ip_fw2.c revision 200027
1/*- 2 * Copyright (c) 2002 Luigi Rizzo, Universita` di Pisa 3 * 4 * Redistribution and use in source and binary forms, with or without 5 * modification, are permitted provided that the following conditions 6 * are met: 7 * 1. Redistributions of source code must retain the above copyright 8 * notice, this list of conditions and the following disclaimer. 9 * 2. Redistributions in binary form must reproduce the above copyright 10 * notice, this list of conditions and the following disclaimer in the 11 * documentation and/or other materials provided with the distribution. 12 * 13 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 14 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 15 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 16 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 17 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 18 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 19 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 20 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 21 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 22 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 23 * SUCH DAMAGE. 24 */ 25 26#include <sys/cdefs.h> 27__FBSDID("$FreeBSD: head/sys/netinet/ipfw/ip_fw2.c 200027 2009-12-02 14:32:01Z ume $"); 28 29#define DEB(x) 30#define DDB(x) x 31 32/* 33 * Implement IP packet firewall (new version) 34 */ 35 36#if !defined(KLD_MODULE) 37#include "opt_ipfw.h" 38#include "opt_ipdivert.h" 39#include "opt_ipdn.h" 40#include "opt_inet.h" 41#ifndef INET 42#error IPFIREWALL requires INET. 43#endif /* INET */ 44#endif 45#include "opt_inet6.h" 46#include "opt_ipsec.h" 47 48#include <sys/param.h> 49#include <sys/systm.h> 50#include <sys/condvar.h> 51#include <sys/eventhandler.h> 52#include <sys/malloc.h> 53#include <sys/mbuf.h> 54#include <sys/kernel.h> 55#include <sys/lock.h> 56#include <sys/jail.h> 57#include <sys/module.h> 58#include <sys/priv.h> 59#include <sys/proc.h> 60#include <sys/rwlock.h> 61#include <sys/socket.h> 62#include <sys/socketvar.h> 63#include <sys/sysctl.h> 64#include <sys/syslog.h> 65#include <sys/ucred.h> 66#include <net/ethernet.h> /* for ETHERTYPE_IP */ 67#include <net/if.h> 68#include <net/radix.h> 69#include <net/route.h> 70#include <net/pf_mtag.h> 71#include <net/vnet.h> 72 73#define IPFW_INTERNAL /* Access to protected data structures in ip_fw.h. */ 74 75#include <netinet/in.h> 76#include <netinet/in_var.h> 77#include <netinet/in_pcb.h> 78#include <netinet/ip.h> 79#include <netinet/ip_var.h> 80#include <netinet/ip_icmp.h> 81#include <netinet/ip_fw.h> 82#include <netinet/ip_divert.h> 83#include <netinet/ip_dummynet.h> 84#include <netinet/ip_carp.h> 85#include <netinet/pim.h> 86#include <netinet/tcp_var.h> 87#include <netinet/udp.h> 88#include <netinet/udp_var.h> 89#include <netinet/sctp.h> 90 91#include <netgraph/ng_ipfw.h> 92 93#include <netinet/ip6.h> 94#include <netinet/icmp6.h> 95#ifdef INET6 96#include <netinet6/scope6_var.h> 97#include <netinet6/ip6_var.h> 98#endif 99 100#include <machine/in_cksum.h> /* XXX for in_cksum */ 101 102#ifdef MAC 103#include <security/mac/mac_framework.h> 104#endif 105 106static VNET_DEFINE(int, ipfw_vnet_ready) = 0; 107#define V_ipfw_vnet_ready VNET(ipfw_vnet_ready) 108/* 109 * set_disable contains one bit per set value (0..31). 110 * If the bit is set, all rules with the corresponding set 111 * are disabled. Set RESVD_SET(31) is reserved for the default rule 112 * and rules that are not deleted by the flush command, 113 * and CANNOT be disabled. 114 * Rules in set RESVD_SET can only be deleted explicitly. 115 */ 116static VNET_DEFINE(u_int32_t, set_disable); 117static VNET_DEFINE(int, fw_verbose); 118static VNET_DEFINE(struct callout, ipfw_timeout); 119static VNET_DEFINE(int, verbose_limit); 120 121#define V_set_disable VNET(set_disable) 122#define V_fw_verbose VNET(fw_verbose) 123#define V_ipfw_timeout VNET(ipfw_timeout) 124#define V_verbose_limit VNET(verbose_limit) 125 126#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 127static int default_to_accept = 1; 128#else 129static int default_to_accept; 130#endif 131static uma_zone_t ipfw_dyn_rule_zone; 132 133struct ip_fw *ip_fw_default_rule; 134 135/* 136 * list of rules for layer 3 137 */ 138VNET_DEFINE(struct ip_fw_chain, layer3_chain); 139 140MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 141MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 142#define IPFW_NAT_LOADED (ipfw_nat_ptr != NULL) 143ipfw_nat_t *ipfw_nat_ptr = NULL; 144ipfw_nat_cfg_t *ipfw_nat_cfg_ptr; 145ipfw_nat_cfg_t *ipfw_nat_del_ptr; 146ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; 147ipfw_nat_cfg_t *ipfw_nat_get_log_ptr; 148 149struct table_entry { 150 struct radix_node rn[2]; 151 struct sockaddr_in addr, mask; 152 u_int32_t value; 153}; 154 155static VNET_DEFINE(int, autoinc_step); 156#define V_autoinc_step VNET(autoinc_step) 157static VNET_DEFINE(int, fw_deny_unknown_exthdrs); 158#define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs) 159 160extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS); 161 162#ifdef SYSCTL_NODE 163SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 164SYSCTL_VNET_PROC(_net_inet_ip_fw, OID_AUTO, enable, 165 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_enable), 0, 166 ipfw_chg_hook, "I", "Enable ipfw"); 167SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, 168 CTLFLAG_RW, &VNET_NAME(autoinc_step), 0, 169 "Rule number auto-increment step"); 170SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 171 CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0, 172 "Only do a single pass through ipfw when using dummynet(4)"); 173SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose, 174 CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0, 175 "Log matches to ipfw rules"); 176SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, 177 CTLFLAG_RW, &VNET_NAME(verbose_limit), 0, 178 "Set upper limit of matches of ipfw rules logged"); 179SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, 180 NULL, IPFW_DEFAULT_RULE, 181 "The default/max possible rule number."); 182SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, tables_max, CTLFLAG_RD, 183 NULL, IPFW_TABLES_MAX, 184 "The maximum number of tables."); 185SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN, 186 &default_to_accept, 0, 187 "Make the default rule accept all packets."); 188TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept); 189#ifdef INET6 190SYSCTL_DECL(_net_inet6_ip6); 191SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 192SYSCTL_VNET_PROC(_net_inet6_ip6_fw, OID_AUTO, enable, 193 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw6_enable), 0, 194 ipfw_chg_hook, "I", "Enable ipfw+6"); 195SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs, 196 CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0, 197 "Deny packets with unknown IPv6 Extension Headers"); 198#endif 199#endif 200 201/* 202 * Description of dynamic rules. 203 * 204 * Dynamic rules are stored in lists accessed through a hash table 205 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 206 * be modified through the sysctl variable dyn_buckets which is 207 * updated when the table becomes empty. 208 * 209 * XXX currently there is only one list, ipfw_dyn. 210 * 211 * When a packet is received, its address fields are first masked 212 * with the mask defined for the rule, then hashed, then matched 213 * against the entries in the corresponding list. 214 * Dynamic rules can be used for different purposes: 215 * + stateful rules; 216 * + enforcing limits on the number of sessions; 217 * + in-kernel NAT (not implemented yet) 218 * 219 * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 220 * measured in seconds and depending on the flags. 221 * 222 * The total number of dynamic rules is stored in dyn_count. 223 * The max number of dynamic rules is dyn_max. When we reach 224 * the maximum number of rules we do not create anymore. This is 225 * done to avoid consuming too much memory, but also too much 226 * time when searching on each packet (ideally, we should try instead 227 * to put a limit on the length of the list on each bucket...). 228 * 229 * Each dynamic rule holds a pointer to the parent ipfw rule so 230 * we know what action to perform. Dynamic rules are removed when 231 * the parent rule is deleted. XXX we should make them survive. 232 * 233 * There are some limitations with dynamic rules -- we do not 234 * obey the 'randomized match', and we do not do multiple 235 * passes through the firewall. XXX check the latter!!! 236 */ 237static VNET_DEFINE(ipfw_dyn_rule **, ipfw_dyn_v); 238static VNET_DEFINE(u_int32_t, dyn_buckets); 239static VNET_DEFINE(u_int32_t, curr_dyn_buckets); 240 241#define V_ipfw_dyn_v VNET(ipfw_dyn_v) 242#define V_dyn_buckets VNET(dyn_buckets) 243#define V_curr_dyn_buckets VNET(curr_dyn_buckets) 244 245static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 246#define IPFW_DYN_LOCK_INIT() \ 247 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 248#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 249#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 250#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 251#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 252 253static struct mbuf *send_pkt(struct mbuf *, struct ipfw_flow_id *, 254 u_int32_t, u_int32_t, int); 255 256 257/* 258 * Timeouts for various events in handing dynamic rules. 259 */ 260static VNET_DEFINE(u_int32_t, dyn_ack_lifetime); 261static VNET_DEFINE(u_int32_t, dyn_syn_lifetime); 262static VNET_DEFINE(u_int32_t, dyn_fin_lifetime); 263static VNET_DEFINE(u_int32_t, dyn_rst_lifetime); 264static VNET_DEFINE(u_int32_t, dyn_udp_lifetime); 265static VNET_DEFINE(u_int32_t, dyn_short_lifetime); 266 267#define V_dyn_ack_lifetime VNET(dyn_ack_lifetime) 268#define V_dyn_syn_lifetime VNET(dyn_syn_lifetime) 269#define V_dyn_fin_lifetime VNET(dyn_fin_lifetime) 270#define V_dyn_rst_lifetime VNET(dyn_rst_lifetime) 271#define V_dyn_udp_lifetime VNET(dyn_udp_lifetime) 272#define V_dyn_short_lifetime VNET(dyn_short_lifetime) 273 274/* 275 * Keepalives are sent if dyn_keepalive is set. They are sent every 276 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 277 * seconds of lifetime of a rule. 278 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 279 * than dyn_keepalive_period. 280 */ 281 282static VNET_DEFINE(u_int32_t, dyn_keepalive_interval); 283static VNET_DEFINE(u_int32_t, dyn_keepalive_period); 284static VNET_DEFINE(u_int32_t, dyn_keepalive); 285 286#define V_dyn_keepalive_interval VNET(dyn_keepalive_interval) 287#define V_dyn_keepalive_period VNET(dyn_keepalive_period) 288#define V_dyn_keepalive VNET(dyn_keepalive) 289 290static VNET_DEFINE(u_int32_t, static_count); /* # of static rules */ 291static VNET_DEFINE(u_int32_t, static_len); /* bytes of static rules */ 292static VNET_DEFINE(u_int32_t, dyn_count); /* # of dynamic rules */ 293static VNET_DEFINE(u_int32_t, dyn_max); /* max # of dynamic rules */ 294 295#define V_static_count VNET(static_count) 296#define V_static_len VNET(static_len) 297#define V_dyn_count VNET(dyn_count) 298#define V_dyn_max VNET(dyn_max) 299 300#ifdef SYSCTL_NODE 301SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, 302 CTLFLAG_RW, &VNET_NAME(dyn_buckets), 0, 303 "Number of dyn. buckets"); 304SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, 305 CTLFLAG_RD, &VNET_NAME(curr_dyn_buckets), 0, 306 "Current Number of dyn. buckets"); 307SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, 308 CTLFLAG_RD, &VNET_NAME(dyn_count), 0, 309 "Number of dyn. rules"); 310SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, 311 CTLFLAG_RW, &VNET_NAME(dyn_max), 0, 312 "Max number of dyn. rules"); 313SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, static_count, 314 CTLFLAG_RD, &VNET_NAME(static_count), 0, 315 "Number of static rules"); 316SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, 317 CTLFLAG_RW, &VNET_NAME(dyn_ack_lifetime), 0, 318 "Lifetime of dyn. rules for acks"); 319SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, 320 CTLFLAG_RW, &VNET_NAME(dyn_syn_lifetime), 0, 321 "Lifetime of dyn. rules for syn"); 322SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, 323 CTLFLAG_RW, &VNET_NAME(dyn_fin_lifetime), 0, 324 "Lifetime of dyn. rules for fin"); 325SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, 326 CTLFLAG_RW, &VNET_NAME(dyn_rst_lifetime), 0, 327 "Lifetime of dyn. rules for rst"); 328SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, 329 CTLFLAG_RW, &VNET_NAME(dyn_udp_lifetime), 0, 330 "Lifetime of dyn. rules for UDP"); 331SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, 332 CTLFLAG_RW, &VNET_NAME(dyn_short_lifetime), 0, 333 "Lifetime of dyn. rules for other situations"); 334SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, 335 CTLFLAG_RW, &VNET_NAME(dyn_keepalive), 0, 336 "Enable keepalives for dyn. rules"); 337#endif /* SYSCTL_NODE */ 338 339/* 340 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 341 * Other macros just cast void * into the appropriate type 342 */ 343#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 344#define TCP(p) ((struct tcphdr *)(p)) 345#define SCTP(p) ((struct sctphdr *)(p)) 346#define UDP(p) ((struct udphdr *)(p)) 347#define ICMP(p) ((struct icmphdr *)(p)) 348#define ICMP6(p) ((struct icmp6_hdr *)(p)) 349 350static __inline int 351icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 352{ 353 int type = icmp->icmp_type; 354 355 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 356} 357 358#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 359 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 360 361static int 362is_icmp_query(struct icmphdr *icmp) 363{ 364 int type = icmp->icmp_type; 365 366 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 367} 368#undef TT 369 370/* 371 * The following checks use two arrays of 8 or 16 bits to store the 372 * bits that we want set or clear, respectively. They are in the 373 * low and high half of cmd->arg1 or cmd->d[0]. 374 * 375 * We scan options and store the bits we find set. We succeed if 376 * 377 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 378 * 379 * The code is sometimes optimized not to store additional variables. 380 */ 381 382static int 383flags_match(ipfw_insn *cmd, u_int8_t bits) 384{ 385 u_char want_clear; 386 bits = ~bits; 387 388 if ( ((cmd->arg1 & 0xff) & bits) != 0) 389 return 0; /* some bits we want set were clear */ 390 want_clear = (cmd->arg1 >> 8) & 0xff; 391 if ( (want_clear & bits) != want_clear) 392 return 0; /* some bits we want clear were set */ 393 return 1; 394} 395 396static int 397ipopts_match(struct ip *ip, ipfw_insn *cmd) 398{ 399 int optlen, bits = 0; 400 u_char *cp = (u_char *)(ip + 1); 401 int x = (ip->ip_hl << 2) - sizeof (struct ip); 402 403 for (; x > 0; x -= optlen, cp += optlen) { 404 int opt = cp[IPOPT_OPTVAL]; 405 406 if (opt == IPOPT_EOL) 407 break; 408 if (opt == IPOPT_NOP) 409 optlen = 1; 410 else { 411 optlen = cp[IPOPT_OLEN]; 412 if (optlen <= 0 || optlen > x) 413 return 0; /* invalid or truncated */ 414 } 415 switch (opt) { 416 417 default: 418 break; 419 420 case IPOPT_LSRR: 421 bits |= IP_FW_IPOPT_LSRR; 422 break; 423 424 case IPOPT_SSRR: 425 bits |= IP_FW_IPOPT_SSRR; 426 break; 427 428 case IPOPT_RR: 429 bits |= IP_FW_IPOPT_RR; 430 break; 431 432 case IPOPT_TS: 433 bits |= IP_FW_IPOPT_TS; 434 break; 435 } 436 } 437 return (flags_match(cmd, bits)); 438} 439 440static int 441tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 442{ 443 int optlen, bits = 0; 444 u_char *cp = (u_char *)(tcp + 1); 445 int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 446 447 for (; x > 0; x -= optlen, cp += optlen) { 448 int opt = cp[0]; 449 if (opt == TCPOPT_EOL) 450 break; 451 if (opt == TCPOPT_NOP) 452 optlen = 1; 453 else { 454 optlen = cp[1]; 455 if (optlen <= 0) 456 break; 457 } 458 459 switch (opt) { 460 461 default: 462 break; 463 464 case TCPOPT_MAXSEG: 465 bits |= IP_FW_TCPOPT_MSS; 466 break; 467 468 case TCPOPT_WINDOW: 469 bits |= IP_FW_TCPOPT_WINDOW; 470 break; 471 472 case TCPOPT_SACK_PERMITTED: 473 case TCPOPT_SACK: 474 bits |= IP_FW_TCPOPT_SACK; 475 break; 476 477 case TCPOPT_TIMESTAMP: 478 bits |= IP_FW_TCPOPT_TS; 479 break; 480 481 } 482 } 483 return (flags_match(cmd, bits)); 484} 485 486static int 487iface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 488{ 489 if (ifp == NULL) /* no iface with this packet, match fails */ 490 return 0; 491 /* Check by name or by IP address */ 492 if (cmd->name[0] != '\0') { /* match by name */ 493 /* Check name */ 494 if (cmd->p.glob) { 495 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 496 return(1); 497 } else { 498 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 499 return(1); 500 } 501 } else { 502 struct ifaddr *ia; 503 504 if_addr_rlock(ifp); 505 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 506 if (ia->ifa_addr->sa_family != AF_INET) 507 continue; 508 if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 509 (ia->ifa_addr))->sin_addr.s_addr) { 510 if_addr_runlock(ifp); 511 return(1); /* match */ 512 } 513 } 514 if_addr_runlock(ifp); 515 } 516 return(0); /* no match, fail ... */ 517} 518 519/* 520 * The verify_path function checks if a route to the src exists and 521 * if it is reachable via ifp (when provided). 