ip_fw2.c revision 141383
11541Srgrimes/*- 21541Srgrimes * Copyright (c) 2002 Luigi Rizzo, Universita` di Pisa 31541Srgrimes * 41541Srgrimes * Redistribution and use in source and binary forms, with or without 51541Srgrimes * modification, are permitted provided that the following conditions 61541Srgrimes * are met: 71541Srgrimes * 1. Redistributions of source code must retain the above copyright 81541Srgrimes * notice, this list of conditions and the following disclaimer. 91541Srgrimes * 2. Redistributions in binary form must reproduce the above copyright 101541Srgrimes * notice, this list of conditions and the following disclaimer in the 111541Srgrimes * documentation and/or other materials provided with the distribution. 121541Srgrimes * 131541Srgrimes * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 141541Srgrimes * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 151541Srgrimes * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 161541Srgrimes * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 171541Srgrimes * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 181541Srgrimes * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 191541Srgrimes * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 201541Srgrimes * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 211541Srgrimes * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 221541Srgrimes * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 231541Srgrimes * SUCH DAMAGE. 241541Srgrimes * 251541Srgrimes * $FreeBSD: head/sys/netinet/ip_fw2.c 141383 2005-02-06 11:13:59Z glebius $ 261541Srgrimes */ 271541Srgrimes 281541Srgrimes#define DEB(x) 291541Srgrimes#define DDB(x) x 3050477Speter 311541Srgrimes/* 321541Srgrimes * Implement IP packet firewall (new version) 33131840Sbrian */ 3416333Sgpalmer 3541793Sluigi#if !defined(KLD_MODULE) 3630966Sjoerg#include "opt_ipfw.h" 3734746Speter#include "opt_ipdn.h" 3850561Sdes#include "opt_inet.h" 3955009Sshin#include "opt_ipsec.h" 40101095Srwatson#ifndef INET 4164060Sdarrenr#error IPFIREWALL requires INET. 4277574Skris#endif /* INET */ 4316333Sgpalmer#endif 441541Srgrimes 451541Srgrimes#define IPFW2 1 46101095Srwatson#if IPFW2 471541Srgrimes#include <sys/param.h> 4841793Sluigi#include <sys/systm.h> 491541Srgrimes#include <sys/condvar.h> 501541Srgrimes#include <sys/malloc.h> 511541Srgrimes#include <sys/mbuf.h> 521541Srgrimes#include <sys/kernel.h> 531541Srgrimes#include <sys/jail.h> 547091Swollman#include <sys/module.h> 557090Sbde#include <sys/proc.h> 561541Srgrimes#include <sys/socket.h> 5764075Sache#include <sys/socketvar.h> 581541Srgrimes#include <sys/sysctl.h> 5983934Sbrooks#include <sys/syslog.h> 6030948Sjulian#include <sys/ucred.h> 6119622Sfenner#include <net/if.h> 621541Srgrimes#include <net/radix.h> 638426Swollman#include <net/route.h> 641541Srgrimes#include <netinet/in.h> 651541Srgrimes#include <netinet/in_systm.h> 661541Srgrimes#include <netinet/in_var.h> 677090Sbde#include <netinet/in_pcb.h> 681541Srgrimes#include <netinet/ip.h> 691541Srgrimes#include <netinet/ip_var.h> 701541Srgrimes#include <netinet/ip_icmp.h> 711541Srgrimes#include <netinet/ip_fw.h> 7218797Swollman#include <netinet/ip_divert.h> 731541Srgrimes#include <netinet/ip_dummynet.h> 742531Swollman#include <netinet/tcp.h> 7515680Sgpalmer#include <netinet/tcp_timer.h> 7615680Sgpalmer#include <netinet/tcp_var.h> 77126239Smlaier#include <netinet/tcpip.h> 7884516Sps#include <netinet/udp.h> 7915680Sgpalmer#include <netinet/udp_var.h> 8055009Sshin 8155009Sshin#include <netgraph/ng_ipfw.h> 8255009Sshin 8355009Sshin#include <altq/if_altq.h> 8455009Sshin 85105199Ssam#ifdef IPSEC 86105199Ssam#include <netinet6/ipsec.h> 87105199Ssam#endif 88105199Ssam 89105199Ssam#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 909209Swollman 912531Swollman#include <machine/in_cksum.h> /* XXX for in_cksum */ 9236192Sdg 9312296Sphk/* 9446381Sbillf * set_disable contains one bit per set value (0..31). 9512296Sphk * If the bit is set, all rules with the corresponding set 9613266Swollman * are disabled. Set RESVD_SET(31) is reserved for the default rule 9712296Sphk * and rules that are not deleted by the flush command, 9846381Sbillf * and CANNOT be disabled. 9912296Sphk * Rules in set RESVD_SET can only be deleted explicitly. 1001541Srgrimes */ 10112296Sphkstatic u_int32_t set_disable; 10246381Sbillf 10312296Sphkstatic int fw_verbose; 10412296Sphkstatic int verbose_limit; 10512296Sphk 10646381Sbillfstatic struct callout ipfw_timeout; 10733440Sguidostatic uma_zone_t ipfw_dyn_rule_zone; 10833440Sguido#define IPFW_DEFAULT_RULE 65535 10946381Sbillf 11046381Sbillf/* 11146381Sbillf * Data structure to cache our ucred related 11255009Sshin * information. This structure only gets used if 113129017Sandre * the user specified UID/GID based constraints in 114129017Sandre * a firewall rule. 115129017Sandre */ 116129017Sandrestruct ip_fw_ugid { 11755009Sshin gid_t fw_groups[NGROUPS]; 11855009Sshin int fw_ngroups; 11955009Sshin uid_t fw_uid; 12055009Sshin int fw_prid; 12155009Sshin}; 122109843Ssilby 123109843Ssilbystruct ip_fw_chain { 12477665Sjesper struct ip_fw *rules; /* list of rules */ 125109843Ssilby struct ip_fw *reap; /* list of rules to reap */ 12677665Sjesper struct mtx mtx; /* lock guarding rule list */ 12777665Sjesper int busy_count; /* busy count for rw locks */ 128111244Ssilby int want_write; 129111244Ssilby struct cv cv; 130111244Ssilby}; 131111244Ssilby#define IPFW_LOCK_INIT(_chain) \ 132111244Ssilby mtx_init(&(_chain)->mtx, "IPFW static rules", NULL, \ 133107081Ssilby MTX_DEF | MTX_RECURSE) 134107081Ssilby#define IPFW_LOCK_DESTROY(_chain) mtx_destroy(&(_chain)->mtx) 135107081Ssilby#define IPFW_WLOCK_ASSERT(_chain) do { \ 136107081Ssilby mtx_assert(&(_chain)->mtx, MA_OWNED); \ 137107081Ssilby NET_ASSERT_GIANT(); \ 13873399Struckman} while (0) 13973399Struckman 14073399Struckmanstatic __inline void 14173791StruckmanIPFW_RLOCK(struct ip_fw_chain *chain) 14273399Struckman{ 14373626Struckman mtx_lock(&chain->mtx); 14473791Struckman chain->busy_count++; 14573791Struckman mtx_unlock(&chain->mtx); 14673626Struckman} 14773626Struckman 14873626Struckmanstatic __inline void 14973626StruckmanIPFW_RUNLOCK(struct ip_fw_chain *chain) 15073399Struckman{ 15173357Sjlemon mtx_lock(&chain->mtx); 15273357Sjlemon chain->busy_count--; 15373357Sjlemon if (chain->busy_count == 0 && chain->want_write) 15473357Sjlemon cv_signal(&chain->cv); 1551541Srgrimes mtx_unlock(&chain->mtx); 15612296Sphk} 1571541Srgrimes 158120386Ssamstatic __inline void 159120386SsamIPFW_WLOCK(struct ip_fw_chain *chain) 160120386Ssam{ 1611541Srgrimes mtx_lock(&chain->mtx); 162111888Sjlemon chain->want_write++; 16384102Sjlemon while (chain->busy_count > 0) 16484102Sjlemon cv_wait(&chain->cv, &chain->mtx); 1651541Srgrimes} 16682884Sjulian 1671541Srgrimesstatic __inline void 16884102SjlemonIPFW_WUNLOCK(struct ip_fw_chain *chain) 16984102Sjlemon{ 17084102Sjlemon chain->want_write--; 17184102Sjlemon cv_signal(&chain->cv); 17247546Sdg mtx_unlock(&chain->mtx); 17346381Sbillf} 17412296Sphk 17546381Sbillf/* 1761541Srgrimes * list of rules for layer 3 1772112Swollman */ 17878667Srustatic struct ip_fw_chain layer3_chain; 17946381Sbillf 1802112SwollmanMALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 18129480SacheMALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 18229480Sache 18329480Sachestruct table_entry { 18429480Sache struct radix_node rn[2]; 18529480Sache struct sockaddr_in addr, mask; 18643305Sdillon u_int32_t value; 18729480Sache}; 18874362Sphk 189119753Ssam#define IPFW_TABLES_MAX 128 19029480Sachestatic struct { 191119753Ssam struct radix_node_head *rnh; 192119753Ssam int modified; 193121091Ssam} ipfw_tables[IPFW_TABLES_MAX]; 194121091Ssam 195119753Ssamstatic int fw_debug = 1; 19612296Sphkstatic int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 19712296Sphk 19846381Sbillf#ifdef SYSCTL_NODE 19912296SphkSYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 20012296SphkSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, enable, 20144219Sdes CTLFLAG_RW | CTLFLAG_SECURE3, 202122723Sandre &fw_enable, 0, "Enable ipfw"); 20344219SdesSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 20444219Sdes &autoinc_step, 0, "Rule number autincrement step"); 20544219SdesSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 20644219Sdes CTLFLAG_RW | CTLFLAG_SECURE3, 20718797Swollman &fw_one_pass, 0, 20815026Sphk "Only do a single pass through ipfw when using dummynet(4)"); 20915026SphkSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 21057117Sluigi &fw_debug, 0, "Enable printing of debug ip_fw statements"); 211107114SluigiSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 21214209Sphk CTLFLAG_RW | CTLFLAG_SECURE3, 21384516Sps &fw_verbose, 0, "Log matches to ipfw rules"); 21484516SpsSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 21541793Sluigi &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 216121093Ssam 21798701Sluigi/* 21898701Sluigi * Description of dynamic rules. 21998701Sluigi * 22098701Sluigi * Dynamic rules are stored in lists accessed through a hash table 22198701Sluigi * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 22298701Sluigi * be modified through the sysctl variable dyn_buckets which is 2231541Srgrimes * updated when the table becomes empty. 2241541Srgrimes * 2251541Srgrimes * XXX currently there is only one list, ipfw_dyn. 2261541Srgrimes * 2271541Srgrimes * When a packet is received, its address fields are first masked 228133390Sandre * with the mask defined for the rule, then hashed, then matched 2291541Srgrimes * against the entries in the corresponding list. 23012296Sphk * Dynamic rules can be used for different purposes: 2311541Srgrimes * + stateful rules; 2321541Srgrimes * + enforcing limits on the number of sessions; 2331541Srgrimes * + in-kernel NAT (not implemented yet) 2341541Srgrimes * 2351541Srgrimes * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 2361541Srgrimes * measured in seconds and depending on the flags. 2371541Srgrimes * 23892723Salfred * The total number of dynamic rules is stored in dyn_count. 23998613Sluigi * The max number of dynamic rules is dyn_max. When we reach 24098613Sluigi * the maximum number of rules we do not create anymore. This is 241122708Sandre * done to avoid consuming too much memory, but also too much 24298613Sluigi * time when searching on each packet (ideally, we should try instead 24392723Salfred * to put a limit on the length of the list on each bucket...). 244133069Sandre * 24517072Sjulian * Each dynamic rule holds a pointer to the parent ipfw rule so 2461541Srgrimes * we know what action to perform. Dynamic rules are removed when 2471541Srgrimes * the parent rule is deleted. XXX we should make them survive. 2481541Srgrimes * 2491541Srgrimes * There are some limitations with dynamic rules -- we do not 2501541Srgrimes * obey the 'randomized match', and we do not do multiple 2511541Srgrimes * passes through the firewall. XXX check the latter!!! 2521541Srgrimes */ 25382884Sjulianstatic ipfw_dyn_rule **ipfw_dyn_v = NULL; 2541541Srgrimesstatic u_int32_t dyn_buckets = 256; /* must be power of 2 */ 2551541Srgrimesstatic u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 25620407Swollman 25784102Sjlemonstatic struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 25882884Sjulian#define IPFW_DYN_LOCK_INIT() \ 2591541Srgrimes mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 2601541Srgrimes#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 2611541Srgrimes#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 2621541Srgrimes#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 26382884Sjulian#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 26482884Sjulian 2651541Srgrimes/* 2661541Srgrimes * Timeouts for various events in handing dynamic rules. 2671541Srgrimes */ 26829480Sachestatic u_int32_t dyn_ack_lifetime = 300; 269120386Ssamstatic u_int32_t dyn_syn_lifetime = 20; 270120386Ssamstatic u_int32_t dyn_fin_lifetime = 1; 271120386Ssamstatic u_int32_t dyn_rst_lifetime = 1; 272120386Ssamstatic u_int32_t dyn_udp_lifetime = 10; 273120386Ssamstatic u_int32_t dyn_short_lifetime = 5; 274120386Ssam 275120386Ssam/* 276120386Ssam * Keepalives are sent if dyn_keepalive is set. They are sent every 277119753Ssam * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 27829480Sache * seconds of lifetime of a rule. 27974362Sphk * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 28029480Sache * than dyn_keepalive_period. 281111244Ssilby */ 282111244Ssilby 28329480Sachestatic u_int32_t dyn_keepalive_interval = 20; 28477574Skrisstatic u_int32_t dyn_keepalive_period = 5; 28534961Sphkstatic u_int32_t dyn_keepalive = 1; /* do send keepalives */ 28677574Skris 2871541Srgrimesstatic u_int32_t static_count; /* # of static rules */ 28893818Sjhbstatic u_int32_t static_len; /* size in bytes of static rules */ 289122320Ssamstatic u_int32_t dyn_count; /* # of dynamic rules */ 2901541Srgrimesstatic u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 2911541Srgrimes 29298701SluigiSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 2931541Srgrimes &dyn_buckets, 0, "Number of dyn. buckets"); 2941541SrgrimesSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 2951541Srgrimes &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 29613929SwollmanSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 29713929Swollman &dyn_count, 0, "Number of dyn. rules"); 2981541SrgrimesSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 299122702Sandre &dyn_max, 0, "Max number of dyn. rules"); 30067334SjoeSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 30184102Sjlemon &static_count, 0, "Number of static rules"); 302122702SandreSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 30322333Sbrian &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 30473172SjlemonSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 305126239Smlaier &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 306126239SmlaierSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 307126239Smlaier &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 30898613SluigiSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 309122708Sandre &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 310121141SsamSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 311121141Ssam &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 312121141SsamSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 313105199Ssam &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 314125952SmlaierSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 315105199Ssam &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 316105199Ssam 317105199Ssam#endif /* SYSCTL_NODE */ 318105199Ssam 3191541Srgrimes 32098613Sluigi/* 32198613Sluigi * This macro maps an ip pointer into a layer3 header pointer of type T 32254175Sarchie */ 323125952Smlaier#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 324126239Smlaier 325128829Sdarrenrstatic __inline int 326126239Smlaiericmptype_match(struct ip *ip, ipfw_insn_u32 *cmd) 327125952Smlaier{ 328126239Smlaier int type = L3HDR(struct icmp,ip)->icmp_type; 329126239Smlaier 330126239Smlaier return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 331125952Smlaier} 332126239Smlaier 333125952Smlaier#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 334125952Smlaier (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 335125952Smlaier 336125952Smlaierstatic int 33798613Sluigiis_icmp_query(struct ip *ip) 33898613Sluigi{ 33998613Sluigi int type = L3HDR(struct icmp, ip)->icmp_type; 3401541Srgrimes return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 34118797Swollman} 34218797Swollman#undef TT 34318797Swollman 34418797Swollman/* 3451541Srgrimes * The following checks use two arrays of 8 or 16 bits to store the 346133481Sandre * bits that we want set or clear, respectively. They are in the 3471541Srgrimes * low and high half of cmd->arg1 or cmd->d[0]. 