ip_fw2.c revision 130311
162587Sitojun/* 262587Sitojun * Copyright (c) 2002 Luigi Rizzo, Universita` di Pisa 362587Sitojun * 453541Sshin * Redistribution and use in source and binary forms, with or without 553541Sshin * modification, are permitted provided that the following conditions 653541Sshin * are met: 753541Sshin * 1. Redistributions of source code must retain the above copyright 853541Sshin * notice, this list of conditions and the following disclaimer. 953541Sshin * 2. Redistributions in binary form must reproduce the above copyright 1053541Sshin * notice, this list of conditions and the following disclaimer in the 1153541Sshin * documentation and/or other materials provided with the distribution. 1253541Sshin * 1353541Sshin * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 1453541Sshin * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 1553541Sshin * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 1653541Sshin * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 1753541Sshin * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 1853541Sshin * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 1953541Sshin * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 2053541Sshin * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 2153541Sshin * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 2253541Sshin * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 2353541Sshin * SUCH DAMAGE. 2453541Sshin * 2553541Sshin * $FreeBSD: head/sys/netinet/ip_fw2.c 130311 2004-06-10 20:20:37Z ru $ 2653541Sshin */ 2753541Sshin 2853541Sshin#define DEB(x) 2953541Sshin#define DDB(x) x 3053541Sshin 3153541Sshin/* 3253541Sshin * Implement IP packet firewall (new version) 3353541Sshin */ 3453541Sshin 3553541Sshin#if !defined(KLD_MODULE) 3653541Sshin#include "opt_ipfw.h" 3753541Sshin#include "opt_ipdn.h" 3853541Sshin#include "opt_ipdivert.h" 3953541Sshin#include "opt_inet.h" 4053541Sshin#include "opt_ipsec.h" 4153541Sshin#ifndef INET 4253541Sshin#error IPFIREWALL requires INET. 4353541Sshin#endif /* INET */ 4453541Sshin#endif 4553541Sshin 4653541Sshin#define IPFW2 1 4753541Sshin#if IPFW2 4853541Sshin#include <sys/param.h> 4953541Sshin#include <sys/systm.h> 5062587Sitojun#include <sys/malloc.h> 5153541Sshin#include <sys/mbuf.h> 5262587Sitojun#include <sys/kernel.h> 5353541Sshin#include <sys/module.h> 5453541Sshin#include <sys/proc.h> 5553541Sshin#include <sys/socket.h> 5653541Sshin#include <sys/socketvar.h> 5753541Sshin#include <sys/sysctl.h> 5853541Sshin#include <sys/syslog.h> 5953541Sshin#include <sys/ucred.h> 6053541Sshin#include <net/if.h> 6162587Sitojun#include <net/radix.h> 6253541Sshin#include <net/route.h> 6362587Sitojun#include <netinet/in.h> 6462587Sitojun#include <netinet/in_systm.h> 6562587Sitojun#include <netinet/in_var.h> 6662587Sitojun#include <netinet/in_pcb.h> 6762587Sitojun#include <netinet/ip.h> 6853541Sshin#include <netinet/ip_var.h> 6962587Sitojun#include <netinet/ip_icmp.h> 7062587Sitojun#include <netinet/ip_fw.h> 7162587Sitojun#include <netinet/ip_divert.h> 7253541Sshin#include <netinet/ip_dummynet.h> 7362587Sitojun#include <netinet/tcp.h> 7469774Sphk#include <netinet/tcp_timer.h> 7562587Sitojun#include <netinet/tcp_var.h> 7653541Sshin#include <netinet/tcpip.h> 7753541Sshin#include <netinet/udp.h> 7853541Sshin#include <netinet/udp_var.h> 7953541Sshin 8053541Sshin#ifdef IPSEC 8153541Sshin#include <netinet6/ipsec.h> 8253541Sshin#endif 8353541Sshin 8453541Sshin#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 8553541Sshin 8653541Sshin#include <machine/in_cksum.h> /* XXX for in_cksum */ 8753541Sshin 8853541Sshin/* 8962587Sitojun * This is used to avoid that a firewall-generated packet 9053541Sshin * loops forever through the firewall. Note that it must 9153541Sshin * be a flag that is unused by other protocols that might 9253541Sshin * be called from ip_output (e.g. IPsec) and it must be 9353541Sshin * listed in M_COPYFLAGS in mbuf.h so that if the mbuf chain 9462587Sitojun * is altered on the way through ip_output it is not lost. 9562587Sitojun * It might be better to add an m_tag since the this happens 9662587Sitojun * infrequently. 9762587Sitojun */ 9862587Sitojun#define M_SKIP_FIREWALL M_PROTO6 9962587Sitojun 10062587Sitojun/* 10162587Sitojun * set_disable contains one bit per set value (0..31). 10262587Sitojun * If the bit is set, all rules with the corresponding set 10362587Sitojun * are disabled. Set RESVD_SET(31) is reserved for the default rule 10462587Sitojun * and rules that are not deleted by the flush command, 10562587Sitojun * and CANNOT be disabled. 10662587Sitojun * Rules in set RESVD_SET can only be deleted explicitly. 10762587Sitojun */ 10862587Sitojunstatic u_int32_t set_disable; 10962587Sitojun 11062587Sitojunstatic int fw_verbose; 11162587Sitojunstatic int verbose_limit; 11262587Sitojun 11362587Sitojunstatic struct callout ipfw_timeout; 11462587Sitojun#define IPFW_DEFAULT_RULE 65535 11562587Sitojun 11662587Sitojunstruct ip_fw_chain { 11762587Sitojun struct ip_fw *rules; /* list of rules */ 11862587Sitojun struct ip_fw *reap; /* list of rules to reap */ 11962587Sitojun struct mtx mtx; /* lock guarding rule list */ 12062587Sitojun}; 12162587Sitojun#define IPFW_LOCK_INIT(_chain) \ 12262587Sitojun mtx_init(&(_chain)->mtx, "IPFW static rules", NULL, \ 12353541Sshin MTX_DEF | MTX_RECURSE) 12453541Sshin#define IPFW_LOCK_DESTROY(_chain) mtx_destroy(&(_chain)->mtx) 12553541Sshin#define IPFW_LOCK(_chain) mtx_lock(&(_chain)->mtx) 12653541Sshin#define IPFW_UNLOCK(_chain) mtx_unlock(&(_chain)->mtx) 12753541Sshin#define IPFW_LOCK_ASSERT(_chain) mtx_assert(&(_chain)->mtx, MA_OWNED) 12853541Sshin 12953541Sshin/* 13053541Sshin * list of rules for layer 3 13153541Sshin */ 13253541Sshinstatic struct ip_fw_chain layer3_chain; 13353541Sshin 13462587SitojunMALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 13553541SshinMALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 13653541Sshin 13762587Sitojunstruct table_entry { 13853541Sshin struct radix_node rn[2]; 13953541Sshin struct sockaddr_in addr, mask; 14053541Sshin u_int32_t value; 14153541Sshin}; 14253541Sshin 14353541Sshin#define IPFW_TABLES_MAX 128 14462587Sitojunstatic struct { 14562587Sitojun struct radix_node_head *rnh; 14653541Sshin int modified; 14753541Sshin} ipfw_tables[IPFW_TABLES_MAX]; 14862587Sitojun 14962587Sitojunstatic int fw_debug = 1; 15062587Sitojunstatic int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 15162587Sitojun 15262587Sitojun#ifdef SYSCTL_NODE 15353541SshinSYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 15453541SshinSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, enable, 15553541Sshin CTLFLAG_RW | CTLFLAG_SECURE3, 15653541Sshin &fw_enable, 0, "Enable ipfw"); 15753541SshinSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 15853541Sshin &autoinc_step, 0, "Rule number autincrement step"); 15953541SshinSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 16053541Sshin CTLFLAG_RW | CTLFLAG_SECURE3, 16153541Sshin &fw_one_pass, 0, 16253541Sshin "Only do a single pass through ipfw when using dummynet(4)"); 16353541SshinSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 16453541Sshin &fw_debug, 0, "Enable printing of debug ip_fw statements"); 16553541SshinSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 16653541Sshin CTLFLAG_RW | CTLFLAG_SECURE3, 16753541Sshin &fw_verbose, 0, "Log matches to ipfw rules"); 16853541SshinSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 16953541Sshin &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 17054350Sshin 17153541Sshin/* 17253541Sshin * Description of dynamic rules. 17353541Sshin * 17453541Sshin * Dynamic rules are stored in lists accessed through a hash table 17553541Sshin * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 17653541Sshin * be modified through the sysctl variable dyn_buckets which is 17753541Sshin * updated when the table becomes empty. 17853541Sshin * 17953541Sshin * XXX currently there is only one list, ipfw_dyn. 18053541Sshin * 18153541Sshin * When a packet is received, its address fields are first masked 18253541Sshin * with the mask defined for the rule, then hashed, then matched 18353541Sshin * against the entries in the corresponding list. 18453541Sshin * Dynamic rules can be used for different purposes: 18553541Sshin * + stateful rules; 18653541Sshin * + enforcing limits on the number of sessions; 18753541Sshin * + in-kernel NAT (not implemented yet) 18853541Sshin * 18953541Sshin * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 19053541Sshin * measured in seconds and depending on the flags. 19153541Sshin * 19253541Sshin * The total number of dynamic rules is stored in dyn_count. 19353541Sshin * The max number of dynamic rules is dyn_max. When we reach 19453541Sshin * the maximum number of rules we do not create anymore. This is 19553541Sshin * done to avoid consuming too much memory, but also too much 19662587Sitojun * time when searching on each packet (ideally, we should try instead 19753541Sshin * to put a limit on the length of the list on each bucket...). 19853541Sshin * 19953541Sshin * Each dynamic rule holds a pointer to the parent ipfw rule so 20053541Sshin * we know what action to perform. Dynamic rules are removed when 20153541Sshin * the parent rule is deleted. XXX we should make them survive. 20253541Sshin * 20353541Sshin * There are some limitations with dynamic rules -- we do not 20462587Sitojun * obey the 'randomized match', and we do not do multiple 20553541Sshin * passes through the firewall. XXX check the latter!!! 20653541Sshin */ 20753541Sshinstatic ipfw_dyn_rule **ipfw_dyn_v = NULL; 20853541Sshinstatic u_int32_t dyn_buckets = 256; /* must be power of 2 */ 20953541Sshinstatic u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 21053541Sshin 21153541Sshinstatic struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 21253541Sshin#define IPFW_DYN_LOCK_INIT() \ 21353541Sshin mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 21453541Sshin#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 21553541Sshin#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 21653541Sshin#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 21753541Sshin#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 21853541Sshin 21953541Sshin/* 22053541Sshin * Timeouts for various events in handing dynamic rules. 