1/*- 2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD 3 * 4 * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 25 * SUCH DAMAGE. 26 */ 27 28#include <sys/cdefs.h> 29__FBSDID("$FreeBSD$"); 30 31/* 32 * The FreeBSD IP packet firewall, main file 33 */ 34 35#include "opt_ipfw.h" 36#include "opt_ipdivert.h" 37#include "opt_inet.h" 38#ifndef INET 39#error "IPFIREWALL requires INET" 40#endif /* INET */ 41#include "opt_inet6.h" 42 43#include <sys/param.h> 44#include <sys/systm.h> 45#include <sys/condvar.h> 46#include <sys/counter.h> 47#include <sys/eventhandler.h> 48#include <sys/malloc.h> 49#include <sys/mbuf.h> 50#include <sys/kernel.h> 51#include <sys/lock.h> 52#include <sys/jail.h> 53#include <sys/module.h> 54#include <sys/priv.h> 55#include <sys/proc.h> 56#include <sys/rwlock.h> 57#include <sys/rmlock.h> 58#include <sys/sdt.h> 59#include <sys/socket.h> 60#include <sys/socketvar.h> 61#include <sys/sysctl.h> 62#include <sys/syslog.h> 63#include <sys/ucred.h> 64#include <net/ethernet.h> /* for ETHERTYPE_IP */ 65#include <net/if.h> 66#include <net/if_var.h> 67#include <net/route.h> 68#include <net/pfil.h> 69#include <net/vnet.h> 70 71#include <netpfil/pf/pf_mtag.h> 72 73#include <netinet/in.h> 74#include <netinet/in_var.h> 75#include <netinet/in_pcb.h> 76#include <netinet/ip.h> 77#include <netinet/ip_var.h> 78#include <netinet/ip_icmp.h> 79#include <netinet/ip_fw.h> 80#include <netinet/ip_carp.h> 81#include <netinet/pim.h> 82#include <netinet/tcp_var.h> 83#include <netinet/udp.h> 84#include <netinet/udp_var.h> 85#include <netinet/sctp.h> 86#include <netinet/sctp_crc32.h> 87#include <netinet/sctp_header.h> 88 89#include <netinet/ip6.h> 90#include <netinet/icmp6.h> 91#include <netinet/in_fib.h> 92#ifdef INET6 93#include <netinet6/in6_fib.h> 94#include <netinet6/in6_pcb.h> 95#include <netinet6/scope6_var.h> 96#include <netinet6/ip6_var.h> 97#endif 98 99#include <net/if_gre.h> /* for struct grehdr */ 100 101#include <netpfil/ipfw/ip_fw_private.h> 102 103#include <machine/in_cksum.h> /* XXX for in_cksum */ 104 105#ifdef MAC 106#include <security/mac/mac_framework.h> 107#endif 108 109#define IPFW_PROBE(probe, arg0, arg1, arg2, arg3, arg4, arg5) \ 110 SDT_PROBE6(ipfw, , , probe, arg0, arg1, arg2, arg3, arg4, arg5) 111 112SDT_PROVIDER_DEFINE(ipfw); 113SDT_PROBE_DEFINE6(ipfw, , , rule__matched, 114 "int", /* retval */ 115 "int", /* af */ 116 "void *", /* src addr */ 117 "void *", /* dst addr */ 118 "struct ip_fw_args *", /* args */ 119 "struct ip_fw *" /* rule */); 120 121/* 122 * static variables followed by global ones. 123 * All ipfw global variables are here. 124 */ 125 126VNET_DEFINE_STATIC(int, fw_deny_unknown_exthdrs); 127#define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs) 128 129VNET_DEFINE_STATIC(int, fw_permit_single_frag6) = 1; 130#define V_fw_permit_single_frag6 VNET(fw_permit_single_frag6) 131 132#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 133static int default_to_accept = 1; 134#else 135static int default_to_accept; 136#endif 137 138VNET_DEFINE(int, autoinc_step); 139VNET_DEFINE(int, fw_one_pass) = 1; 140 141VNET_DEFINE(unsigned int, fw_tables_max); 142VNET_DEFINE(unsigned int, fw_tables_sets) = 0; /* Don't use set-aware tables */ 143/* Use 128 tables by default */ 144static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT; 145 146#ifndef LINEAR_SKIPTO 147static int jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num, 148 int tablearg, int jump_backwards); 149#define JUMP(ch, f, num, targ, back) jump_fast(ch, f, num, targ, back) 150#else 151static int jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num, 152 int tablearg, int jump_backwards); 153#define JUMP(ch, f, num, targ, back) jump_linear(ch, f, num, targ, back) 154#endif 155 156/* 157 * Each rule belongs to one of 32 different sets (0..31). 158 * The variable set_disable contains one bit per set. 159 * If the bit is set, all rules in the corresponding set 160 * are disabled. Set RESVD_SET(31) is reserved for the default rule 161 * and rules that are not deleted by the flush command, 162 * and CANNOT be disabled. 163 * Rules in set RESVD_SET can only be deleted individually. 164 */ 165VNET_DEFINE(u_int32_t, set_disable); 166#define V_set_disable VNET(set_disable) 167 168VNET_DEFINE(int, fw_verbose); 169/* counter for ipfw_log(NULL...) */ 170VNET_DEFINE(u_int64_t, norule_counter); 171VNET_DEFINE(int, verbose_limit); 172 173/* layer3_chain contains the list of rules for layer 3 */ 174VNET_DEFINE(struct ip_fw_chain, layer3_chain); 175 176/* ipfw_vnet_ready controls when we are open for business */ 177VNET_DEFINE(int, ipfw_vnet_ready) = 0; 178 179VNET_DEFINE(int, ipfw_nat_ready) = 0; 180 181ipfw_nat_t *ipfw_nat_ptr = NULL; 182struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int); 183ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; 184ipfw_nat_cfg_t *ipfw_nat_del_ptr; 185ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; 186ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; 187 188#ifdef SYSCTL_NODE 189uint32_t dummy_def = IPFW_DEFAULT_RULE; 190static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS); 191static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS); 192 193SYSBEGIN(f3) 194 195SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 196SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 197 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0, 198 "Only do a single pass through ipfw when using dummynet(4)"); 199SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, 200 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0, 201 "Rule number auto-increment step"); 202SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 203 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0, 204 "Log matches to ipfw rules"); 205SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, 206 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(verbose_limit), 0, 207 "Set upper limit of matches of ipfw rules logged"); 208SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, 209 &dummy_def, 0, 210 "The default/max possible rule number."); 211SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_max, 212 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU", 213 "Maximum number of concurrently used tables"); 214SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets, 215 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW, 216 0, 0, sysctl_ipfw_tables_sets, "IU", 217 "Use per-set namespace for tables"); 218SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN, 219 &default_to_accept, 0, 220 "Make the default rule accept all packets."); 221TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables); 222SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, 223 CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0, 224 "Number of static rules"); 225 226#ifdef INET6 227SYSCTL_DECL(_net_inet6_ip6); 228SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 229SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs, 230 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, 231 &VNET_NAME(fw_deny_unknown_exthdrs), 0, 232 "Deny packets with unknown IPv6 Extension Headers"); 233SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6, 234 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, 235 &VNET_NAME(fw_permit_single_frag6), 0, 236 "Permit single packet IPv6 fragments"); 237#endif /* INET6 */ 238 239SYSEND 240 241#endif /* SYSCTL_NODE */ 242 243 244/* 245 * Some macros used in the various matching options. 246 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 247 * Other macros just cast void * into the appropriate type 248 */ 249#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 250#define TCP(p) ((struct tcphdr *)(p)) 251#define SCTP(p) ((struct sctphdr *)(p)) 252#define UDP(p) ((struct udphdr *)(p)) 253#define ICMP(p) ((struct icmphdr *)(p)) 254#define ICMP6(p) ((struct icmp6_hdr *)(p)) 255 256static __inline int 257icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 258{ 259 int type = icmp->icmp_type; 260 261 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 262} 263 264#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 265 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 266 267static int 268is_icmp_query(struct icmphdr *icmp) 269{ 270 int type = icmp->icmp_type; 271 272 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 273} 274#undef TT 275 276/* 277 * The following checks use two arrays of 8 or 16 bits to store the 278 * bits that we want set or clear, respectively. They are in the 279 * low and high half of cmd->arg1 or cmd->d[0]. 280 * 281 * We scan options and store the bits we find set. We succeed if 282 * 283 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 284 * 285 * The code is sometimes optimized not to store additional variables. 286 */ 287 288static int 289flags_match(ipfw_insn *cmd, u_int8_t bits) 290{ 291 u_char want_clear; 292 bits = ~bits; 293 294 if ( ((cmd->arg1 & 0xff) & bits) != 0) 295 return 0; /* some bits we want set were clear */ 296 want_clear = (cmd->arg1 >> 8) & 0xff; 297 if ( (want_clear & bits) != want_clear) 298 return 0; /* some bits we want clear were set */ 299 return 1; 300} 301 302static int 303ipopts_match(struct ip *ip, ipfw_insn *cmd) 304{ 305 int optlen, bits = 0; 306 u_char *cp = (u_char *)(ip + 1); 307 int x = (ip->ip_hl << 2) - sizeof (struct ip); 308 309 for (; x > 0; x -= optlen, cp += optlen) { 310 int opt = cp[IPOPT_OPTVAL]; 311 312 if (opt == IPOPT_EOL) 313 break; 314 if (opt == IPOPT_NOP) 315 optlen = 1; 316 else { 317 optlen = cp[IPOPT_OLEN]; 318 if (optlen <= 0 || optlen > x) 319 return 0; /* invalid or truncated */ 320 } 321 switch (opt) { 322 323 default: 324 break; 325 326 case IPOPT_LSRR: 327 bits |= IP_FW_IPOPT_LSRR; 328 break; 329 330 case IPOPT_SSRR: 331 bits |= IP_FW_IPOPT_SSRR; 332 break; 333 334 case IPOPT_RR: 335 bits |= IP_FW_IPOPT_RR; 336 break; 337 338 case IPOPT_TS: 339 bits |= IP_FW_IPOPT_TS; 340 break; 341 } 342 } 343 return (flags_match(cmd, bits)); 344} 345 346/* 347 * Parse TCP options. The logic copied from tcp_dooptions(). 348 */ 349static int 350tcpopts_parse(const struct tcphdr *tcp, uint16_t *mss) 351{ 352 const u_char *cp = (const u_char *)(tcp + 1); 353 int optlen, bits = 0; 354 int cnt = (tcp->th_off << 2) - sizeof(struct tcphdr); 355 356 for (; cnt > 0; cnt -= optlen, cp += optlen) { 357 int opt = cp[0]; 358 if (opt == TCPOPT_EOL) 359 break; 360 if (opt == TCPOPT_NOP) 361 optlen = 1; 362 else { 363 if (cnt < 2) 364 break; 365 optlen = cp[1]; 366 if (optlen < 2 || optlen > cnt) 367 break; 368 } 369 370 switch (opt) { 371 default: 372 break; 373 374 case TCPOPT_MAXSEG: 375 if (optlen != TCPOLEN_MAXSEG) 376 break; 377 bits |= IP_FW_TCPOPT_MSS; 378 if (mss != NULL) 379 *mss = be16dec(cp + 2); 380 break; 381 382 case TCPOPT_WINDOW: 383 if (optlen == TCPOLEN_WINDOW) 384 bits |= IP_FW_TCPOPT_WINDOW; 385 break; 386 387 case TCPOPT_SACK_PERMITTED: 388 if (optlen == TCPOLEN_SACK_PERMITTED) 389 bits |= IP_FW_TCPOPT_SACK; 390 break; 391 392 case TCPOPT_SACK: 393 if (optlen > 2 && (optlen - 2) % TCPOLEN_SACK == 0) 394 bits |= IP_FW_TCPOPT_SACK; 395 break; 396 397 case TCPOPT_TIMESTAMP: 398 if (optlen == TCPOLEN_TIMESTAMP) 399 bits |= IP_FW_TCPOPT_TS; 400 break; 401 } 402 } 403 return (bits); 404} 405 406static int 407tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 408{ 409 410 return (flags_match(cmd, tcpopts_parse(tcp, NULL))); 411} 412 413static int 414iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain, 415 uint32_t *tablearg) 416{ 417 418 if (ifp == NULL) /* no iface with this packet, match fails */ 419 return (0); 420 421 /* Check by name or by IP address */ 422 if (cmd->name[0] != '\0') { /* match by name */ 423 if (cmd->name[0] == '\1') /* use tablearg to match */ 424 return ipfw_lookup_table(chain, cmd->p.kidx, 0, 425 &ifp->if_index, tablearg); 426 /* Check name */ 427 if (cmd->p.glob) { 428 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 429 return(1); 430 } else { 431 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 432 return(1); 433 } 434 } else { 435#if !defined(USERSPACE) && defined(__FreeBSD__) /* and OSX too ? */ 436 struct ifaddr *ia; 437 438 if_addr_rlock(ifp); 439 CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 440 if (ia->ifa_addr->sa_family != AF_INET) 441 continue; 442 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 443 (ia->ifa_addr))->sin_addr.s_addr) { 444 if_addr_runlock(ifp); 445 return(1); /* match */ 446 } 447 } 448 if_addr_runlock(ifp); 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 */ 475static int 476verify_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 nhop4_basic nh4; 482 483 if (fib4_lookup_nh_basic(fib, src, NHR_IFAIF, 0, &nh4) != 0) 484 return (0); 485 486 /* 487 * If ifp is provided, check for equality with rtentry. 