522 * 523 * The 'verrevpath' option checks that the interface that an IP packet 524 * arrives on is the same interface that traffic destined for the 525 * packet's source address would be routed out of. The 'versrcreach' 526 * option just checks that the source address is reachable via any route 527 * (except default) in the routing table. These two are a measure to block 528 * forged packets. This is also commonly known as "anti-spoofing" or Unicast 529 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 530 * is purposely reminiscent of the Cisco IOS command, 531 * 532 * ip verify unicast reverse-path 533 * ip verify unicast source reachable-via any 534 * 535 * which implements the same functionality. But note that syntax is 536 * misleading. The check may be performed on all IP packets whether unicast, 537 * multicast, or broadcast. 538 */ 539static int 540verify_path(struct in_addr src, struct ifnet *ifp, u_int fib) 541{ 542 struct route ro; 543 struct sockaddr_in *dst; 544 545 bzero(&ro, sizeof(ro)); 546 547 dst = (struct sockaddr_in *)&(ro.ro_dst); 548 dst->sin_family = AF_INET; 549 dst->sin_len = sizeof(*dst); 550 dst->sin_addr = src; 551 in_rtalloc_ign(&ro, 0, fib); 552 553 if (ro.ro_rt == NULL) 554 return 0; 555 556 /* 557 * If ifp is provided, check for equality with rtentry. 558 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 559 * in order to pass packets injected back by if_simloop(): 560 * if useloopback == 1 routing entry (via lo0) for our own address 561 * may exist, so we need to handle routing assymetry. 562 */ 563 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 564 RTFREE(ro.ro_rt); 565 return 0; 566 } 567 568 /* if no ifp provided, check if rtentry is not default route */ 569 if (ifp == NULL && 570 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 571 RTFREE(ro.ro_rt); 572 return 0; 573 } 574 575 /* or if this is a blackhole/reject route */ 576 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 577 RTFREE(ro.ro_rt); 578 return 0; 579 } 580 581 /* found valid route */ 582 RTFREE(ro.ro_rt); 583 return 1; 584} 585 586#ifdef INET6 587/* 588 * ipv6 specific rules here... 589 */ 590static __inline int 591icmp6type_match (int type, ipfw_insn_u32 *cmd) 592{ 593 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 594} 595 596static int 597flow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 598{ 599 int i; 600 for (i=0; i <= cmd->o.arg1; ++i ) 601 if (curr_flow == cmd->d[i] ) 602 return 1; 603 return 0; 604} 605 606/* support for IP6_*_ME opcodes */ 607static int 608search_ip6_addr_net (struct in6_addr * ip6_addr) 609{ 610 struct ifnet *mdc; 611 struct ifaddr *mdc2; 612 struct in6_ifaddr *fdm; 613 struct in6_addr copia; 614 615 TAILQ_FOREACH(mdc, &V_ifnet, if_link) { 616 if_addr_rlock(mdc); 617 TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) { 618 if (mdc2->ifa_addr->sa_family == AF_INET6) { 619 fdm = (struct in6_ifaddr *)mdc2; 620 copia = fdm->ia_addr.sin6_addr; 621 /* need for leaving scope_id in the sock_addr */ 622 in6_clearscope(&copia); 623 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) { 624 if_addr_runlock(mdc); 625 return 1; 626 } 627 } 628 } 629 if_addr_runlock(mdc); 630 } 631 return 0; 632} 633 634static int 635verify_path6(struct in6_addr *src, struct ifnet *ifp) 636{ 637 struct route_in6 ro; 638 struct sockaddr_in6 *dst; 639 640 bzero(&ro, sizeof(ro)); 641 642 dst = (struct sockaddr_in6 * )&(ro.ro_dst); 643 dst->sin6_family = AF_INET6; 644 dst->sin6_len = sizeof(*dst); 645 dst->sin6_addr = *src; 646 /* XXX MRT 0 for ipv6 at this time */ 647 rtalloc_ign((struct route *)&ro, 0); 648 649 if (ro.ro_rt == NULL) 650 return 0; 651 652 /* 653 * if ifp is provided, check for equality with rtentry 654 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 655 * to support the case of sending packets to an address of our own. 656 * (where the former interface is the first argument of if_simloop() 657 * (=ifp), the latter is lo0) 658 */ 659 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 660 RTFREE(ro.ro_rt); 661 return 0; 662 } 663 664 /* if no ifp provided, check if rtentry is not default route */ 665 if (ifp == NULL && 666 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 667 RTFREE(ro.ro_rt); 668 return 0; 669 } 670 671 /* or if this is a blackhole/reject route */ 672 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 673 RTFREE(ro.ro_rt); 674 return 0; 675 } 676 677 /* found valid route */ 678 RTFREE(ro.ro_rt); 679 return 1; 680 681} 682static __inline int 683hash_packet6(struct ipfw_flow_id *id) 684{ 685 u_int32_t i; 686 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 687 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 688 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 689 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 690 (id->dst_port) ^ (id->src_port); 691 return i; 692} 693 694static int 695is_icmp6_query(int icmp6_type) 696{ 697 if ((icmp6_type <= ICMP6_MAXTYPE) && 698 (icmp6_type == ICMP6_ECHO_REQUEST || 699 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 700 icmp6_type == ICMP6_WRUREQUEST || 701 icmp6_type == ICMP6_FQDN_QUERY || 702 icmp6_type == ICMP6_NI_QUERY)) 703 return (1); 704 705 return (0); 706} 707 708static void 709send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 710{ 711 struct mbuf *m; 712 713 m = args->m; 714 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 715 struct tcphdr *tcp; 716 tcp = (struct tcphdr *)((char *)ip6 + hlen); 717 718 if ((tcp->th_flags & TH_RST) == 0) { 719 struct mbuf *m0; 720 m0 = send_pkt(args->m, &(args->f_id), 721 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 722 tcp->th_flags | TH_RST); 723 if (m0 != NULL) 724 ip6_output(m0, NULL, NULL, 0, NULL, NULL, 725 NULL); 726 } 727 m_freem(m); 728 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 729#if 0 730 /* 731 * Unlike above, the mbufs need to line up with the ip6 hdr, 732 * as the contents are read. We need to m_adj() the 733 * needed amount. 734 * The mbuf will however be thrown away so we can adjust it. 735 * Remember we did an m_pullup on it already so we 736 * can make some assumptions about contiguousness. 737 */ 738 if (args->L3offset) 739 m_adj(m, args->L3offset); 740#endif 741 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 742 } else 743 m_freem(m); 744 745 args->m = NULL; 746} 747 748#endif /* INET6 */ 749 750/* counter for ipfw_log(NULL...) */ 751static VNET_DEFINE(u_int64_t, norule_counter); 752#define V_norule_counter VNET(norule_counter) 753 754#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 755#define SNP(buf) buf, sizeof(buf) 756 757/* 758 * We enter here when we have a rule with O_LOG. 759 * XXX this function alone takes about 2Kbytes of code! 760 */ 761static void 762ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 763 struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg, 764 struct ip *ip) 765{ 766 struct ether_header *eh = args->eh; 767 char *action; 768 int limit_reached = 0; 769 char action2[40], proto[128], fragment[32]; 770 771 fragment[0] = '\0'; 772 proto[0] = '\0'; 773 774 if (f == NULL) { /* bogus pkt */ 775 if (V_verbose_limit != 0 && V_norule_counter >= V_verbose_limit) 776 return; 777 V_norule_counter++; 778 if (V_norule_counter == V_verbose_limit) 779 limit_reached = V_verbose_limit; 780 action = "Refuse"; 781 } else { /* O_LOG is the first action, find the real one */ 782 ipfw_insn *cmd = ACTION_PTR(f); 783 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 784 785 if (l->max_log != 0 && l->log_left == 0) 786 return; 787 l->log_left--; 788 if (l->log_left == 0) 789 limit_reached = l->max_log; 790 cmd += F_LEN(cmd); /* point to first action */ 791 if (cmd->opcode == O_ALTQ) { 792 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 793 794 snprintf(SNPARGS(action2, 0), "Altq %d", 795 altq->qid); 796 cmd += F_LEN(cmd); 797 } 798 if (cmd->opcode == O_PROB) 799 cmd += F_LEN(cmd); 800 801 if (cmd->opcode == O_TAG) 802 cmd += F_LEN(cmd); 803 804 action = action2; 805 switch (cmd->opcode) { 806 case O_DENY: 807 action = "Deny"; 808 break; 809 810 case O_REJECT: 811 if (cmd->arg1==ICMP_REJECT_RST) 812 action = "Reset"; 813 else if (cmd->arg1==ICMP_UNREACH_HOST) 814 action = "Reject"; 815 else 816 snprintf(SNPARGS(action2, 0), "Unreach %d", 817 cmd->arg1); 818 break; 819 820 case O_UNREACH6: 821 if (cmd->arg1==ICMP6_UNREACH_RST) 822 action = "Reset"; 823 else 824 snprintf(SNPARGS(action2, 0), "Unreach %d", 825 cmd->arg1); 826 break; 827 828 case O_ACCEPT: 829 action = "Accept"; 830 break; 831 case O_COUNT: 832 action = "Count"; 833 break; 834 case O_DIVERT: 835 snprintf(SNPARGS(action2, 0), "Divert %d", 836 cmd->arg1); 837 break; 838 case O_TEE: 839 snprintf(SNPARGS(action2, 0), "Tee %d", 840 cmd->arg1); 841 break; 842 case O_SETFIB: 843 snprintf(SNPARGS(action2, 0), "SetFib %d", 844 cmd->arg1); 845 break; 846 case O_SKIPTO: 847 snprintf(SNPARGS(action2, 0), "SkipTo %d", 848 cmd->arg1); 849 break; 850 case O_PIPE: 851 snprintf(SNPARGS(action2, 0), "Pipe %d", 852 cmd->arg1); 853 break; 854 case O_QUEUE: 855 snprintf(SNPARGS(action2, 0), "Queue %d", 856 cmd->arg1); 857 break; 858 case O_FORWARD_IP: { 859 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 860 int len; 861 struct in_addr dummyaddr; 862 if (sa->sa.sin_addr.s_addr == INADDR_ANY) 863 dummyaddr.s_addr = htonl(tablearg); 864 else 865 dummyaddr.s_addr = sa->sa.sin_addr.s_addr; 866 867 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 868 inet_ntoa(dummyaddr)); 869 870 if (sa->sa.sin_port) 871 snprintf(SNPARGS(action2, len), ":%d", 872 sa->sa.sin_port); 873 } 874 break; 875 case O_NETGRAPH: 876 snprintf(SNPARGS(action2, 0), "Netgraph %d", 877 cmd->arg1); 878 break; 879 case O_NGTEE: 880 snprintf(SNPARGS(action2, 0), "Ngtee %d", 881 cmd->arg1); 882 break; 883 case O_NAT: 884 action = "Nat"; 885 break; 886 case O_REASS: 887 action = "Reass"; 888 break; 889 default: 890 action = "UNKNOWN"; 891 break; 892 } 893 } 894 895 if (hlen == 0) { /* non-ip */ 896 snprintf(SNPARGS(proto, 0), "MAC"); 897 898 } else { 899 int len; 900 char src[48], dst[48]; 901 struct icmphdr *icmp; 902 struct tcphdr *tcp; 903 struct udphdr *udp; 904#ifdef INET6 905 struct ip6_hdr *ip6 = NULL; 906 struct icmp6_hdr *icmp6; 907#endif 908 src[0] = '\0'; 909 dst[0] = '\0'; 910#ifdef INET6 911 if (IS_IP6_FLOW_ID(&(args->f_id))) { 912 char ip6buf[INET6_ADDRSTRLEN]; 913 snprintf(src, sizeof(src), "[%s]", 914 ip6_sprintf(ip6buf, &args->f_id.src_ip6)); 915 snprintf(dst, sizeof(dst), "[%s]", 916 ip6_sprintf(ip6buf, &args->f_id.dst_ip6)); 917 918 ip6 = (struct ip6_hdr *)ip; 919 tcp = (struct tcphdr *)(((char *)ip) + hlen); 920 udp = (struct udphdr *)(((char *)ip) + hlen); 921 } else 922#endif 923 { 924 tcp = L3HDR(struct tcphdr, ip); 925 udp = L3HDR(struct udphdr, ip); 926 927 inet_ntoa_r(ip->ip_src, src); 928 inet_ntoa_r(ip->ip_dst, dst); 929 } 930 931 switch (args->f_id.proto) { 932 case IPPROTO_TCP: 933 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 934 if (offset == 0) 935 snprintf(SNPARGS(proto, len), ":%d %s:%d", 936 ntohs(tcp->th_sport), 937 dst, 938 ntohs(tcp->th_dport)); 939 else 940 snprintf(SNPARGS(proto, len), " %s", dst); 941 break; 942 943 case IPPROTO_UDP: 944 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 945 if (offset == 0) 946 snprintf(SNPARGS(proto, len), ":%d %s:%d", 947 ntohs(udp->uh_sport), 948 dst, 949 ntohs(udp->uh_dport)); 950 else 951 snprintf(SNPARGS(proto, len), " %s", dst); 952 break; 953 954 case IPPROTO_ICMP: 955 icmp = L3HDR(struct icmphdr, ip); 956 if (offset == 0) 957 len = snprintf(SNPARGS(proto, 0), 958 "ICMP:%u.%u ", 959 icmp->icmp_type, icmp->icmp_code); 960 else 961 len = snprintf(SNPARGS(proto, 0), "ICMP "); 962 len += snprintf(SNPARGS(proto, len), "%s", src); 963 snprintf(SNPARGS(proto, len), " %s", dst); 964 break; 965#ifdef INET6 966 case IPPROTO_ICMPV6: 967 icmp6 = (struct icmp6_hdr *)(((char *)ip) + hlen); 968 if (offset == 0) 969 len = snprintf(SNPARGS(proto, 0), 970 "ICMPv6:%u.%u ", 971 icmp6->icmp6_type, icmp6->icmp6_code); 972 else 973 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 974 len += snprintf(SNPARGS(proto, len), "%s", src); 975 snprintf(SNPARGS(proto, len), " %s", dst); 976 break; 977#endif 978 default: 979 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 980 args->f_id.proto, src); 981 snprintf(SNPARGS(proto, len), " %s", dst); 982 break; 983 } 984 985#ifdef INET6 986 if (IS_IP6_FLOW_ID(&(args->f_id))) { 987 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 988 snprintf(SNPARGS(fragment, 0), 989 " (frag %08x:%d@%d%s)", 990 args->f_id.frag_id6, 991 ntohs(ip6->ip6_plen) - hlen, 992 ntohs(offset & IP6F_OFF_MASK) << 3, 993 (offset & IP6F_MORE_FRAG) ? "+" : ""); 994 } else 995#endif 996 { 997 int ip_off, ip_len; 998 if (eh != NULL) { /* layer 2 packets are as on the wire */ 999 ip_off = ntohs(ip->ip_off); 1000 ip_len = ntohs(ip->ip_len); 1001 } else { 1002 ip_off = ip->ip_off; 1003 ip_len = ip->ip_len; 1004 } 1005 if (ip_off & (IP_MF | IP_OFFMASK)) 1006 snprintf(SNPARGS(fragment, 0), 1007 " (frag %d:%d@%d%s)", 1008 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 1009 offset << 3, 1010 (ip_off & IP_MF) ? "+" : ""); 1011 } 1012 } 1013 if (oif || m->m_pkthdr.rcvif) 1014 log(LOG_SECURITY | LOG_INFO, 1015 "ipfw: %d %s %s %s via %s%s\n", 1016 f ? f->rulenum : -1, 1017 action, proto, oif ? "out" : "in", 1018 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 1019 fragment); 1020 else 1021 log(LOG_SECURITY | LOG_INFO, 1022 "ipfw: %d %s %s [no if info]%s\n", 1023 f ? f->rulenum : -1, 1024 action, proto, fragment); 1025 if (limit_reached) 1026 log(LOG_SECURITY | LOG_NOTICE, 1027 "ipfw: limit %d reached on entry %d\n", 1028 limit_reached, f ? f->rulenum : -1); 1029} 1030 1031/* 1032 * IMPORTANT: the hash function for dynamic rules must be commutative 1033 * in source and destination (ip,port), because rules are bidirectional 1034 * and we want to find both in the same bucket. 1035 */ 1036static __inline int 1037hash_packet(struct ipfw_flow_id *id) 1038{ 1039 u_int32_t i; 1040 1041#ifdef INET6 1042 if (IS_IP6_FLOW_ID(id)) 1043 i = hash_packet6(id); 1044 else 1045#endif /* INET6 */ 1046 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1047 i &= (V_curr_dyn_buckets - 1); 1048 return i; 1049} 1050 1051/** 1052 * unlink a dynamic rule from a chain. prev is a pointer to 1053 * the previous one, q is a pointer to the rule to delete, 1054 * head is a pointer to the head of the queue. 1055 * Modifies q and potentially also head. 1056 */ 1057#define UNLINK_DYN_RULE(prev, head, q) { \ 1058 ipfw_dyn_rule *old_q = q; \ 1059 \ 1060 /* remove a refcount to the parent */ \ 1061 if (q->dyn_type == O_LIMIT) \ 1062 q->parent->count--; \ 1063 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1064 (q->id.src_ip), (q->id.src_port), \ 1065 (q->id.dst_ip), (q->id.dst_port), V_dyn_count-1 ); ) \ 1066 if (prev != NULL) \ 1067 prev->next = q = q->next; \ 1068 else \ 1069 head = q = q->next; \ 1070 V_dyn_count--; \ 1071 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1072 1073#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1074 1075/** 1076 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1077 * 1078 * If keep_me == NULL, rules are deleted even if not expired, 1079 * otherwise only expired rules are removed. 1080 * 1081 * The value of the second parameter is also used to point to identify 1082 * a rule we absolutely do not want to remove (e.g. because we are 1083 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1084 * rules). The pointer is only used for comparison, so any non-null 1085 * value will do. 1086 */ 1087static void 1088remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1089{ 1090 static u_int32_t last_remove = 0; 1091 1092#define FORCE (keep_me == NULL) 1093 1094 ipfw_dyn_rule *prev, *q; 1095 int i, pass = 0, max_pass = 0; 1096 1097 IPFW_DYN_LOCK_ASSERT(); 1098 1099 if (V_ipfw_dyn_v == NULL || V_dyn_count == 0) 1100 return; 1101 /* do not expire more than once per second, it is useless */ 1102 if (!FORCE && last_remove == time_uptime) 1103 return; 1104 last_remove = time_uptime; 1105 1106 /* 1107 * because O_LIMIT refer to parent rules, during the first pass only 1108 * remove child and mark any pending LIMIT_PARENT, and remove 1109 * them in a second pass. 1110 */ 1111next_pass: 1112 for (i = 0 ; i < V_curr_dyn_buckets ; i++) { 1113 for (prev=NULL, q = V_ipfw_dyn_v[i] ; q ; ) { 1114 /* 1115 * Logic can become complex here, so we split tests. 