENUM */ 3149 f_pos = ipfw_find_rule(chain, 3150 chain->map[ 3151 jmpto]->rulenum + 1, 0); 3152 } 3153 3154 /* 3155 * Skip disabled rules, and re-enter 3156 * the inner loop with the correct 3157 * f_pos, f, l and cmd. 3158 * Also clear cmdlen and skip_or 3159 */ 3160 MPASS(f_pos < chain->n_rules - 1); 3161 for (; f_pos < chain->n_rules - 1 && 3162 (V_set_disable & 3163 (1 << chain->map[f_pos]->set)); f_pos++) 3164 ; 3165 /* 3166 * Re-enter the inner loop at the dest 3167 * rule. 3168 */ 3169 f = chain->map[f_pos]; 3170 l = f->cmd_len; 3171 cmd = f->cmd; 3172 cmdlen = 0; 3173 skip_or = 0; 3174 continue; 3175 break; /* NOTREACHED */ 3176 } 3177 #undef IS_CALL 3178 #undef IS_RETURN 3179 3180 case O_REJECT: 3181 /* 3182 * Drop the packet and send a reject notice 3183 * if the packet is not ICMP (or is an ICMP 3184 * query), and it is not multicast/broadcast. 3185 */ 3186 if (hlen > 0 && is_ipv4 && offset == 0 && 3187 (proto != IPPROTO_ICMP || 3188 is_icmp_query(ICMP(ulp))) && 3189 !(m->m_flags & (M_BCAST|M_MCAST)) && 3190 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3191 KASSERT(!need_send_reject, 3192 ("o_reject - need_send_reject was set previously")); 3193 if ((reject_code = cmd->arg1) == ICMP_UNREACH_NEEDFRAG && 3194 cmd->len == F_INSN_SIZE(ipfw_insn_u16)) { 3195 reject_mtu = 3196 ((ipfw_insn_u16 *)cmd)->ports[0]; 3197 } else { 3198 reject_mtu = 0; 3199 } 3200 need_send_reject = true; 3201 } 3202 /* FALLTHROUGH */ 3203 #ifdef INET6 3204 case O_UNREACH6: 3205 if (hlen > 0 && is_ipv6 && 3206 ((offset & IP6F_OFF_MASK) == 0) && 3207 (proto != IPPROTO_ICMPV6 || 3208 (is_icmp6_query(icmp6_type) == 1)) && 3209 !(m->m_flags & (M_BCAST|M_MCAST)) && 3210 !IN6_IS_ADDR_MULTICAST( 3211 &args->f_id.dst_ip6)) { 3212 KASSERT(!need_send_reject, 3213 ("o_unreach6 - need_send_reject was set previously")); 3214 reject_code = cmd->arg1; 3215 if (cmd->opcode == O_REJECT) { 3216 reject_code = 3217 map_icmp_unreach(reject_code); 3218 } 3219 need_send_reject = true; 3220 } 3221 /* FALLTHROUGH */ 3222 #endif 3223 case O_DENY: 3224 retval = IP_FW_DENY; 3225 l = 0; /* exit inner loop */ 3226 done = 1; /* exit outer loop */ 3227 break; 3228 3229 case O_FORWARD_IP: 3230 if (args->flags & IPFW_ARGS_ETHER) 3231 break; /* not valid on layer2 pkts */ 3232 if (q != f || 3233 dyn_info.direction == MATCH_FORWARD) { 3234 struct sockaddr_in *sa; 3235 3236 sa = &(((ipfw_insn_sa *)cmd)->sa); 3237 if (sa->sin_addr.s_addr == INADDR_ANY) { 3238 #ifdef INET6 3239 /* 3240 * We use O_FORWARD_IP opcode for 3241 * fwd rule with tablearg, but tables 3242 * now support IPv6 addresses. And 3243 * when we are inspecting IPv6 packet, 3244 * we can use nh6 field from 3245 * table_value as next_hop6 address. 