488 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 489 * in order to pass packets injected back by if_simloop(): 490 * routing entry (via lo0) for our own address 491 * may exist, so we need to handle routing assymetry. 492 */ 493 if (ifp != NULL && ifp != nh4.nh_ifp) 494 return (0); 495 496 /* if no ifp provided, check if rtentry is not default route */ 497 if (ifp == NULL && (nh4.nh_flags & NHF_DEFAULT) != 0) 498 return (0); 499 500 /* or if this is a blackhole/reject route */ 501 if (ifp == NULL && (nh4.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0) 502 return (0); 503 504 /* found valid route */ 505 return 1; 506#endif /* __FreeBSD__ */ 507} 508 509/* 510 * Generate an SCTP packet containing an ABORT chunk. The verification tag 511 * is given by vtag. The T-bit is set in the ABORT chunk if and only if 512 * reflected is not 0. 513 */ 514 515static struct mbuf * 516ipfw_send_abort(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t vtag, 517 int reflected) 518{ 519 struct mbuf *m; 520 struct ip *ip; 521#ifdef INET6 522 struct ip6_hdr *ip6; 523#endif 524 struct sctphdr *sctp; 525 struct sctp_chunkhdr *chunk; 526 u_int16_t hlen, plen, tlen; 527 528 MGETHDR(m, M_NOWAIT, MT_DATA); 529 if (m == NULL) 530 return (NULL); 531 532 M_SETFIB(m, id->fib); 533#ifdef MAC 534 if (replyto != NULL) 535 mac_netinet_firewall_reply(replyto, m); 536 else 537 mac_netinet_firewall_send(m); 538#else 539 (void)replyto; /* don't warn about unused arg */ 540#endif 541 542 switch (id->addr_type) { 543 case 4: 544 hlen = sizeof(struct ip); 545 break; 546#ifdef INET6 547 case 6: 548 hlen = sizeof(struct ip6_hdr); 549 break; 550#endif 551 default: 552 /* XXX: log me?!? */ 553 FREE_PKT(m); 554 return (NULL); 555 } 556 plen = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr); 557 tlen = hlen + plen; 558 m->m_data += max_linkhdr; 559 m->m_flags |= M_SKIP_FIREWALL; 560 m->m_pkthdr.len = m->m_len = tlen; 561 m->m_pkthdr.rcvif = NULL; 562 bzero(m->m_data, tlen); 563 564 switch (id->addr_type) { 565 case 4: 566 ip = mtod(m, struct ip *); 567 568 ip->ip_v = 4; 569 ip->ip_hl = sizeof(struct ip) >> 2; 570 ip->ip_tos = IPTOS_LOWDELAY; 571 ip->ip_len = htons(tlen); 572 ip->ip_id = htons(0); 573 ip->ip_off = htons(0); 574 ip->ip_ttl = V_ip_defttl; 575 ip->ip_p = IPPROTO_SCTP; 576 ip->ip_sum = 0; 577 ip->ip_src.s_addr = htonl(id->dst_ip); 578 ip->ip_dst.s_addr = htonl(id->src_ip); 579 580 sctp = (struct sctphdr *)(ip + 1); 581 break; 582#ifdef INET6 583 case 6: 584 ip6 = mtod(m, struct ip6_hdr *); 585 586 ip6->ip6_vfc = IPV6_VERSION; 587 ip6->ip6_plen = htons(plen); 588 ip6->ip6_nxt = IPPROTO_SCTP; 589 ip6->ip6_hlim = IPV6_DEFHLIM; 590 ip6->ip6_src = id->dst_ip6; 591 ip6->ip6_dst = id->src_ip6; 592 593 sctp = (struct sctphdr *)(ip6 + 1); 594 break; 595#endif 596 } 597 598 sctp->src_port = htons(id->dst_port); 599 sctp->dest_port = htons(id->src_port); 600 sctp->v_tag = htonl(vtag); 601 sctp->checksum = htonl(0); 602 603 chunk = (struct sctp_chunkhdr *)(sctp + 1); 604 chunk->chunk_type = SCTP_ABORT_ASSOCIATION; 605 chunk->chunk_flags = 0; 606 if (reflected != 0) { 607 chunk->chunk_flags |= SCTP_HAD_NO_TCB; 608 } 609 chunk->chunk_length = htons(sizeof(struct sctp_chunkhdr)); 610 611 sctp->checksum = sctp_calculate_cksum(m, hlen); 612 613 return (m); 614} 615 616/* 617 * Generate a TCP packet, containing either a RST or a keepalive. 618 * When flags & TH_RST, we are sending a RST packet, because of a 619 * "reset" action matched the packet. 620 * Otherwise we are sending a keepalive, and flags & TH_ 621 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 622 * so that MAC can label the reply appropriately. 623 */ 624struct mbuf * 625ipfw_send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 626 u_int32_t ack, int flags) 627{ 628 struct mbuf *m = NULL; /* stupid compiler */ 629 struct ip *h = NULL; /* stupid compiler */ 630#ifdef INET6 631 struct ip6_hdr *h6 = NULL; 632#endif 633 struct tcphdr *th = NULL; 634 int len, dir; 635 636 MGETHDR(m, M_NOWAIT, MT_DATA); 637 if (m == NULL) 638 return (NULL); 639 640 M_SETFIB(m, id->fib); 641#ifdef MAC 642 if (replyto != NULL) 643 mac_netinet_firewall_reply(replyto, m); 644 else 645 mac_netinet_firewall_send(m); 646#else 647 (void)replyto; /* don't warn about unused arg */ 648#endif 649 650 switch (id->addr_type) { 651 case 4: 652 len = sizeof(struct ip) + sizeof(struct tcphdr); 653 break; 654#ifdef INET6 655 case 6: 656 len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); 657 break; 658#endif 659 default: 660 /* XXX: log me?!? */ 661 FREE_PKT(m); 662 return (NULL); 663 } 664 dir = ((flags & (TH_SYN | TH_RST)) == TH_SYN); 665 666 m->m_data += max_linkhdr; 667 m->m_flags |= M_SKIP_FIREWALL; 668 m->m_pkthdr.len = m->m_len = len; 669 m->m_pkthdr.rcvif = NULL; 670 bzero(m->m_data, len); 671 672 switch (id->addr_type) { 673 case 4: 674 h = mtod(m, struct ip *); 675 676 /* prepare for checksum */ 677 h->ip_p = IPPROTO_TCP; 678 h->ip_len = htons(sizeof(struct tcphdr)); 679 if (dir) { 680 h->ip_src.s_addr = htonl(id->src_ip); 681 h->ip_dst.s_addr = htonl(id->dst_ip); 682 } else { 683 h->ip_src.s_addr = htonl(id->dst_ip); 684 h->ip_dst.s_addr = htonl(id->src_ip); 685 } 686 687 th = (struct tcphdr *)(h + 1); 688 break; 689#ifdef INET6 690 case 6: 691 h6 = mtod(m, struct ip6_hdr *); 692 693 /* prepare for checksum */ 694 h6->ip6_nxt = IPPROTO_TCP; 695 h6->ip6_plen = htons(sizeof(struct tcphdr)); 696 if (dir) { 697 h6->ip6_src = id->src_ip6; 698 h6->ip6_dst = id->dst_ip6; 699 } else { 700 h6->ip6_src = id->dst_ip6; 701 h6->ip6_dst = id->src_ip6; 702 } 703 704 th = (struct tcphdr *)(h6 + 1); 705 break; 706#endif 707 } 708 709 if (dir) { 710 th->th_sport = htons(id->src_port); 711 th->th_dport = htons(id->dst_port); 712 } else { 713 th->th_sport = htons(id->dst_port); 714 th->th_dport = htons(id->src_port); 715 } 716 th->th_off = sizeof(struct tcphdr) >> 2; 717 718 if (flags & TH_RST) { 719 if (flags & TH_ACK) { 720 th->th_seq = htonl(ack); 721 th->th_flags = TH_RST; 722 } else { 723 if (flags & TH_SYN) 724 seq++; 725 th->th_ack = htonl(seq); 726 th->th_flags = TH_RST | TH_ACK; 727 } 728 } else { 729 /* 730 * Keepalive - use caller provided sequence numbers 731 */ 732 th->th_seq = htonl(seq); 733 th->th_ack = htonl(ack); 734 th->th_flags = TH_ACK; 735 } 736 737 switch (id->addr_type) { 738 case 4: 739 th->th_sum = in_cksum(m, len); 740 741 /* finish the ip header */ 742 h->ip_v = 4; 743 h->ip_hl = sizeof(*h) >> 2; 744 h->ip_tos = IPTOS_LOWDELAY; 745 h->ip_off = htons(0); 746 h->ip_len = htons(len); 747 h->ip_ttl = V_ip_defttl; 748 h->ip_sum = 0; 749 break; 750#ifdef INET6 751 case 6: 752 th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(*h6), 753 sizeof(struct tcphdr)); 754 755 /* finish the ip6 header */ 756 h6->ip6_vfc |= IPV6_VERSION; 757 h6->ip6_hlim = IPV6_DEFHLIM; 758 break; 759#endif 760 } 761 762 return (m); 763} 764 765#ifdef INET6 766/* 767 * ipv6 specific rules here... 768 */ 769static __inline int 770icmp6type_match(int type, ipfw_insn_u32 *cmd) 771{ 772 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 773} 774 775static int 776flow6id_match(int curr_flow, ipfw_insn_u32 *cmd) 777{ 778 int i; 779 for (i=0; i <= cmd->o.arg1; ++i) 780 if (curr_flow == cmd->d[i]) 781 return 1; 782 return 0; 783} 784 785/* support for IP6_*_ME opcodes */ 786static const struct in6_addr lla_mask = {{{ 787 0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff, 788 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff 789}}}; 790 791static int 792ipfw_localip6(struct in6_addr *in6) 793{ 794 struct rm_priotracker in6_ifa_tracker; 795 struct in6_ifaddr *ia; 796 797 if (IN6_IS_ADDR_MULTICAST(in6)) 798 return (0); 799 800 if (!IN6_IS_ADDR_LINKLOCAL(in6)) 801 return (in6_localip(in6)); 802 803 IN6_IFADDR_RLOCK(&in6_ifa_tracker); 804 CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) { 805 if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr)) 806 continue; 807 if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr, 808 in6, &lla_mask)) { 809 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); 810 return (1); 811 } 812 } 813 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker); 814 return (0); 815} 816 817static int 818verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib) 819{ 820 struct nhop6_basic nh6; 821 822 if (IN6_IS_SCOPE_LINKLOCAL(src)) 823 return (1); 824 825 if (fib6_lookup_nh_basic(fib, src, 0, NHR_IFAIF, 0, &nh6) != 0) 826 return (0); 827 828 /* If ifp is provided, check for equality with route table. */ 829 if (ifp != NULL && ifp != nh6.nh_ifp) 830 return (0); 831 832 /* if no ifp provided, check if rtentry is not default route */ 833 if (ifp == NULL && (nh6.nh_flags & NHF_DEFAULT) != 0) 834 return (0); 835 836 /* or if this is a blackhole/reject route */ 837 if (ifp == NULL && (nh6.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0) 838 return (0); 839 840 /* found valid route */ 841 return 1; 842} 843 844static int 845is_icmp6_query(int icmp6_type) 846{ 847 if ((icmp6_type <= ICMP6_MAXTYPE) && 848 (icmp6_type == ICMP6_ECHO_REQUEST || 849 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 850 icmp6_type == ICMP6_WRUREQUEST || 851 icmp6_type == ICMP6_FQDN_QUERY || 852 icmp6_type == ICMP6_NI_QUERY)) 853 return (1); 854 855 return (0); 856} 857 858static int 859map_icmp_unreach(int code) 860{ 861 862 /* RFC 7915 p4.2 */ 863 switch (code) { 864 case ICMP_UNREACH_NET: 865 case ICMP_UNREACH_HOST: 866 case ICMP_UNREACH_SRCFAIL: 867 case ICMP_UNREACH_NET_UNKNOWN: 868 case ICMP_UNREACH_HOST_UNKNOWN: 869 case ICMP_UNREACH_TOSNET: 870 case ICMP_UNREACH_TOSHOST: 871 return (ICMP6_DST_UNREACH_NOROUTE); 872 case ICMP_UNREACH_PORT: 873 return (ICMP6_DST_UNREACH_NOPORT); 874 default: 875 /* 876 * Map the rest of codes into admit prohibited. 877 * XXX: unreach proto should be mapped into ICMPv6 878 * parameter problem, but we use only unreach type. 879 */ 880 return (ICMP6_DST_UNREACH_ADMIN); 881 } 882} 883 884static void 885send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 886{ 887 struct mbuf *m; 888 889 m = args->m; 890 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 891 struct tcphdr *tcp; 892 tcp = (struct tcphdr *)((char *)ip6 + hlen); 893 894 if ((tcp->th_flags & TH_RST) == 0) { 895 struct mbuf *m0; 896 m0 = ipfw_send_pkt(args->m, &(args->f_id), 897 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 898 tcp->th_flags | TH_RST); 899 if (m0 != NULL) 900 ip6_output(m0, NULL, NULL, 0, NULL, NULL, 901 NULL); 902 } 903 FREE_PKT(m); 904 } else if (code == ICMP6_UNREACH_ABORT && 905 args->f_id.proto == IPPROTO_SCTP) { 906 struct mbuf *m0; 907 struct sctphdr *sctp; 908 u_int32_t v_tag; 909 int reflected; 910 911 sctp = (struct sctphdr *)((char *)ip6 + hlen); 912 reflected = 1; 913 v_tag = ntohl(sctp->v_tag); 914 /* Investigate the first chunk header if available */ 915 if (m->m_len >= hlen + sizeof(struct sctphdr) + 916 sizeof(struct sctp_chunkhdr)) { 917 struct sctp_chunkhdr *chunk; 918 919 chunk = (struct sctp_chunkhdr *)(sctp + 1); 920 switch (chunk->chunk_type) { 921 case SCTP_INITIATION: 922 /* 923 * Packets containing an INIT chunk MUST have 924 * a zero v-tag. 925 */ 926 if (v_tag != 0) { 927 v_tag = 0; 928 break; 929 } 930 /* INIT chunk MUST NOT be bundled */ 931 if (m->m_pkthdr.len > 932 hlen + sizeof(struct sctphdr) + 933 ntohs(chunk->chunk_length) + 3) { 934 break; 935 } 936 /* Use the initiate tag if available */ 937 if ((m->m_len >= hlen + sizeof(struct sctphdr) + 938 sizeof(struct sctp_chunkhdr) + 939 offsetof(struct sctp_init, a_rwnd))) { 940 struct sctp_init *init; 941 942 init = (struct sctp_init *)(chunk + 1); 943 v_tag = ntohl(init->initiate_tag); 944 reflected = 0; 945 } 946 break; 947 case SCTP_ABORT_ASSOCIATION: 948 /* 949 * If the packet contains an ABORT chunk, don't 950 * reply. 