1116 */ 1117 if (q == keep_me) 1118 goto next; 1119 if (rule != NULL && rule != q->rule) 1120 goto next; /* not the one we are looking for */ 1121 if (q->dyn_type == O_LIMIT_PARENT) { 1122 /* 1123 * handle parent in the second pass, 1124 * record we need one. 1125 */ 1126 max_pass = 1; 1127 if (pass == 0) 1128 goto next; 1129 if (FORCE && q->count != 0 ) { 1130 /* XXX should not happen! */ 1131 printf("ipfw: OUCH! cannot remove rule," 1132 " count %d\n", q->count); 1133 } 1134 } else { 1135 if (!FORCE && 1136 !TIME_LEQ( q->expire, time_uptime )) 1137 goto next; 1138 } 1139 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1140 UNLINK_DYN_RULE(prev, V_ipfw_dyn_v[i], q); 1141 continue; 1142 } 1143next: 1144 prev=q; 1145 q=q->next; 1146 } 1147 } 1148 if (pass++ < max_pass) 1149 goto next_pass; 1150} 1151 1152 1153/** 1154 * lookup a dynamic rule. 1155 */ 1156static ipfw_dyn_rule * 1157lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1158 struct tcphdr *tcp) 1159{ 1160 /* 1161 * stateful ipfw extensions. 1162 * Lookup into dynamic session queue 1163 */ 1164#define MATCH_REVERSE 0 1165#define MATCH_FORWARD 1 1166#define MATCH_NONE 2 1167#define MATCH_UNKNOWN 3 1168 int i, dir = MATCH_NONE; 1169 ipfw_dyn_rule *prev, *q=NULL; 1170 1171 IPFW_DYN_LOCK_ASSERT(); 1172 1173 if (V_ipfw_dyn_v == NULL) 1174 goto done; /* not found */ 1175 i = hash_packet( pkt ); 1176 for (prev=NULL, q = V_ipfw_dyn_v[i] ; q != NULL ; ) { 1177 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1178 goto next; 1179 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1180 UNLINK_DYN_RULE(prev, V_ipfw_dyn_v[i], q); 1181 continue; 1182 } 1183 if (pkt->proto == q->id.proto && 1184 q->dyn_type != O_LIMIT_PARENT) { 1185 if (IS_IP6_FLOW_ID(pkt)) { 1186 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1187 &(q->id.src_ip6)) && 1188 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1189 &(q->id.dst_ip6)) && 1190 pkt->src_port == q->id.src_port && 1191 pkt->dst_port == q->id.dst_port ) { 1192 dir = MATCH_FORWARD; 1193 break; 1194 } 1195 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1196 &(q->id.dst_ip6)) && 1197 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1198 &(q->id.src_ip6)) && 1199 pkt->src_port == q->id.dst_port && 1200 pkt->dst_port == q->id.src_port ) { 1201 dir = MATCH_REVERSE; 1202 break; 1203 } 1204 } else { 1205 if (pkt->src_ip == q->id.src_ip && 1206 pkt->dst_ip == q->id.dst_ip && 1207 pkt->src_port == q->id.src_port && 1208 pkt->dst_port == q->id.dst_port ) { 1209 dir = MATCH_FORWARD; 1210 break; 1211 } 1212 if (pkt->src_ip == q->id.dst_ip && 1213 pkt->dst_ip == q->id.src_ip && 1214 pkt->src_port == q->id.dst_port && 1215 pkt->dst_port == q->id.src_port ) { 1216 dir = MATCH_REVERSE; 1217 break; 1218 } 1219 } 1220 } 1221next: 1222 prev = q; 1223 q = q->next; 1224 } 1225 if (q == NULL) 1226 goto done; /* q = NULL, not found */ 1227 1228 if ( prev != NULL) { /* found and not in front */ 1229 prev->next = q->next; 1230 q->next = V_ipfw_dyn_v[i]; 1231 V_ipfw_dyn_v[i] = q; 1232 } 1233 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1234 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1235 1236#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1237#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1238 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1239 switch (q->state) { 1240 case TH_SYN: /* opening */ 1241 q->expire = time_uptime + V_dyn_syn_lifetime; 1242 break; 1243 1244 case BOTH_SYN: /* move to established */ 1245 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1246 case BOTH_SYN | (TH_FIN << 8) : 1247 if (tcp) { 1248#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1249 u_int32_t ack = ntohl(tcp->th_ack); 1250 if (dir == MATCH_FORWARD) { 1251 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1252 q->ack_fwd = ack; 1253 else { /* ignore out-of-sequence */ 1254 break; 1255 } 1256 } else { 1257 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1258 q->ack_rev = ack; 1259 else { /* ignore out-of-sequence */ 1260 break; 1261 } 1262 } 1263 } 1264 q->expire = time_uptime + V_dyn_ack_lifetime; 1265 break; 1266 1267 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1268 if (V_dyn_fin_lifetime >= V_dyn_keepalive_period) 1269 V_dyn_fin_lifetime = V_dyn_keepalive_period - 1; 1270 q->expire = time_uptime + V_dyn_fin_lifetime; 1271 break; 1272 1273 default: 1274#if 0 1275 /* 1276 * reset or some invalid combination, but can also 1277 * occur if we use keep-state the wrong way. 1278 */ 1279 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1280 printf("invalid state: 0x%x\n", q->state); 1281#endif 1282 if (V_dyn_rst_lifetime >= V_dyn_keepalive_period) 1283 V_dyn_rst_lifetime = V_dyn_keepalive_period - 1; 1284 q->expire = time_uptime + V_dyn_rst_lifetime; 1285 break; 1286 } 1287 } else if (pkt->proto == IPPROTO_UDP) { 1288 q->expire = time_uptime + V_dyn_udp_lifetime; 1289 } else { 1290 /* other protocols */ 1291 q->expire = time_uptime + V_dyn_short_lifetime; 1292 } 1293done: 1294 if (match_direction) 1295 *match_direction = dir; 1296 return q; 1297} 1298 1299static ipfw_dyn_rule * 1300lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1301 struct tcphdr *tcp) 1302{ 1303 ipfw_dyn_rule *q; 1304 1305 IPFW_DYN_LOCK(); 1306 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1307 if (q == NULL) 1308 IPFW_DYN_UNLOCK(); 1309 /* NB: return table locked when q is not NULL */ 1310 return q; 1311} 1312 1313static void 1314realloc_dynamic_table(void) 1315{ 1316 IPFW_DYN_LOCK_ASSERT(); 1317 1318 /* 1319 * Try reallocation, make sure we have a power of 2 and do 1320 * not allow more than 64k entries. In case of overflow, 1321 * default to 1024. 1322 */ 1323 1324 if (V_dyn_buckets > 65536) 1325 V_dyn_buckets = 1024; 1326 if ((V_dyn_buckets & (V_dyn_buckets-1)) != 0) { /* not a power of 2 */ 1327 V_dyn_buckets = V_curr_dyn_buckets; /* reset */ 1328 return; 1329 } 1330 V_curr_dyn_buckets = V_dyn_buckets; 1331 if (V_ipfw_dyn_v != NULL) 1332 free(V_ipfw_dyn_v, M_IPFW); 1333 for (;;) { 1334 V_ipfw_dyn_v = malloc(V_curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1335 M_IPFW, M_NOWAIT | M_ZERO); 1336 if (V_ipfw_dyn_v != NULL || V_curr_dyn_buckets <= 2) 1337 break; 1338 V_curr_dyn_buckets /= 2; 1339 } 1340} 1341 1342/** 1343 * Install state of type 'type' for a dynamic session. 1344 * The hash table contains two type of rules: 1345 * - regular rules (O_KEEP_STATE) 1346 * - rules for sessions with limited number of sess per user 1347 * (O_LIMIT). When they are created, the parent is 1348 * increased by 1, and decreased on delete. In this case, 1349 * the third parameter is the parent rule and not the chain. 1350 * - "parent" rules for the above (O_LIMIT_PARENT). 1351 */ 1352static ipfw_dyn_rule * 1353add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1354{ 1355 ipfw_dyn_rule *r; 1356 int i; 1357 1358 IPFW_DYN_LOCK_ASSERT(); 1359 1360 if (V_ipfw_dyn_v == NULL || 1361 (V_dyn_count == 0 && V_dyn_buckets != V_curr_dyn_buckets)) { 1362 realloc_dynamic_table(); 1363 if (V_ipfw_dyn_v == NULL) 1364 return NULL; /* failed ! */ 1365 } 1366 i = hash_packet(id); 1367 1368 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1369 if (r == NULL) { 1370 printf ("ipfw: sorry cannot allocate state\n"); 1371 return NULL; 1372 } 1373 1374 /* increase refcount on parent, and set pointer */ 1375 if (dyn_type == O_LIMIT) { 1376 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1377 if ( parent->dyn_type != O_LIMIT_PARENT) 1378 panic("invalid parent"); 1379 parent->count++; 1380 r->parent = parent; 1381 rule = parent->rule; 1382 } 1383 1384 r->id = *id; 1385 r->expire = time_uptime + V_dyn_syn_lifetime; 1386 r->rule = rule; 1387 r->dyn_type = dyn_type; 1388 r->pcnt = r->bcnt = 0; 1389 r->count = 0; 1390 1391 r->bucket = i; 1392 r->next = V_ipfw_dyn_v[i]; 1393 V_ipfw_dyn_v[i] = r; 1394 V_dyn_count++; 1395 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1396 dyn_type, 1397 (r->id.src_ip), (r->id.src_port), 1398 (r->id.dst_ip), (r->id.dst_port), 1399 V_dyn_count ); ) 1400 return r; 1401} 1402 1403/** 1404 * lookup dynamic parent rule using pkt and rule as search keys. 1405 * If the lookup fails, then install one. 1406 */ 1407static ipfw_dyn_rule * 1408lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1409{ 1410 ipfw_dyn_rule *q; 1411 int i; 1412 1413 IPFW_DYN_LOCK_ASSERT(); 1414 1415 if (V_ipfw_dyn_v) { 1416 int is_v6 = IS_IP6_FLOW_ID(pkt); 1417 i = hash_packet( pkt ); 1418 for (q = V_ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1419 if (q->dyn_type == O_LIMIT_PARENT && 1420 rule== q->rule && 1421 pkt->proto == q->id.proto && 1422 pkt->src_port == q->id.src_port && 1423 pkt->dst_port == q->id.dst_port && 1424 ( 1425 (is_v6 && 1426 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1427 &(q->id.src_ip6)) && 1428 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1429 &(q->id.dst_ip6))) || 1430 (!is_v6 && 1431 pkt->src_ip == q->id.src_ip && 1432 pkt->dst_ip == q->id.dst_ip) 1433 ) 1434 ) { 1435 q->expire = time_uptime + V_dyn_short_lifetime; 1436 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1437 return q; 1438 } 1439 } 1440 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1441} 1442 1443/** 1444 * Install dynamic state for rule type cmd->o.opcode 1445 * 1446 * Returns 1 (failure) if state is not installed because of errors or because 1447 * session limitations are enforced. 1448 */ 1449static int 1450install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1451 struct ip_fw_args *args, uint32_t tablearg) 1452{ 1453 static int last_log; 1454 ipfw_dyn_rule *q; 1455 struct in_addr da; 1456 char src[48], dst[48]; 1457 1458 src[0] = '\0'; 1459 dst[0] = '\0'; 1460 1461 DEB( 1462 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1463 __func__, cmd->o.opcode, 1464 (args->f_id.src_ip), (args->f_id.src_port), 1465 (args->f_id.dst_ip), (args->f_id.dst_port)); 1466 ) 1467 1468 IPFW_DYN_LOCK(); 1469 1470 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1471 1472 if (q != NULL) { /* should never occur */ 1473 if (last_log != time_uptime) { 1474 last_log = time_uptime; 1475 printf("ipfw: %s: entry already present, done\n", 1476 __func__); 1477 } 1478 IPFW_DYN_UNLOCK(); 1479 return (0); 1480 } 1481 1482 if (V_dyn_count >= V_dyn_max) 1483 /* Run out of slots, try to remove any expired rule. */ 1484 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1485 1486 if (V_dyn_count >= V_dyn_max) { 1487 if (last_log != time_uptime) { 1488 last_log = time_uptime; 1489 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1490 } 1491 IPFW_DYN_UNLOCK(); 1492 return (1); /* cannot install, notify caller */ 1493 } 1494 1495 switch (cmd->o.opcode) { 1496 case O_KEEP_STATE: /* bidir rule */ 1497 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1498 break; 1499 1500 case O_LIMIT: { /* limit number of sessions */ 1501 struct ipfw_flow_id id; 1502 ipfw_dyn_rule *parent; 1503 uint32_t conn_limit; 1504 uint16_t limit_mask = cmd->limit_mask; 1505 1506 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1507 tablearg : cmd->conn_limit; 1508 1509 DEB( 1510 if (cmd->conn_limit == IP_FW_TABLEARG) 1511 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1512 "(tablearg)\n", __func__, conn_limit); 1513 else 1514 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1515 __func__, conn_limit); 1516 ) 1517 1518 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1519 id.proto = args->f_id.proto; 1520 id.addr_type = args->f_id.addr_type; 1521 id.fib = M_GETFIB(args->m); 1522 1523 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1524 if (limit_mask & DYN_SRC_ADDR) 1525 id.src_ip6 = args->f_id.src_ip6; 1526 if (limit_mask & DYN_DST_ADDR) 1527 id.dst_ip6 = args->f_id.dst_ip6; 1528 } else { 1529 if (limit_mask & DYN_SRC_ADDR) 1530 id.src_ip = args->f_id.src_ip; 1531 if (limit_mask & DYN_DST_ADDR) 1532 id.dst_ip = args->f_id.dst_ip; 1533 } 1534 if (limit_mask & DYN_SRC_PORT) 1535 id.src_port = args->f_id.src_port; 1536 if (limit_mask & DYN_DST_PORT) 1537 id.dst_port = args->f_id.dst_port; 1538 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1539 printf("ipfw: %s: add parent failed\n", __func__); 1540 IPFW_DYN_UNLOCK(); 1541 return (1); 1542 } 1543 1544 if (parent->count >= conn_limit) { 1545 /* See if we can remove some expired rule. */ 1546 remove_dyn_rule(rule, parent); 1547 if (parent->count >= conn_limit) { 1548 if (V_fw_verbose && last_log != time_uptime) { 1549 last_log = time_uptime; 1550#ifdef INET6 1551 /* 1552 * XXX IPv6 flows are not 1553 * supported yet. 1554 */ 1555 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1556 char ip6buf[INET6_ADDRSTRLEN]; 1557 snprintf(src, sizeof(src), 1558 "[%s]", ip6_sprintf(ip6buf, 1559 &args->f_id.src_ip6)); 1560 snprintf(dst, sizeof(dst), 1561 "[%s]", ip6_sprintf(ip6buf, 1562 &args->f_id.dst_ip6)); 1563 } else 1564#endif 1565 { 1566 da.s_addr = 1567 htonl(args->f_id.src_ip); 1568 inet_ntoa_r(da, src); 1569 da.s_addr = 1570 htonl(args->f_id.dst_ip); 1571 inet_ntoa_r(da, dst); 1572 } 1573 log(LOG_SECURITY | LOG_DEBUG, 1574 "ipfw: %d %s %s:%u -> %s:%u, %s\n", 1575 parent->rule->rulenum, 1576 "drop session", 1577 src, (args->f_id.src_port), 1578 dst, (args->f_id.dst_port), 1579 "too many entries"); 1580 } 1581 IPFW_DYN_UNLOCK(); 1582 return (1); 1583 } 1584 } 1585 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1586 break; 1587 } 1588 default: 1589 printf("ipfw: %s: unknown dynamic rule type %u\n", 1590 __func__, cmd->o.opcode); 1591 IPFW_DYN_UNLOCK(); 1592 return (1); 1593 } 1594 1595 /* XXX just set lifetime */ 1596 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1597 1598 IPFW_DYN_UNLOCK(); 1599 return (0); 1600} 1601 1602/* 1603 * Generate a TCP packet, containing either a RST or a keepalive. 1604 * When flags & TH_RST, we are sending a RST packet, because of a 1605 * "reset" action matched the packet. 1606 * Otherwise we are sending a keepalive, and flags & TH_ 1607 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 1608 * so that MAC can label the reply appropriately. 1609 */ 1610static struct mbuf * 1611send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 1612 u_int32_t ack, int flags) 1613{ 1614 struct mbuf *m; 1615 int len, dir; 1616 struct ip *h = NULL; /* stupid compiler */ 1617#ifdef INET6 1618 struct ip6_hdr *h6 = NULL; 1619#endif 1620 struct tcphdr *th = NULL; 1621 1622 MGETHDR(m, M_DONTWAIT, MT_DATA); 1623 if (m == NULL) 1624 return (NULL); 1625 1626 M_SETFIB(m, id->fib); 1627#ifdef MAC 1628 if (replyto != NULL) 1629 mac_netinet_firewall_reply(replyto, m); 1630 else 1631 mac_netinet_firewall_send(m); 1632#else 1633 (void)replyto; /* don't warn about unused arg */ 1634#endif 1635 1636 switch (id->addr_type) { 1637 case 4: 1638 len = sizeof(struct ip) + sizeof(struct tcphdr); 1639 break; 1640#ifdef INET6 1641 case 6: 1642 len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr); 1643 break; 1644#endif 1645 default: 1646 /* XXX: log me?!? */ 1647 m_freem(m); 1648 return (NULL); 1649 } 1650 dir = ((flags & (TH_SYN | TH_RST)) == TH_SYN); 1651 1652 m->m_data += max_linkhdr; 1653 m->m_flags |= M_SKIP_FIREWALL; 1654 m->m_pkthdr.len = m->m_len = len; 1655 m->m_pkthdr.rcvif = NULL; 1656 bzero(m->m_data, len); 1657 1658 switch (id->addr_type) { 1659 case 4: 1660 h = mtod(m, struct ip *); 1661 1662 /* prepare for checksum */ 1663 h->ip_p = IPPROTO_TCP; 1664 h->ip_len = htons(sizeof(struct tcphdr)); 1665 if (dir) { 1666 h->ip_src.s_addr = htonl(id->src_ip); 1667 h->ip_dst.s_addr = htonl(id->dst_ip); 1668 } else { 1669 h->ip_src.s_addr = htonl(id->dst_ip); 1670 h->ip_dst.s_addr = htonl(id->src_ip); 1671 } 1672 1673 th = (struct tcphdr *)(h + 1); 1674 break; 1675#ifdef INET6 1676 case 6: 1677 h6 = mtod(m, struct ip6_hdr *); 1678 1679 /* prepare for checksum */ 1680 h6->ip6_nxt = IPPROTO_TCP; 1681 h6->ip6_plen = htons(sizeof(struct tcphdr)); 1682 if (dir) { 1683 h6->ip6_src = id->src_ip6; 1684 h6->ip6_dst = id->dst_ip6; 1685 } else { 1686 h6->ip6_src = id->dst_ip6; 1687 h6->ip6_dst = id->src_ip6; 1688 } 1689 1690 th = (struct tcphdr *)(h6 + 1); 1691 break; 1692#endif 1693 } 1694 1695 if (dir) { 1696 th->th_sport = htons(id->src_port); 1697 th->th_dport = htons(id->dst_port); 1698 } else { 1699 th->th_sport = htons(id->dst_port); 1700 th->th_dport = htons(id->src_port); 1701 } 1702 th->th_off = sizeof(struct tcphdr) >> 2; 1703 1704 if (flags & TH_RST) { 1705 if (flags & TH_ACK) { 1706 th->th_seq = htonl(ack); 1707 th->th_flags = TH_RST; 1708 } else { 1709 if (flags & TH_SYN) 1710 seq++; 1711 th->th_ack = htonl(seq); 1712 th->th_flags = TH_RST | TH_ACK; 1713 } 1714 } else { 1715 /* 1716 * Keepalive - use caller provided sequence numbers 1717 */ 1718 th->th_seq = htonl(seq); 1719 th->th_ack = htonl(ack); 1720 th->th_flags = TH_ACK; 1721 } 1722 1723 switch (id->addr_type) { 1724 case 4: 1725 th->th_sum = in_cksum(m, len); 1726 1727 /* finish the ip header */ 1728 h->ip_v = 4; 1729 h->ip_hl = sizeof(*h) >> 2; 1730 h->ip_tos = IPTOS_LOWDELAY; 1731 h->ip_off = 0; 1732 h->ip_len = len; 1733 h->ip_ttl = V_ip_defttl; 1734 h->ip_sum = 0; 1735 break; 1736#ifdef INET6 1737 case 6: 1738 th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(*h6), 1739 sizeof(struct tcphdr)); 1740 1741 /* finish the ip6 header */ 1742 h6->ip6_vfc |= IPV6_VERSION; 1743 h6->ip6_hlim = IPV6_DEFHLIM; 1744 break; 1745#endif 1746 } 1747 1748 return (m); 1749} 1750 1751/* 1752 * sends a reject message, consuming the mbuf passed as an argument. 1753 */ 1754static void 1755send_reject(struct ip_fw_args *args, int code, int ip_len, struct ip *ip) 1756{ 1757 1758#if 0 1759 /* XXX When ip is not guaranteed to be at mtod() we will 1760 * need to account for this */ 1761 * The mbuf will however be thrown away so we can adjust it. 1762 * Remember we did an m_pullup on it already so we 1763 * can make some assumptions about contiguousness. 