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 static int jump_lookup_pos(struct ip_fw_chain *chain, struct ip_fw *f, int num, 148 int tablearg, int jump_backwards); 149 #ifndef LINEAR_SKIPTO 150 static int jump_cached(struct ip_fw_chain *chain, struct ip_fw *f, int num, 151 int tablearg, int jump_backwards); 152 #define JUMP(ch, f, num, targ, back) jump_cached(ch, f, num, targ, back) 153 #else 154 #define JUMP(ch, f, num, targ, back) jump_lookup_pos(ch, f, num, targ, back) 155 #endif 156 157 /* 158 * Each rule belongs to one of 32 different sets (0..31). 159 * The variable set_disable contains one bit per set. 160 * If the bit is set, all rules in the corresponding set 161 * are disabled. Set RESVD_SET(31) is reserved for the default rule 162 * and rules that are not deleted by the flush command, 163 * and CANNOT be disabled. 164 * Rules in set RESVD_SET can only be deleted individually. 165 */ 166 VNET_DEFINE(u_int32_t, set_disable); 167 #define V_set_disable VNET(set_disable) 168 169 VNET_DEFINE(int, fw_verbose); 170 /* counter for ipfw_log(NULL...) */ 171 VNET_DEFINE(u_int64_t, norule_counter); 172 VNET_DEFINE(int, verbose_limit); 173 174 /* layer3_chain contains the list of rules for layer 3 */ 175 VNET_DEFINE(struct ip_fw_chain, layer3_chain); 176 177 /* ipfw_vnet_ready controls when we are open for business */ 178 VNET_DEFINE(int, ipfw_vnet_ready) = 0; 179 180 VNET_DEFINE(int, ipfw_nat_ready) = 0; 181 182 ipfw_nat_t *ipfw_nat_ptr = NULL; 183 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int); 184 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; 185 ipfw_nat_cfg_t *ipfw_nat_del_ptr; 186 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; 187 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; 188 189 #ifdef SYSCTL_NODE 190 uint32_t dummy_def = IPFW_DEFAULT_RULE; 191 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS); 192 static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS); 193 194 SYSBEGIN(f3) 195 196 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 197 "Firewall"); 198 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 199 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0, 200 "Only do a single pass through ipfw when using dummynet(4)"); 201 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, 202 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0, 203 "Rule number auto-increment step"); 204 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 205 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0, 206 "Log matches to ipfw rules"); 207 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, 208 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(verbose_limit), 0, 209 "Set upper limit of matches of ipfw rules logged"); 210 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, 211 &dummy_def, 0, 212 "The default/max possible rule number."); 213 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_max, 214 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, 215 0, 0, sysctl_ipfw_table_num, "IU", 216 "Maximum number of concurrently used tables"); 217 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets, 218 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_MPSAFE, 219 0, 0, sysctl_ipfw_tables_sets, "IU", 220 "Use per-set namespace for tables"); 221 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN, 222 &default_to_accept, 0, 223 "Make the default rule accept all packets."); 224 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables); 225 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, 226 CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0, 227 "Number of static rules"); 228 229 #ifdef INET6 230 SYSCTL_DECL(_net_inet6_ip6); 231 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 232 "Firewall"); 233 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs, 234 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, 235 &VNET_NAME(fw_deny_unknown_exthdrs), 0, 236 "Deny packets with unknown IPv6 Extension Headers"); 237 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6, 238 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, 239 &VNET_NAME(fw_permit_single_frag6), 0, 240 "Permit single packet IPv6 fragments"); 241 #endif /* INET6 */ 242 243 SYSEND 244 245 #endif /* SYSCTL_NODE */ 246 247 /* 248 * Some macros used in the various matching options. 249 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 250 * Other macros just cast void * into the appropriate type 251 */ 252 #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 253 #define TCP(p) ((struct tcphdr *)(p)) 254 #define SCTP(p) ((struct sctphdr *)(p)) 255 #define UDP(p) ((struct udphdr *)(p)) 256 #define ICMP(p) ((struct icmphdr *)(p)) 257 #define ICMP6(p) ((struct icmp6_hdr *)(p)) 258 259 static __inline int 260 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 261 { 262 int type = icmp->icmp_type; 263 264 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 265 } 266 267 #define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 268 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 269 270 static int 271 is_icmp_query(struct icmphdr *icmp) 272 { 273 int type = icmp->icmp_type; 274 275 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 276 } 277 #undef TT 278 279 /* 280 * The following checks use two arrays of 8 or 16 bits to store the 281 * bits that we want set or clear, respectively. They are in the 282 * low and high half of cmd->arg1 or cmd->d[0]. 283 * 284 * We scan options and store the bits we find set. We succeed if 285 * 286 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 287 * 288 * The code is sometimes optimized not to store additional variables. 289 */ 290 291 static int 292 flags_match(ipfw_insn *cmd, u_int8_t bits) 293 { 294 u_char want_clear; 295 bits = ~bits; 296 297 if ( ((cmd->arg1 & 0xff) & bits) != 0) 298 return 0; /* some bits we want set were clear */ 299 want_clear = (cmd->arg1 >> 8) & 0xff; 300 if ( (want_clear & bits) != want_clear) 301 return 0; /* some bits we want clear were set */ 302 return 1; 303 } 304 305 static int 306 ipopts_match(struct ip *ip, ipfw_insn *cmd) 307 { 308 int optlen, bits = 0; 309 u_char *cp = (u_char *)(ip + 1); 310 int x = (ip->ip_hl << 2) - sizeof (struct ip); 311 312 for (; x > 0; x -= optlen, cp += optlen) { 313 int opt = cp[IPOPT_OPTVAL]; 314 315 if (opt == IPOPT_EOL) 316 break; 317 if (opt == IPOPT_NOP) 318 optlen = 1; 319 else { 320 optlen = cp[IPOPT_OLEN]; 321 if (optlen <= 0 || optlen > x) 322 return 0; /* invalid or truncated */ 323 } 324 switch (opt) { 325 default: 326 break; 327 328 case IPOPT_LSRR: 329 bits |= IP_FW_IPOPT_LSRR; 330 break; 331 332 case IPOPT_SSRR: 333 bits |= IP_FW_IPOPT_SSRR; 334 break; 335 336 case IPOPT_RR: 337 bits |= IP_FW_IPOPT_RR; 338 break; 339 340 case IPOPT_TS: 341 bits |= IP_FW_IPOPT_TS; 342 break; 343 } 344 } 345 return (flags_match(cmd, bits)); 346 } 347 348 /* 349 * Parse TCP options. The logic copied from tcp_dooptions(). 350 */ 351 static int 352 tcpopts_parse(const struct tcphdr *tcp, uint16_t *mss) 353 { 354 const u_char *cp = (const u_char *)(tcp + 1); 355 int optlen, bits = 0; 356 int cnt = (tcp->th_off << 2) - sizeof(struct tcphdr); 357 358 for (; cnt > 0; cnt -= optlen, cp += optlen) { 359 int opt = cp[0]; 360 if (opt == TCPOPT_EOL) 361 break; 362 if (opt == TCPOPT_NOP) 363 optlen = 1; 364 else { 365 if (cnt < 2) 366 break; 367 optlen = cp[1]; 368 if (optlen < 2 || optlen > cnt) 369 break; 370 } 371 372 switch (opt) { 373 default: 374 break; 375 376 case TCPOPT_MAXSEG: 377 if (optlen != TCPOLEN_MAXSEG) 378 break; 379 bits |= IP_FW_TCPOPT_MSS; 380 if (mss != NULL) 381 *mss = be16dec(cp + 2); 382 break; 383 384 case TCPOPT_WINDOW: 385 if (optlen == TCPOLEN_WINDOW) 386 bits |= IP_FW_TCPOPT_WINDOW; 387 break; 388 389 case TCPOPT_SACK_PERMITTED: 390 if (optlen == TCPOLEN_SACK_PERMITTED) 391 bits |= IP_FW_TCPOPT_SACK; 392 break; 393 394 case TCPOPT_SACK: 395 if (optlen > 2 && (optlen - 2) % TCPOLEN_SACK == 0) 396 bits |= IP_FW_TCPOPT_SACK; 397 break; 398 399 case TCPOPT_TIMESTAMP: 400 if (optlen == TCPOLEN_TIMESTAMP) 401 bits |= IP_FW_TCPOPT_TS; 402 break; 403 } 404 } 405 return (bits); 406 } 407 408 static int 409 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 410 { 411 412 return (flags_match(cmd, tcpopts_parse(tcp, NULL))); 413 } 414 415 static int 416 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain, 417 uint32_t *tablearg) 418 { 419 420 if (ifp == NULL) /* no iface with this packet, match fails */ 421 return (0); 422 423 /* Check by name or by IP address */ 424 if (cmd->name[0] != '\0') { /* match by name */ 425 if (cmd->name[0] == '\1') /* use tablearg to match */ 426 return ipfw_lookup_table(chain, cmd->p.kidx, 0, 427 &ifp->if_index, tablearg); 428 /* Check name */ 429 if (cmd->p.glob) { 430 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 431 return(1); 432 } else { 433 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 434 return(1); 435 } 436 } else { 437 #if !defined(USERSPACE) && defined(__FreeBSD__) /* and OSX too ? */ 438 struct ifaddr *ia; 439 440 NET_EPOCH_ASSERT(); 441 442 CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 443 if (ia->ifa_addr->sa_family != AF_INET) 444 continue; 445 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 446 (ia->ifa_addr))->sin_addr.s_addr) 447 return (1); /* match */ 448 } 449 #endif /* __FreeBSD__ */ 450 } 451 return(0); /* no match, fail ... */ 452 } 453 454 /* 455 * The verify_path function checks if a route to the src exists and 456 * if it is reachable via ifp (when provided). 457 * 458 * The 'verrevpath' option checks that the interface that an IP packet 459 * arrives on is the same interface that traffic destined for the 460 * packet's source address would be routed out of. 461 * The 'versrcreach' option just checks that the source address is 462 * reachable via any route (except default) in the routing table. 463 * These two are a measure to block forged packets. This is also 464 * commonly known as "anti-spoofing" or Unicast Reverse Path 465 * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 466 * is purposely reminiscent of the Cisco IOS command, 467 * 468 * ip verify unicast reverse-path 469 * ip verify unicast source reachable-via any 470 * 471 * which implements the same functionality. But note that the syntax 472 * is misleading, and the check may be performed on all IP packets 473 * whether unicast, multicast, or broadcast. 474 */ 475 static int 476 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) 477 { 478 #if defined(USERSPACE) || !defined(__FreeBSD__) 479 return 0; 480 #else 481 struct nhop_object *nh; 482 483 nh = fib4_lookup(fib, src, 0, NHR_NONE, 0); 484 if (nh == NULL) 485 return (0); 486 487 /* 488 * If ifp is provided, check for equality with rtentry. 489 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 490 * in order to pass packets injected back by if_simloop(): 491 * routing entry (via lo0) for our own address 492 * may exist, so we need to handle routing assymetry. 493 */ 494 if (ifp != NULL && ifp != nh->nh_aifp) 495 return (0); 496 497 /* if no ifp provided, check if rtentry is not default route */ 498 if (ifp == NULL && (nh->nh_flags & NHF_DEFAULT) != 0) 499 return (0); 500 501 /* or if this is a blackhole/reject route */ 502 if (ifp == NULL && (nh->nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0) 503 return (0); 504 505 /* found valid route */ 506 return 1; 507 #endif /* __FreeBSD__ */ 508 } 509 510 /* 511 * Generate an SCTP packet containing an ABORT chunk. The verification tag 512 * is given by vtag. The T-bit is set in the ABORT chunk if and only if 513 * reflected is not 0. 