22153541Sshin */ 22253541Sshinstatic u_int32_t dyn_ack_lifetime = 300; 22353541Sshinstatic u_int32_t dyn_syn_lifetime = 20; 22453541Sshinstatic u_int32_t dyn_fin_lifetime = 1; 22553541Sshinstatic u_int32_t dyn_rst_lifetime = 1; 22653541Sshinstatic u_int32_t dyn_udp_lifetime = 10; 22753541Sshinstatic u_int32_t dyn_short_lifetime = 5; 22853541Sshin 22953541Sshin/* 23053541Sshin * Keepalives are sent if dyn_keepalive is set. They are sent every 23153541Sshin * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 23253541Sshin * seconds of lifetime of a rule. 23353541Sshin * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 23453541Sshin * than dyn_keepalive_period. 23562587Sitojun */ 23653541Sshin 23753541Sshinstatic u_int32_t dyn_keepalive_interval = 20; 23853541Sshinstatic u_int32_t dyn_keepalive_period = 5; 23962587Sitojunstatic u_int32_t dyn_keepalive = 1; /* do send keepalives */ 24053541Sshin 24162587Sitojunstatic u_int32_t static_count; /* # of static rules */ 24262587Sitojunstatic u_int32_t static_len; /* size in bytes of static rules */ 24362587Sitojunstatic u_int32_t dyn_count; /* # of dynamic rules */ 24453541Sshinstatic u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 24553541Sshin 24653541SshinSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 24753541Sshin &dyn_buckets, 0, "Number of dyn. buckets"); 24853541SshinSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 24953541Sshin &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 25053541SshinSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 25153541Sshin &dyn_count, 0, "Number of dyn. rules"); 25253541SshinSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 25353541Sshin &dyn_max, 0, "Max number of dyn. rules"); 25453541SshinSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 25553541Sshin &static_count, 0, "Number of static rules"); 25653541SshinSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 25753541Sshin &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 25853541SshinSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 25953541Sshin &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 26053541SshinSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 26153541Sshin &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 26253541SshinSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 26353541Sshin &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 26453541SshinSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 26553541Sshin &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 26653541SshinSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 26753541Sshin &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 26862587SitojunSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 26953541Sshin &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 27053541Sshin 27153541Sshin#endif /* SYSCTL_NODE */ 27253541Sshin 27362587Sitojun 27462587Sitojunstatic ip_fw_chk_t ipfw_chk; 27553541Sshin 27653541Sshinip_dn_ruledel_t *ip_dn_ruledel_ptr = NULL; /* hook into dummynet */ 27753541Sshin 27853541Sshin/* 27953541Sshin * This macro maps an ip pointer into a layer3 header pointer of type T 28062587Sitojun */ 28162587Sitojun#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 28253541Sshin 28353541Sshinstatic __inline int 28453541Sshinicmptype_match(struct ip *ip, ipfw_insn_u32 *cmd) 28553541Sshin{ 28653541Sshin int type = L3HDR(struct icmp,ip)->icmp_type; 28753541Sshin 28853541Sshin return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 28953541Sshin} 29053541Sshin 29153541Sshin#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 29253541Sshin (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 29353541Sshin 29453541Sshinstatic int 29553541Sshinis_icmp_query(struct ip *ip) 29653541Sshin{ 29753541Sshin int type = L3HDR(struct icmp, ip)->icmp_type; 29853541Sshin return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 29953541Sshin} 30053541Sshin#undef TT 30162587Sitojun 30253541Sshin/* 30353541Sshin * The following checks use two arrays of 8 or 16 bits to store the 30453541Sshin * bits that we want set or clear, respectively. They are in the 30553541Sshin * low and high half of cmd->arg1 or cmd->d[0]. 30653541Sshin * 30753541Sshin * We scan options and store the bits we find set. We succeed if 30853541Sshin * 30953541Sshin * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 31053541Sshin * 31153541Sshin * The code is sometimes optimized not to store additional variables. 31253541Sshin */ 31353541Sshin 31453541Sshinstatic int 31562587Sitojunflags_match(ipfw_insn *cmd, u_int8_t bits) 31662587Sitojun{ 31753541Sshin u_char want_clear; 31853541Sshin bits = ~bits; 31953541Sshin 32053541Sshin if ( ((cmd->arg1 & 0xff) & bits) != 0) 32162587Sitojun return 0; /* some bits we want set were clear */ 32262587Sitojun want_clear = (cmd->arg1 >> 8) & 0xff; 32362587Sitojun if ( (want_clear & bits) != want_clear) 32462587Sitojun return 0; /* some bits we want clear were set */ 32562587Sitojun return 1; 32662587Sitojun} 32762587Sitojun 32862587Sitojunstatic int 32962587Sitojunipopts_match(struct ip *ip, ipfw_insn *cmd) 33053541Sshin{ 33153541Sshin int optlen, bits = 0; 33253541Sshin u_char *cp = (u_char *)(ip + 1); 33353541Sshin int x = (ip->ip_hl << 2) - sizeof (struct ip); 33453541Sshin 33553541Sshin for (; x > 0; x -= optlen, cp += optlen) { 33653541Sshin int opt = cp[IPOPT_OPTVAL]; 33753541Sshin 33853541Sshin if (opt == IPOPT_EOL) 33953541Sshin break; 34053541Sshin if (opt == IPOPT_NOP) 34153541Sshin optlen = 1; 34253541Sshin else { 34353541Sshin optlen = cp[IPOPT_OLEN]; 34453541Sshin if (optlen <= 0 || optlen > x) 34553541Sshin return 0; /* invalid or truncated */ 34653541Sshin } 34753541Sshin switch (opt) { 34853541Sshin 34962587Sitojun default: 35053541Sshin break; 35162587Sitojun 35262587Sitojun case IPOPT_LSRR: 35362587Sitojun bits |= IP_FW_IPOPT_LSRR; 35462587Sitojun break; 35562587Sitojun 35662587Sitojun case IPOPT_SSRR: 35762587Sitojun bits |= IP_FW_IPOPT_SSRR; 35862587Sitojun break; 35962587Sitojun 36062587Sitojun case IPOPT_RR: 36162587Sitojun bits |= IP_FW_IPOPT_RR; 36262587Sitojun break; 36362587Sitojun 36462587Sitojun case IPOPT_TS: 36562587Sitojun bits |= IP_FW_IPOPT_TS; 36662587Sitojun break; 36762587Sitojun } 36862587Sitojun } 36962587Sitojun return (flags_match(cmd, bits)); 37062587Sitojun} 37162587Sitojun 37262587Sitojunstatic int 37362587Sitojuntcpopts_match(struct ip *ip, ipfw_insn *cmd) 37462587Sitojun{ 37562587Sitojun int optlen, bits = 0; 37662587Sitojun struct tcphdr *tcp = L3HDR(struct tcphdr,ip); 37762587Sitojun u_char *cp = (u_char *)(tcp + 1); 37862587Sitojun int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 37962587Sitojun 38062587Sitojun for (; x > 0; x -= optlen, cp += optlen) { 38162587Sitojun int opt = cp[0]; 38262587Sitojun if (opt == TCPOPT_EOL) 38362587Sitojun break; 38462587Sitojun if (opt == TCPOPT_NOP) 38562587Sitojun optlen = 1; 38653541Sshin else { 38753541Sshin optlen = cp[1]; 38853541Sshin if (optlen <= 0) 38953541Sshin break; 39053541Sshin } 39153541Sshin 39253541Sshin switch (opt) { 39353541Sshin 39476899Ssumikawa default: 39553541Sshin break; 39653541Sshin 39753541Sshin case TCPOPT_MAXSEG: 39876899Ssumikawa bits |= IP_FW_TCPOPT_MSS; 39976899Ssumikawa break; 40053541Sshin 40153541Sshin case TCPOPT_WINDOW: 40253541Sshin bits |= IP_FW_TCPOPT_WINDOW; 40353541Sshin break; 40453541Sshin 40553541Sshin case TCPOPT_SACK_PERMITTED: 40676899Ssumikawa case TCPOPT_SACK: 40753541Sshin bits |= IP_FW_TCPOPT_SACK; 40853541Sshin break; 40953541Sshin 41076899Ssumikawa case TCPOPT_TIMESTAMP: 41176899Ssumikawa bits |= IP_FW_TCPOPT_TS; 41253541Sshin break; 41353541Sshin 41453541Sshin case TCPOPT_CC: 41562587Sitojun case TCPOPT_CCNEW: 41653541Sshin case TCPOPT_CCECHO: 41753541Sshin bits |= IP_FW_TCPOPT_CC; 41853541Sshin break; 41953541Sshin } 42053541Sshin } 42153541Sshin return (flags_match(cmd, bits)); 42253541Sshin} 42353541Sshin 42453541Sshinstatic int 42553541Sshiniface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 42662587Sitojun{ 42762587Sitojun if (ifp == NULL) /* no iface with this packet, match fails */ 42862587Sitojun return 0; 42962587Sitojun /* Check by name or by IP address */ 43062587Sitojun if (cmd->name[0] != '\0') { /* match by name */ 43162587Sitojun /* Check name */ 43253541Sshin if (cmd->p.glob) { 43353541Sshin if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 43453541Sshin return(1); 43553541Sshin } else { 43653541Sshin if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 43753541Sshin return(1); 43853541Sshin } 43953541Sshin } else { 44053541Sshin struct ifaddr *ia; 44153541Sshin 44253541Sshin /* XXX lock? */ 44353541Sshin TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 44453541Sshin if (ia->ifa_addr == NULL) 44553541Sshin continue; 44653541Sshin if (ia->ifa_addr->sa_family != AF_INET) 44753541Sshin continue; 44853541Sshin if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 44953541Sshin (ia->ifa_addr))->sin_addr.s_addr) 45053541Sshin return(1); /* match */ 45153541Sshin } 45262587Sitojun } 45353541Sshin return(0); /* no match, fail ... */ 45462587Sitojun} 45562587Sitojun 45653541Sshin/* 45753541Sshin * The verify_path function checks if a route to the src exists and 45853541Sshin * if it is reachable via ifp (when provided). 45962587Sitojun * 46062587Sitojun * The 'verrevpath' option checks that the interface that an IP packet 46162587Sitojun * arrives on is the same interface that traffic destined for the 46253541Sshin * packet's source address would be routed out of. The 'versrcreach' 46353541Sshin * option just checks that the source address is reachable via any route 46453541Sshin * (except default) in the routing table. These two are a measure to block 46553541Sshin * forged packets. This is also commonly known as "anti-spoofing" or Unicast 46662587Sitojun * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 46762587Sitojun * is purposely reminiscent of the Cisco IOS command, 46853541Sshin * 46953541Sshin * ip verify unicast reverse-path 47053541Sshin * ip verify unicast source reachable-via any 47153541Sshin * 47262587Sitojun * which implements the same functionality. But note that syntax is 47362587Sitojun * misleading. The check may be performed on all IP packets whether unicast, 47462587Sitojun * multicast, or broadcast. 