3246 */ 3247 if (is_ipv6) { 3248 struct ip_fw_nh6 *nh6; 3249 3250 args->flags |= IPFW_ARGS_NH6; 3251 nh6 = &args->hopstore6; 3252 nh6->sin6_addr = TARG_VAL( 3253 chain, tablearg, nh6); 3254 nh6->sin6_port = sa->sin_port; 3255 nh6->sin6_scope_id = TARG_VAL( 3256 chain, tablearg, zoneid); 3257 } else 3258 #endif 3259 { 3260 args->flags |= IPFW_ARGS_NH4; 3261 args->hopstore.sin_port = 3262 sa->sin_port; 3263 sa = &args->hopstore; 3264 sa->sin_family = AF_INET; 3265 sa->sin_len = sizeof(*sa); 3266 sa->sin_addr.s_addr = htonl( 3267 TARG_VAL(chain, tablearg, 3268 nh4)); 3269 } 3270 } else { 3271 args->flags |= IPFW_ARGS_NH4PTR; 3272 args->next_hop = sa; 3273 } 3274 } 3275 retval = IP_FW_PASS; 3276 l = 0; /* exit inner loop */ 3277 done = 1; /* exit outer loop */ 3278 break; 3279 3280 #ifdef INET6 3281 case O_FORWARD_IP6: 3282 if (args->flags & IPFW_ARGS_ETHER) 3283 break; /* not valid on layer2 pkts */ 3284 if (q != f || 3285 dyn_info.direction == MATCH_FORWARD) { 3286 struct sockaddr_in6 *sin6; 3287 3288 sin6 = &(((ipfw_insn_sa6 *)cmd)->sa); 3289 args->flags |= IPFW_ARGS_NH6PTR; 3290 args->next_hop6 = sin6; 3291 } 3292 retval = IP_FW_PASS; 3293 l = 0; /* exit inner loop */ 3294 done = 1; /* exit outer loop */ 3295 break; 3296 #endif 3297 3298 case O_NETGRAPH: 3299 case O_NGTEE: 3300 set_match(args, f_pos, chain); 3301 args->rule.info = TARG(cmd->arg1, netgraph); 3302 if (V_fw_one_pass) 3303 args->rule.info |= IPFW_ONEPASS; 3304 retval = (cmd->opcode == O_NETGRAPH) ? 3305 IP_FW_NETGRAPH : IP_FW_NGTEE; 3306 l = 0; /* exit inner loop */ 3307 done = 1; /* exit outer loop */ 3308 break; 3309 3310 case O_SETFIB: { 3311 uint32_t fib; 3312 3313 IPFW_INC_RULE_COUNTER(f, pktlen); 3314 fib = TARG(cmd->arg1, fib) & 0x7FFF; 3315 if (fib >= rt_numfibs) 3316 fib = 0; 3317 M_SETFIB(m, fib); 3318 args->f_id.fib = fib; /* XXX */ 3319 l = 0; /* exit inner loop */ 3320 break; 3321 } 3322 3323 case O_SETDSCP: { 3324 uint16_t code; 3325 3326 code = TARG(cmd->arg1, dscp) & 0x3F; 3327 l = 0; /* exit inner loop */ 3328 if (is_ipv4) { 3329 uint16_t old; 3330 3331 old = *(uint16_t *)ip; 3332 ip->ip_tos = (code << 2) | 3333 (ip->ip_tos & 0x03); 3334 ip->ip_sum = cksum_adjust(ip->ip_sum, 3335 old, *(uint16_t *)ip); 3336 } else if (is_ipv6) { 3337 /* update cached value */ 3338 args->f_id.flow_id6 = 3339 ntohl(*(uint32_t *)ip) & ~0x0FC00000; 3340 args->f_id.flow_id6 |= code << 22; 3341 3342 *((uint32_t *)ip) = 3343 htonl(args->f_id.flow_id6); 3344 } else 3345 break; 3346 3347 IPFW_INC_RULE_COUNTER(f, pktlen); 3348 break; 3349 } 3350 3351 case O_NAT: 3352 l = 0; /* exit inner loop */ 3353 done = 1; /* exit outer loop */ 3354 /* 3355 * Ensure that we do not invoke NAT handler for 3356 * non IPv4 packets. Libalias expects only IPv4. 3357 */ 3358 if (!is_ipv4 || !IPFW_NAT_LOADED) { 3359 retval = IP_FW_DENY; 3360 break; 3361 } 3362 3363 struct cfg_nat *t; 3364 int nat_id; 3365 3366 args->rule.info = 0; 3367 set_match(args, f_pos, chain); 3368 /* Check if this is 'global' nat rule */ 3369 if (cmd->arg1 == IP_FW_NAT44_GLOBAL) { 3370 retval = ipfw_nat_ptr(args, NULL, m); 3371 break; 3372 } 3373 t = ((ipfw_insn_nat *)cmd)->nat; 3374 if (t == NULL) { 3375 nat_id = TARG(cmd->arg1, nat); 3376 t = (*lookup_nat_ptr)(&chain->nat, nat_id); 3377 3378 if (t == NULL) { 3379 retval = IP_FW_DENY; 3380 break; 3381 } 3382 if (cmd->arg1 != IP_FW_TARG) 3383 ((ipfw_insn_nat *)cmd)->nat = t; 3384 } 3385 retval = ipfw_nat_ptr(args, t, m); 3386 break; 3387 3388 case O_REASS: { 3389 int ip_off; 3390 3391 l = 0; /* in any case exit inner loop */ 3392 if (is_ipv6) /* IPv6 is not supported yet */ 3393 break; 3394 IPFW_INC_RULE_COUNTER(f, pktlen); 3395 ip_off = ntohs(ip->ip_off); 3396 3397 /* if not fragmented, go to next rule */ 3398 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0) 3399 break; 3400 3401 args->m = m = ip_reass(m); 3402 3403 /* 3404 * do IP header checksum fixup. 