1764 */ 1765 if (args->L3offset) 1766 m_adj(m, args->L3offset); 1767#endif 1768 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1769 /* We need the IP header in host order for icmp_error(). */ 1770 if (args->eh != NULL) { 1771 ip->ip_len = ntohs(ip->ip_len); 1772 ip->ip_off = ntohs(ip->ip_off); 1773 } 1774 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1775 } else if (args->f_id.proto == IPPROTO_TCP) { 1776 struct tcphdr *const tcp = 1777 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1778 if ( (tcp->th_flags & TH_RST) == 0) { 1779 struct mbuf *m; 1780 m = send_pkt(args->m, &(args->f_id), 1781 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1782 tcp->th_flags | TH_RST); 1783 if (m != NULL) 1784 ip_output(m, NULL, NULL, 0, NULL, NULL); 1785 } 1786 m_freem(args->m); 1787 } else 1788 m_freem(args->m); 1789 args->m = NULL; 1790} 1791 1792/** 1793 * 1794 * Given an ip_fw *, lookup_next_rule will return a pointer 1795 * to the next rule, which can be either the jump 1796 * target (for skipto instructions) or the next one in the list (in 1797 * all other cases including a missing jump target). 1798 * The result is also written in the "next_rule" field of the rule. 1799 * Backward jumps are not allowed, so start looking from the next 1800 * rule... 1801 * 1802 * This never returns NULL -- in case we do not have an exact match, 1803 * the next rule is returned. When the ruleset is changed, 1804 * pointers are flushed so we are always correct. 1805 */ 1806 1807static struct ip_fw * 1808lookup_next_rule(struct ip_fw *me, u_int32_t tablearg) 1809{ 1810 struct ip_fw *rule = NULL; 1811 ipfw_insn *cmd; 1812 u_int16_t rulenum; 1813 1814 /* look for action, in case it is a skipto */ 1815 cmd = ACTION_PTR(me); 1816 if (cmd->opcode == O_LOG) 1817 cmd += F_LEN(cmd); 1818 if (cmd->opcode == O_ALTQ) 1819 cmd += F_LEN(cmd); 1820 if (cmd->opcode == O_TAG) 1821 cmd += F_LEN(cmd); 1822 if (cmd->opcode == O_SKIPTO ) { 1823 if (tablearg != 0) { 1824 rulenum = (u_int16_t)tablearg; 1825 } else { 1826 rulenum = cmd->arg1; 1827 } 1828 for (rule = me->next; rule ; rule = rule->next) { 1829 if (rule->rulenum >= rulenum) { 1830 break; 1831 } 1832 } 1833 } 1834 if (rule == NULL) /* failure or not a skipto */ 1835 rule = me->next; 1836 me->next_rule = rule; 1837 return rule; 1838} 1839 1840static int 1841add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1842 uint8_t mlen, uint32_t value) 1843{ 1844 struct radix_node_head *rnh; 1845 struct table_entry *ent; 1846 struct radix_node *rn; 1847 1848 if (tbl >= IPFW_TABLES_MAX) 1849 return (EINVAL); 1850 rnh = ch->tables[tbl]; 1851 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1852 if (ent == NULL) 1853 return (ENOMEM); 1854 ent->value = value; 1855 ent->addr.sin_len = ent->mask.sin_len = 8; 1856 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1857 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1858 IPFW_WLOCK(ch); 1859 rn = rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent); 1860 if (rn == NULL) { 1861 IPFW_WUNLOCK(ch); 1862 free(ent, M_IPFW_TBL); 1863 return (EEXIST); 1864 } 1865 IPFW_WUNLOCK(ch); 1866 return (0); 1867} 1868 1869static int 1870del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1871 uint8_t mlen) 1872{ 1873 struct radix_node_head *rnh; 1874 struct table_entry *ent; 1875 struct sockaddr_in sa, mask; 1876 1877 if (tbl >= IPFW_TABLES_MAX) 1878 return (EINVAL); 1879 rnh = ch->tables[tbl]; 1880 sa.sin_len = mask.sin_len = 8; 1881 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1882 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1883 IPFW_WLOCK(ch); 1884 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1885 if (ent == NULL) { 1886 IPFW_WUNLOCK(ch); 1887 return (ESRCH); 1888 } 1889 IPFW_WUNLOCK(ch); 1890 free(ent, M_IPFW_TBL); 1891 return (0); 1892} 1893 1894static int 1895flush_table_entry(struct radix_node *rn, void *arg) 1896{ 1897 struct radix_node_head * const rnh = arg; 1898 struct table_entry *ent; 1899 1900 ent = (struct table_entry *) 1901 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1902 if (ent != NULL) 1903 free(ent, M_IPFW_TBL); 1904 return (0); 1905} 1906 1907static int 1908flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1909{ 1910 struct radix_node_head *rnh; 1911 1912 IPFW_WLOCK_ASSERT(ch); 1913 1914 if (tbl >= IPFW_TABLES_MAX) 1915 return (EINVAL); 1916 rnh = ch->tables[tbl]; 1917 KASSERT(rnh != NULL, ("NULL IPFW table")); 1918 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1919 return (0); 1920} 1921 1922static void 1923flush_tables(struct ip_fw_chain *ch) 1924{ 1925 uint16_t tbl; 1926 1927 IPFW_WLOCK_ASSERT(ch); 1928 1929 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1930 flush_table(ch, tbl); 1931} 1932 1933static int 1934init_tables(struct ip_fw_chain *ch) 1935{ 1936 int i; 1937 uint16_t j; 1938 1939 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1940 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1941 for (j = 0; j < i; j++) { 1942 (void) flush_table(ch, j); 1943 } 1944 return (ENOMEM); 1945 } 1946 } 1947 return (0); 1948} 1949 1950static int 1951lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1952 uint32_t *val) 1953{ 1954 struct radix_node_head *rnh; 1955 struct table_entry *ent; 1956 struct sockaddr_in sa; 1957 1958 if (tbl >= IPFW_TABLES_MAX) 1959 return (0); 1960 rnh = ch->tables[tbl]; 1961 sa.sin_len = 8; 1962 sa.sin_addr.s_addr = addr; 1963 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1964 if (ent != NULL) { 1965 *val = ent->value; 1966 return (1); 1967 } 1968 return (0); 1969} 1970 1971static int 1972count_table_entry(struct radix_node *rn, void *arg) 1973{ 1974 u_int32_t * const cnt = arg; 1975 1976 (*cnt)++; 1977 return (0); 1978} 1979 1980static int 1981count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1982{ 1983 struct radix_node_head *rnh; 1984 1985 if (tbl >= IPFW_TABLES_MAX) 1986 return (EINVAL); 1987 rnh = ch->tables[tbl]; 1988 *cnt = 0; 1989 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1990 return (0); 1991} 1992 1993static int 1994dump_table_entry(struct radix_node *rn, void *arg) 1995{ 1996 struct table_entry * const n = (struct table_entry *)rn; 1997 ipfw_table * const tbl = arg; 1998 ipfw_table_entry *ent; 1999 2000 if (tbl->cnt == tbl->size) 2001 return (1); 2002 ent = &tbl->ent[tbl->cnt]; 2003 ent->tbl = tbl->tbl; 2004 if (in_nullhost(n->mask.sin_addr)) 2005 ent->masklen = 0; 2006 else 2007 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 2008 ent->addr = n->addr.sin_addr.s_addr; 2009 ent->value = n->value; 2010 tbl->cnt++; 2011 return (0); 2012} 2013 2014static int 2015dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 2016{ 2017 struct radix_node_head *rnh; 2018 2019 if (tbl->tbl >= IPFW_TABLES_MAX) 2020 return (EINVAL); 2021 rnh = ch->tables[tbl->tbl]; 2022 tbl->cnt = 0; 2023 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 2024 return (0); 2025} 2026 2027static int 2028check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif, 2029 struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip, 2030 u_int16_t src_port, struct ucred **uc, int *ugid_lookupp, 2031 struct inpcb *inp) 2032{ 2033 struct inpcbinfo *pi; 2034 int wildcard; 2035 struct inpcb *pcb; 2036 int match; 2037 2038 /* 2039 * Check to see if the UDP or TCP stack supplied us with 2040 * the PCB. If so, rather then holding a lock and looking 2041 * up the PCB, we can use the one that was supplied. 2042 */ 2043 if (inp && *ugid_lookupp == 0) { 2044 INP_LOCK_ASSERT(inp); 2045 if (inp->inp_socket != NULL) { 2046 *uc = crhold(inp->inp_cred); 2047 *ugid_lookupp = 1; 2048 } else 2049 *ugid_lookupp = -1; 2050 } 2051 /* 2052 * If we have already been here and the packet has no 2053 * PCB entry associated with it, then we can safely 2054 * assume that this is a no match. 2055 */ 2056 if (*ugid_lookupp == -1) 2057 return (0); 2058 if (proto == IPPROTO_TCP) { 2059 wildcard = 0; 2060 pi = &V_tcbinfo; 2061 } else if (proto == IPPROTO_UDP) { 2062 wildcard = INPLOOKUP_WILDCARD; 2063 pi = &V_udbinfo; 2064 } else 2065 return 0; 2066 match = 0; 2067 if (*ugid_lookupp == 0) { 2068 INP_INFO_RLOCK(pi); 2069 pcb = (oif) ? 2070 in_pcblookup_hash(pi, 2071 dst_ip, htons(dst_port), 2072 src_ip, htons(src_port), 2073 wildcard, oif) : 2074 in_pcblookup_hash(pi, 2075 src_ip, htons(src_port), 2076 dst_ip, htons(dst_port), 2077 wildcard, NULL); 2078 if (pcb != NULL) { 2079 *uc = crhold(pcb->inp_cred); 2080 *ugid_lookupp = 1; 2081 } 2082 INP_INFO_RUNLOCK(pi); 2083 if (*ugid_lookupp == 0) { 2084 /* 2085 * If the lookup did not yield any results, there 2086 * is no sense in coming back and trying again. So 2087 * we can set lookup to -1 and ensure that we wont 2088 * bother the pcb system again. 2089 */ 2090 *ugid_lookupp = -1; 2091 return (0); 2092 } 2093 } 2094 if (insn->o.opcode == O_UID) 2095 match = ((*uc)->cr_uid == (uid_t)insn->d[0]); 2096 else if (insn->o.opcode == O_GID) 2097 match = groupmember((gid_t)insn->d[0], *uc); 2098 else if (insn->o.opcode == O_JAIL) 2099 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]); 2100 return match; 2101} 2102 2103/* 2104 * The main check routine for the firewall. 2105 * 2106 * All arguments are in args so we can modify them and return them 2107 * back to the caller. 2108 * 2109 * Parameters: 2110 * 2111 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 2112 * Starts with the IP header. 2113 * args->eh (in) Mac header if present, or NULL for layer3 packet. 2114 * args->L3offset Number of bytes bypassed if we came from L2. 2115 * e.g. often sizeof(eh) ** NOTYET ** 2116 * args->oif Outgoing interface, or NULL if packet is incoming. 2117 * The incoming interface is in the mbuf. (in) 2118 * args->divert_rule (in/out) 2119 * Skip up to the first rule past this rule number; 2120 * upon return, non-zero port number for divert or tee. 2121 * 2122 * args->rule Pointer to the last matching rule (in/out) 2123 * args->next_hop Socket we are forwarding to (out). 2124 * args->f_id Addresses grabbed from the packet (out) 2125 * args->cookie a cookie depending on rule action 2126 * 2127 * Return value: 2128 * 2129 * IP_FW_PASS the packet must be accepted 2130 * IP_FW_DENY the packet must be dropped 2131 * IP_FW_DIVERT divert packet, port in m_tag 2132 * IP_FW_TEE tee packet, port in m_tag 2133 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2134 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2135 * 2136 */ 2137int 2138ipfw_chk(struct ip_fw_args *args) 2139{ 2140 2141 /* 2142 * Local variables holding state during the processing of a packet: 2143 * 2144 * IMPORTANT NOTE: to speed up the processing of rules, there 2145 * are some assumption on the values of the variables, which 2146 * are documented here. Should you change them, please check 2147 * the implementation of the various instructions to make sure 2148 * that they still work. 2149 * 2150 * args->eh The MAC header. It is non-null for a layer2 2151 * packet, it is NULL for a layer-3 packet. 2152 * **notyet** 2153 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. 2154 * 2155 * m | args->m Pointer to the mbuf, as received from the caller. 2156 * It may change if ipfw_chk() does an m_pullup, or if it 2157 * consumes the packet because it calls send_reject(). 2158 * XXX This has to change, so that ipfw_chk() never modifies 2159 * or consumes the buffer. 2160 * ip is the beginning of the ip(4 or 6) header. 2161 * Calculated by adding the L3offset to the start of data. 2162 * (Until we start using L3offset, the packet is 2163 * supposed to start with the ip header). 2164 */ 2165 struct mbuf *m = args->m; 2166 struct ip *ip = mtod(m, struct ip *); 2167 2168 /* 2169 * For rules which contain uid/gid or jail constraints, cache 2170 * a copy of the users credentials after the pcb lookup has been 2171 * executed. This will speed up the processing of rules with 2172 * these types of constraints, as well as decrease contention 2173 * on pcb related locks. 2174 */ 2175 struct ucred *ucred_cache = NULL; 2176 int ucred_lookup = 0; 2177 2178 /* 2179 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2180 * associated with a packet input on a divert socket. This 2181 * will allow to distinguish traffic and its direction when 2182 * it originates from a divert socket. 2183 */ 2184 u_int divinput_flags = 0; 2185 2186 /* 2187 * oif | args->oif If NULL, ipfw_chk has been called on the 2188 * inbound path (ether_input, ip_input). 2189 * If non-NULL, ipfw_chk has been called on the outbound path 2190 * (ether_output, ip_output). 2191 */ 2192 struct ifnet *oif = args->oif; 2193 2194 struct ip_fw *f = NULL; /* matching rule */ 2195 int retval = 0; 2196 2197 /* 2198 * hlen The length of the IP header. 2199 */ 2200 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2201 2202 /* 2203 * offset The offset of a fragment. offset != 0 means that 2204 * we have a fragment at this offset of an IPv4 packet. 2205 * offset == 0 means that (if this is an IPv4 packet) 2206 * this is the first or only fragment. 2207 * For IPv6 offset == 0 means there is no Fragment Header. 2208 * If offset != 0 for IPv6 always use correct mask to 2209 * get the correct offset because we add IP6F_MORE_FRAG 2210 * to be able to dectect the first fragment which would 2211 * otherwise have offset = 0. 2212 */ 2213 u_short offset = 0; 2214 2215 /* 2216 * Local copies of addresses. They are only valid if we have 2217 * an IP packet. 2218 * 2219 * proto The protocol. Set to 0 for non-ip packets, 2220 * or to the protocol read from the packet otherwise. 2221 * proto != 0 means that we have an IPv4 packet. 2222 * 2223 * src_port, dst_port port numbers, in HOST format. Only 2224 * valid for TCP and UDP packets. 2225 * 2226 * src_ip, dst_ip ip addresses, in NETWORK format. 2227 * Only valid for IPv4 packets. 2228 */ 2229 u_int8_t proto; 2230 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2231 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2232 u_int16_t ip_len=0; 2233 int pktlen; 2234 u_int16_t etype = 0; /* Host order stored ether type */ 2235 2236 /* 2237 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2238 * MATCH_NONE when checked and not matched (q = NULL), 2239 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2240 */ 2241 int dyn_dir = MATCH_UNKNOWN; 2242 ipfw_dyn_rule *q = NULL; 2243 struct ip_fw_chain *chain = &V_layer3_chain; 2244 struct m_tag *mtag; 2245 2246 /* 2247 * We store in ulp a pointer to the upper layer protocol header. 2248 * In the ipv4 case this is easy to determine from the header, 2249 * but for ipv6 we might have some additional headers in the middle. 2250 * ulp is NULL if not found. 2251 */ 2252 void *ulp = NULL; /* upper layer protocol pointer. */ 2253 /* XXX ipv6 variables */ 2254 int is_ipv6 = 0; 2255 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2256 /* end of ipv6 variables */ 2257 int is_ipv4 = 0; 2258 2259 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready)) 2260 return (IP_FW_PASS); /* accept */ 2261 2262 dst_ip.s_addr = 0; /* make sure it is initialized */ 2263 pktlen = m->m_pkthdr.len; 2264 args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */ 2265 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2266 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2267 2268/* 2269 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2270 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2271 * pointer might become stale after other pullups (but we never use it 2272 * this way). 