951 * XXX: We should search through all chunks, 952 * but do not do that to avoid attacks. 953 */ 954 v_tag = 0; 955 break; 956 } 957 } 958 if (v_tag == 0) { 959 m0 = NULL; 960 } else { 961 m0 = ipfw_send_abort(args->m, &(args->f_id), v_tag, 962 reflected); 963 } 964 if (m0 != NULL) 965 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL); 966 FREE_PKT(m); 967 } else if (code != ICMP6_UNREACH_RST && code != ICMP6_UNREACH_ABORT) { 968 /* Send an ICMPv6 unreach. */ 969#if 0 970 /* 971 * Unlike above, the mbufs need to line up with the ip6 hdr, 972 * as the contents are read. We need to m_adj() the 973 * needed amount. 974 * The mbuf will however be thrown away so we can adjust it. 975 * Remember we did an m_pullup on it already so we 976 * can make some assumptions about contiguousness. 977 */ 978 if (args->L3offset) 979 m_adj(m, args->L3offset); 980#endif 981 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 982 } else 983 FREE_PKT(m); 984 985 args->m = NULL; 986} 987 988#endif /* INET6 */ 989 990 991/* 992 * sends a reject message, consuming the mbuf passed as an argument. 993 */ 994static void 995send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip) 996{ 997 998#if 0 999 /* XXX When ip is not guaranteed to be at mtod() we will 1000 * need to account for this */ 1001 * The mbuf will however be thrown away so we can adjust it. 1002 * Remember we did an m_pullup on it already so we 1003 * can make some assumptions about contiguousness. 1004 */ 1005 if (args->L3offset) 1006 m_adj(m, args->L3offset); 1007#endif 1008 if (code != ICMP_REJECT_RST && code != ICMP_REJECT_ABORT) { 1009 /* Send an ICMP unreach */ 1010 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1011 } else if (code == ICMP_REJECT_RST && args->f_id.proto == IPPROTO_TCP) { 1012 struct tcphdr *const tcp = 1013 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1014 if ( (tcp->th_flags & TH_RST) == 0) { 1015 struct mbuf *m; 1016 m = ipfw_send_pkt(args->m, &(args->f_id), 1017 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1018 tcp->th_flags | TH_RST); 1019 if (m != NULL) 1020 ip_output(m, NULL, NULL, 0, NULL, NULL); 1021 } 1022 FREE_PKT(args->m); 1023 } else if (code == ICMP_REJECT_ABORT && 1024 args->f_id.proto == IPPROTO_SCTP) { 1025 struct mbuf *m; 1026 struct sctphdr *sctp; 1027 struct sctp_chunkhdr *chunk; 1028 struct sctp_init *init; 1029 u_int32_t v_tag; 1030 int reflected; 1031 1032 sctp = L3HDR(struct sctphdr, mtod(args->m, struct ip *)); 1033 reflected = 1; 1034 v_tag = ntohl(sctp->v_tag); 1035 if (iplen >= (ip->ip_hl << 2) + sizeof(struct sctphdr) + 1036 sizeof(struct sctp_chunkhdr)) { 1037 /* Look at the first chunk header if available */ 1038 chunk = (struct sctp_chunkhdr *)(sctp + 1); 1039 switch (chunk->chunk_type) { 1040 case SCTP_INITIATION: 1041 /* 1042 * Packets containing an INIT chunk MUST have 1043 * a zero v-tag. 1044 */ 1045 if (v_tag != 0) { 1046 v_tag = 0; 1047 break; 1048 } 1049 /* INIT chunk MUST NOT be bundled */ 1050 if (iplen > 1051 (ip->ip_hl << 2) + sizeof(struct sctphdr) + 1052 ntohs(chunk->chunk_length) + 3) { 1053 break; 1054 } 1055 /* Use the initiate tag if available */ 1056 if ((iplen >= (ip->ip_hl << 2) + 1057 sizeof(struct sctphdr) + 1058 sizeof(struct sctp_chunkhdr) + 1059 offsetof(struct sctp_init, a_rwnd))) { 1060 init = (struct sctp_init *)(chunk + 1); 1061 v_tag = ntohl(init->initiate_tag); 1062 reflected = 0; 1063 } 1064 break; 1065 case SCTP_ABORT_ASSOCIATION: 1066 /* 1067 * If the packet contains an ABORT chunk, don't 1068 * reply. 1069 * XXX: We should search through all chunks, 1070 * but do not do that to avoid attacks. 1071 */ 1072 v_tag = 0; 1073 break; 1074 } 1075 } 1076 if (v_tag == 0) { 1077 m = NULL; 1078 } else { 1079 m = ipfw_send_abort(args->m, &(args->f_id), v_tag, 1080 reflected); 1081 } 1082 if (m != NULL) 1083 ip_output(m, NULL, NULL, 0, NULL, NULL); 1084 FREE_PKT(args->m); 1085 } else 1086 FREE_PKT(args->m); 1087 args->m = NULL; 1088} 1089 1090/* 1091 * Support for uid/gid/jail lookup. These tests are expensive 1092 * (because we may need to look into the list of active sockets) 1093 * so we cache the results. ugid_lookupp is 0 if we have not 1094 * yet done a lookup, 1 if we succeeded, and -1 if we tried 1095 * and failed. The function always returns the match value. 1096 * We could actually spare the variable and use *uc, setting 1097 * it to '(void *)check_uidgid if we have no info, NULL if 1098 * we tried and failed, or any other value if successful. 1099 */ 1100static int 1101check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp, 1102 struct ucred **uc) 1103{ 1104#if defined(USERSPACE) 1105 return 0; // not supported in userspace 1106#else 1107#ifndef __FreeBSD__ 1108 /* XXX */ 1109 return cred_check(insn, proto, oif, 1110 dst_ip, dst_port, src_ip, src_port, 1111 (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb); 1112#else /* FreeBSD */ 1113 struct in_addr src_ip, dst_ip; 1114 struct inpcbinfo *pi; 1115 struct ipfw_flow_id *id; 1116 struct inpcb *pcb, *inp; 1117 struct ifnet *oif; 1118 int lookupflags; 1119 int match; 1120 1121 id = &args->f_id; 1122 inp = args->inp; 1123 oif = args->oif; 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 (oif == NULL) 1162 pcb = in6_pcblookup_mbuf(pi, 1163 &id->src_ip6, htons(id->src_port), 1164 &id->dst_ip6, htons(id->dst_port), 1165 lookupflags, oif, 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, oif, 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 (oif == NULL) 1179 pcb = in_pcblookup_mbuf(pi, 1180 src_ip, htons(id->src_port), 1181 dst_ip, htons(id->dst_port), 1182 lookupflags, oif, 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, oif, 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 */ 1220static inline void 1221set_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#ifndef LINEAR_SKIPTO 1232/* 1233 * Helper function to enable cached rule lookups using 1234 * cached_id and cached_pos fields in ipfw rule. 1235 */ 1236static int 1237jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num, 1238 int tablearg, int jump_backwards) 1239{ 1240 int f_pos; 1241 1242 /* If possible use cached f_pos (in f->cached_pos), 1243 * whose version is written in f->cached_id 1244 * (horrible hacks to avoid changing the ABI). 1245 */ 1246 if (num != IP_FW_TARG && f->cached_id == chain->id) 1247 f_pos = f->cached_pos; 1248 else { 1249 int i = IP_FW_ARG_TABLEARG(chain, num, skipto); 1250 /* make sure we do not jump backward */ 1251 if (jump_backwards == 0 && i <= f->rulenum) 1252 i = f->rulenum + 1; 1253 if (chain->idxmap != NULL) 1254 f_pos = chain->idxmap[i]; 1255 else 1256 f_pos = ipfw_find_rule(chain, i, 0); 1257 /* update the cache */ 1258 if (num != IP_FW_TARG) { 1259 f->cached_id = chain->id; 1260 f->cached_pos = f_pos; 1261 } 1262 } 1263 1264 return (f_pos); 1265} 1266#else 1267/* 1268 * Helper function to enable real fast rule lookups. 1269 */ 1270static int 1271jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num, 1272 int tablearg, int jump_backwards) 1273{ 1274 int f_pos; 1275 1276 num = IP_FW_ARG_TABLEARG(chain, num, skipto); 1277 /* make sure we do not jump backward */ 1278 if (jump_backwards == 0 && num <= f->rulenum) 1279 num = f->rulenum + 1; 1280 f_pos = chain->idxmap[num]; 1281 1282 return (f_pos); 1283} 1284#endif 1285 1286#define TARG(k, f) IP_FW_ARG_TABLEARG(chain, k, f) 1287/* 1288 * The main check routine for the firewall. 1289 * 1290 * All arguments are in args so we can modify them and return them 1291 * back to the caller. 1292 * 1293 * Parameters: 1294 * 1295 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1296 * Starts with the IP header. 1297 * args->eh (in) Mac header if present, NULL for layer3 packet. 1298 * args->L3offset Number of bytes bypassed if we came from L2. 1299 * e.g. often sizeof(eh) ** NOTYET ** 1300 * args->oif Outgoing interface, NULL if packet is incoming. 1301 * The incoming interface is in the mbuf. (in) 1302 * args->divert_rule (in/out) 1303 * Skip up to the first rule past this rule number; 1304 * upon return, non-zero port number for divert or tee. 1305 * 1306 * args->rule Pointer to the last matching rule (in/out) 1307 * args->next_hop Socket we are forwarding to (out). 1308 * args->next_hop6 IPv6 next hop we are forwarding to (out). 1309 * args->f_id Addresses grabbed from the packet (out) 1310 * args->rule.info a cookie depending on rule action 1311 * 1312 * Return value: 1313 * 1314 * IP_FW_PASS the packet must be accepted 1315 * IP_FW_DENY the packet must be dropped 1316 * IP_FW_DIVERT divert packet, port in m_tag 1317 * IP_FW_TEE tee packet, port in m_tag 1318 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 1319 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 1320 * args->rule contains the matching rule, 1321 * args->rule.info has additional information. 1322 * 1323 */ 1324int 1325ipfw_chk(struct ip_fw_args *args) 1326{ 1327 1328 /* 1329 * Local variables holding state while processing a packet: 1330 * 1331 * IMPORTANT NOTE: to speed up the processing of rules, there 1332 * are some assumption on the values of the variables, which 1333 * are documented here. Should you change them, please check 1334 * the implementation of the various instructions to make sure 1335 * that they still work. 1336 * 1337 * args->eh The MAC header. It is non-null for a layer2 1338 * packet, it is NULL for a layer-3 packet. 1339 * **notyet** 1340 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. 1341 * 1342 * m | args->m Pointer to the mbuf, as received from the caller. 1343 * It may change if ipfw_chk() does an m_pullup, or if it 1344 * consumes the packet because it calls send_reject(). 1345 * XXX This has to change, so that ipfw_chk() never modifies 1346 * or consumes the buffer. 1347 * ip is the beginning of the ip(4 or 6) header. 1348 * Calculated by adding the L3offset to the start of data. 1349 * (Until we start using L3offset, the packet is 1350 * supposed to start with the ip header). 1351 */ 1352 struct mbuf *m = args->m; 1353 struct ip *ip = mtod(m, struct ip *); 1354 1355 /* 1356 * For rules which contain uid/gid or jail constraints, cache 1357 * a copy of the users credentials after the pcb lookup has been 1358 * executed. This will speed up the processing of rules with 1359 * these types of constraints, as well as decrease contention 1360 * on pcb related locks. 1361 */ 1362#ifndef __FreeBSD__ 1363 struct bsd_ucred ucred_cache; 1364#else 1365 struct ucred *ucred_cache = NULL; 1366#endif 1367 int ucred_lookup = 0; 1368 1369 /* 1370 * oif | args->oif If NULL, ipfw_chk has been called on the 1371 * inbound path (ether_input, ip_input). 1372 * If non-NULL, ipfw_chk has been called on the outbound path 1373 * (ether_output, ip_output). 