3405 */ 3406 if (m == NULL) { /* fragment got swallowed */ 3407 retval = IP_FW_DENY; 3408 } else { /* good, packet complete */ 3409 int hlen; 3410 3411 ip = mtod(m, struct ip *); 3412 hlen = ip->ip_hl << 2; 3413 ip->ip_sum = 0; 3414 if (hlen == sizeof(struct ip)) 3415 ip->ip_sum = in_cksum_hdr(ip); 3416 else 3417 ip->ip_sum = in_cksum(m, hlen); 3418 retval = IP_FW_REASS; 3419 args->rule.info = 0; 3420 set_match(args, f_pos, chain); 3421 } 3422 done = 1; /* exit outer loop */ 3423 break; 3424 } 3425 3426 case O_SETMARK: { 3427 l = 0; /* exit inner loop */ 3428 args->rule.pkt_mark = ( 3429 (cmd->arg1 == IP_FW_TARG) ? 3430 TARG_VAL(chain, tablearg, mark) : 3431 insntoc(cmd, u32)->d[0]); 3432 3433 IPFW_INC_RULE_COUNTER(f, pktlen); 3434 break; 3435 } 3436 3437 case O_EXTERNAL_ACTION: 3438 l = 0; /* in any case exit inner loop */ 3439 retval = ipfw_run_eaction(chain, args, 3440 cmd, &done); 3441 /* 3442 * If both @retval and @done are zero, 3443 * consider this as rule matching and 3444 * update counters. 3445 */ 3446 if (retval == 0 && done == 0) { 3447 IPFW_INC_RULE_COUNTER(f, pktlen); 3448 /* 3449 * Reset the result of the last 3450 * dynamic state lookup. 3451 * External action can change 3452 * @args content, and it may be 3453 * used for new state lookup later. 3454 */ 3455 DYN_INFO_INIT(&dyn_info); 3456 } 3457 break; 3458 3459 default: 3460 panic("ipfw: rule %u: unknown opcode %d\n", 3461 f->rulenum, cmd->opcode); 3462 } /* end of switch() on opcodes */ 3463 /* 3464 * if we get here with l=0, then match is irrelevant. 3465 */ 3466 3467 if (cmd->len & F_NOT) 3468 match = !match; 3469 3470 if (match) { 3471 if (cmd->len & F_OR) 3472 skip_or = 1; 3473 } else { 3474 if (!(cmd->len & F_OR)) /* not an OR block, */ 3475 break; /* try next rule */ 3476 } 3477 3478 } /* end of inner loop, scan opcodes */ 3479 #undef PULLUP_LEN 3480 #undef PULLUP_LEN_LOCKED 3481 3482 if (done) 3483 break; 3484 3485 /* next_rule:; */ /* try next rule */ 3486 3487 } /* end of outer for, scan rules */ 3488 3489 if (done) { 3490 struct ip_fw *rule = chain->map[f_pos]; 3491 /* Update statistics */ 3492 IPFW_INC_RULE_COUNTER(rule, pktlen); 3493 IPFW_PROBE(rule__matched, retval, 3494 is_ipv4 ? AF_INET : AF_INET6, 3495 is_ipv4 ? (uintptr_t)&src_ip : 3496 (uintptr_t)&args->f_id.src_ip6, 3497 is_ipv4 ? (uintptr_t)&dst_ip : 3498 (uintptr_t)&args->f_id.dst_ip6, 3499 args, rule); 3500 } else { 3501 retval = IP_FW_DENY; 3502 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3503 } 3504 IPFW_PF_RUNLOCK(chain); 3505 if (need_send_reject) { 3506 #ifdef INET6 3507 if (is_ipv6) 3508 send_reject6(args, reject_code, hlen, 3509 (struct ip6_hdr *)ip); 3510 else 3511 #endif 3512 send_reject(args, reject_code, reject_mtu, 3513 iplen, ip); 3514 } 3515 #ifdef __FreeBSD__ 3516 if (ucred_cache != NULL) 3517 crfree(ucred_cache); 3518 #endif 3519 return (retval); 3520 3521 pullup_failed: 3522 if (V_fw_verbose) 3523 printf("ipfw: pullup failed\n"); 3524 return (IP_FW_DENY); 3525 } 3526 3527 /* 3528 * Set maximum number of tables that can be used in given VNET ipfw instance. 