514 */ 515 516 static struct mbuf * 517 ipfw_send_abort(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t vtag, 518 int reflected) 519 { 520 struct mbuf *m; 521 struct ip *ip; 522 #ifdef INET6 523 struct ip6_hdr *ip6; 524 #endif 525 struct sctphdr *sctp; 526 struct sctp_chunkhdr *chunk; 527 u_int16_t hlen, plen, tlen; 528 529 MGETHDR(m, M_NOWAIT, MT_DATA); 530 if (m == NULL) 531 return (NULL); 532 533 M_SETFIB(m, id->fib); 534 #ifdef MAC 535 if (replyto != NULL) 536 mac_netinet_firewall_reply(replyto, m); 537 else 538 mac_netinet_firewall_send(m); 539 #else 540 (void)replyto; /* don't warn about unused arg */ 541 #endif 542 543 switch (id->addr_type) { 544 case 4: 545 hlen = sizeof(struct ip); 546 break; 547 #ifdef INET6 548 case 6: 549 hlen = sizeof(struct ip6_hdr); 550 break; 551 #endif 552 default: 553 /* XXX: log me?!? */ 554 FREE_PKT(m); 555 return (NULL); 556 } 557 plen = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); 558 tlen = hlen + plen; 559 m->m_data += max_linkhdr; 560 m->m_flags |= M_SKIP_FIREWALL; 561 m->m_pkthdr.len = m->m_len = tlen; 562 m->m_pkthdr.rcvif = NULL; 563 bzero(m->m_data, tlen); 564 565 switch (id->addr_type) { 566 case 4: 567 ip = mtod(m, struct ip *); 568 569 ip->ip_v = 4; 570 ip->ip_hl = sizeof(struct ip) >> 2; 571 ip->ip_tos = IPTOS_LOWDELAY; 572 ip->ip_len = htons(tlen); 573 ip->ip_id = htons(0); 574 ip->ip_off = htons(0); 575 ip->ip_ttl = V_ip_defttl; 576 ip->ip_p = IPPROTO_SCTP; 577 ip->ip_sum = 0; 578 ip->ip_src.s_addr = htonl(id->dst_ip); 579 ip->ip_dst.s_addr = htonl(id->src_ip); 580 581 sctp = (struct sctphdr *)(ip + 1); 582 break; 583 #ifdef INET6 584 case 6: 585 ip6 = mtod(m, struct ip6_hdr *); 586 587 ip6->ip6_vfc = IPV6_VERSION; 588 ip6->ip6_plen = htons(plen); 589 ip6->ip6_nxt = IPPROTO_SCTP; 590 ip6->ip6_hlim = IPV6_DEFHLIM; 591 ip6->ip6_src = id->dst_ip6; 592 ip6->ip6_dst = id->src_ip6; 593 594 sctp = (struct sctphdr *)(ip6 + 1); 595 break; 596 #endif 597 } 598 599 sctp->src_port = htons(id->dst_port); 600 sctp->dest_port = htons(id->src_port); 601 sctp->v_tag = htonl(vtag); 602 sctp->checksum = htonl(0); 603 604 chunk = (struct sctp_chunkhdr *)(sctp + 1); 605 chunk->chunk_type = SCTP_ABORT_ASSOCIATION; 606 chunk->chunk_flags = 0; 607 if (reflected != 0) { 608 chunk->chunk_flags |= SCTP_HAD_NO_TCB; 609 } 610 chunk->chunk_length = htons(sizeof(struct sctp_chunkhdr)); 611 612 sctp->checksum = sctp_calculate_cksum(m, hlen); 613 614 return (m); 615 } 616 617 /* 618 * Generate a TCP packet, containing either a RST or a keepalive. 619 * When flags & TH_RST, we are sending a RST packet, because of a 620 * "reset" action matched the packet. 621 * Otherwise we are sending a keepalive, and flags & TH_ 622 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 623 * so that MAC can label the reply appropriately. 624 */ 625 struct mbuf * 626 ipfw_send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 627 u_int32_t ack, int flags) 628 { 629 struct mbuf *m = NULL; /* stupid compiler */ 630 struct ip *h = NULL; /* stupid compiler */ 631 #ifdef INET6 632 struct ip6_hdr *h6 = NULL; 633 #endif 634 struct tcphdr *th = NULL; 635 int len, dir; 636 637 MGETHDR(m, M_NOWAIT, MT_DATA); 638 if (m == NULL) 639 return (NULL); 640 641 M_SETFIB(m, id->fib); 642 #ifdef MAC 643 if (replyto != NULL) 644 mac_netinet_firewall_reply(replyto, m); 645 else 646 mac_netinet_firewall_send(m); 647 #else 648 (void)replyto; /* don't warn about unused arg */ 649 #endif 650 651 switch (id->addr_type) { 652 case 4: 653 len = sizeof(struct ip) + sizeof(struct tcphdr); 654 break; 655 #ifdef INET6 656 case 6: 657 len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); 658 break; 659 #endif 660 default: 661 /* XXX: log me?!? */ 662 FREE_PKT(m); 663 return (NULL); 664 } 665 dir = ((flags & (TH_SYN | TH_RST)) == TH_SYN); 666 667 m->m_data += max_linkhdr; 668 m->m_flags |= M_SKIP_FIREWALL; 669 m->m_pkthdr.len = m->m_len = len; 670 m->m_pkthdr.rcvif = NULL; 671 bzero(m->m_data, len); 672 673 switch (id->addr_type) { 674 case 4: 675 h = mtod(m, struct ip *); 676 677 /* prepare for checksum */ 678 h->ip_p = IPPROTO_TCP; 679 h->ip_len = htons(sizeof(struct tcphdr)); 680 if (dir) { 681 h->ip_src.s_addr = htonl(id->src_ip); 682 h->ip_dst.s_addr = htonl(id->dst_ip); 683 } else { 684 h->ip_src.s_addr = htonl(id->dst_ip); 685 h->ip_dst.s_addr = htonl(id->src_ip); 686 } 687 688 th = (struct tcphdr *)(h + 1); 689 break; 690 #ifdef INET6 691 case 6: 692 h6 = mtod(m, struct ip6_hdr *); 693 694 /* prepare for checksum */ 695 h6->ip6_nxt = IPPROTO_TCP; 696 h6->ip6_plen = htons(sizeof(struct tcphdr)); 697 if (dir) { 698 h6->ip6_src = id->src_ip6; 699 h6->ip6_dst = id->dst_ip6; 700 } else { 701 h6->ip6_src = id->dst_ip6; 702 h6->ip6_dst = id->src_ip6; 703 } 704 705 th = (struct tcphdr *)(h6 + 1); 706 break; 707 #endif 708 } 709 710 if (dir) { 711 th->th_sport = htons(id->src_port); 712 th->th_dport = htons(id->dst_port); 713 } else { 714 th->th_sport = htons(id->dst_port); 715 th->th_dport = htons(id->src_port); 716 } 717 th->th_off = sizeof(struct tcphdr) >> 2; 718 719 if (flags & TH_RST) { 720 if (flags & TH_ACK) { 721 th->th_seq = htonl(ack); 722 tcp_set_flags(th, TH_RST); 723 } else { 724 if (flags & TH_SYN) 725 seq++; 726 th->th_ack = htonl(seq); 727 tcp_set_flags(th, TH_RST | TH_ACK); 728 } 729 } else { 730 /* 731 * Keepalive - use caller provided sequence numbers 732 */ 733 th->th_seq = htonl(seq); 734 th->th_ack = htonl(ack); 735 tcp_set_flags(th, TH_ACK); 736 } 737 738 switch (id->addr_type) { 739 case 4: 740 th->th_sum = in_cksum(m, len); 741 742 /* finish the ip header */ 743 h->ip_v = 4; 744 h->ip_hl = sizeof(*h) >> 2; 745 h->ip_tos = IPTOS_LOWDELAY; 746 h->ip_off = htons(0); 747 h->ip_len = htons(len); 748 h->ip_ttl = V_ip_defttl; 749 h->ip_sum = 0; 750 break; 751 #ifdef INET6 752 case 6: 753 th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(*h6), 754 sizeof(struct tcphdr)); 755 756 /* finish the ip6 header */ 757 h6->ip6_vfc |= IPV6_VERSION; 758 h6->ip6_hlim = IPV6_DEFHLIM; 759 break; 760 #endif 761 } 762 763 return (m); 764 } 765 766 #ifdef INET6 767 /* 768 * ipv6 specific rules here... 769 */ 770 static __inline int 771 icmp6type_match(int type, ipfw_insn_u32 *cmd) 772 { 773 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 774 } 775 776 static int 777 flow6id_match(int curr_flow, ipfw_insn_u32 *cmd) 778 { 779 int i; 780 for (i=0; i <= cmd->o.arg1; ++i) 781 if (curr_flow == cmd->d[i]) 782 return 1; 783 return 0; 784 } 785 786 /* support for IP6_*_ME opcodes */ 787 static const struct in6_addr lla_mask = {{{ 788 0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff, 789 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff 790 }}}; 791 792 static int 793 ipfw_localip6(struct in6_addr *in6) 794 { 795 struct rm_priotracker in6_ifa_tracker; 796 struct in6_ifaddr *ia; 797 798 if (IN6_IS_ADDR_MULTICAST(in6)) 799 return (0); 800 801 if (!IN6_IS_ADDR_LINKLOCAL(in6)) 802 return (in6_localip(in6)); 803 804 IN6_IFADDR_RLOCK(&in6_ifa_tracker); 805 CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) { 806 if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) 807 continue; 808 if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr, 809 in6, &lla_mask)) { 810 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); 811 return (1); 812 } 813 } 814 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); 815 return (0); 816 } 817 818 static int 819 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib) 820 { 821 struct nhop_object *nh; 822 823 if (IN6_IS_SCOPE_LINKLOCAL(src)) 824 return (1); 825 826 nh = fib6_lookup(fib, src, 0, NHR_NONE, 0); 827 if (nh == NULL) 828 return (0); 829 830 /* If ifp is provided, check for equality with route table. */ 831 if (ifp != NULL && ifp != nh->nh_aifp) 832 return (0); 833 834 /* if no ifp provided, check if rtentry is not default route */ 835 if (ifp == NULL && (nh->nh_flags & NHF_DEFAULT) != 0) 836 return (0); 837 838 /* or if this is a blackhole/reject route */ 839 if (ifp == NULL && (nh->nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0) 840 return (0); 841 842 /* found valid route */ 843 return 1; 844 } 845 846 static int 847 is_icmp6_query(int icmp6_type) 848 { 849 if ((icmp6_type <= ICMP6_MAXTYPE) && 850 (icmp6_type == ICMP6_ECHO_REQUEST || 851 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 852 icmp6_type == ICMP6_WRUREQUEST || 853 icmp6_type == ICMP6_FQDN_QUERY || 854 icmp6_type == ICMP6_NI_QUERY)) 855 return (1); 856 857 return (0); 858 } 859 860 static int 861 map_icmp_unreach(int code) 862 { 863 864 /* RFC 7915 p4.2 */ 865 switch (code) { 866 case ICMP_UNREACH_NET: 867 case ICMP_UNREACH_HOST: 868 case ICMP_UNREACH_SRCFAIL: 869 case ICMP_UNREACH_NET_UNKNOWN: 870 case ICMP_UNREACH_HOST_UNKNOWN: 871 case ICMP_UNREACH_TOSNET: 872 case ICMP_UNREACH_TOSHOST: 873 return (ICMP6_DST_UNREACH_NOROUTE); 874 case ICMP_UNREACH_PORT: 875 return (ICMP6_DST_UNREACH_NOPORT); 876 default: 877 /* 878 * Map the rest of codes into admit prohibited. 879 * XXX: unreach proto should be mapped into ICMPv6 880 * parameter problem, but we use only unreach type. 881 */ 882 return (ICMP6_DST_UNREACH_ADMIN); 883 } 884 } 885 886 static void 887 send_reject6(struct ip_fw_args *args, int code, u_int hlen, 888 const struct ip6_hdr *ip6) 889 { 890 struct mbuf *m; 891 892 m = args->m; 893 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 894 const struct tcphdr * tcp; 895 tcp = (const struct tcphdr *)((const char *)ip6 + hlen); 896 897 if ((tcp_get_flags(tcp) & TH_RST) == 0) { 898 struct mbuf *m0; 899 m0 = ipfw_send_pkt(args->m, &(args->f_id), 900 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 901 tcp_get_flags(tcp) | TH_RST); 902 if (m0 != NULL) 903 ip6_output(m0, NULL, NULL, 0, NULL, NULL, 904 NULL); 905 } 906 FREE_PKT(m); 907 } else if (code == ICMP6_UNREACH_ABORT && 908 args->f_id.proto == IPPROTO_SCTP) { 909 struct mbuf *m0; 910 const struct sctphdr *sctp; 911 u_int32_t v_tag; 912 int reflected; 913 914 sctp = (const struct sctphdr *)((const char *)ip6 + hlen); 915 reflected = 1; 916 v_tag = ntohl(sctp->v_tag); 917 /* Investigate the first chunk header if available */ 918 if (m->m_len >= hlen + sizeof(struct sctphdr) + 919 sizeof(struct sctp_chunkhdr)) { 920 const struct sctp_chunkhdr *chunk; 921 922 chunk = (const struct sctp_chunkhdr *)(sctp + 1); 923 switch (chunk->chunk_type) { 924 case SCTP_INITIATION: 925 /* 926 * Packets containing an INIT chunk MUST have 927 * a zero v-tag. 928 */ 929 if (v_tag != 0) { 930 v_tag = 0; 931 break; 932 } 933 /* INIT chunk MUST NOT be bundled */ 934 if (m->m_pkthdr.len > 935 hlen + sizeof(struct sctphdr) + 936 ntohs(chunk->chunk_length) + 3) { 937 break; 938 } 939 /* Use the initiate tag if available */ 940 if ((m->m_len >= hlen + sizeof(struct sctphdr) + 941 sizeof(struct sctp_chunkhdr) + 942 offsetof(struct sctp_init, a_rwnd))) { 943 const struct sctp_init *init; 944 945 init = (const struct sctp_init *)(chunk + 1); 946 v_tag = ntohl(init->initiate_tag); 947 reflected = 0; 948 } 949 break; 950 case SCTP_ABORT_ASSOCIATION: 951 /* 952 * If the packet contains an ABORT chunk, don't 953 * reply. 954 * XXX: We should search through all chunks, 955 * but do not do that to avoid attacks. 956 */ 957 v_tag = 0; 958 break; 959 } 960 } 961 if (v_tag == 0) { 962 m0 = NULL; 963 } else { 964 m0 = ipfw_send_abort(args->m, &(args->f_id), v_tag, 965 reflected); 966 } 967 if (m0 != NULL) 968 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL); 969 FREE_PKT(m); 970 } else if (code != ICMP6_UNREACH_RST && code != ICMP6_UNREACH_ABORT) { 971 /* Send an ICMPv6 unreach. */ 972 #if 0 973 /* 974 * Unlike above, the mbufs need to line up with the ip6 hdr, 975 * as the contents are read. We need to m_adj() the 976 * needed amount. 977 * The mbuf will however be thrown away so we can adjust it. 978 * Remember we did an m_pullup on it already so we 979 * can make some assumptions about contiguousness. 980 */ 981 if (args->L3offset) 982 m_adj(m, args->L3offset); 983 #endif 984 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 985 } else 986 FREE_PKT(m); 987 988 args->m = NULL; 989 } 990 991 #endif /* INET6 */ 992 993 /* 994 * sends a reject message, consuming the mbuf passed as an argument. 995 */ 996 static void 997 send_reject(struct ip_fw_args *args, int code, uint16_t mtu, int iplen, 998 const struct ip *ip) 999 { 1000 #if 0 1001 /* XXX When ip is not guaranteed to be at mtod() we will 1002 * need to account for this */ 1003 * The mbuf will however be thrown away so we can adjust it. 1004 * Remember we did an m_pullup on it already so we 1005 * can make some assumptions about contiguousness. 