2273 */ 2274#define PULLUP_TO(len, p, T) \ 2275do { \ 2276 int x = (len) + sizeof(T); \ 2277 if ((m)->m_len < x) { \ 2278 args->m = m = m_pullup(m, x); \ 2279 if (m == NULL) \ 2280 goto pullup_failed; \ 2281 } \ 2282 p = (mtod(m, char *) + (len)); \ 2283} while (0) 2284 2285 /* 2286 * if we have an ether header, 2287 */ 2288 if (args->eh) 2289 etype = ntohs(args->eh->ether_type); 2290 2291 /* Identify IP packets and fill up variables. */ 2292 if (pktlen >= sizeof(struct ip6_hdr) && 2293 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { 2294 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 2295 is_ipv6 = 1; 2296 args->f_id.addr_type = 6; 2297 hlen = sizeof(struct ip6_hdr); 2298 proto = ip6->ip6_nxt; 2299 2300 /* Search extension headers to find upper layer protocols */ 2301 while (ulp == NULL) { 2302 switch (proto) { 2303 case IPPROTO_ICMPV6: 2304 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2305 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2306 break; 2307 2308 case IPPROTO_TCP: 2309 PULLUP_TO(hlen, ulp, struct tcphdr); 2310 dst_port = TCP(ulp)->th_dport; 2311 src_port = TCP(ulp)->th_sport; 2312 args->f_id.flags = TCP(ulp)->th_flags; 2313 break; 2314 2315 case IPPROTO_SCTP: 2316 PULLUP_TO(hlen, ulp, struct sctphdr); 2317 src_port = SCTP(ulp)->src_port; 2318 dst_port = SCTP(ulp)->dest_port; 2319 break; 2320 2321 case IPPROTO_UDP: 2322 PULLUP_TO(hlen, ulp, struct udphdr); 2323 dst_port = UDP(ulp)->uh_dport; 2324 src_port = UDP(ulp)->uh_sport; 2325 break; 2326 2327 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2328 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2329 ext_hd |= EXT_HOPOPTS; 2330 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2331 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2332 ulp = NULL; 2333 break; 2334 2335 case IPPROTO_ROUTING: /* RFC 2460 */ 2336 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2337 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2338 case 0: 2339 ext_hd |= EXT_RTHDR0; 2340 break; 2341 case 2: 2342 ext_hd |= EXT_RTHDR2; 2343 break; 2344 default: 2345 printf("IPFW2: IPV6 - Unknown Routing " 2346 "Header type(%d)\n", 2347 ((struct ip6_rthdr *)ulp)->ip6r_type); 2348 if (V_fw_deny_unknown_exthdrs) 2349 return (IP_FW_DENY); 2350 break; 2351 } 2352 ext_hd |= EXT_ROUTING; 2353 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2354 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2355 ulp = NULL; 2356 break; 2357 2358 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2359 PULLUP_TO(hlen, ulp, struct ip6_frag); 2360 ext_hd |= EXT_FRAGMENT; 2361 hlen += sizeof (struct ip6_frag); 2362 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2363 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2364 IP6F_OFF_MASK; 2365 /* Add IP6F_MORE_FRAG for offset of first 2366 * fragment to be != 0. */ 2367 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2368 IP6F_MORE_FRAG; 2369 if (offset == 0) { 2370 printf("IPFW2: IPV6 - Invalid Fragment " 2371 "Header\n"); 2372 if (V_fw_deny_unknown_exthdrs) 2373 return (IP_FW_DENY); 2374 break; 2375 } 2376 args->f_id.frag_id6 = 2377 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2378 ulp = NULL; 2379 break; 2380 2381 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2382 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2383 ext_hd |= EXT_DSTOPTS; 2384 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2385 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2386 ulp = NULL; 2387 break; 2388 2389 case IPPROTO_AH: /* RFC 2402 */ 2390 PULLUP_TO(hlen, ulp, struct ip6_ext); 2391 ext_hd |= EXT_AH; 2392 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2393 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2394 ulp = NULL; 2395 break; 2396 2397 case IPPROTO_ESP: /* RFC 2406 */ 2398 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2399 /* Anything past Seq# is variable length and 2400 * data past this ext. header is encrypted. */ 2401 ext_hd |= EXT_ESP; 2402 break; 2403 2404 case IPPROTO_NONE: /* RFC 2460 */ 2405 /* 2406 * Packet ends here, and IPv6 header has 2407 * already been pulled up. If ip6e_len!=0 2408 * then octets must be ignored. 2409 */ 2410 ulp = ip; /* non-NULL to get out of loop. */ 2411 break; 2412 2413 case IPPROTO_OSPFIGP: 2414 /* XXX OSPF header check? */ 2415 PULLUP_TO(hlen, ulp, struct ip6_ext); 2416 break; 2417 2418 case IPPROTO_PIM: 2419 /* XXX PIM header check? */ 2420 PULLUP_TO(hlen, ulp, struct pim); 2421 break; 2422 2423 case IPPROTO_CARP: 2424 PULLUP_TO(hlen, ulp, struct carp_header); 2425 if (((struct carp_header *)ulp)->carp_version != 2426 CARP_VERSION) 2427 return (IP_FW_DENY); 2428 if (((struct carp_header *)ulp)->carp_type != 2429 CARP_ADVERTISEMENT) 2430 return (IP_FW_DENY); 2431 break; 2432 2433 case IPPROTO_IPV6: /* RFC 2893 */ 2434 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2435 break; 2436 2437 case IPPROTO_IPV4: /* RFC 2893 */ 2438 PULLUP_TO(hlen, ulp, struct ip); 2439 break; 2440 2441 default: 2442 printf("IPFW2: IPV6 - Unknown Extension " 2443 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2444 if (V_fw_deny_unknown_exthdrs) 2445 return (IP_FW_DENY); 2446 PULLUP_TO(hlen, ulp, struct ip6_ext); 2447 break; 2448 } /*switch */ 2449 } 2450 ip = mtod(m, struct ip *); 2451 ip6 = (struct ip6_hdr *)ip; 2452 args->f_id.src_ip6 = ip6->ip6_src; 2453 args->f_id.dst_ip6 = ip6->ip6_dst; 2454 args->f_id.src_ip = 0; 2455 args->f_id.dst_ip = 0; 2456 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 2457 } else if (pktlen >= sizeof(struct ip) && 2458 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { 2459 is_ipv4 = 1; 2460 hlen = ip->ip_hl << 2; 2461 args->f_id.addr_type = 4; 2462 2463 /* 2464 * Collect parameters into local variables for faster matching. 2465 */ 2466 proto = ip->ip_p; 2467 src_ip = ip->ip_src; 2468 dst_ip = ip->ip_dst; 2469 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2470 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2471 ip_len = ntohs(ip->ip_len); 2472 } else { 2473 offset = ip->ip_off & IP_OFFMASK; 2474 ip_len = ip->ip_len; 2475 } 2476 pktlen = ip_len < pktlen ? ip_len : pktlen; 2477 2478 if (offset == 0) { 2479 switch (proto) { 2480 case IPPROTO_TCP: 2481 PULLUP_TO(hlen, ulp, struct tcphdr); 2482 dst_port = TCP(ulp)->th_dport; 2483 src_port = TCP(ulp)->th_sport; 2484 args->f_id.flags = TCP(ulp)->th_flags; 2485 break; 2486 2487 case IPPROTO_UDP: 2488 PULLUP_TO(hlen, ulp, struct udphdr); 2489 dst_port = UDP(ulp)->uh_dport; 2490 src_port = UDP(ulp)->uh_sport; 2491 break; 2492 2493 case IPPROTO_ICMP: 2494 PULLUP_TO(hlen, ulp, struct icmphdr); 2495 args->f_id.flags = ICMP(ulp)->icmp_type; 2496 break; 2497 2498 default: 2499 break; 2500 } 2501 } 2502 2503 ip = mtod(m, struct ip *); 2504 args->f_id.src_ip = ntohl(src_ip.s_addr); 2505 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2506 } 2507#undef PULLUP_TO 2508 if (proto) { /* we may have port numbers, store them */ 2509 args->f_id.proto = proto; 2510 args->f_id.src_port = src_port = ntohs(src_port); 2511 args->f_id.dst_port = dst_port = ntohs(dst_port); 2512 } 2513 2514 IPFW_RLOCK(chain); 2515 if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */ 2516 IPFW_RUNLOCK(chain); 2517 return (IP_FW_PASS); /* accept */ 2518 } 2519 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2520 if (args->rule) { 2521 /* 2522 * Packet has already been tagged. Look for the next rule 2523 * to restart processing. Make sure that args->rule still 2524 * exists and not changed. 2525 */ 2526 if (chain->id != args->chain_id) { 2527 for (f = chain->rules; f != NULL; f = f->next) 2528 if (f == args->rule && f->id == args->rule_id) 2529 break; 2530 2531 if (f != NULL) 2532 f = f->next_rule; 2533 else 2534 f = ip_fw_default_rule; 2535 } else 2536 f = args->rule->next_rule; 2537 2538 if (f == NULL) 2539 f = lookup_next_rule(args->rule, 0); 2540 } else { 2541 /* 2542 * Find the starting rule. It can be either the first 2543 * one, or the one after divert_rule if asked so. 2544 */ 2545 int skipto = mtag ? divert_cookie(mtag) : 0; 2546 2547 f = chain->rules; 2548 if (args->eh == NULL && skipto != 0) { 2549 if (skipto >= IPFW_DEFAULT_RULE) { 2550 IPFW_RUNLOCK(chain); 2551 return (IP_FW_DENY); /* invalid */ 2552 } 2553 while (f && f->rulenum <= skipto) 2554 f = f->next; 2555 if (f == NULL) { /* drop packet */ 2556 IPFW_RUNLOCK(chain); 2557 return (IP_FW_DENY); 2558 } 2559 } 2560 } 2561 /* reset divert rule to avoid confusion later */ 2562 if (mtag) { 2563 divinput_flags = divert_info(mtag) & 2564 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2565 m_tag_delete(m, mtag); 2566 } 2567 2568 /* 2569 * Now scan the rules, and parse microinstructions for each rule. 2570 */ 2571 for (; f; f = f->next) { 2572 ipfw_insn *cmd; 2573 uint32_t tablearg = 0; 2574 int l, cmdlen, skip_or; /* skip rest of OR block */ 2575 2576again: 2577 if (V_set_disable & (1 << f->set) ) 2578 continue; 2579 2580 skip_or = 0; 2581 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2582 l -= cmdlen, cmd += cmdlen) { 2583 int match; 2584 2585 /* 2586 * check_body is a jump target used when we find a 2587 * CHECK_STATE, and need to jump to the body of 2588 * the target rule. 2589 */ 2590 2591check_body: 2592 cmdlen = F_LEN(cmd); 2593 /* 2594 * An OR block (insn_1 || .. || insn_n) has the 2595 * F_OR bit set in all but the last instruction. 2596 * The first match will set "skip_or", and cause 2597 * the following instructions to be skipped until 2598 * past the one with the F_OR bit clear. 2599 */ 2600 if (skip_or) { /* skip this instruction */ 2601 if ((cmd->len & F_OR) == 0) 2602 skip_or = 0; /* next one is good */ 2603 continue; 2604 } 2605 match = 0; /* set to 1 if we succeed */ 2606 2607 switch (cmd->opcode) { 2608 /* 2609 * The first set of opcodes compares the packet's 2610 * fields with some pattern, setting 'match' if a 2611 * match is found. At the end of the loop there is 2612 * logic to deal with F_NOT and F_OR flags associated 2613 * with the opcode. 2614 */ 2615 case O_NOP: 2616 match = 1; 2617 break; 2618 2619 case O_FORWARD_MAC: 2620 printf("ipfw: opcode %d unimplemented\n", 2621 cmd->opcode); 2622 break; 2623 2624 case O_GID: 2625 case O_UID: 2626 case O_JAIL: 2627 /* 2628 * We only check offset == 0 && proto != 0, 2629 * as this ensures that we have a 2630 * packet with the ports info. 2631 */ 2632 if (offset!=0) 2633 break; 2634 if (is_ipv6) /* XXX to be fixed later */ 2635 break; 2636 if (proto == IPPROTO_TCP || 2637 proto == IPPROTO_UDP) 2638 match = check_uidgid( 2639 (ipfw_insn_u32 *)cmd, 2640 proto, oif, 2641 dst_ip, dst_port, 2642 src_ip, src_port, &ucred_cache, 2643 &ucred_lookup, args->inp); 2644 break; 2645 2646 case O_RECV: 2647 match = iface_match(m->m_pkthdr.rcvif, 2648 (ipfw_insn_if *)cmd); 2649 break; 2650 2651 case O_XMIT: 2652 match = iface_match(oif, (ipfw_insn_if *)cmd); 2653 break; 2654 2655 case O_VIA: 2656 match = iface_match(oif ? oif : 2657 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2658 break; 2659 2660 case O_MACADDR2: 2661 if (args->eh != NULL) { /* have MAC header */ 2662 u_int32_t *want = (u_int32_t *) 2663 ((ipfw_insn_mac *)cmd)->addr; 2664 u_int32_t *mask = (u_int32_t *) 2665 ((ipfw_insn_mac *)cmd)->mask; 2666 u_int32_t *hdr = (u_int32_t *)args->eh; 2667 2668 match = 2669 ( want[0] == (hdr[0] & mask[0]) && 2670 want[1] == (hdr[1] & mask[1]) && 2671 want[2] == (hdr[2] & mask[2]) ); 2672 } 2673 break; 2674 2675 case O_MAC_TYPE: 2676 if (args->eh != NULL) { 2677 u_int16_t *p = 2678 ((ipfw_insn_u16 *)cmd)->ports; 2679 int i; 2680 2681 for (i = cmdlen - 1; !match && i>0; 2682 i--, p += 2) 2683 match = (etype >= p[0] && 2684 etype <= p[1]); 2685 } 2686 break; 2687 2688 case O_FRAG: 2689 match = (offset != 0); 2690 break; 2691 2692 case O_IN: /* "out" is "not in" */ 2693 match = (oif == NULL); 2694 break; 2695 2696 case O_LAYER2: 2697 match = (args->eh != NULL); 2698 break; 2699 2700 case O_DIVERTED: 2701 match = (cmd->arg1 & 1 && divinput_flags & 2702 IP_FW_DIVERT_LOOPBACK_FLAG) || 2703 (cmd->arg1 & 2 && divinput_flags & 2704 IP_FW_DIVERT_OUTPUT_FLAG); 2705 break; 2706 2707 case O_PROTO: 2708 /* 2709 * We do not allow an arg of 0 so the 2710 * check of "proto" only suffices. 2711 */ 2712 match = (proto == cmd->arg1); 2713 break; 2714 2715 case O_IP_SRC: 2716 match = is_ipv4 && 2717 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2718 src_ip.s_addr); 2719 break; 2720 2721 case O_IP_SRC_LOOKUP: 2722 case O_IP_DST_LOOKUP: 2723 if (is_ipv4) { 2724 uint32_t a = 2725 (cmd->opcode == O_IP_DST_LOOKUP) ? 2726 dst_ip.s_addr : src_ip.s_addr; 2727 uint32_t v = 0; 2728 2729 match = lookup_table(chain, cmd->arg1, a, 2730 &v); 2731 if (!match) 2732 break; 2733 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2734 match = 2735 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2736 else 2737 tablearg = v; 2738 } 2739 break; 2740 2741 case O_IP_SRC_MASK: 2742 case O_IP_DST_MASK: 2743 if (is_ipv4) { 2744 uint32_t a = 2745 (cmd->opcode == O_IP_DST_MASK) ? 2746 dst_ip.s_addr : src_ip.s_addr; 2747 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2748 int i = cmdlen-1; 2749 2750 for (; !match && i>0; i-= 2, p+= 2) 2751 match = (p[0] == (a & p[1])); 2752 } 2753 break; 2754 2755 case O_IP_SRC_ME: 2756 if (is_ipv4) { 2757 struct ifnet *tif; 2758 2759 INADDR_TO_IFP(src_ip, tif); 2760 match = (tif != NULL); 2761 } 2762 break; 2763 2764 case O_IP_DST_SET: 2765 case O_IP_SRC_SET: 2766 if (is_ipv4) { 2767 u_int32_t *d = (u_int32_t *)(cmd+1); 2768 u_int32_t addr = 2769 cmd->opcode == O_IP_DST_SET ? 2770 args->f_id.dst_ip : 2771 args->f_id.src_ip; 2772 2773 if (addr < d[0]) 2774 break; 2775 addr -= d[0]; /* subtract base */ 2776 match = (addr < cmd->arg1) && 2777 ( d[ 1 + (addr>>5)] & 2778 (1<<(addr & 0x1f)) ); 2779 } 2780 break; 2781 2782 case O_IP_DST: 2783 match = is_ipv4 && 2784 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2785 dst_ip.s_addr); 2786 break; 2787 2788 case O_IP_DST_ME: 2789 if (is_ipv4) { 2790 struct ifnet *tif; 2791 2792 INADDR_TO_IFP(dst_ip, tif); 2793 match = (tif != NULL); 2794 } 2795 break; 2796 2797 case O_IP_SRCPORT: 2798 case O_IP_DSTPORT: 2799 /* 2800 * offset == 0 && proto != 0 is enough 2801 * to guarantee that we have a 2802 * packet with port info. 2803 */ 2804 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2805 && offset == 0) { 2806 u_int16_t x = 2807 (cmd->opcode == O_IP_SRCPORT) ? 2808 src_port : dst_port ; 2809 u_int16_t *p = 2810 ((ipfw_insn_u16 *)cmd)->ports; 2811 int i; 2812 2813 for (i = cmdlen - 1; !match && i>0; 2814 i--, p += 2) 2815 match = (x>=p[0] && x<=p[1]); 2816 } 2817 break; 2818 2819 case O_ICMPTYPE: 2820 match = (offset == 0 && proto==IPPROTO_ICMP && 2821 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2822 break; 2823 2824#ifdef INET6 2825 case O_ICMP6TYPE: 2826 match = is_ipv6 && offset == 0 && 2827 proto==IPPROTO_ICMPV6 && 2828 icmp6type_match( 2829 ICMP6(ulp)->icmp6_type, 2830 (ipfw_insn_u32 *)cmd); 2831 break; 2832#endif /* INET6 */ 2833 2834 case O_IPOPT: 2835 match = (is_ipv4 && 2836 ipopts_match(ip, cmd) ); 2837 break; 2838 2839 case O_IPVER: 2840 match = (is_ipv4 && 2841 cmd->arg1 == ip->ip_v); 2842 break; 2843 2844 case O_IPID: 2845 case O_IPLEN: 2846 case O_IPTTL: 2847 if (is_ipv4) { /* only for IP packets */ 2848 uint16_t x; 2849 uint16_t *p; 2850 int i; 2851 2852 if (cmd->opcode == O_IPLEN) 2853 x = ip_len; 2854 else if (cmd->opcode == O_IPTTL) 2855 x = ip->ip_ttl; 2856 else /* must be IPID */ 2857 x = ntohs(ip->ip_id); 2858 if (cmdlen == 1) { 2859 match = (cmd->arg1 == x); 2860 break; 2861 } 2862 /* otherwise we have ranges */ 2863 p = ((ipfw_insn_u16 *)cmd)->ports; 2864 i = cmdlen - 1; 2865 for (; !match && i>0; i--, p += 2) 2866 match = (x >= p[0] && x <= p[1]); 2867 } 2868 break; 2869 2870 case O_IPPRECEDENCE: 2871 match = (is_ipv4 && 2872 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2873 break; 2874 2875 case O_IPTOS: 2876 match = (is_ipv4 && 2877 flags_match(cmd, ip->ip_tos)); 2878 break; 2879 2880 case O_TCPDATALEN: 2881 if (proto == IPPROTO_TCP && offset == 0) { 2882 struct tcphdr *tcp; 2883 uint16_t x; 2884 uint16_t *p; 2885 int i; 2886 2887 tcp = TCP(ulp); 2888 x = ip_len - 2889 ((ip->ip_hl + tcp->th_off) << 2); 2890 if (cmdlen == 1) { 2891 match = (cmd->arg1 == x); 2892 break; 2893 } 2894 /* otherwise we have ranges */ 2895 p = ((ipfw_insn_u16 *)cmd)->ports; 2896 i = cmdlen - 1; 2897 for (; !match && i>0; i--, p += 2) 2898 match = (x >= p[0] && x <= p[1]); 2899 } 2900 break; 2901 2902 case O_TCPFLAGS: 2903 match = (proto == IPPROTO_TCP && offset == 0 && 2904 flags_match(cmd, TCP(ulp)->th_flags)); 2905 break; 2906 2907 case O_TCPOPTS: 2908 match = (proto == IPPROTO_TCP && offset == 0 && 2909 tcpopts_match(TCP(ulp), cmd)); 2910 break; 2911 2912 case O_TCPSEQ: 2913 match = (proto == IPPROTO_TCP && offset == 0 && 2914 ((ipfw_insn_u32 *)cmd)->d[0] == 2915 TCP(ulp)->th_seq); 2916 break; 2917 2918 case O_TCPACK: 2919 match = (proto == IPPROTO_TCP && offset == 0 && 2920 ((ipfw_insn_u32 *)cmd)->d[0] == 2921 TCP(ulp)->th_ack); 2922 break; 2923 2924 case O_TCPWIN: 2925 match = (proto == IPPROTO_TCP && offset == 0 && 2926 cmd->arg1 == TCP(ulp)->th_win); 2927 break; 2928 2929 case O_ESTAB: 2930 /* reject packets which have SYN only */ 2931 /* XXX should i also check for TH_ACK ? */ 2932 match = (proto == IPPROTO_TCP && offset == 0 && 2933 (TCP(ulp)->th_flags & 2934 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2935 break; 2936 2937 case O_ALTQ: { 2938 struct pf_mtag *at; 2939 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2940 2941 match = 1; 2942 at = pf_find_mtag(m); 2943 if (at != NULL && at->qid != 0) 2944 break; 2945 at = pf_get_mtag(m); 2946 if (at == NULL) { 2947 /* 2948 * Let the packet fall back to the 2949 * default ALTQ. 2950 */ 2951 break; 2952 } 2953 at->qid = altq->qid; 2954 if (is_ipv4) 2955 at->af = AF_INET; 2956 else 2957 at->af = AF_LINK; 2958 at->hdr = ip; 2959 break; 2960 } 2961 2962 case O_LOG: 2963 if (V_fw_verbose) 2964 ipfw_log(f, hlen, args, m, 2965 oif, offset, tablearg, ip); 2966 match = 1; 2967 break; 2968 2969 case O_PROB: 2970 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2971 break; 2972 2973 case O_VERREVPATH: 2974 /* Outgoing packets automatically pass/match */ 2975 match = ((oif != NULL) || 2976 (m->m_pkthdr.rcvif == NULL) || 2977 ( 2978#ifdef INET6 2979 is_ipv6 ? 2980 verify_path6(&(args->f_id.src_ip6), 2981 m->m_pkthdr.rcvif) : 2982#endif 2983 verify_path(src_ip, m->m_pkthdr.rcvif, 2984 args->f_id.fib))); 2985 break; 2986 2987 case O_VERSRCREACH: 2988 /* Outgoing packets automatically pass/match */ 2989 match = (hlen > 0 && ((oif != NULL) || 2990#ifdef INET6 2991 is_ipv6 ? 