47553541Sshin */ 47653541Sshinstatic int 47753541Sshinverify_path(struct in_addr src, struct ifnet *ifp) 47853541Sshin{ 47962587Sitojun struct route ro; 48053541Sshin struct sockaddr_in *dst; 48153541Sshin 48262587Sitojun bzero(&ro, sizeof(ro)); 48362587Sitojun 48462587Sitojun dst = (struct sockaddr_in *)&(ro.ro_dst); 48562587Sitojun dst->sin_family = AF_INET; 48662587Sitojun dst->sin_len = sizeof(*dst); 48762587Sitojun dst->sin_addr = src; 48862587Sitojun rtalloc_ign(&ro, RTF_CLONING); 48962587Sitojun 49062587Sitojun if (ro.ro_rt == NULL) 49162587Sitojun return 0; 49262587Sitojun 49362587Sitojun /* if ifp is provided, check for equality with rtentry */ 49453541Sshin if (ifp != NULL && ro.ro_rt->rt_ifp != ifp) { 49553541Sshin RTFREE(ro.ro_rt); 49653541Sshin return 0; 49753541Sshin } 49853541Sshin 49953541Sshin /* if no ifp provided, check if rtentry is not default route */ 50053541Sshin if (ifp == NULL && 50153541Sshin satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 50253541Sshin RTFREE(ro.ro_rt); 50353541Sshin return 0; 50453541Sshin } 50553541Sshin 50653541Sshin /* found valid route */ 50753541Sshin RTFREE(ro.ro_rt); 50853541Sshin return 1; 50953541Sshin} 51053541Sshin 51153541Sshin 51253541Sshinstatic u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 51353541Sshin 51453541Sshin#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 51553541Sshin#define SNP(buf) buf, sizeof(buf) 51653541Sshin 51753541Sshin/* 51853541Sshin * We enter here when we have a rule with O_LOG. 51953541Sshin * XXX this function alone takes about 2Kbytes of code! 52053541Sshin */ 52153541Sshinstatic void 52253541Sshinipfw_log(struct ip_fw *f, u_int hlen, struct ether_header *eh, 52353541Sshin struct mbuf *m, struct ifnet *oif) 52453541Sshin{ 52553541Sshin char *action; 52653541Sshin int limit_reached = 0; 52753541Sshin char action2[40], proto[48], fragment[28]; 52853541Sshin 52953541Sshin fragment[0] = '\0'; 53053541Sshin proto[0] = '\0'; 53153541Sshin 53253541Sshin if (f == NULL) { /* bogus pkt */ 53353541Sshin if (verbose_limit != 0 && norule_counter >= verbose_limit) 53453541Sshin return; 53553541Sshin norule_counter++; 53653541Sshin if (norule_counter == verbose_limit) 53753541Sshin limit_reached = verbose_limit; 53853541Sshin action = "Refuse"; 53953541Sshin } else { /* O_LOG is the first action, find the real one */ 54053541Sshin ipfw_insn *cmd = ACTION_PTR(f); 54153541Sshin ipfw_insn_log *l = (ipfw_insn_log *)cmd; 54253541Sshin 54353541Sshin if (l->max_log != 0 && l->log_left == 0) 54453541Sshin return; 54553541Sshin l->log_left--; 54653541Sshin if (l->log_left == 0) 54753541Sshin limit_reached = l->max_log; 54853541Sshin cmd += F_LEN(cmd); /* point to first action */ 54953541Sshin if (cmd->opcode == O_PROB) 55053541Sshin cmd += F_LEN(cmd); 55153541Sshin 55253541Sshin action = action2; 55353541Sshin switch (cmd->opcode) { 55453541Sshin case O_DENY: 55553541Sshin action = "Deny"; 55653541Sshin break; 55753541Sshin 55853541Sshin case O_REJECT: 55953541Sshin if (cmd->arg1==ICMP_REJECT_RST) 56053541Sshin action = "Reset"; 56153541Sshin else if (cmd->arg1==ICMP_UNREACH_HOST) 56253541Sshin action = "Reject"; 56353541Sshin else 56453541Sshin snprintf(SNPARGS(action2, 0), "Unreach %d", 56553541Sshin cmd->arg1); 56653541Sshin break; 56753541Sshin 56862587Sitojun case O_ACCEPT: 56953541Sshin action = "Accept"; 57062587Sitojun break; 57153541Sshin case O_COUNT: 57253541Sshin action = "Count"; 57353541Sshin break; 57453541Sshin case O_DIVERT: 57553541Sshin snprintf(SNPARGS(action2, 0), "Divert %d", 57653541Sshin cmd->arg1); 57753541Sshin break; 57853541Sshin case O_TEE: 57953541Sshin snprintf(SNPARGS(action2, 0), "Tee %d", 58053541Sshin cmd->arg1); 58153541Sshin break; 58253541Sshin case O_SKIPTO: 58353541Sshin snprintf(SNPARGS(action2, 0), "SkipTo %d", 58453541Sshin cmd->arg1); 58553541Sshin break; 58653541Sshin case O_PIPE: 58753541Sshin snprintf(SNPARGS(action2, 0), "Pipe %d", 58853541Sshin cmd->arg1); 58953541Sshin break; 59053541Sshin case O_QUEUE: 59153541Sshin snprintf(SNPARGS(action2, 0), "Queue %d", 59253541Sshin cmd->arg1); 59353541Sshin break; 59453541Sshin case O_FORWARD_IP: { 59553541Sshin ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 59653541Sshin int len; 59753541Sshin 59853541Sshin len = snprintf(SNPARGS(action2, 0), "Forward to %s", 59953541Sshin inet_ntoa(sa->sa.sin_addr)); 60053541Sshin if (sa->sa.sin_port) 60153541Sshin snprintf(SNPARGS(action2, len), ":%d", 60253541Sshin sa->sa.sin_port); 60353541Sshin } 60453541Sshin break; 60553541Sshin default: 60653541Sshin action = "UNKNOWN"; 60753541Sshin break; 60853541Sshin } 60953541Sshin } 61053541Sshin 61153541Sshin if (hlen == 0) { /* non-ip */ 61253541Sshin snprintf(SNPARGS(proto, 0), "MAC"); 61353541Sshin } else { 61453541Sshin struct ip *ip = mtod(m, struct ip *); 61553541Sshin /* these three are all aliases to the same thing */ 61653541Sshin struct icmp *const icmp = L3HDR(struct icmp, ip); 61753541Sshin struct tcphdr *const tcp = (struct tcphdr *)icmp; 61853541Sshin struct udphdr *const udp = (struct udphdr *)icmp; 61953541Sshin 62053541Sshin int ip_off, offset, ip_len; 62153541Sshin 62253541Sshin int len; 62353541Sshin 62453541Sshin if (eh != NULL) { /* layer 2 packets are as on the wire */ 62553541Sshin ip_off = ntohs(ip->ip_off); 62653541Sshin ip_len = ntohs(ip->ip_len); 62753541Sshin } else { 62853541Sshin ip_off = ip->ip_off; 62953541Sshin ip_len = ip->ip_len; 63062587Sitojun } 63162587Sitojun offset = ip_off & IP_OFFMASK; 63262587Sitojun switch (ip->ip_p) { 63353541Sshin case IPPROTO_TCP: 63453541Sshin len = snprintf(SNPARGS(proto, 0), "TCP %s", 63553541Sshin inet_ntoa(ip->ip_src)); 63653541Sshin if (offset == 0) 63753541Sshin snprintf(SNPARGS(proto, len), ":%d %s:%d", 63853541Sshin ntohs(tcp->th_sport), 63953541Sshin inet_ntoa(ip->ip_dst), 64053541Sshin ntohs(tcp->th_dport)); 64153541Sshin else 64253541Sshin snprintf(SNPARGS(proto, len), " %s", 64353541Sshin inet_ntoa(ip->ip_dst)); 64453541Sshin break; 64553541Sshin 64653541Sshin case IPPROTO_UDP: 64753541Sshin len = snprintf(SNPARGS(proto, 0), "UDP %s", 64853541Sshin inet_ntoa(ip->ip_src)); 64953541Sshin if (offset == 0) 65053541Sshin snprintf(SNPARGS(proto, len), ":%d %s:%d", 65153541Sshin ntohs(udp->uh_sport), 65253541Sshin inet_ntoa(ip->ip_dst), 65353541Sshin ntohs(udp->uh_dport)); 65453541Sshin else 65553541Sshin snprintf(SNPARGS(proto, len), " %s", 65653541Sshin inet_ntoa(ip->ip_dst)); 65762587Sitojun break; 65862587Sitojun 65962587Sitojun case IPPROTO_ICMP: 66062587Sitojun if (offset == 0) 66162587Sitojun len = snprintf(SNPARGS(proto, 0), 66262587Sitojun "ICMP:%u.%u ", 66362587Sitojun icmp->icmp_type, icmp->icmp_code); 66462587Sitojun else 66562587Sitojun len = snprintf(SNPARGS(proto, 0), "ICMP "); 66662587Sitojun len += snprintf(SNPARGS(proto, len), "%s", 66762587Sitojun inet_ntoa(ip->ip_src)); 66862587Sitojun snprintf(SNPARGS(proto, len), " %s", 66962587Sitojun inet_ntoa(ip->ip_dst)); 67062587Sitojun break; 67162587Sitojun 67262587Sitojun default: 67362587Sitojun len = snprintf(SNPARGS(proto, 0), "P:%d %s", ip->ip_p, 67453541Sshin inet_ntoa(ip->ip_src)); 67553541Sshin snprintf(SNPARGS(proto, len), " %s", 67653541Sshin inet_ntoa(ip->ip_dst)); 67753541Sshin break; 67853541Sshin } 67953541Sshin 68053541Sshin if (ip_off & (IP_MF | IP_OFFMASK)) 68153541Sshin snprintf(SNPARGS(fragment, 0), " (frag %d:%d@%d%s)", 68253541Sshin ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 68353541Sshin offset << 3, 68453541Sshin (ip_off & IP_MF) ? "+" : ""); 68553541Sshin } 68653541Sshin if (oif || m->m_pkthdr.rcvif) 68753541Sshin log(LOG_SECURITY | LOG_INFO, 68853541Sshin "ipfw: %d %s %s %s via %s%s\n", 68953541Sshin f ? f->rulenum : -1, 69053541Sshin action, proto, oif ? "out" : "in", 691 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 692 fragment); 693 else 694 log(LOG_SECURITY | LOG_INFO, 695 "ipfw: %d %s %s [no if info]%s\n", 696 f ? f->rulenum : -1, 697 action, proto, fragment); 698 if (limit_reached) 699 log(LOG_SECURITY | LOG_NOTICE, 700 "ipfw: limit %d reached on entry %d\n", 701 limit_reached, f ? f->rulenum : -1); 702} 703 704/* 705 * IMPORTANT: the hash function for dynamic rules must be commutative 706 * in source and destination (ip,port), because rules are bidirectional 707 * and we want to find both in the same bucket. 708 */ 709static __inline int 710hash_packet(struct ipfw_flow_id *id) 711{ 712 u_int32_t i; 713 714 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 715 i &= (curr_dyn_buckets - 1); 716 return i; 717} 718 719/** 720 * unlink a dynamic rule from a chain. prev is a pointer to 721 * the previous one, q is a pointer to the rule to delete, 722 * head is a pointer to the head of the queue. 723 * Modifies q and potentially also head. 724 */ 725#define UNLINK_DYN_RULE(prev, head, q) { \ 726 ipfw_dyn_rule *old_q = q; \ 727 \ 728 /* remove a refcount to the parent */ \ 729 if (q->dyn_type == O_LIMIT) \ 730 q->parent->count--; \ 731 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 732 (q->id.src_ip), (q->id.src_port), \ 733 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 734 if (prev != NULL) \ 735 prev->next = q = q->next; \ 736 else \ 737 head = q = q->next; \ 738 dyn_count--; \ 739 free(old_q, M_IPFW); } 740 741#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 742 743/** 744 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 745 * 746 * If keep_me == NULL, rules are deleted even if not expired, 747 * otherwise only expired rules are removed. 748 * 749 * The value of the second parameter is also used to point to identify 750 * a rule we absolutely do not want to remove (e.g. because we are 751 * holding a reference to it -- this is the case with O_LIMIT_PARENT 752 * rules). The pointer is only used for comparison, so any non-null 753 * value will do. 754 */ 755static void 756remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 757{ 758 static u_int32_t last_remove = 0; 759 760#define FORCE (keep_me == NULL) 761 762 ipfw_dyn_rule *prev, *q; 763 int i, pass = 0, max_pass = 0; 764 765 IPFW_DYN_LOCK_ASSERT(); 766 767 if (ipfw_dyn_v == NULL || dyn_count == 0) 768 return; 769 /* do not expire more than once per second, it is useless */ 770 if (!FORCE && last_remove == time_second) 771 return; 772 last_remove = time_second; 773 774 /* 775 * because O_LIMIT refer to parent rules, during the first pass only 776 * remove child and mark any pending LIMIT_PARENT, and remove 777 * them in a second pass. 778 */ 779next_pass: 780 for (i = 0 ; i < curr_dyn_buckets ; i++) { 781 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 782 /* 783 * Logic can become complex here, so we split tests. 