1006 */ 1007 if (args->L3offset) 1008 m_adj(m, args->L3offset); 1009 #endif 1010 if (code != ICMP_REJECT_RST && code != ICMP_REJECT_ABORT) { 1011 /* Send an ICMP unreach */ 1012 icmp_error(args->m, ICMP_UNREACH, code, 0L, mtu); 1013 } else if (code == ICMP_REJECT_RST && args->f_id.proto == IPPROTO_TCP) { 1014 struct tcphdr *const tcp = 1015 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1016 if ( (tcp_get_flags(tcp) & TH_RST) == 0) { 1017 struct mbuf *m; 1018 m = ipfw_send_pkt(args->m, &(args->f_id), 1019 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1020 tcp_get_flags(tcp) | TH_RST); 1021 if (m != NULL) 1022 ip_output(m, NULL, NULL, 0, NULL, NULL); 1023 } 1024 FREE_PKT(args->m); 1025 } else if (code == ICMP_REJECT_ABORT && 1026 args->f_id.proto == IPPROTO_SCTP) { 1027 struct mbuf *m; 1028 struct sctphdr *sctp; 1029 struct sctp_chunkhdr *chunk; 1030 struct sctp_init *init; 1031 u_int32_t v_tag; 1032 int reflected; 1033 1034 sctp = L3HDR(struct sctphdr, mtod(args->m, struct ip *)); 1035 reflected = 1; 1036 v_tag = ntohl(sctp->v_tag); 1037 if (iplen >= (ip->ip_hl << 2) + sizeof(struct sctphdr) + 1038 sizeof(struct sctp_chunkhdr)) { 1039 /* Look at the first chunk header if available */ 1040 chunk = (struct sctp_chunkhdr *)(sctp + 1); 1041 switch (chunk->chunk_type) { 1042 case SCTP_INITIATION: 1043 /* 1044 * Packets containing an INIT chunk MUST have 1045 * a zero v-tag. 1046 */ 1047 if (v_tag != 0) { 1048 v_tag = 0; 1049 break; 1050 } 1051 /* INIT chunk MUST NOT be bundled */ 1052 if (iplen > 1053 (ip->ip_hl << 2) + sizeof(struct sctphdr) + 1054 ntohs(chunk->chunk_length) + 3) { 1055 break; 1056 } 1057 /* Use the initiate tag if available */ 1058 if ((iplen >= (ip->ip_hl << 2) + 1059 sizeof(struct sctphdr) + 1060 sizeof(struct sctp_chunkhdr) + 1061 offsetof(struct sctp_init, a_rwnd))) { 1062 init = (struct sctp_init *)(chunk + 1); 1063 v_tag = ntohl(init->initiate_tag); 1064 reflected = 0; 1065 } 1066 break; 1067 case SCTP_ABORT_ASSOCIATION: 1068 /* 1069 * If the packet contains an ABORT chunk, don't 1070 * reply. 1071 * XXX: We should search through all chunks, 1072 * but do not do that to avoid attacks. 1073 */ 1074 v_tag = 0; 1075 break; 1076 } 1077 } 1078 if (v_tag == 0) { 1079 m = NULL; 1080 } else { 1081 m = ipfw_send_abort(args->m, &(args->f_id), v_tag, 1082 reflected); 1083 } 1084 if (m != NULL) 1085 ip_output(m, NULL, NULL, 0, NULL, NULL); 1086 FREE_PKT(args->m); 1087 } else 1088 FREE_PKT(args->m); 1089 args->m = NULL; 1090 } 1091 1092 /* 1093 * Support for uid/gid/jail lookup. These tests are expensive 1094 * (because we may need to look into the list of active sockets) 1095 * so we cache the results. ugid_lookupp is 0 if we have not 1096 * yet done a lookup, 1 if we succeeded, and -1 if we tried 1097 * and failed. The function always returns the match value. 1098 * We could actually spare the variable and use *uc, setting 1099 * it to '(void *)check_uidgid if we have no info, NULL if 1100 * we tried and failed, or any other value if successful. 1101 */ 1102 static int 1103 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp, 1104 struct ucred **uc) 1105 { 1106 #if defined(USERSPACE) 1107 return 0; // not supported in userspace 1108 #else 1109 #ifndef __FreeBSD__ 1110 /* XXX */ 1111 return cred_check(insn, proto, oif, 1112 dst_ip, dst_port, src_ip, src_port, 1113 (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb); 1114 #else /* FreeBSD */ 1115 struct in_addr src_ip, dst_ip; 1116 struct inpcbinfo *pi; 1117 struct ipfw_flow_id *id; 1118 struct inpcb *pcb, *inp; 1119 int lookupflags; 1120 int match; 1121 1122 id = &args->f_id; 1123 inp = args->inp; 1124 1125 /* 1126 * Check to see if the UDP or TCP stack supplied us with 1127 * the PCB. If so, rather then holding a lock and looking 1128 * up the PCB, we can use the one that was supplied. 1129 */ 1130 if (inp && *ugid_lookupp == 0) { 1131 INP_LOCK_ASSERT(inp); 1132 if (inp->inp_socket != NULL) { 1133 *uc = crhold(inp->inp_cred); 1134 *ugid_lookupp = 1; 1135 } else 1136 *ugid_lookupp = -1; 1137 } 1138 /* 1139 * If we have already been here and the packet has no 1140 * PCB entry associated with it, then we can safely 1141 * assume that this is a no match. 1142 */ 1143 if (*ugid_lookupp == -1) 1144 return (0); 1145 if (id->proto == IPPROTO_TCP) { 1146 lookupflags = 0; 1147 pi = &V_tcbinfo; 1148 } else if (id->proto == IPPROTO_UDP) { 1149 lookupflags = INPLOOKUP_WILDCARD; 1150 pi = &V_udbinfo; 1151 } else if (id->proto == IPPROTO_UDPLITE) { 1152 lookupflags = INPLOOKUP_WILDCARD; 1153 pi = &V_ulitecbinfo; 1154 } else 1155 return 0; 1156 lookupflags |= INPLOOKUP_RLOCKPCB; 1157 match = 0; 1158 if (*ugid_lookupp == 0) { 1159 if (id->addr_type == 6) { 1160 #ifdef INET6 1161 if (args->flags & IPFW_ARGS_IN) 1162 pcb = in6_pcblookup_mbuf(pi, 1163 &id->src_ip6, htons(id->src_port), 1164 &id->dst_ip6, htons(id->dst_port), 1165 lookupflags, NULL, args->m); 1166 else 1167 pcb = in6_pcblookup_mbuf(pi, 1168 &id->dst_ip6, htons(id->dst_port), 1169 &id->src_ip6, htons(id->src_port), 1170 lookupflags, args->ifp, args->m); 1171 #else 1172 *ugid_lookupp = -1; 1173 return (0); 1174 #endif 1175 } else { 1176 src_ip.s_addr = htonl(id->src_ip); 1177 dst_ip.s_addr = htonl(id->dst_ip); 1178 if (args->flags & IPFW_ARGS_IN) 1179 pcb = in_pcblookup_mbuf(pi, 1180 src_ip, htons(id->src_port), 1181 dst_ip, htons(id->dst_port), 1182 lookupflags, NULL, args->m); 1183 else 1184 pcb = in_pcblookup_mbuf(pi, 1185 dst_ip, htons(id->dst_port), 1186 src_ip, htons(id->src_port), 1187 lookupflags, args->ifp, args->m); 1188 } 1189 if (pcb != NULL) { 1190 INP_RLOCK_ASSERT(pcb); 1191 *uc = crhold(pcb->inp_cred); 1192 *ugid_lookupp = 1; 1193 INP_RUNLOCK(pcb); 1194 } 1195 if (*ugid_lookupp == 0) { 1196 /* 1197 * We tried and failed, set the variable to -1 1198 * so we will not try again on this packet. 1199 */ 1200 *ugid_lookupp = -1; 1201 return (0); 1202 } 1203 } 1204 if (insn->o.opcode == O_UID) 1205 match = ((*uc)->cr_uid == (uid_t)insn->d[0]); 1206 else if (insn->o.opcode == O_GID) 1207 match = groupmember((gid_t)insn->d[0], *uc); 1208 else if (insn->o.opcode == O_JAIL) 1209 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]); 1210 return (match); 1211 #endif /* __FreeBSD__ */ 1212 #endif /* not supported in userspace */ 1213 } 1214 1215 /* 1216 * Helper function to set args with info on the rule after the matching 1217 * one. slot is precise, whereas we guess rule_id as they are 1218 * assigned sequentially. 1219 */ 1220 static inline void 1221 set_match(struct ip_fw_args *args, int slot, 1222 struct ip_fw_chain *chain) 1223 { 1224 args->rule.chain_id = chain->id; 1225 args->rule.slot = slot + 1; /* we use 0 as a marker */ 1226 args->rule.rule_id = 1 + chain->map[slot]->id; 1227 args->rule.rulenum = chain->map[slot]->rulenum; 1228 args->flags |= IPFW_ARGS_REF; 1229 } 1230 1231 static int 1232 jump_lookup_pos(struct ip_fw_chain *chain, struct ip_fw *f, int num, 1233 int tablearg, int jump_backwards) 1234 { 1235 int f_pos, i; 1236 1237 i = IP_FW_ARG_TABLEARG(chain, num, skipto); 1238 /* make sure we do not jump backward */ 1239 if (jump_backwards == 0 && i <= f->rulenum) 1240 i = f->rulenum + 1; 1241 1242 #ifndef LINEAR_SKIPTO 1243 if (chain->idxmap != NULL) 1244 f_pos = chain->idxmap[i]; 1245 else 1246 f_pos = ipfw_find_rule(chain, i, 0); 1247 #else 1248 f_pos = chain->idxmap[i]; 1249 #endif /* LINEAR_SKIPTO */ 1250 1251 return (f_pos); 1252 } 1253 1254 1255 #ifndef LINEAR_SKIPTO 1256 /* 1257 * Helper function to enable cached rule lookups using 1258 * cache.id and cache.pos fields in ipfw rule. 1259 */ 1260 static int 1261 jump_cached(struct ip_fw_chain *chain, struct ip_fw *f, int num, 1262 int tablearg, int 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 #endif /* !LINEAR_SKIPTO */ 1308 1309 #define TARG(k, f) IP_FW_ARG_TABLEARG(chain, k, f) 1310 /* 1311 * The main check routine for the firewall. 1312 * 1313 * All arguments are in args so we can modify them and return them 1314 * back to the caller. 1315 * 1316 * Parameters: 1317 * 1318 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1319 * Starts with the IP header. 1320 * args->L3offset Number of bytes bypassed if we came from L2. 1321 * e.g. often sizeof(eh) ** NOTYET ** 1322 * args->ifp Incoming or outgoing interface. 1323 * args->divert_rule (in/out) 1324 * Skip up to the first rule past this rule number; 1325 * upon return, non-zero port number for divert or tee. 1326 * 1327 * args->rule Pointer to the last matching rule (in/out) 1328 * args->next_hop Socket we are forwarding to (out). 1329 * args->next_hop6 IPv6 next hop we are forwarding to (out). 1330 * args->f_id Addresses grabbed from the packet (out) 1331 * args->rule.info a cookie depending on rule action 1332 * 1333 * Return value: 1334 * 1335 * IP_FW_PASS the packet must be accepted 1336 * IP_FW_DENY the packet must be dropped 1337 * IP_FW_DIVERT divert packet, port in m_tag 1338 * IP_FW_TEE tee packet, port in m_tag 1339 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 1340 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 1341 * args->rule contains the matching rule, 1342 * args->rule.info has additional information. 1343 * 1344 */ 1345 int 1346 ipfw_chk(struct ip_fw_args *args) 1347 { 1348 1349 /* 1350 * Local variables holding state while processing a packet: 1351 * 1352 * IMPORTANT NOTE: to speed up the processing of rules, there 1353 * are some assumption on the values of the variables, which 1354 * are documented here. Should you change them, please check 1355 * the implementation of the various instructions to make sure 1356 * that they still work. 1357 * 1358 * m | args->m Pointer to the mbuf, as received from the caller. 1359 * It may change if ipfw_chk() does an m_pullup, or if it 1360 * consumes the packet because it calls send_reject(). 1361 * XXX This has to change, so that ipfw_chk() never modifies 1362 * or consumes the buffer. 1363 * OR 1364 * args->mem Pointer to contigous memory chunk. 1365 * ip Is the beginning of the ip(4 or 6) header. 1366 * eh Ethernet header in case if input is Layer2. 1367 */ 1368 struct mbuf *m; 1369 struct ip *ip; 1370 struct ether_header *eh; 1371 1372 /* 1373 * For rules which contain uid/gid or jail constraints, cache 1374 * a copy of the users credentials after the pcb lookup has been 1375 * executed. This will speed up the processing of rules with 1376 * these types of constraints, as well as decrease contention 1377 * on pcb related locks. 1378 */ 1379 #ifndef __FreeBSD__ 1380 struct bsd_ucred ucred_cache; 1381 #else 1382 struct ucred *ucred_cache = NULL; 1383 #endif 1384 int ucred_lookup = 0; 1385 int f_pos = 0; /* index of current rule in the array */ 1386 int retval = 0; 1387 struct ifnet *oif, *iif; 1388 1389 /* 1390 * hlen The length of the IP header. 1391 */ 1392 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 1393 1394 /* 1395 * offset The offset of a fragment. offset != 0 means that 1396 * we have a fragment at this offset of an IPv4 packet. 1397 * offset == 0 means that (if this is an IPv4 packet) 1398 * this is the first or only fragment. 1399 * For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header 1400 * or there is a single packet fragment (fragment header added 1401 * without needed). We will treat a single packet fragment as if 1402 * there was no fragment header (or log/block depending on the 1403 * V_fw_permit_single_frag6 sysctl setting). 1404 */ 1405 u_short offset = 0; 1406 u_short ip6f_mf = 0; 1407 1408 /* 1409 * Local copies of addresses. They are only valid if we have 1410 * an IP packet. 1411 * 1412 * proto The protocol. Set to 0 for non-ip packets, 1413 * or to the protocol read from the packet otherwise. 1414 * proto != 0 means that we have an IPv4 packet. 1415 * 1416 * src_port, dst_port port numbers, in HOST format. Only 1417 * valid for TCP and UDP packets. 1418 * 1419 * src_ip, dst_ip ip addresses, in NETWORK format. 1420 * Only valid for IPv4 packets. 1421 */ 1422 uint8_t proto; 1423 uint16_t src_port, dst_port; /* NOTE: host format */ 1424 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 1425 int iplen = 0; 1426 int pktlen; 1427 1428 struct ipfw_dyn_info dyn_info; 1429 struct ip_fw *q = NULL; 1430 struct ip_fw_chain *chain = &V_layer3_chain; 1431 1432 /* 1433 * We store in ulp a pointer to the upper layer protocol header. 1434 * In the ipv4 case this is easy to determine from the header, 1435 * but for ipv6 we might have some additional headers in the middle. 