2992 verify_path6(&(args->f_id.src_ip6), 2993 NULL) : 2994#endif 2995 verify_path(src_ip, NULL, args->f_id.fib))); 2996 break; 2997 2998 case O_ANTISPOOF: 2999 /* Outgoing packets automatically pass/match */ 3000 if (oif == NULL && hlen > 0 && 3001 ( (is_ipv4 && in_localaddr(src_ip)) 3002#ifdef INET6 3003 || (is_ipv6 && 3004 in6_localaddr(&(args->f_id.src_ip6))) 3005#endif 3006 )) 3007 match = 3008#ifdef INET6 3009 is_ipv6 ? verify_path6( 3010 &(args->f_id.src_ip6), 3011 m->m_pkthdr.rcvif) : 3012#endif 3013 verify_path(src_ip, 3014 m->m_pkthdr.rcvif, 3015 args->f_id.fib); 3016 else 3017 match = 1; 3018 break; 3019 3020 case O_IPSEC: 3021#ifdef IPSEC 3022 match = (m_tag_find(m, 3023 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 3024#endif 3025 /* otherwise no match */ 3026 break; 3027 3028#ifdef INET6 3029 case O_IP6_SRC: 3030 match = is_ipv6 && 3031 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 3032 &((ipfw_insn_ip6 *)cmd)->addr6); 3033 break; 3034 3035 case O_IP6_DST: 3036 match = is_ipv6 && 3037 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 3038 &((ipfw_insn_ip6 *)cmd)->addr6); 3039 break; 3040 case O_IP6_SRC_MASK: 3041 case O_IP6_DST_MASK: 3042 if (is_ipv6) { 3043 int i = cmdlen - 1; 3044 struct in6_addr p; 3045 struct in6_addr *d = 3046 &((ipfw_insn_ip6 *)cmd)->addr6; 3047 3048 for (; !match && i > 0; d += 2, 3049 i -= F_INSN_SIZE(struct in6_addr) 3050 * 2) { 3051 p = (cmd->opcode == 3052 O_IP6_SRC_MASK) ? 3053 args->f_id.src_ip6: 3054 args->f_id.dst_ip6; 3055 APPLY_MASK(&p, &d[1]); 3056 match = 3057 IN6_ARE_ADDR_EQUAL(&d[0], 3058 &p); 3059 } 3060 } 3061 break; 3062 3063 case O_IP6_SRC_ME: 3064 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 3065 break; 3066 3067 case O_IP6_DST_ME: 3068 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 3069 break; 3070 3071 case O_FLOW6ID: 3072 match = is_ipv6 && 3073 flow6id_match(args->f_id.flow_id6, 3074 (ipfw_insn_u32 *) cmd); 3075 break; 3076 3077 case O_EXT_HDR: 3078 match = is_ipv6 && 3079 (ext_hd & ((ipfw_insn *) cmd)->arg1); 3080 break; 3081 3082 case O_IP6: 3083 match = is_ipv6; 3084 break; 3085#endif 3086 3087 case O_IP4: 3088 match = is_ipv4; 3089 break; 3090 3091 case O_TAG: { 3092 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3093 tablearg : cmd->arg1; 3094 3095 /* Packet is already tagged with this tag? */ 3096 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 3097 3098 /* We have `untag' action when F_NOT flag is 3099 * present. And we must remove this mtag from 3100 * mbuf and reset `match' to zero (`match' will 3101 * be inversed later). 3102 * Otherwise we should allocate new mtag and 3103 * push it into mbuf. 3104 */ 3105 if (cmd->len & F_NOT) { /* `untag' action */ 3106 if (mtag != NULL) 3107 m_tag_delete(m, mtag); 3108 } else if (mtag == NULL) { 3109 if ((mtag = m_tag_alloc(MTAG_IPFW, 3110 tag, 0, M_NOWAIT)) != NULL) 3111 m_tag_prepend(m, mtag); 3112 } 3113 match = (cmd->len & F_NOT) ? 0: 1; 3114 break; 3115 } 3116 3117 case O_FIB: /* try match the specified fib */ 3118 if (args->f_id.fib == cmd->arg1) 3119 match = 1; 3120 break; 3121 3122 case O_TAGGED: { 3123 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3124 tablearg : cmd->arg1; 3125 3126 if (cmdlen == 1) { 3127 match = m_tag_locate(m, MTAG_IPFW, 3128 tag, NULL) != NULL; 3129 break; 3130 } 3131 3132 /* we have ranges */ 3133 for (mtag = m_tag_first(m); 3134 mtag != NULL && !match; 3135 mtag = m_tag_next(m, mtag)) { 3136 uint16_t *p; 3137 int i; 3138 3139 if (mtag->m_tag_cookie != MTAG_IPFW) 3140 continue; 3141 3142 p = ((ipfw_insn_u16 *)cmd)->ports; 3143 i = cmdlen - 1; 3144 for(; !match && i > 0; i--, p += 2) 3145 match = 3146 mtag->m_tag_id >= p[0] && 3147 mtag->m_tag_id <= p[1]; 3148 } 3149 break; 3150 } 3151 3152 /* 3153 * The second set of opcodes represents 'actions', 3154 * i.e. the terminal part of a rule once the packet 3155 * matches all previous patterns. 3156 * Typically there is only one action for each rule, 3157 * and the opcode is stored at the end of the rule 3158 * (but there are exceptions -- see below). 3159 * 3160 * In general, here we set retval and terminate the 3161 * outer loop (would be a 'break 3' in some language, 3162 * but we need to do a 'goto done'). 3163 * 3164 * Exceptions: 3165 * O_COUNT and O_SKIPTO actions: 3166 * instead of terminating, we jump to the next rule 3167 * ('goto next_rule', equivalent to a 'break 2'), 3168 * or to the SKIPTO target ('goto again' after 3169 * having set f, cmd and l), respectively. 3170 * 3171 * O_TAG, O_LOG and O_ALTQ action parameters: 3172 * perform some action and set match = 1; 3173 * 3174 * O_LIMIT and O_KEEP_STATE: these opcodes are 3175 * not real 'actions', and are stored right 3176 * before the 'action' part of the rule. 3177 * These opcodes try to install an entry in the 3178 * state tables; if successful, we continue with 3179 * the next opcode (match=1; break;), otherwise 3180 * the packet * must be dropped 3181 * ('goto done' after setting retval); 3182 * 3183 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3184 * cause a lookup of the state table, and a jump 3185 * to the 'action' part of the parent rule 3186 * ('goto check_body') if an entry is found, or 3187 * (CHECK_STATE only) a jump to the next rule if 3188 * the entry is not found ('goto next_rule'). 3189 * The result of the lookup is cached to make 3190 * further instances of these opcodes are 3191 * effectively NOPs. 3192 */ 3193 case O_LIMIT: 3194 case O_KEEP_STATE: 3195 if (install_state(f, 3196 (ipfw_insn_limit *)cmd, args, tablearg)) { 3197 retval = IP_FW_DENY; 3198 goto done; /* error/limit violation */ 3199 } 3200 match = 1; 3201 break; 3202 3203 case O_PROBE_STATE: 3204 case O_CHECK_STATE: 3205 /* 3206 * dynamic rules are checked at the first 3207 * keep-state or check-state occurrence, 3208 * with the result being stored in dyn_dir. 3209 * The compiler introduces a PROBE_STATE 3210 * instruction for us when we have a 3211 * KEEP_STATE (because PROBE_STATE needs 3212 * to be run first). 3213 */ 3214 if (dyn_dir == MATCH_UNKNOWN && 3215 (q = lookup_dyn_rule(&args->f_id, 3216 &dyn_dir, proto == IPPROTO_TCP ? 3217 TCP(ulp) : NULL)) 3218 != NULL) { 3219 /* 3220 * Found dynamic entry, update stats 3221 * and jump to the 'action' part of 3222 * the parent rule. 3223 */ 3224 q->pcnt++; 3225 q->bcnt += pktlen; 3226 f = q->rule; 3227 cmd = ACTION_PTR(f); 3228 l = f->cmd_len - f->act_ofs; 3229 IPFW_DYN_UNLOCK(); 3230 goto check_body; 3231 } 3232 /* 3233 * Dynamic entry not found. If CHECK_STATE, 3234 * skip to next rule, if PROBE_STATE just 3235 * ignore and continue with next opcode. 3236 */ 3237 if (cmd->opcode == O_CHECK_STATE) 3238 goto next_rule; 3239 match = 1; 3240 break; 3241 3242 case O_ACCEPT: 3243 retval = 0; /* accept */ 3244 goto done; 3245 3246 case O_PIPE: 3247 case O_QUEUE: 3248 args->rule = f; /* report matching rule */ 3249 args->rule_id = f->id; 3250 args->chain_id = chain->id; 3251 if (cmd->arg1 == IP_FW_TABLEARG) 3252 args->cookie = tablearg; 3253 else 3254 args->cookie = cmd->arg1; 3255 retval = IP_FW_DUMMYNET; 3256 goto done; 3257 3258 case O_DIVERT: 3259 case O_TEE: { 3260 struct divert_tag *dt; 3261 3262 if (args->eh) /* not on layer 2 */ 3263 break; 3264 mtag = m_tag_get(PACKET_TAG_DIVERT, 3265 sizeof(struct divert_tag), 3266 M_NOWAIT); 3267 if (mtag == NULL) { 3268 /* XXX statistic */ 3269 /* drop packet */ 3270 IPFW_RUNLOCK(chain); 3271 if (ucred_cache != NULL) 3272 crfree(ucred_cache); 3273 return (IP_FW_DENY); 3274 } 3275 dt = (struct divert_tag *)(mtag+1); 3276 dt->cookie = f->rulenum; 3277 if (cmd->arg1 == IP_FW_TABLEARG) 3278 dt->info = tablearg; 3279 else 3280 dt->info = cmd->arg1; 3281 m_tag_prepend(m, mtag); 3282 retval = (cmd->opcode == O_DIVERT) ? 3283 IP_FW_DIVERT : IP_FW_TEE; 3284 goto done; 3285 } 3286 case O_COUNT: 3287 case O_SKIPTO: 3288 f->pcnt++; /* update stats */ 3289 f->bcnt += pktlen; 3290 f->timestamp = time_uptime; 3291 if (cmd->opcode == O_COUNT) 3292 goto next_rule; 3293 /* handle skipto */ 3294 if (cmd->arg1 == IP_FW_TABLEARG) { 3295 f = lookup_next_rule(f, tablearg); 3296 } else { 3297 if (f->next_rule == NULL) 3298 lookup_next_rule(f, 0); 3299 f = f->next_rule; 3300 } 3301 goto again; 3302 3303 case O_REJECT: 3304 /* 3305 * Drop the packet and send a reject notice 3306 * if the packet is not ICMP (or is an ICMP 3307 * query), and it is not multicast/broadcast. 3308 */ 3309 if (hlen > 0 && is_ipv4 && offset == 0 && 3310 (proto != IPPROTO_ICMP || 3311 is_icmp_query(ICMP(ulp))) && 3312 !(m->m_flags & (M_BCAST|M_MCAST)) && 3313 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3314 send_reject(args, cmd->arg1, ip_len, ip); 3315 m = args->m; 3316 } 3317 /* FALLTHROUGH */ 3318#ifdef INET6 3319 case O_UNREACH6: 3320 if (hlen > 0 && is_ipv6 && 3321 ((offset & IP6F_OFF_MASK) == 0) && 3322 (proto != IPPROTO_ICMPV6 || 3323 (is_icmp6_query(args->f_id.flags) == 1)) && 3324 !(m->m_flags & (M_BCAST|M_MCAST)) && 3325 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3326 send_reject6( 3327 args, cmd->arg1, hlen, 3328 (struct ip6_hdr *)ip); 3329 m = args->m; 3330 } 3331 /* FALLTHROUGH */ 3332#endif 3333 case O_DENY: 3334 retval = IP_FW_DENY; 3335 goto done; 3336 3337 case O_FORWARD_IP: { 3338 struct sockaddr_in *sa; 3339 sa = &(((ipfw_insn_sa *)cmd)->sa); 3340 if (args->eh) /* not valid on layer2 pkts */ 3341 break; 3342 if (!q || dyn_dir == MATCH_FORWARD) { 3343 if (sa->sin_addr.s_addr == INADDR_ANY) { 3344 bcopy(sa, &args->hopstore, 3345 sizeof(*sa)); 3346 args->hopstore.sin_addr.s_addr = 3347 htonl(tablearg); 3348 args->next_hop = 3349 &args->hopstore; 3350 } else { 3351 args->next_hop = sa; 3352 } 3353 } 3354 retval = IP_FW_PASS; 3355 } 3356 goto done; 3357 3358 case O_NETGRAPH: 3359 case O_NGTEE: 3360 args->rule = f; /* report matching rule */ 3361 args->rule_id = f->id; 3362 args->chain_id = chain->id; 3363 if (cmd->arg1 == IP_FW_TABLEARG) 3364 args->cookie = tablearg; 3365 else 3366 args->cookie = cmd->arg1; 3367 retval = (cmd->opcode == O_NETGRAPH) ? 3368 IP_FW_NETGRAPH : IP_FW_NGTEE; 3369 goto done; 3370 3371 case O_SETFIB: 3372 f->pcnt++; /* update stats */ 3373 f->bcnt += pktlen; 3374 f->timestamp = time_uptime; 3375 M_SETFIB(m, cmd->arg1); 3376 args->f_id.fib = cmd->arg1; 3377 goto next_rule; 3378 3379 case O_NAT: { 3380 struct cfg_nat *t; 3381 int nat_id; 3382 3383 if (IPFW_NAT_LOADED) { 3384 args->rule = f; /* Report matching rule. */ 3385 args->rule_id = f->id; 3386 args->chain_id = chain->id; 3387 t = ((ipfw_insn_nat *)cmd)->nat; 3388 if (t == NULL) { 3389 nat_id = (cmd->arg1 == IP_FW_TABLEARG) ? 3390 tablearg : cmd->arg1; 3391 LOOKUP_NAT(V_layer3_chain, nat_id, t); 3392 if (t == NULL) { 3393 retval = IP_FW_DENY; 3394 goto done; 3395 } 3396 if (cmd->arg1 != IP_FW_TABLEARG) 3397 ((ipfw_insn_nat *)cmd)->nat = t; 3398 } 3399 retval = ipfw_nat_ptr(args, t, m); 3400 } else 3401 retval = IP_FW_DENY; 3402 goto done; 3403 } 3404 3405 case O_REASS: { 3406 int ip_off; 3407 3408 f->pcnt++; 3409 f->bcnt += pktlen; 3410 ip_off = (args->eh != NULL) ? ntohs(ip->ip_off) : ip->ip_off; 3411 if (ip_off & (IP_MF | IP_OFFMASK)) { 3412 /* 3413 * ip_reass() expects len & off in host 3414 * byte order: fix them in case we come 3415 * from layer2. 3416 */ 3417 if (args->eh != NULL) { 3418 ip->ip_len = ntohs(ip->ip_len); 3419 ip->ip_off = ntohs(ip->ip_off); 3420 } 3421 3422 m = ip_reass(m); 3423 args->m = m; 3424 3425 /* 3426 * IP header checksum fixup after 3427 * reassembly and leave header 3428 * in network byte order. 3429 */ 3430 if (m != NULL) { 3431 int hlen; 3432 3433 ip = mtod(m, struct ip *); 3434 hlen = ip->ip_hl << 2; 3435 /* revert len & off for layer2 pkts */ 3436 if (args->eh != NULL) 3437 ip->ip_len = htons(ip->ip_len); 3438 ip->ip_sum = 0; 3439 if (hlen == sizeof(struct ip)) 3440 ip->ip_sum = in_cksum_hdr(ip); 3441 else 3442 ip->ip_sum = in_cksum(m, hlen); 3443 retval = IP_FW_REASS; 3444 args->rule = f; 3445 args->rule_id = f->id; 3446 args->chain_id = chain->id; 3447 goto done; 3448 } else { 3449 retval = IP_FW_DENY; 3450 goto done; 3451 } 3452 } 3453 goto next_rule; 3454 } 3455 3456 default: 3457 panic("-- unknown opcode %d\n", cmd->opcode); 3458 } /* end of switch() on opcodes */ 3459 3460 if (cmd->len & F_NOT) 3461 match = !match; 3462 3463 if (match) { 3464 if (cmd->len & F_OR) 3465 skip_or = 1; 3466 } else { 3467 if (!(cmd->len & F_OR)) /* not an OR block, */ 3468 break; /* try next rule */ 3469 } 3470 3471 } /* end of inner for, scan opcodes */ 3472 3473next_rule:; /* try next rule */ 3474 3475 } /* end of outer for, scan rules */ 3476 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3477 IPFW_RUNLOCK(chain); 3478 if (ucred_cache != NULL) 3479 crfree(ucred_cache); 3480 return (IP_FW_DENY); 3481 3482done: 3483 /* Update statistics */ 3484 f->pcnt++; 3485 f->bcnt += pktlen; 3486 f->timestamp = time_uptime; 3487 IPFW_RUNLOCK(chain); 3488 if (ucred_cache != NULL) 3489 crfree(ucred_cache); 3490 return (retval); 3491 3492pullup_failed: 3493 if (V_fw_verbose) 3494 printf("ipfw: pullup failed\n"); 3495 return (IP_FW_DENY); 3496} 3497 3498/* 3499 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3500 * These will be reconstructed on the fly as packets are matched. 3501 */ 3502static void 3503flush_rule_ptrs(struct ip_fw_chain *chain) 3504{ 3505 struct ip_fw *rule; 3506 3507 IPFW_WLOCK_ASSERT(chain); 3508 3509 chain->id++; 3510 3511 for (rule = chain->rules; rule; rule = rule->next) 3512 rule->next_rule = NULL; 3513} 3514 3515/* 3516 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3517 * possibly create a rule number and add the rule to the list. 3518 * Update the rule_number in the input struct so the caller knows it as well. 3519 */ 3520static int 3521add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3522{ 3523 struct ip_fw *rule, *f, *prev; 3524 int l = RULESIZE(input_rule); 3525 3526 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3527 return (EINVAL); 3528 3529 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3530 if (rule == NULL) 3531 return (ENOSPC); 3532 3533 bcopy(input_rule, rule, l); 3534 3535 rule->next = NULL; 3536 rule->next_rule = NULL; 3537 3538 rule->pcnt = 0; 3539 rule->bcnt = 0; 3540 rule->timestamp = 0; 3541 3542 IPFW_WLOCK(chain); 3543 3544 if (chain->rules == NULL) { /* default rule */ 3545 chain->rules = rule; 3546 rule->id = ++chain->id; 3547 goto done; 3548 } 3549 3550 /* 3551 * If rulenum is 0, find highest numbered rule before the 3552 * default rule, and add autoinc_step 3553 */ 3554 if (V_autoinc_step < 1) 3555 V_autoinc_step = 1; 3556 else if (V_autoinc_step > 1000) 3557 V_autoinc_step = 1000; 3558 if (rule->rulenum == 0) { 3559 /* 3560 * locate the highest numbered rule before default 3561 */ 3562 for (f = chain->rules; f; f = f->next) { 3563 if (f->rulenum == IPFW_DEFAULT_RULE) 3564 break; 3565 rule->rulenum = f->rulenum; 3566 } 3567 if (rule->rulenum < IPFW_DEFAULT_RULE - V_autoinc_step) 3568 rule->rulenum += V_autoinc_step; 3569 input_rule->rulenum = rule->rulenum; 3570 } 3571 3572 /* 3573 * Now insert the new rule in the right place in the sorted list. 3574 */ 3575 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3576 if (f->rulenum > rule->rulenum) { /* found the location */ 3577 if (prev) { 3578 rule->next = f; 3579 prev->next = rule; 3580 } else { /* head insert */ 3581 rule->next = chain->rules; 3582 chain->rules = rule; 3583 } 3584 break; 3585 } 3586 } 3587 flush_rule_ptrs(chain); 3588 /* chain->id incremented inside flush_rule_ptrs() */ 3589 rule->id = chain->id; 3590done: 3591 V_static_count++; 3592 V_static_len += l; 3593 IPFW_WUNLOCK(chain); 3594 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3595 rule->rulenum, V_static_count);) 3596 return (0); 3597} 3598 3599/** 3600 * Remove a static rule (including derived * dynamic rules) 3601 * and place it on the ``reap list'' for later reclamation. 3602 * The caller is in charge of clearing rule pointers to avoid 3603 * dangling pointers. 3604 * @return a pointer to the next entry. 3605 * Arguments are not checked, so they better be correct. 3606 */ 3607static struct ip_fw * 3608remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, 3609 struct ip_fw *prev) 3610{ 3611 struct ip_fw *n; 3612 int l = RULESIZE(rule); 3613 3614 IPFW_WLOCK_ASSERT(chain); 3615 3616 n = rule->next; 3617 IPFW_DYN_LOCK(); 3618 remove_dyn_rule(rule, NULL /* force removal */); 3619 IPFW_DYN_UNLOCK(); 3620 if (prev == NULL) 3621 chain->rules = n; 3622 else 3623 prev->next = n; 3624 V_static_count--; 3625 V_static_len -= l; 3626 3627 rule->next = chain->reap; 3628 chain->reap = rule; 3629 3630 return n; 3631} 3632 3633/* 3634 * Reclaim storage associated with a list of rules. This is 3635 * typically the list created using remove_rule. 3636 * A NULL pointer on input is handled correctly. 