784 */ 785 if (q == keep_me) 786 goto next; 787 if (rule != NULL && rule != q->rule) 788 goto next; /* not the one we are looking for */ 789 if (q->dyn_type == O_LIMIT_PARENT) { 790 /* 791 * handle parent in the second pass, 792 * record we need one. 793 */ 794 max_pass = 1; 795 if (pass == 0) 796 goto next; 797 if (FORCE && q->count != 0 ) { 798 /* XXX should not happen! */ 799 printf("ipfw: OUCH! cannot remove rule," 800 " count %d\n", q->count); 801 } 802 } else { 803 if (!FORCE && 804 !TIME_LEQ( q->expire, time_second )) 805 goto next; 806 } 807 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 808 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 809 continue; 810 } 811next: 812 prev=q; 813 q=q->next; 814 } 815 } 816 if (pass++ < max_pass) 817 goto next_pass; 818} 819 820 821/** 822 * lookup a dynamic rule. 823 */ 824static ipfw_dyn_rule * 825lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 826 struct tcphdr *tcp) 827{ 828 /* 829 * stateful ipfw extensions. 830 * Lookup into dynamic session queue 831 */ 832#define MATCH_REVERSE 0 833#define MATCH_FORWARD 1 834#define MATCH_NONE 2 835#define MATCH_UNKNOWN 3 836 int i, dir = MATCH_NONE; 837 ipfw_dyn_rule *prev, *q=NULL; 838 839 IPFW_DYN_LOCK_ASSERT(); 840 841 if (ipfw_dyn_v == NULL) 842 goto done; /* not found */ 843 i = hash_packet( pkt ); 844 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 845 if (q->dyn_type == O_LIMIT_PARENT && q->count) 846 goto next; 847 if (TIME_LEQ( q->expire, time_second)) { /* expire entry */ 848 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 849 continue; 850 } 851 if (pkt->proto == q->id.proto && 852 q->dyn_type != O_LIMIT_PARENT) { 853 if (pkt->src_ip == q->id.src_ip && 854 pkt->dst_ip == q->id.dst_ip && 855 pkt->src_port == q->id.src_port && 856 pkt->dst_port == q->id.dst_port ) { 857 dir = MATCH_FORWARD; 858 break; 859 } 860 if (pkt->src_ip == q->id.dst_ip && 861 pkt->dst_ip == q->id.src_ip && 862 pkt->src_port == q->id.dst_port && 863 pkt->dst_port == q->id.src_port ) { 864 dir = MATCH_REVERSE; 865 break; 866 } 867 } 868next: 869 prev = q; 870 q = q->next; 871 } 872 if (q == NULL) 873 goto done; /* q = NULL, not found */ 874 875 if ( prev != NULL) { /* found and not in front */ 876 prev->next = q->next; 877 q->next = ipfw_dyn_v[i]; 878 ipfw_dyn_v[i] = q; 879 } 880 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 881 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 882 883#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 884#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 885 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 886 switch (q->state) { 887 case TH_SYN: /* opening */ 888 q->expire = time_second + dyn_syn_lifetime; 889 break; 890 891 case BOTH_SYN: /* move to established */ 892 case BOTH_SYN | TH_FIN : /* one side tries to close */ 893 case BOTH_SYN | (TH_FIN << 8) : 894 if (tcp) { 895#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 896 u_int32_t ack = ntohl(tcp->th_ack); 897 if (dir == MATCH_FORWARD) { 898 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 899 q->ack_fwd = ack; 900 else { /* ignore out-of-sequence */ 901 break; 902 } 903 } else { 904 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 905 q->ack_rev = ack; 906 else { /* ignore out-of-sequence */ 907 break; 908 } 909 } 910 } 911 q->expire = time_second + dyn_ack_lifetime; 912 break; 913 914 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 915 if (dyn_fin_lifetime >= dyn_keepalive_period) 916 dyn_fin_lifetime = dyn_keepalive_period - 1; 917 q->expire = time_second + dyn_fin_lifetime; 918 break; 919 920 default: 921#if 0 922 /* 923 * reset or some invalid combination, but can also 924 * occur if we use keep-state the wrong way. 925 */ 926 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 927 printf("invalid state: 0x%x\n", q->state); 928#endif 929 if (dyn_rst_lifetime >= dyn_keepalive_period) 930 dyn_rst_lifetime = dyn_keepalive_period - 1; 931 q->expire = time_second + dyn_rst_lifetime; 932 break; 933 } 934 } else if (pkt->proto == IPPROTO_UDP) { 935 q->expire = time_second + dyn_udp_lifetime; 936 } else { 937 /* other protocols */ 938 q->expire = time_second + dyn_short_lifetime; 939 } 940done: 941 if (match_direction) 942 *match_direction = dir; 943 return q; 944} 945 946static ipfw_dyn_rule * 947lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 948 struct tcphdr *tcp) 949{ 950 ipfw_dyn_rule *q; 951 952 IPFW_DYN_LOCK(); 953 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 954 if (q == NULL) 955 IPFW_DYN_UNLOCK(); 956 /* NB: return table locked when q is not NULL */ 957 return q; 958} 959 960static void 961realloc_dynamic_table(void) 962{ 963 IPFW_DYN_LOCK_ASSERT(); 964 965 /* 966 * Try reallocation, make sure we have a power of 2 and do 967 * not allow more than 64k entries. In case of overflow, 968 * default to 1024. 969 */ 970 971 if (dyn_buckets > 65536) 972 dyn_buckets = 1024; 973 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 974 dyn_buckets = curr_dyn_buckets; /* reset */ 975 return; 976 } 977 curr_dyn_buckets = dyn_buckets; 978 if (ipfw_dyn_v != NULL) 979 free(ipfw_dyn_v, M_IPFW); 980 for (;;) { 981 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 982 M_IPFW, M_NOWAIT | M_ZERO); 983 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 984 break; 985 curr_dyn_buckets /= 2; 986 } 987} 988 989/** 990 * Install state of type 'type' for a dynamic session. 991 * The hash table contains two type of rules: 992 * - regular rules (O_KEEP_STATE) 993 * - rules for sessions with limited number of sess per user 994 * (O_LIMIT). When they are created, the parent is 995 * increased by 1, and decreased on delete. In this case, 996 * the third parameter is the parent rule and not the chain. 997 * - "parent" rules for the above (O_LIMIT_PARENT). 998 */ 999static ipfw_dyn_rule * 1000add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1001{ 1002 ipfw_dyn_rule *r; 1003 int i; 1004 1005 IPFW_DYN_LOCK_ASSERT(); 1006 1007 if (ipfw_dyn_v == NULL || 1008 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 1009 realloc_dynamic_table(); 1010 if (ipfw_dyn_v == NULL) 1011 return NULL; /* failed ! */ 1012 } 1013 i = hash_packet(id); 1014 1015 r = malloc(sizeof *r, M_IPFW, M_NOWAIT | M_ZERO); 1016 if (r == NULL) { 1017 printf ("ipfw: sorry cannot allocate state\n"); 1018 return NULL; 1019 } 1020 1021 /* increase refcount on parent, and set pointer */ 1022 if (dyn_type == O_LIMIT) { 1023 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1024 if ( parent->dyn_type != O_LIMIT_PARENT) 1025 panic("invalid parent"); 1026 parent->count++; 1027 r->parent = parent; 1028 rule = parent->rule; 1029 } 1030 1031 r->id = *id; 1032 r->expire = time_second + dyn_syn_lifetime; 1033 r->rule = rule; 1034 r->dyn_type = dyn_type; 1035 r->pcnt = r->bcnt = 0; 1036 r->count = 0; 1037 1038 r->bucket = i; 1039 r->next = ipfw_dyn_v[i]; 1040 ipfw_dyn_v[i] = r; 1041 dyn_count++; 1042 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1043 dyn_type, 1044 (r->id.src_ip), (r->id.src_port), 1045 (r->id.dst_ip), (r->id.dst_port), 1046 dyn_count ); ) 1047 return r; 1048} 1049 1050/** 1051 * lookup dynamic parent rule using pkt and rule as search keys. 1052 * If the lookup fails, then install one. 1053 */ 1054static ipfw_dyn_rule * 1055lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1056{ 1057 ipfw_dyn_rule *q; 1058 int i; 1059 1060 IPFW_DYN_LOCK_ASSERT(); 1061 1062 if (ipfw_dyn_v) { 1063 i = hash_packet( pkt ); 1064 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1065 if (q->dyn_type == O_LIMIT_PARENT && 1066 rule== q->rule && 1067 pkt->proto == q->id.proto && 1068 pkt->src_ip == q->id.src_ip && 1069 pkt->dst_ip == q->id.dst_ip && 1070 pkt->src_port == q->id.src_port && 1071 pkt->dst_port == q->id.dst_port) { 1072 q->expire = time_second + dyn_short_lifetime; 1073 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1074 return q; 1075 } 1076 } 1077 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1078} 1079 1080/** 1081 * Install dynamic state for rule type cmd->o.opcode 1082 * 1083 * Returns 1 (failure) if state is not installed because of errors or because 1084 * session limitations are enforced. 1085 */ 1086static int 1087install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1088 struct ip_fw_args *args) 1089{ 1090 static int last_log; 1091 1092 ipfw_dyn_rule *q; 1093 1094 DEB(printf("ipfw: install state type %d 0x%08x %u -> 0x%08x %u\n", 1095 cmd->o.opcode, 1096 (args->f_id.src_ip), (args->f_id.src_port), 1097 (args->f_id.dst_ip), (args->f_id.dst_port) );) 1098 1099 IPFW_DYN_LOCK(); 1100 1101 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1102 1103 if (q != NULL) { /* should never occur */ 1104 if (last_log != time_second) { 1105 last_log = time_second; 1106 printf("ipfw: install_state: entry already present, done\n"); 1107 } 1108 IPFW_DYN_UNLOCK(); 1109 return 0; 1110 } 1111 1112 if (dyn_count >= dyn_max) 1113 /* 1114 * Run out of slots, try to remove any expired rule. 1115 */ 1116 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1117 1118 if (dyn_count >= dyn_max) { 1119 if (last_log != time_second) { 1120 last_log = time_second; 1121 printf("ipfw: install_state: Too many dynamic rules\n"); 1122 } 1123 IPFW_DYN_UNLOCK(); 1124 return 1; /* cannot install, notify caller */ 1125 } 1126 1127 switch (cmd->o.opcode) { 1128 case O_KEEP_STATE: /* bidir rule */ 1129 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1130 break; 1131 1132 case O_LIMIT: /* limit number of sessions */ 1133 { 1134 u_int16_t limit_mask = cmd->limit_mask; 1135 struct ipfw_flow_id id; 1136 ipfw_dyn_rule *parent; 1137 1138 DEB(printf("ipfw: installing dyn-limit rule %d\n", 1139 cmd->conn_limit);) 1140 1141 id.dst_ip = id.src_ip = 0; 1142 id.dst_port = id.src_port = 0; 1143 id.proto = args->f_id.proto; 1144 1145 if (limit_mask & DYN_SRC_ADDR) 1146 id.src_ip = args->f_id.src_ip; 1147 if (limit_mask & DYN_DST_ADDR) 1148 id.dst_ip = args->f_id.dst_ip; 1149 if (limit_mask & DYN_SRC_PORT) 1150 id.src_port = args->f_id.src_port; 1151 if (limit_mask & DYN_DST_PORT) 1152 id.dst_port = args->f_id.dst_port; 1153 parent = lookup_dyn_parent(&id, rule); 1154 if (parent == NULL) { 1155 printf("ipfw: add parent failed\n"); 1156 return 1; 1157 } 1158 if (parent->count >= cmd->conn_limit) { 1159 /* 1160 * See if we can remove some expired rule. 1161 */ 1162 remove_dyn_rule(rule, parent); 1163 if (parent->count >= cmd->conn_limit) { 1164 if (fw_verbose && last_log != time_second) { 1165 last_log = time_second; 1166 log(LOG_SECURITY | LOG_DEBUG, 1167 "drop session, too many entries\n"); 1168 } 1169 IPFW_DYN_UNLOCK(); 1170 return 1; 1171 } 1172 } 1173 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1174 } 1175 break; 1176 default: 1177 printf("ipfw: unknown dynamic rule type %u\n", cmd->o.opcode); 1178 IPFW_DYN_UNLOCK(); 1179 return 1; 1180 } 1181 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); /* XXX just set lifetime */ 1182 IPFW_DYN_UNLOCK(); 1183 return 0; 1184} 1185 1186/* 1187 * Transmit a TCP packet, containing either a RST or a keepalive. 