1436 * ulp is NULL if not found. 1437 */ 1438 void *ulp = NULL; /* upper layer protocol pointer. */ 1439 1440 /* XXX ipv6 variables */ 1441 int is_ipv6 = 0; 1442 #ifdef INET6 1443 uint8_t icmp6_type = 0; 1444 #endif 1445 uint16_t ext_hd = 0; /* bits vector for extension header filtering */ 1446 /* end of ipv6 variables */ 1447 1448 int is_ipv4 = 0; 1449 1450 int done = 0; /* flag to exit the outer loop */ 1451 IPFW_RLOCK_TRACKER; 1452 bool mem; 1453 bool need_send_reject = false; 1454 int reject_code; 1455 uint16_t reject_mtu; 1456 1457 if ((mem = (args->flags & IPFW_ARGS_LENMASK))) { 1458 if (args->flags & IPFW_ARGS_ETHER) { 1459 eh = (struct ether_header *)args->mem; 1460 if (eh->ether_type == htons(ETHERTYPE_VLAN)) 1461 ip = (struct ip *) 1462 ((struct ether_vlan_header *)eh + 1); 1463 else 1464 ip = (struct ip *)(eh + 1); 1465 } else { 1466 eh = NULL; 1467 ip = (struct ip *)args->mem; 1468 } 1469 pktlen = IPFW_ARGS_LENGTH(args->flags); 1470 args->f_id.fib = args->ifp->if_fib; /* best guess */ 1471 } else { 1472 m = args->m; 1473 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready)) 1474 return (IP_FW_PASS); /* accept */ 1475 if (args->flags & IPFW_ARGS_ETHER) { 1476 /* We need some amount of data to be contiguous. */ 1477 if (m->m_len < min(m->m_pkthdr.len, max_protohdr) && 1478 (args->m = m = m_pullup(m, min(m->m_pkthdr.len, 1479 max_protohdr))) == NULL) 1480 goto pullup_failed; 1481 eh = mtod(m, struct ether_header *); 1482 ip = (struct ip *)(eh + 1); 1483 } else { 1484 eh = NULL; 1485 ip = mtod(m, struct ip *); 1486 } 1487 pktlen = m->m_pkthdr.len; 1488 args->f_id.fib = M_GETFIB(m); /* mbuf not altered */ 1489 } 1490 1491 dst_ip.s_addr = 0; /* make sure it is initialized */ 1492 src_ip.s_addr = 0; /* make sure it is initialized */ 1493 src_port = dst_port = 0; 1494 1495 DYN_INFO_INIT(&dyn_info); 1496 /* 1497 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 1498 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 1499 * pointer might become stale after other pullups (but we never use it 1500 * this way). 1501 */ 1502 #define PULLUP_TO(_len, p, T) PULLUP_LEN(_len, p, sizeof(T)) 1503 #define EHLEN (eh != NULL ? ((char *)ip - (char *)eh) : 0) 1504 #define _PULLUP_LOCKED(_len, p, T, unlock) \ 1505 do { \ 1506 int x = (_len) + T + EHLEN; \ 1507 if (mem) { \ 1508 if (__predict_false(pktlen < x)) { \ 1509 unlock; \ 1510 goto pullup_failed; \ 1511 } \ 1512 p = (char *)args->mem + (_len) + EHLEN; \ 1513 } else { \ 1514 if (__predict_false((m)->m_len < x)) { \ 1515 args->m = m = m_pullup(m, x); \ 1516 if (m == NULL) { \ 1517 unlock; \ 1518 goto pullup_failed; \ 1519 } \ 1520 } \ 1521 p = mtod(m, char *) + (_len) + EHLEN; \ 1522 } \ 1523 } while (0) 1524 1525 #define PULLUP_LEN(_len, p, T) _PULLUP_LOCKED(_len, p, T, ) 1526 #define PULLUP_LEN_LOCKED(_len, p, T) \ 1527 _PULLUP_LOCKED(_len, p, T, IPFW_PF_RUNLOCK(chain)); \ 1528 UPDATE_POINTERS() 1529 /* 1530 * In case pointers got stale after pullups, update them. 1531 */ 1532 #define UPDATE_POINTERS() \ 1533 do { \ 1534 if (!mem) { \ 1535 if (eh != NULL) { \ 1536 eh = mtod(m, struct ether_header *); \ 1537 ip = (struct ip *)(eh + 1); \ 1538 } else \ 1539 ip = mtod(m, struct ip *); \ 1540 args->m = m; \ 1541 } \ 1542 } while (0) 1543 1544 /* Identify IP packets and fill up variables. */ 1545 if (pktlen >= sizeof(struct ip6_hdr) && 1546 (eh == NULL || eh->ether_type == htons(ETHERTYPE_IPV6)) && 1547 ip->ip_v == 6) { 1548 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 1549 1550 is_ipv6 = 1; 1551 args->flags |= IPFW_ARGS_IP6; 1552 hlen = sizeof(struct ip6_hdr); 1553 proto = ip6->ip6_nxt; 1554 /* Search extension headers to find upper layer protocols */ 1555 while (ulp == NULL && offset == 0) { 1556 switch (proto) { 1557 case IPPROTO_ICMPV6: 1558 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 1559 #ifdef INET6 1560 icmp6_type = ICMP6(ulp)->icmp6_type; 1561 #endif 1562 break; 1563 1564 case IPPROTO_TCP: 1565 PULLUP_TO(hlen, ulp, struct tcphdr); 1566 dst_port = TCP(ulp)->th_dport; 1567 src_port = TCP(ulp)->th_sport; 1568 /* save flags for dynamic rules */ 1569 args->f_id._flags = tcp_get_flags(TCP(ulp)); 1570 break; 1571 1572 case IPPROTO_SCTP: 1573 if (pktlen >= hlen + sizeof(struct sctphdr) + 1574 sizeof(struct sctp_chunkhdr) + 1575 offsetof(struct sctp_init, a_rwnd)) 1576 PULLUP_LEN(hlen, ulp, 1577 sizeof(struct sctphdr) + 1578 sizeof(struct sctp_chunkhdr) + 1579 offsetof(struct sctp_init, a_rwnd)); 1580 else if (pktlen >= hlen + sizeof(struct sctphdr)) 1581 PULLUP_LEN(hlen, ulp, pktlen - hlen); 1582 else 1583 PULLUP_LEN(hlen, ulp, 1584 sizeof(struct sctphdr)); 1585 src_port = SCTP(ulp)->src_port; 1586 dst_port = SCTP(ulp)->dest_port; 1587 break; 1588 1589 case IPPROTO_UDP: 1590 case IPPROTO_UDPLITE: 1591 PULLUP_TO(hlen, ulp, struct udphdr); 1592 dst_port = UDP(ulp)->uh_dport; 1593 src_port = UDP(ulp)->uh_sport; 1594 break; 1595 1596 case IPPROTO_HOPOPTS: /* RFC 2460 */ 1597 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1598 ext_hd |= EXT_HOPOPTS; 1599 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1600 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1601 ulp = NULL; 1602 break; 1603 1604 case IPPROTO_ROUTING: /* RFC 2460 */ 1605 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 1606 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 1607 case 0: 1608 ext_hd |= EXT_RTHDR0; 1609 break; 1610 case 2: 1611 ext_hd |= EXT_RTHDR2; 1612 break; 1613 default: 1614 if (V_fw_verbose) 1615 printf("IPFW2: IPV6 - Unknown " 1616 "Routing Header type(%d)\n", 1617 ((struct ip6_rthdr *) 1618 ulp)->ip6r_type); 1619 if (V_fw_deny_unknown_exthdrs) 1620 return (IP_FW_DENY); 1621 break; 1622 } 1623 ext_hd |= EXT_ROUTING; 1624 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 1625 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 1626 ulp = NULL; 1627 break; 1628 1629 case IPPROTO_FRAGMENT: /* RFC 2460 */ 1630 PULLUP_TO(hlen, ulp, struct ip6_frag); 1631 ext_hd |= EXT_FRAGMENT; 1632 hlen += sizeof (struct ip6_frag); 1633 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 1634 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 1635 IP6F_OFF_MASK; 1636 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg & 1637 IP6F_MORE_FRAG; 1638 if (V_fw_permit_single_frag6 == 0 && 1639 offset == 0 && ip6f_mf == 0) { 1640 if (V_fw_verbose) 1641 printf("IPFW2: IPV6 - Invalid " 1642 "Fragment Header\n"); 1643 if (V_fw_deny_unknown_exthdrs) 1644 return (IP_FW_DENY); 1645 break; 1646 } 1647 args->f_id.extra = 1648 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 1649 ulp = NULL; 1650 break; 1651 1652 case IPPROTO_DSTOPTS: /* RFC 2460 */ 1653 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1654 ext_hd |= EXT_DSTOPTS; 1655 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1656 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1657 ulp = NULL; 1658 break; 1659 1660 case IPPROTO_AH: /* RFC 2402 */ 1661 PULLUP_TO(hlen, ulp, struct ip6_ext); 1662 ext_hd |= EXT_AH; 1663 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 1664 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 1665 ulp = NULL; 1666 break; 1667 1668 case IPPROTO_ESP: /* RFC 2406 */ 1669 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 1670 /* Anything past Seq# is variable length and 1671 * data past this ext. header is encrypted. */ 1672 ext_hd |= EXT_ESP; 1673 break; 1674 1675 case IPPROTO_NONE: /* RFC 2460 */ 1676 /* 1677 * Packet ends here, and IPv6 header has 1678 * already been pulled up. If ip6e_len!=0 1679 * then octets must be ignored. 1680 */ 1681 ulp = ip; /* non-NULL to get out of loop. */ 1682 break; 1683 1684 case IPPROTO_OSPFIGP: 1685 /* XXX OSPF header check? */ 1686 PULLUP_TO(hlen, ulp, struct ip6_ext); 1687 break; 1688 1689 case IPPROTO_PIM: 1690 /* XXX PIM header check? */ 1691 PULLUP_TO(hlen, ulp, struct pim); 1692 break; 1693 1694 case IPPROTO_GRE: /* RFC 1701 */ 1695 /* XXX GRE header check? */ 1696 PULLUP_TO(hlen, ulp, struct grehdr); 1697 break; 1698 1699 case IPPROTO_CARP: 1700 PULLUP_TO(hlen, ulp, offsetof( 1701 struct carp_header, carp_counter)); 1702 if (CARP_ADVERTISEMENT != 1703 ((struct carp_header *)ulp)->carp_type) 1704 return (IP_FW_DENY); 1705 break; 1706 1707 case IPPROTO_IPV6: /* RFC 2893 */ 1708 PULLUP_TO(hlen, ulp, struct ip6_hdr); 1709 break; 1710 1711 case IPPROTO_IPV4: /* RFC 2893 */ 1712 PULLUP_TO(hlen, ulp, struct ip); 1713 break; 1714 1715 case IPPROTO_PFSYNC: 1716 PULLUP_TO(hlen, ulp, struct pfsync_header); 1717 break; 1718 1719 default: 1720 if (V_fw_verbose) 1721 printf("IPFW2: IPV6 - Unknown " 1722 "Extension Header(%d), ext_hd=%x\n", 1723 proto, ext_hd); 1724 if (V_fw_deny_unknown_exthdrs) 1725 return (IP_FW_DENY); 1726 PULLUP_TO(hlen, ulp, struct ip6_ext); 1727 break; 1728 } /*switch */ 1729 } 1730 UPDATE_POINTERS(); 1731 ip6 = (struct ip6_hdr *)ip; 1732 args->f_id.addr_type = 6; 1733 args->f_id.src_ip6 = ip6->ip6_src; 1734 args->f_id.dst_ip6 = ip6->ip6_dst; 1735 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 1736 iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6); 1737 } else if (pktlen >= sizeof(struct ip) && 1738 (eh == NULL || eh->ether_type == htons(ETHERTYPE_IP)) && 1739 ip->ip_v == 4) { 1740 is_ipv4 = 1; 1741 args->flags |= IPFW_ARGS_IP4; 1742 hlen = ip->ip_hl << 2; 1743 /* 1744 * Collect parameters into local variables for faster 1745 * matching. 1746 */ 1747 proto = ip->ip_p; 1748 src_ip = ip->ip_src; 1749 dst_ip = ip->ip_dst; 1750 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1751 iplen = ntohs(ip->ip_len); 1752 1753 if (offset == 0) { 1754 switch (proto) { 1755 case IPPROTO_TCP: 1756 PULLUP_TO(hlen, ulp, struct tcphdr); 1757 dst_port = TCP(ulp)->th_dport; 1758 src_port = TCP(ulp)->th_sport; 1759 /* save flags for dynamic rules */ 1760 args->f_id._flags = tcp_get_flags(TCP(ulp)); 1761 break; 1762 1763 case IPPROTO_SCTP: 1764 if (pktlen >= hlen + sizeof(struct sctphdr) + 1765 sizeof(struct sctp_chunkhdr) + 1766 offsetof(struct sctp_init, a_rwnd)) 1767 PULLUP_LEN(hlen, ulp, 1768 sizeof(struct sctphdr) + 1769 sizeof(struct sctp_chunkhdr) + 1770 offsetof(struct sctp_init, a_rwnd)); 1771 else if (pktlen >= hlen + sizeof(struct sctphdr)) 1772 PULLUP_LEN(hlen, ulp, pktlen - hlen); 1773 else 1774 PULLUP_LEN(hlen, ulp, 1775 sizeof(struct sctphdr)); 1776 src_port = SCTP(ulp)->src_port; 1777 dst_port = SCTP(ulp)->dest_port; 1778 break; 1779 1780 case IPPROTO_UDP: 1781 case IPPROTO_UDPLITE: 1782 PULLUP_TO(hlen, ulp, struct udphdr); 1783 dst_port = UDP(ulp)->uh_dport; 1784 src_port = UDP(ulp)->uh_sport; 1785 break; 1786 1787 case IPPROTO_ICMP: 1788 PULLUP_TO(hlen, ulp, struct icmphdr); 1789 //args->f_id.flags = ICMP(ulp)->icmp_type; 1790 break; 1791 1792 default: 1793 break; 1794 } 1795 } else { 1796 if (offset == 1 && proto == IPPROTO_TCP) { 1797 /* RFC 3128 */ 1798 goto pullup_failed; 1799 } 1800 } 1801 1802 UPDATE_POINTERS(); 1803 args->f_id.addr_type = 4; 1804 args->f_id.src_ip = ntohl(src_ip.s_addr); 1805 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1806 } else { 1807 proto = 0; 1808 dst_ip.s_addr = src_ip.s_addr = 0; 1809 1810 args->f_id.addr_type = 1; /* XXX */ 1811 } 1812 #undef PULLUP_TO 1813 pktlen = iplen < pktlen ? iplen: pktlen; 1814 1815 /* Properly initialize the rest of f_id */ 1816 args->f_id.proto = proto; 1817 args->f_id.src_port = src_port = ntohs(src_port); 1818 args->f_id.dst_port = dst_port = ntohs(dst_port); 1819 1820 IPFW_PF_RLOCK(chain); 1821 if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */ 1822 IPFW_PF_RUNLOCK(chain); 1823 return (IP_FW_PASS); /* accept */ 1824 } 1825 if (args->flags & IPFW_ARGS_REF) { 1826 /* 1827 * Packet has already been tagged as a result of a previous 1828 * match on rule args->rule aka args->rule_id (PIPE, QUEUE, 1829 * REASS, NETGRAPH, DIVERT/TEE...) 1830 * Validate the slot and continue from the next one 1831 * if still present, otherwise do a lookup. 1832 */ 1833 f_pos = (args->rule.chain_id == chain->id) ? 