34813929Swollman * 3491541Srgrimes * We scan options and store the bits we find set. We succeed if 3501541Srgrimes * 35118797Swollman * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 352105586Sphk * 3531541Srgrimes * The code is sometimes optimized not to store additional variables. 3541541Srgrimes */ 3551541Srgrimes 35618797Swollmanstatic int 357105586Sphkflags_match(ipfw_insn *cmd, u_int8_t bits) 3581541Srgrimes{ 3591541Srgrimes u_char want_clear; 3601541Srgrimes bits = ~bits; 3611541Srgrimes 3621541Srgrimes if ( ((cmd->arg1 & 0xff) & bits) != 0) 363133481Sandre return 0; /* some bits we want set were clear */ 3641541Srgrimes want_clear = (cmd->arg1 >> 8) & 0xff; 36513929Swollman if ( (want_clear & bits) != want_clear) 3661541Srgrimes return 0; /* some bits we want clear were set */ 3671541Srgrimes return 1; 3681541Srgrimes} 36978064Sume 37078064Sumestatic int 37178064Sumeipopts_match(struct ip *ip, ipfw_insn *cmd) 37278064Sume{ 37378064Sume int optlen, bits = 0; 37478064Sume u_char *cp = (u_char *)(ip + 1); 37578064Sume int x = (ip->ip_hl << 2) - sizeof (struct ip); 37678064Sume 37778064Sume for (; x > 0; x -= optlen, cp += optlen) { 37878064Sume int opt = cp[IPOPT_OPTVAL]; 37958698Sjlemon 38058698Sjlemon if (opt == IPOPT_EOL) 38118797Swollman break; 38258698Sjlemon if (opt == IPOPT_NOP) 38358698Sjlemon optlen = 1; 38458698Sjlemon else { 38558698Sjlemon optlen = cp[IPOPT_OLEN]; 38658698Sjlemon if (optlen <= 0 || optlen > x) 38718797Swollman return 0; /* invalid or truncated */ 38822333Sbrian } 3891541Srgrimes switch (opt) { 3901541Srgrimes 3911541Srgrimes default: 3921541Srgrimes break; 393130416Smlaier 394130416Smlaier case IPOPT_LSRR: 395130416Smlaier bits |= IP_FW_IPOPT_LSRR; 396130416Smlaier break; 397130416Smlaier 398130416Smlaier case IPOPT_SSRR: 3991541Srgrimes bits |= IP_FW_IPOPT_SSRR; 4001541Srgrimes break; 4011541Srgrimes 40290868Smike case IPOPT_RR: 4031541Srgrimes bits |= IP_FW_IPOPT_RR; 4041541Srgrimes break; 4051541Srgrimes 4061541Srgrimes case IPOPT_TS: 40790868Smike bits |= IP_FW_IPOPT_TS; 4081541Srgrimes break; 4091541Srgrimes } 4101541Srgrimes } 4111541Srgrimes return (flags_match(cmd, bits)); 4121541Srgrimes} 4131541Srgrimes 4141541Srgrimesstatic int 4151541Srgrimestcpopts_match(struct ip *ip, ipfw_insn *cmd) 41618797Swollman{ 4171541Srgrimes int optlen, bits = 0; 4181541Srgrimes struct tcphdr *tcp = L3HDR(struct tcphdr,ip); 4191541Srgrimes u_char *cp = (u_char *)(tcp + 1); 4201541Srgrimes int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 4211541Srgrimes 4221541Srgrimes for (; x > 0; x -= optlen, cp += optlen) { 4231541Srgrimes int opt = cp[0]; 4241541Srgrimes if (opt == TCPOPT_EOL) 4251541Srgrimes break; 4261541Srgrimes if (opt == TCPOPT_NOP) 427111275Ssam optlen = 1; 428111275Ssam else { 429111275Ssam optlen = cp[1]; 430111275Ssam if (optlen <= 0) 431122062Sume break; 432111275Ssam } 433111275Ssam 434117897Ssam switch (opt) { 435117897Ssam 436117897Ssam default: 437117897Ssam break; 438117897Ssam 439117897Ssam case TCPOPT_MAXSEG: 440117897Ssam bits |= IP_FW_TCPOPT_MSS; 44173626Struckman break; 44273626Struckman 4435543Sugen case TCPOPT_WINDOW: 4445543Sugen bits |= IP_FW_TCPOPT_WINDOW; 4455543Sugen break; 4465543Sugen 44717072Sjulian case TCPOPT_SACK_PERMITTED: 44817758Ssos case TCPOPT_SACK: 44941793Sluigi bits |= IP_FW_TCPOPT_SACK; 4505543Sugen break; 4515543Sugen 4525543Sugen case TCPOPT_TIMESTAMP: 45341793Sluigi bits |= IP_FW_TCPOPT_TS; 45441793Sluigi break; 45564060Sdarrenr 45664060Sdarrenr } 45724590Sdarrenr } 458120386Ssam return (flags_match(cmd, bits)); 459121141Ssam} 460121141Ssam 461121141Ssamstatic int 462121141Ssamiface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 46324590Sdarrenr{ 464121141Ssam if (ifp == NULL) /* no iface with this packet, match fails */ 465120386Ssam return 0; 466120386Ssam /* Check by name or by IP address */ 467120386Ssam if (cmd->name[0] != '\0') { /* match by name */ 468120386Ssam /* Check name */ 469120386Ssam if (cmd->p.glob) { 470120386Ssam if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 471122708Sandre return(1); 47264060Sdarrenr } else { 4731541Srgrimes if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 47486047Sluigi return(1); 47537409Sjulian } 47637409Sjulian } else { 47737409Sjulian struct ifaddr *ia; 47837409Sjulian 47998613Sluigi /* XXX lock? */ 48037409Sjulian TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 48198613Sluigi if (ia->ifa_addr == NULL) 48298613Sluigi continue; 48398613Sluigi if (ia->ifa_addr->sa_family != AF_INET) 48498613Sluigi continue; 48598613Sluigi if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 48696245Sluigi (ia->ifa_addr))->sin_addr.s_addr) 48787915Sjlemon return(1); /* match */ 48887915Sjlemon } 48987915Sjlemon } 49071909Sluigi return(0); /* no match, fail ... */ 49196245Sluigi} 49298613Sluigi 49354175Sarchie/* 49486047Sluigi * The verify_path function checks if a route to the src exists and 49598666Sluigi * if it is reachable via ifp (when provided). 49698666Sluigi * 49754175Sarchie * The 'verrevpath' option checks that the interface that an IP packet 49841793Sluigi * arrives on is the same interface that traffic destined for the 49926359Sjulian * packet's source address would be routed out of. The 'versrcreach' 50054175Sarchie * option just checks that the source address is reachable via any route 50154175Sarchie * (except default) in the routing table. These two are a measure to block 50226359Sjulian * forged packets. This is also commonly known as "anti-spoofing" or Unicast 50326359Sjulian * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 50441793Sluigi * is purposely reminiscent of the Cisco IOS command, 50598613Sluigi * 50654175Sarchie * ip verify unicast reverse-path 50737409Sjulian * ip verify unicast source reachable-via any 50841793Sluigi * 50937409Sjulian * which implements the same functionality. But note that syntax is 51054175Sarchie * misleading. The check may be performed on all IP packets whether unicast, 51154175Sarchie * multicast, or broadcast. 51217072Sjulian */ 51341793Sluigistatic int 51417072Sjulianverify_path(struct in_addr src, struct ifnet *ifp) 5151541Srgrimes{ 5161541Srgrimes struct route ro; 5171541Srgrimes struct sockaddr_in *dst; 5181541Srgrimes 5191541Srgrimes bzero(&ro, sizeof(ro)); 5201541Srgrimes 5211541Srgrimes dst = (struct sockaddr_in *)&(ro.ro_dst); 52298613Sluigi dst->sin_family = AF_INET; 52313929Swollman dst->sin_len = sizeof(*dst); 5241541Srgrimes dst->sin_addr = src; 5252531Swollman rtalloc_ign(&ro, RTF_CLONING); 5269209Swollman 5272531Swollman if (ro.ro_rt == NULL) 5282531Swollman return 0; 5292531Swollman 5302531Swollman /* if ifp is provided, check for equality with rtentry */ 5319209Swollman if (ifp != NULL && ro.ro_rt->rt_ifp != ifp) { 5322531Swollman RTFREE(ro.ro_rt); 5332531Swollman return 0; 5341541Srgrimes } 5351541Srgrimes 53643802Swollman /* if no ifp provided, check if rtentry is not default route */ 53743802Swollman if (ifp == NULL && 53843802Swollman satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 5391541Srgrimes RTFREE(ro.ro_rt); 54043802Swollman return 0; 54143802Swollman } 54243802Swollman 54343802Swollman /* or if this is a blackhole/reject route */ 54473172Sjlemon if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 54573172Sjlemon RTFREE(ro.ro_rt); 54673172Sjlemon return 0; 54773172Sjlemon } 54898613Sluigi 54973172Sjlemon /* found valid route */ 55073399Struckman RTFREE(ro.ro_rt); 55173399Struckman return 1; 55273399Struckman} 55373399Struckman 55473402Struckman 55573402Struckmanstatic u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 55673626Struckman 55786047Sluigi#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 55873626Struckman#define SNP(buf) buf, sizeof(buf) 55973626Struckman 56073791Struckman/* 56173626Struckman * We enter here when we have a rule with O_LOG. 56273626Struckman * XXX this function alone takes about 2Kbytes of code! 56373399Struckman */ 56473399Struckmanstatic void 56583934Sbrooksipfw_log(struct ip_fw *f, u_int hlen, struct ether_header *eh, 56673402Struckman struct mbuf *m, struct ifnet *oif) 567125785Smlaier{ 56873399Struckman char *action; 56984102Sjlemon int limit_reached = 0; 57084102Sjlemon char action2[40], proto[48], fragment[28]; 57184102Sjlemon 57284102Sjlemon fragment[0] = '\0'; 57337409Sjulian proto[0] = '\0'; 57473399Struckman 57573399Struckman if (f == NULL) { /* bogus pkt */ 57673399Struckman if (verbose_limit != 0 && norule_counter >= verbose_limit) 57737409Sjulian return; 57873399Struckman norule_counter++; 57973399Struckman if (norule_counter == verbose_limit) 58037409Sjulian limit_reached = verbose_limit; 58184102Sjlemon action = "Refuse"; 58284102Sjlemon } else { /* O_LOG is the first action, find the real one */ 58384102Sjlemon ipfw_insn *cmd = ACTION_PTR(f); 58484102Sjlemon ipfw_insn_log *l = (ipfw_insn_log *)cmd; 58584102Sjlemon 58684102Sjlemon if (l->max_log != 0 && l->log_left == 0) 58784102Sjlemon return; 58884102Sjlemon l->log_left--; 58984102Sjlemon if (l->log_left == 0) 590130685Sbms limit_reached = l->max_log; 591130685Sbms cmd += F_LEN(cmd); /* point to first action */ 59284102Sjlemon if (cmd->opcode == O_ALTQ) { 59384102Sjlemon ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 59484102Sjlemon 59584102Sjlemon snprintf(SNPARGS(action2, 0), "Altq %d", 5961541Srgrimes altq->qid); 59773172Sjlemon cmd += F_LEN(cmd); 5981541Srgrimes } 59973172Sjlemon if (cmd->opcode == O_PROB) 6001541Srgrimes cmd += F_LEN(cmd); 601131840Sbrian 602131840Sbrian action = action2; 603131840Sbrian switch (cmd->opcode) { 604131840Sbrian case O_DENY: 6051541Srgrimes action = "Deny"; 6061541Srgrimes break; 6071541Srgrimes 6081541Srgrimes case O_REJECT: 6091541Srgrimes if (cmd->arg1==ICMP_REJECT_RST) 6101541Srgrimes action = "Reset"; 6111541Srgrimes else if (cmd->arg1==ICMP_UNREACH_HOST) 6121541Srgrimes action = "Reject"; 6131541Srgrimes else 6141541Srgrimes snprintf(SNPARGS(action2, 0), "Unreach %d", 6151541Srgrimes cmd->arg1); 6161541Srgrimes break; 6171541Srgrimes 618106968Sluigi case O_ACCEPT: 619106968Sluigi action = "Accept"; 6201541Srgrimes break; 6211541Srgrimes case O_COUNT: 62213929Swollman action = "Count"; 6231541Srgrimes break; 6241541Srgrimes case O_DIVERT: 6251541Srgrimes snprintf(SNPARGS(action2, 0), "Divert %d", 62696432Sdd cmd->arg1); 6271541Srgrimes break; 6281541Srgrimes case O_TEE: 6291541Srgrimes snprintf(SNPARGS(action2, 0), "Tee %d", 6301541Srgrimes cmd->arg1); 6311541Srgrimes break; 6321541Srgrimes case O_SKIPTO: 6331541Srgrimes snprintf(SNPARGS(action2, 0), "SkipTo %d", 6341541Srgrimes cmd->arg1); 6351541Srgrimes break; 6361541Srgrimes case O_PIPE: 6371541Srgrimes snprintf(SNPARGS(action2, 0), "Pipe %d", 6381541Srgrimes cmd->arg1); 6391541Srgrimes break; 64021932Swollman case O_QUEUE: 6411541Srgrimes snprintf(SNPARGS(action2, 0), "Queue %d", 64213929Swollman cmd->arg1); 6431541Srgrimes break; 6441541Srgrimes case O_FORWARD_IP: { 6451541Srgrimes ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 6461541Srgrimes int len; 6471541Srgrimes 6481541Srgrimes len = snprintf(SNPARGS(action2, 0), "Forward to %s", 6491541Srgrimes inet_ntoa(sa->sa.sin_addr)); 6501541Srgrimes if (sa->sa.sin_port) 6511541Srgrimes snprintf(SNPARGS(action2, len), ":%d", 65255009Sshin sa->sa.sin_port); 65355009Sshin } 65455009Sshin break; 65555009Sshin case O_NETGRAPH: 65655009Sshin snprintf(SNPARGS(action2, 0), "Netgraph %d", 65755009Sshin cmd->arg1); 65855009Sshin break; 65955009Sshin case O_NGTEE: 66055009Sshin snprintf(SNPARGS(action2, 0), "Ngtee %d", 66155009Sshin cmd->arg1); 66283934Sbrooks break; 66355009Sshin default: 6641541Srgrimes action = "UNKNOWN"; 6651541Srgrimes break; 6661541Srgrimes } 6671541Srgrimes } 66872803Sjesper 66991271Sjedgar if (hlen == 0) { /* non-ip */ 67091271Sjedgar snprintf(SNPARGS(proto, 0), "MAC"); 67191271Sjedgar } else { 67291271Sjedgar struct ip *ip = mtod(m, struct ip *); 67391271Sjedgar /* these three are all aliases to the same thing */ 67491271Sjedgar struct icmp *const icmp = L3HDR(struct icmp, ip); 67591271Sjedgar struct tcphdr *const tcp = (struct tcphdr *)icmp; 67691271Sjedgar struct udphdr *const udp = (struct udphdr *)icmp; 67791271Sjedgar 67891271Sjedgar int ip_off, offset, ip_len; 679105199Ssam 680105199Ssam int len; 681105199Ssam 682105199Ssam if (eh != NULL) { /* layer 2 packets are as on the wire */ 683105199Ssam ip_off = ntohs(ip->ip_off); 684105199Ssam ip_len = ntohs(ip->ip_len); 685105199Ssam } else { 686105199Ssam ip_off = ip->ip_off; 687105199Ssam ip_len = ip->ip_len; 688105199Ssam } 689105199Ssam offset = ip_off & IP_OFFMASK; 690105199Ssam switch (ip->ip_p) { 691105199Ssam case IPPROTO_TCP: 692105199Ssam len = snprintf(SNPARGS(proto, 0), "TCP %s", 693105199Ssam inet_ntoa(ip->ip_src)); 694105199Ssam if (offset == 0) 695105199Ssam snprintf(SNPARGS(proto, len), ":%d %s:%d", 696105199Ssam ntohs(tcp->th_sport), 697105199Ssam inet_ntoa(ip->ip_dst), 698105199Ssam ntohs(tcp->th_dport)); 699105199Ssam else 700105199Ssam snprintf(SNPARGS(proto, len), " %s", 701105199Ssam inet_ntoa(ip->ip_dst)); 702105199Ssam break; 703105199Ssam 704105199Ssam case IPPROTO_UDP: 705105199Ssam len = snprintf(SNPARGS(proto, 0), "UDP %s", 706105199Ssam inet_ntoa(ip->ip_src)); 707122708Sandre if (offset == 0) 70891271Sjedgar snprintf(SNPARGS(proto, len), ":%d %s:%d", 70913929Swollman ntohs(udp->uh_sport), 7101541Srgrimes inet_ntoa(ip->ip_dst), 7111541Srgrimes ntohs(udp->uh_dport)); 71288665Syar else 71388665Syar snprintf(SNPARGS(proto, len), " %s", 71488665Syar inet_ntoa(ip->ip_dst)); 71588665Syar break; 71688665Syar 71798613Sluigi case IPPROTO_ICMP: 71898613Sluigi if (offset == 0) 71988665Syar len = snprintf(SNPARGS(proto, 0), 72088665Syar "ICMP:%u.%u ", 72188665Syar icmp->icmp_type, icmp->icmp_code); 72267334Sjoe else 72367334Sjoe len = snprintf(SNPARGS(proto, 0), "ICMP "); 72467334Sjoe len += snprintf(SNPARGS(proto, len), "%s", 72567334Sjoe inet_ntoa(ip->ip_src)); 72667334Sjoe snprintf(SNPARGS(proto, len), " %s", 7273969Sjkh inet_ntoa(ip->ip_dst)); 7281541Srgrimes break; 729133069Sandre 730133069Sandre default: 7311541Srgrimes len = snprintf(SNPARGS(proto, 0), "P:%d %s", ip->ip_p, 73267620Sru inet_ntoa(ip->ip_src)); 733133069Sandre snprintf(SNPARGS(proto, len), " %s", 734133069Sandre inet_ntoa(ip->ip_dst)); 73567620Sru break; 73667620Sru } 73767620Sru 738105586Sphk if (ip_off & (IP_MF | IP_OFFMASK)) 73927669Sbrian snprintf(SNPARGS(fragment, 0), " (frag %d:%d@%d%s)", 74067620Sru ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 741126239Smlaier offset << 3, 74290868Smike (ip_off & IP_MF) ? "+" : ""); 74390868Smike } 74467620Sru if (oif || m->m_pkthdr.rcvif) 74568169Sru log(LOG_SECURITY | LOG_INFO, 74668169Sru "ipfw: %d %s %s %s via %s%s\n", 74768169Sru f ? f->rulenum : -1, 74868169Sru action, proto, oif ? "out" : "in", 74990868Smike oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 75090868Smike fragment); 75167620Sru else 75227669Sbrian log(LOG_SECURITY | LOG_INFO, 753133069Sandre "ipfw: %d %s %s [no if info]%s\n", 7541541Srgrimes f ? f->rulenum : -1, 755133069Sandre action, proto, fragment); 756133069Sandre if (limit_reached) 757133069Sandre log(LOG_SECURITY | LOG_NOTICE, 758133069Sandre "ipfw: limit %d reached on entry %d\n", 759133069Sandre limit_reached, f ? f->rulenum : -1); 760133069Sandre} 76117072Sjulian 7621541Srgrimes/* 76354175Sarchie * IMPORTANT: the hash function for dynamic rules must be commutative 76417072Sjulian * in source and destination (ip,port), because rules are bidirectional 765126239Smlaier * and we want to find both in the same bucket. 