1188 * When flags & TH_RST, we are sending a RST packet, because of a 1189 * "reset" action matched the packet. 1190 * Otherwise we are sending a keepalive, and flags & TH_ 1191 */ 1192static void 1193send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags) 1194{ 1195 struct mbuf *m; 1196 struct ip *ip; 1197 struct tcphdr *tcp; 1198 1199 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1200 if (m == 0) 1201 return; 1202 m->m_pkthdr.rcvif = (struct ifnet *)0; 1203 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1204 m->m_data += max_linkhdr; 1205 1206 ip = mtod(m, struct ip *); 1207 bzero(ip, m->m_len); 1208 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1209 ip->ip_p = IPPROTO_TCP; 1210 tcp->th_off = 5; 1211 /* 1212 * Assume we are sending a RST (or a keepalive in the reverse 1213 * direction), swap src and destination addresses and ports. 1214 */ 1215 ip->ip_src.s_addr = htonl(id->dst_ip); 1216 ip->ip_dst.s_addr = htonl(id->src_ip); 1217 tcp->th_sport = htons(id->dst_port); 1218 tcp->th_dport = htons(id->src_port); 1219 if (flags & TH_RST) { /* we are sending a RST */ 1220 if (flags & TH_ACK) { 1221 tcp->th_seq = htonl(ack); 1222 tcp->th_ack = htonl(0); 1223 tcp->th_flags = TH_RST; 1224 } else { 1225 if (flags & TH_SYN) 1226 seq++; 1227 tcp->th_seq = htonl(0); 1228 tcp->th_ack = htonl(seq); 1229 tcp->th_flags = TH_RST | TH_ACK; 1230 } 1231 } else { 1232 /* 1233 * We are sending a keepalive. flags & TH_SYN determines 1234 * the direction, forward if set, reverse if clear. 1235 * NOTE: seq and ack are always assumed to be correct 1236 * as set by the caller. This may be confusing... 1237 */ 1238 if (flags & TH_SYN) { 1239 /* 1240 * we have to rewrite the correct addresses! 1241 */ 1242 ip->ip_dst.s_addr = htonl(id->dst_ip); 1243 ip->ip_src.s_addr = htonl(id->src_ip); 1244 tcp->th_dport = htons(id->dst_port); 1245 tcp->th_sport = htons(id->src_port); 1246 } 1247 tcp->th_seq = htonl(seq); 1248 tcp->th_ack = htonl(ack); 1249 tcp->th_flags = TH_ACK; 1250 } 1251 /* 1252 * set ip_len to the payload size so we can compute 1253 * the tcp checksum on the pseudoheader 1254 * XXX check this, could save a couple of words ? 1255 */ 1256 ip->ip_len = htons(sizeof(struct tcphdr)); 1257 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1258 /* 1259 * now fill fields left out earlier 1260 */ 1261 ip->ip_ttl = ip_defttl; 1262 ip->ip_len = m->m_pkthdr.len; 1263 m->m_flags |= M_SKIP_FIREWALL; 1264 ip_output(m, NULL, NULL, 0, NULL, NULL); 1265} 1266 1267/* 1268 * sends a reject message, consuming the mbuf passed as an argument. 1269 */ 1270static void 1271send_reject(struct ip_fw_args *args, int code, int offset, int ip_len) 1272{ 1273 1274 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1275 /* We need the IP header in host order for icmp_error(). */ 1276 if (args->eh != NULL) { 1277 struct ip *ip = mtod(args->m, struct ip *); 1278 ip->ip_len = ntohs(ip->ip_len); 1279 ip->ip_off = ntohs(ip->ip_off); 1280 } 1281 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1282 } else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) { 1283 struct tcphdr *const tcp = 1284 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1285 if ( (tcp->th_flags & TH_RST) == 0) 1286 send_pkt(&(args->f_id), ntohl(tcp->th_seq), 1287 ntohl(tcp->th_ack), 1288 tcp->th_flags | TH_RST); 1289 m_freem(args->m); 1290 } else 1291 m_freem(args->m); 1292 args->m = NULL; 1293} 1294 1295/** 1296 * 1297 * Given an ip_fw *, lookup_next_rule will return a pointer 1298 * to the next rule, which can be either the jump 1299 * target (for skipto instructions) or the next one in the list (in 1300 * all other cases including a missing jump target). 1301 * The result is also written in the "next_rule" field of the rule. 1302 * Backward jumps are not allowed, so start looking from the next 1303 * rule... 1304 * 1305 * This never returns NULL -- in case we do not have an exact match, 1306 * the next rule is returned. When the ruleset is changed, 1307 * pointers are flushed so we are always correct. 1308 */ 1309 1310static struct ip_fw * 1311lookup_next_rule(struct ip_fw *me) 1312{ 1313 struct ip_fw *rule = NULL; 1314 ipfw_insn *cmd; 1315 1316 /* look for action, in case it is a skipto */ 1317 cmd = ACTION_PTR(me); 1318 if (cmd->opcode == O_LOG) 1319 cmd += F_LEN(cmd); 1320 if ( cmd->opcode == O_SKIPTO ) 1321 for (rule = me->next; rule ; rule = rule->next) 1322 if (rule->rulenum >= cmd->arg1) 1323 break; 1324 if (rule == NULL) /* failure or not a skipto */ 1325 rule = me->next; 1326 me->next_rule = rule; 1327 return rule; 1328} 1329 1330static void 1331init_tables(void) 1332{ 1333 int i; 1334 1335 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1336 rn_inithead((void **)&ipfw_tables[i].rnh, 32); 1337 ipfw_tables[i].modified = 1; 1338 } 1339} 1340 1341static int 1342add_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen, u_int32_t value) 1343{ 1344 struct radix_node_head *rnh; 1345 struct table_entry *ent; 1346 1347 if (tbl >= IPFW_TABLES_MAX) 1348 return (EINVAL); 1349 rnh = ipfw_tables[tbl].rnh; 1350 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1351 if (ent == NULL) 1352 return (ENOMEM); 1353 ent->value = value; 1354 ent->addr.sin_len = ent->mask.sin_len = 8; 1355 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1356 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1357 RADIX_NODE_HEAD_LOCK(rnh); 1358 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) == 1359 NULL) { 1360 RADIX_NODE_HEAD_UNLOCK(rnh); 1361 free(ent, M_IPFW_TBL); 1362 return (EEXIST); 1363 } 1364 ipfw_tables[tbl].modified = 1; 1365 RADIX_NODE_HEAD_UNLOCK(rnh); 1366 return (0); 1367} 1368 1369static int 1370del_table_entry(u_int16_t tbl, in_addr_t addr, u_int8_t mlen) 1371{ 1372 struct radix_node_head *rnh; 1373 struct table_entry *ent; 1374 struct sockaddr_in sa, mask; 1375 1376 if (tbl >= IPFW_TABLES_MAX) 1377 return (EINVAL); 1378 rnh = ipfw_tables[tbl].rnh; 1379 sa.sin_len = mask.sin_len = 8; 1380 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1381 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1382 RADIX_NODE_HEAD_LOCK(rnh); 1383 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1384 if (ent == NULL) { 1385 RADIX_NODE_HEAD_UNLOCK(rnh); 1386 return (ESRCH); 1387 } 1388 ipfw_tables[tbl].modified = 1; 1389 RADIX_NODE_HEAD_UNLOCK(rnh); 1390 free(ent, M_IPFW_TBL); 1391 return (0); 1392} 1393 1394static int 1395flush_table_entry(struct radix_node *rn, void *arg) 1396{ 1397 struct radix_node_head * const rnh = arg; 1398 struct table_entry *ent; 1399 1400 ent = (struct table_entry *) 1401 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1402 if (ent != NULL) 1403 free(ent, M_IPFW_TBL); 1404 return (0); 1405} 1406 1407static int 1408flush_table(u_int16_t tbl) 1409{ 1410 struct radix_node_head *rnh; 1411 1412 if (tbl >= IPFW_TABLES_MAX) 1413 return (EINVAL); 1414 rnh = ipfw_tables[tbl].rnh; 1415 RADIX_NODE_HEAD_LOCK(rnh); 1416 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1417 ipfw_tables[tbl].modified = 1; 1418 RADIX_NODE_HEAD_UNLOCK(rnh); 1419 return (0); 1420} 1421 1422static void 1423flush_tables(void) 1424{ 1425 u_int16_t tbl; 1426 1427 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1428 flush_table(tbl); 1429} 1430 1431static int 1432lookup_table(u_int16_t tbl, in_addr_t addr, u_int32_t *val) 1433{ 1434 struct radix_node_head *rnh; 1435 struct table_entry *ent; 1436 struct sockaddr_in sa; 1437 static in_addr_t last_addr; 1438 static int last_tbl; 1439 static int last_match; 1440 static u_int32_t last_value; 1441 1442 if (tbl >= IPFW_TABLES_MAX) 1443 return (0); 1444 if (tbl == last_tbl && addr == last_addr && 1445 !ipfw_tables[tbl].modified) { 1446 if (last_match) 1447 *val = last_value; 1448 return (last_match); 1449 } 1450 rnh = ipfw_tables[tbl].rnh; 1451 sa.sin_len = 8; 1452 sa.sin_addr.s_addr = addr; 1453 RADIX_NODE_HEAD_LOCK(rnh); 1454 ipfw_tables[tbl].modified = 0; 1455 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1456 RADIX_NODE_HEAD_UNLOCK(rnh); 1457 last_addr = addr; 1458 last_tbl = tbl; 1459 if (ent != NULL) { 1460 last_value = *val = ent->value; 1461 last_match = 1; 1462 return (1); 1463 } 1464 last_match = 0; 1465 return (0); 1466} 1467 1468static int 1469count_table_entry(struct radix_node *rn, void *arg) 1470{ 1471 u_int32_t * const cnt = arg; 1472 1473 (*cnt)++; 1474 return (0); 1475} 1476 1477static int 1478count_table(u_int32_t tbl, u_int32_t *cnt) 1479{ 1480 struct radix_node_head *rnh; 1481 1482 if (tbl >= IPFW_TABLES_MAX) 1483 return (EINVAL); 1484 rnh = ipfw_tables[tbl].rnh; 1485 *cnt = 0; 1486 RADIX_NODE_HEAD_LOCK(rnh); 1487 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1488 RADIX_NODE_HEAD_UNLOCK(rnh); 1489 return (0); 1490} 1491 1492static int 1493dump_table_entry(struct radix_node *rn, void *arg) 1494{ 1495 struct table_entry * const n = (struct table_entry *)rn; 1496 ipfw_table * const tbl = arg; 1497 ipfw_table_entry *ent; 1498 1499 if (tbl->cnt == tbl->size) 1500 return (1); 1501 ent = &tbl->ent[tbl->cnt]; 1502 ent->tbl = tbl->tbl; 1503 if (in_nullhost(n->mask.sin_addr)) 1504 ent->masklen = 0; 1505 else 1506 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1507 ent->addr = n->addr.sin_addr.s_addr; 1508 ent->value = n->value; 1509 tbl->cnt++; 1510 return (0); 1511} 1512 1513static int 1514dump_table(ipfw_table *tbl) 1515{ 1516 struct radix_node_head *rnh; 1517 1518 if (tbl->tbl >= IPFW_TABLES_MAX) 1519 return (EINVAL); 1520 rnh = ipfw_tables[tbl->tbl].rnh; 1521 tbl->cnt = 0; 1522 RADIX_NODE_HEAD_LOCK(rnh); 1523 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 1524 RADIX_NODE_HEAD_UNLOCK(rnh); 1525 return (0); 1526} 1527 1528static int 1529check_uidgid(ipfw_insn_u32 *insn, 1530 int proto, struct ifnet *oif, 1531 struct in_addr dst_ip, u_int16_t dst_port, 1532 struct in_addr src_ip, u_int16_t src_port) 1533{ 1534 struct inpcbinfo *pi; 1535 int wildcard; 1536 struct inpcb *pcb; 1537 int match; 1538 1539 if (proto == IPPROTO_TCP) { 1540 wildcard = 0; 1541 pi = &tcbinfo; 1542 } else if (proto == IPPROTO_UDP) { 1543 wildcard = 1; 1544 pi = &udbinfo; 1545 } else 1546 return 0; 1547 1548 match = 0; 1549 1550 INP_INFO_RLOCK(pi); /* XXX LOR with IPFW */ 1551 pcb = (oif) ? 