1374 */ 1375 struct ifnet *oif = args->oif; 1376 1377 int f_pos = 0; /* index of current rule in the array */ 1378 int retval = 0; 1379 1380 /* 1381 * hlen The length of the IP header. 1382 */ 1383 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 1384 1385 /* 1386 * offset The offset of a fragment. offset != 0 means that 1387 * we have a fragment at this offset of an IPv4 packet. 1388 * offset == 0 means that (if this is an IPv4 packet) 1389 * this is the first or only fragment. 1390 * For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header 1391 * or there is a single packet fragment (fragment header added 1392 * without needed). We will treat a single packet fragment as if 1393 * there was no fragment header (or log/block depending on the 1394 * V_fw_permit_single_frag6 sysctl setting). 1395 */ 1396 u_short offset = 0; 1397 u_short ip6f_mf = 0; 1398 1399 /* 1400 * Local copies of addresses. They are only valid if we have 1401 * an IP packet. 1402 * 1403 * proto The protocol. Set to 0 for non-ip packets, 1404 * or to the protocol read from the packet otherwise. 1405 * proto != 0 means that we have an IPv4 packet. 1406 * 1407 * src_port, dst_port port numbers, in HOST format. Only 1408 * valid for TCP and UDP packets. 1409 * 1410 * src_ip, dst_ip ip addresses, in NETWORK format. 1411 * Only valid for IPv4 packets. 1412 */ 1413 uint8_t proto; 1414 uint16_t src_port, dst_port; /* NOTE: host format */ 1415 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 1416 int iplen = 0; 1417 int pktlen; 1418 uint16_t etype; /* Host order stored ether type */ 1419 1420 struct ipfw_dyn_info dyn_info; 1421 struct ip_fw *q = NULL; 1422 struct ip_fw_chain *chain = &V_layer3_chain; 1423 1424 /* 1425 * We store in ulp a pointer to the upper layer protocol header. 1426 * In the ipv4 case this is easy to determine from the header, 1427 * but for ipv6 we might have some additional headers in the middle. 1428 * ulp is NULL if not found. 1429 */ 1430 void *ulp = NULL; /* upper layer protocol pointer. */ 1431 1432 /* XXX ipv6 variables */ 1433 int is_ipv6 = 0; 1434 uint8_t icmp6_type = 0; 1435 uint16_t ext_hd = 0; /* bits vector for extension header filtering */ 1436 /* end of ipv6 variables */ 1437 1438 int is_ipv4 = 0; 1439 1440 int done = 0; /* flag to exit the outer loop */ 1441 1442 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready)) 1443 return (IP_FW_PASS); /* accept */ 1444 1445 dst_ip.s_addr = 0; /* make sure it is initialized */ 1446 src_ip.s_addr = 0; /* make sure it is initialized */ 1447 src_port = dst_port = 0; 1448 pktlen = m->m_pkthdr.len; 1449 1450 DYN_INFO_INIT(&dyn_info); 1451/* 1452 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 1453 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 1454 * pointer might become stale after other pullups (but we never use it 1455 * this way). 1456 */ 1457#define PULLUP_TO(_len, p, T) PULLUP_LEN(_len, p, sizeof(T)) 1458#define _PULLUP_LOCKED(_len, p, T, unlock) \ 1459do { \ 1460 int x = (_len) + T; \ 1461 if ((m)->m_len < x) { \ 1462 args->m = m = m_pullup(m, x); \ 1463 if (m == NULL) { \ 1464 unlock; \ 1465 goto pullup_failed; \ 1466 } \ 1467 } \ 1468 p = (mtod(m, char *) + (_len)); \ 1469} while (0) 1470 1471#define PULLUP_LEN(_len, p, T) _PULLUP_LOCKED(_len, p, T, ) 1472#define PULLUP_LEN_LOCKED(_len, p, T) \ 1473 _PULLUP_LOCKED(_len, p, T, IPFW_PF_RUNLOCK(chain)); \ 1474 UPDATE_POINTERS() 1475/* 1476 * In case pointers got stale after pullups, update them. 1477 */ 1478#define UPDATE_POINTERS() \ 1479do { \ 1480 ip = mtod(m, struct ip *); \ 1481} while (0) 1482 1483 /* 1484 * if we have an ether header, 1485 */ 1486 if (args->flags & IPFW_ARGS_ETHER) 1487 etype = ntohs(args->eh->ether_type); 1488 else 1489 etype = 0; 1490 1491 /* Identify IP packets and fill up variables. */ 1492 if (pktlen >= sizeof(struct ip6_hdr) && 1493 (etype == 0 || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { 1494 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 1495 1496 is_ipv6 = 1; 1497 hlen = sizeof(struct ip6_hdr); 1498 proto = ip6->ip6_nxt; 1499 /* Search extension headers to find upper layer protocols */ 1500 while (ulp == NULL && offset == 0) { 1501 switch (proto) { 1502 case IPPROTO_ICMPV6: 1503 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 1504 icmp6_type = ICMP6(ulp)->icmp6_type; 1505 break; 1506 1507 case IPPROTO_TCP: 1508 PULLUP_TO(hlen, ulp, struct tcphdr); 1509 dst_port = TCP(ulp)->th_dport; 1510 src_port = TCP(ulp)->th_sport; 1511 /* save flags for dynamic rules */ 1512 args->f_id._flags = TCP(ulp)->th_flags; 1513 break; 1514 1515 case IPPROTO_SCTP: 1516 if (pktlen >= hlen + sizeof(struct sctphdr) + 1517 sizeof(struct sctp_chunkhdr) + 1518 offsetof(struct sctp_init, a_rwnd)) 1519 PULLUP_LEN(hlen, ulp, 1520 sizeof(struct sctphdr) + 1521 sizeof(struct sctp_chunkhdr) + 1522 offsetof(struct sctp_init, a_rwnd)); 1523 else if (pktlen >= hlen + sizeof(struct sctphdr)) 1524 PULLUP_LEN(hlen, ulp, pktlen - hlen); 1525 else 1526 PULLUP_LEN(hlen, ulp, 1527 sizeof(struct sctphdr)); 1528 src_port = SCTP(ulp)->src_port; 1529 dst_port = SCTP(ulp)->dest_port; 1530 break; 1531 1532 case IPPROTO_UDP: 1533 case IPPROTO_UDPLITE: 1534 PULLUP_TO(hlen, ulp, struct udphdr); 1535 dst_port = UDP(ulp)->uh_dport; 1536 src_port = UDP(ulp)->uh_sport; 1537 break; 1538 1539 case IPPROTO_HOPOPTS: /* RFC 2460 */ 1540 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1541 ext_hd |= EXT_HOPOPTS; 1542 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1543 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1544 ulp = NULL; 1545 break; 1546 1547 case IPPROTO_ROUTING: /* RFC 2460 */ 1548 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 1549 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 1550 case 0: 1551 ext_hd |= EXT_RTHDR0; 1552 break; 1553 case 2: 1554 ext_hd |= EXT_RTHDR2; 1555 break; 1556 default: 1557 if (V_fw_verbose) 1558 printf("IPFW2: IPV6 - Unknown " 1559 "Routing Header type(%d)\n", 1560 ((struct ip6_rthdr *) 1561 ulp)->ip6r_type); 1562 if (V_fw_deny_unknown_exthdrs) 1563 return (IP_FW_DENY); 1564 break; 1565 } 1566 ext_hd |= EXT_ROUTING; 1567 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 1568 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 1569 ulp = NULL; 1570 break; 1571 1572 case IPPROTO_FRAGMENT: /* RFC 2460 */ 1573 PULLUP_TO(hlen, ulp, struct ip6_frag); 1574 ext_hd |= EXT_FRAGMENT; 1575 hlen += sizeof (struct ip6_frag); 1576 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 1577 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 1578 IP6F_OFF_MASK; 1579 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg & 1580 IP6F_MORE_FRAG; 1581 if (V_fw_permit_single_frag6 == 0 && 1582 offset == 0 && ip6f_mf == 0) { 1583 if (V_fw_verbose) 1584 printf("IPFW2: IPV6 - Invalid " 1585 "Fragment Header\n"); 1586 if (V_fw_deny_unknown_exthdrs) 1587 return (IP_FW_DENY); 1588 break; 1589 } 1590 args->f_id.extra = 1591 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 1592 ulp = NULL; 1593 break; 1594 1595 case IPPROTO_DSTOPTS: /* RFC 2460 */ 1596 PULLUP_TO(hlen, ulp, struct ip6_hbh); 1597 ext_hd |= EXT_DSTOPTS; 1598 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 1599 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 1600 ulp = NULL; 1601 break; 1602 1603 case IPPROTO_AH: /* RFC 2402 */ 1604 PULLUP_TO(hlen, ulp, struct ip6_ext); 1605 ext_hd |= EXT_AH; 1606 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 1607 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 1608 ulp = NULL; 1609 break; 1610 1611 case IPPROTO_ESP: /* RFC 2406 */ 1612 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 1613 /* Anything past Seq# is variable length and 1614 * data past this ext. header is encrypted. */ 1615 ext_hd |= EXT_ESP; 1616 break; 1617 1618 case IPPROTO_NONE: /* RFC 2460 */ 1619 /* 1620 * Packet ends here, and IPv6 header has 1621 * already been pulled up. If ip6e_len!=0 1622 * then octets must be ignored. 1623 */ 1624 ulp = ip; /* non-NULL to get out of loop. */ 1625 break; 1626 1627 case IPPROTO_OSPFIGP: 1628 /* XXX OSPF header check? */ 1629 PULLUP_TO(hlen, ulp, struct ip6_ext); 1630 break; 1631 1632 case IPPROTO_PIM: 1633 /* XXX PIM header check? */ 1634 PULLUP_TO(hlen, ulp, struct pim); 1635 break; 1636 1637 case IPPROTO_GRE: /* RFC 1701 */ 1638 /* XXX GRE header check? */ 1639 PULLUP_TO(hlen, ulp, struct grehdr); 1640 break; 1641 1642 case IPPROTO_CARP: 1643 PULLUP_TO(hlen, ulp, offsetof( 1644 struct carp_header, carp_counter)); 1645 if (CARP_ADVERTISEMENT != 1646 ((struct carp_header *)ulp)->carp_type) 1647 return (IP_FW_DENY); 1648 break; 1649 1650 case IPPROTO_IPV6: /* RFC 2893 */ 1651 PULLUP_TO(hlen, ulp, struct ip6_hdr); 1652 break; 1653 1654 case IPPROTO_IPV4: /* RFC 2893 */ 1655 PULLUP_TO(hlen, ulp, struct ip); 1656 break; 1657 1658 default: 1659 if (V_fw_verbose) 1660 printf("IPFW2: IPV6 - Unknown " 1661 "Extension Header(%d), ext_hd=%x\n", 1662 proto, ext_hd); 1663 if (V_fw_deny_unknown_exthdrs) 1664 return (IP_FW_DENY); 1665 PULLUP_TO(hlen, ulp, struct ip6_ext); 1666 break; 1667 } /*switch */ 1668 } 1669 ip = mtod(m, struct ip *); 1670 ip6 = (struct ip6_hdr *)ip; 1671 args->f_id.addr_type = 6; 1672 args->f_id.src_ip6 = ip6->ip6_src; 1673 args->f_id.dst_ip6 = ip6->ip6_dst; 1674 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 1675 iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6); 1676 } else if (pktlen >= sizeof(struct ip) && 1677 (etype == 0 || etype == ETHERTYPE_IP) && ip->ip_v == 4) { 1678 is_ipv4 = 1; 1679 hlen = ip->ip_hl << 2; 1680 /* 1681 * Collect parameters into local variables for faster 1682 * matching. 1683 */ 1684 proto = ip->ip_p; 1685 src_ip = ip->ip_src; 1686 dst_ip = ip->ip_dst; 1687 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1688 iplen = ntohs(ip->ip_len); 1689 1690 if (offset == 0) { 1691 switch (proto) { 1692 case IPPROTO_TCP: 1693 PULLUP_TO(hlen, ulp, struct tcphdr); 1694 dst_port = TCP(ulp)->th_dport; 1695 src_port = TCP(ulp)->th_sport; 1696 /* save flags for dynamic rules */ 1697 args->f_id._flags = TCP(ulp)->th_flags; 1698 break; 1699 1700 case IPPROTO_SCTP: 1701 if (pktlen >= hlen + sizeof(struct sctphdr) + 1702 sizeof(struct sctp_chunkhdr) + 1703 offsetof(struct sctp_init, a_rwnd)) 1704 PULLUP_LEN(hlen, ulp, 1705 sizeof(struct sctphdr) + 1706 sizeof(struct sctp_chunkhdr) + 1707 offsetof(struct sctp_init, a_rwnd)); 1708 else if (pktlen >= hlen + sizeof(struct sctphdr)) 1709 PULLUP_LEN(hlen, ulp, pktlen - hlen); 1710 else 1711 PULLUP_LEN(hlen, ulp, 1712 sizeof(struct sctphdr)); 1713 src_port = SCTP(ulp)->src_port; 1714 dst_port = SCTP(ulp)->dest_port; 1715 break; 1716 1717 case IPPROTO_UDP: 1718 case IPPROTO_UDPLITE: 1719 PULLUP_TO(hlen, ulp, struct udphdr); 1720 dst_port = UDP(ulp)->uh_dport; 1721 src_port = UDP(ulp)->uh_sport; 1722 break; 1723 1724 case IPPROTO_ICMP: 1725 PULLUP_TO(hlen, ulp, struct icmphdr); 1726 //args->f_id.flags = ICMP(ulp)->icmp_type; 1727 break; 1728 1729 default: 1730 break; 1731 } 1732 } else { 1733 if (offset == 1 && proto == IPPROTO_TCP) { 1734 /* RFC 3128 */ 1735 goto pullup_failed; 1736 } 1737 } 1738 1739 ip = mtod(m, struct ip *); 1740 args->f_id.addr_type = 4; 1741 args->f_id.src_ip = ntohl(src_ip.s_addr); 1742 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1743 } else { 1744 proto = 0; 1745 dst_ip.s_addr = src_ip.s_addr = 0; 1746 1747 args->f_id.addr_type = 1; /* XXX */ 1748 } 1749#undef PULLUP_TO 1750 pktlen = iplen < pktlen ? iplen: pktlen; 1751 1752 /* Properly initialize the rest of f_id */ 1753 args->f_id.proto = proto; 1754 args->f_id.src_port = src_port = ntohs(src_port); 1755 args->f_id.dst_port = dst_port = ntohs(dst_port); 1756 args->f_id.fib = M_GETFIB(m); 1757 1758 IPFW_PF_RLOCK(chain); 1759 if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */ 1760 IPFW_PF_RUNLOCK(chain); 1761 return (IP_FW_PASS); /* accept */ 1762 } 1763 if (args->flags & IPFW_ARGS_REF) { 1764 /* 1765 * Packet has already been tagged as a result of a previous 1766 * match on rule args->rule aka args->rule_id (PIPE, QUEUE, 1767 * REASS, NETGRAPH, DIVERT/TEE...) 1768 * Validate the slot and continue from the next one 1769 * if still present, otherwise do a lookup. 