76654175Sarchie */ 767126239Smlaierstatic __inline int 76854175Sarchiehash_packet(struct ipfw_flow_id *id) 76954175Sarchie{ 77054175Sarchie u_int32_t i; 771126239Smlaier 772126239Smlaier i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 773126239Smlaier i &= (curr_dyn_buckets - 1); 77454175Sarchie return i; 77554175Sarchie} 77654175Sarchie 77790868Smike/** 77890868Smike * unlink a dynamic rule from a chain. prev is a pointer to 77954175Sarchie * the previous one, q is a pointer to the rule to delete, 78054175Sarchie * head is a pointer to the head of the queue. 781126239Smlaier * Modifies q and potentially also head. 78226359Sjulian */ 78331163Sjulian#define UNLINK_DYN_RULE(prev, head, q) { \ 78454175Sarchie ipfw_dyn_rule *old_q = q; \ 78554175Sarchie \ 78654175Sarchie /* remove a refcount to the parent */ \ 78754175Sarchie if (q->dyn_type == O_LIMIT) \ 78854175Sarchie q->parent->count--; \ 78989809Scjc DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 79098613Sluigi (q->id.src_ip), (q->id.src_port), \ 79198613Sluigi (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 792126239Smlaier if (prev != NULL) \ 793126239Smlaier prev->next = q = q->next; \ 794126239Smlaier else \ 795126239Smlaier head = q = q->next; \ 79698613Sluigi dyn_count--; \ 79789809Scjc uma_zfree(ipfw_dyn_rule_zone, old_q); } 79831163Sjulian 79917072Sjulian#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 80017072Sjulian 80178064Sume/** 80217072Sjulian * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 80378064Sume * 80478064Sume * If keep_me == NULL, rules are deleted even if not expired, 80578064Sume * otherwise only expired rules are removed. 80678064Sume * 80778064Sume * The value of the second parameter is also used to point to identify 80878064Sume * a rule we absolutely do not want to remove (e.g. because we are 80978064Sume * holding a reference to it -- this is the case with O_LIMIT_PARENT 81078064Sume * rules). The pointer is only used for comparison, so any non-null 81178064Sume * value will do. 81278064Sume */ 813105199Ssamstatic void 814105199Ssamremove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 815105199Ssam{ 816105199Ssam static u_int32_t last_remove = 0; 817105199Ssam 818105199Ssam#define FORCE (keep_me == NULL) 819105199Ssam 820105199Ssam ipfw_dyn_rule *prev, *q; 821105199Ssam int i, pass = 0, max_pass = 0; 822105199Ssam 823105199Ssam IPFW_DYN_LOCK_ASSERT(); 824105199Ssam 825105199Ssam if (ipfw_dyn_v == NULL || dyn_count == 0) 826105199Ssam return; 827105199Ssam /* do not expire more than once per second, it is useless */ 828105199Ssam if (!FORCE && last_remove == time_second) 829105199Ssam return; 830105199Ssam last_remove = time_second; 831105199Ssam 832105199Ssam /* 833105199Ssam * because O_LIMIT refer to parent rules, during the first pass only 834105199Ssam * remove child and mark any pending LIMIT_PARENT, and remove 835105199Ssam * them in a second pass. 836105199Ssam */ 837105199Ssamnext_pass: 838105199Ssam for (i = 0 ; i < curr_dyn_buckets ; i++) { 839105199Ssam for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 840105199Ssam /* 841105199Ssam * Logic can become complex here, so we split tests. 842105199Ssam */ 843105199Ssam if (q == keep_me) 844105199Ssam goto next; 845105199Ssam if (rule != NULL && rule != q->rule) 846105199Ssam goto next; /* not the one we are looking for */ 847105199Ssam if (q->dyn_type == O_LIMIT_PARENT) { 848105199Ssam /* 849105199Ssam * handle parent in the second pass, 850105199Ssam * record we need one. 851105199Ssam */ 85278064Sume max_pass = 1; 85378064Sume if (pass == 0) 8541541Srgrimes goto next; 8551541Srgrimes if (FORCE && q->count != 0 ) { 8561541Srgrimes /* XXX should not happen! */ 857126239Smlaier printf("ipfw: OUCH! cannot remove rule," 858126239Smlaier " count %d\n", q->count); 859126239Smlaier } 860126239Smlaier } else { 861126239Smlaier if (!FORCE && 862126239Smlaier !TIME_LEQ( q->expire, time_second )) 863126239Smlaier goto next; 864126239Smlaier } 865126239Smlaier if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 866126239Smlaier UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 86713929Swollman continue; 8681541Srgrimes } 8691541Srgrimesnext: 8701541Srgrimes prev=q; 8711541Srgrimes q=q->next; 87213929Swollman } 87354175Sarchie } 874133069Sandre if (pass++ < max_pass) 875133069Sandre goto next_pass; 876133069Sandre} 877133069Sandre 878133069Sandre 879133069Sandre/** 880133069Sandre * lookup a dynamic rule. 8811541Srgrimes */ 88254175Sarchiestatic ipfw_dyn_rule * 883133069Sandrelookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 884133069Sandre struct tcphdr *tcp) 8851541Srgrimes{ 886133069Sandre /* 887133069Sandre * stateful ipfw extensions. 888133069Sandre * Lookup into dynamic session queue 889133069Sandre */ 890133069Sandre#define MATCH_REVERSE 0 891121684Sume#define MATCH_FORWARD 1 892133069Sandre#define MATCH_NONE 2 8931541Srgrimes#define MATCH_UNKNOWN 3 894133069Sandre int i, dir = MATCH_NONE; 895133069Sandre ipfw_dyn_rule *prev, *q=NULL; 896133069Sandre 897133069Sandre IPFW_DYN_LOCK_ASSERT(); 898133557Sandre 899133557Sandre if (ipfw_dyn_v == NULL) 900133069Sandre goto done; /* not found */ 901119753Ssam i = hash_packet( pkt ); 902133069Sandre for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 903133069Sandre if (q->dyn_type == O_LIMIT_PARENT && q->count) 904133069Sandre goto next; 905133069Sandre if (TIME_LEQ( q->expire, time_second)) { /* expire entry */ 906133069Sandre UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 907133069Sandre continue; 908133069Sandre } 9091541Srgrimes if (pkt->proto == q->id.proto && 910133069Sandre q->dyn_type != O_LIMIT_PARENT) { 911133069Sandre if (pkt->src_ip == q->id.src_ip && 912133069Sandre pkt->dst_ip == q->id.dst_ip && 913133069Sandre pkt->src_port == q->id.src_port && 914133069Sandre pkt->dst_port == q->id.dst_port ) { 915133069Sandre dir = MATCH_FORWARD; 916133069Sandre break; 917133069Sandre } 918133069Sandre if (pkt->src_ip == q->id.dst_ip && 919133069Sandre pkt->dst_ip == q->id.src_ip && 920133069Sandre pkt->src_port == q->id.dst_port && 921133069Sandre pkt->dst_port == q->id.src_port ) { 922133069Sandre dir = MATCH_REVERSE; 923133069Sandre break; 924133069Sandre } 925133069Sandre } 926133069Sandrenext: 927133069Sandre prev = q; 928133069Sandre q = q->next; 929133069Sandre } 930133069Sandre if (q == NULL) 931133069Sandre goto done; /* q = NULL, not found */ 932133069Sandre 933133069Sandre if ( prev != NULL) { /* found and not in front */ 934133069Sandre prev->next = q->next; 935133069Sandre q->next = ipfw_dyn_v[i]; 936133069Sandre ipfw_dyn_v[i] = q; 937133069Sandre } 938133069Sandre if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 939133069Sandre u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 940133069Sandre 941133069Sandre#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 942133069Sandre#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 943133069Sandre q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 944133069Sandre switch (q->state) { 945133069Sandre case TH_SYN: /* opening */ 946133069Sandre q->expire = time_second + dyn_syn_lifetime; 947133069Sandre break; 948133069Sandre 949133069Sandre case BOTH_SYN: /* move to established */ 950133069Sandre case BOTH_SYN | TH_FIN : /* one side tries to close */ 951133069Sandre case BOTH_SYN | (TH_FIN << 8) : 952133069Sandre if (tcp) { 953133069Sandre#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 954133069Sandre u_int32_t ack = ntohl(tcp->th_ack); 955133069Sandre if (dir == MATCH_FORWARD) { 956133069Sandre if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 957133069Sandre q->ack_fwd = ack; 958133069Sandre else { /* ignore out-of-sequence */ 959133069Sandre break; 960133069Sandre } 961133069Sandre } else { 962133069Sandre if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 963133069Sandre q->ack_rev = ack; 964133069Sandre else { /* ignore out-of-sequence */ 965133069Sandre break; 966133069Sandre } 967133069Sandre } 968133069Sandre } 969133069Sandre q->expire = time_second + dyn_ack_lifetime; 970133069Sandre break; 971133069Sandre 972133069Sandre case BOTH_SYN | BOTH_FIN: /* both sides closed */ 973133069Sandre if (dyn_fin_lifetime >= dyn_keepalive_period) 974133069Sandre dyn_fin_lifetime = dyn_keepalive_period - 1; 975133069Sandre q->expire = time_second + dyn_fin_lifetime; 976133069Sandre break; 977133069Sandre 978133069Sandre default: 979133069Sandre#if 0 980133069Sandre /* 9811541Srgrimes * reset or some invalid combination, but can also 9821541Srgrimes * occur if we use keep-state the wrong way. 9831541Srgrimes */ 9841541Srgrimes if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 9851541Srgrimes printf("invalid state: 0x%x\n", q->state); 9861541Srgrimes#endif 9871541Srgrimes if (dyn_rst_lifetime >= dyn_keepalive_period) 9881541Srgrimes dyn_rst_lifetime = dyn_keepalive_period - 1; 9891541Srgrimes q->expire = time_second + dyn_rst_lifetime; 990115909Srwatson break; 991111119Simp } 9921541Srgrimes } else if (pkt->proto == IPPROTO_UDP) { 9931541Srgrimes q->expire = time_second + dyn_udp_lifetime; 994101095Srwatson } else { 995112675Srwatson /* other protocols */ 996112675Srwatson q->expire = time_second + dyn_short_lifetime; 997112675Srwatson } 998112675Srwatsondone: 999101095Srwatson if (match_direction) 1000101095Srwatson *match_direction = dir; 100174362Sphk return q; 100229480Sache} 1003111244Ssilby 10041541Srgrimesstatic ipfw_dyn_rule * 10051541Srgrimeslookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 10061541Srgrimes struct tcphdr *tcp) 100738513Sdfr{ 100838513Sdfr ipfw_dyn_rule *q; 100941993Sluigi 101041993Sluigi IPFW_DYN_LOCK(); 101141993Sluigi q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1012101095Srwatson if (q == NULL) 1013111244Ssilby IPFW_DYN_UNLOCK(); 1014101095Srwatson /* NB: return table locked when q is not NULL */ 1015101095Srwatson return q; 1016101095Srwatson} 10171541Srgrimes 10181541Srgrimesstatic void 101938513Sdfrrealloc_dynamic_table(void) 102038513Sdfr{ 10211541Srgrimes IPFW_DYN_LOCK_ASSERT(); 1022121684Sume 1023121684Sume /* 1024121684Sume * Try reallocation, make sure we have a power of 2 and do 1025121684Sume * not allow more than 64k entries. In case of overflow, 1026121684Sume * default to 1024. 1027121684Sume */ 1028121684Sume 1029121684Sume if (dyn_buckets > 65536) 1030121684Sume dyn_buckets = 1024; 1031121684Sume if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 1032121684Sume dyn_buckets = curr_dyn_buckets; /* reset */ 1033121684Sume return; 1034121684Sume } 1035121684Sume curr_dyn_buckets = dyn_buckets; 1036121684Sume if (ipfw_dyn_v != NULL) 1037121684Sume free(ipfw_dyn_v, M_IPFW); 10381541Srgrimes for (;;) { 10391541Srgrimes ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 104038513Sdfr M_IPFW, M_NOWAIT | M_ZERO); 104138513Sdfr if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 10421541Srgrimes break; 10431541Srgrimes curr_dyn_buckets /= 2; 10441541Srgrimes } 10451541Srgrimes} 10461541Srgrimes 104741993Sluigi/** 104841993Sluigi * Install state of type 'type' for a dynamic session. 104958698Sjlemon * The hash table contains two type of rules: 105058698Sjlemon * - regular rules (O_KEEP_STATE) 105158698Sjlemon * - rules for sessions with limited number of sess per user 10521541Srgrimes * (O_LIMIT). When they are created, the parent is 105338513Sdfr * increased by 1, and decreased on delete. In this case, 105438513Sdfr * the third parameter is the parent rule and not the chain. 10551541Srgrimes * - "parent" rules for the above (O_LIMIT_PARENT). 10561541Srgrimes */ 10571541Srgrimesstatic ipfw_dyn_rule * 105855009Sshinadd_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 105958698Sjlemon{ 10601541Srgrimes ipfw_dyn_rule *r; 10611541Srgrimes int i; 10621541Srgrimes 106341993Sluigi IPFW_DYN_LOCK_ASSERT(); 106441993Sluigi 106541993Sluigi if (ipfw_dyn_v == NULL || 106641993Sluigi (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 106741993Sluigi realloc_dynamic_table(); 10681541Srgrimes if (ipfw_dyn_v == NULL) 10691541Srgrimes return NULL; /* failed ! */ 10701541Srgrimes } 10711541Srgrimes i = hash_packet(id); 10721541Srgrimes 10731541Srgrimes r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 107438513Sdfr if (r == NULL) { 107541993Sluigi printf ("ipfw: sorry cannot allocate state\n"); 1076111479Smaxim return NULL; 107738513Sdfr } 107838513Sdfr 107938513Sdfr /* increase refcount on parent, and set pointer */ 108038513Sdfr if (dyn_type == O_LIMIT) { 108158698Sjlemon ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 10821541Srgrimes if ( parent->dyn_type != O_LIMIT_PARENT) 10831541Srgrimes panic("invalid parent"); 108438513Sdfr parent->count++; 108541993Sluigi r->parent = parent; 1086111478Smaxim rule = parent->rule; 1087111244Ssilby } 108838513Sdfr 10891541Srgrimes r->id = *id; 10901541Srgrimes r->expire = time_second + dyn_syn_lifetime; 109141993Sluigi r->rule = rule; 109217072Sjulian r->dyn_type = dyn_type; 109317072Sjulian r->pcnt = r->bcnt = 0; 1094126239Smlaier r->count = 0; 1095126239Smlaier 1096126239Smlaier r->bucket = i; 1097126239Smlaier r->next = ipfw_dyn_v[i]; 1098126239Smlaier ipfw_dyn_v[i] = r; 1099126239Smlaier dyn_count++; 1100126239Smlaier DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1101126239Smlaier dyn_type, 110298613Sluigi (r->id.src_ip), (r->id.src_port), 110317072Sjulian (r->id.dst_ip), (r->id.dst_port), 110417072Sjulian dyn_count ); ) 110517072Sjulian return r; 1106111244Ssilby} 1107111244Ssilby 1108111244Ssilby/** 1109111244Ssilby * lookup dynamic parent rule using pkt and rule as search keys. 1110111244Ssilby * If the lookup fails, then install one. 1111111244Ssilby */ 1112111244Ssilbystatic ipfw_dyn_rule * 1113111244Ssilbylookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 11141541Srgrimes{ 11151541Srgrimes ipfw_dyn_rule *q; 111638513Sdfr int i; 1117111244Ssilby 1118111478Smaxim IPFW_DYN_LOCK_ASSERT(); 1119111478Smaxim 1120111244Ssilby if (ipfw_dyn_v) { 1121111478Smaxim i = hash_packet( pkt ); 1122133069Sandre for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1123111244Ssilby if (q->dyn_type == O_LIMIT_PARENT && 112438513Sdfr rule== q->rule && 11251541Srgrimes pkt->proto == q->id.proto && 112638513Sdfr pkt->src_ip == q->id.src_ip && 1127111244Ssilby pkt->dst_ip == q->id.dst_ip && 1128111478Smaxim pkt->src_port == q->id.src_port && 1129111478Smaxim pkt->dst_port == q->id.dst_port) { 1130111244Ssilby q->expire = time_second + dyn_short_lifetime; 1131111478Smaxim DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1132133069Sandre return q; 1133111244Ssilby } 11341541Srgrimes } 11351541Srgrimes return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 113619183Sfenner} 11371541Srgrimes 113838513Sdfr/** 113938513Sdfr * Install dynamic state for rule type cmd->o.opcode 1140105586Sphk * 114119183Sfenner * Returns 1 (failure) if state is not installed because of errors or because 1142111478Smaxim * session limitations are enforced. 