1552 in_pcblookup_hash(pi, 1553 dst_ip, htons(dst_port), 1554 src_ip, htons(src_port), 1555 wildcard, oif) : 1556 in_pcblookup_hash(pi, 1557 src_ip, htons(src_port), 1558 dst_ip, htons(dst_port), 1559 wildcard, NULL); 1560 if (pcb != NULL) { 1561 INP_LOCK(pcb); 1562 if (pcb->inp_socket != NULL) { 1563#if __FreeBSD_version < 500034 1564#define socheckuid(a,b) ((a)->so_cred->cr_uid != (b)) 1565#endif 1566 if (insn->o.opcode == O_UID) { 1567 match = !socheckuid(pcb->inp_socket, 1568 (uid_t)insn->d[0]); 1569 } else { 1570 match = groupmember((uid_t)insn->d[0], 1571 pcb->inp_socket->so_cred); 1572 } 1573 } 1574 INP_UNLOCK(pcb); 1575 } 1576 INP_INFO_RUNLOCK(pi); 1577 1578 return match; 1579} 1580 1581/* 1582 * The main check routine for the firewall. 1583 * 1584 * All arguments are in args so we can modify them and return them 1585 * back to the caller. 1586 * 1587 * Parameters: 1588 * 1589 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1590 * Starts with the IP header. 1591 * args->eh (in) Mac header if present, or NULL for layer3 packet. 1592 * args->oif Outgoing interface, or NULL if packet is incoming. 1593 * The incoming interface is in the mbuf. (in) 1594 * args->divert_rule (in/out) 1595 * Skip up to the first rule past this rule number; 1596 * upon return, non-zero port number for divert or tee. 1597 * 1598 * args->rule Pointer to the last matching rule (in/out) 1599 * args->next_hop Socket we are forwarding to (out). 1600 * args->f_id Addresses grabbed from the packet (out) 1601 * 1602 * Return value: 1603 * 1604 * IP_FW_PORT_DENY_FLAG the packet must be dropped. 1605 * 0 The packet is to be accepted and routed normally OR 1606 * the packet was denied/rejected and has been dropped; 1607 * in the latter case, *m is equal to NULL upon return. 1608 * port Divert the packet to port, with these caveats: 1609 * 1610 * - If IP_FW_PORT_TEE_FLAG is set, tee the packet instead 1611 * of diverting it (ie, 'ipfw tee'). 1612 * 1613 * - If IP_FW_PORT_DYNT_FLAG is set, interpret the lower 1614 * 16 bits as a dummynet pipe number instead of diverting 1615 */ 1616 1617static int 1618ipfw_chk(struct ip_fw_args *args) 1619{ 1620 /* 1621 * Local variables hold state during the processing of a packet. 1622 * 1623 * IMPORTANT NOTE: to speed up the processing of rules, there 1624 * are some assumption on the values of the variables, which 1625 * are documented here. Should you change them, please check 1626 * the implementation of the various instructions to make sure 1627 * that they still work. 1628 * 1629 * args->eh The MAC header. It is non-null for a layer2 1630 * packet, it is NULL for a layer-3 packet. 1631 * 1632 * m | args->m Pointer to the mbuf, as received from the caller. 1633 * It may change if ipfw_chk() does an m_pullup, or if it 1634 * consumes the packet because it calls send_reject(). 1635 * XXX This has to change, so that ipfw_chk() never modifies 1636 * or consumes the buffer. 1637 * ip is simply an alias of the value of m, and it is kept 1638 * in sync with it (the packet is supposed to start with 1639 * the ip header). 1640 */ 1641 struct mbuf *m = args->m; 1642 struct ip *ip = mtod(m, struct ip *); 1643 1644 /* 1645 * oif | args->oif If NULL, ipfw_chk has been called on the 1646 * inbound path (ether_input, bdg_forward, ip_input). 1647 * If non-NULL, ipfw_chk has been called on the outbound path 1648 * (ether_output, ip_output). 1649 */ 1650 struct ifnet *oif = args->oif; 1651 1652 struct ip_fw *f = NULL; /* matching rule */ 1653 int retval = 0; 1654 1655 /* 1656 * hlen The length of the IPv4 header. 1657 * hlen >0 means we have an IPv4 packet. 1658 */ 1659 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 1660 1661 /* 1662 * offset The offset of a fragment. offset != 0 means that 1663 * we have a fragment at this offset of an IPv4 packet. 1664 * offset == 0 means that (if this is an IPv4 packet) 1665 * this is the first or only fragment. 1666 */ 1667 u_short offset = 0; 1668 1669 /* 1670 * Local copies of addresses. They are only valid if we have 1671 * an IP packet. 1672 * 1673 * proto The protocol. Set to 0 for non-ip packets, 1674 * or to the protocol read from the packet otherwise. 1675 * proto != 0 means that we have an IPv4 packet. 1676 * 1677 * src_port, dst_port port numbers, in HOST format. Only 1678 * valid for TCP and UDP packets. 1679 * 1680 * src_ip, dst_ip ip addresses, in NETWORK format. 1681 * Only valid for IPv4 packets. 1682 */ 1683 u_int8_t proto; 1684 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 1685 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 1686 u_int16_t ip_len=0; 1687 int pktlen; 1688 int dyn_dir = MATCH_UNKNOWN; 1689 ipfw_dyn_rule *q = NULL; 1690 struct ip_fw_chain *chain = &layer3_chain; 1691 struct m_tag *mtag; 1692 1693 if (m->m_flags & M_SKIP_FIREWALL) 1694 return 0; /* accept */ 1695 /* 1696 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 1697 * MATCH_NONE when checked and not matched (q = NULL), 1698 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 1699 */ 1700 1701 pktlen = m->m_pkthdr.len; 1702 if (args->eh == NULL || /* layer 3 packet */ 1703 ( m->m_pkthdr.len >= sizeof(struct ip) && 1704 ntohs(args->eh->ether_type) == ETHERTYPE_IP)) 1705 hlen = ip->ip_hl << 2; 1706 1707 /* 1708 * Collect parameters into local variables for faster matching. 1709 */ 1710 if (hlen == 0) { /* do not grab addresses for non-ip pkts */ 1711 proto = args->f_id.proto = 0; /* mark f_id invalid */ 1712 goto after_ip_checks; 1713 } 1714 1715 proto = args->f_id.proto = ip->ip_p; 1716 src_ip = ip->ip_src; 1717 dst_ip = ip->ip_dst; 1718 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 1719 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1720 ip_len = ntohs(ip->ip_len); 1721 } else { 1722 offset = ip->ip_off & IP_OFFMASK; 1723 ip_len = ip->ip_len; 1724 } 1725 pktlen = ip_len < pktlen ? ip_len : pktlen; 1726 1727#define PULLUP_TO(len) \ 1728 do { \ 1729 if ((m)->m_len < (len)) { \ 1730 args->m = m = m_pullup(m, (len)); \ 1731 if (m == 0) \ 1732 goto pullup_failed; \ 1733 ip = mtod(m, struct ip *); \ 1734 } \ 1735 } while (0) 1736 1737 if (offset == 0) { 1738 switch (proto) { 1739 case IPPROTO_TCP: 1740 { 1741 struct tcphdr *tcp; 1742 1743 PULLUP_TO(hlen + sizeof(struct tcphdr)); 1744 tcp = L3HDR(struct tcphdr, ip); 1745 dst_port = tcp->th_dport; 1746 src_port = tcp->th_sport; 1747 args->f_id.flags = tcp->th_flags; 1748 } 1749 break; 1750 1751 case IPPROTO_UDP: 1752 { 1753 struct udphdr *udp; 1754 1755 PULLUP_TO(hlen + sizeof(struct udphdr)); 1756 udp = L3HDR(struct udphdr, ip); 1757 dst_port = udp->uh_dport; 1758 src_port = udp->uh_sport; 1759 } 1760 break; 1761 1762 case IPPROTO_ICMP: 1763 PULLUP_TO(hlen + 4); /* type, code and checksum. */ 1764 args->f_id.flags = L3HDR(struct icmp, ip)->icmp_type; 1765 break; 1766 1767 default: 1768 break; 1769 } 1770#undef PULLUP_TO 1771 } 1772 1773 args->f_id.src_ip = ntohl(src_ip.s_addr); 1774 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1775 args->f_id.src_port = src_port = ntohs(src_port); 1776 args->f_id.dst_port = dst_port = ntohs(dst_port); 1777 1778after_ip_checks: 1779 IPFW_LOCK(chain); /* XXX expensive? can we run lock free? */ 1780 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 1781 if (args->rule) { 1782 /* 1783 * Packet has already been tagged. Look for the next rule 1784 * to restart processing. 1785 * 1786 * If fw_one_pass != 0 then just accept it. 1787 * XXX should not happen here, but optimized out in 1788 * the caller. 1789 */ 1790 if (fw_one_pass) { 1791 IPFW_UNLOCK(chain); /* XXX optimize */ 1792 return 0; 1793 } 1794 1795 f = args->rule->next_rule; 1796 if (f == NULL) 1797 f = lookup_next_rule(args->rule); 1798 } else { 1799 /* 1800 * Find the starting rule. It can be either the first 1801 * one, or the one after divert_rule if asked so. 1802 */ 1803 int skipto = mtag ? divert_cookie(mtag) : 0; 1804 1805 f = chain->rules; 1806 if (args->eh == NULL && skipto != 0) { 1807 if (skipto >= IPFW_DEFAULT_RULE) { 1808 IPFW_UNLOCK(chain); 1809 return(IP_FW_PORT_DENY_FLAG); /* invalid */ 1810 } 1811 while (f && f->rulenum <= skipto) 1812 f = f->next; 1813 if (f == NULL) { /* drop packet */ 1814 IPFW_UNLOCK(chain); 1815 return(IP_FW_PORT_DENY_FLAG); 1816 } 1817 } 1818 } 1819 /* reset divert rule to avoid confusion later */ 1820 if (mtag) 1821 m_tag_delete(m, mtag); 1822 1823 /* 1824 * Now scan the rules, and parse microinstructions for each rule. 1825 */ 1826 for (; f; f = f->next) { 1827 int l, cmdlen; 1828 ipfw_insn *cmd; 1829 int skip_or; /* skip rest of OR block */ 1830 1831again: 1832 if (set_disable & (1 << f->set) ) 1833 continue; 1834 1835 skip_or = 0; 1836 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1837 l -= cmdlen, cmd += cmdlen) { 1838 int match; 1839 1840 /* 1841 * check_body is a jump target used when we find a 1842 * CHECK_STATE, and need to jump to the body of 1843 * the target rule. 1844 */ 1845 1846check_body: 1847 cmdlen = F_LEN(cmd); 1848 /* 1849 * An OR block (insn_1 || .. || insn_n) has the 1850 * F_OR bit set in all but the last instruction. 1851 * The first match will set "skip_or", and cause 1852 * the following instructions to be skipped until 1853 * past the one with the F_OR bit clear. 