1834 args->rule.slot : 1835 ipfw_find_rule(chain, args->rule.rulenum, 1836 args->rule.rule_id); 1837 } else { 1838 f_pos = 0; 1839 } 1840 1841 if (args->flags & IPFW_ARGS_IN) { 1842 iif = args->ifp; 1843 oif = NULL; 1844 } else { 1845 MPASS(args->flags & IPFW_ARGS_OUT); 1846 iif = mem ? NULL : m_rcvif(m); 1847 oif = args->ifp; 1848 } 1849 1850 /* 1851 * Now scan the rules, and parse microinstructions for each rule. 1852 * We have two nested loops and an inner switch. Sometimes we 1853 * need to break out of one or both loops, or re-enter one of 1854 * the loops with updated variables. Loop variables are: 1855 * 1856 * f_pos (outer loop) points to the current rule. 1857 * On output it points to the matching rule. 1858 * done (outer loop) is used as a flag to break the loop. 1859 * l (inner loop) residual length of current rule. 1860 * cmd points to the current microinstruction. 1861 * 1862 * We break the inner loop by setting l=0 and possibly 1863 * cmdlen=0 if we don't want to advance cmd. 1864 * We break the outer loop by setting done=1 1865 * We can restart the inner loop by setting l>0 and f_pos, f, cmd 1866 * as needed. 1867 */ 1868 for (; f_pos < chain->n_rules; f_pos++) { 1869 ipfw_insn *cmd; 1870 uint32_t tablearg = 0; 1871 int l, cmdlen, skip_or; /* skip rest of OR block */ 1872 struct ip_fw *f; 1873 1874 f = chain->map[f_pos]; 1875 if (V_set_disable & (1 << f->set) ) 1876 continue; 1877 1878 skip_or = 0; 1879 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1880 l -= cmdlen, cmd += cmdlen) { 1881 int match; 1882 1883 /* 1884 * check_body is a jump target used when we find a 1885 * CHECK_STATE, and need to jump to the body of 1886 * the target rule. 1887 */ 1888 1889 /* check_body: */ 1890 cmdlen = F_LEN(cmd); 1891 /* 1892 * An OR block (insn_1 || .. || insn_n) has the 1893 * F_OR bit set in all but the last instruction. 1894 * The first match will set "skip_or", and cause 1895 * the following instructions to be skipped until 1896 * past the one with the F_OR bit clear. 1897 */ 1898 if (skip_or) { /* skip this instruction */ 1899 if ((cmd->len & F_OR) == 0) 1900 skip_or = 0; /* next one is good */ 1901 continue; 1902 } 1903 match = 0; /* set to 1 if we succeed */ 1904 1905 switch (cmd->opcode) { 1906 /* 1907 * The first set of opcodes compares the packet's 1908 * fields with some pattern, setting 'match' if a 1909 * match is found. At the end of the loop there is 1910 * logic to deal with F_NOT and F_OR flags associated 1911 * with the opcode. 1912 */ 1913 case O_NOP: 1914 match = 1; 1915 break; 1916 1917 case O_FORWARD_MAC: 1918 printf("ipfw: opcode %d unimplemented\n", 1919 cmd->opcode); 1920 break; 1921 1922 case O_GID: 1923 case O_UID: 1924 case O_JAIL: 1925 /* 1926 * We only check offset == 0 && proto != 0, 1927 * as this ensures that we have a 1928 * packet with the ports info. 1929 */ 1930 if (offset != 0) 1931 break; 1932 if (proto == IPPROTO_TCP || 1933 proto == IPPROTO_UDP || 1934 proto == IPPROTO_UDPLITE) 1935 match = check_uidgid( 1936 (ipfw_insn_u32 *)cmd, 1937 args, &ucred_lookup, 1938 #ifdef __FreeBSD__ 1939 &ucred_cache); 1940 #else 1941 (void *)&ucred_cache); 1942 #endif 1943 break; 1944 1945 case O_RECV: 1946 match = iface_match(iif, (ipfw_insn_if *)cmd, 1947 chain, &tablearg); 1948 break; 1949 1950 case O_XMIT: 1951 match = iface_match(oif, (ipfw_insn_if *)cmd, 1952 chain, &tablearg); 1953 break; 1954 1955 case O_VIA: 1956 match = iface_match(args->ifp, 1957 (ipfw_insn_if *)cmd, chain, &tablearg); 1958 break; 1959 1960 case O_MACADDR2: 1961 if (args->flags & IPFW_ARGS_ETHER) { 1962 u_int32_t *want = (u_int32_t *) 1963 ((ipfw_insn_mac *)cmd)->addr; 1964 u_int32_t *mask = (u_int32_t *) 1965 ((ipfw_insn_mac *)cmd)->mask; 1966 u_int32_t *hdr = (u_int32_t *)eh; 1967 1968 match = 1969 ( want[0] == (hdr[0] & mask[0]) && 1970 want[1] == (hdr[1] & mask[1]) && 1971 want[2] == (hdr[2] & mask[2]) ); 1972 } 1973 break; 1974 1975 case O_MAC_TYPE: 1976 if (args->flags & IPFW_ARGS_ETHER) { 1977 u_int16_t *p = 1978 ((ipfw_insn_u16 *)cmd)->ports; 1979 int i; 1980 1981 for (i = cmdlen - 1; !match && i>0; 1982 i--, p += 2) 1983 match = 1984 (ntohs(eh->ether_type) >= 1985 p[0] && 1986 ntohs(eh->ether_type) <= 1987 p[1]); 1988 } 1989 break; 1990 1991 case O_FRAG: 1992 if (is_ipv4) { 1993 /* 1994 * Since flags_match() works with 1995 * uint8_t we pack ip_off into 8 bits. 1996 * For this match offset is a boolean. 1997 */ 1998 match = flags_match(cmd, 1999 ((ntohs(ip->ip_off) & ~IP_OFFMASK) 2000 >> 8) | (offset != 0)); 2001 } else { 2002 /* 2003 * Compatibility: historically bare 2004 * "frag" would match IPv6 fragments. 2005 */ 2006 match = (cmd->arg1 == 0x1 && 2007 (offset != 0)); 2008 } 2009 break; 2010 2011 case O_IN: /* "out" is "not in" */ 2012 match = (oif == NULL); 2013 break; 2014 2015 case O_LAYER2: 2016 match = (args->flags & IPFW_ARGS_ETHER); 2017 break; 2018 2019 case O_DIVERTED: 2020 if ((args->flags & IPFW_ARGS_REF) == 0) 2021 break; 2022 /* 2023 * For diverted packets, args->rule.info 2024 * contains the divert port (in host format) 2025 * reason and direction. 2026 */ 2027 match = ((args->rule.info & IPFW_IS_MASK) == 2028 IPFW_IS_DIVERT) && ( 2029 ((args->rule.info & IPFW_INFO_IN) ? 2030 1: 2) & cmd->arg1); 2031 break; 2032 2033 case O_PROTO: 2034 /* 2035 * We do not allow an arg of 0 so the 2036 * check of "proto" only suffices. 2037 */ 2038 match = (proto == cmd->arg1); 2039 break; 2040 2041 case O_IP_SRC: 2042 match = is_ipv4 && 2043 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2044 src_ip.s_addr); 2045 break; 2046 2047 case O_IP_DST_LOOKUP: 2048 { 2049 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) { 2050 void *pkey; 2051 uint32_t vidx, key; 2052 uint16_t keylen = 0; /* zero if can't match the packet */ 2053 2054 /* Determine lookup key type */ 2055 vidx = ((ipfw_insn_u32 *)cmd)->d[1]; 2056 switch (vidx) { 2057 case LOOKUP_DST_IP: 2058 case LOOKUP_SRC_IP: 2059 /* Need IP frame */ 2060 if (is_ipv6 == 0 && is_ipv4 == 0) 2061 break; 2062 if (vidx == LOOKUP_DST_IP) 2063 pkey = is_ipv6 ? 2064 (void *)&args->f_id.dst_ip6: 2065 (void *)&dst_ip; 2066 else 2067 pkey = is_ipv6 ? 2068 (void *)&args->f_id.src_ip6: 2069 (void *)&src_ip; 2070 keylen = is_ipv6 ? 2071 sizeof(struct in6_addr): 2072 sizeof(in_addr_t); 2073 break; 2074 case LOOKUP_DST_PORT: 2075 case LOOKUP_SRC_PORT: 2076 /* Need IP frame */ 2077 if (is_ipv6 == 0 && is_ipv4 == 0) 2078 break; 2079 /* Skip fragments */ 2080 if (offset != 0) 2081 break; 2082 /* Skip proto without ports */ 2083 if (proto != IPPROTO_TCP && 2084 proto != IPPROTO_UDP && 2085 proto != IPPROTO_UDPLITE && 2086 proto != IPPROTO_SCTP) 2087 break; 2088 key = vidx == LOOKUP_DST_PORT ? 2089 dst_port: 2090 src_port; 2091 pkey = &key; 2092 keylen = sizeof(key); 2093 break; 2094 case LOOKUP_UID: 2095 case LOOKUP_JAIL: 2096 check_uidgid( 2097 (ipfw_insn_u32 *)cmd, 2098 args, &ucred_lookup, 2099 &ucred_cache); 2100 key = vidx == LOOKUP_UID ? 2101 ucred_cache->cr_uid: 2102 ucred_cache->cr_prison->pr_id; 2103 pkey = &key; 2104 keylen = sizeof(key); 2105 break; 2106 case LOOKUP_DSCP: 2107 /* Need IP frame */ 2108 if (is_ipv6 == 0 && is_ipv4 == 0) 2109 break; 2110 if (is_ipv6) 2111 key = IPV6_DSCP( 2112 (struct ip6_hdr *)ip) >> 2; 2113 else 2114 key = ip->ip_tos >> 2; 2115 pkey = &key; 2116 keylen = sizeof(key); 2117 break; 2118 case LOOKUP_DST_MAC: 2119 case LOOKUP_SRC_MAC: 2120 /* Need ether frame */ 2121 if ((args->flags & IPFW_ARGS_ETHER) == 0) 2122 break; 2123 pkey = vidx == LOOKUP_DST_MAC ? 2124 eh->ether_dhost: 2125 eh->ether_shost; 2126 keylen = ETHER_ADDR_LEN; 2127 break; 2128 case LOOKUP_MARK: 2129 key = args->rule.pkt_mark; 2130 pkey = &key; 2131 keylen = sizeof(key); 2132 break; 2133 } 2134 if (keylen == 0) 2135 break; 2136 match = ipfw_lookup_table(chain, 2137 cmd->arg1, keylen, pkey, &vidx); 2138 if (!match) 2139 break; 2140 tablearg = vidx; 2141 break; 2142 } 2143 /* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */ 2144 /* FALLTHROUGH */ 2145 } 2146 case O_IP_SRC_LOOKUP: 2147 { 2148 void *pkey; 2149 uint32_t vidx; 2150 uint16_t keylen; 2151 2152 if (is_ipv4) { 2153 keylen = sizeof(in_addr_t); 2154 if (cmd->opcode == O_IP_DST_LOOKUP) 2155 pkey = &dst_ip; 2156 else 2157 pkey = &src_ip; 2158 } else if (is_ipv6) { 2159 keylen = sizeof(struct in6_addr); 2160 if (cmd->opcode == O_IP_DST_LOOKUP) 2161 pkey = &args->f_id.dst_ip6; 2162 else 2163 pkey = &args->f_id.src_ip6; 2164 } else 2165 break; 2166 match = ipfw_lookup_table(chain, cmd->arg1, 2167 keylen, pkey, &vidx); 2168 if (!match) 2169 break; 2170 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) { 2171 match = ((ipfw_insn_u32 *)cmd)->d[0] == 2172 TARG_VAL(chain, vidx, tag); 2173 if (!match) 2174 break; 2175 } 2176 tablearg = vidx; 2177 break; 2178 } 2179 2180 case O_MAC_SRC_LOOKUP: 2181 case O_MAC_DST_LOOKUP: 2182 { 2183 void *pkey; 2184 uint32_t vidx; 2185 uint16_t keylen = ETHER_ADDR_LEN; 2186 2187 /* Need ether frame */ 2188 if ((args->flags & IPFW_ARGS_ETHER) == 0) 2189 break; 2190 2191 if (cmd->opcode == O_MAC_DST_LOOKUP) 2192 pkey = eh->ether_dhost; 2193 else 2194 pkey = eh->ether_shost; 2195 2196 match = ipfw_lookup_table(chain, cmd->arg1, 2197 keylen, pkey, &vidx); 2198 if (!match) 2199 break; 2200 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) { 2201 match = ((ipfw_insn_u32 *)cmd)->d[0] == 2202 TARG_VAL(chain, vidx, tag); 2203 if (!match) 2204 break; 2205 } 2206 tablearg = vidx; 2207 break; 2208 } 2209 2210 case O_IP_FLOW_LOOKUP: 2211 { 2212 uint32_t v = 0; 2213 match = ipfw_lookup_table(chain, 2214 cmd->arg1, 0, &args->f_id, &v); 2215 if (!match) 2216 break; 2217 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2218 match = ((ipfw_insn_u32 *)cmd)->d[0] == 2219 TARG_VAL(chain, v, tag); 2220 if (match) 2221 tablearg = v; 2222 } 2223 break; 2224 case O_IP_SRC_MASK: 2225 case O_IP_DST_MASK: 2226 if (is_ipv4) { 2227 uint32_t a = 2228 (cmd->opcode == O_IP_DST_MASK) ? 2229 dst_ip.s_addr : src_ip.s_addr; 2230 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2231 int i = cmdlen-1; 2232 2233 for (; !match && i>0; i-= 2, p+= 2) 2234 match = (p[0] == (a & p[1])); 2235 } 2236 break; 2237 2238 case O_IP_SRC_ME: 2239 if (is_ipv4) { 2240 match = in_localip(src_ip); 2241 break; 2242 } 2243 #ifdef INET6 2244 /* FALLTHROUGH */ 2245 case O_IP6_SRC_ME: 2246 match = is_ipv6 && 2247 ipfw_localip6(&args->f_id.src_ip6); 2248 #endif 2249 break; 2250 2251 case O_IP_DST_SET: 2252 case O_IP_SRC_SET: 2253 if (is_ipv4) { 2254 u_int32_t *d = (u_int32_t *)(cmd+1); 2255 u_int32_t addr = 2256 cmd->opcode == O_IP_DST_SET ? 2257 args->f_id.dst_ip : 2258 args->f_id.src_ip; 2259 2260 if (addr < d[0]) 2261 break; 2262 addr -= d[0]; /* subtract base */ 2263 match = (addr < cmd->arg1) && 2264 ( d[ 1 + (addr>>5)] & 2265 (1<<(addr & 0x1f)) ); 2266 } 2267 break; 2268 2269 case O_IP_DST: 2270 match = is_ipv4 && 2271 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2272 dst_ip.s_addr); 2273 break; 2274 2275 case O_IP_DST_ME: 2276 if (is_ipv4) { 2277 match = in_localip(dst_ip); 2278 break; 2279 } 2280 #ifdef INET6 2281 /* FALLTHROUGH */ 2282 case O_IP6_DST_ME: 2283 match = is_ipv6 && 2284 ipfw_localip6(&args->f_id.dst_ip6); 2285 #endif 2286 break; 2287 2288 case O_IP_SRCPORT: 2289 case O_IP_DSTPORT: 2290 /* 2291 * offset == 0 && proto != 0 is enough 2292 * to guarantee that we have a 2293 * packet with port info. 2294 */ 2295 if ((proto == IPPROTO_UDP || 2296 proto == IPPROTO_UDPLITE || 2297 proto == IPPROTO_TCP || 2298 proto == IPPROTO_SCTP) && offset == 0) { 2299 u_int16_t x = 2300 (cmd->opcode == O_IP_SRCPORT) ? 