114374362Sphk */ 1144133069Sandrestatic int 114519183Sfennerinstall_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 114619183Sfenner struct ip_fw_args *args) 114719183Sfenner{ 114819183Sfenner static int last_log; 114919183Sfenner 115038513Sdfr ipfw_dyn_rule *q; 11511541Srgrimes 11521541Srgrimes DEB(printf("ipfw: install state type %d 0x%08x %u -> 0x%08x %u\n", 11531541Srgrimes cmd->o.opcode, 115438513Sdfr (args->f_id.src_ip), (args->f_id.src_port), 115539043Sdfr (args->f_id.dst_ip), (args->f_id.dst_port) );) 115638513Sdfr 115738513Sdfr IPFW_DYN_LOCK(); 115839043Sdfr 115958698Sjlemon q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 116058698Sjlemon 116165859Sjlemon if (q != NULL) { /* should never occur */ 11621541Srgrimes if (last_log != time_second) { 1163101095Srwatson last_log = time_second; 1164101095Srwatson printf("ipfw: install_state: entry already present, done\n"); 1165101095Srwatson } 1166101095Srwatson IPFW_DYN_UNLOCK(); 11671541Srgrimes return 0; 11681541Srgrimes } 1169133069Sandre 1170133069Sandre if (dyn_count >= dyn_max) 11711541Srgrimes /* 11721541Srgrimes * Run out of slots, try to remove any expired rule. 1173133069Sandre */ 117438513Sdfr remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 117538513Sdfr 117674362Sphk if (dyn_count >= dyn_max) { 117729480Sache if (last_log != time_second) { 11781541Srgrimes last_log = time_second; 1179105586Sphk printf("ipfw: install_state: Too many dynamic rules\n"); 1180105586Sphk } 11811541Srgrimes IPFW_DYN_UNLOCK(); 1182103553Sphk return 1; /* cannot install, notify caller */ 1183103553Sphk } 1184133069Sandre 1185133069Sandre switch (cmd->o.opcode) { 118655009Sshin case O_KEEP_STATE: /* bidir rule */ 11871541Srgrimes add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 11881541Srgrimes break; 11891541Srgrimes 1190115909Srwatson case O_LIMIT: /* limit number of sessions */ 1191111244Ssilby { 11921541Srgrimes u_int16_t limit_mask = cmd->limit_mask; 1193133069Sandre struct ipfw_flow_id id; 1194133069Sandre ipfw_dyn_rule *parent; 1195133069Sandre 119638513Sdfr DEB(printf("ipfw: installing dyn-limit rule %d\n", 119738513Sdfr cmd->conn_limit);) 11981541Srgrimes 11991541Srgrimes id.dst_ip = id.src_ip = 0; 12001541Srgrimes id.dst_port = id.src_port = 0; 12011541Srgrimes id.proto = args->f_id.proto; 12021541Srgrimes 12031541Srgrimes if (limit_mask & DYN_SRC_ADDR) 120412296Sphk id.src_ip = args->f_id.src_ip; 120574362Sphk if (limit_mask & DYN_DST_ADDR) 120674362Sphk id.dst_ip = args->f_id.dst_ip; 12071541Srgrimes if (limit_mask & DYN_SRC_PORT) 12081541Srgrimes id.src_port = args->f_id.src_port; 120938513Sdfr if (limit_mask & DYN_DST_PORT) 12101541Srgrimes id.dst_port = args->f_id.dst_port; 1211119753Ssam parent = lookup_dyn_parent(&id, rule); 1212119753Ssam if (parent == NULL) { 121338513Sdfr printf("ipfw: add parent failed\n"); 121438513Sdfr return 1; 121538513Sdfr } 121638513Sdfr if (parent->count >= cmd->conn_limit) { 12171541Srgrimes /* 121874362Sphk * See if we can remove some expired rule. 12191541Srgrimes */ 122029480Sache remove_dyn_rule(rule, parent); 12211541Srgrimes if (parent->count >= cmd->conn_limit) { 12221541Srgrimes if (fw_verbose && last_log != time_second) { 12231541Srgrimes last_log = time_second; 12241541Srgrimes log(LOG_SECURITY | LOG_DEBUG, 12251541Srgrimes "drop session, too many entries\n"); 12261541Srgrimes } 12271541Srgrimes IPFW_DYN_UNLOCK(); 12281541Srgrimes return 1; 12291541Srgrimes } 12301541Srgrimes } 12311541Srgrimes add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 12321541Srgrimes } 123329480Sache break; 12341541Srgrimes default: 1235119753Ssam printf("ipfw: unknown dynamic rule type %u\n", cmd->o.opcode); 123629480Sache IPFW_DYN_UNLOCK(); 123774362Sphk return 1; 123874362Sphk } 123974362Sphk lookup_dyn_rule_locked(&args->f_id, NULL, NULL); /* XXX just set lifetime */ 124074362Sphk IPFW_DYN_UNLOCK(); 124174362Sphk return 0; 124274362Sphk} 1243111478Smaxim 124474362Sphk/* 124529480Sache * Transmit a TCP packet, containing either a RST or a keepalive. 12461541Srgrimes * When flags & TH_RST, we are sending a RST packet, because of a 12471541Srgrimes * "reset" action matched the packet. 124877665Sjesper * Otherwise we are sending a keepalive, and flags & TH_ 124977665Sjesper */ 125077665Sjesperstatic void 125177665Sjespersend_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags) 125277665Sjesper{ 1253111541Ssilby struct mbuf *m; 1254110178Ssilby struct ip *ip; 1255111479Smaxim struct tcphdr *tcp; 1256111478Smaxim 1257111478Smaxim MGETHDR(m, M_DONTWAIT, MT_HEADER); 125877665Sjesper if (m == 0) 125977665Sjesper return; 126077665Sjesper m->m_pkthdr.rcvif = (struct ifnet *)0; 126177665Sjesper m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1262119753Ssam m->m_data += max_linkhdr; 12631541Srgrimes 12641541Srgrimes ip = mtod(m, struct ip *); 12651541Srgrimes bzero(ip, m->m_len); 12661541Srgrimes tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 12671541Srgrimes ip->ip_p = IPPROTO_TCP; 12681541Srgrimes tcp->th_off = 5; 12691541Srgrimes /* 12701541Srgrimes * Assume we are sending a RST (or a keepalive in the reverse 12711541Srgrimes * direction), swap src and destination addresses and ports. 127229480Sache */ 127329480Sache ip->ip_src.s_addr = htonl(id->dst_ip); 1274119753Ssam ip->ip_dst.s_addr = htonl(id->src_ip); 127529480Sache tcp->th_sport = htons(id->dst_port); 127674362Sphk tcp->th_dport = htons(id->src_port); 1277111478Smaxim if (flags & TH_RST) { /* we are sending a RST */ 1278111478Smaxim if (flags & TH_ACK) { 127974362Sphk tcp->th_seq = htonl(ack); 128029480Sache tcp->th_ack = htonl(0); 12811541Srgrimes tcp->th_flags = TH_RST; 1282119753Ssam } else { 128312933Swollman if (flags & TH_SYN) 12841541Srgrimes seq++; 12851541Srgrimes tcp->th_seq = htonl(0); 12861541Srgrimes tcp->th_ack = htonl(seq); 12871541Srgrimes tcp->th_flags = TH_RST | TH_ACK; 12881541Srgrimes } 12891541Srgrimes } else { 129088665Syar /* 129188665Syar * We are sending a keepalive. flags & TH_SYN determines 129288665Syar * the direction, forward if set, reverse if clear. 129388665Syar * NOTE: seq and ack are always assumed to be correct 129488665Syar * as set by the caller. This may be confusing... 12951541Srgrimes */ 12961541Srgrimes if (flags & TH_SYN) { 12971541Srgrimes /* 129812296Sphk * we have to rewrite the correct addresses! 129998613Sluigi */ 13001541Srgrimes ip->ip_dst.s_addr = htonl(id->dst_ip); 130198613Sluigi ip->ip_src.s_addr = htonl(id->src_ip); 130298613Sluigi tcp->th_dport = htons(id->dst_port); 130398613Sluigi tcp->th_sport = htons(id->src_port); 13041541Srgrimes } 13051541Srgrimes tcp->th_seq = htonl(seq); 13061541Srgrimes tcp->th_ack = htonl(ack); 130798701Sluigi tcp->th_flags = TH_ACK; 13081541Srgrimes } 1309129017Sandre /* 1310129017Sandre * set ip_len to the payload size so we can compute 1311129017Sandre * the tcp checksum on the pseudoheader 1312129017Sandre * XXX check this, could save a couple of words ? 1313129017Sandre */ 1314129017Sandre ip->ip_len = htons(sizeof(struct tcphdr)); 1315129017Sandre tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1316129017Sandre /* 1317129017Sandre * now fill fields left out earlier 13181541Srgrimes */ 13191541Srgrimes ip->ip_ttl = ip_defttl; 1320105586Sphk ip->ip_len = m->m_pkthdr.len; 13211541Srgrimes m->m_flags |= M_SKIP_FIREWALL; 13221541Srgrimes ip_output(m, NULL, NULL, 0, NULL, NULL); 13231541Srgrimes} 13241541Srgrimes 13251541Srgrimes/* 13261541Srgrimes * sends a reject message, consuming the mbuf passed as an argument. 13271541Srgrimes */ 132860304Sitojunstatic void 132960304Sitojunsend_reject(struct ip_fw_args *args, int code, int offset, int ip_len) 133060304Sitojun{ 133160304Sitojun 13321541Srgrimes if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 133361183Sjlemon /* We need the IP header in host order for icmp_error(). */ 13341541Srgrimes if (args->eh != NULL) { 13351541Srgrimes struct ip *ip = mtod(args->m, struct ip *); 13361541Srgrimes ip->ip_len = ntohs(ip->ip_len); 13371541Srgrimes ip->ip_off = ntohs(ip->ip_off); 13381541Srgrimes } 13391541Srgrimes icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 13401541Srgrimes } else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) { 13411541Srgrimes struct tcphdr *const tcp = 13421541Srgrimes L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 13431541Srgrimes if ( (tcp->th_flags & TH_RST) == 0) 13441541Srgrimes send_pkt(&(args->f_id), ntohl(tcp->th_seq), 13451541Srgrimes ntohl(tcp->th_ack), 13461541Srgrimes tcp->th_flags | TH_RST); 13471541Srgrimes m_freem(args->m); 13481541Srgrimes } else 13491541Srgrimes m_freem(args->m); 13501541Srgrimes args->m = NULL; 13511541Srgrimes} 13521541Srgrimes 13531541Srgrimes/** 135488665Syar * 135588665Syar * Given an ip_fw *, lookup_next_rule will return a pointer 135688665Syar * to the next rule, which can be either the jump 135788665Syar * target (for skipto instructions) or the next one in the list (in 135878064Sume * all other cases including a missing jump target). 135978064Sume * The result is also written in the "next_rule" field of the rule. 136078064Sume * Backward jumps are not allowed, so start looking from the next 136178064Sume * rule... 13621541Srgrimes * 13631541Srgrimes * This never returns NULL -- in case we do not have an exact match, 13641541Srgrimes * the next rule is returned. When the ruleset is changed, 13651541Srgrimes * pointers are flushed so we are always correct. 13661541Srgrimes */ 13671541Srgrimes 13681541Srgrimesstatic struct ip_fw * 1369133481Sandrelookup_next_rule(struct ip_fw *me) 13701541Srgrimes{ 13711541Srgrimes struct ip_fw *rule = NULL; 13721541Srgrimes ipfw_insn *cmd; 13731541Srgrimes 13741541Srgrimes /* look for action, in case it is a skipto */ 137530816Sguido cmd = ACTION_PTR(me); 137630816Sguido if (cmd->opcode == O_LOG) 13771541Srgrimes cmd += F_LEN(cmd); 13781541Srgrimes if (cmd->opcode == O_ALTQ) 13791541Srgrimes cmd += F_LEN(cmd); 13801541Srgrimes if ( cmd->opcode == O_SKIPTO ) 13811541Srgrimes for (rule = me->next; rule ; rule = rule->next) 13821541Srgrimes if (rule->rulenum >= cmd->arg1) 13831541Srgrimes break; 138460661Sjlemon if (rule == NULL) /* failure or not a skipto */ 13851541Srgrimes rule = me->next; 13861541Srgrimes me->next_rule = rule; 13871541Srgrimes return rule; 138833440Sguido} 138933440Sguido 13901541Srgrimesstatic void 13911541Srgrimesinit_tables(void) 13921541Srgrimes{ 139388665Syar int i; 139488665Syar 139588665Syar for (i = 0; i < IPFW_TABLES_MAX; i++) { 139688665Syar rn_inithead((void **)&ipfw_tables[i].rnh, 32); 13977091Swollman ipfw_tables[i].modified = 1; 139837498Sdg } 139937498Sdg} 140037498Sdg 140137498Sdgstatic int 140237498Sdgadd_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen, u_int32_t value) 140330790Sguido{ 140437498Sdg struct radix_node_head *rnh; 140537498Sdg struct table_entry *ent; 140637498Sdg 140737498Sdg if (tbl >= IPFW_TABLES_MAX) 140837498Sdg return (EINVAL); 140937498Sdg rnh = ipfw_tables[tbl].rnh; 141037498Sdg ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 141137498Sdg if (ent == NULL) 141237498Sdg return (ENOMEM); 141337498Sdg ent->value = value; 141437498Sdg ent->addr.sin_len = ent->mask.sin_len = 8; 141588665Syar ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 141688665Syar ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 141788665Syar RADIX_NODE_HEAD_LOCK(rnh); 141837498Sdg if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 141937498Sdg NULL) { 142037498Sdg RADIX_NODE_HEAD_UNLOCK(rnh); 142137498Sdg free(ent, M_IPFW_TBL); 14227091Swollman return (EEXIST); 14237091Swollman } 14241541Srgrimes ipfw_tables[tbl].modified = 1; 14251541Srgrimes RADIX_NODE_HEAD_UNLOCK(rnh); 14261541Srgrimes return (0); 14278384Sdg} 14281541Srgrimes 14297091Swollmanstatic int 14301541Srgrimesdel_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen) 14311541Srgrimes{ 14321541Srgrimes struct radix_node_head *rnh; 1433133481Sandre struct table_entry *ent; 14341541Srgrimes struct sockaddr_in sa, mask; 14351541Srgrimes 1436122708Sandre if (tbl >= IPFW_TABLES_MAX) 1437133481Sandre return (EINVAL); 14381541Srgrimes rnh = ipfw_tables[tbl].rnh; 14391541Srgrimes sa.sin_len = mask.sin_len = 8; 14401541Srgrimes mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 14411541Srgrimes sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 14421541Srgrimes RADIX_NODE_HEAD_LOCK(rnh); 14438384Sdg ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 14448384Sdg if (ent == NULL) { 14451541Srgrimes RADIX_NODE_HEAD_UNLOCK(rnh); 14461541Srgrimes return (ESRCH); 14471541Srgrimes } 14481541Srgrimes ipfw_tables[tbl].modified = 1; 14491541Srgrimes RADIX_NODE_HEAD_UNLOCK(rnh); 14501541Srgrimes free(ent, M_IPFW_TBL); 14511541Srgrimes return (0); 14521541Srgrimes} 145388665Syar 145488665Syarstatic int 145588665Syarflush_table_entry(struct radix_node *rn, void *arg) 145688665Syar{ 145761183Sjlemon struct radix_node_head * const rnh = arg; 145861183Sjlemon struct table_entry *ent; 145961183Sjlemon 146061183Sjlemon ent = (struct table_entry *) 14611541Srgrimes rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 14621541Srgrimes if (ent != NULL) 14631541Srgrimes free(ent, M_IPFW_TBL); 14641541Srgrimes return (0); 14651541Srgrimes} 14661541Srgrimes 14671541Srgrimesstatic int 14681541Srgrimesflush_table(u_int16_t tbl) 146960661Sjlemon{ 14701541Srgrimes struct radix_node_head *rnh; 14718384Sdg 14721541Srgrimes if (tbl >= IPFW_TABLES_MAX) 14731541Srgrimes return (EINVAL); 14741541Srgrimes rnh = ipfw_tables[tbl].rnh; 14751541Srgrimes RADIX_NODE_HEAD_LOCK(rnh); 14761541Srgrimes rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1477133481Sandre ipfw_tables[tbl].modified = 1; 1478133481Sandre RADIX_NODE_HEAD_UNLOCK(rnh); 14791541Srgrimes return (0); 14801541Srgrimes} 14811541Srgrimes 14821541Srgrimesstatic void 14838384Sdgflush_tables(void) 14848384Sdg{ 14851541Srgrimes u_int16_t tbl; 14861541Srgrimes 14871541Srgrimes for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 14881541Srgrimes flush_table(tbl); 148988665Syar} 149088665Syar 149188665Syarstatic int 149288665Syarlookup_table(u_int16_t tbl, in_addr_t addr, u_int32_t *val) 14931541Srgrimes{ 149485467Sjlemon struct radix_node_head *rnh; 149585467Sjlemon struct table_entry *ent; 14961541Srgrimes struct sockaddr_in sa; 149778064Sume static in_addr_t last_addr; 149885467Sjlemon static int last_tbl; 149985467Sjlemon static int last_match; 150078064Sume static u_int32_t last_value; 150178064Sume 150285467Sjlemon if (tbl >= IPFW_TABLES_MAX) 150385467Sjlemon return (0); 150485467Sjlemon if (tbl == last_tbl && addr == last_addr && 150585467Sjlemon !ipfw_tables[tbl].modified) { 15061541Srgrimes if (last_match) 150778064Sume *val = last_value; 15081541Srgrimes return (last_match); 15091541Srgrimes } 151085467Sjlemon rnh = ipfw_tables[tbl].rnh; 151185467Sjlemon sa.sin_len = 8; 151285467Sjlemon sa.sin_addr.s_addr = addr; 15131541Srgrimes RADIX_NODE_HEAD_LOCK(rnh); 15141541Srgrimes ipfw_tables[tbl].modified = 0; 15151541Srgrimes ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 15161541Srgrimes RADIX_NODE_HEAD_UNLOCK(rnh); 15171541Srgrimes last_addr = addr; 151885467Sjlemon last_tbl = tbl; 151985467Sjlemon if (ent != NULL) { 152085467Sjlemon last_value = *val = ent->value; 15211541Srgrimes last_match = 1; 152278064Sume return (1); 15231541Srgrimes } 15241541Srgrimes last_match = 0; 15251541Srgrimes return (0); 1526133481Sandre} 15271541Srgrimes 15288384Sdgstatic int 15298384Sdgcount_table_entry(struct radix_node *rn, void *arg) 153085467Sjlemon{ 1531104774Smaxim u_int32_t * const cnt = arg; 15321541Srgrimes 15331541Srgrimes (*cnt)++; 15341541Srgrimes return (0); 153585467Sjlemon} 153685467Sjlemon 153785467Sjlemonstatic int 15381541Srgrimescount_table(u_int32_t tbl, u_int32_t *cnt) 153978064Sume{ 15408384Sdg struct radix_node_head *rnh; 15411541Srgrimes 1542133481Sandre if (tbl >= IPFW_TABLES_MAX) 15431541Srgrimes return (EINVAL); 154485467Sjlemon rnh = ipfw_tables[tbl].rnh; 1545104774Smaxim *cnt = 0; 15461541Srgrimes RADIX_NODE_HEAD_LOCK(rnh); 15471541Srgrimes rnh->rnh_walktree(rnh, count_table_entry, cnt); 15481541Srgrimes RADIX_NODE_HEAD_UNLOCK(rnh); 154985467Sjlemon return (0); 15501541Srgrimes} 15511541Srgrimes 15521541Srgrimesstatic int 155385467Sjlemondump_table_entry(struct radix_node *rn, void *arg) 155485467Sjlemon{ 15551541Srgrimes struct table_entry * const n = (struct table_entry *)rn; 15561541Srgrimes ipfw_table * const tbl = arg; 155733268Sache ipfw_table_entry *ent; 1558122708Sandre 15591541Srgrimes if (tbl->cnt == tbl->size) 15601541Srgrimes return (1); 15611541Srgrimes ent = &tbl->ent[tbl->cnt]; 15621541Srgrimes ent->tbl = tbl->tbl; 15631541Srgrimes if (in_nullhost(n->mask.sin_addr)) 15641541Srgrimes ent->masklen = 0; 15651541Srgrimes else 15661541Srgrimes ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 15671541Srgrimes ent->addr = n->addr.sin_addr.s_addr; 15681541Srgrimes ent->value = n->value; 15691541Srgrimes tbl->cnt++; 15701541Srgrimes return (0); 15711541Srgrimes} 157287120Sru 1573122708Sandrestatic int 157487120Srudump_table(ipfw_table *tbl) 15751541Srgrimes{ 1576122922Sandre struct radix_node_head *rnh; 1577122708Sandre 1578122708Sandre if (tbl->tbl >= IPFW_TABLES_MAX) 15791541Srgrimes return (EINVAL); 1580122996Sandre rnh = ipfw_tables[tbl->tbl].rnh; 1581122922Sandre tbl->cnt = 0; 1582122708Sandre RADIX_NODE_HEAD_LOCK(rnh); 1583122708Sandre rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1584122708Sandre RADIX_NODE_HEAD_UNLOCK(rnh); 1585122922Sandre return (0); 15861541Srgrimes} 1587122922Sandre 1588122708Sandrestatic void 15891541Srgrimesfill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp) 1590122922Sandre{ 1591122922Sandre struct ucred *cr; 1592122708Sandre 15931541Srgrimes if (inp->inp_socket != NULL) { 15941541Srgrimes cr = inp->inp_socket->so_cred; 15951541Srgrimes ugp->fw_prid = jailed(cr) ? 