1770 */ 1771 f_pos = (args->rule.chain_id == chain->id) ? 1772 args->rule.slot : 1773 ipfw_find_rule(chain, args->rule.rulenum, 1774 args->rule.rule_id); 1775 } else { 1776 f_pos = 0; 1777 } 1778 1779 /* 1780 * Now scan the rules, and parse microinstructions for each rule. 1781 * We have two nested loops and an inner switch. Sometimes we 1782 * need to break out of one or both loops, or re-enter one of 1783 * the loops with updated variables. Loop variables are: 1784 * 1785 * f_pos (outer loop) points to the current rule. 1786 * On output it points to the matching rule. 1787 * done (outer loop) is used as a flag to break the loop. 1788 * l (inner loop) residual length of current rule. 1789 * cmd points to the current microinstruction. 1790 * 1791 * We break the inner loop by setting l=0 and possibly 1792 * cmdlen=0 if we don't want to advance cmd. 1793 * We break the outer loop by setting done=1 1794 * We can restart the inner loop by setting l>0 and f_pos, f, cmd 1795 * as needed. 1796 */ 1797 for (; f_pos < chain->n_rules; f_pos++) { 1798 ipfw_insn *cmd; 1799 uint32_t tablearg = 0; 1800 int l, cmdlen, skip_or; /* skip rest of OR block */ 1801 struct ip_fw *f; 1802 1803 f = chain->map[f_pos]; 1804 if (V_set_disable & (1 << f->set) ) 1805 continue; 1806 1807 skip_or = 0; 1808 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1809 l -= cmdlen, cmd += cmdlen) { 1810 int match; 1811 1812 /* 1813 * check_body is a jump target used when we find a 1814 * CHECK_STATE, and need to jump to the body of 1815 * the target rule. 1816 */ 1817 1818/* check_body: */ 1819 cmdlen = F_LEN(cmd); 1820 /* 1821 * An OR block (insn_1 || .. || insn_n) has the 1822 * F_OR bit set in all but the last instruction. 1823 * The first match will set "skip_or", and cause 1824 * the following instructions to be skipped until 1825 * past the one with the F_OR bit clear. 1826 */ 1827 if (skip_or) { /* skip this instruction */ 1828 if ((cmd->len & F_OR) == 0) 1829 skip_or = 0; /* next one is good */ 1830 continue; 1831 } 1832 match = 0; /* set to 1 if we succeed */ 1833 1834 switch (cmd->opcode) { 1835 /* 1836 * The first set of opcodes compares the packet's 1837 * fields with some pattern, setting 'match' if a 1838 * match is found. At the end of the loop there is 1839 * logic to deal with F_NOT and F_OR flags associated 1840 * with the opcode. 1841 */ 1842 case O_NOP: 1843 match = 1; 1844 break; 1845 1846 case O_FORWARD_MAC: 1847 printf("ipfw: opcode %d unimplemented\n", 1848 cmd->opcode); 1849 break; 1850 1851 case O_GID: 1852 case O_UID: 1853 case O_JAIL: 1854 /* 1855 * We only check offset == 0 && proto != 0, 1856 * as this ensures that we have a 1857 * packet with the ports info. 1858 */ 1859 if (offset != 0) 1860 break; 1861 if (proto == IPPROTO_TCP || 1862 proto == IPPROTO_UDP || 1863 proto == IPPROTO_UDPLITE) 1864 match = check_uidgid( 1865 (ipfw_insn_u32 *)cmd, 1866 args, &ucred_lookup, 1867#ifdef __FreeBSD__ 1868 &ucred_cache); 1869#else 1870 (void *)&ucred_cache); 1871#endif 1872 break; 1873 1874 case O_RECV: 1875 match = iface_match(m->m_pkthdr.rcvif, 1876 (ipfw_insn_if *)cmd, chain, &tablearg); 1877 break; 1878 1879 case O_XMIT: 1880 match = iface_match(oif, (ipfw_insn_if *)cmd, 1881 chain, &tablearg); 1882 break; 1883 1884 case O_VIA: 1885 match = iface_match(oif ? oif : 1886 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd, 1887 chain, &tablearg); 1888 break; 1889 1890 case O_MACADDR2: 1891 if (args->flags & IPFW_ARGS_ETHER) { 1892 u_int32_t *want = (u_int32_t *) 1893 ((ipfw_insn_mac *)cmd)->addr; 1894 u_int32_t *mask = (u_int32_t *) 1895 ((ipfw_insn_mac *)cmd)->mask; 1896 u_int32_t *hdr = (u_int32_t *)args->eh; 1897 1898 match = 1899 ( want[0] == (hdr[0] & mask[0]) && 1900 want[1] == (hdr[1] & mask[1]) && 1901 want[2] == (hdr[2] & mask[2]) ); 1902 } 1903 break; 1904 1905 case O_MAC_TYPE: 1906 if (args->flags & IPFW_ARGS_ETHER) { 1907 u_int16_t *p = 1908 ((ipfw_insn_u16 *)cmd)->ports; 1909 int i; 1910 1911 for (i = cmdlen - 1; !match && i>0; 1912 i--, p += 2) 1913 match = (etype >= p[0] && 1914 etype <= p[1]); 1915 } 1916 break; 1917 1918 case O_FRAG: 1919 if (is_ipv4) { 1920 /* 1921 * Since flags_match() works with 1922 * uint8_t we pack ip_off into 8 bits. 1923 * For this match offset is a boolean. 1924 */ 1925 match = flags_match(cmd, 1926 ((ntohs(ip->ip_off) & ~IP_OFFMASK) 1927 >> 8) | (offset != 0)); 1928 } else { 1929 /* 1930 * Compatiblity: historically bare 1931 * "frag" would match IPv6 fragments. 1932 */ 1933 match = (cmd->arg1 == 0x1 && 1934 (offset != 0)); 1935 } 1936 break; 1937 1938 case O_IN: /* "out" is "not in" */ 1939 match = (oif == NULL); 1940 break; 1941 1942 case O_LAYER2: 1943 match = (args->flags & IPFW_ARGS_ETHER); 1944 break; 1945 1946 case O_DIVERTED: 1947 if ((args->flags & IPFW_ARGS_REF) == 0) 1948 break; 1949 /* 1950 * For diverted packets, args->rule.info 1951 * contains the divert port (in host format) 1952 * reason and direction. 1953 */ 1954 match = ((args->rule.info & IPFW_IS_MASK) == 1955 IPFW_IS_DIVERT) && ( 1956 ((args->rule.info & IPFW_INFO_IN) ? 1957 1: 2) & cmd->arg1); 1958 break; 1959 1960 case O_PROTO: 1961 /* 1962 * We do not allow an arg of 0 so the 1963 * check of "proto" only suffices. 1964 */ 1965 match = (proto == cmd->arg1); 1966 break; 1967 1968 case O_IP_SRC: 1969 match = is_ipv4 && 1970 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1971 src_ip.s_addr); 1972 break; 1973 1974 case O_IP_DST_LOOKUP: 1975 { 1976 void *pkey; 1977 uint32_t vidx, key; 1978 uint16_t keylen; 1979 1980 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) { 1981 /* Determine lookup key type */ 1982 vidx = ((ipfw_insn_u32 *)cmd)->d[1]; 1983 if (vidx != 4 /* uid */ && 1984 vidx != 5 /* jail */ && 1985 is_ipv6 == 0 && is_ipv4 == 0) 1986 break; 1987 /* Determine key length */ 1988 if (vidx == 0 /* dst-ip */ || 1989 vidx == 1 /* src-ip */) 1990 keylen = is_ipv6 ? 1991 sizeof(struct in6_addr): 1992 sizeof(in_addr_t); 1993 else { 1994 keylen = sizeof(key); 1995 pkey = &key; 1996 } 1997 if (vidx == 0 /* dst-ip */) 1998 pkey = is_ipv4 ? (void *)&dst_ip: 1999 (void *)&args->f_id.dst_ip6; 2000 else if (vidx == 1 /* src-ip */) 2001 pkey = is_ipv4 ? (void *)&src_ip: 2002 (void *)&args->f_id.src_ip6; 2003 else if (vidx == 6 /* dscp */) { 2004 if (is_ipv4) 2005 key = ip->ip_tos >> 2; 2006 else { 2007 key = args->f_id.flow_id6; 2008 key = (key & 0x0f) << 2 | 2009 (key & 0xf000) >> 14; 2010 } 2011 key &= 0x3f; 2012 } else if (vidx == 2 /* dst-port */ || 2013 vidx == 3 /* src-port */) { 2014 /* Skip fragments */ 2015 if (offset != 0) 2016 break; 2017 /* Skip proto without ports */ 2018 if (proto != IPPROTO_TCP && 2019 proto != IPPROTO_UDP && 2020 proto != IPPROTO_UDPLITE && 2021 proto != IPPROTO_SCTP) 2022 break; 2023 if (vidx == 2 /* dst-port */) 2024 key = dst_port; 2025 else 2026 key = src_port; 2027 } 2028#ifndef USERSPACE 2029 else if (vidx == 4 /* uid */ || 2030 vidx == 5 /* jail */) { 2031 check_uidgid( 2032 (ipfw_insn_u32 *)cmd, 2033 args, &ucred_lookup, 2034#ifdef __FreeBSD__ 2035 &ucred_cache); 2036 if (vidx == 4 /* uid */) 2037 key = ucred_cache->cr_uid; 2038 else if (vidx == 5 /* jail */) 2039 key = ucred_cache->cr_prison->pr_id; 2040#else /* !__FreeBSD__ */ 2041 (void *)&ucred_cache); 2042 if (vidx == 4 /* uid */) 2043 key = ucred_cache.uid; 2044 else if (vidx == 5 /* jail */) 2045 key = ucred_cache.xid; 2046#endif /* !__FreeBSD__ */ 2047 } 2048#endif /* !USERSPACE */ 2049 else 2050 break; 2051 match = ipfw_lookup_table(chain, 2052 cmd->arg1, keylen, pkey, &vidx); 2053 if (!match) 2054 break; 2055 tablearg = vidx; 2056 break; 2057 } 2058 /* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */ 2059 /* FALLTHROUGH */ 2060 } 2061 case O_IP_SRC_LOOKUP: 2062 { 2063 void *pkey; 2064 uint32_t vidx; 2065 uint16_t keylen; 2066 2067 if (is_ipv4) { 2068 keylen = sizeof(in_addr_t); 2069 if (cmd->opcode == O_IP_DST_LOOKUP) 2070 pkey = &dst_ip; 2071 else 2072 pkey = &src_ip; 2073 } else if (is_ipv6) { 2074 keylen = sizeof(struct in6_addr); 2075 if (cmd->opcode == O_IP_DST_LOOKUP) 2076 pkey = &args->f_id.dst_ip6; 2077 else 2078 pkey = &args->f_id.src_ip6; 2079 } else 2080 break; 2081 match = ipfw_lookup_table(chain, cmd->arg1, 2082 keylen, pkey, &vidx); 2083 if (!match) 2084 break; 2085 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) { 2086 match = ((ipfw_insn_u32 *)cmd)->d[0] == 2087 TARG_VAL(chain, vidx, tag); 2088 if (!match) 2089 break; 2090 } 2091 tablearg = vidx; 2092 break; 2093 } 2094 2095 case O_IP_FLOW_LOOKUP: 2096 { 2097 uint32_t v = 0; 2098 match = ipfw_lookup_table(chain, 2099 cmd->arg1, 0, &args->f_id, &v); 2100 if (!match) 2101 break; 2102 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2103 match = ((ipfw_insn_u32 *)cmd)->d[0] == 2104 TARG_VAL(chain, v, tag); 2105 if (match) 2106 tablearg = v; 2107 } 2108 break; 2109 case O_IP_SRC_MASK: 2110 case O_IP_DST_MASK: 2111 if (is_ipv4) { 2112 uint32_t a = 2113 (cmd->opcode == O_IP_DST_MASK) ? 2114 dst_ip.s_addr : src_ip.s_addr; 2115 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2116 int i = cmdlen-1; 2117 2118 for (; !match && i>0; i-= 2, p+= 2) 2119 match = (p[0] == (a & p[1])); 2120 } 2121 break; 2122 2123 case O_IP_SRC_ME: 2124 if (is_ipv4) { 2125 match = in_localip(src_ip); 2126 break; 2127 } 2128#ifdef INET6 2129 /* FALLTHROUGH */ 2130 case O_IP6_SRC_ME: 2131 match = is_ipv6 && 2132 ipfw_localip6(&args->f_id.src_ip6); 2133#endif 2134 break; 2135 2136 case O_IP_DST_SET: 2137 case O_IP_SRC_SET: 2138 if (is_ipv4) { 2139 u_int32_t *d = (u_int32_t *)(cmd+1); 2140 u_int32_t addr = 2141 cmd->opcode == O_IP_DST_SET ? 2142 args->f_id.dst_ip : 2143 args->f_id.src_ip; 2144 2145 if (addr < d[0]) 2146 break; 2147 addr -= d[0]; /* subtract base */ 2148 match = (addr < cmd->arg1) && 2149 ( d[ 1 + (addr>>5)] & 2150 (1<<(addr & 0x1f)) ); 2151 } 2152 break; 2153 2154 case O_IP_DST: 2155 match = is_ipv4 && 2156 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2157 dst_ip.s_addr); 2158 break; 2159 2160 case O_IP_DST_ME: 2161 if (is_ipv4) { 2162 match = in_localip(dst_ip); 2163 break; 2164 } 2165#ifdef INET6 2166 /* FALLTHROUGH */ 2167 case O_IP6_DST_ME: 2168 match = is_ipv6 && 2169 ipfw_localip6(&args->f_id.dst_ip6); 2170#endif 2171 break; 2172 2173 2174 case O_IP_SRCPORT: 2175 case O_IP_DSTPORT: 2176 /* 2177 * offset == 0 && proto != 0 is enough 2178 * to guarantee that we have a 2179 * packet with port info. 2180 */ 2181 if ((proto == IPPROTO_UDP || 2182 proto == IPPROTO_UDPLITE || 2183 proto == IPPROTO_TCP || 2184 proto == IPPROTO_SCTP) && offset == 0) { 2185 u_int16_t x = 2186 (cmd->opcode == O_IP_SRCPORT) ? 