1854 */ 1855 if (skip_or) { /* skip this instruction */ 1856 if ((cmd->len & F_OR) == 0) 1857 skip_or = 0; /* next one is good */ 1858 continue; 1859 } 1860 match = 0; /* set to 1 if we succeed */ 1861 1862 switch (cmd->opcode) { 1863 /* 1864 * The first set of opcodes compares the packet's 1865 * fields with some pattern, setting 'match' if a 1866 * match is found. At the end of the loop there is 1867 * logic to deal with F_NOT and F_OR flags associated 1868 * with the opcode. 1869 */ 1870 case O_NOP: 1871 match = 1; 1872 break; 1873 1874 case O_FORWARD_MAC: 1875 printf("ipfw: opcode %d unimplemented\n", 1876 cmd->opcode); 1877 break; 1878 1879 case O_GID: 1880 case O_UID: 1881 /* 1882 * We only check offset == 0 && proto != 0, 1883 * as this ensures that we have an IPv4 1884 * packet with the ports info. 1885 */ 1886 if (offset!=0) 1887 break; 1888 if (proto == IPPROTO_TCP || 1889 proto == IPPROTO_UDP) 1890 match = check_uidgid( 1891 (ipfw_insn_u32 *)cmd, 1892 proto, oif, 1893 dst_ip, dst_port, 1894 src_ip, src_port); 1895 break; 1896 1897 case O_RECV: 1898 match = iface_match(m->m_pkthdr.rcvif, 1899 (ipfw_insn_if *)cmd); 1900 break; 1901 1902 case O_XMIT: 1903 match = iface_match(oif, (ipfw_insn_if *)cmd); 1904 break; 1905 1906 case O_VIA: 1907 match = iface_match(oif ? oif : 1908 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 1909 break; 1910 1911 case O_MACADDR2: 1912 if (args->eh != NULL) { /* have MAC header */ 1913 u_int32_t *want = (u_int32_t *) 1914 ((ipfw_insn_mac *)cmd)->addr; 1915 u_int32_t *mask = (u_int32_t *) 1916 ((ipfw_insn_mac *)cmd)->mask; 1917 u_int32_t *hdr = (u_int32_t *)args->eh; 1918 1919 match = 1920 ( want[0] == (hdr[0] & mask[0]) && 1921 want[1] == (hdr[1] & mask[1]) && 1922 want[2] == (hdr[2] & mask[2]) ); 1923 } 1924 break; 1925 1926 case O_MAC_TYPE: 1927 if (args->eh != NULL) { 1928 u_int16_t t = 1929 ntohs(args->eh->ether_type); 1930 u_int16_t *p = 1931 ((ipfw_insn_u16 *)cmd)->ports; 1932 int i; 1933 1934 for (i = cmdlen - 1; !match && i>0; 1935 i--, p += 2) 1936 match = (t>=p[0] && t<=p[1]); 1937 } 1938 break; 1939 1940 case O_FRAG: 1941 match = (hlen > 0 && offset != 0); 1942 break; 1943 1944 case O_IN: /* "out" is "not in" */ 1945 match = (oif == NULL); 1946 break; 1947 1948 case O_LAYER2: 1949 match = (args->eh != NULL); 1950 break; 1951 1952 case O_PROTO: 1953 /* 1954 * We do not allow an arg of 0 so the 1955 * check of "proto" only suffices. 1956 */ 1957 match = (proto == cmd->arg1); 1958 break; 1959 1960 case O_IP_SRC: 1961 match = (hlen > 0 && 1962 ((ipfw_insn_ip *)cmd)->addr.s_addr == 1963 src_ip.s_addr); 1964 break; 1965 1966 case O_IP_SRC_LOOKUP: 1967 case O_IP_DST_LOOKUP: 1968 if (hlen > 0) { 1969 uint32_t a = 1970 (cmd->opcode == O_IP_DST_LOOKUP) ? 1971 dst_ip.s_addr : src_ip.s_addr; 1972 uint32_t v; 1973 1974 match = lookup_table(cmd->arg1, a, &v); 1975 if (!match) 1976 break; 1977 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 1978 match = 1979 ((ipfw_insn_u32 *)cmd)->d[0] == v; 1980 } 1981 break; 1982 1983 case O_IP_SRC_MASK: 1984 case O_IP_DST_MASK: 1985 if (hlen > 0) { 1986 uint32_t a = 1987 (cmd->opcode == O_IP_DST_MASK) ? 1988 dst_ip.s_addr : src_ip.s_addr; 1989 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 1990 int i = cmdlen-1; 1991 1992 for (; !match && i>0; i-= 2, p+= 2) 1993 match = (p[0] == (a & p[1])); 1994 } 1995 break; 1996 1997 case O_IP_SRC_ME: 1998 if (hlen > 0) { 1999 struct ifnet *tif; 2000 2001 INADDR_TO_IFP(src_ip, tif); 2002 match = (tif != NULL); 2003 } 2004 break; 2005 2006 case O_IP_DST_SET: 2007 case O_IP_SRC_SET: 2008 if (hlen > 0) { 2009 u_int32_t *d = (u_int32_t *)(cmd+1); 2010 u_int32_t addr = 2011 cmd->opcode == O_IP_DST_SET ? 2012 args->f_id.dst_ip : 2013 args->f_id.src_ip; 2014 2015 if (addr < d[0]) 2016 break; 2017 addr -= d[0]; /* subtract base */ 2018 match = (addr < cmd->arg1) && 2019 ( d[ 1 + (addr>>5)] & 2020 (1<<(addr & 0x1f)) ); 2021 } 2022 break; 2023 2024 case O_IP_DST: 2025 match = (hlen > 0 && 2026 ((ipfw_insn_ip *)cmd)->addr.s_addr == 2027 dst_ip.s_addr); 2028 break; 2029 2030 case O_IP_DST_ME: 2031 if (hlen > 0) { 2032 struct ifnet *tif; 2033 2034 INADDR_TO_IFP(dst_ip, tif); 2035 match = (tif != NULL); 2036 } 2037 break; 2038 2039 case O_IP_SRCPORT: 2040 case O_IP_DSTPORT: 2041 /* 2042 * offset == 0 && proto != 0 is enough 2043 * to guarantee that we have an IPv4 2044 * packet with port info. 2045 */ 2046 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2047 && offset == 0) { 2048 u_int16_t x = 2049 (cmd->opcode == O_IP_SRCPORT) ? 2050 src_port : dst_port ; 2051 u_int16_t *p = 2052 ((ipfw_insn_u16 *)cmd)->ports; 2053 int i; 2054 2055 for (i = cmdlen - 1; !match && i>0; 2056 i--, p += 2) 2057 match = (x>=p[0] && x<=p[1]); 2058 } 2059 break; 2060 2061 case O_ICMPTYPE: 2062 match = (offset == 0 && proto==IPPROTO_ICMP && 2063 icmptype_match(ip, (ipfw_insn_u32 *)cmd) ); 2064 break; 2065 2066 case O_IPOPT: 2067 match = (hlen > 0 && ipopts_match(ip, cmd) ); 2068 break; 2069 2070 case O_IPVER: 2071 match = (hlen > 0 && cmd->arg1 == ip->ip_v); 2072 break; 2073 2074 case O_IPID: 2075 case O_IPLEN: 2076 case O_IPTTL: 2077 if (hlen > 0) { /* only for IP packets */ 2078 uint16_t x; 2079 uint16_t *p; 2080 int i; 2081 2082 if (cmd->opcode == O_IPLEN) 2083 x = ip_len; 2084 else if (cmd->opcode == O_IPTTL) 2085 x = ip->ip_ttl; 2086 else /* must be IPID */ 2087 x = ntohs(ip->ip_id); 2088 if (cmdlen == 1) { 2089 match = (cmd->arg1 == x); 2090 break; 2091 } 2092 /* otherwise we have ranges */ 2093 p = ((ipfw_insn_u16 *)cmd)->ports; 2094 i = cmdlen - 1; 2095 for (; !match && i>0; i--, p += 2) 2096 match = (x >= p[0] && x <= p[1]); 2097 } 2098 break; 2099 2100 case O_IPPRECEDENCE: 2101 match = (hlen > 0 && 2102 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2103 break; 2104 2105 case O_IPTOS: 2106 match = (hlen > 0 && 2107 flags_match(cmd, ip->ip_tos)); 2108 break; 2109 2110 case O_TCPFLAGS: 2111 match = (proto == IPPROTO_TCP && offset == 0 && 2112 flags_match(cmd, 2113 L3HDR(struct tcphdr,ip)->th_flags)); 2114 break; 2115 2116 case O_TCPOPTS: 2117 match = (proto == IPPROTO_TCP && offset == 0 && 2118 tcpopts_match(ip, cmd)); 2119 break; 2120 2121 case O_TCPSEQ: 2122 match = (proto == IPPROTO_TCP && offset == 0 && 2123 ((ipfw_insn_u32 *)cmd)->d[0] == 2124 L3HDR(struct tcphdr,ip)->th_seq); 2125 break; 2126 2127 case O_TCPACK: 2128 match = (proto == IPPROTO_TCP && offset == 0 && 2129 ((ipfw_insn_u32 *)cmd)->d[0] == 2130 L3HDR(struct tcphdr,ip)->th_ack); 2131 break; 2132 2133 case O_TCPWIN: 2134 match = (proto == IPPROTO_TCP && offset == 0 && 2135 cmd->arg1 == 2136 L3HDR(struct tcphdr,ip)->th_win); 2137 break; 2138 2139 case O_ESTAB: 2140 /* reject packets which have SYN only */ 2141 /* XXX should i also check for TH_ACK ? */ 2142 match = (proto == IPPROTO_TCP && offset == 0 && 2143 (L3HDR(struct tcphdr,ip)->th_flags & 2144 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2145 break; 2146 2147 case O_LOG: 2148 if (fw_verbose) 2149 ipfw_log(f, hlen, args->eh, m, oif); 2150 match = 1; 2151 break; 2152 2153 case O_PROB: 2154 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2155 break; 2156 2157 case O_VERREVPATH: 2158 /* Outgoing packets automatically pass/match */ 2159 match = ((oif != NULL) || 2160 (m->m_pkthdr.rcvif == NULL) || 2161 verify_path(src_ip, m->m_pkthdr.rcvif)); 2162 break; 2163 2164 case O_VERSRCREACH: 2165 /* Outgoing packets automatically pass/match */ 2166 match = ((oif != NULL) || 2167 verify_path(src_ip, NULL)); 2168 break; 2169 2170 case O_IPSEC: 2171#ifdef FAST_IPSEC 2172 match = (m_tag_find(m, 2173 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 2174#endif 2175#ifdef IPSEC 2176 match = (ipsec_getnhist(m) != 0); 2177#endif 2178 /* otherwise no match */ 2179 break; 2180 2181 /* 2182 * The second set of opcodes represents 'actions', 2183 * i.e. the terminal part of a rule once the packet 2184 * matches all previous patterns. 2185 * Typically there is only one action for each rule, 2186 * and the opcode is stored at the end of the rule 2187 * (but there are exceptions -- see below). 2188 * 2189 * In general, here we set retval and terminate the 2190 * outer loop (would be a 'break 3' in some language, 2191 * but we need to do a 'goto done'). 2192 * 2193 * Exceptions: 2194 * O_COUNT and O_SKIPTO actions: 2195 * instead of terminating, we jump to the next rule 2196 * ('goto next_rule', equivalent to a 'break 2'), 2197 * or to the SKIPTO target ('goto again' after 2198 * having set f, cmd and l), respectively. 2199 * 2200 * O_LIMIT and O_KEEP_STATE: these opcodes are 2201 * not real 'actions', and are stored right 2202 * before the 'action' part of the rule. 2203 * These opcodes try to install an entry in the 2204 * state tables; if successful, we continue with 2205 * the next opcode (match=1; break;), otherwise 2206 * the packet * must be dropped 2207 * ('goto done' after setting retval); 2208 * 2209 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 2210 * cause a lookup of the state table, and a jump 2211 * to the 'action' part of the parent rule 2212 * ('goto check_body') if an entry is found, or 2213 * (CHECK_STATE only) a jump to the next rule if 2214 * the entry is not found ('goto next_rule'). 2215 * The result of the lookup is cached to make 2216 * further instances of these opcodes are 2217 * effectively NOPs. 2218 */ 2219 case O_LIMIT: 2220 case O_KEEP_STATE: 2221 if (install_state(f, 2222 (ipfw_insn_limit *)cmd, args)) { 2223 retval = IP_FW_PORT_DENY_FLAG; 2224 goto done; /* error/limit violation */ 2225 } 2226 match = 1; 2227 break; 2228 2229 case O_PROBE_STATE: 2230 case O_CHECK_STATE: 2231 /* 2232 * dynamic rules are checked at the first 2233 * keep-state or check-state occurrence, 2234 * with the result being stored in dyn_dir. 2235 * The compiler introduces a PROBE_STATE 2236 * instruction for us when we have a 2237 * KEEP_STATE (because PROBE_STATE needs 2238 * to be run first). 2239 */ 2240 if (dyn_dir == MATCH_UNKNOWN && 2241 (q = lookup_dyn_rule(&args->f_id, 2242 &dyn_dir, proto == IPPROTO_TCP ? 2243 L3HDR(struct tcphdr, ip) : NULL)) 2244 != NULL) { 2245 /* 2246 * Found dynamic entry, update stats 2247 * and jump to the 'action' part of 2248 * the parent rule. 2249 */ 2250 q->pcnt++; 2251 q->bcnt += pktlen; 2252 f = q->rule; 2253 cmd = ACTION_PTR(f); 2254 l = f->cmd_len - f->act_ofs; 2255 IPFW_DYN_UNLOCK(); 2256 goto check_body; 2257 } 2258 /* 2259 * Dynamic entry not found. If CHECK_STATE, 2260 * skip to next rule, if PROBE_STATE just 2261 * ignore and continue with next opcode. 