15961541Srgrimes cr->cr_prison->pr_id : -1; 15971541Srgrimes ugp->fw_uid = cr->cr_uid; 15981541Srgrimes ugp->fw_ngroups = cr->cr_ngroups; 1599104094Sphk bcopy(cr->cr_groups, ugp->fw_groups, 16001541Srgrimes sizeof(ugp->fw_groups)); 16011541Srgrimes } 16021541Srgrimes} 16031541Srgrimes 16041541Srgrimesstatic int 16051541Srgrimescheck_uidgid(ipfw_insn_u32 *insn, 16061541Srgrimes int proto, struct ifnet *oif, 16071541Srgrimes struct in_addr dst_ip, u_int16_t dst_port, 16081541Srgrimes struct in_addr src_ip, u_int16_t src_port, 16091541Srgrimes struct ip_fw_ugid *ugp, int *lookup, struct inpcb *inp) 16101541Srgrimes{ 16111541Srgrimes struct inpcbinfo *pi; 16121541Srgrimes int wildcard; 161316206Sbde struct inpcb *pcb; 16141541Srgrimes int match; 16151541Srgrimes gid_t *gp; 16161541Srgrimes 16171541Srgrimes /* 16181541Srgrimes * Check to see if the UDP or TCP stack supplied us with 16191541Srgrimes * the PCB. If so, rather then holding a lock and looking 16201541Srgrimes * up the PCB, we can use the one that was supplied. 16211541Srgrimes */ 16221541Srgrimes if (inp && *lookup == 0) { 16231541Srgrimes INP_LOCK_ASSERT(inp); 16241541Srgrimes if (inp->inp_socket != NULL) { 16251541Srgrimes fill_ugid_cache(inp, ugp); 16261541Srgrimes *lookup = 1; 16271541Srgrimes } 16281541Srgrimes } 16291541Srgrimes /* 16301541Srgrimes * If we have already been here and the packet has no 1631111119Simp * PCB entry associated with it, then we can safely 1632133481Sandre * assume that this is a no match. 16331541Srgrimes */ 16341541Srgrimes if (*lookup == -1) 16351541Srgrimes return (0); 16361541Srgrimes if (proto == IPPROTO_TCP) { 16371541Srgrimes wildcard = 0; 16381541Srgrimes pi = &tcbinfo; 16391541Srgrimes } else if (proto == IPPROTO_UDP) { 16401541Srgrimes wildcard = 1; 16411541Srgrimes pi = &udbinfo; 16421541Srgrimes } else 16431541Srgrimes return 0; 16441541Srgrimes match = 0; 16451541Srgrimes if (*lookup == 0) { 16461541Srgrimes INP_INFO_RLOCK(pi); 16471541Srgrimes pcb = (oif) ? 16481541Srgrimes in_pcblookup_hash(pi, 16491541Srgrimes dst_ip, htons(dst_port), 16501541Srgrimes src_ip, htons(src_port), 16511541Srgrimes wildcard, oif) : 165241993Sluigi in_pcblookup_hash(pi, 16531541Srgrimes src_ip, htons(src_port), 16541541Srgrimes dst_ip, htons(dst_port), 16551541Srgrimes wildcard, NULL); 16561541Srgrimes if (pcb != NULL) { 16571541Srgrimes INP_LOCK(pcb); 16581541Srgrimes if (pcb->inp_socket != NULL) { 16591541Srgrimes fill_ugid_cache(pcb, ugp); 16608384Sdg *lookup = 1; 16618384Sdg } 16621541Srgrimes INP_UNLOCK(pcb); 16631541Srgrimes } 16641541Srgrimes INP_INFO_RUNLOCK(pi); 16651541Srgrimes if (*lookup == 0) { 16661541Srgrimes /* 16671541Srgrimes * If the lookup did not yield any results, there 16681541Srgrimes * is no sense in coming back and trying again. So 16691541Srgrimes * we can set lookup to -1 and ensure that we wont 16701541Srgrimes * bother the pcb system again. 16711541Srgrimes */ 167241993Sluigi *lookup = -1; 16731541Srgrimes return (0); 16741541Srgrimes } 16751541Srgrimes } 16761541Srgrimes if (insn->o.opcode == O_UID) 16771541Srgrimes match = (ugp->fw_uid == (uid_t)insn->d[0]); 16781541Srgrimes else if (insn->o.opcode == O_GID) { 16791541Srgrimes for (gp = ugp->fw_groups; 16801541Srgrimes gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++) 16811541Srgrimes if (*gp == (gid_t)insn->d[0]) { 168241993Sluigi match = 1; 16831541Srgrimes break; 16841541Srgrimes } 16851541Srgrimes } else if (insn->o.opcode == O_JAIL) 16861541Srgrimes match = (ugp->fw_prid == (int)insn->d[0]); 16871541Srgrimes return match; 16881541Srgrimes} 16891541Srgrimes 16901541Srgrimes/* 16911541Srgrimes * The main check routine for the firewall. 16921541Srgrimes * 16931541Srgrimes * All arguments are in args so we can modify them and return them 16941541Srgrimes * back to the caller. 16951541Srgrimes * 16961541Srgrimes * Parameters: 16971541Srgrimes * 16981541Srgrimes * args->m (in/out) The packet; we set to NULL when/if we nuke it. 16991541Srgrimes * Starts with the IP header. 17001541Srgrimes * args->eh (in) Mac header if present, or NULL for layer3 packet. 17011541Srgrimes * args->oif Outgoing interface, or NULL if packet is incoming. 17021541Srgrimes * The incoming interface is in the mbuf. (in) 17031541Srgrimes * args->divert_rule (in/out) 1704105586Sphk * Skip up to the first rule past this rule number; 17051541Srgrimes * upon return, non-zero port number for divert or tee. 17061541Srgrimes * 17078384Sdg * args->rule Pointer to the last matching rule (in/out) 17081541Srgrimes * args->next_hop Socket we are forwarding to (out). 17091541Srgrimes * args->f_id Addresses grabbed from the packet (out) 17101541Srgrimes * args->cookie a cookie depending on rule action 1711105586Sphk * 1712105586Sphk * Return value: 17131541Srgrimes * 17141541Srgrimes * IP_FW_PASS the packet must be accepted 17151541Srgrimes * IP_FW_DENY the packet must be dropped 17161541Srgrimes * IP_FW_DIVERT divert packet, port in m_tag 17171541Srgrimes * IP_FW_TEE tee packet, port in m_tag 17181541Srgrimes * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 17191541Srgrimes * IP_FW_NETGRAPH into netgraph, cookie args->cookie 1720116462Ssilby * 172182445Sjesper */ 17221541Srgrimes 17231541Srgrimesint 17241541Srgrimesipfw_chk(struct ip_fw_args *args) 17251541Srgrimes{ 17261541Srgrimes /* 17271541Srgrimes * Local variables hold state during the processing of a packet. 17281541Srgrimes * 17291541Srgrimes * IMPORTANT NOTE: to speed up the processing of rules, there 17301541Srgrimes * are some assumption on the values of the variables, which 17311541Srgrimes * are documented here. Should you change them, please check 17321541Srgrimes * the implementation of the various instructions to make sure 17331541Srgrimes * that they still work. 17341541Srgrimes * 17351541Srgrimes * args->eh The MAC header. It is non-null for a layer2 17361541Srgrimes * packet, it is NULL for a layer-3 packet. 17371541Srgrimes * 173812296Sphk * m | args->m Pointer to the mbuf, as received from the caller. 1739122708Sandre * It may change if ipfw_chk() does an m_pullup, or if it 17401541Srgrimes * consumes the packet because it calls send_reject(). 174198613Sluigi * XXX This has to change, so that ipfw_chk() never modifies 1742122708Sandre * or consumes the buffer. 17431549Srgrimes * ip is simply an alias of the value of m, and it is kept 17441541Srgrimes * in sync with it (the packet is supposed to start with 17451541Srgrimes * the ip header). 174688593Sjulian */ 17471541Srgrimes struct mbuf *m = args->m; 1748105199Ssam struct ip *ip = mtod(m, struct ip *); 174955009Sshin 175055009Sshin /* 17511541Srgrimes * For rules which contain uid/gid or jail constraints, cache 175288593Sjulian * a copy of the users credentials after the pcb lookup has been 175388593Sjulian * executed. This will speed up the processing of rules with 175488593Sjulian * these types of constraints, as well as decrease contention 175588593Sjulian * on pcb related locks. 175698613Sluigi */ 175788593Sjulian struct ip_fw_ugid fw_ugid_cache; 17581541Srgrimes int ugid_lookup = 0; 17591541Srgrimes 17603969Sjkh /* 176188593Sjulian * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 176238373Sbde * associated with a packet input on a divert socket. This 17631541Srgrimes * will allow to distinguish traffic and its direction when 17643969Sjkh * it originates from a divert socket. 17653969Sjkh */ 176688593Sjulian u_int divinput_flags = 0; 17671541Srgrimes 17681541Srgrimes /* 17691541Srgrimes * oif | args->oif If NULL, ipfw_chk has been called on the 17701541Srgrimes * inbound path (ether_input, bdg_forward, ip_input). 177144219Sdes * If non-NULL, ipfw_chk has been called on the outbound path 177244219Sdes * (ether_output, ip_output). 177344219Sdes */ 177444219Sdes struct ifnet *oif = args->oif; 177544219Sdes 1776122708Sandre struct ip_fw *f = NULL; /* matching rule */ 177744219Sdes int retval = 0; 177844219Sdes 177944219Sdes /* 17801541Srgrimes * hlen The length of the IPv4 header. 178144219Sdes * hlen >0 means we have an IPv4 packet. 17821541Srgrimes */ 1783133481Sandre u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 1784122708Sandre 178574415Sru /* 1786122708Sandre * offset The offset of a fragment. offset != 0 means that 17871541Srgrimes * we have a fragment at this offset of an IPv4 packet. 17881541Srgrimes * offset == 0 means that (if this is an IPv4 packet) 178973996Siedowse * this is the first or only fragment. 179073996Siedowse */ 179173996Siedowse u_short offset = 0; 179298701Sluigi 179398701Sluigi /* 179498701Sluigi * Local copies of addresses. They are only valid if we have 179598701Sluigi * an IP packet. 179698701Sluigi * 179798701Sluigi * proto The protocol. Set to 0 for non-ip packets, 179898701Sluigi * or to the protocol read from the packet otherwise. 179973996Siedowse * proto != 0 means that we have an IPv4 packet. 180073996Siedowse * 180173996Siedowse * src_port, dst_port port numbers, in HOST format. Only 180273996Siedowse * valid for TCP and UDP packets. 18031541Srgrimes * 1804111119Simp * src_ip, dst_ip ip addresses, in NETWORK format. 1805111119Simp * Only valid for IPv4 packets. 1806108466Ssam */ 1807108466Ssam u_int8_t proto; 1808108466Ssam u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 1809108466Ssam struct in_addr src_ip, dst_ip; /* NOTE: network format */ 1810108466Ssam u_int16_t ip_len=0; 1811108466Ssam int pktlen; 1812108466Ssam int dyn_dir = MATCH_UNKNOWN; 1813108466Ssam ipfw_dyn_rule *q = NULL; 1814108466Ssam struct ip_fw_chain *chain = &layer3_chain; 181573996Siedowse struct m_tag *mtag; 1816105586Sphk 181773996Siedowse if (m->m_flags & M_SKIP_FIREWALL) 1818130580Sbms return (IP_FW_PASS); /* accept */ 181973996Siedowse /* 182073996Siedowse * dyn_dir = MATCH_UNKNOWN when rules unchecked, 18211541Srgrimes * MATCH_NONE when checked and not matched (q = NULL), 182265327Sru * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 182365327Sru */ 182465327Sru 182565327Sru pktlen = m->m_pkthdr.len; 182665327Sru if (args->eh == NULL || /* layer 3 packet */ 182765327Sru ( m->m_pkthdr.len >= sizeof(struct ip) && 182865327Sru ntohs(args->eh->ether_type) == ETHERTYPE_IP)) 182965327Sru hlen = ip->ip_hl << 2; 18301541Srgrimes 18311541Srgrimes /* 18321541Srgrimes * Collect parameters into local variables for faster matching. 18331541Srgrimes */ 18341541Srgrimes if (hlen == 0) { /* do not grab addresses for non-ip pkts */ 18351541Srgrimes proto = args->f_id.proto = 0; /* mark f_id invalid */ 18361541Srgrimes goto after_ip_checks; 18371541Srgrimes } 1838122708Sandre 1839122708Sandre proto = args->f_id.proto = ip->ip_p; 1840122708Sandre src_ip = ip->ip_src; 1841122708Sandre dst_ip = ip->ip_dst; 1842122708Sandre if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 1843122708Sandre offset = ntohs(ip->ip_off) & IP_OFFMASK; 1844122996Sandre ip_len = ntohs(ip->ip_len); 1845122708Sandre } else { 1846122708Sandre offset = ip->ip_off & IP_OFFMASK; 1847122708Sandre ip_len = ip->ip_len; 1848122708Sandre } 1849122921Sandre pktlen = ip_len < pktlen ? ip_len : pktlen; 1850122708Sandre 1851122708Sandre#define PULLUP_TO(len) \ 1852122708Sandre do { \ 1853122708Sandre if ((m)->m_len < (len)) { \ 1854122708Sandre args->m = m = m_pullup(m, (len)); \ 1855122708Sandre if (m == 0) \ 18561541Srgrimes goto pullup_failed; \ 1857122708Sandre ip = mtod(m, struct ip *); \ 18581541Srgrimes } \ 1859122708Sandre } while (0) 1860122708Sandre 1861122708Sandre if (offset == 0) { 1862122708Sandre switch (proto) { 1863122708Sandre case IPPROTO_TCP: 1864122708Sandre { 1865122708Sandre struct tcphdr *tcp; 1866122708Sandre 1867122708Sandre PULLUP_TO(hlen + sizeof(struct tcphdr)); 18681541Srgrimes tcp = L3HDR(struct tcphdr, ip); 1869122708Sandre dst_port = tcp->th_dport; 1870122708Sandre src_port = tcp->th_sport; 18711541Srgrimes args->f_id.flags = tcp->th_flags; 1872122708Sandre } 18731541Srgrimes break; 1874122708Sandre 1875122708Sandre case IPPROTO_UDP: 18761541Srgrimes { 18771541Srgrimes struct udphdr *udp; 1878101268Sluigi 1879126239Smlaier PULLUP_TO(hlen + sizeof(struct udphdr)); 1880126239Smlaier udp = L3HDR(struct udphdr, ip); 1881126239Smlaier dst_port = udp->uh_dport; 1882126239Smlaier src_port = udp->uh_sport; 1883126239Smlaier } 1884126239Smlaier break; 1885126239Smlaier 1886126239Smlaier case IPPROTO_ICMP: 1887101268Sluigi PULLUP_TO(hlen + 4); /* type, code and checksum. */ 1888122708Sandre args->f_id.flags = L3HDR(struct icmp, ip)->icmp_type; 18891541Srgrimes break; 18901541Srgrimes 18911541Srgrimes default: 18921541Srgrimes break; 18931541Srgrimes } 18941541Srgrimes#undef PULLUP_TO 18951541Srgrimes } 1896122702Sandre 18971541Srgrimes args->f_id.src_ip = ntohl(src_ip.s_addr); 18981541Srgrimes args->f_id.dst_ip = ntohl(dst_ip.s_addr); 18991541Srgrimes args->f_id.src_port = src_port = ntohs(src_port); 19001541Srgrimes args->f_id.dst_port = dst_port = ntohs(dst_port); 19011541Srgrimes 19021541Srgrimesafter_ip_checks: 19031541Srgrimes IPFW_RLOCK(chain); 19041541Srgrimes mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 19051541Srgrimes if (args->rule) { 19061541Srgrimes /* 19071541Srgrimes * Packet has already been tagged. Look for the next rule 19081541Srgrimes * to restart processing. 19091541Srgrimes * 19101541Srgrimes * If fw_one_pass != 0 then just accept it. 19111541Srgrimes * XXX should not happen here, but optimized out in 19121541Srgrimes * the caller. 