2301 src_port : dst_port ; 2302 u_int16_t *p = 2303 ((ipfw_insn_u16 *)cmd)->ports; 2304 int i; 2305 2306 for (i = cmdlen - 1; !match && i>0; 2307 i--, p += 2) 2308 match = (x>=p[0] && x<=p[1]); 2309 } 2310 break; 2311 2312 case O_ICMPTYPE: 2313 match = (offset == 0 && proto==IPPROTO_ICMP && 2314 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2315 break; 2316 2317 #ifdef INET6 2318 case O_ICMP6TYPE: 2319 match = is_ipv6 && offset == 0 && 2320 proto==IPPROTO_ICMPV6 && 2321 icmp6type_match( 2322 ICMP6(ulp)->icmp6_type, 2323 (ipfw_insn_u32 *)cmd); 2324 break; 2325 #endif /* INET6 */ 2326 2327 case O_IPOPT: 2328 match = (is_ipv4 && 2329 ipopts_match(ip, cmd) ); 2330 break; 2331 2332 case O_IPVER: 2333 match = ((is_ipv4 || is_ipv6) && 2334 cmd->arg1 == ip->ip_v); 2335 break; 2336 2337 case O_IPID: 2338 case O_IPTTL: 2339 if (!is_ipv4) 2340 break; 2341 case O_IPLEN: 2342 { /* only for IP packets */ 2343 uint16_t x; 2344 uint16_t *p; 2345 int i; 2346 2347 if (cmd->opcode == O_IPLEN) 2348 x = iplen; 2349 else if (cmd->opcode == O_IPTTL) 2350 x = ip->ip_ttl; 2351 else /* must be IPID */ 2352 x = ntohs(ip->ip_id); 2353 if (cmdlen == 1) { 2354 match = (cmd->arg1 == x); 2355 break; 2356 } 2357 /* otherwise we have ranges */ 2358 p = ((ipfw_insn_u16 *)cmd)->ports; 2359 i = cmdlen - 1; 2360 for (; !match && i>0; i--, p += 2) 2361 match = (x >= p[0] && x <= p[1]); 2362 } 2363 break; 2364 2365 case O_IPPRECEDENCE: 2366 match = (is_ipv4 && 2367 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2368 break; 2369 2370 case O_IPTOS: 2371 match = (is_ipv4 && 2372 flags_match(cmd, ip->ip_tos)); 2373 break; 2374 2375 case O_DSCP: 2376 { 2377 uint32_t *p; 2378 uint16_t x; 2379 2380 p = ((ipfw_insn_u32 *)cmd)->d; 2381 2382 if (is_ipv4) 2383 x = ip->ip_tos >> 2; 2384 else if (is_ipv6) { 2385 x = IPV6_DSCP( 2386 (struct ip6_hdr *)ip) >> 2; 2387 x &= 0x3f; 2388 } else 2389 break; 2390 2391 /* DSCP bitmask is stored as low_u32 high_u32 */ 2392 if (x >= 32) 2393 match = *(p + 1) & (1 << (x - 32)); 2394 else 2395 match = *p & (1 << x); 2396 } 2397 break; 2398 2399 case O_TCPDATALEN: 2400 if (proto == IPPROTO_TCP && offset == 0) { 2401 struct tcphdr *tcp; 2402 uint16_t x; 2403 uint16_t *p; 2404 int i; 2405 #ifdef INET6 2406 if (is_ipv6) { 2407 struct ip6_hdr *ip6; 2408 2409 ip6 = (struct ip6_hdr *)ip; 2410 if (ip6->ip6_plen == 0) { 2411 /* 2412 * Jumbo payload is not 2413 * supported by this 2414 * opcode. 2415 */ 2416 break; 2417 } 2418 x = iplen - hlen; 2419 } else 2420 #endif /* INET6 */ 2421 x = iplen - (ip->ip_hl << 2); 2422 tcp = TCP(ulp); 2423 x -= tcp->th_off << 2; 2424 if (cmdlen == 1) { 2425 match = (cmd->arg1 == x); 2426 break; 2427 } 2428 /* otherwise we have ranges */ 2429 p = ((ipfw_insn_u16 *)cmd)->ports; 2430 i = cmdlen - 1; 2431 for (; !match && i>0; i--, p += 2) 2432 match = (x >= p[0] && x <= p[1]); 2433 } 2434 break; 2435 2436 case O_TCPFLAGS: 2437 /* 2438 * Note that this is currently only set up to 2439 * match the lower 8 TCP header flag bits, not 2440 * the full compliment of all 12 flags. 2441 */ 2442 match = (proto == IPPROTO_TCP && offset == 0 && 2443 flags_match(cmd, tcp_get_flags(TCP(ulp)))); 2444 break; 2445 2446 case O_TCPOPTS: 2447 if (proto == IPPROTO_TCP && offset == 0 && ulp){ 2448 PULLUP_LEN_LOCKED(hlen, ulp, 2449 (TCP(ulp)->th_off << 2)); 2450 match = tcpopts_match(TCP(ulp), cmd); 2451 } 2452 break; 2453 2454 case O_TCPSEQ: 2455 match = (proto == IPPROTO_TCP && offset == 0 && 2456 ((ipfw_insn_u32 *)cmd)->d[0] == 2457 TCP(ulp)->th_seq); 2458 break; 2459 2460 case O_TCPACK: 2461 match = (proto == IPPROTO_TCP && offset == 0 && 2462 ((ipfw_insn_u32 *)cmd)->d[0] == 2463 TCP(ulp)->th_ack); 2464 break; 2465 2466 case O_TCPMSS: 2467 if (proto == IPPROTO_TCP && 2468 (args->f_id._flags & TH_SYN) != 0 && 2469 ulp != NULL) { 2470 uint16_t mss, *p; 2471 int i; 2472 2473 PULLUP_LEN_LOCKED(hlen, ulp, 2474 (TCP(ulp)->th_off << 2)); 2475 if ((tcpopts_parse(TCP(ulp), &mss) & 2476 IP_FW_TCPOPT_MSS) == 0) 2477 break; 2478 if (cmdlen == 1) { 2479 match = (cmd->arg1 == mss); 2480 break; 2481 } 2482 /* Otherwise we have ranges. */ 2483 p = ((ipfw_insn_u16 *)cmd)->ports; 2484 i = cmdlen - 1; 2485 for (; !match && i > 0; i--, p += 2) 2486 match = (mss >= p[0] && 2487 mss <= p[1]); 2488 } 2489 break; 2490 2491 case O_TCPWIN: 2492 if (proto == IPPROTO_TCP && offset == 0) { 2493 uint16_t x; 2494 uint16_t *p; 2495 int i; 2496 2497 x = ntohs(TCP(ulp)->th_win); 2498 if (cmdlen == 1) { 2499 match = (cmd->arg1 == x); 2500 break; 2501 } 2502 /* Otherwise we have ranges. */ 2503 p = ((ipfw_insn_u16 *)cmd)->ports; 2504 i = cmdlen - 1; 2505 for (; !match && i > 0; i--, p += 2) 2506 match = (x >= p[0] && x <= p[1]); 2507 } 2508 break; 2509 2510 case O_ESTAB: 2511 /* reject packets which have SYN only */ 2512 /* XXX should i also check for TH_ACK ? */ 2513 match = (proto == IPPROTO_TCP && offset == 0 && 2514 (tcp_get_flags(TCP(ulp)) & 2515 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2516 break; 2517 2518 case O_ALTQ: { 2519 struct pf_mtag *at; 2520 struct m_tag *mtag; 2521 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2522 2523 /* 2524 * ALTQ uses mbuf tags from another 2525 * packet filtering system - pf(4). 2526 * We allocate a tag in its format 2527 * and fill it in, pretending to be pf(4). 2528 */ 2529 match = 1; 2530 at = pf_find_mtag(m); 2531 if (at != NULL && at->qid != 0) 2532 break; 2533 mtag = m_tag_get(PACKET_TAG_PF, 2534 sizeof(struct pf_mtag), M_NOWAIT | M_ZERO); 2535 if (mtag == NULL) { 2536 /* 2537 * Let the packet fall back to the 2538 * default ALTQ. 2539 */ 2540 break; 2541 } 2542 m_tag_prepend(m, mtag); 2543 at = (struct pf_mtag *)(mtag + 1); 2544 at->qid = altq->qid; 2545 at->hdr = ip; 2546 break; 2547 } 2548 2549 case O_LOG: 2550 ipfw_log(chain, f, hlen, args, 2551 offset | ip6f_mf, tablearg, ip); 2552 match = 1; 2553 break; 2554 2555 case O_PROB: 2556 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2557 break; 2558 2559 case O_VERREVPATH: 2560 /* Outgoing packets automatically pass/match */ 2561 match = (args->flags & IPFW_ARGS_OUT || 2562 ( 2563 #ifdef INET6 2564 is_ipv6 ? 2565 verify_path6(&(args->f_id.src_ip6), 2566 iif, args->f_id.fib) : 2567 #endif 2568 verify_path(src_ip, iif, args->f_id.fib))); 2569 break; 2570 2571 case O_VERSRCREACH: 2572 /* Outgoing packets automatically pass/match */ 2573 match = (hlen > 0 && ((oif != NULL) || ( 2574 #ifdef INET6 2575 is_ipv6 ? 2576 verify_path6(&(args->f_id.src_ip6), 2577 NULL, args->f_id.fib) : 2578 #endif 2579 verify_path(src_ip, NULL, args->f_id.fib)))); 2580 break; 2581 2582 case O_ANTISPOOF: 2583 /* Outgoing packets automatically pass/match */ 2584 if (oif == NULL && hlen > 0 && 2585 ( (is_ipv4 && in_localaddr(src_ip)) 2586 #ifdef INET6 2587 || (is_ipv6 && 2588 in6_localaddr(&(args->f_id.src_ip6))) 2589 #endif 2590 )) 2591 match = 2592 #ifdef INET6 2593 is_ipv6 ? verify_path6( 2594 &(args->f_id.src_ip6), iif, 2595 args->f_id.fib) : 2596 #endif 2597 verify_path(src_ip, iif, 2598 args->f_id.fib); 2599 else 2600 match = 1; 2601 break; 2602 2603 case O_IPSEC: 2604 match = (m_tag_find(m, 2605 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2606 /* otherwise no match */ 2607 break; 2608 2609 #ifdef INET6 2610 case O_IP6_SRC: 2611 match = is_ipv6 && 2612 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2613 &((ipfw_insn_ip6 *)cmd)->addr6); 2614 break; 2615 2616 case O_IP6_DST: 2617 match = is_ipv6 && 2618 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2619 &((ipfw_insn_ip6 *)cmd)->addr6); 2620 break; 2621 case O_IP6_SRC_MASK: 2622 case O_IP6_DST_MASK: 2623 if (is_ipv6) { 2624 int i = cmdlen - 1; 2625 struct in6_addr p; 2626 struct in6_addr *d = 2627 &((ipfw_insn_ip6 *)cmd)->addr6; 2628 2629 for (; !match && i > 0; d += 2, 2630 i -= F_INSN_SIZE(struct in6_addr) 2631 * 2) { 2632 p = (cmd->opcode == 2633 O_IP6_SRC_MASK) ? 2634 args->f_id.src_ip6: 2635 args->f_id.dst_ip6; 2636 APPLY_MASK(&p, &d[1]); 2637 match = 2638 IN6_ARE_ADDR_EQUAL(&d[0], 2639 &p); 2640 } 2641 } 2642 break; 2643 2644 case O_FLOW6ID: 2645 match = is_ipv6 && 2646 flow6id_match(args->f_id.flow_id6, 2647 (ipfw_insn_u32 *) cmd); 2648 break; 2649 2650 case O_EXT_HDR: 2651 match = is_ipv6 && 2652 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2653 break; 2654 2655 case O_IP6: 2656 match = is_ipv6; 2657 break; 2658 #endif 2659 2660 case O_IP4: 2661 match = is_ipv4; 2662 break; 2663 2664 case O_TAG: { 2665 struct m_tag *mtag; 2666 uint32_t tag = TARG(cmd->arg1, tag); 2667 2668 /* Packet is already tagged with this tag? */ 2669 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 2670 2671 /* We have `untag' action when F_NOT flag is 2672 * present. And we must remove this mtag from 2673 * mbuf and reset `match' to zero (`match' will 2674 * be inversed later). 2675 * Otherwise we should allocate new mtag and 2676 * push it into mbuf. 2677 */ 2678 if (cmd->len & F_NOT) { /* `untag' action */ 2679 if (mtag != NULL) 2680 m_tag_delete(m, mtag); 2681 match = 0; 2682 } else { 2683 if (mtag == NULL) { 2684 mtag = m_tag_alloc( MTAG_IPFW, 2685 tag, 0, M_NOWAIT); 2686 if (mtag != NULL) 2687 m_tag_prepend(m, mtag); 2688 } 2689 match = 1; 2690 } 2691 break; 2692 } 2693 2694 case O_FIB: /* try match the specified fib */ 2695 if (args->f_id.fib == cmd->arg1) 2696 match = 1; 2697 break; 2698 2699 case O_SOCKARG: { 2700 #ifndef USERSPACE /* not supported in userspace */ 2701 struct inpcb *inp = args->inp; 2702 struct inpcbinfo *pi; 2703 bool inp_locked = false; 2704 2705 if (proto == IPPROTO_TCP) 2706 pi = &V_tcbinfo; 2707 else if (proto == IPPROTO_UDP) 2708 pi = &V_udbinfo; 2709 else if (proto == IPPROTO_UDPLITE) 2710 pi = &V_ulitecbinfo; 2711 else 2712 break; 2713 2714 /* 2715 * XXXRW: so_user_cookie should almost 2716 * certainly be inp_user_cookie? 2717 */ 2718 2719 /* 2720 * For incoming packet lookup the inpcb 2721 * using the src/dest ip/port tuple. 2722 */ 2723 if (is_ipv4 && inp == NULL) { 2724 inp = in_pcblookup(pi, 2725 src_ip, htons(src_port), 2726 dst_ip, htons(dst_port), 2727 INPLOOKUP_RLOCKPCB, NULL); 2728 inp_locked = true; 2729 } 2730 #ifdef INET6 2731 if (is_ipv6 && inp == NULL) { 2732 inp = in6_pcblookup(pi, 2733 &args->f_id.src_ip6, 2734 htons(src_port), 2735 &args->f_id.dst_ip6, 2736 htons(dst_port), 2737 INPLOOKUP_RLOCKPCB, NULL); 2738 inp_locked = true; 2739 } 2740 #endif /* INET6 */ 2741 if (inp != NULL) { 2742 if (inp->inp_socket) { 2743 tablearg = 2744 inp->inp_socket->so_user_cookie; 2745 if (tablearg) 2746 match = 1; 2747 } 2748 if (inp_locked) 2749 INP_RUNLOCK(inp); 2750 } 2751 #endif /* !USERSPACE */ 2752 break; 2753 } 2754 2755 case O_TAGGED: { 2756 struct m_tag *mtag; 2757 uint32_t tag = TARG(cmd->arg1, tag); 2758 2759 if (cmdlen == 1) { 2760 match = m_tag_locate(m, MTAG_IPFW, 2761 tag, NULL) != NULL; 2762 break; 2763 } 2764 2765 /* we have ranges */ 2766 for (mtag = m_tag_first(m); 2767 mtag != NULL && !match; 2768 mtag = m_tag_next(m, mtag)) { 2769 uint16_t *p; 2770 int i; 2771 2772 if (mtag->m_tag_cookie != MTAG_IPFW) 2773 continue; 2774 2775 p = ((ipfw_insn_u16 *)cmd)->ports; 2776 i = cmdlen - 1; 2777 for(; !match && i > 0; i--, p += 2) 2778 match = 2779 mtag->m_tag_id >= p[0] && 2780 mtag->m_tag_id <= p[1]; 2781 } 2782 break; 2783 } 2784 2785 case O_MARK: { 2786 uint32_t mark; 2787 if (cmd->arg1 == IP_FW_TARG) 2788 mark = TARG_VAL(chain, tablearg, mark); 2789 else 2790 mark = ((ipfw_insn_u32 *)cmd)->d[0]; 2791 match = 2792 (args->rule.pkt_mark & 2793 ((ipfw_insn_u32 *)cmd)->d[1]) == 2794 (mark & ((ipfw_insn_u32 *)cmd)->d[1]); 2795 break; 2796 } 2797 2798 /* 2799 * The second set of opcodes represents 'actions', 2800 * i.e. the terminal part of a rule once the packet 2801 * matches all previous patterns. 2802 * Typically there is only one action for each rule, 2803 * and the opcode is stored at the end of the rule 2804 * (but there are exceptions -- see below). 2805 * 2806 * In general, here we set retval and terminate the 2807 * outer loop (would be a 'break 3' in some language, 2808 * but we need to set l=0, done=1) 2809 * 2810 * Exceptions: 2811 * O_COUNT and O_SKIPTO actions: 2812 * instead of terminating, we jump to the next rule 2813 * (setting l=0), or to the SKIPTO target (setting 2814 * f/f_len, cmd and l as needed), respectively. 2815 * 2816 * O_TAG, O_LOG and O_ALTQ action parameters: 2817 * perform some action and set match = 1; 2818 * 2819 * O_LIMIT and O_KEEP_STATE: these opcodes are 2820 * not real 'actions', and are stored right 2821 * before the 'action' part of the rule (one 2822 * exception is O_SKIP_ACTION which could be 2823 * between these opcodes and 'action' one). 2824 * These opcodes try to install an entry in the 2825 * state tables; if successful, we continue with 2826 * the next opcode (match=1; break;), otherwise 2827 * the packet must be dropped (set retval, 2828 * break loops with l=0, done=1) 2829 * 2830 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2831 * cause a lookup of the state table, and a jump 2832 * to the 'action' part of the parent rule 2833 * if an entry is found, or 2834 * (CHECK_STATE only) a jump to the next rule if 2835 * the entry is not found. 