3529 */ 3530 #ifdef SYSCTL_NODE 3531 static int 3532 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS) 3533 { 3534 int error; 3535 unsigned int ntables; 3536 3537 ntables = V_fw_tables_max; 3538 3539 error = sysctl_handle_int(oidp, &ntables, 0, req); 3540 /* Read operation or some error */ 3541 if ((error != 0) || (req->newptr == NULL)) 3542 return (error); 3543 3544 return (ipfw_resize_tables(&V_layer3_chain, ntables)); 3545 } 3546 3547 /* 3548 * Switches table namespace between global and per-set. 3549 */ 3550 static int 3551 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS) 3552 { 3553 int error; 3554 unsigned int sets; 3555 3556 sets = V_fw_tables_sets; 3557 3558 error = sysctl_handle_int(oidp, &sets, 0, req); 3559 /* Read operation or some error */ 3560 if ((error != 0) || (req->newptr == NULL)) 3561 return (error); 3562 3563 return (ipfw_switch_tables_namespace(&V_layer3_chain, sets)); 3564 } 3565 #endif 3566 3567 /* 3568 * Module and VNET glue 3569 */ 3570 3571 /* 3572 * Stuff that must be initialised only on boot or module load 3573 */ 3574 static int 3575 ipfw_init(void) 3576 { 3577 int error = 0; 3578 3579 /* 3580 * Only print out this stuff the first time around, 3581 * when called from the sysinit code. 3582 */ 3583 printf("ipfw2 " 3584 #ifdef INET6 3585 "(+ipv6) " 3586 #endif 3587 "initialized, divert %s, nat %s, " 3588 "default to %s, logging ", 3589 #ifdef IPDIVERT 3590 "enabled", 3591 #else 3592 "loadable", 3593 #endif 3594 #ifdef IPFIREWALL_NAT 3595 "enabled", 3596 #else 3597 "loadable", 3598 #endif 3599 default_to_accept ? "accept" : "deny"); 3600 3601 /* 3602 * Note: V_xxx variables can be accessed here but the vnet specific 3603 * initializer may not have been called yet for the VIMAGE case. 3604 * Tuneables will have been processed. We will print out values for 3605 * the default vnet. 3606 * XXX This should all be rationalized AFTER 8.0 3607 */ 3608 if (V_fw_verbose == 0) 3609 printf("disabled\n"); 3610 else if (V_verbose_limit == 0) 3611 printf("unlimited\n"); 3612 else 3613 printf("limited to %d packets/entry by default\n", 3614 V_verbose_limit); 3615 3616 /* Check user-supplied table count for validness */ 3617 if (default_fw_tables > IPFW_TABLES_MAX) 3618 default_fw_tables = IPFW_TABLES_MAX; 3619 3620 ipfw_init_sopt_handler(); 3621 ipfw_init_obj_rewriter(); 3622 ipfw_iface_init(); 3623 return (error); 3624 } 3625 3626 /* 3627 * Called for the removal of the last instance only on module unload. 3628 */ 3629 static void 3630 ipfw_destroy(void) 3631 { 3632 3633 ipfw_iface_destroy(); 3634 ipfw_destroy_sopt_handler(); 3635 ipfw_destroy_obj_rewriter(); 3636 printf("IP firewall unloaded\n"); 3637 } 3638 3639 /* 3640 * Stuff that must be initialized for every instance 3641 * (including the first of course). 