19131541Srgrimes */ 19141541Srgrimes if (fw_one_pass) { 19151541Srgrimes IPFW_RUNLOCK(chain); 19161541Srgrimes return (IP_FW_PASS); 19171541Srgrimes } 19181541Srgrimes 19191541Srgrimes f = args->rule->next_rule; 19201541Srgrimes if (f == NULL) 19211541Srgrimes f = lookup_next_rule(args->rule); 19221541Srgrimes } else { 1923122708Sandre /* 192455009Sshin * Find the starting rule. It can be either the first 192555009Sshin * one, or the one after divert_rule if asked so. 192655009Sshin */ 192755009Sshin int skipto = mtag ? divert_cookie(mtag) : 0; 192855009Sshin 192955009Sshin f = chain->rules; 1930122708Sandre if (args->eh == NULL && skipto != 0) { 193155009Sshin if (skipto >= IPFW_DEFAULT_RULE) { 193255009Sshin IPFW_RUNLOCK(chain); 193355009Sshin return (IP_FW_DENY); /* invalid */ 1934122708Sandre } 193555009Sshin while (f && f->rulenum <= skipto) 1936122708Sandre f = f->next; 193755009Sshin if (f == NULL) { /* drop packet */ 193855009Sshin IPFW_RUNLOCK(chain); 1939122708Sandre return (IP_FW_DENY); 1940122708Sandre } 1941122708Sandre } 1942105199Ssam } 1943105199Ssam /* reset divert rule to avoid confusion later */ 1944122708Sandre if (mtag) { 1945122708Sandre divinput_flags = divert_info(mtag) & 1946122708Sandre (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 1947122708Sandre m_tag_delete(m, mtag); 1948105199Ssam } 1949105199Ssam 1950105199Ssam /* 1951105199Ssam * Now scan the rules, and parse microinstructions for each rule. 1952105199Ssam */ 1953105199Ssam for (; f; f = f->next) { 1954105199Ssam int l, cmdlen; 1955105199Ssam ipfw_insn *cmd; 1956105199Ssam int skip_or; /* skip rest of OR block */ 1957105199Ssam 1958105199Ssamagain: 1959105199Ssam if (set_disable & (1 << f->set) ) 1960105199Ssam continue; 1961105199Ssam 1962105199Ssam skip_or = 0; 1963105199Ssam for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1964105199Ssam l -= cmdlen, cmd += cmdlen) { 1965105199Ssam int match; 1966122708Sandre 1967122708Sandre /* 1968105199Ssam * check_body is a jump target used when we find a 1969130581Sbms * CHECK_STATE, and need to jump to the body of 1970130581Sbms * the target rule. 1971122708Sandre */ 1972105199Ssam 1973105199Ssamcheck_body: 1974105199Ssam cmdlen = F_LEN(cmd); 1975105199Ssam /* 1976105199Ssam * An OR block (insn_1 || .. || insn_n) has the 1977122708Sandre * F_OR bit set in all but the last instruction. 1978122708Sandre * The first match will set "skip_or", and cause 1979122708Sandre * the following instructions to be skipped until 1980105199Ssam * past the one with the F_OR bit clear. 1981122708Sandre */ 1982122708Sandre if (skip_or) { /* skip this instruction */ 1983122708Sandre if ((cmd->len & F_OR) == 0) 1984122708Sandre skip_or = 0; /* next one is good */ 1985122708Sandre continue; 1986122708Sandre } 1987122708Sandre match = 0; /* set to 1 if we succeed */ 1988105199Ssam 1989122708Sandre switch (cmd->opcode) { 19901541Srgrimes /* 19911541Srgrimes * The first set of opcodes compares the packet's 19921541Srgrimes * fields with some pattern, setting 'match' if a 19931541Srgrimes * match is found. At the end of the loop there is 1994107081Ssilby * logic to deal with F_NOT and F_OR flags associated 1995107081Ssilby * with the opcode. 1996107081Ssilby */ 1997107081Ssilby case O_NOP: 1998107081Ssilby match = 1; 1999107081Ssilby break; 2000107081Ssilby 2001107081Ssilby case O_FORWARD_MAC: 2002107081Ssilby printf("ipfw: opcode %d unimplemented\n", 2003107081Ssilby cmd->opcode); 2004107081Ssilby break; 2005107081Ssilby 2006107081Ssilby case O_GID: 2007107081Ssilby case O_UID: 2008107081Ssilby case O_JAIL: 20091541Srgrimes /* 201060612Sru * We only check offset == 0 && proto != 0, 201160612Sru * as this ensures that we have an IPv4 201260612Sru * packet with the ports info. 201360612Sru */ 20141541Srgrimes if (offset!=0) 20151541Srgrimes break; 20161541Srgrimes if (proto == IPPROTO_TCP || 20171541Srgrimes proto == IPPROTO_UDP) 201819622Sfenner match = check_uidgid( 201919622Sfenner (ipfw_insn_u32 *)cmd, 202019622Sfenner proto, oif, 202119622Sfenner dst_ip, dst_port, 202219622Sfenner src_ip, src_port, &fw_ugid_cache, 202319622Sfenner &ugid_lookup, args->inp); 202419622Sfenner break; 2025125264Sphk 2026125264Sphk case O_RECV: 202719622Sfenner match = iface_match(m->m_pkthdr.rcvif, 2028125264Sphk (ipfw_insn_if *)cmd); 2029125264Sphk break; 2030125264Sphk 2031125264Sphk case O_XMIT: 2032125264Sphk match = iface_match(oif, (ipfw_insn_if *)cmd); 2033125264Sphk break; 2034125264Sphk 2035125264Sphk case O_VIA: 2036125264Sphk match = iface_match(oif ? oif : 2037125264Sphk m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2038125264Sphk break; 2039125264Sphk 2040125264Sphk case O_MACADDR2: 2041125264Sphk if (args->eh != NULL) { /* have MAC header */ 2042125264Sphk u_int32_t *want = (u_int32_t *) 2043125264Sphk ((ipfw_insn_mac *)cmd)->addr; 204497658Stanimura u_int32_t *mask = (u_int32_t *) 204519622Sfenner ((ipfw_insn_mac *)cmd)->mask; 204619622Sfenner u_int32_t *hdr = (u_int32_t *)args->eh; 204719622Sfenner 204819622Sfenner match = 204919622Sfenner ( want[0] == (hdr[0] & mask[0]) && 205019622Sfenner want[1] == (hdr[1] & mask[1]) && 2051114258Smdodd want[2] == (hdr[2] & mask[2]) ); 2052114258Smdodd } 2053114258Smdodd break; 2054114258Smdodd 2055114258Smdodd case O_MAC_TYPE: 2056114258Smdodd if (args->eh != NULL) { 205719622Sfenner u_int16_t t = 205819622Sfenner ntohs(args->eh->ether_type); 205919622Sfenner u_int16_t *p = 206019622Sfenner ((ipfw_insn_u16 *)cmd)->ports; 206119622Sfenner int i; 206219622Sfenner 206319622Sfenner for (i = cmdlen - 1; !match && i>0; 206419622Sfenner i--, p += 2) 206519622Sfenner match = (t>=p[0] && t<=p[1]); 206619622Sfenner } 206719622Sfenner break; 206819622Sfenner 206919622Sfenner case O_FRAG: 207019622Sfenner match = (hlen > 0 && offset != 0); 207119622Sfenner break; 207219622Sfenner 207319622Sfenner case O_IN: /* "out" is "not in" */ 207419622Sfenner match = (oif == NULL); 207519622Sfenner break; 207619622Sfenner 207719622Sfenner case O_LAYER2: 207830948Sjulian match = (args->eh != NULL); 207930948Sjulian break; 208030948Sjulian 208130948Sjulian case O_DIVERTED: 208230948Sjulian match = (cmd->arg1 & 1 && divinput_flags & 208330948Sjulian IP_FW_DIVERT_LOOPBACK_FLAG) || 208430948Sjulian (cmd->arg1 & 2 && divinput_flags & 208519622Sfenner IP_FW_DIVERT_OUTPUT_FLAG); 208630948Sjulian break; 208730948Sjulian 208883130Sjlemon case O_PROTO: 208983130Sjlemon /* 209030948Sjulian * We do not allow an arg of 0 so the 209130948Sjulian * check of "proto" only suffices. 209230948Sjulian */ 209330948Sjulian match = (proto == cmd->arg1); 209430948Sjulian break; 209530948Sjulian 209630948Sjulian case O_IP_SRC: 209730948Sjulian match = (hlen > 0 && 209830948Sjulian ((ipfw_insn_ip *)cmd)->addr.s_addr == 209930948Sjulian src_ip.s_addr); 210030948Sjulian break; 210130948Sjulian 210230948Sjulian case O_IP_SRC_LOOKUP: 210330948Sjulian case O_IP_DST_LOOKUP: 210430948Sjulian if (hlen > 0) { 210530948Sjulian uint32_t a = 210630948Sjulian (cmd->opcode == O_IP_DST_LOOKUP) ? 210719622Sfenner dst_ip.s_addr : src_ip.s_addr; 210819622Sfenner uint32_t v; 210919622Sfenner 211019622Sfenner match = lookup_table(cmd->arg1, a, &v); 211119622Sfenner if (!match) 211219622Sfenner break; 211398701Sluigi if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 211498701Sluigi match = 211598701Sluigi ((ipfw_insn_u32 *)cmd)->d[0] == v; 211698701Sluigi } 211798701Sluigi break; 211898701Sluigi 211998701Sluigi case O_IP_SRC_MASK: 212098701Sluigi case O_IP_DST_MASK: 212198701Sluigi if (hlen > 0) { 21221541Srgrimes uint32_t a = 21232531Swollman (cmd->opcode == O_IP_DST_MASK) ? 21242531Swollman dst_ip.s_addr : src_ip.s_addr; 21252531Swollman uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 21262531Swollman int i = cmdlen-1; 2127106968Sluigi 21282531Swollman for (; !match && i>0; i-= 2, p+= 2) 21292531Swollman match = (p[0] == (a & p[1])); 2130106968Sluigi } 21312531Swollman break; 21322531Swollman 21339209Swollman case O_IP_SRC_ME: 21349209Swollman if (hlen > 0) { 21359209Swollman struct ifnet *tif; 21369209Swollman 21379209Swollman INADDR_TO_IFP(src_ip, tif); 21389209Swollman match = (tif != NULL); 21399209Swollman } 21409209Swollman break; 21412531Swollman 21422531Swollman case O_IP_DST_SET: 21432531Swollman case O_IP_SRC_SET: 21442531Swollman if (hlen > 0) { 21452531Swollman u_int32_t *d = (u_int32_t *)(cmd+1); 21462531Swollman u_int32_t addr = 21472531Swollman cmd->opcode == O_IP_DST_SET ? 21482531Swollman args->f_id.dst_ip : 21499209Swollman args->f_id.src_ip; 21509209Swollman 21519209Swollman if (addr < d[0]) 21529209Swollman break; 21539209Swollman addr -= d[0]; /* subtract base */ 21549209Swollman match = (addr < cmd->arg1) && 21559209Swollman ( d[ 1 + (addr>>5)] & 21569209Swollman (1<<(addr & 0x1f)) ); 21572531Swollman } 21582531Swollman break; 2159106968Sluigi 2160106968Sluigi case O_IP_DST: 2161106968Sluigi match = (hlen > 0 && 2162106968Sluigi ((ipfw_insn_ip *)cmd)->addr.s_addr == 2163106968Sluigi dst_ip.s_addr); 2164106968Sluigi break; 2165106968Sluigi 2166106968Sluigi case O_IP_DST_ME: 2167106968Sluigi if (hlen > 0) { 2168106968Sluigi struct ifnet *tif; 2169106968Sluigi 2170106968Sluigi INADDR_TO_IFP(dst_ip, tif); 2171106968Sluigi match = (tif != NULL); 2172106968Sluigi } 2173106968Sluigi break; 2174106968Sluigi 2175106968Sluigi case O_IP_SRCPORT: 2176106968Sluigi case O_IP_DSTPORT: 2177106968Sluigi /* 2178106968Sluigi * offset == 0 && proto != 0 is enough 2179106968Sluigi * to guarantee that we have an IPv4 2180106968Sluigi * packet with port info. 2181106968Sluigi */ 2182106968Sluigi if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2183106968Sluigi && offset == 0) { 2184106968Sluigi u_int16_t x = 2185 (cmd->opcode == O_IP_SRCPORT) ? 2186 src_port : dst_port ; 2187 u_int16_t *p = 2188 ((ipfw_insn_u16 *)cmd)->ports; 2189 int i; 2190 2191 for (i = cmdlen - 1; !match && i>0; 2192 i--, p += 2) 2193 match = (x>=p[0] && x<=p[1]); 2194 } 2195 break; 2196 2197 case O_ICMPTYPE: 2198 match = (offset == 0 && proto==IPPROTO_ICMP && 2199 icmptype_match(ip, (ipfw_insn_u32 *)cmd) ); 2200 break; 2201 2202 case O_IPOPT: 2203 match = (hlen > 0 && ipopts_match(ip, cmd) ); 2204 break; 2205 2206 case O_IPVER: 2207 match = (hlen > 0 && cmd->arg1 == ip->ip_v); 2208 break; 2209 2210 case O_IPID: 2211 case O_IPLEN: 2212 case O_IPTTL: 2213 if (hlen > 0) { /* only for IP packets */ 2214 uint16_t x; 2215 uint16_t *p; 2216 int i; 2217 2218 if (cmd->opcode == O_IPLEN) 2219 x = ip_len; 2220 else if (cmd->opcode == O_IPTTL) 2221 x = ip->ip_ttl; 2222 else /* must be IPID */ 2223 x = ntohs(ip->ip_id); 2224 if (cmdlen == 1) { 2225 match = (cmd->arg1 == x); 2226 break; 2227 } 2228 /* otherwise we have ranges */ 2229 p = ((ipfw_insn_u16 *)cmd)->ports; 2230 i = cmdlen - 1; 2231 for (; !match && i>0; i--, p += 2) 2232 match = (x >= p[0] && x <= p[1]); 2233 } 2234 break; 2235 2236 case O_IPPRECEDENCE: 2237 match = (hlen > 0 && 2238 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2239 break; 2240 2241 case O_IPTOS: 2242 match = (hlen > 0 && 2243 flags_match(cmd, ip->ip_tos)); 2244 break; 2245 2246 case O_TCPDATALEN: 2247 if (proto == IPPROTO_TCP && offset == 0) { 2248 struct tcphdr *tcp; 2249 uint16_t x; 2250 uint16_t *p; 2251 int i; 2252 2253 tcp = L3HDR(struct tcphdr,ip); 2254 x = ip_len - 2255 ((ip->ip_hl + tcp->th_off) << 2); 2256 if (cmdlen == 1) { 2257 match = (cmd->arg1 == x); 2258 break; 2259 } 2260 /* otherwise we have ranges */ 2261 p = ((ipfw_insn_u16 *)cmd)->ports; 2262 i = cmdlen - 1; 2263 for (; !match && i>0; i--, p += 2) 2264 match = (x >= p[0] && x <= p[1]); 2265 } 2266 break; 2267 2268 case O_TCPFLAGS: 2269 match = (proto == IPPROTO_TCP && offset == 0 && 2270 flags_match(cmd, 2271 L3HDR(struct tcphdr,ip)->th_flags)); 2272 break; 2273 2274 case O_TCPOPTS: 2275 match = (proto == IPPROTO_TCP && offset == 0 && 2276 tcpopts_match(ip, cmd)); 2277 break; 2278 2279 case O_TCPSEQ: 2280 match = (proto == IPPROTO_TCP && offset == 0 && 2281 ((ipfw_insn_u32 *)cmd)->d[0] == 2282 L3HDR(struct tcphdr,ip)->th_seq); 2283 break; 2284 2285 case O_TCPACK: 2286 match = (proto == IPPROTO_TCP && offset == 0 && 2287 ((ipfw_insn_u32 *)cmd)->d[0] == 2288 L3HDR(struct tcphdr,ip)->th_ack); 2289 break; 2290 2291 case O_TCPWIN: 2292 match = (proto == IPPROTO_TCP && offset == 0 && 2293 cmd->arg1 == 2294 L3HDR(struct tcphdr,ip)->th_win); 2295 break; 2296 2297 case O_ESTAB: 2298 /* reject packets which have SYN only */ 2299 /* XXX should i also check for TH_ACK ? */ 2300 match = (proto == IPPROTO_TCP && offset == 0 && 2301 (L3HDR(struct tcphdr,ip)->th_flags & 2302 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2303 break; 2304 2305 case O_ALTQ: { 2306 struct altq_tag *at; 2307 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2308 2309 match = 1; 2310 mtag = m_tag_get(PACKET_TAG_PF_QID, 2311 sizeof(struct altq_tag), 2312 M_NOWAIT); 2313 if (mtag == NULL) { 2314 /* 2315 * Let the packet fall back to the 2316 * default ALTQ. 2317 */ 2318 break; 2319 } 2320 at = (struct altq_tag *)(mtag+1); 2321 at->qid = altq->qid; 2322 if (hlen != 0) 2323 at->af = AF_INET; 2324 else 2325 at->af = AF_LINK; 2326 at->hdr = ip; 2327 m_tag_prepend(m, mtag); 2328 break; 2329 } 2330 2331 case O_LOG: 2332 if (fw_verbose) 2333 ipfw_log(f, hlen, args->eh, m, oif); 2334 match = 1; 2335 break; 2336 2337 case O_PROB: 2338 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2339 break; 2340 2341 case O_VERREVPATH: 2342 /* Outgoing packets automatically pass/match */ 2343 match = (hlen > 0 && ((oif != NULL) || 2344 (m->m_pkthdr.rcvif == NULL) || 2345 verify_path(src_ip, m->m_pkthdr.rcvif))); 2346 break; 2347 2348 case O_VERSRCREACH: 2349 /* Outgoing packets automatically pass/match */ 2350 match = (hlen > 0 && ((oif != NULL) || 2351 verify_path(src_ip, NULL))); 2352 break; 2353 2354 case O_ANTISPOOF: 2355 /* Outgoing packets automatically pass/match */ 2356 if (oif == NULL && hlen > 0 && 2357 in_localaddr(src_ip)) 2358 match = verify_path(src_ip, 2359 m->m_pkthdr.rcvif); 2360 else 2361 match = 1; 2362 break; 2363 2364 case O_IPSEC: 2365#ifdef FAST_IPSEC 2366 match = (m_tag_find(m, 2367 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2368#endif 2369#ifdef IPSEC 2370 match = (ipsec_getnhist(m) != 0); 2371#endif 2372 /* otherwise no match */ 2373 break; 2374 2375 /* 2376 * The second set of opcodes represents 'actions', 2377 * i.e. the terminal part of a rule once the packet 2378 * matches all previous patterns. 2379 * Typically there is only one action for each rule, 2380 * and the opcode is stored at the end of the rule 2381 * (but there are exceptions -- see below). 2382 * 2383 * In general, here we set retval and terminate the 2384 * outer loop (would be a 'break 3' in some language, 2385 * but we need to do a 'goto done'). 2386 * 2387 * Exceptions: 2388 * O_COUNT and O_SKIPTO actions: 2389 * instead of terminating, we jump to the next rule 2390 * ('goto next_rule', equivalent to a 'break 2'), 2391 * or to the SKIPTO target ('goto again' after 2392 * having set f, cmd and l), respectively. 2393 * 2394 * O_LOG and O_ALTQ action parameters: 2395 * perform some action and set match = 1; 2396 * 2397 * O_LIMIT and O_KEEP_STATE: these opcodes are 2398 * not real 'actions', and are stored right 2399 * before the 'action' part of the rule. 2400 * These opcodes try to install an entry in the 2401 * state tables; if successful, we continue with 2402 * the next opcode (match=1; break;), otherwise 2403 * the packet * must be dropped 2404 * ('goto done' after setting retval); 2405 * 2406 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2407 * cause a lookup of the state table, and a jump 2408 * to the 'action' part of the parent rule 2409 * ('goto check_body') if an entry is found, or 2410 * (CHECK_STATE only) a jump to the next rule if 2411 * the entry is not found ('goto next_rule'). 2412 * The result of the lookup is cached to make 2413 * further instances of these opcodes are 2414 * effectively NOPs. 2415 */ 2416 case O_LIMIT: 2417 case O_KEEP_STATE: 2418 if (install_state(f, 2419 (ipfw_insn_limit *)cmd, args)) { 2420 retval = IP_FW_DENY; 2421 goto done; /* error/limit violation */ 2422 } 2423 match = 1; 2424 break; 2425 2426 case O_PROBE_STATE: 2427 case O_CHECK_STATE: 2428 /* 2429 * dynamic rules are checked at the first 2430 * keep-state or check-state occurrence, 2431 * with the result being stored in dyn_dir. 2432 * The compiler introduces a PROBE_STATE 2433 * instruction for us when we have a 2434 * KEEP_STATE (because PROBE_STATE needs 2435 * to be run first). 2436 */ 2437 if (dyn_dir == MATCH_UNKNOWN && 2438 (q = lookup_dyn_rule(&args->f_id, 2439 &dyn_dir, proto == IPPROTO_TCP ? 