3637 */ 3638static void 3639reap_rules(struct ip_fw *head) 3640{ 3641 struct ip_fw *rule; 3642 3643 while ((rule = head) != NULL) { 3644 head = head->next; 3645 free(rule, M_IPFW); 3646 } 3647} 3648 3649/* 3650 * Remove all rules from a chain (except rules in set RESVD_SET 3651 * unless kill_default = 1). The caller is responsible for 3652 * reclaiming storage for the rules left in chain->reap. 3653 */ 3654static void 3655free_chain(struct ip_fw_chain *chain, int kill_default) 3656{ 3657 struct ip_fw *prev, *rule; 3658 3659 IPFW_WLOCK_ASSERT(chain); 3660 3661 chain->reap = NULL; 3662 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3663 for (prev = NULL, rule = chain->rules; rule ; ) 3664 if (kill_default || rule->set != RESVD_SET) 3665 rule = remove_rule(chain, rule, prev); 3666 else { 3667 prev = rule; 3668 rule = rule->next; 3669 } 3670} 3671 3672/** 3673 * Remove all rules with given number, and also do set manipulation. 3674 * Assumes chain != NULL && *chain != NULL. 3675 * 3676 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3677 * the next 8 bits are the new set, the top 8 bits are the command: 3678 * 3679 * 0 delete rules with given number 3680 * 1 delete rules with given set number 3681 * 2 move rules with given number to new set 3682 * 3 move rules with given set number to new set 3683 * 4 swap sets with given numbers 3684 * 5 delete rules with given number and with given set number 3685 */ 3686static int 3687del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3688{ 3689 struct ip_fw *prev = NULL, *rule; 3690 u_int16_t rulenum; /* rule or old_set */ 3691 u_int8_t cmd, new_set; 3692 3693 rulenum = arg & 0xffff; 3694 cmd = (arg >> 24) & 0xff; 3695 new_set = (arg >> 16) & 0xff; 3696 3697 if (cmd > 5 || new_set > RESVD_SET) 3698 return EINVAL; 3699 if (cmd == 0 || cmd == 2 || cmd == 5) { 3700 if (rulenum >= IPFW_DEFAULT_RULE) 3701 return EINVAL; 3702 } else { 3703 if (rulenum > RESVD_SET) /* old_set */ 3704 return EINVAL; 3705 } 3706 3707 IPFW_WLOCK(chain); 3708 rule = chain->rules; /* common starting point */ 3709 chain->reap = NULL; /* prepare for deletions */ 3710 switch (cmd) { 3711 case 0: /* delete rules with given number */ 3712 /* 3713 * locate first rule to delete 3714 */ 3715 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3716 ; 3717 if (rule->rulenum != rulenum) { 3718 IPFW_WUNLOCK(chain); 3719 return EINVAL; 3720 } 3721 3722 /* 3723 * flush pointers outside the loop, then delete all matching 3724 * rules. prev remains the same throughout the cycle. 3725 */ 3726 flush_rule_ptrs(chain); 3727 while (rule->rulenum == rulenum) 3728 rule = remove_rule(chain, rule, prev); 3729 break; 3730 3731 case 1: /* delete all rules with given set number */ 3732 flush_rule_ptrs(chain); 3733 while (rule->rulenum < IPFW_DEFAULT_RULE) { 3734 if (rule->set == rulenum) 3735 rule = remove_rule(chain, rule, prev); 3736 else { 3737 prev = rule; 3738 rule = rule->next; 3739 } 3740 } 3741 break; 3742 3743 case 2: /* move rules with given number to new set */ 3744 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3745 if (rule->rulenum == rulenum) 3746 rule->set = new_set; 3747 break; 3748 3749 case 3: /* move rules with given set number to new set */ 3750 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3751 if (rule->set == rulenum) 3752 rule->set = new_set; 3753 break; 3754 3755 case 4: /* swap two sets */ 3756 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3757 if (rule->set == rulenum) 3758 rule->set = new_set; 3759 else if (rule->set == new_set) 3760 rule->set = rulenum; 3761 break; 3762 3763 case 5: /* delete rules with given number and with given set number. 3764 * rulenum - given rule number; 3765 * new_set - given set number. 3766 */ 3767 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3768 ; 3769 if (rule->rulenum != rulenum) { 3770 IPFW_WUNLOCK(chain); 3771 return (EINVAL); 3772 } 3773 flush_rule_ptrs(chain); 3774 while (rule->rulenum == rulenum) { 3775 if (rule->set == new_set) 3776 rule = remove_rule(chain, rule, prev); 3777 else { 3778 prev = rule; 3779 rule = rule->next; 3780 } 3781 } 3782 } 3783 /* 3784 * Look for rules to reclaim. We grab the list before 3785 * releasing the lock then reclaim them w/o the lock to 3786 * avoid a LOR with dummynet. 3787 */ 3788 rule = chain->reap; 3789 IPFW_WUNLOCK(chain); 3790 reap_rules(rule); 3791 return 0; 3792} 3793 3794/* 3795 * Clear counters for a specific rule. 3796 * The enclosing "table" is assumed locked. 3797 */ 3798static void 3799clear_counters(struct ip_fw *rule, int log_only) 3800{ 3801 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3802 3803 if (log_only == 0) { 3804 rule->bcnt = rule->pcnt = 0; 3805 rule->timestamp = 0; 3806 } 3807 if (l->o.opcode == O_LOG) 3808 l->log_left = l->max_log; 3809} 3810 3811/** 3812 * Reset some or all counters on firewall rules. 3813 * The argument `arg' is an u_int32_t. The low 16 bit are the rule number, 3814 * the next 8 bits are the set number, the top 8 bits are the command: 3815 * 0 work with rules from all set's; 3816 * 1 work with rules only from specified set. 3817 * Specified rule number is zero if we want to clear all entries. 3818 * log_only is 1 if we only want to reset logs, zero otherwise. 3819 */ 3820static int 3821zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only) 3822{ 3823 struct ip_fw *rule; 3824 char *msg; 3825 3826 uint16_t rulenum = arg & 0xffff; 3827 uint8_t set = (arg >> 16) & 0xff; 3828 uint8_t cmd = (arg >> 24) & 0xff; 3829 3830 if (cmd > 1) 3831 return (EINVAL); 3832 if (cmd == 1 && set > RESVD_SET) 3833 return (EINVAL); 3834 3835 IPFW_WLOCK(chain); 3836 if (rulenum == 0) { 3837 V_norule_counter = 0; 3838 for (rule = chain->rules; rule; rule = rule->next) { 3839 /* Skip rules from another set. */ 3840 if (cmd == 1 && rule->set != set) 3841 continue; 3842 clear_counters(rule, log_only); 3843 } 3844 msg = log_only ? "All logging counts reset" : 3845 "Accounting cleared"; 3846 } else { 3847 int cleared = 0; 3848 /* 3849 * We can have multiple rules with the same number, so we 3850 * need to clear them all. 3851 */ 3852 for (rule = chain->rules; rule; rule = rule->next) 3853 if (rule->rulenum == rulenum) { 3854 while (rule && rule->rulenum == rulenum) { 3855 if (cmd == 0 || rule->set == set) 3856 clear_counters(rule, log_only); 3857 rule = rule->next; 3858 } 3859 cleared = 1; 3860 break; 3861 } 3862 if (!cleared) { /* we did not find any matching rules */ 3863 IPFW_WUNLOCK(chain); 3864 return (EINVAL); 3865 } 3866 msg = log_only ? "logging count reset" : "cleared"; 3867 } 3868 IPFW_WUNLOCK(chain); 3869 3870 if (V_fw_verbose) { 3871 int lev = LOG_SECURITY | LOG_NOTICE; 3872 3873 if (rulenum) 3874 log(lev, "ipfw: Entry %d %s.\n", rulenum, msg); 3875 else 3876 log(lev, "ipfw: %s.\n", msg); 3877 } 3878 return (0); 3879} 3880 3881/* 3882 * Check validity of the structure before insert. 3883 * Fortunately rules are simple, so this mostly need to check rule sizes. 3884 */ 3885static int 3886check_ipfw_struct(struct ip_fw *rule, int size) 3887{ 3888 int l, cmdlen = 0; 3889 int have_action=0; 3890 ipfw_insn *cmd; 3891 3892 if (size < sizeof(*rule)) { 3893 printf("ipfw: rule too short\n"); 3894 return (EINVAL); 3895 } 3896 /* first, check for valid size */ 3897 l = RULESIZE(rule); 3898 if (l != size) { 3899 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3900 return (EINVAL); 3901 } 3902 if (rule->act_ofs >= rule->cmd_len) { 3903 printf("ipfw: bogus action offset (%u > %u)\n", 3904 rule->act_ofs, rule->cmd_len - 1); 3905 return (EINVAL); 3906 } 3907 /* 3908 * Now go for the individual checks. Very simple ones, basically only 3909 * instruction sizes. 3910 */ 3911 for (l = rule->cmd_len, cmd = rule->cmd ; 3912 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3913 cmdlen = F_LEN(cmd); 3914 if (cmdlen > l) { 3915 printf("ipfw: opcode %d size truncated\n", 3916 cmd->opcode); 3917 return EINVAL; 3918 } 3919 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3920 switch (cmd->opcode) { 3921 case O_PROBE_STATE: 3922 case O_KEEP_STATE: 3923 case O_PROTO: 3924 case O_IP_SRC_ME: 3925 case O_IP_DST_ME: 3926 case O_LAYER2: 3927 case O_IN: 3928 case O_FRAG: 3929 case O_DIVERTED: 3930 case O_IPOPT: 3931 case O_IPTOS: 3932 case O_IPPRECEDENCE: 3933 case O_IPVER: 3934 case O_TCPWIN: 3935 case O_TCPFLAGS: 3936 case O_TCPOPTS: 3937 case O_ESTAB: 3938 case O_VERREVPATH: 3939 case O_VERSRCREACH: 3940 case O_ANTISPOOF: 3941 case O_IPSEC: 3942#ifdef INET6 3943 case O_IP6_SRC_ME: 3944 case O_IP6_DST_ME: 3945 case O_EXT_HDR: 3946 case O_IP6: 3947#endif 3948 case O_IP4: 3949 case O_TAG: 3950 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3951 goto bad_size; 3952 break; 3953 3954 case O_FIB: 3955 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3956 goto bad_size; 3957 if (cmd->arg1 >= rt_numfibs) { 3958 printf("ipfw: invalid fib number %d\n", 3959 cmd->arg1); 3960 return EINVAL; 3961 } 3962 break; 3963 3964 case O_SETFIB: 3965 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3966 goto bad_size; 3967 if (cmd->arg1 >= rt_numfibs) { 3968 printf("ipfw: invalid fib number %d\n", 3969 cmd->arg1); 3970 return EINVAL; 3971 } 3972 goto check_action; 3973 3974 case O_UID: 3975 case O_GID: 3976 case O_JAIL: 3977 case O_IP_SRC: 3978 case O_IP_DST: 3979 case O_TCPSEQ: 3980 case O_TCPACK: 3981 case O_PROB: 3982 case O_ICMPTYPE: 3983 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3984 goto bad_size; 3985 break; 3986 3987 case O_LIMIT: 3988 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3989 goto bad_size; 3990 break; 3991 3992 case O_LOG: 3993 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3994 goto bad_size; 3995 3996 ((ipfw_insn_log *)cmd)->log_left = 3997 ((ipfw_insn_log *)cmd)->max_log; 3998 3999 break; 4000 4001 case O_IP_SRC_MASK: 4002 case O_IP_DST_MASK: 4003 /* only odd command lengths */ 4004 if ( !(cmdlen & 1) || cmdlen > 31) 4005 goto bad_size; 4006 break; 4007 4008 case O_IP_SRC_SET: 4009 case O_IP_DST_SET: 4010 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 4011 printf("ipfw: invalid set size %d\n", 4012 cmd->arg1); 4013 return EINVAL; 4014 } 4015 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 4016 (cmd->arg1+31)/32 ) 4017 goto bad_size; 4018 break; 4019 4020 case O_IP_SRC_LOOKUP: 4021 case O_IP_DST_LOOKUP: 4022 if (cmd->arg1 >= IPFW_TABLES_MAX) { 4023 printf("ipfw: invalid table number %d\n", 4024 cmd->arg1); 4025 return (EINVAL); 4026 } 4027 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 4028 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 4029 goto bad_size; 4030 break; 4031 4032 case O_MACADDR2: 4033 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 4034 goto bad_size; 4035 break; 4036 4037 case O_NOP: 4038 case O_IPID: 4039 case O_IPTTL: 4040 case O_IPLEN: 4041 case O_TCPDATALEN: 4042 case O_TAGGED: 4043 if (cmdlen < 1 || cmdlen > 31) 4044 goto bad_size; 4045 break; 4046 4047 case O_MAC_TYPE: 4048 case O_IP_SRCPORT: 4049 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 4050 if (cmdlen < 2 || cmdlen > 31) 4051 goto bad_size; 4052 break; 4053 4054 case O_RECV: 4055 case O_XMIT: 4056 case O_VIA: 4057 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 4058 goto bad_size; 4059 break; 4060 4061 case O_ALTQ: 4062 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 4063 goto bad_size; 4064 break; 4065 4066 case O_PIPE: 4067 case O_QUEUE: 4068 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 4069 goto bad_size; 4070 goto check_action; 4071 4072 case O_FORWARD_IP: 4073#ifdef IPFIREWALL_FORWARD 4074 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 4075 goto bad_size; 4076 goto check_action; 4077#else 4078 return EINVAL; 4079#endif 4080 4081 case O_DIVERT: 4082 case O_TEE: 4083 if (ip_divert_ptr == NULL) 4084 return EINVAL; 4085 else 4086 goto check_size; 4087 case O_NETGRAPH: 4088 case O_NGTEE: 4089 if (!NG_IPFW_LOADED) 4090 return EINVAL; 4091 else 4092 goto check_size; 4093 case O_NAT: 4094 if (!IPFW_NAT_LOADED) 4095 return EINVAL; 4096 if (cmdlen != F_INSN_SIZE(ipfw_insn_nat)) 4097 goto bad_size; 4098 goto check_action; 4099 case O_FORWARD_MAC: /* XXX not implemented yet */ 4100 case O_CHECK_STATE: 4101 case O_COUNT: 4102 case O_ACCEPT: 4103 case O_DENY: 4104 case O_REJECT: 4105#ifdef INET6 4106 case O_UNREACH6: 4107#endif 4108 case O_SKIPTO: 4109 case O_REASS: 4110check_size: 4111 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 4112 goto bad_size; 4113check_action: 4114 if (have_action) { 4115 printf("ipfw: opcode %d, multiple actions" 4116 " not allowed\n", 4117 cmd->opcode); 4118 return EINVAL; 4119 } 4120 have_action = 1; 4121 if (l != cmdlen) { 4122 printf("ipfw: opcode %d, action must be" 4123 " last opcode\n", 4124 cmd->opcode); 4125 return EINVAL; 4126 } 4127 break; 4128#ifdef INET6 4129 case O_IP6_SRC: 4130 case O_IP6_DST: 4131 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 4132 F_INSN_SIZE(ipfw_insn)) 4133 goto bad_size; 4134 break; 4135 4136 case O_FLOW6ID: 4137 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 4138 ((ipfw_insn_u32 *)cmd)->o.arg1) 4139 goto bad_size; 4140 break; 4141 4142 case O_IP6_SRC_MASK: 4143 case O_IP6_DST_MASK: 4144 if ( !(cmdlen & 1) || cmdlen > 127) 4145 goto bad_size; 4146 break; 4147 case O_ICMP6TYPE: 4148 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 4149 goto bad_size; 4150 break; 4151#endif 4152 4153 default: 4154 switch (cmd->opcode) { 4155#ifndef INET6 4156 case O_IP6_SRC_ME: 4157 case O_IP6_DST_ME: 4158 case O_EXT_HDR: 4159 case O_IP6: 4160 case O_UNREACH6: 4161 case O_IP6_SRC: 4162 case O_IP6_DST: 4163 case O_FLOW6ID: 4164 case O_IP6_SRC_MASK: 4165 case O_IP6_DST_MASK: 4166 case O_ICMP6TYPE: 4167 printf("ipfw: no IPv6 support in kernel\n"); 4168 return EPROTONOSUPPORT; 4169#endif 4170 default: 4171 printf("ipfw: opcode %d, unknown opcode\n", 4172 cmd->opcode); 4173 return EINVAL; 4174 } 4175 } 4176 } 4177 if (have_action == 0) { 4178 printf("ipfw: missing action\n"); 4179 return EINVAL; 4180 } 4181 return 0; 4182 4183bad_size: 4184 printf("ipfw: opcode %d size %d wrong\n", 4185 cmd->opcode, cmdlen); 4186 return EINVAL; 4187} 4188 4189/* 4190 * Copy the static and dynamic rules to the supplied buffer 4191 * and return the amount of space actually used. 4192 */ 4193static size_t 4194ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 4195{ 4196 char *bp = buf; 4197 char *ep = bp + space; 4198 struct ip_fw *rule; 4199 int i; 4200 time_t boot_seconds; 4201 4202 boot_seconds = boottime.tv_sec; 4203 /* XXX this can take a long time and locking will block packet flow */ 4204 IPFW_RLOCK(chain); 4205 for (rule = chain->rules; rule ; rule = rule->next) { 4206 /* 4207 * Verify the entry fits in the buffer in case the 4208 * rules changed between calculating buffer space and 4209 * now. This would be better done using a generation 4210 * number but should suffice for now. 4211 */ 4212 i = RULESIZE(rule); 4213 if (bp + i <= ep) { 4214 bcopy(rule, bp, i); 4215 /* 4216 * XXX HACK. Store the disable mask in the "next" 4217 * pointer in a wild attempt to keep the ABI the same. 4218 * Why do we do this on EVERY rule? 4219 */ 4220 bcopy(&V_set_disable, 4221 &(((struct ip_fw *)bp)->next_rule), 4222 sizeof(V_set_disable)); 4223 if (((struct ip_fw *)bp)->timestamp) 4224 ((struct ip_fw *)bp)->timestamp += boot_seconds; 4225 bp += i; 4226 } 4227 } 4228 IPFW_RUNLOCK(chain); 4229 if (V_ipfw_dyn_v) { 4230 ipfw_dyn_rule *p, *last = NULL; 4231 4232 IPFW_DYN_LOCK(); 4233 for (i = 0 ; i < V_curr_dyn_buckets; i++) 4234 for (p = V_ipfw_dyn_v[i] ; p != NULL; p = p->next) { 4235 if (bp + sizeof *p <= ep) { 4236 ipfw_dyn_rule *dst = 4237 (ipfw_dyn_rule *)bp; 4238 bcopy(p, dst, sizeof *p); 4239 bcopy(&(p->rule->rulenum), &(dst->rule), 4240 sizeof(p->rule->rulenum)); 4241 /* 4242 * store set number into high word of 4243 * dst->rule pointer. 4244 */ 4245 bcopy(&(p->rule->set), 4246 (char *)&dst->rule + 4247 sizeof(p->rule->rulenum), 4248 sizeof(p->rule->set)); 4249 /* 4250 * store a non-null value in "next". 4251 * The userland code will interpret a 4252 * NULL here as a marker 4253 * for the last dynamic rule. 4254 */ 4255 bcopy(&dst, &dst->next, sizeof(dst)); 4256 last = dst; 4257 dst->expire = 4258 TIME_LEQ(dst->expire, time_uptime) ? 4259 0 : dst->expire - time_uptime ; 4260 bp += sizeof(ipfw_dyn_rule); 4261 } 4262 } 4263 IPFW_DYN_UNLOCK(); 4264 if (last != NULL) /* mark last dynamic rule */ 4265 bzero(&last->next, sizeof(last)); 4266 } 4267 return (bp - (char *)buf); 4268} 4269 4270 4271/** 4272 * {set|get}sockopt parser. 4273 */ 4274static int 4275ipfw_ctl(struct sockopt *sopt) 4276{ 4277#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 4278 int error; 4279 size_t size; 4280 struct ip_fw *buf, *rule; 4281 u_int32_t rulenum[2]; 4282 4283 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW); 4284 if (error) 4285 return (error); 4286 4287 /* 4288 * Disallow modifications in really-really secure mode, but still allow 4289 * the logging counters to be reset. 4290 */ 4291 if (sopt->sopt_name == IP_FW_ADD || 4292 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 4293 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 4294 if (error) 4295 return (error); 4296 } 4297 4298 error = 0; 4299 4300 switch (sopt->sopt_name) { 4301 case IP_FW_GET: 4302 /* 4303 * pass up a copy of the current rules. Static rules 4304 * come first (the last of which has number IPFW_DEFAULT_RULE), 4305 * followed by a possibly empty list of dynamic rule. 4306 * The last dynamic rule has NULL in the "next" field. 