2187 src_port : dst_port ; 2188 u_int16_t *p = 2189 ((ipfw_insn_u16 *)cmd)->ports; 2190 int i; 2191 2192 for (i = cmdlen - 1; !match && i>0; 2193 i--, p += 2) 2194 match = (x>=p[0] && x<=p[1]); 2195 } 2196 break; 2197 2198 case O_ICMPTYPE: 2199 match = (offset == 0 && proto==IPPROTO_ICMP && 2200 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2201 break; 2202 2203#ifdef INET6 2204 case O_ICMP6TYPE: 2205 match = is_ipv6 && offset == 0 && 2206 proto==IPPROTO_ICMPV6 && 2207 icmp6type_match( 2208 ICMP6(ulp)->icmp6_type, 2209 (ipfw_insn_u32 *)cmd); 2210 break; 2211#endif /* INET6 */ 2212 2213 case O_IPOPT: 2214 match = (is_ipv4 && 2215 ipopts_match(ip, cmd) ); 2216 break; 2217 2218 case O_IPVER: 2219 match = ((is_ipv4 || is_ipv6) && 2220 cmd->arg1 == ip->ip_v); 2221 break; 2222 2223 case O_IPID: 2224 case O_IPTTL: 2225 if (!is_ipv4) 2226 break; 2227 case O_IPLEN: 2228 { /* only for IP packets */ 2229 uint16_t x; 2230 uint16_t *p; 2231 int i; 2232 2233 if (cmd->opcode == O_IPLEN) 2234 x = iplen; 2235 else if (cmd->opcode == O_IPTTL) 2236 x = ip->ip_ttl; 2237 else /* must be IPID */ 2238 x = ntohs(ip->ip_id); 2239 if (cmdlen == 1) { 2240 match = (cmd->arg1 == x); 2241 break; 2242 } 2243 /* otherwise we have ranges */ 2244 p = ((ipfw_insn_u16 *)cmd)->ports; 2245 i = cmdlen - 1; 2246 for (; !match && i>0; i--, p += 2) 2247 match = (x >= p[0] && x <= p[1]); 2248 } 2249 break; 2250 2251 case O_IPPRECEDENCE: 2252 match = (is_ipv4 && 2253 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2254 break; 2255 2256 case O_IPTOS: 2257 match = (is_ipv4 && 2258 flags_match(cmd, ip->ip_tos)); 2259 break; 2260 2261 case O_DSCP: 2262 { 2263 uint32_t *p; 2264 uint16_t x; 2265 2266 p = ((ipfw_insn_u32 *)cmd)->d; 2267 2268 if (is_ipv4) 2269 x = ip->ip_tos >> 2; 2270 else if (is_ipv6) { 2271 uint8_t *v; 2272 v = &((struct ip6_hdr *)ip)->ip6_vfc; 2273 x = (*v & 0x0F) << 2; 2274 v++; 2275 x |= *v >> 6; 2276 } else 2277 break; 2278 2279 /* DSCP bitmask is stored as low_u32 high_u32 */ 2280 if (x >= 32) 2281 match = *(p + 1) & (1 << (x - 32)); 2282 else 2283 match = *p & (1 << x); 2284 } 2285 break; 2286 2287 case O_TCPDATALEN: 2288 if (proto == IPPROTO_TCP && offset == 0) { 2289 struct tcphdr *tcp; 2290 uint16_t x; 2291 uint16_t *p; 2292 int i; 2293#ifdef INET6 2294 if (is_ipv6) { 2295 struct ip6_hdr *ip6; 2296 2297 ip6 = (struct ip6_hdr *)ip; 2298 if (ip6->ip6_plen == 0) { 2299 /* 2300 * Jumbo payload is not 2301 * supported by this 2302 * opcode. 2303 */ 2304 break; 2305 } 2306 x = iplen - hlen; 2307 } else 2308#endif /* INET6 */ 2309 x = iplen - (ip->ip_hl << 2); 2310 tcp = TCP(ulp); 2311 x -= tcp->th_off << 2; 2312 if (cmdlen == 1) { 2313 match = (cmd->arg1 == x); 2314 break; 2315 } 2316 /* otherwise we have ranges */ 2317 p = ((ipfw_insn_u16 *)cmd)->ports; 2318 i = cmdlen - 1; 2319 for (; !match && i>0; i--, p += 2) 2320 match = (x >= p[0] && x <= p[1]); 2321 } 2322 break; 2323 2324 case O_TCPFLAGS: 2325 match = (proto == IPPROTO_TCP && offset == 0 && 2326 flags_match(cmd, TCP(ulp)->th_flags)); 2327 break; 2328 2329 case O_TCPOPTS: 2330 if (proto == IPPROTO_TCP && offset == 0 && ulp){ 2331 PULLUP_LEN_LOCKED(hlen, ulp, 2332 (TCP(ulp)->th_off << 2)); 2333 match = tcpopts_match(TCP(ulp), cmd); 2334 } 2335 break; 2336 2337 case O_TCPSEQ: 2338 match = (proto == IPPROTO_TCP && offset == 0 && 2339 ((ipfw_insn_u32 *)cmd)->d[0] == 2340 TCP(ulp)->th_seq); 2341 break; 2342 2343 case O_TCPACK: 2344 match = (proto == IPPROTO_TCP && offset == 0 && 2345 ((ipfw_insn_u32 *)cmd)->d[0] == 2346 TCP(ulp)->th_ack); 2347 break; 2348 2349 case O_TCPMSS: 2350 if (proto == IPPROTO_TCP && 2351 (args->f_id._flags & TH_SYN) != 0 && 2352 ulp != NULL) { 2353 uint16_t mss, *p; 2354 int i; 2355 2356 PULLUP_LEN_LOCKED(hlen, ulp, 2357 (TCP(ulp)->th_off << 2)); 2358 if ((tcpopts_parse(TCP(ulp), &mss) & 2359 IP_FW_TCPOPT_MSS) == 0) 2360 break; 2361 if (cmdlen == 1) { 2362 match = (cmd->arg1 == mss); 2363 break; 2364 } 2365 /* Otherwise we have ranges. */ 2366 p = ((ipfw_insn_u16 *)cmd)->ports; 2367 i = cmdlen - 1; 2368 for (; !match && i > 0; i--, p += 2) 2369 match = (mss >= p[0] && 2370 mss <= p[1]); 2371 } 2372 break; 2373 2374 case O_TCPWIN: 2375 if (proto == IPPROTO_TCP && offset == 0) { 2376 uint16_t x; 2377 uint16_t *p; 2378 int i; 2379 2380 x = ntohs(TCP(ulp)->th_win); 2381 if (cmdlen == 1) { 2382 match = (cmd->arg1 == x); 2383 break; 2384 } 2385 /* Otherwise we have ranges. */ 2386 p = ((ipfw_insn_u16 *)cmd)->ports; 2387 i = cmdlen - 1; 2388 for (; !match && i > 0; i--, p += 2) 2389 match = (x >= p[0] && x <= p[1]); 2390 } 2391 break; 2392 2393 case O_ESTAB: 2394 /* reject packets which have SYN only */ 2395 /* XXX should i also check for TH_ACK ? */ 2396 match = (proto == IPPROTO_TCP && offset == 0 && 2397 (TCP(ulp)->th_flags & 2398 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2399 break; 2400 2401 case O_ALTQ: { 2402 struct pf_mtag *at; 2403 struct m_tag *mtag; 2404 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2405 2406 /* 2407 * ALTQ uses mbuf tags from another 2408 * packet filtering system - pf(4). 2409 * We allocate a tag in its format 2410 * and fill it in, pretending to be pf(4). 2411 */ 2412 match = 1; 2413 at = pf_find_mtag(m); 2414 if (at != NULL && at->qid != 0) 2415 break; 2416 mtag = m_tag_get(PACKET_TAG_PF, 2417 sizeof(struct pf_mtag), M_NOWAIT | M_ZERO); 2418 if (mtag == NULL) { 2419 /* 2420 * Let the packet fall back to the 2421 * default ALTQ. 2422 */ 2423 break; 2424 } 2425 m_tag_prepend(m, mtag); 2426 at = (struct pf_mtag *)(mtag + 1); 2427 at->qid = altq->qid; 2428 at->hdr = ip; 2429 break; 2430 } 2431 2432 case O_LOG: 2433 ipfw_log(chain, f, hlen, args, m, 2434 oif, offset | ip6f_mf, tablearg, ip); 2435 match = 1; 2436 break; 2437 2438 case O_PROB: 2439 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2440 break; 2441 2442 case O_VERREVPATH: 2443 /* Outgoing packets automatically pass/match */ 2444 match = ((oif != NULL) || 2445 (m->m_pkthdr.rcvif == NULL) || 2446 ( 2447#ifdef INET6 2448 is_ipv6 ? 2449 verify_path6(&(args->f_id.src_ip6), 2450 m->m_pkthdr.rcvif, args->f_id.fib) : 2451#endif 2452 verify_path(src_ip, m->m_pkthdr.rcvif, 2453 args->f_id.fib))); 2454 break; 2455 2456 case O_VERSRCREACH: 2457 /* Outgoing packets automatically pass/match */ 2458 match = (hlen > 0 && ((oif != NULL) || ( 2459#ifdef INET6 2460 is_ipv6 ? 2461 verify_path6(&(args->f_id.src_ip6), 2462 NULL, args->f_id.fib) : 2463#endif 2464 verify_path(src_ip, NULL, args->f_id.fib)))); 2465 break; 2466 2467 case O_ANTISPOOF: 2468 /* Outgoing packets automatically pass/match */ 2469 if (oif == NULL && hlen > 0 && 2470 ( (is_ipv4 && in_localaddr(src_ip)) 2471#ifdef INET6 2472 || (is_ipv6 && 2473 in6_localaddr(&(args->f_id.src_ip6))) 2474#endif 2475 )) 2476 match = 2477#ifdef INET6 2478 is_ipv6 ? verify_path6( 2479 &(args->f_id.src_ip6), 2480 m->m_pkthdr.rcvif, 2481 args->f_id.fib) : 2482#endif 2483 verify_path(src_ip, 2484 m->m_pkthdr.rcvif, 2485 args->f_id.fib); 2486 else 2487 match = 1; 2488 break; 2489 2490 case O_IPSEC: 2491 match = (m_tag_find(m, 2492 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2493 /* otherwise no match */ 2494 break; 2495 2496#ifdef INET6 2497 case O_IP6_SRC: 2498 match = is_ipv6 && 2499 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 2500 &((ipfw_insn_ip6 *)cmd)->addr6); 2501 break; 2502 2503 case O_IP6_DST: 2504 match = is_ipv6 && 2505 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 2506 &((ipfw_insn_ip6 *)cmd)->addr6); 2507 break; 2508 case O_IP6_SRC_MASK: 2509 case O_IP6_DST_MASK: 2510 if (is_ipv6) { 2511 int i = cmdlen - 1; 2512 struct in6_addr p; 2513 struct in6_addr *d = 2514 &((ipfw_insn_ip6 *)cmd)->addr6; 2515 2516 for (; !match && i > 0; d += 2, 2517 i -= F_INSN_SIZE(struct in6_addr) 2518 * 2) { 2519 p = (cmd->opcode == 2520 O_IP6_SRC_MASK) ? 2521 args->f_id.src_ip6: 2522 args->f_id.dst_ip6; 2523 APPLY_MASK(&p, &d[1]); 2524 match = 2525 IN6_ARE_ADDR_EQUAL(&d[0], 2526 &p); 2527 } 2528 } 2529 break; 2530 2531 case O_FLOW6ID: 2532 match = is_ipv6 && 2533 flow6id_match(args->f_id.flow_id6, 2534 (ipfw_insn_u32 *) cmd); 2535 break; 2536 2537 case O_EXT_HDR: 2538 match = is_ipv6 && 2539 (ext_hd & ((ipfw_insn *) cmd)->arg1); 2540 break; 2541 2542 case O_IP6: 2543 match = is_ipv6; 2544 break; 2545#endif 2546 2547 case O_IP4: 2548 match = is_ipv4; 2549 break; 2550 2551 case O_TAG: { 2552 struct m_tag *mtag; 2553 uint32_t tag = TARG(cmd->arg1, tag); 2554 2555 /* Packet is already tagged with this tag? */ 2556 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 2557 2558 /* We have `untag' action when F_NOT flag is 2559 * present. And we must remove this mtag from 2560 * mbuf and reset `match' to zero (`match' will 2561 * be inversed later). 2562 * Otherwise we should allocate new mtag and 2563 * push it into mbuf. 2564 */ 2565 if (cmd->len & F_NOT) { /* `untag' action */ 2566 if (mtag != NULL) 2567 m_tag_delete(m, mtag); 2568 match = 0; 2569 } else { 2570 if (mtag == NULL) { 2571 mtag = m_tag_alloc( MTAG_IPFW, 2572 tag, 0, M_NOWAIT); 2573 if (mtag != NULL) 2574 m_tag_prepend(m, mtag); 2575 } 2576 match = 1; 2577 } 2578 break; 2579 } 2580 2581 case O_FIB: /* try match the specified fib */ 2582 if (args->f_id.fib == cmd->arg1) 2583 match = 1; 2584 break; 2585 2586 case O_SOCKARG: { 2587#ifndef USERSPACE /* not supported in userspace */ 2588 struct inpcb *inp = args->inp; 2589 struct inpcbinfo *pi; 2590 bool inp_locked = false; 2591 2592 if (proto == IPPROTO_TCP) 2593 pi = &V_tcbinfo; 2594 else if (proto == IPPROTO_UDP) 2595 pi = &V_udbinfo; 2596 else if (proto == IPPROTO_UDPLITE) 2597 pi = &V_ulitecbinfo; 2598 else 2599 break; 2600 2601 /* 2602 * XXXRW: so_user_cookie should almost 2603 * certainly be inp_user_cookie? 2604 */ 2605 2606 /* 2607 * For incoming packet lookup the inpcb 2608 * using the src/dest ip/port tuple. 2609 */ 2610 if (is_ipv4 && inp == NULL) { 2611 inp = in_pcblookup(pi, 2612 src_ip, htons(src_port), 2613 dst_ip, htons(dst_port), 2614 INPLOOKUP_RLOCKPCB, NULL); 2615 inp_locked = true; 2616 } 2617#ifdef INET6 2618 if (is_ipv6 && inp == NULL) { 2619 inp = in6_pcblookup(pi, 2620 &args->f_id.src_ip6, 2621 htons(src_port), 2622 &args->f_id.dst_ip6, 2623 htons(dst_port), 2624 INPLOOKUP_RLOCKPCB, NULL); 2625 inp_locked = true; 2626 } 2627#endif /* INET6 */ 2628 if (inp != NULL) { 2629 if (inp->inp_socket) { 2630 tablearg = 2631 inp->inp_socket->so_user_cookie; 2632 if (tablearg) 2633 match = 1; 2634 } 2635 if (inp_locked) 2636 INP_RUNLOCK(inp); 2637 } 2638#endif /* !USERSPACE */ 2639 break; 2640 } 2641 2642 case O_TAGGED: { 2643 struct m_tag *mtag; 2644 uint32_t tag = TARG(cmd->arg1, tag); 2645 2646 if (cmdlen == 1) { 2647 match = m_tag_locate(m, MTAG_IPFW, 2648 tag, NULL) != NULL; 2649 break; 2650 } 2651 2652 /* we have ranges */ 2653 for (mtag = m_tag_first(m); 2654 mtag != NULL && !match; 2655 mtag = m_tag_next(m, mtag)) { 2656 uint16_t *p; 2657 int i; 2658 2659 if (mtag->m_tag_cookie != MTAG_IPFW) 2660 continue; 2661 2662 p = ((ipfw_insn_u16 *)cmd)->ports; 2663 i = cmdlen - 1; 2664 for(; !match && i > 0; i--, p += 2) 2665 match = 2666 mtag->m_tag_id >= p[0] && 2667 mtag->m_tag_id <= p[1]; 2668 } 2669 break; 2670 } 2671 2672 /* 2673 * The second set of opcodes represents 'actions', 2674 * i.e. the terminal part of a rule once the packet 2675 * matches all previous patterns. 2676 * Typically there is only one action for each rule, 2677 * and the opcode is stored at the end of the rule 2678 * (but there are exceptions -- see below). 2679 * 2680 * In general, here we set retval and terminate the 2681 * outer loop (would be a 'break 3' in some language, 2682 * but we need to set l=0, done=1) 2683 * 2684 * Exceptions: 2685 * O_COUNT and O_SKIPTO actions: 2686 * instead of terminating, we jump to the next rule 2687 * (setting l=0), or to the SKIPTO target (setting 2688 * f/f_len, cmd and l as needed), respectively. 2689 * 2690 * O_TAG, O_LOG and O_ALTQ action parameters: 2691 * perform some action and set match = 1; 2692 * 2693 * O_LIMIT and O_KEEP_STATE: these opcodes are 2694 * not real 'actions', and are stored right 2695 * before the 'action' part of the rule (one 2696 * exception is O_SKIP_ACTION which could be 2697 * between these opcodes and 'action' one). 2698 * These opcodes try to install an entry in the 2699 * state tables; if successful, we continue with 2700 * the next opcode (match=1; break;), otherwise 2701 * the packet must be dropped (set retval, 2702 * break loops with l=0, done=1) 2703 * 2704 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2705 * cause a lookup of the state table, and a jump 2706 * to the 'action' part of the parent rule 2707 * if an entry is found, or 2708 * (CHECK_STATE only) a jump to the next rule if 2709 * the entry is not found. 