2440 L3HDR(struct tcphdr, ip) : NULL)) 2441 != NULL) { 2442 /* 2443 * Found dynamic entry, update stats 2444 * and jump to the 'action' part of 2445 * the parent rule. 2446 */ 2447 q->pcnt++; 2448 q->bcnt += pktlen; 2449 f = q->rule; 2450 cmd = ACTION_PTR(f); 2451 l = f->cmd_len - f->act_ofs; 2452 IPFW_DYN_UNLOCK(); 2453 goto check_body; 2454 } 2455 /* 2456 * Dynamic entry not found. If CHECK_STATE, 2457 * skip to next rule, if PROBE_STATE just 2458 * ignore and continue with next opcode. 2459 */ 2460 if (cmd->opcode == O_CHECK_STATE) 2461 goto next_rule; 2462 match = 1; 2463 break; 2464 2465 case O_ACCEPT: 2466 retval = 0; /* accept */ 2467 goto done; 2468 2469 case O_PIPE: 2470 case O_QUEUE: 2471 args->rule = f; /* report matching rule */ 2472 args->cookie = cmd->arg1; 2473 retval = IP_FW_DUMMYNET; 2474 goto done; 2475 2476 case O_DIVERT: 2477 case O_TEE: { 2478 struct divert_tag *dt; 2479 2480 if (args->eh) /* not on layer 2 */ 2481 break; 2482 mtag = m_tag_get(PACKET_TAG_DIVERT, 2483 sizeof(struct divert_tag), 2484 M_NOWAIT); 2485 if (mtag == NULL) { 2486 /* XXX statistic */ 2487 /* drop packet */ 2488 IPFW_RUNLOCK(chain); 2489 return (IP_FW_DENY); 2490 } 2491 dt = (struct divert_tag *)(mtag+1); 2492 dt->cookie = f->rulenum; 2493 dt->info = cmd->arg1; 2494 m_tag_prepend(m, mtag); 2495 retval = (cmd->opcode == O_DIVERT) ? 2496 IP_FW_DIVERT : IP_FW_TEE; 2497 goto done; 2498 } 2499 2500 case O_COUNT: 2501 case O_SKIPTO: 2502 f->pcnt++; /* update stats */ 2503 f->bcnt += pktlen; 2504 f->timestamp = time_second; 2505 if (cmd->opcode == O_COUNT) 2506 goto next_rule; 2507 /* handle skipto */ 2508 if (f->next_rule == NULL) 2509 lookup_next_rule(f); 2510 f = f->next_rule; 2511 goto again; 2512 2513 case O_REJECT: 2514 /* 2515 * Drop the packet and send a reject notice 2516 * if the packet is not ICMP (or is an ICMP 2517 * query), and it is not multicast/broadcast. 2518 */ 2519 if (hlen > 0 && 2520 (proto != IPPROTO_ICMP || 2521 is_icmp_query(ip)) && 2522 !(m->m_flags & (M_BCAST|M_MCAST)) && 2523 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 2524 send_reject(args, cmd->arg1, 2525 offset,ip_len); 2526 m = args->m; 2527 } 2528 /* FALLTHROUGH */ 2529 case O_DENY: 2530 retval = IP_FW_DENY; 2531 goto done; 2532 2533 case O_FORWARD_IP: 2534 if (args->eh) /* not valid on layer2 pkts */ 2535 break; 2536 if (!q || dyn_dir == MATCH_FORWARD) 2537 args->next_hop = 2538 &((ipfw_insn_sa *)cmd)->sa; 2539 retval = IP_FW_PASS; 2540 goto done; 2541 2542 case O_NETGRAPH: 2543 case O_NGTEE: 2544 args->rule = f; /* report matching rule */ 2545 args->cookie = cmd->arg1; 2546 retval = (cmd->opcode == O_NETGRAPH) ? 2547 IP_FW_NETGRAPH : IP_FW_NGTEE; 2548 goto done; 2549 2550 default: 2551 panic("-- unknown opcode %d\n", cmd->opcode); 2552 } /* end of switch() on opcodes */ 2553 2554 if (cmd->len & F_NOT) 2555 match = !match; 2556 2557 if (match) { 2558 if (cmd->len & F_OR) 2559 skip_or = 1; 2560 } else { 2561 if (!(cmd->len & F_OR)) /* not an OR block, */ 2562 break; /* try next rule */ 2563 } 2564 2565 } /* end of inner for, scan opcodes */ 2566 2567next_rule:; /* try next rule */ 2568 2569 } /* end of outer for, scan rules */ 2570 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 2571 IPFW_RUNLOCK(chain); 2572 return (IP_FW_DENY); 2573 2574done: 2575 /* Update statistics */ 2576 f->pcnt++; 2577 f->bcnt += pktlen; 2578 f->timestamp = time_second; 2579 IPFW_RUNLOCK(chain); 2580 return (retval); 2581 2582pullup_failed: 2583 if (fw_verbose) 2584 printf("ipfw: pullup failed\n"); 2585 return (IP_FW_DENY); 2586} 2587 2588/* 2589 * When a rule is added/deleted, clear the next_rule pointers in all rules. 2590 * These will be reconstructed on the fly as packets are matched. 2591 */ 2592static void 2593flush_rule_ptrs(struct ip_fw_chain *chain) 2594{ 2595 struct ip_fw *rule; 2596 2597 IPFW_WLOCK_ASSERT(chain); 2598 2599 for (rule = chain->rules; rule; rule = rule->next) 2600 rule->next_rule = NULL; 2601} 2602 2603/* 2604 * When pipes/queues are deleted, clear the "pipe_ptr" pointer to a given 2605 * pipe/queue, or to all of them (match == NULL). 2606 */ 2607void 2608flush_pipe_ptrs(struct dn_flow_set *match) 2609{ 2610 struct ip_fw *rule; 2611 2612 IPFW_WLOCK(&layer3_chain); 2613 for (rule = layer3_chain.rules; rule; rule = rule->next) { 2614 ipfw_insn_pipe *cmd = (ipfw_insn_pipe *)ACTION_PTR(rule); 2615 2616 if (cmd->o.opcode != O_PIPE && cmd->o.opcode != O_QUEUE) 2617 continue; 2618 /* 2619 * XXX Use bcmp/bzero to handle pipe_ptr to overcome 2620 * possible alignment problems on 64-bit architectures. 2621 * This code is seldom used so we do not worry too 2622 * much about efficiency. 2623 */ 2624 if (match == NULL || 2625 !bcmp(&cmd->pipe_ptr, &match, sizeof(match)) ) 2626 bzero(&cmd->pipe_ptr, sizeof(cmd->pipe_ptr)); 2627 } 2628 IPFW_WUNLOCK(&layer3_chain); 2629} 2630 2631/* 2632 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 2633 * possibly create a rule number and add the rule to the list. 2634 * Update the rule_number in the input struct so the caller knows it as well. 2635 */ 2636static int 2637add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 2638{ 2639 struct ip_fw *rule, *f, *prev; 2640 int l = RULESIZE(input_rule); 2641 2642 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 2643 return (EINVAL); 2644 2645 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 2646 if (rule == NULL) 2647 return (ENOSPC); 2648 2649 bcopy(input_rule, rule, l); 2650 2651 rule->next = NULL; 2652 rule->next_rule = NULL; 2653 2654 rule->pcnt = 0; 2655 rule->bcnt = 0; 2656 rule->timestamp = 0; 2657 2658 IPFW_WLOCK(chain); 2659 2660 if (chain->rules == NULL) { /* default rule */ 2661 chain->rules = rule; 2662 goto done; 2663 } 2664 2665 /* 2666 * If rulenum is 0, find highest numbered rule before the 2667 * default rule, and add autoinc_step 2668 */ 2669 if (autoinc_step < 1) 2670 autoinc_step = 1; 2671 else if (autoinc_step > 1000) 2672 autoinc_step = 1000; 2673 if (rule->rulenum == 0) { 2674 /* 2675 * locate the highest numbered rule before default 2676 */ 2677 for (f = chain->rules; f; f = f->next) { 2678 if (f->rulenum == IPFW_DEFAULT_RULE) 2679 break; 2680 rule->rulenum = f->rulenum; 2681 } 2682 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 2683 rule->rulenum += autoinc_step; 2684 input_rule->rulenum = rule->rulenum; 2685 } 2686 2687 /* 2688 * Now insert the new rule in the right place in the sorted list. 2689 */ 2690 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 2691 if (f->rulenum > rule->rulenum) { /* found the location */ 2692 if (prev) { 2693 rule->next = f; 2694 prev->next = rule; 2695 } else { /* head insert */ 2696 rule->next = chain->rules; 2697 chain->rules = rule; 2698 } 2699 break; 2700 } 2701 } 2702 flush_rule_ptrs(chain); 2703done: 2704 static_count++; 2705 static_len += l; 2706 IPFW_WUNLOCK(chain); 2707 DEB(printf("ipfw: installed rule %d, static count now %d\n", 2708 rule->rulenum, static_count);) 2709 return (0); 2710} 2711 2712/** 2713 * Remove a static rule (including derived * dynamic rules) 2714 * and place it on the ``reap list'' for later reclamation. 2715 * The caller is in charge of clearing rule pointers to avoid 2716 * dangling pointers. 2717 * @return a pointer to the next entry. 2718 * Arguments are not checked, so they better be correct. 2719 */ 2720static struct ip_fw * 2721remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 2722{ 2723 struct ip_fw *n; 2724 int l = RULESIZE(rule); 2725 2726 IPFW_WLOCK_ASSERT(chain); 2727 2728 n = rule->next; 2729 IPFW_DYN_LOCK(); 2730 remove_dyn_rule(rule, NULL /* force removal */); 2731 IPFW_DYN_UNLOCK(); 2732 if (prev == NULL) 2733 chain->rules = n; 2734 else 2735 prev->next = n; 2736 static_count--; 2737 static_len -= l; 2738 2739 rule->next = chain->reap; 2740 chain->reap = rule; 2741 2742 return n; 2743} 2744 2745/** 2746 * Reclaim storage associated with a list of rules. This is 2747 * typically the list created using remove_rule. 2748 */ 2749static void 2750reap_rules(struct ip_fw *head) 2751{ 2752 struct ip_fw *rule; 2753 2754 while ((rule = head) != NULL) { 2755 head = head->next; 2756 if (DUMMYNET_LOADED) 2757 ip_dn_ruledel_ptr(rule); 2758 free(rule, M_IPFW); 2759 } 2760} 2761 2762/* 2763 * Remove all rules from a chain (except rules in set RESVD_SET 2764 * unless kill_default = 1). The caller is responsible for 2765 * reclaiming storage for the rules left in chain->reap. 2766 */ 2767static void 2768free_chain(struct ip_fw_chain *chain, int kill_default) 2769{ 2770 struct ip_fw *prev, *rule; 2771 2772 IPFW_WLOCK_ASSERT(chain); 2773 2774 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 2775 for (prev = NULL, rule = chain->rules; rule ; ) 2776 if (kill_default || rule->set != RESVD_SET) 2777 rule = remove_rule(chain, rule, prev); 2778 else { 2779 prev = rule; 2780 rule = rule->next; 2781 } 2782} 2783 2784/** 2785 * Remove all rules with given number, and also do set manipulation. 2786 * Assumes chain != NULL && *chain != NULL. 2787 * 2788 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 2789 * the next 8 bits are the new set, the top 8 bits are the command: 2790 * 2791 * 0 delete rules with given number 2792 * 1 delete rules with given set number 2793 * 2 move rules with given number to new set 2794 * 3 move rules with given set number to new set 2795 * 4 swap sets with given numbers 2796 */ 2797static int 2798del_entry(struct ip_fw_chain *chain, u_int32_t arg) 2799{ 2800 struct ip_fw *prev = NULL, *rule; 2801 u_int16_t rulenum; /* rule or old_set */ 2802 u_int8_t cmd, new_set; 2803 2804 rulenum = arg & 0xffff; 2805 cmd = (arg >> 24) & 0xff; 2806 new_set = (arg >> 16) & 0xff; 2807 2808 if (cmd > 4) 2809 return EINVAL; 2810 if (new_set > RESVD_SET) 2811 return EINVAL; 2812 if (cmd == 0 || cmd == 2) { 2813 if (rulenum >= IPFW_DEFAULT_RULE) 2814 return EINVAL; 2815 } else { 2816 if (rulenum > RESVD_SET) /* old_set */ 2817 return EINVAL; 2818 } 2819 2820 IPFW_WLOCK(chain); 2821 rule = chain->rules; 2822 chain->reap = NULL; 2823 switch (cmd) { 2824 case 0: /* delete rules with given number */ 2825 /* 2826 * locate first rule to delete 2827 */ 2828 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 2829 ; 2830 if (rule->rulenum != rulenum) { 2831 IPFW_WUNLOCK(chain); 2832 return EINVAL; 2833 } 2834 2835 /* 2836 * flush pointers outside the loop, then delete all matching 2837 * rules. prev remains the same throughout the cycle. 2838 */ 2839 flush_rule_ptrs(chain); 2840 while (rule->rulenum == rulenum) 2841 rule = remove_rule(chain, rule, prev); 2842 break; 2843 2844 case 1: /* delete all rules with given set number */ 2845 flush_rule_ptrs(chain); 2846 rule = chain->rules; 2847 while (rule->rulenum < IPFW_DEFAULT_RULE) 2848 if (rule->set == rulenum) 2849 rule = remove_rule(chain, rule, prev); 2850 else { 2851 prev = rule; 2852 rule = rule->next; 2853 } 2854 break; 2855 2856 case 2: /* move rules with given number to new set */ 2857 rule = chain->rules; 2858 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 2859 if (rule->rulenum == rulenum) 2860 rule->set = new_set; 2861 break; 2862 2863 case 3: /* move rules with given set number to new set */ 2864 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 2865 if (rule->set == rulenum) 2866 rule->set = new_set; 2867 break; 2868 2869 case 4: /* swap two sets */ 2870 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 2871 if (rule->set == rulenum) 2872 rule->set = new_set; 2873 else if (rule->set == new_set) 2874 rule->set = rulenum; 2875 break; 2876 } 2877 /* 2878 * Look for rules to reclaim. We grab the list before 2879 * releasing the lock then reclaim them w/o the lock to 2880 * avoid a LOR with dummynet. 2881 */ 2882 rule = chain->reap; 2883 chain->reap = NULL; 2884 IPFW_WUNLOCK(chain); 2885 if (rule) 2886 reap_rules(rule); 2887 return 0; 2888} 2889 2890/* 2891 * Clear counters for a specific rule. 2892 * The enclosing "table" is assumed locked. 2893 */ 2894static void 2895clear_counters(struct ip_fw *rule, int log_only) 2896{ 2897 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 2898 2899 if (log_only == 0) { 2900 rule->bcnt = rule->pcnt = 0; 2901 rule->timestamp = 0; 2902 } 2903 if (l->o.opcode == O_LOG) 2904 l->log_left = l->max_log; 2905} 2906 2907/** 2908 * Reset some or all counters on firewall rules. 2909 * @arg frwl is null to clear all entries, or contains a specific 2910 * rule number. 2911 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 2912 */ 2913static int 2914zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 2915{ 2916 struct ip_fw *rule; 2917 char *msg; 2918 2919 IPFW_WLOCK(chain); 2920 if (rulenum == 0) { 2921 norule_counter = 0; 2922 for (rule = chain->rules; rule; rule = rule->next) 2923 clear_counters(rule, log_only); 2924 msg = log_only ? "ipfw: All logging counts reset.\n" : 2925 "ipfw: Accounting cleared.\n"; 2926 } else { 2927 int cleared = 0; 2928 /* 2929 * We can have multiple rules with the same number, so we 2930 * need to clear them all. 2931 */ 2932 for (rule = chain->rules; rule; rule = rule->next) 2933 if (rule->rulenum == rulenum) { 2934 while (rule && rule->rulenum == rulenum) { 2935 clear_counters(rule, log_only); 2936 rule = rule->next; 2937 } 2938 cleared = 1; 2939 break; 2940 } 2941 if (!cleared) { /* we did not find any matching rules */ 2942 IPFW_WUNLOCK(chain); 2943 return (EINVAL); 2944 } 2945 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 2946 "ipfw: Entry %d cleared.\n"; 2947 } 2948 IPFW_WUNLOCK(chain); 2949 2950 if (fw_verbose) 2951 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 2952 return (0); 2953} 2954 2955/* 2956 * Check validity of the structure before insert. 2957 * Fortunately rules are simple, so this mostly need to check rule sizes. 2958 */ 2959static int 2960check_ipfw_struct(struct ip_fw *rule, int size) 2961{ 2962 int l, cmdlen = 0; 2963 int have_action=0; 2964 ipfw_insn *cmd; 2965 2966 if (size < sizeof(*rule)) { 2967 printf("ipfw: rule too short\n"); 2968 return (EINVAL); 2969 } 2970 /* first, check for valid size */ 2971 l = RULESIZE(rule); 2972 if (l != size) { 2973 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 2974 return (EINVAL); 2975 } 2976 if (rule->act_ofs >= rule->cmd_len) { 2977 printf("ipfw: bogus action offset (%u > %u)\n", 2978 rule->act_ofs, rule->cmd_len - 1); 2979 return (EINVAL); 2980 } 2981 /* 2982 * Now go for the individual checks. Very simple ones, basically only 2983 * instruction sizes. 2984 */ 2985 for (l = rule->cmd_len, cmd = rule->cmd ; 2986 l > 0 ; l -= cmdlen, cmd += cmdlen) { 2987 cmdlen = F_LEN(cmd); 2988 if (cmdlen > l) { 2989 printf("ipfw: opcode %d size truncated\n", 2990 cmd->opcode); 2991 return EINVAL; 2992 } 2993 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 2994 switch (cmd->opcode) { 2995 case O_PROBE_STATE: 2996 case O_KEEP_STATE: 2997 case O_PROTO: 2998 case O_IP_SRC_ME: 2999 case O_IP_DST_ME: 3000 case O_LAYER2: 3001 case O_IN: 3002 case O_FRAG: 3003 case O_DIVERTED: 3004 case O_IPOPT: 3005 case O_IPTOS: 3006 case O_IPPRECEDENCE: 3007 case O_IPVER: 3008 case O_TCPWIN: 3009 case O_TCPFLAGS: 3010 case O_TCPOPTS: 3011 case O_ESTAB: 3012 case O_VERREVPATH: 3013 case O_VERSRCREACH: 3014 case O_ANTISPOOF: 3015 case O_IPSEC: 3016 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3017 goto bad_size; 3018 break; 3019 3020 case O_UID: 3021 case O_GID: 3022 case O_JAIL: 3023 case O_IP_SRC: 3024 case O_IP_DST: 3025 case O_TCPSEQ: 3026 case O_TCPACK: 3027 case O_PROB: 3028 case O_ICMPTYPE: 3029 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3030 goto bad_size; 3031 break; 3032 3033 case O_LIMIT: 3034 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3035 goto bad_size; 3036 break; 3037 3038 case O_LOG: 3039 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3040 goto bad_size; 3041 3042 ((ipfw_insn_log *)cmd)->log_left = 3043 ((ipfw_insn_log *)cmd)->max_log; 3044 3045 break; 3046 3047 case O_IP_SRC_MASK: 3048 case O_IP_DST_MASK: 3049 /* only odd command lengths */ 3050 if ( !