4307 * 4308 * Note that the calculated size is used to bound the 4309 * amount of data returned to the user. The rule set may 4310 * change between calculating the size and returning the 4311 * data in which case we'll just return what fits. 4312 */ 4313 size = V_static_len; /* size of static rules */ 4314 if (V_ipfw_dyn_v) /* add size of dyn.rules */ 4315 size += (V_dyn_count * sizeof(ipfw_dyn_rule)); 4316 4317 if (size >= sopt->sopt_valsize) 4318 break; 4319 /* 4320 * XXX todo: if the user passes a short length just to know 4321 * how much room is needed, do not bother filling up the 4322 * buffer, just jump to the sooptcopyout. 4323 */ 4324 buf = malloc(size, M_TEMP, M_WAITOK); 4325 error = sooptcopyout(sopt, buf, 4326 ipfw_getrules(&V_layer3_chain, buf, size)); 4327 free(buf, M_TEMP); 4328 break; 4329 4330 case IP_FW_FLUSH: 4331 /* 4332 * Normally we cannot release the lock on each iteration. 4333 * We could do it here only because we start from the head all 4334 * the times so there is no risk of missing some entries. 4335 * On the other hand, the risk is that we end up with 4336 * a very inconsistent ruleset, so better keep the lock 4337 * around the whole cycle. 4338 * 4339 * XXX this code can be improved by resetting the head of 4340 * the list to point to the default rule, and then freeing 4341 * the old list without the need for a lock. 4342 */ 4343 4344 IPFW_WLOCK(&V_layer3_chain); 4345 free_chain(&V_layer3_chain, 0 /* keep default rule */); 4346 rule = V_layer3_chain.reap; 4347 IPFW_WUNLOCK(&V_layer3_chain); 4348 reap_rules(rule); 4349 break; 4350 4351 case IP_FW_ADD: 4352 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 4353 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 4354 sizeof(struct ip_fw) ); 4355 if (error == 0) 4356 error = check_ipfw_struct(rule, sopt->sopt_valsize); 4357 if (error == 0) { 4358 error = add_rule(&V_layer3_chain, rule); 4359 size = RULESIZE(rule); 4360 if (!error && sopt->sopt_dir == SOPT_GET) 4361 error = sooptcopyout(sopt, rule, size); 4362 } 4363 free(rule, M_TEMP); 4364 break; 4365 4366 case IP_FW_DEL: 4367 /* 4368 * IP_FW_DEL is used for deleting single rules or sets, 4369 * and (ab)used to atomically manipulate sets. Argument size 4370 * is used to distinguish between the two: 4371 * sizeof(u_int32_t) 4372 * delete single rule or set of rules, 4373 * or reassign rules (or sets) to a different set. 4374 * 2*sizeof(u_int32_t) 4375 * atomic disable/enable sets. 4376 * first u_int32_t contains sets to be disabled, 4377 * second u_int32_t contains sets to be enabled. 4378 */ 4379 error = sooptcopyin(sopt, rulenum, 4380 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4381 if (error) 4382 break; 4383 size = sopt->sopt_valsize; 4384 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4385 error = del_entry(&V_layer3_chain, rulenum[0]); 4386 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4387 V_set_disable = 4388 (V_set_disable | rulenum[0]) & ~rulenum[1] & 4389 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4390 else 4391 error = EINVAL; 4392 break; 4393 4394 case IP_FW_ZERO: 4395 case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */ 4396 rulenum[0] = 0; 4397 if (sopt->sopt_val != 0) { 4398 error = sooptcopyin(sopt, rulenum, 4399 sizeof(u_int32_t), sizeof(u_int32_t)); 4400 if (error) 4401 break; 4402 } 4403 error = zero_entry(&V_layer3_chain, rulenum[0], 4404 sopt->sopt_name == IP_FW_RESETLOG); 4405 break; 4406 4407 case IP_FW_TABLE_ADD: 4408 { 4409 ipfw_table_entry ent; 4410 4411 error = sooptcopyin(sopt, &ent, 4412 sizeof(ent), sizeof(ent)); 4413 if (error) 4414 break; 4415 error = add_table_entry(&V_layer3_chain, ent.tbl, 4416 ent.addr, ent.masklen, ent.value); 4417 } 4418 break; 4419 4420 case IP_FW_TABLE_DEL: 4421 { 4422 ipfw_table_entry ent; 4423 4424 error = sooptcopyin(sopt, &ent, 4425 sizeof(ent), sizeof(ent)); 4426 if (error) 4427 break; 4428 error = del_table_entry(&V_layer3_chain, ent.tbl, 4429 ent.addr, ent.masklen); 4430 } 4431 break; 4432 4433 case IP_FW_TABLE_FLUSH: 4434 { 4435 u_int16_t tbl; 4436 4437 error = sooptcopyin(sopt, &tbl, 4438 sizeof(tbl), sizeof(tbl)); 4439 if (error) 4440 break; 4441 IPFW_WLOCK(&V_layer3_chain); 4442 error = flush_table(&V_layer3_chain, tbl); 4443 IPFW_WUNLOCK(&V_layer3_chain); 4444 } 4445 break; 4446 4447 case IP_FW_TABLE_GETSIZE: 4448 { 4449 u_int32_t tbl, cnt; 4450 4451 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4452 sizeof(tbl)))) 4453 break; 4454 IPFW_RLOCK(&V_layer3_chain); 4455 error = count_table(&V_layer3_chain, tbl, &cnt); 4456 IPFW_RUNLOCK(&V_layer3_chain); 4457 if (error) 4458 break; 4459 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4460 } 4461 break; 4462 4463 case IP_FW_TABLE_LIST: 4464 { 4465 ipfw_table *tbl; 4466 4467 if (sopt->sopt_valsize < sizeof(*tbl)) { 4468 error = EINVAL; 4469 break; 4470 } 4471 size = sopt->sopt_valsize; 4472 tbl = malloc(size, M_TEMP, M_WAITOK); 4473 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4474 if (error) { 4475 free(tbl, M_TEMP); 4476 break; 4477 } 4478 tbl->size = (size - sizeof(*tbl)) / 4479 sizeof(ipfw_table_entry); 4480 IPFW_RLOCK(&V_layer3_chain); 4481 error = dump_table(&V_layer3_chain, tbl); 4482 IPFW_RUNLOCK(&V_layer3_chain); 4483 if (error) { 4484 free(tbl, M_TEMP); 4485 break; 4486 } 4487 error = sooptcopyout(sopt, tbl, size); 4488 free(tbl, M_TEMP); 4489 } 4490 break; 4491 4492 case IP_FW_NAT_CFG: 4493 if (IPFW_NAT_LOADED) 4494 error = ipfw_nat_cfg_ptr(sopt); 4495 else { 4496 printf("IP_FW_NAT_CFG: %s\n", 4497 "ipfw_nat not present, please load it"); 4498 error = EINVAL; 4499 } 4500 break; 4501 4502 case IP_FW_NAT_DEL: 4503 if (IPFW_NAT_LOADED) 4504 error = ipfw_nat_del_ptr(sopt); 4505 else { 4506 printf("IP_FW_NAT_DEL: %s\n", 4507 "ipfw_nat not present, please load it"); 4508 error = EINVAL; 4509 } 4510 break; 4511 4512 case IP_FW_NAT_GET_CONFIG: 4513 if (IPFW_NAT_LOADED) 4514 error = ipfw_nat_get_cfg_ptr(sopt); 4515 else { 4516 printf("IP_FW_NAT_GET_CFG: %s\n", 4517 "ipfw_nat not present, please load it"); 4518 error = EINVAL; 4519 } 4520 break; 4521 4522 case IP_FW_NAT_GET_LOG: 4523 if (IPFW_NAT_LOADED) 4524 error = ipfw_nat_get_log_ptr(sopt); 4525 else { 4526 printf("IP_FW_NAT_GET_LOG: %s\n", 4527 "ipfw_nat not present, please load it"); 4528 error = EINVAL; 4529 } 4530 break; 4531 4532 default: 4533 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4534 error = EINVAL; 4535 } 4536 4537 return (error); 4538#undef RULE_MAXSIZE 4539} 4540 4541 4542/* 4543 * This procedure is only used to handle keepalives. It is invoked 4544 * every dyn_keepalive_period 4545 */ 4546static void 4547ipfw_tick(void * vnetx) 4548{ 4549 struct mbuf *m0, *m, *mnext, **mtailp; 4550#ifdef INET6 4551 struct mbuf *m6, **m6_tailp; 4552#endif 4553 int i; 4554 ipfw_dyn_rule *q; 4555#ifdef VIMAGE 4556 struct vnet *vp = vnetx; 4557#endif 4558 4559 CURVNET_SET(vp); 4560 if (V_dyn_keepalive == 0 || V_ipfw_dyn_v == NULL || V_dyn_count == 0) 4561 goto done; 4562 4563 /* 4564 * We make a chain of packets to go out here -- not deferring 4565 * until after we drop the IPFW dynamic rule lock would result 4566 * in a lock order reversal with the normal packet input -> ipfw 4567 * call stack. 4568 */ 4569 m0 = NULL; 4570 mtailp = &m0; 4571#ifdef INET6 4572 m6 = NULL; 4573 m6_tailp = &m6; 4574#endif 4575 IPFW_DYN_LOCK(); 4576 for (i = 0 ; i < V_curr_dyn_buckets ; i++) { 4577 for (q = V_ipfw_dyn_v[i] ; q ; q = q->next ) { 4578 if (q->dyn_type == O_LIMIT_PARENT) 4579 continue; 4580 if (q->id.proto != IPPROTO_TCP) 4581 continue; 4582 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4583 continue; 4584 if (TIME_LEQ(time_uptime + V_dyn_keepalive_interval, 4585 q->expire)) 4586 continue; /* too early */ 4587 if (TIME_LEQ(q->expire, time_uptime)) 4588 continue; /* too late, rule expired */ 4589 4590 m = send_pkt(NULL, &(q->id), q->ack_rev - 1, 4591 q->ack_fwd, TH_SYN); 4592 mnext = send_pkt(NULL, &(q->id), q->ack_fwd - 1, 4593 q->ack_rev, 0); 4594 4595 switch (q->id.addr_type) { 4596 case 4: 4597 if (m != NULL) { 4598 *mtailp = m; 4599 mtailp = &(*mtailp)->m_nextpkt; 4600 } 4601 if (mnext != NULL) { 4602 *mtailp = mnext; 4603 mtailp = &(*mtailp)->m_nextpkt; 4604 } 4605 break; 4606#ifdef INET6 4607 case 6: 4608 if (m != NULL) { 4609 *m6_tailp = m; 4610 m6_tailp = &(*m6_tailp)->m_nextpkt; 4611 } 4612 if (mnext != NULL) { 4613 *m6_tailp = mnext; 4614 m6_tailp = &(*m6_tailp)->m_nextpkt; 4615 } 4616 break; 4617#endif 4618 } 4619 4620 m = mnext = NULL; 4621 } 4622 } 4623 IPFW_DYN_UNLOCK(); 4624 for (m = mnext = m0; m != NULL; m = mnext) { 4625 mnext = m->m_nextpkt; 4626 m->m_nextpkt = NULL; 4627 ip_output(m, NULL, NULL, 0, NULL, NULL); 4628 } 4629#ifdef INET6 4630 for (m = mnext = m6; m != NULL; m = mnext) { 4631 mnext = m->m_nextpkt; 4632 m->m_nextpkt = NULL; 4633 ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL); 4634 } 4635#endif 4636done: 4637 callout_reset(&V_ipfw_timeout, V_dyn_keepalive_period * hz, 4638 ipfw_tick, vnetx); 4639 CURVNET_RESTORE(); 4640} 4641 4642/**************** 4643 * Stuff that must be initialised only on boot or module load 4644 */ 4645static int 4646ipfw_init(void) 4647{ 4648 int error = 0; 4649 4650 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule", 4651 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4652 UMA_ALIGN_PTR, 0); 4653 4654 IPFW_DYN_LOCK_INIT(); 4655 /* 4656 * Only print out this stuff the first time around, 4657 * when called from the sysinit code. 4658 */ 4659 printf("ipfw2 " 4660#ifdef INET6 4661 "(+ipv6) " 4662#endif 4663 "initialized, divert %s, nat %s, " 4664 "rule-based forwarding " 4665#ifdef IPFIREWALL_FORWARD 4666 "enabled, " 4667#else 4668 "disabled, " 4669#endif 4670 "default to %s, logging ", 4671#ifdef IPDIVERT 4672 "enabled", 4673#else 4674 "loadable", 4675#endif 4676#ifdef IPFIREWALL_NAT 4677 "enabled", 4678#else 4679 "loadable", 4680#endif 4681 default_to_accept ? "accept" : "deny"); 4682 4683 /* 4684 * Note: V_xxx variables can be accessed here but the vnet specific 4685 * initializer may not have been called yet for the VIMAGE case. 4686 * Tuneables will have been processed. We will print out values for 4687 * the default vnet. 4688 * XXX This should all be rationalized AFTER 8.0 4689 */ 4690 if (V_fw_verbose == 0) 4691 printf("disabled\n"); 4692 else if (V_verbose_limit == 0) 4693 printf("unlimited\n"); 4694 else 4695 printf("limited to %d packets/entry by default\n", 4696 V_verbose_limit); 4697 4698 return (error); 4699} 4700 4701/********************** 4702 * Called for the removal of the last instance only on module unload. 4703 */ 4704static void 4705ipfw_destroy(void) 4706{ 4707 4708 uma_zdestroy(ipfw_dyn_rule_zone); 4709 IPFW_DYN_LOCK_DESTROY(); 4710 printf("IP firewall unloaded\n"); 4711} 4712 4713/**************** 4714 * Stuff that must be initialized for every instance 4715 * (including the first of course). 4716 */ 4717static int 4718vnet_ipfw_init(const void *unused) 4719{ 4720 int error; 4721 struct ip_fw default_rule; 4722 4723 /* First set up some values that are compile time options */ 4724#ifdef IPFIREWALL_VERBOSE 4725 V_fw_verbose = 1; 4726#endif 4727#ifdef IPFIREWALL_VERBOSE_LIMIT 4728 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4729#endif 4730 4731 error = init_tables(&V_layer3_chain); 4732 if (error) { 4733 panic("init_tables"); /* XXX Marko fix this ! */ 4734 } 4735#ifdef IPFIREWALL_NAT 4736 LIST_INIT(&V_layer3_chain.nat); 4737#endif 4738 4739 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 4740 4741 V_ipfw_dyn_v = NULL; 4742 V_dyn_buckets = 256; /* must be power of 2 */ 4743 V_curr_dyn_buckets = 256; /* must be power of 2 */ 4744 4745 V_dyn_ack_lifetime = 300; 4746 V_dyn_syn_lifetime = 20; 4747 V_dyn_fin_lifetime = 1; 4748 V_dyn_rst_lifetime = 1; 4749 V_dyn_udp_lifetime = 10; 4750 V_dyn_short_lifetime = 5; 4751 4752 V_dyn_keepalive_interval = 20; 4753 V_dyn_keepalive_period = 5; 4754 V_dyn_keepalive = 1; /* do send keepalives */ 4755 4756 V_dyn_max = 4096; /* max # of dynamic rules */ 4757 4758 V_fw_deny_unknown_exthdrs = 1; 4759 4760 V_layer3_chain.rules = NULL; 4761 IPFW_LOCK_INIT(&V_layer3_chain); 4762 callout_init(&V_ipfw_timeout, CALLOUT_MPSAFE); 4763 4764 bzero(&default_rule, sizeof default_rule); 4765 default_rule.act_ofs = 0; 4766 default_rule.rulenum = IPFW_DEFAULT_RULE; 4767 default_rule.cmd_len = 1; 4768 default_rule.set = RESVD_SET; 4769 default_rule.cmd[0].len = 1; 4770 default_rule.cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; 4771 error = add_rule(&V_layer3_chain, &default_rule); 4772 4773 if (error != 0) { 4774 printf("ipfw2: error %u initializing default rule " 4775 "(support disabled)\n", error); 4776 IPFW_LOCK_DESTROY(&V_layer3_chain); 4777 printf("leaving ipfw_iattach (1) with error %d\n", error); 4778 return (error); 4779 } 4780 4781 ip_fw_default_rule = V_layer3_chain.rules; 4782 4783 if (error) { 4784 IPFW_LOCK_DESTROY(&V_layer3_chain); 4785 printf("leaving ipfw_iattach (2) with error %d\n", error); 4786 return (error); 4787 } 4788#ifdef VIMAGE /* want a better way to do this */ 4789 callout_reset(&V_ipfw_timeout, hz, ipfw_tick, curvnet); 4790#else 4791 callout_reset(&V_ipfw_timeout, hz, ipfw_tick, NULL); 4792#endif 4793 4794 /* First set up some values that are compile time options */ 4795 V_ipfw_vnet_ready = 1; /* Open for business */ 4796 4797 /* Hook up the raw inputs */ 4798 V_ip_fw_ctl_ptr = ipfw_ctl; 4799 V_ip_fw_chk_ptr = ipfw_chk; 4800 4801 /* 4802 * Hook us up to pfil. 4803 */ 4804 if (V_fw_enable) { 4805 if ((error = ipfw_hook()) != 0) { 4806 printf("ipfw_hook() error\n"); 4807 return (error); 4808 } 4809 } 4810#ifdef INET6 4811 if (V_fw6_enable) { 4812 if ((error = ipfw6_hook()) != 0) { 4813 printf("ipfw6_hook() error\n"); 4814 /* XXX should we unhook everything else? */ 4815 return (error); 4816 } 4817 } 4818#endif 4819 return (0); 4820} 4821 4822/*********************** 4823 * Called for the removal of each instance. 4824 */ 4825static int 4826vnet_ipfw_uninit(const void *unused) 4827{ 4828 struct ip_fw *reap; 4829 4830 V_ipfw_vnet_ready = 0; /* tell new callers to go away */ 4831 ipfw_unhook(); 4832#ifdef INET6 4833 ipfw6_unhook(); 4834#endif 4835 /* layer2 and other entrypoints still come in this way. */ 4836 V_ip_fw_chk_ptr = NULL; 4837 V_ip_fw_ctl_ptr = NULL; 4838 IPFW_WLOCK(&V_layer3_chain); 4839 /* We wait on the wlock here until the last user leaves */ 4840 IPFW_WUNLOCK(&V_layer3_chain); 4841 IPFW_WLOCK(&V_layer3_chain); 4842 callout_drain(&V_ipfw_timeout); 4843 flush_tables(&V_layer3_chain); 4844 V_layer3_chain.reap = NULL; 4845 free_chain(&V_layer3_chain, 1 /* kill default rule */); 4846 reap = V_layer3_chain.reap; 4847 V_layer3_chain.reap = NULL; 4848 IPFW_WUNLOCK(&V_layer3_chain); 4849 if (reap != NULL) 4850 reap_rules(reap); 4851 IPFW_LOCK_DESTROY(&V_layer3_chain); 4852 if (V_ipfw_dyn_v != NULL) 4853 free(V_ipfw_dyn_v, M_IPFW); 4854 return 0; 4855} 4856 4857/* 4858 * Module event handler. 4859 * In general we have the choice of handling most of these events by the 4860 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to 4861 * use the SYSINIT handlers as they are more capable of expressing the 4862 * flow of control during module and vnet operations, so this is just 4863 * a skeleton. Note there is no SYSINIT equivalent of the module 4864 * SHUTDOWN handler, but we don't have anything to do in that case anyhow. 4865 */ 4866static int 4867ipfw_modevent(module_t mod, int type, void *unused) 4868{ 4869 int err = 0; 4870 4871 switch (type) { 4872 case MOD_LOAD: 4873 /* Called once at module load or 4874 * system boot if compiled in. */ 4875 break; 4876 case MOD_QUIESCE: 4877 /* Called before unload. May veto unloading. */ 4878 break; 4879 case MOD_UNLOAD: 4880 /* Called during unload. */ 4881 break; 4882 case MOD_SHUTDOWN: 4883 /* Called during system shutdown. */ 4884 break; 4885 default: 4886 err = EOPNOTSUPP; 4887 break; 4888 } 4889 return err; 4890} 4891 4892static moduledata_t ipfwmod = { 4893 "ipfw", 4894 ipfw_modevent, 4895 0 4896}; 4897 4898/* Define startup order. */ 4899#define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_IFATTACHDOMAIN 4900#define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */ 4901#define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */ 4902#define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */ 4903 4904DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER); 4905MODULE_VERSION(ipfw, 2); 4906/* should declare some dependencies here */ 4907 4908/* 4909 * Starting up. Done in order after ipfwmod() has been called. 4910 * VNET_SYSINIT is also called for each existing vnet and each new vnet. 4911 */ 4912SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 4913 ipfw_init, NULL); 4914VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 4915 vnet_ipfw_init, NULL); 4916 4917/* 4918 * Closing up shop. These are done in REVERSE ORDER, but still 4919 * after ipfwmod() has been called. Not called on reboot. 4920 * VNET_SYSUNINIT is also called for each exiting vnet as it exits. 4921 * or when the module is unloaded. 4922 */ 4923SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 4924 ipfw_destroy, NULL); 4925VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 4926 vnet_ipfw_uninit, NULL); 4927 4928