2710 * The result of the lookup is cached so that 2711 * further instances of these opcodes become NOPs. 2712 * The jump to the next rule is done by setting 2713 * l=0, cmdlen=0. 2714 * 2715 * O_SKIP_ACTION: this opcode is not a real 'action' 2716 * either, and is stored right before the 'action' 2717 * part of the rule, right after the O_KEEP_STATE 2718 * opcode. It causes match failure so the real 2719 * 'action' could be executed only if the rule 2720 * is checked via dynamic rule from the state 2721 * table, as in such case execution starts 2722 * from the true 'action' opcode directly. 2723 * 2724 */ 2725 case O_LIMIT: 2726 case O_KEEP_STATE: 2727 if (ipfw_dyn_install_state(chain, f, 2728 (ipfw_insn_limit *)cmd, args, ulp, 2729 pktlen, &dyn_info, tablearg)) { 2730 /* error or limit violation */ 2731 retval = IP_FW_DENY; 2732 l = 0; /* exit inner loop */ 2733 done = 1; /* exit outer loop */ 2734 } 2735 match = 1; 2736 break; 2737 2738 case O_PROBE_STATE: 2739 case O_CHECK_STATE: 2740 /* 2741 * dynamic rules are checked at the first 2742 * keep-state or check-state occurrence, 2743 * with the result being stored in dyn_info. 2744 * The compiler introduces a PROBE_STATE 2745 * instruction for us when we have a 2746 * KEEP_STATE (because PROBE_STATE needs 2747 * to be run first). 2748 */ 2749 if (DYN_LOOKUP_NEEDED(&dyn_info, cmd) && 2750 (q = ipfw_dyn_lookup_state(args, ulp, 2751 pktlen, cmd, &dyn_info)) != NULL) { 2752 /* 2753 * Found dynamic entry, jump to the 2754 * 'action' part of the parent rule 2755 * by setting f, cmd, l and clearing 2756 * cmdlen. 2757 */ 2758 f = q; 2759 f_pos = dyn_info.f_pos; 2760 cmd = ACTION_PTR(f); 2761 l = f->cmd_len - f->act_ofs; 2762 cmdlen = 0; 2763 match = 1; 2764 break; 2765 } 2766 /* 2767 * Dynamic entry not found. If CHECK_STATE, 2768 * skip to next rule, if PROBE_STATE just 2769 * ignore and continue with next opcode. 2770 */ 2771 if (cmd->opcode == O_CHECK_STATE) 2772 l = 0; /* exit inner loop */ 2773 match = 1; 2774 break; 2775 2776 case O_SKIP_ACTION: 2777 match = 0; /* skip to the next rule */ 2778 l = 0; /* exit inner loop */ 2779 break; 2780 2781 case O_ACCEPT: 2782 retval = 0; /* accept */ 2783 l = 0; /* exit inner loop */ 2784 done = 1; /* exit outer loop */ 2785 break; 2786 2787 case O_PIPE: 2788 case O_QUEUE: 2789 set_match(args, f_pos, chain); 2790 args->rule.info = TARG(cmd->arg1, pipe); 2791 if (cmd->opcode == O_PIPE) 2792 args->rule.info |= IPFW_IS_PIPE; 2793 if (V_fw_one_pass) 2794 args->rule.info |= IPFW_ONEPASS; 2795 retval = IP_FW_DUMMYNET; 2796 l = 0; /* exit inner loop */ 2797 done = 1; /* exit outer loop */ 2798 break; 2799 2800 case O_DIVERT: 2801 case O_TEE: 2802 if (args->flags & IPFW_ARGS_ETHER) 2803 break; /* not on layer 2 */ 2804 /* otherwise this is terminal */ 2805 l = 0; /* exit inner loop */ 2806 done = 1; /* exit outer loop */ 2807 retval = (cmd->opcode == O_DIVERT) ? 2808 IP_FW_DIVERT : IP_FW_TEE; 2809 set_match(args, f_pos, chain); 2810 args->rule.info = TARG(cmd->arg1, divert); 2811 break; 2812 2813 case O_COUNT: 2814 IPFW_INC_RULE_COUNTER(f, pktlen); 2815 l = 0; /* exit inner loop */ 2816 break; 2817 2818 case O_SKIPTO: 2819 IPFW_INC_RULE_COUNTER(f, pktlen); 2820 f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0); 2821 /* 2822 * Skip disabled rules, and re-enter 2823 * the inner loop with the correct 2824 * f_pos, f, l and cmd. 2825 * Also clear cmdlen and skip_or 2826 */ 2827 for (; f_pos < chain->n_rules - 1 && 2828 (V_set_disable & 2829 (1 << chain->map[f_pos]->set)); 2830 f_pos++) 2831 ; 2832 /* Re-enter the inner loop at the skipto rule. */ 2833 f = chain->map[f_pos]; 2834 l = f->cmd_len; 2835 cmd = f->cmd; 2836 match = 1; 2837 cmdlen = 0; 2838 skip_or = 0; 2839 continue; 2840 break; /* not reached */ 2841 2842 case O_CALLRETURN: { 2843 /* 2844 * Implementation of `subroutine' call/return, 2845 * in the stack carried in an mbuf tag. This 2846 * is different from `skipto' in that any call 2847 * address is possible (`skipto' must prevent 2848 * backward jumps to avoid endless loops). 2849 * We have `return' action when F_NOT flag is 2850 * present. The `m_tag_id' field is used as 2851 * stack pointer. 2852 */ 2853 struct m_tag *mtag; 2854 uint16_t jmpto, *stack; 2855 2856#define IS_CALL ((cmd->len & F_NOT) == 0) 2857#define IS_RETURN ((cmd->len & F_NOT) != 0) 2858 /* 2859 * Hand-rolled version of m_tag_locate() with 2860 * wildcard `type'. 2861 * If not already tagged, allocate new tag. 2862 */ 2863 mtag = m_tag_first(m); 2864 while (mtag != NULL) { 2865 if (mtag->m_tag_cookie == 2866 MTAG_IPFW_CALL) 2867 break; 2868 mtag = m_tag_next(m, mtag); 2869 } 2870 if (mtag == NULL && IS_CALL) { 2871 mtag = m_tag_alloc(MTAG_IPFW_CALL, 0, 2872 IPFW_CALLSTACK_SIZE * 2873 sizeof(uint16_t), M_NOWAIT); 2874 if (mtag != NULL) 2875 m_tag_prepend(m, mtag); 2876 } 2877 2878 /* 2879 * On error both `call' and `return' just 2880 * continue with next rule. 2881 */ 2882 if (IS_RETURN && (mtag == NULL || 2883 mtag->m_tag_id == 0)) { 2884 l = 0; /* exit inner loop */ 2885 break; 2886 } 2887 if (IS_CALL && (mtag == NULL || 2888 mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) { 2889 printf("ipfw: call stack error, " 2890 "go to next rule\n"); 2891 l = 0; /* exit inner loop */ 2892 break; 2893 } 2894 2895 IPFW_INC_RULE_COUNTER(f, pktlen); 2896 stack = (uint16_t *)(mtag + 1); 2897 2898 /* 2899 * The `call' action may use cached f_pos 2900 * (in f->next_rule), whose version is written 2901 * in f->next_rule. 2902 * The `return' action, however, doesn't have 2903 * fixed jump address in cmd->arg1 and can't use 2904 * cache. 2905 */ 2906 if (IS_CALL) { 2907 stack[mtag->m_tag_id] = f->rulenum; 2908 mtag->m_tag_id++; 2909 f_pos = JUMP(chain, f, cmd->arg1, 2910 tablearg, 1); 2911 } else { /* `return' action */ 2912 mtag->m_tag_id--; 2913 jmpto = stack[mtag->m_tag_id] + 1; 2914 f_pos = ipfw_find_rule(chain, jmpto, 0); 2915 } 2916 2917 /* 2918 * Skip disabled rules, and re-enter 2919 * the inner loop with the correct 2920 * f_pos, f, l and cmd. 2921 * Also clear cmdlen and skip_or 2922 */ 2923 for (; f_pos < chain->n_rules - 1 && 2924 (V_set_disable & 2925 (1 << chain->map[f_pos]->set)); f_pos++) 2926 ; 2927 /* Re-enter the inner loop at the dest rule. */ 2928 f = chain->map[f_pos]; 2929 l = f->cmd_len; 2930 cmd = f->cmd; 2931 cmdlen = 0; 2932 skip_or = 0; 2933 continue; 2934 break; /* NOTREACHED */ 2935 } 2936#undef IS_CALL 2937#undef IS_RETURN 2938 2939 case O_REJECT: 2940 /* 2941 * Drop the packet and send a reject notice 2942 * if the packet is not ICMP (or is an ICMP 2943 * query), and it is not multicast/broadcast. 2944 */ 2945 if (hlen > 0 && is_ipv4 && offset == 0 && 2946 (proto != IPPROTO_ICMP || 2947 is_icmp_query(ICMP(ulp))) && 2948 !(m->m_flags & (M_BCAST|M_MCAST)) && 2949 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 2950 send_reject(args, cmd->arg1, iplen, ip); 2951 m = args->m; 2952 } 2953 /* FALLTHROUGH */ 2954#ifdef INET6 2955 case O_UNREACH6: 2956 if (hlen > 0 && is_ipv6 && 2957 ((offset & IP6F_OFF_MASK) == 0) && 2958 (proto != IPPROTO_ICMPV6 || 2959 (is_icmp6_query(icmp6_type) == 1)) && 2960 !(m->m_flags & (M_BCAST|M_MCAST)) && 2961 !IN6_IS_ADDR_MULTICAST( 2962 &args->f_id.dst_ip6)) { 2963 send_reject6(args, 2964 cmd->opcode == O_REJECT ? 2965 map_icmp_unreach(cmd->arg1): 2966 cmd->arg1, hlen, 2967 (struct ip6_hdr *)ip); 2968 m = args->m; 2969 } 2970 /* FALLTHROUGH */ 2971#endif 2972 case O_DENY: 2973 retval = IP_FW_DENY; 2974 l = 0; /* exit inner loop */ 2975 done = 1; /* exit outer loop */ 2976 break; 2977 2978 case O_FORWARD_IP: 2979 if (args->flags & IPFW_ARGS_ETHER) 2980 break; /* not valid on layer2 pkts */ 2981 if (q != f || 2982 dyn_info.direction == MATCH_FORWARD) { 2983 struct sockaddr_in *sa; 2984 2985 sa = &(((ipfw_insn_sa *)cmd)->sa); 2986 if (sa->sin_addr.s_addr == INADDR_ANY) { 2987#ifdef INET6 2988 /* 2989 * We use O_FORWARD_IP opcode for 2990 * fwd rule with tablearg, but tables 2991 * now support IPv6 addresses. And 2992 * when we are inspecting IPv6 packet, 2993 * we can use nh6 field from 2994 * table_value as next_hop6 address. 2995 */ 2996 if (is_ipv6) { 2997 struct ip_fw_nh6 *nh6; 2998 2999 args->flags |= IPFW_ARGS_NH6; 3000 nh6 = &args->hopstore6; 3001 nh6->sin6_addr = TARG_VAL( 3002 chain, tablearg, nh6); 3003 nh6->sin6_port = sa->sin_port; 3004 nh6->sin6_scope_id = TARG_VAL( 3005 chain, tablearg, zoneid); 3006 } else 3007#endif 3008 { 3009 args->flags |= IPFW_ARGS_NH4; 3010 args->hopstore.sin_port = 3011 sa->sin_port; 3012 sa = &args->hopstore; 3013 sa->sin_family = AF_INET; 3014 sa->sin_len = sizeof(*sa); 3015 sa->sin_addr.s_addr = htonl( 3016 TARG_VAL(chain, tablearg, 3017 nh4)); 3018 } 3019 } else { 3020 args->flags |= IPFW_ARGS_NH4PTR; 3021 args->next_hop = sa; 3022 } 3023 } 3024 retval = IP_FW_PASS; 3025 l = 0; /* exit inner loop */ 3026 done = 1; /* exit outer loop */ 3027 break; 3028 3029#ifdef INET6 3030 case O_FORWARD_IP6: 3031 if (args->flags & IPFW_ARGS_ETHER) 3032 break; /* not valid on layer2 pkts */ 3033 if (q != f || 3034 dyn_info.direction == MATCH_FORWARD) { 3035 struct sockaddr_in6 *sin6; 3036 3037 sin6 = &(((ipfw_insn_sa6 *)cmd)->sa); 3038 args->flags |= IPFW_ARGS_NH6PTR; 3039 args->next_hop6 = sin6; 3040 } 3041 retval = IP_FW_PASS; 3042 l = 0; /* exit inner loop */ 3043 done = 1; /* exit outer loop */ 3044 break; 3045#endif 3046 3047 case O_NETGRAPH: 3048 case O_NGTEE: 3049 set_match(args, f_pos, chain); 3050 args->rule.info = TARG(cmd->arg1, netgraph); 3051 if (V_fw_one_pass) 3052 args->rule.info |= IPFW_ONEPASS; 3053 retval = (cmd->opcode == O_NETGRAPH) ? 3054 IP_FW_NETGRAPH : IP_FW_NGTEE; 3055 l = 0; /* exit inner loop */ 3056 done = 1; /* exit outer loop */ 3057 break; 3058 3059 case O_SETFIB: { 3060 uint32_t fib; 3061 3062 IPFW_INC_RULE_COUNTER(f, pktlen); 3063 fib = TARG(cmd->arg1, fib) & 0x7FFF; 3064 if (fib >= rt_numfibs) 3065 fib = 0; 3066 M_SETFIB(m, fib); 3067 args->f_id.fib = fib; /* XXX */ 3068 l = 0; /* exit inner loop */ 3069 break; 3070 } 3071 3072 case O_SETDSCP: { 3073 uint16_t code; 3074 3075 code = TARG(cmd->arg1, dscp) & 0x3F; 3076 l = 0; /* exit inner loop */ 3077 if (is_ipv4) { 3078 uint16_t old; 3079 3080 old = *(uint16_t *)ip; 3081 ip->ip_tos = (code << 2) | 3082 (ip->ip_tos & 0x03); 3083 ip->ip_sum = cksum_adjust(ip->ip_sum, 3084 old, *(uint16_t *)ip); 3085 } else if (is_ipv6) { 3086 uint8_t *v; 3087 3088 v = &((struct ip6_hdr *)ip)->ip6_vfc; 3089 *v = (*v & 0xF0) | (code >> 2); 3090 v++; 3091 *v = (*v & 0x3F) | ((code & 0x03) << 6); 3092 } else 3093 break; 3094 3095 IPFW_INC_RULE_COUNTER(f, pktlen); 3096 break; 3097 } 3098 3099 case O_NAT: 3100 l = 0; /* exit inner loop */ 3101 done = 1; /* exit outer loop */ 3102 /* 3103 * Ensure that we do not invoke NAT handler for 3104 * non IPv4 packets. Libalias expects only IPv4. 3105 */ 3106 if (!is_ipv4 || !IPFW_NAT_LOADED) { 3107 retval = IP_FW_DENY; 3108 break; 3109 } 3110 3111 struct cfg_nat *t; 3112 int nat_id; 3113 3114 args->rule.info = 0; 3115 set_match(args, f_pos, chain); 3116 /* Check if this is 'global' nat rule */ 3117 if (cmd->arg1 == IP_FW_NAT44_GLOBAL) { 3118 retval = ipfw_nat_ptr(args, NULL, m); 3119 break; 3120 } 3121 t = ((ipfw_insn_nat *)cmd)->nat; 3122 if (t == NULL) { 3123 nat_id = TARG(cmd->arg1, nat); 3124 t = (*lookup_nat_ptr)(&chain->nat, nat_id); 3125 3126 if (t == NULL) { 3127 retval = IP_FW_DENY; 3128 break; 3129 } 3130 if (cmd->arg1 != IP_FW_TARG) 3131 ((ipfw_insn_nat *)cmd)->nat = t; 3132 } 3133 retval = ipfw_nat_ptr(args, t, m); 3134 break; 3135 3136 case O_REASS: { 3137 int ip_off; 3138 3139 l = 0; /* in any case exit inner loop */ 3140 if (is_ipv6) /* IPv6 is not supported yet */ 3141 break; 3142 IPFW_INC_RULE_COUNTER(f, pktlen); 3143 ip_off = ntohs(ip->ip_off); 3144 3145 /* if not fragmented, go to next rule */ 3146 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0) 3147 break; 3148 3149 args->m = m = ip_reass(m); 3150 3151 /* 3152 * do IP header checksum fixup. 