2836 * The result of the lookup is cached so that 2837 * further instances of these opcodes become NOPs. 2838 * The jump to the next rule is done by setting 2839 * l=0, cmdlen=0. 2840 * 2841 * O_SKIP_ACTION: this opcode is not a real 'action' 2842 * either, and is stored right before the 'action' 2843 * part of the rule, right after the O_KEEP_STATE 2844 * opcode. It causes match failure so the real 2845 * 'action' could be executed only if the rule 2846 * is checked via dynamic rule from the state 2847 * table, as in such case execution starts 2848 * from the true 'action' opcode directly. 2849 * 2850 */ 2851 case O_LIMIT: 2852 case O_KEEP_STATE: 2853 if (ipfw_dyn_install_state(chain, f, 2854 (ipfw_insn_limit *)cmd, args, ulp, 2855 pktlen, &dyn_info, tablearg)) { 2856 /* error or limit violation */ 2857 retval = IP_FW_DENY; 2858 l = 0; /* exit inner loop */ 2859 done = 1; /* exit outer loop */ 2860 } 2861 match = 1; 2862 break; 2863 2864 case O_PROBE_STATE: 2865 case O_CHECK_STATE: 2866 /* 2867 * dynamic rules are checked at the first 2868 * keep-state or check-state occurrence, 2869 * with the result being stored in dyn_info. 2870 * The compiler introduces a PROBE_STATE 2871 * instruction for us when we have a 2872 * KEEP_STATE (because PROBE_STATE needs 2873 * to be run first). 2874 */ 2875 if (DYN_LOOKUP_NEEDED(&dyn_info, cmd) && 2876 (q = ipfw_dyn_lookup_state(args, ulp, 2877 pktlen, cmd, &dyn_info)) != NULL) { 2878 /* 2879 * Found dynamic entry, jump to the 2880 * 'action' part of the parent rule 2881 * by setting f, cmd, l and clearing 2882 * cmdlen. 2883 */ 2884 f = q; 2885 f_pos = dyn_info.f_pos; 2886 cmd = ACTION_PTR(f); 2887 l = f->cmd_len - f->act_ofs; 2888 cmdlen = 0; 2889 continue; 2890 } 2891 /* 2892 * Dynamic entry not found. If CHECK_STATE, 2893 * skip to next rule, if PROBE_STATE just 2894 * ignore and continue with next opcode. 2895 */ 2896 if (cmd->opcode == O_CHECK_STATE) 2897 l = 0; /* exit inner loop */ 2898 match = 1; 2899 break; 2900 2901 case O_SKIP_ACTION: 2902 match = 0; /* skip to the next rule */ 2903 l = 0; /* exit inner loop */ 2904 break; 2905 2906 case O_ACCEPT: 2907 retval = 0; /* accept */ 2908 l = 0; /* exit inner loop */ 2909 done = 1; /* exit outer loop */ 2910 break; 2911 2912 case O_PIPE: 2913 case O_QUEUE: 2914 set_match(args, f_pos, chain); 2915 args->rule.info = TARG(cmd->arg1, pipe); 2916 if (cmd->opcode == O_PIPE) 2917 args->rule.info |= IPFW_IS_PIPE; 2918 if (V_fw_one_pass) 2919 args->rule.info |= IPFW_ONEPASS; 2920 retval = IP_FW_DUMMYNET; 2921 l = 0; /* exit inner loop */ 2922 done = 1; /* exit outer loop */ 2923 break; 2924 2925 case O_DIVERT: 2926 case O_TEE: 2927 if (args->flags & IPFW_ARGS_ETHER) 2928 break; /* not on layer 2 */ 2929 /* otherwise this is terminal */ 2930 l = 0; /* exit inner loop */ 2931 done = 1; /* exit outer loop */ 2932 retval = (cmd->opcode == O_DIVERT) ? 2933 IP_FW_DIVERT : IP_FW_TEE; 2934 set_match(args, f_pos, chain); 2935 args->rule.info = TARG(cmd->arg1, divert); 2936 break; 2937 2938 case O_COUNT: 2939 IPFW_INC_RULE_COUNTER(f, pktlen); 2940 l = 0; /* exit inner loop */ 2941 break; 2942 2943 case O_SKIPTO: 2944 IPFW_INC_RULE_COUNTER(f, pktlen); 2945 f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0); 2946 /* 2947 * Skip disabled rules, and re-enter 2948 * the inner loop with the correct 2949 * f_pos, f, l and cmd. 2950 * Also clear cmdlen and skip_or 2951 */ 2952 for (; f_pos < chain->n_rules - 1 && 2953 (V_set_disable & 2954 (1 << chain->map[f_pos]->set)); 2955 f_pos++) 2956 ; 2957 /* Re-enter the inner loop at the skipto rule. */ 2958 f = chain->map[f_pos]; 2959 l = f->cmd_len; 2960 cmd = f->cmd; 2961 match = 1; 2962 cmdlen = 0; 2963 skip_or = 0; 2964 continue; 2965 break; /* not reached */ 2966 2967 case O_CALLRETURN: { 2968 /* 2969 * Implementation of `subroutine' call/return, 2970 * in the stack carried in an mbuf tag. This 2971 * is different from `skipto' in that any call 2972 * address is possible (`skipto' must prevent 2973 * backward jumps to avoid endless loops). 2974 * We have `return' action when F_NOT flag is 2975 * present. The `m_tag_id' field is used as 2976 * stack pointer. 2977 */ 2978 struct m_tag *mtag; 2979 uint16_t jmpto, *stack; 2980 2981 #define IS_CALL ((cmd->len & F_NOT) == 0) 2982 #define IS_RETURN ((cmd->len & F_NOT) != 0) 2983 /* 2984 * Hand-rolled version of m_tag_locate() with 2985 * wildcard `type'. 2986 * If not already tagged, allocate new tag. 2987 */ 2988 mtag = m_tag_first(m); 2989 while (mtag != NULL) { 2990 if (mtag->m_tag_cookie == 2991 MTAG_IPFW_CALL) 2992 break; 2993 mtag = m_tag_next(m, mtag); 2994 } 2995 if (mtag == NULL && IS_CALL) { 2996 mtag = m_tag_alloc(MTAG_IPFW_CALL, 0, 2997 IPFW_CALLSTACK_SIZE * 2998 sizeof(uint16_t), M_NOWAIT); 2999 if (mtag != NULL) 3000 m_tag_prepend(m, mtag); 3001 } 3002 3003 /* 3004 * On error both `call' and `return' just 3005 * continue with next rule. 3006 */ 3007 if (IS_RETURN && (mtag == NULL || 3008 mtag->m_tag_id == 0)) { 3009 l = 0; /* exit inner loop */ 3010 break; 3011 } 3012 if (IS_CALL && (mtag == NULL || 3013 mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) { 3014 printf("ipfw: call stack error, " 3015 "go to next rule\n"); 3016 l = 0; /* exit inner loop */ 3017 break; 3018 } 3019 3020 IPFW_INC_RULE_COUNTER(f, pktlen); 3021 stack = (uint16_t *)(mtag + 1); 3022 3023 /* 3024 * The `call' action may use cached f_pos 3025 * (in f->next_rule), whose version is written 3026 * in f->next_rule. 3027 * The `return' action, however, doesn't have 3028 * fixed jump address in cmd->arg1 and can't use 3029 * cache. 3030 */ 3031 if (IS_CALL) { 3032 stack[mtag->m_tag_id] = f->rulenum; 3033 mtag->m_tag_id++; 3034 f_pos = JUMP(chain, f, cmd->arg1, 3035 tablearg, 1); 3036 } else { /* `return' action */ 3037 mtag->m_tag_id--; 3038 jmpto = stack[mtag->m_tag_id] + 1; 3039 f_pos = ipfw_find_rule(chain, jmpto, 0); 3040 } 3041 3042 /* 3043 * Skip disabled rules, and re-enter 3044 * the inner loop with the correct 3045 * f_pos, f, l and cmd. 3046 * Also clear cmdlen and skip_or 3047 */ 3048 for (; f_pos < chain->n_rules - 1 && 3049 (V_set_disable & 3050 (1 << chain->map[f_pos]->set)); f_pos++) 3051 ; 3052 /* Re-enter the inner loop at the dest rule. */ 3053 f = chain->map[f_pos]; 3054 l = f->cmd_len; 3055 cmd = f->cmd; 3056 cmdlen = 0; 3057 skip_or = 0; 3058 continue; 3059 break; /* NOTREACHED */ 3060 } 3061 #undef IS_CALL 3062 #undef IS_RETURN 3063 3064 case O_REJECT: 3065 /* 3066 * Drop the packet and send a reject notice 3067 * if the packet is not ICMP (or is an ICMP 3068 * query), and it is not multicast/broadcast. 3069 */ 3070 if (hlen > 0 && is_ipv4 && offset == 0 && 3071 (proto != IPPROTO_ICMP || 3072 is_icmp_query(ICMP(ulp))) && 3073 !(m->m_flags & (M_BCAST|M_MCAST)) && 3074 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3075 KASSERT(!need_send_reject, 3076 ("o_reject - need_send_reject was set previously")); 3077 if ((reject_code = cmd->arg1) == ICMP_UNREACH_NEEDFRAG && 3078 cmd->len == F_INSN_SIZE(ipfw_insn_u16)) { 3079 reject_mtu = 3080 ((ipfw_insn_u16 *)cmd)->ports[0]; 3081 } else { 3082 reject_mtu = 0; 3083 } 3084 need_send_reject = true; 3085 } 3086 /* FALLTHROUGH */ 3087 #ifdef INET6 3088 case O_UNREACH6: 3089 if (hlen > 0 && is_ipv6 && 3090 ((offset & IP6F_OFF_MASK) == 0) && 3091 (proto != IPPROTO_ICMPV6 || 3092 (is_icmp6_query(icmp6_type) == 1)) && 3093 !(m->m_flags & (M_BCAST|M_MCAST)) && 3094 !IN6_IS_ADDR_MULTICAST( 3095 &args->f_id.dst_ip6)) { 3096 KASSERT(!need_send_reject, 3097 ("o_unreach6 - need_send_reject was set previously")); 3098 reject_code = cmd->arg1; 3099 if (cmd->opcode == O_REJECT) { 3100 reject_code = 3101 map_icmp_unreach(reject_code); 3102 } 3103 need_send_reject = true; 3104 } 3105 /* FALLTHROUGH */ 3106 #endif 3107 case O_DENY: 3108 retval = IP_FW_DENY; 3109 l = 0; /* exit inner loop */ 3110 done = 1; /* exit outer loop */ 3111 break; 3112 3113 case O_FORWARD_IP: 3114 if (args->flags & IPFW_ARGS_ETHER) 3115 break; /* not valid on layer2 pkts */ 3116 if (q != f || 3117 dyn_info.direction == MATCH_FORWARD) { 3118 struct sockaddr_in *sa; 3119 3120 sa = &(((ipfw_insn_sa *)cmd)->sa); 3121 if (sa->sin_addr.s_addr == INADDR_ANY) { 3122 #ifdef INET6 3123 /* 3124 * We use O_FORWARD_IP opcode for 3125 * fwd rule with tablearg, but tables 3126 * now support IPv6 addresses. And 3127 * when we are inspecting IPv6 packet, 3128 * we can use nh6 field from 3129 * table_value as next_hop6 address. 3130 */ 3131 if (is_ipv6) { 3132 struct ip_fw_nh6 *nh6; 3133 3134 args->flags |= IPFW_ARGS_NH6; 3135 nh6 = &args->hopstore6; 3136 nh6->sin6_addr = TARG_VAL( 3137 chain, tablearg, nh6); 3138 nh6->sin6_port = sa->sin_port; 3139 nh6->sin6_scope_id = TARG_VAL( 3140 chain, tablearg, zoneid); 3141 } else 3142 #endif 3143 { 3144 args->flags |= IPFW_ARGS_NH4; 3145 args->hopstore.sin_port = 3146 sa->sin_port; 3147 sa = &args->hopstore; 3148 sa->sin_family = AF_INET; 3149 sa->sin_len = sizeof(*sa); 3150 sa->sin_addr.s_addr = htonl( 3151 TARG_VAL(chain, tablearg, 3152 nh4)); 3153 } 3154 } else { 3155 args->flags |= IPFW_ARGS_NH4PTR; 3156 args->next_hop = sa; 3157 } 3158 } 3159 retval = IP_FW_PASS; 3160 l = 0; /* exit inner loop */ 3161 done = 1; /* exit outer loop */ 3162 break; 3163 3164 #ifdef INET6 3165 case O_FORWARD_IP6: 3166 if (args->flags & IPFW_ARGS_ETHER) 3167 break; /* not valid on layer2 pkts */ 3168 if (q != f || 3169 dyn_info.direction == MATCH_FORWARD) { 3170 struct sockaddr_in6 *sin6; 3171 3172 sin6 = &(((ipfw_insn_sa6 *)cmd)->sa); 3173 args->flags |= IPFW_ARGS_NH6PTR; 3174 args->next_hop6 = sin6; 3175 } 3176 retval = IP_FW_PASS; 3177 l = 0; /* exit inner loop */ 3178 done = 1; /* exit outer loop */ 3179 break; 3180 #endif 3181 3182 case O_NETGRAPH: 3183 case O_NGTEE: 3184 set_match(args, f_pos, chain); 3185 args->rule.info = TARG(cmd->arg1, netgraph); 3186 if (V_fw_one_pass) 3187 args->rule.info |= IPFW_ONEPASS; 3188 retval = (cmd->opcode == O_NETGRAPH) ? 3189 IP_FW_NETGRAPH : IP_FW_NGTEE; 3190 l = 0; /* exit inner loop */ 3191 done = 1; /* exit outer loop */ 3192 break; 3193 3194 case O_SETFIB: { 3195 uint32_t fib; 3196 3197 IPFW_INC_RULE_COUNTER(f, pktlen); 3198 fib = TARG(cmd->arg1, fib) & 0x7FFF; 3199 if (fib >= rt_numfibs) 3200 fib = 0; 3201 M_SETFIB(m, fib); 3202 args->f_id.fib = fib; /* XXX */ 3203 l = 0; /* exit inner loop */ 3204 break; 3205 } 3206 3207 case O_SETDSCP: { 3208 uint16_t code; 3209 3210 code = TARG(cmd->arg1, dscp) & 0x3F; 3211 l = 0; /* exit inner loop */ 3212 if (is_ipv4) { 3213 uint16_t old; 3214 3215 old = *(uint16_t *)ip; 3216 ip->ip_tos = (code << 2) | 3217 (ip->ip_tos & 0x03); 3218 ip->ip_sum = cksum_adjust(ip->ip_sum, 3219 old, *(uint16_t *)ip); 3220 } else if (is_ipv6) { 3221 /* update cached value */ 3222 args->f_id.flow_id6 = 3223 ntohl(*(uint32_t *)ip) & ~0x0FC00000; 3224 args->f_id.flow_id6 |= code << 22; 3225 3226 *((uint32_t *)ip) = 3227 htonl(args->f_id.flow_id6); 3228 } else 3229 break; 3230 3231 IPFW_INC_RULE_COUNTER(f, pktlen); 3232 break; 3233 } 3234 3235 case O_NAT: 3236 l = 0; /* exit inner loop */ 3237 done = 1; /* exit outer loop */ 3238 /* 3239 * Ensure that we do not invoke NAT handler for 3240 * non IPv4 packets. Libalias expects only IPv4. 3241 */ 3242 if (!is_ipv4 || !IPFW_NAT_LOADED) { 3243 retval = IP_FW_DENY; 3244 break; 3245 } 3246 3247 struct cfg_nat *t; 3248 int nat_id; 3249 3250 args->rule.info = 0; 3251 set_match(args, f_pos, chain); 3252 /* Check if this is 'global' nat rule */ 3253 if (cmd->arg1 == IP_FW_NAT44_GLOBAL) { 3254 retval = ipfw_nat_ptr(args, NULL, m); 3255 break; 3256 } 3257 t = ((ipfw_insn_nat *)cmd)->nat; 3258 if (t == NULL) { 3259 nat_id = TARG(cmd->arg1, nat); 3260 t = (*lookup_nat_ptr)(&chain->nat, nat_id); 3261 3262 if (t == NULL) { 3263 retval = IP_FW_DENY; 3264 break; 3265 } 3266 if (cmd->arg1 != IP_FW_TARG) 3267 ((ipfw_insn_nat *)cmd)->nat = t; 3268 } 3269 retval = ipfw_nat_ptr(args, t, m); 3270 break; 3271 3272 case O_REASS: { 3273 int ip_off; 3274 3275 l = 0; /* in any case exit inner loop */ 3276 if (is_ipv6) /* IPv6 is not supported yet */ 3277 break; 3278 IPFW_INC_RULE_COUNTER(f, pktlen); 3279 ip_off = ntohs(ip->ip_off); 3280 3281 /* if not fragmented, go to next rule */ 3282 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0) 3283 break; 3284 3285 args->m = m = ip_reass(m); 3286 3287 /* 3288 * do IP header checksum fixup. 