(cmdlen & 1) || cmdlen > 31) 3051 goto bad_size; 3052 break; 3053 3054 case O_IP_SRC_SET: 3055 case O_IP_DST_SET: 3056 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3057 printf("ipfw: invalid set size %d\n", 3058 cmd->arg1); 3059 return EINVAL; 3060 } 3061 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3062 (cmd->arg1+31)/32 ) 3063 goto bad_size; 3064 break; 3065 3066 case O_IP_SRC_LOOKUP: 3067 case O_IP_DST_LOOKUP: 3068 if (cmd->arg1 >= IPFW_TABLES_MAX) { 3069 printf("ipfw: invalid table number %d\n", 3070 cmd->arg1); 3071 return (EINVAL); 3072 } 3073 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 3074 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3075 goto bad_size; 3076 break; 3077 3078 case O_MACADDR2: 3079 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 3080 goto bad_size; 3081 break; 3082 3083 case O_NOP: 3084 case O_IPID: 3085 case O_IPTTL: 3086 case O_IPLEN: 3087 case O_TCPDATALEN: 3088 if (cmdlen < 1 || cmdlen > 31) 3089 goto bad_size; 3090 break; 3091 3092 case O_MAC_TYPE: 3093 case O_IP_SRCPORT: 3094 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 3095 if (cmdlen < 2 || cmdlen > 31) 3096 goto bad_size; 3097 break; 3098 3099 case O_RECV: 3100 case O_XMIT: 3101 case O_VIA: 3102 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 3103 goto bad_size; 3104 break; 3105 3106 case O_ALTQ: 3107 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 3108 goto bad_size; 3109 break; 3110 3111 case O_PIPE: 3112 case O_QUEUE: 3113 if (cmdlen != F_INSN_SIZE(ipfw_insn_pipe)) 3114 goto bad_size; 3115 goto check_action; 3116 3117 case O_FORWARD_IP: 3118#ifdef IPFIREWALL_FORWARD 3119 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 3120 goto bad_size; 3121 goto check_action; 3122#else 3123 return EINVAL; 3124#endif 3125 3126 case O_DIVERT: 3127 case O_TEE: 3128 if (ip_divert_ptr == NULL) 3129 return EINVAL; 3130 else 3131 goto check_size; 3132 case O_NETGRAPH: 3133 case O_NGTEE: 3134 if (!NG_IPFW_LOADED) 3135 return EINVAL; 3136 else 3137 goto check_size; 3138 case O_FORWARD_MAC: /* XXX not implemented yet */ 3139 case O_CHECK_STATE: 3140 case O_COUNT: 3141 case O_ACCEPT: 3142 case O_DENY: 3143 case O_REJECT: 3144 case O_SKIPTO: 3145check_size: 3146 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3147 goto bad_size; 3148check_action: 3149 if (have_action) { 3150 printf("ipfw: opcode %d, multiple actions" 3151 " not allowed\n", 3152 cmd->opcode); 3153 return EINVAL; 3154 } 3155 have_action = 1; 3156 if (l != cmdlen) { 3157 printf("ipfw: opcode %d, action must be" 3158 " last opcode\n", 3159 cmd->opcode); 3160 return EINVAL; 3161 } 3162 break; 3163 default: 3164 printf("ipfw: opcode %d, unknown opcode\n", 3165 cmd->opcode); 3166 return EINVAL; 3167 } 3168 } 3169 if (have_action == 0) { 3170 printf("ipfw: missing action\n"); 3171 return EINVAL; 3172 } 3173 return 0; 3174 3175bad_size: 3176 printf("ipfw: opcode %d size %d wrong\n", 3177 cmd->opcode, cmdlen); 3178 return EINVAL; 3179} 3180 3181/* 3182 * Copy the static and dynamic rules to the supplied buffer 3183 * and return the amount of space actually used. 3184 */ 3185static size_t 3186ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 3187{ 3188 char *bp = buf; 3189 char *ep = bp + space; 3190 struct ip_fw *rule; 3191 int i; 3192 3193 /* XXX this can take a long time and locking will block packet flow */ 3194 IPFW_RLOCK(chain); 3195 for (rule = chain->rules; rule ; rule = rule->next) { 3196 /* 3197 * Verify the entry fits in the buffer in case the 3198 * rules changed between calculating buffer space and 3199 * now. This would be better done using a generation 3200 * number but should suffice for now. 3201 */ 3202 i = RULESIZE(rule); 3203 if (bp + i <= ep) { 3204 bcopy(rule, bp, i); 3205 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 3206 sizeof(set_disable)); 3207 bp += i; 3208 } 3209 } 3210 IPFW_RUNLOCK(chain); 3211 if (ipfw_dyn_v) { 3212 ipfw_dyn_rule *p, *last = NULL; 3213 3214 IPFW_DYN_LOCK(); 3215 for (i = 0 ; i < curr_dyn_buckets; i++) 3216 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 3217 if (bp + sizeof *p <= ep) { 3218 ipfw_dyn_rule *dst = 3219 (ipfw_dyn_rule *)bp; 3220 bcopy(p, dst, sizeof *p); 3221 bcopy(&(p->rule->rulenum), &(dst->rule), 3222 sizeof(p->rule->rulenum)); 3223 /* 3224 * store a non-null value in "next". 3225 * The userland code will interpret a 3226 * NULL here as a marker 3227 * for the last dynamic rule. 3228 */ 3229 bcopy(&dst, &dst->next, sizeof(dst)); 3230 last = dst; 3231 dst->expire = 3232 TIME_LEQ(dst->expire, time_second) ? 3233 0 : dst->expire - time_second ; 3234 bp += sizeof(ipfw_dyn_rule); 3235 } 3236 } 3237 IPFW_DYN_UNLOCK(); 3238 if (last != NULL) /* mark last dynamic rule */ 3239 bzero(&last->next, sizeof(last)); 3240 } 3241 return (bp - (char *)buf); 3242} 3243 3244 3245/** 3246 * {set|get}sockopt parser. 3247 */ 3248static int 3249ipfw_ctl(struct sockopt *sopt) 3250{ 3251#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3252 int error, rule_num; 3253 size_t size; 3254 struct ip_fw *buf, *rule; 3255 u_int32_t rulenum[2]; 3256 3257 error = suser(sopt->sopt_td); 3258 if (error) 3259 return (error); 3260 3261 /* 3262 * Disallow modifications in really-really secure mode, but still allow 3263 * the logging counters to be reset. 3264 */ 3265 if (sopt->sopt_name == IP_FW_ADD || 3266 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 3267#if __FreeBSD_version >= 500034 3268 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3269 if (error) 3270 return (error); 3271#else /* FreeBSD 4.x */ 3272 if (securelevel >= 3) 3273 return (EPERM); 3274#endif 3275 } 3276 3277 error = 0; 3278 3279 switch (sopt->sopt_name) { 3280 case IP_FW_GET: 3281 /* 3282 * pass up a copy of the current rules. Static rules 3283 * come first (the last of which has number IPFW_DEFAULT_RULE), 3284 * followed by a possibly empty list of dynamic rule. 3285 * The last dynamic rule has NULL in the "next" field. 3286 * 3287 * Note that the calculated size is used to bound the 3288 * amount of data returned to the user. The rule set may 3289 * change between calculating the size and returning the 3290 * data in which case we'll just return what fits. 3291 */ 3292 size = static_len; /* size of static rules */ 3293 if (ipfw_dyn_v) /* add size of dyn.rules */ 3294 size += (dyn_count * sizeof(ipfw_dyn_rule)); 3295 3296 /* 3297 * XXX todo: if the user passes a short length just to know 3298 * how much room is needed, do not bother filling up the 3299 * buffer, just jump to the sooptcopyout. 3300 */ 3301 buf = malloc(size, M_TEMP, M_WAITOK); 3302 error = sooptcopyout(sopt, buf, 3303 ipfw_getrules(&layer3_chain, buf, size)); 3304 free(buf, M_TEMP); 3305 break; 3306 3307 case IP_FW_FLUSH: 3308 /* 3309 * Normally we cannot release the lock on each iteration. 3310 * We could do it here only because we start from the head all 3311 * the times so there is no risk of missing some entries. 3312 * On the other hand, the risk is that we end up with 3313 * a very inconsistent ruleset, so better keep the lock 3314 * around the whole cycle. 3315 * 3316 * XXX this code can be improved by resetting the head of 3317 * the list to point to the default rule, and then freeing 3318 * the old list without the need for a lock. 3319 */ 3320 3321 IPFW_WLOCK(&layer3_chain); 3322 layer3_chain.reap = NULL; 3323 free_chain(&layer3_chain, 0 /* keep default rule */); 3324 rule = layer3_chain.reap, layer3_chain.reap = NULL; 3325 IPFW_WUNLOCK(&layer3_chain); 3326 if (layer3_chain.reap != NULL) 3327 reap_rules(rule); 3328 break; 3329 3330 case IP_FW_ADD: 3331 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 3332 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 3333 sizeof(struct ip_fw) ); 3334 if (error == 0) 3335 error = check_ipfw_struct(rule, sopt->sopt_valsize); 3336 if (error == 0) { 3337 error = add_rule(&layer3_chain, rule); 3338 size = RULESIZE(rule); 3339 if (!error && sopt->sopt_dir == SOPT_GET) 3340 error = sooptcopyout(sopt, rule, size); 3341 } 3342 free(rule, M_TEMP); 3343 break; 3344 3345 case IP_FW_DEL: 3346 /* 3347 * IP_FW_DEL is used for deleting single rules or sets, 3348 * and (ab)used to atomically manipulate sets. Argument size 3349 * is used to distinguish between the two: 3350 * sizeof(u_int32_t) 3351 * delete single rule or set of rules, 3352 * or reassign rules (or sets) to a different set. 3353 * 2*sizeof(u_int32_t) 3354 * atomic disable/enable sets. 3355 * first u_int32_t contains sets to be disabled, 3356 * second u_int32_t contains sets to be enabled. 3357 */ 3358 error = sooptcopyin(sopt, rulenum, 3359 2*sizeof(u_int32_t), sizeof(u_int32_t)); 3360 if (error) 3361 break; 3362 size = sopt->sopt_valsize; 3363 if (size == sizeof(u_int32_t)) /* delete or reassign */ 3364 error = del_entry(&layer3_chain, rulenum[0]); 3365 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 3366 set_disable = 3367 (set_disable | rulenum[0]) & ~rulenum[1] & 3368 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 3369 else 3370 error = EINVAL; 3371 break; 3372 3373 case IP_FW_ZERO: 3374 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 3375 rule_num = 0; 3376 if (sopt->sopt_val != 0) { 3377 error = sooptcopyin(sopt, &rule_num, 3378 sizeof(int), sizeof(int)); 3379 if (error) 3380 break; 3381 } 3382 error = zero_entry(&layer3_chain, rule_num, 3383 sopt->sopt_name == IP_FW_RESETLOG); 3384 break; 3385 3386 case IP_FW_TABLE_ADD: 3387 { 3388 ipfw_table_entry ent; 3389 3390 error = sooptcopyin(sopt, &ent, 3391 sizeof(ent), sizeof(ent)); 3392 if (error) 3393 break; 3394 error = add_table_entry(ent.tbl, ent.addr, 3395 ent.masklen, ent.value); 3396 } 3397 break; 3398 3399 case IP_FW_TABLE_DEL: 3400 { 3401 ipfw_table_entry ent; 3402 3403 error = sooptcopyin(sopt, &ent, 3404 sizeof(ent), sizeof(ent)); 3405 if (error) 3406 break; 3407 error = del_table_entry(ent.tbl, ent.addr, ent.masklen); 3408 } 3409 break; 3410 3411 case IP_FW_TABLE_FLUSH: 3412 { 3413 u_int16_t tbl; 3414 3415 error = sooptcopyin(sopt, &tbl, 3416 sizeof(tbl), sizeof(tbl)); 3417 if (error) 3418 break; 3419 error = flush_table(tbl); 3420 } 3421 break; 3422 3423 case IP_FW_TABLE_GETSIZE: 3424 { 3425 u_int32_t tbl, cnt; 3426 3427 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 3428 sizeof(tbl)))) 3429 break; 3430 if ((error = count_table(tbl, &cnt))) 3431 break; 3432 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 3433 } 3434 break; 3435 3436 case IP_FW_TABLE_LIST: 3437 { 3438 ipfw_table *tbl; 3439 3440 if (sopt->sopt_valsize < sizeof(*tbl)) { 3441 error = EINVAL; 3442 break; 3443 } 3444 size = sopt->sopt_valsize; 3445 tbl = malloc(size, M_TEMP, M_WAITOK); 3446 if (tbl == NULL) { 3447 error = ENOMEM; 3448 break; 3449 } 3450 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 3451 if (error) { 3452 free(tbl, M_TEMP); 3453 break; 3454 } 3455 tbl->size = (size - sizeof(*tbl)) / 3456 sizeof(ipfw_table_entry); 3457 error = dump_table(tbl); 3458 if (error) { 3459 free(tbl, M_TEMP); 3460 break; 3461 } 3462 error = sooptcopyout(sopt, tbl, size); 3463 free(tbl, M_TEMP); 3464 } 3465 break; 3466 3467 default: 3468 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 3469 error = EINVAL; 3470 } 3471 3472 return (error); 3473#undef RULE_MAXSIZE 3474} 3475 3476/** 3477 * dummynet needs a reference to the default rule, because rules can be 3478 * deleted while packets hold a reference to them. When this happens, 3479 * dummynet changes the reference to the default rule (it could well be a 3480 * NULL pointer, but this way we do not need to check for the special 3481 * case, plus here he have info on the default behaviour). 3482 */ 3483struct ip_fw *ip_fw_default_rule; 3484 3485/* 3486 * This procedure is only used to handle keepalives. It is invoked 3487 * every dyn_keepalive_period 3488 */ 3489static void 3490ipfw_tick(void * __unused unused) 3491{ 3492 int i; 3493 ipfw_dyn_rule *q; 3494 3495 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 3496 goto done; 3497 3498 IPFW_DYN_LOCK(); 3499 for (i = 0 ; i < curr_dyn_buckets ; i++) { 3500 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 3501 if (q->dyn_type == O_LIMIT_PARENT) 3502 continue; 3503 if (q->id.proto != IPPROTO_TCP) 3504 continue; 3505 if ( (q->state & BOTH_SYN) != BOTH_SYN) 3506 continue; 3507 if (TIME_LEQ( time_second+dyn_keepalive_interval, 3508 q->expire)) 3509 continue; /* too early */ 3510 if (TIME_LEQ(q->expire, time_second)) 3511 continue; /* too late, rule expired */ 3512 3513 send_pkt(&(q->id), q->ack_rev - 1, q->ack_fwd, TH_SYN); 3514 send_pkt(&(q->id), q->ack_fwd - 1, q->ack_rev, 0); 3515 } 3516 } 3517 IPFW_DYN_UNLOCK(); 3518done: 3519 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 3520} 3521 3522int 3523ipfw_init(void) 3524{ 3525 struct ip_fw default_rule; 3526 int error; 3527 3528 layer3_chain.rules = NULL; 3529 layer3_chain.want_write = 0; 3530 layer3_chain.busy_count = 0; 3531 cv_init(&layer3_chain.cv, "Condition variable for IPFW rw locks"); 3532 IPFW_LOCK_INIT(&layer3_chain); 3533 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule zone", 3534 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 3535 UMA_ALIGN_PTR, 0); 3536 IPFW_DYN_LOCK_INIT(); 3537 callout_init(&ipfw_timeout, debug_mpsafenet ? CALLOUT_MPSAFE : 0); 3538 3539 bzero(&default_rule, sizeof default_rule); 3540 3541 default_rule.act_ofs = 0; 3542 default_rule.rulenum = IPFW_DEFAULT_RULE; 3543 default_rule.cmd_len = 1; 3544 default_rule.set = RESVD_SET; 3545 3546 default_rule.cmd[0].len = 1; 3547 default_rule.cmd[0].opcode = 3548#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 3549 1 ? O_ACCEPT : 3550#endif 3551 O_DENY; 3552 3553 error = add_rule(&layer3_chain, &default_rule); 3554 if (error != 0) { 3555 printf("ipfw2: error %u initializing default rule " 3556 "(support disabled)\n", error); 3557 IPFW_DYN_LOCK_DESTROY(); 3558 IPFW_LOCK_DESTROY(&layer3_chain); 3559 return (error); 3560 } 3561 3562 ip_fw_default_rule = layer3_chain.rules; 3563 printf("ipfw2 initialized, divert %s, " 3564 "rule-based forwarding " 3565#ifdef IPFIREWALL_FORWARD 3566 "enabled, " 3567#else 3568 "disabled, " 3569#endif 3570 "default to %s, logging ", 3571#ifdef IPDIVERT 3572 "enabled", 3573#else 3574 "loadable", 3575#endif 3576 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 3577 3578#ifdef IPFIREWALL_VERBOSE 3579 fw_verbose = 1; 3580#endif 3581#ifdef IPFIREWALL_VERBOSE_LIMIT 3582 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 3583#endif 3584 if (fw_verbose == 0) 3585 printf("disabled\n"); 3586 else if (verbose_limit == 0) 3587 printf("unlimited\n"); 3588 else 3589 printf("limited to %d packets/entry by default\n", 3590 verbose_limit); 3591 3592 init_tables(); 3593 ip_fw_ctl_ptr = ipfw_ctl; 3594 ip_fw_chk_ptr = ipfw_chk; 3595 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 3596 3597 return (0); 3598} 3599 3600void 3601ipfw_destroy(void) 3602{ 3603 struct ip_fw *reap; 3604 3605 ip_fw_chk_ptr = NULL; 3606 ip_fw_ctl_ptr = NULL; 3607 callout_drain(&ipfw_timeout); 3608 IPFW_WLOCK(&layer3_chain); 3609 layer3_chain.reap = NULL; 3610 free_chain(&layer3_chain, 1 /* kill default rule */); 3611 reap = layer3_chain.reap, layer3_chain.reap = NULL; 3612 IPFW_WUNLOCK(&layer3_chain); 3613 if (reap != NULL) 3614 reap_rules(reap); 3615 flush_tables(); 3616 IPFW_DYN_LOCK_DESTROY(); 3617 uma_zdestroy(ipfw_dyn_rule_zone); 3618 IPFW_LOCK_DESTROY(&layer3_chain); 3619 printf("IP firewall unloaded\n"); 3620} 3621 3622#endif /* IPFW2 */ 3623