3642 */ 3643 static int 3644 vnet_ipfw_init(const void *unused) 3645 { 3646 int error, first; 3647 struct ip_fw *rule = NULL; 3648 struct ip_fw_chain *chain; 3649 3650 chain = &V_layer3_chain; 3651 3652 first = IS_DEFAULT_VNET(curvnet) ? 1 : 0; 3653 3654 /* First set up some values that are compile time options */ 3655 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 3656 V_fw_deny_unknown_exthdrs = 1; 3657 #ifdef IPFIREWALL_VERBOSE 3658 V_fw_verbose = 1; 3659 #endif 3660 #ifdef IPFIREWALL_VERBOSE_LIMIT 3661 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 3662 #endif 3663 #ifdef IPFIREWALL_NAT 3664 LIST_INIT(&chain->nat); 3665 #endif 3666 3667 /* Init shared services hash table */ 3668 ipfw_init_srv(chain); 3669 3670 ipfw_init_counters(); 3671 /* Set initial number of tables */ 3672 V_fw_tables_max = default_fw_tables; 3673 error = ipfw_init_tables(chain, first); 3674 if (error) { 3675 printf("ipfw2: setting up tables failed\n"); 3676 free(chain->map, M_IPFW); 3677 free(rule, M_IPFW); 3678 return (ENOSPC); 3679 } 3680 3681 IPFW_LOCK_INIT(chain); 3682 3683 ipfw_dyn_init(chain); 3684 /* fill and insert the default rule */ 3685 rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw)); 3686 rule->flags |= IPFW_RULE_NOOPT; 3687 rule->cmd_len = 1; 3688 rule->cmd[0].len = 1; 3689 rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; 3690 chain->default_rule = rule; 3691 ipfw_add_protected_rule(chain, rule, 0); 3692 3693 ipfw_eaction_init(chain, first); 3694 ipfw_init_skipto_cache(chain); 3695 ipfw_bpf_init(first); 3696 3697 /* First set up some values that are compile time options */ 3698 V_ipfw_vnet_ready = 1; /* Open for business */ 3699 3700 /* 3701 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6. 3702 * Even if the latter two fail we still keep the module alive 3703 * because the sockopt and layer2 paths are still useful. 3704 * ipfw[6]_hook return 0 on success, ENOENT on failure, 3705 * so we can ignore the exact return value and just set a flag. 3706 * 3707 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so 3708 * changes in the underlying (per-vnet) variables trigger 3709 * immediate hook()/unhook() calls. 3710 * In layer2 we have the same behaviour, except that V_ether_ipfw 3711 * is checked on each packet because there are no pfil hooks. 3712 */ 3713 V_ip_fw_ctl_ptr = ipfw_ctl3; 3714 error = ipfw_attach_hooks(); 3715 return (error); 3716 } 3717 3718 /* 3719 * Called for the removal of each instance. 3720 */ 3721 static int 3722 vnet_ipfw_uninit(const void *unused) 3723 { 3724 struct ip_fw *reap; 3725 struct ip_fw_chain *chain = &V_layer3_chain; 3726 int i, last; 3727 3728 V_ipfw_vnet_ready = 0; /* tell new callers to go away */ 3729 /* 3730 * disconnect from ipv4, ipv6, layer2 and sockopt. 3731 * Then grab, release and grab again the WLOCK so we make 3732 * sure the update is propagated and nobody will be in. 3733 */ 3734 ipfw_detach_hooks(); 3735 V_ip_fw_ctl_ptr = NULL; 3736 3737 last = IS_DEFAULT_VNET(curvnet) ? 