3289 */ 3290 if (m == NULL) { /* fragment got swallowed */ 3291 retval = IP_FW_DENY; 3292 } else { /* good, packet complete */ 3293 int hlen; 3294 3295 ip = mtod(m, struct ip *); 3296 hlen = ip->ip_hl << 2; 3297 ip->ip_sum = 0; 3298 if (hlen == sizeof(struct ip)) 3299 ip->ip_sum = in_cksum_hdr(ip); 3300 else 3301 ip->ip_sum = in_cksum(m, hlen); 3302 retval = IP_FW_REASS; 3303 args->rule.info = 0; 3304 set_match(args, f_pos, chain); 3305 } 3306 done = 1; /* exit outer loop */ 3307 break; 3308 } 3309 3310 case O_SETMARK: { 3311 l = 0; /* exit inner loop */ 3312 args->rule.pkt_mark = ( 3313 (cmd->arg1 == IP_FW_TARG) ? 3314 TARG_VAL(chain, tablearg, mark) : 3315 ((ipfw_insn_u32 *)cmd)->d[0]); 3316 3317 IPFW_INC_RULE_COUNTER(f, pktlen); 3318 break; 3319 } 3320 3321 case O_EXTERNAL_ACTION: 3322 l = 0; /* in any case exit inner loop */ 3323 retval = ipfw_run_eaction(chain, args, 3324 cmd, &done); 3325 /* 3326 * If both @retval and @done are zero, 3327 * consider this as rule matching and 3328 * update counters. 3329 */ 3330 if (retval == 0 && done == 0) { 3331 IPFW_INC_RULE_COUNTER(f, pktlen); 3332 /* 3333 * Reset the result of the last 3334 * dynamic state lookup. 3335 * External action can change 3336 * @args content, and it may be 3337 * used for new state lookup later. 3338 */ 3339 DYN_INFO_INIT(&dyn_info); 3340 } 3341 break; 3342 3343 default: 3344 panic("-- unknown opcode %d\n", cmd->opcode); 3345 } /* end of switch() on opcodes */ 3346 /* 3347 * if we get here with l=0, then match is irrelevant. 3348 */ 3349 3350 if (cmd->len & F_NOT) 3351 match = !match; 3352 3353 if (match) { 3354 if (cmd->len & F_OR) 3355 skip_or = 1; 3356 } else { 3357 if (!(cmd->len & F_OR)) /* not an OR block, */ 3358 break; /* try next rule */ 3359 } 3360 3361 } /* end of inner loop, scan opcodes */ 3362 #undef PULLUP_LEN 3363 #undef PULLUP_LEN_LOCKED 3364 3365 if (done) 3366 break; 3367 3368 /* next_rule:; */ /* try next rule */ 3369 3370 } /* end of outer for, scan rules */ 3371 3372 if (done) { 3373 struct ip_fw *rule = chain->map[f_pos]; 3374 /* Update statistics */ 3375 IPFW_INC_RULE_COUNTER(rule, pktlen); 3376 IPFW_PROBE(rule__matched, retval, 3377 is_ipv4 ? AF_INET : AF_INET6, 3378 is_ipv4 ? (uintptr_t)&src_ip : 3379 (uintptr_t)&args->f_id.src_ip6, 3380 is_ipv4 ? (uintptr_t)&dst_ip : 3381 (uintptr_t)&args->f_id.dst_ip6, 3382 args, rule); 3383 } else { 3384 retval = IP_FW_DENY; 3385 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3386 } 3387 IPFW_PF_RUNLOCK(chain); 3388 if (need_send_reject) { 3389 #ifdef INET6 3390 if (is_ipv6) 3391 send_reject6(args, reject_code, hlen, 3392 (struct ip6_hdr *)ip); 3393 else 3394 #endif 3395 send_reject(args, reject_code, reject_mtu, 3396 iplen, ip); 3397 } 3398 #ifdef __FreeBSD__ 3399 if (ucred_cache != NULL) 3400 crfree(ucred_cache); 3401 #endif 3402 return (retval); 3403 3404 pullup_failed: 3405 if (V_fw_verbose) 3406 printf("ipfw: pullup failed\n"); 3407 return (IP_FW_DENY); 3408 } 3409 3410 /* 3411 * Set maximum number of tables that can be used in given VNET ipfw instance. 3412 */ 3413 #ifdef SYSCTL_NODE 3414 static int 3415 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS) 3416 { 3417 int error; 3418 unsigned int ntables; 3419 3420 ntables = V_fw_tables_max; 3421 3422 error = sysctl_handle_int(oidp, &ntables, 0, req); 3423 /* Read operation or some error */ 3424 if ((error != 0) || (req->newptr == NULL)) 3425 return (error); 3426 3427 return (ipfw_resize_tables(&V_layer3_chain, ntables)); 3428 } 3429 3430 /* 3431 * Switches table namespace between global and per-set. 3432 */ 3433 static int 3434 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS) 3435 { 3436 int error; 3437 unsigned int sets; 3438 3439 sets = V_fw_tables_sets; 3440 3441 error = sysctl_handle_int(oidp, &sets, 0, req); 3442 /* Read operation or some error */ 3443 if ((error != 0) || (req->newptr == NULL)) 3444 return (error); 3445 3446 return (ipfw_switch_tables_namespace(&V_layer3_chain, sets)); 3447 } 3448 #endif 3449 3450 /* 3451 * Module and VNET glue 3452 */ 3453 3454 /* 3455 * Stuff that must be initialised only on boot or module load 3456 */ 3457 static int 3458 ipfw_init(void) 3459 { 3460 int error = 0; 3461 3462 /* 3463 * Only print out this stuff the first time around, 3464 * when called from the sysinit code. 3465 */ 3466 printf("ipfw2 " 3467 #ifdef INET6 3468 "(+ipv6) " 3469 #endif 3470 "initialized, divert %s, nat %s, " 3471 "default to %s, logging ", 3472 #ifdef IPDIVERT 3473 "enabled", 3474 #else 3475 "loadable", 3476 #endif 3477 #ifdef IPFIREWALL_NAT 3478 "enabled", 3479 #else 3480 "loadable", 3481 #endif 3482 default_to_accept ? "accept" : "deny"); 3483 3484 /* 3485 * Note: V_xxx variables can be accessed here but the vnet specific 3486 * initializer may not have been called yet for the VIMAGE case. 3487 * Tuneables will have been processed. We will print out values for 3488 * the default vnet. 3489 * XXX This should all be rationalized AFTER 8.0 3490 */ 3491 if (V_fw_verbose == 0) 3492 printf("disabled\n"); 3493 else if (V_verbose_limit == 0) 3494 printf("unlimited\n"); 3495 else 3496 printf("limited to %d packets/entry by default\n", 3497 V_verbose_limit); 3498 3499 /* Check user-supplied table count for validness */ 3500 if (default_fw_tables > IPFW_TABLES_MAX) 3501 default_fw_tables = IPFW_TABLES_MAX; 3502 3503 ipfw_init_sopt_handler(); 3504 ipfw_init_obj_rewriter(); 3505 ipfw_iface_init(); 3506 return (error); 3507 } 3508 3509 /* 3510 * Called for the removal of the last instance only on module unload. 3511 */ 3512 static void 3513 ipfw_destroy(void) 3514 { 3515 3516 ipfw_iface_destroy(); 3517 ipfw_destroy_sopt_handler(); 3518 ipfw_destroy_obj_rewriter(); 3519 printf("IP firewall unloaded\n"); 3520 } 3521 3522 /* 3523 * Stuff that must be initialized for every instance 3524 * (including the first of course). 3525 */ 3526 static int 3527 vnet_ipfw_init(const void *unused) 3528 { 3529 int error, first; 3530 struct ip_fw *rule = NULL; 3531 struct ip_fw_chain *chain; 3532 3533 chain = &V_layer3_chain; 3534 3535 first = IS_DEFAULT_VNET(curvnet) ? 1 : 0; 3536 3537 /* First set up some values that are compile time options */ 3538 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 3539 V_fw_deny_unknown_exthdrs = 1; 3540 #ifdef IPFIREWALL_VERBOSE 3541 V_fw_verbose = 1; 3542 #endif 3543 #ifdef IPFIREWALL_VERBOSE_LIMIT 3544 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 3545 #endif 3546 #ifdef IPFIREWALL_NAT 3547 LIST_INIT(&chain->nat); 3548 #endif 3549 3550 /* Init shared services hash table */ 3551 ipfw_init_srv(chain); 3552 3553 ipfw_init_counters(); 3554 /* Set initial number of tables */ 3555 V_fw_tables_max = default_fw_tables; 3556 error = ipfw_init_tables(chain, first); 3557 if (error) { 3558 printf("ipfw2: setting up tables failed\n"); 3559 free(chain->map, M_IPFW); 3560 free(rule, M_IPFW); 3561 return (ENOSPC); 3562 } 3563 3564 IPFW_LOCK_INIT(chain); 3565 3566 /* fill and insert the default rule */ 3567 rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw)); 3568 rule->flags |= IPFW_RULE_NOOPT; 3569 rule->cmd_len = 1; 3570 rule->cmd[0].len = 1; 3571 rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; 3572 chain->default_rule = rule; 3573 ipfw_add_protected_rule(chain, rule, 0); 3574 3575 ipfw_dyn_init(chain); 3576 ipfw_eaction_init(chain, first); 3577 #ifdef LINEAR_SKIPTO 3578 ipfw_init_skipto_cache(chain); 3579 #endif 3580 ipfw_bpf_init(first); 3581 3582 /* First set up some values that are compile time options */ 3583 V_ipfw_vnet_ready = 1; /* Open for business */ 3584 3585 /* 3586 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6. 3587 * Even if the latter two fail we still keep the module alive 3588 * because the sockopt and layer2 paths are still useful. 3589 * ipfw[6]_hook return 0 on success, ENOENT on failure, 3590 * so we can ignore the exact return value and just set a flag. 3591 * 3592 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so 3593 * changes in the underlying (per-vnet) variables trigger 3594 * immediate hook()/unhook() calls. 3595 * In layer2 we have the same behaviour, except that V_ether_ipfw 3596 * is checked on each packet because there are no pfil hooks. 3597 */ 3598 V_ip_fw_ctl_ptr = ipfw_ctl3; 3599 error = ipfw_attach_hooks(); 3600 return (error); 3601 } 3602 3603 /* 3604 * Called for the removal of each instance. 3605 */ 3606 static int 3607 vnet_ipfw_uninit(const void *unused) 3608 { 3609 struct ip_fw *reap; 3610 struct ip_fw_chain *chain = &V_layer3_chain; 3611 int i, last; 3612 3613 V_ipfw_vnet_ready = 0; /* tell new callers to go away */ 3614 /* 3615 * disconnect from ipv4, ipv6, layer2 and sockopt. 3616 * Then grab, release and grab again the WLOCK so we make 3617 * sure the update is propagated and nobody will be in. 3618 */ 3619 ipfw_detach_hooks(); 3620 V_ip_fw_ctl_ptr = NULL; 3621 3622 last = IS_DEFAULT_VNET(curvnet) ? 1 : 0; 3623 3624 IPFW_UH_WLOCK(chain); 3625 IPFW_UH_WUNLOCK(chain); 3626 3627 ipfw_dyn_uninit(0); /* run the callout_drain */ 3628 3629 IPFW_UH_WLOCK(chain); 3630 3631 reap = NULL; 3632 IPFW_WLOCK(chain); 3633 for (i = 0; i < chain->n_rules; i++) 3634 ipfw_reap_add(chain, &reap, chain->map[i]); 3635 free(chain->map, M_IPFW); 3636 #ifdef LINEAR_SKIPTO 3637 ipfw_destroy_skipto_cache(chain); 3638 #endif 3639 IPFW_WUNLOCK(chain); 3640 IPFW_UH_WUNLOCK(chain); 3641 ipfw_destroy_tables(chain, last); 3642 ipfw_eaction_uninit(chain, last); 3643 if (reap != NULL) 3644 ipfw_reap_rules(reap); 3645 vnet_ipfw_iface_destroy(chain); 3646 ipfw_destroy_srv(chain); 3647 IPFW_LOCK_DESTROY(chain); 3648 ipfw_dyn_uninit(1); /* free the remaining parts */ 3649 ipfw_destroy_counters(); 3650 ipfw_bpf_uninit(last); 3651 return (0); 3652 } 3653 3654 /* 3655 * Module event handler. 3656 * In general we have the choice of handling most of these events by the 3657 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to 3658 * use the SYSINIT handlers as they are more capable of expressing the 3659 * flow of control during module and vnet operations, so this is just 3660 * a skeleton. Note there is no SYSINIT equivalent of the module 3661 * SHUTDOWN handler, but we don't have anything to do in that case anyhow. 3662 */ 3663 static int 3664 ipfw_modevent(module_t mod, int type, void *unused) 3665 { 3666 int err = 0; 3667 3668 switch (type) { 3669 case MOD_LOAD: 3670 /* Called once at module load or 3671 * system boot if compiled in. */ 3672 break; 3673 case MOD_QUIESCE: 3674 /* Called before unload. May veto unloading. */ 3675 break; 3676 case MOD_UNLOAD: 3677 /* Called during unload. */ 3678 break; 3679 case MOD_SHUTDOWN: 3680 /* Called during system shutdown. */ 3681 break; 3682 default: 3683 err = EOPNOTSUPP; 3684 break; 3685 } 3686 return err; 3687 } 3688 3689 static moduledata_t ipfwmod = { 3690 "ipfw", 3691 ipfw_modevent, 3692 0 3693 }; 3694 3695 /* Define startup order. */ 3696 #define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_FIREWALL 3697 #define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */ 3698 #define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */ 3699 #define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */ 3700 3701 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER); 3702 FEATURE(ipfw_ctl3, "ipfw new sockopt calls"); 3703 MODULE_VERSION(ipfw, 3); 3704 /* should declare some dependencies here */ 3705 3706 /* 3707 * Starting up. Done in order after ipfwmod() has been called. 3708 * VNET_SYSINIT is also called for each existing vnet and each new vnet. 3709 */ 3710 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 3711 ipfw_init, NULL); 3712 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 3713 vnet_ipfw_init, NULL); 3714 3715 /* 3716 * Closing up shop. These are done in REVERSE ORDER, but still 3717 * after ipfwmod() has been called. Not called on reboot. 3718 * VNET_SYSUNINIT is also called for each exiting vnet as it exits. 3719 * or when the module is unloaded. 3720 */ 3721 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 3722 ipfw_destroy, NULL); 3723 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 3724 vnet_ipfw_uninit, NULL); 3725 /* end of file */ 3726