3153 */ 3154 if (m == NULL) { /* fragment got swallowed */ 3155 retval = IP_FW_DENY; 3156 } else { /* good, packet complete */ 3157 int hlen; 3158 3159 ip = mtod(m, struct ip *); 3160 hlen = ip->ip_hl << 2; 3161 ip->ip_sum = 0; 3162 if (hlen == sizeof(struct ip)) 3163 ip->ip_sum = in_cksum_hdr(ip); 3164 else 3165 ip->ip_sum = in_cksum(m, hlen); 3166 retval = IP_FW_REASS; 3167 args->rule.info = 0; 3168 set_match(args, f_pos, chain); 3169 } 3170 done = 1; /* exit outer loop */ 3171 break; 3172 } 3173 case O_EXTERNAL_ACTION: 3174 l = 0; /* in any case exit inner loop */ 3175 retval = ipfw_run_eaction(chain, args, 3176 cmd, &done); 3177 /* 3178 * If both @retval and @done are zero, 3179 * consider this as rule matching and 3180 * update counters. 3181 */ 3182 if (retval == 0 && done == 0) { 3183 IPFW_INC_RULE_COUNTER(f, pktlen); 3184 /* 3185 * Reset the result of the last 3186 * dynamic state lookup. 3187 * External action can change 3188 * @args content, and it may be 3189 * used for new state lookup later. 3190 */ 3191 DYN_INFO_INIT(&dyn_info); 3192 } 3193 break; 3194 3195 default: 3196 panic("-- unknown opcode %d\n", cmd->opcode); 3197 } /* end of switch() on opcodes */ 3198 /* 3199 * if we get here with l=0, then match is irrelevant. 3200 */ 3201 3202 if (cmd->len & F_NOT) 3203 match = !match; 3204 3205 if (match) { 3206 if (cmd->len & F_OR) 3207 skip_or = 1; 3208 } else { 3209 if (!(cmd->len & F_OR)) /* not an OR block, */ 3210 break; /* try next rule */ 3211 } 3212 3213 } /* end of inner loop, scan opcodes */ 3214#undef PULLUP_LEN 3215#undef PULLUP_LEN_LOCKED 3216 3217 if (done) 3218 break; 3219 3220/* next_rule:; */ /* try next rule */ 3221 3222 } /* end of outer for, scan rules */ 3223 3224 if (done) { 3225 struct ip_fw *rule = chain->map[f_pos]; 3226 /* Update statistics */ 3227 IPFW_INC_RULE_COUNTER(rule, pktlen); 3228 IPFW_PROBE(rule__matched, retval, 3229 is_ipv4 ? AF_INET : AF_INET6, 3230 is_ipv4 ? (uintptr_t)&src_ip : 3231 (uintptr_t)&args->f_id.src_ip6, 3232 is_ipv4 ? (uintptr_t)&dst_ip : 3233 (uintptr_t)&args->f_id.dst_ip6, 3234 args, rule); 3235 } else { 3236 retval = IP_FW_DENY; 3237 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3238 } 3239 IPFW_PF_RUNLOCK(chain); 3240#ifdef __FreeBSD__ 3241 if (ucred_cache != NULL) 3242 crfree(ucred_cache); 3243#endif 3244 return (retval); 3245 3246pullup_failed: 3247 if (V_fw_verbose) 3248 printf("ipfw: pullup failed\n"); 3249 return (IP_FW_DENY); 3250} 3251 3252/* 3253 * Set maximum number of tables that can be used in given VNET ipfw instance. 3254 */ 3255#ifdef SYSCTL_NODE 3256static int 3257sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS) 3258{ 3259 int error; 3260 unsigned int ntables; 3261 3262 ntables = V_fw_tables_max; 3263 3264 error = sysctl_handle_int(oidp, &ntables, 0, req); 3265 /* Read operation or some error */ 3266 if ((error != 0) || (req->newptr == NULL)) 3267 return (error); 3268 3269 return (ipfw_resize_tables(&V_layer3_chain, ntables)); 3270} 3271 3272/* 3273 * Switches table namespace between global and per-set. 3274 */ 3275static int 3276sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS) 3277{ 3278 int error; 3279 unsigned int sets; 3280 3281 sets = V_fw_tables_sets; 3282 3283 error = sysctl_handle_int(oidp, &sets, 0, req); 3284 /* Read operation or some error */ 3285 if ((error != 0) || (req->newptr == NULL)) 3286 return (error); 3287 3288 return (ipfw_switch_tables_namespace(&V_layer3_chain, sets)); 3289} 3290#endif 3291 3292/* 3293 * Module and VNET glue 3294 */ 3295 3296/* 3297 * Stuff that must be initialised only on boot or module load 3298 */ 3299static int 3300ipfw_init(void) 3301{ 3302 int error = 0; 3303 3304 /* 3305 * Only print out this stuff the first time around, 3306 * when called from the sysinit code. 3307 */ 3308 printf("ipfw2 " 3309#ifdef INET6 3310 "(+ipv6) " 3311#endif 3312 "initialized, divert %s, nat %s, " 3313 "default to %s, logging ", 3314#ifdef IPDIVERT 3315 "enabled", 3316#else 3317 "loadable", 3318#endif 3319#ifdef IPFIREWALL_NAT 3320 "enabled", 3321#else 3322 "loadable", 3323#endif 3324 default_to_accept ? "accept" : "deny"); 3325 3326 /* 3327 * Note: V_xxx variables can be accessed here but the vnet specific 3328 * initializer may not have been called yet for the VIMAGE case. 3329 * Tuneables will have been processed. We will print out values for 3330 * the default vnet. 3331 * XXX This should all be rationalized AFTER 8.0 3332 */ 3333 if (V_fw_verbose == 0) 3334 printf("disabled\n"); 3335 else if (V_verbose_limit == 0) 3336 printf("unlimited\n"); 3337 else 3338 printf("limited to %d packets/entry by default\n", 3339 V_verbose_limit); 3340 3341 /* Check user-supplied table count for validness */ 3342 if (default_fw_tables > IPFW_TABLES_MAX) 3343 default_fw_tables = IPFW_TABLES_MAX; 3344 3345 ipfw_init_sopt_handler(); 3346 ipfw_init_obj_rewriter(); 3347 ipfw_iface_init(); 3348 return (error); 3349} 3350 3351/* 3352 * Called for the removal of the last instance only on module unload. 3353 */ 3354static void 3355ipfw_destroy(void) 3356{ 3357 3358 ipfw_iface_destroy(); 3359 ipfw_destroy_sopt_handler(); 3360 ipfw_destroy_obj_rewriter(); 3361 printf("IP firewall unloaded\n"); 3362} 3363 3364/* 3365 * Stuff that must be initialized for every instance 3366 * (including the first of course). 3367 */ 3368static int 3369vnet_ipfw_init(const void *unused) 3370{ 3371 int error, first; 3372 struct ip_fw *rule = NULL; 3373 struct ip_fw_chain *chain; 3374 3375 chain = &V_layer3_chain; 3376 3377 first = IS_DEFAULT_VNET(curvnet) ? 1 : 0; 3378 3379 /* First set up some values that are compile time options */ 3380 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 3381 V_fw_deny_unknown_exthdrs = 1; 3382#ifdef IPFIREWALL_VERBOSE 3383 V_fw_verbose = 1; 3384#endif 3385#ifdef IPFIREWALL_VERBOSE_LIMIT 3386 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 3387#endif 3388#ifdef IPFIREWALL_NAT 3389 LIST_INIT(&chain->nat); 3390#endif 3391 3392 /* Init shared services hash table */ 3393 ipfw_init_srv(chain); 3394 3395 ipfw_init_counters(); 3396 /* Set initial number of tables */ 3397 V_fw_tables_max = default_fw_tables; 3398 error = ipfw_init_tables(chain, first); 3399 if (error) { 3400 printf("ipfw2: setting up tables failed\n"); 3401 free(chain->map, M_IPFW); 3402 free(rule, M_IPFW); 3403 return (ENOSPC); 3404 } 3405 3406 IPFW_LOCK_INIT(chain); 3407 3408 /* fill and insert the default rule */ 3409 rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw)); 3410 rule->flags |= IPFW_RULE_NOOPT; 3411 rule->cmd_len = 1; 3412 rule->cmd[0].len = 1; 3413 rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; 3414 chain->default_rule = rule; 3415 ipfw_add_protected_rule(chain, rule, 0); 3416 3417 ipfw_dyn_init(chain); 3418 ipfw_eaction_init(chain, first); 3419#ifdef LINEAR_SKIPTO 3420 ipfw_init_skipto_cache(chain); 3421#endif 3422 ipfw_bpf_init(first); 3423 3424 /* First set up some values that are compile time options */ 3425 V_ipfw_vnet_ready = 1; /* Open for business */ 3426 3427 /* 3428 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6. 3429 * Even if the latter two fail we still keep the module alive 3430 * because the sockopt and layer2 paths are still useful. 3431 * ipfw[6]_hook return 0 on success, ENOENT on failure, 3432 * so we can ignore the exact return value and just set a flag. 3433 * 3434 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so 3435 * changes in the underlying (per-vnet) variables trigger 3436 * immediate hook()/unhook() calls. 3437 * In layer2 we have the same behaviour, except that V_ether_ipfw 3438 * is checked on each packet because there are no pfil hooks. 3439 */ 3440 V_ip_fw_ctl_ptr = ipfw_ctl3; 3441 error = ipfw_attach_hooks(1); 3442 return (error); 3443} 3444 3445/* 3446 * Called for the removal of each instance. 3447 */ 3448static int 3449vnet_ipfw_uninit(const void *unused) 3450{ 3451 struct ip_fw *reap; 3452 struct ip_fw_chain *chain = &V_layer3_chain; 3453 int i, last; 3454 3455 V_ipfw_vnet_ready = 0; /* tell new callers to go away */ 3456 /* 3457 * disconnect from ipv4, ipv6, layer2 and sockopt. 3458 * Then grab, release and grab again the WLOCK so we make 3459 * sure the update is propagated and nobody will be in. 3460 */ 3461 (void)ipfw_attach_hooks(0 /* detach */); 3462 V_ip_fw_ctl_ptr = NULL; 3463 3464 last = IS_DEFAULT_VNET(curvnet) ? 1 : 0; 3465 3466 IPFW_UH_WLOCK(chain); 3467 IPFW_UH_WUNLOCK(chain); 3468 3469 ipfw_dyn_uninit(0); /* run the callout_drain */ 3470 3471 IPFW_UH_WLOCK(chain); 3472 3473 reap = NULL; 3474 IPFW_WLOCK(chain); 3475 for (i = 0; i < chain->n_rules; i++) 3476 ipfw_reap_add(chain, &reap, chain->map[i]); 3477 free(chain->map, M_IPFW); 3478#ifdef LINEAR_SKIPTO 3479 ipfw_destroy_skipto_cache(chain); 3480#endif 3481 IPFW_WUNLOCK(chain); 3482 IPFW_UH_WUNLOCK(chain); 3483 ipfw_destroy_tables(chain, last); 3484 ipfw_eaction_uninit(chain, last); 3485 if (reap != NULL) 3486 ipfw_reap_rules(reap); 3487 vnet_ipfw_iface_destroy(chain); 3488 ipfw_destroy_srv(chain); 3489 IPFW_LOCK_DESTROY(chain); 3490 ipfw_dyn_uninit(1); /* free the remaining parts */ 3491 ipfw_destroy_counters(); 3492 ipfw_bpf_uninit(last); 3493 return (0); 3494} 3495 3496/* 3497 * Module event handler. 3498 * In general we have the choice of handling most of these events by the 3499 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to 3500 * use the SYSINIT handlers as they are more capable of expressing the 3501 * flow of control during module and vnet operations, so this is just 3502 * a skeleton. Note there is no SYSINIT equivalent of the module 3503 * SHUTDOWN handler, but we don't have anything to do in that case anyhow. 3504 */ 3505static int 3506ipfw_modevent(module_t mod, int type, void *unused) 3507{ 3508 int err = 0; 3509 3510 switch (type) { 3511 case MOD_LOAD: 3512 /* Called once at module load or 3513 * system boot if compiled in. */ 3514 break; 3515 case MOD_QUIESCE: 3516 /* Called before unload. May veto unloading. */ 3517 break; 3518 case MOD_UNLOAD: 3519 /* Called during unload. */ 3520 break; 3521 case MOD_SHUTDOWN: 3522 /* Called during system shutdown. */ 3523 break; 3524 default: 3525 err = EOPNOTSUPP; 3526 break; 3527 } 3528 return err; 3529} 3530 3531static moduledata_t ipfwmod = { 3532 "ipfw", 3533 ipfw_modevent, 3534 0 3535}; 3536 3537/* Define startup order. */ 3538#define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_FIREWALL 3539#define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */ 3540#define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */ 3541#define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */ 3542 3543DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER); 3544FEATURE(ipfw_ctl3, "ipfw new sockopt calls"); 3545MODULE_VERSION(ipfw, 3); 3546/* should declare some dependencies here */ 3547 3548/* 3549 * Starting up. Done in order after ipfwmod() has been called. 3550 * VNET_SYSINIT is also called for each existing vnet and each new vnet. 3551 */ 3552SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 3553 ipfw_init, NULL); 3554VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 3555 vnet_ipfw_init, NULL); 3556 3557/* 3558 * Closing up shop. These are done in REVERSE ORDER, but still 3559 * after ipfwmod() has been called. Not called on reboot. 3560 * VNET_SYSUNINIT is also called for each exiting vnet as it exits. 3561 * or when the module is unloaded. 3562 */ 3563SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 3564 ipfw_destroy, NULL); 3565VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 3566 vnet_ipfw_uninit, NULL); 3567/* end of file */ 3568