1 : 0; 3738 3739 IPFW_UH_WLOCK(chain); 3740 IPFW_UH_WUNLOCK(chain); 3741 3742 ipfw_dyn_uninit(0); /* run the callout_drain */ 3743 3744 IPFW_UH_WLOCK(chain); 3745 3746 reap = NULL; 3747 IPFW_WLOCK(chain); 3748 for (i = 0; i < chain->n_rules; i++) 3749 ipfw_reap_add(chain, &reap, chain->map[i]); 3750 free(chain->map, M_IPFW); 3751 ipfw_destroy_skipto_cache(chain); 3752 IPFW_WUNLOCK(chain); 3753 IPFW_UH_WUNLOCK(chain); 3754 ipfw_destroy_tables(chain, last); 3755 ipfw_eaction_uninit(chain, last); 3756 if (reap != NULL) 3757 ipfw_reap_rules(reap); 3758 vnet_ipfw_iface_destroy(chain); 3759 ipfw_destroy_srv(chain); 3760 IPFW_LOCK_DESTROY(chain); 3761 ipfw_dyn_uninit(1); /* free the remaining parts */ 3762 ipfw_destroy_counters(); 3763 ipfw_bpf_uninit(last); 3764 return (0); 3765 } 3766 3767 /* 3768 * Module event handler. 3769 * In general we have the choice of handling most of these events by the 3770 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to 3771 * use the SYSINIT handlers as they are more capable of expressing the 3772 * flow of control during module and vnet operations, so this is just 3773 * a skeleton. Note there is no SYSINIT equivalent of the module 3774 * SHUTDOWN handler, but we don't have anything to do in that case anyhow. 3775 */ 3776 static int 3777 ipfw_modevent(module_t mod, int type, void *unused) 3778 { 3779 int err = 0; 3780 3781 switch (type) { 3782 case MOD_LOAD: 3783 /* Called once at module load or 3784 * system boot if compiled in. */ 3785 break; 3786 case MOD_QUIESCE: 3787 /* Called before unload. May veto unloading. */ 3788 break; 3789 case MOD_UNLOAD: 3790 /* Called during unload. */ 3791 break; 3792 case MOD_SHUTDOWN: 3793 /* Called during system shutdown. */ 3794 break; 3795 default: 3796 err = EOPNOTSUPP; 3797 break; 3798 } 3799 return err; 3800 } 3801 3802 static moduledata_t ipfwmod = { 3803 "ipfw", 3804 ipfw_modevent, 3805 0 3806 }; 3807 3808 /* Define startup order. */ 3809 #define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_FIREWALL 3810 #define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */ 3811 #define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */ 3812 #define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */ 3813 3814 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER); 3815 FEATURE(ipfw_ctl3, "ipfw new sockopt calls"); 3816 MODULE_VERSION(ipfw, 3); 3817 /* should declare some dependencies here */ 3818 3819 /* 3820 * Starting up. Done in order after ipfwmod() has been called. 3821 * VNET_SYSINIT is also called for each existing vnet and each new vnet. 3822 */ 3823 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 3824 ipfw_init, NULL); 3825 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 3826 vnet_ipfw_init, NULL); 3827 3828 /* 3829 * Closing up shop. These are done in REVERSE ORDER, but still 3830 * after ipfwmod() has been called. Not called on reboot. 3831 * VNET_SYSUNINIT is also called for each exiting vnet as it exits. 3832 * or when the module is unloaded. 3833 */ 3834 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 3835 ipfw_destroy, NULL); 3836 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 3837 vnet_ipfw_uninit, NULL); 3838 /* end of file */ 3839