2262 */ 2263 if (cmd->opcode == O_CHECK_STATE) 2264 goto next_rule; 2265 match = 1; 2266 break; 2267 2268 case O_ACCEPT: 2269 retval = 0; /* accept */ 2270 goto done; 2271 2272 case O_PIPE: 2273 case O_QUEUE: 2274 args->rule = f; /* report matching rule */ 2275 retval = cmd->arg1 | IP_FW_PORT_DYNT_FLAG; 2276 goto done; 2277 2278 case O_DIVERT: 2279 case O_TEE: { 2280 struct divert_tag *dt; 2281 2282 if (args->eh) /* not on layer 2 */ 2283 break; 2284 mtag = m_tag_get(PACKET_TAG_DIVERT, 2285 sizeof(struct divert_tag), 2286 M_NOWAIT); 2287 if (mtag == NULL) { 2288 /* XXX statistic */ 2289 /* drop packet */ 2290 IPFW_UNLOCK(chain); 2291 return IP_FW_PORT_DENY_FLAG; 2292 } 2293 dt = (struct divert_tag *)(mtag+1); 2294 dt->cookie = f->rulenum; 2295 dt->info = (cmd->opcode == O_DIVERT) ? 2296 cmd->arg1 : 2297 cmd->arg1 | IP_FW_PORT_TEE_FLAG; 2298 m_tag_prepend(m, mtag); 2299 retval = dt->info; 2300 goto done; 2301 } 2302 2303 case O_COUNT: 2304 case O_SKIPTO: 2305 f->pcnt++; /* update stats */ 2306 f->bcnt += pktlen; 2307 f->timestamp = time_second; 2308 if (cmd->opcode == O_COUNT) 2309 goto next_rule; 2310 /* handle skipto */ 2311 if (f->next_rule == NULL) 2312 lookup_next_rule(f); 2313 f = f->next_rule; 2314 goto again; 2315 2316 case O_REJECT: 2317 /* 2318 * Drop the packet and send a reject notice 2319 * if the packet is not ICMP (or is an ICMP 2320 * query), and it is not multicast/broadcast. 2321 */ 2322 if (hlen > 0 && 2323 (proto != IPPROTO_ICMP || 2324 is_icmp_query(ip)) && 2325 !(m->m_flags & (M_BCAST|M_MCAST)) && 2326 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 2327 send_reject(args, cmd->arg1, 2328 offset,ip_len); 2329 m = args->m; 2330 } 2331 /* FALLTHROUGH */ 2332 case O_DENY: 2333 retval = IP_FW_PORT_DENY_FLAG; 2334 goto done; 2335 2336 case O_FORWARD_IP: 2337 if (args->eh) /* not valid on layer2 pkts */ 2338 break; 2339 if (!q || dyn_dir == MATCH_FORWARD) 2340 args->next_hop = 2341 &((ipfw_insn_sa *)cmd)->sa; 2342 retval = 0; 2343 goto done; 2344 2345 default: 2346 panic("-- unknown opcode %d\n", cmd->opcode); 2347 } /* end of switch() on opcodes */ 2348 2349 if (cmd->len & F_NOT) 2350 match = !match; 2351 2352 if (match) { 2353 if (cmd->len & F_OR) 2354 skip_or = 1; 2355 } else { 2356 if (!(cmd->len & F_OR)) /* not an OR block, */ 2357 break; /* try next rule */ 2358 } 2359 2360 } /* end of inner for, scan opcodes */ 2361 2362next_rule:; /* try next rule */ 2363 2364 } /* end of outer for, scan rules */ 2365 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 2366 IPFW_UNLOCK(chain); 2367 return(IP_FW_PORT_DENY_FLAG); 2368 2369done: 2370 /* Update statistics */ 2371 f->pcnt++; 2372 f->bcnt += pktlen; 2373 f->timestamp = time_second; 2374 IPFW_UNLOCK(chain); 2375 return retval; 2376 2377pullup_failed: 2378 if (fw_verbose) 2379 printf("ipfw: pullup failed\n"); 2380 return(IP_FW_PORT_DENY_FLAG); 2381} 2382 2383/* 2384 * When a rule is added/deleted, clear the next_rule pointers in all rules. 2385 * These will be reconstructed on the fly as packets are matched. 2386 */ 2387static void 2388flush_rule_ptrs(struct ip_fw_chain *chain) 2389{ 2390 struct ip_fw *rule; 2391 2392 IPFW_LOCK_ASSERT(chain); 2393 2394 for (rule = chain->rules; rule; rule = rule->next) 2395 rule->next_rule = NULL; 2396} 2397 2398/* 2399 * When pipes/queues are deleted, clear the "pipe_ptr" pointer to a given 2400 * pipe/queue, or to all of them (match == NULL). 2401 */ 2402void 2403flush_pipe_ptrs(struct dn_flow_set *match) 2404{ 2405 struct ip_fw *rule; 2406 2407 IPFW_LOCK(&layer3_chain); 2408 for (rule = layer3_chain.rules; rule; rule = rule->next) { 2409 ipfw_insn_pipe *cmd = (ipfw_insn_pipe *)ACTION_PTR(rule); 2410 2411 if (cmd->o.opcode != O_PIPE && cmd->o.opcode != O_QUEUE) 2412 continue; 2413 /* 2414 * XXX Use bcmp/bzero to handle pipe_ptr to overcome 2415 * possible alignment problems on 64-bit architectures. 2416 * This code is seldom used so we do not worry too 2417 * much about efficiency. 2418 */ 2419 if (match == NULL || 2420 !bcmp(&cmd->pipe_ptr, &match, sizeof(match)) ) 2421 bzero(&cmd->pipe_ptr, sizeof(cmd->pipe_ptr)); 2422 } 2423 IPFW_UNLOCK(&layer3_chain); 2424} 2425 2426/* 2427 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 2428 * possibly create a rule number and add the rule to the list. 2429 * Update the rule_number in the input struct so the caller knows it as well. 2430 */ 2431static int 2432add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 2433{ 2434 struct ip_fw *rule, *f, *prev; 2435 int l = RULESIZE(input_rule); 2436 2437 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 2438 return (EINVAL); 2439 2440 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 2441 if (rule == NULL) 2442 return (ENOSPC); 2443 2444 bcopy(input_rule, rule, l); 2445 2446 rule->next = NULL; 2447 rule->next_rule = NULL; 2448 2449 rule->pcnt = 0; 2450 rule->bcnt = 0; 2451 rule->timestamp = 0; 2452 2453 IPFW_LOCK(chain); 2454 2455 if (chain->rules == NULL) { /* default rule */ 2456 chain->rules = rule; 2457 goto done; 2458 } 2459 2460 /* 2461 * If rulenum is 0, find highest numbered rule before the 2462 * default rule, and add autoinc_step 2463 */ 2464 if (autoinc_step < 1) 2465 autoinc_step = 1; 2466 else if (autoinc_step > 1000) 2467 autoinc_step = 1000; 2468 if (rule->rulenum == 0) { 2469 /* 2470 * locate the highest numbered rule before default 2471 */ 2472 for (f = chain->rules; f; f = f->next) { 2473 if (f->rulenum == IPFW_DEFAULT_RULE) 2474 break; 2475 rule->rulenum = f->rulenum; 2476 } 2477 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 2478 rule->rulenum += autoinc_step; 2479 input_rule->rulenum = rule->rulenum; 2480 } 2481 2482 /* 2483 * Now insert the new rule in the right place in the sorted list. 2484 */ 2485 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 2486 if (f->rulenum > rule->rulenum) { /* found the location */ 2487 if (prev) { 2488 rule->next = f; 2489 prev->next = rule; 2490 } else { /* head insert */ 2491 rule->next = chain->rules; 2492 chain->rules = rule; 2493 } 2494 break; 2495 } 2496 } 2497 flush_rule_ptrs(chain); 2498done: 2499 static_count++; 2500 static_len += l; 2501 IPFW_UNLOCK(chain); 2502 DEB(printf("ipfw: installed rule %d, static count now %d\n", 2503 rule->rulenum, static_count);) 2504 return (0); 2505} 2506 2507/** 2508 * Remove a static rule (including derived * dynamic rules) 2509 * and place it on the ``reap list'' for later reclamation. 2510 * The caller is in charge of clearing rule pointers to avoid 2511 * dangling pointers. 2512 * @return a pointer to the next entry. 2513 * Arguments are not checked, so they better be correct. 2514 */ 2515static struct ip_fw * 2516remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, struct ip_fw *prev) 2517{ 2518 struct ip_fw *n; 2519 int l = RULESIZE(rule); 2520 2521 IPFW_LOCK_ASSERT(chain); 2522 2523 n = rule->next; 2524 IPFW_DYN_LOCK(); 2525 remove_dyn_rule(rule, NULL /* force removal */); 2526 IPFW_DYN_UNLOCK(); 2527 if (prev == NULL) 2528 chain->rules = n; 2529 else 2530 prev->next = n; 2531 static_count--; 2532 static_len -= l; 2533 2534 rule->next = chain->reap; 2535 chain->reap = rule; 2536 2537 return n; 2538} 2539 2540/** 2541 * Reclaim storage associated with a list of rules. This is 2542 * typically the list created using remove_rule. 2543 */ 2544static void 2545reap_rules(struct ip_fw *head) 2546{ 2547 struct ip_fw *rule; 2548 2549 while ((rule = head) != NULL) { 2550 head = head->next; 2551 if (DUMMYNET_LOADED) 2552 ip_dn_ruledel_ptr(rule); 2553 free(rule, M_IPFW); 2554 } 2555} 2556 2557/* 2558 * Remove all rules from a chain (except rules in set RESVD_SET 2559 * unless kill_default = 1). The caller is responsible for 2560 * reclaiming storage for the rules left in chain->reap. 2561 */ 2562static void 2563free_chain(struct ip_fw_chain *chain, int kill_default) 2564{ 2565 struct ip_fw *prev, *rule; 2566 2567 IPFW_LOCK_ASSERT(chain); 2568 2569 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 2570 for (prev = NULL, rule = chain->rules; rule ; ) 2571 if (kill_default || rule->set != RESVD_SET) 2572 rule = remove_rule(chain, rule, prev); 2573 else { 2574 prev = rule; 2575 rule = rule->next; 2576 } 2577} 2578 2579/** 2580 * Remove all rules with given number, and also do set manipulation. 2581 * Assumes chain != NULL && *chain != NULL. 2582 * 2583 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 2584 * the next 8 bits are the new set, the top 8 bits are the command: 2585 * 2586 * 0 delete rules with given number 2587 * 1 delete rules with given set number 2588 * 2 move rules with given number to new set 2589 * 3 move rules with given set number to new set 2590 * 4 swap sets with given numbers 2591 */ 2592static int 2593del_entry(struct ip_fw_chain *chain, u_int32_t arg) 2594{ 2595 struct ip_fw *prev = NULL, *rule; 2596 u_int16_t rulenum; /* rule or old_set */ 2597 u_int8_t cmd, new_set; 2598 2599 rulenum = arg & 0xffff; 2600 cmd = (arg >> 24) & 0xff; 2601 new_set = (arg >> 16) & 0xff; 2602 2603 if (cmd > 4) 2604 return EINVAL; 2605 if (new_set > RESVD_SET) 2606 return EINVAL; 2607 if (cmd == 0 || cmd == 2) { 2608 if (rulenum >= IPFW_DEFAULT_RULE) 2609 return EINVAL; 2610 } else { 2611 if (rulenum > RESVD_SET) /* old_set */ 2612 return EINVAL; 2613 } 2614 2615 IPFW_LOCK(chain); 2616 rule = chain->rules; 2617 chain->reap = NULL; 2618 switch (cmd) { 2619 case 0: /* delete rules with given number */ 2620 /* 2621 * locate first rule to delete 2622 */ 2623 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 2624 ; 2625 if (rule->rulenum != rulenum) { 2626 IPFW_UNLOCK(chain); 2627 return EINVAL; 2628 } 2629 2630 /* 2631 * flush pointers outside the loop, then delete all matching 2632 * rules. prev remains the same throughout the cycle. 2633 */ 2634 flush_rule_ptrs(chain); 2635 while (rule->rulenum == rulenum) 2636 rule = remove_rule(chain, rule, prev); 2637 break; 2638 2639 case 1: /* delete all rules with given set number */ 2640 flush_rule_ptrs(chain); 2641 rule = chain->rules; 2642 while (rule->rulenum < IPFW_DEFAULT_RULE) 2643 if (rule->set == rulenum) 2644 rule = remove_rule(chain, rule, prev); 2645 else { 2646 prev = rule; 2647 rule = rule->next; 2648 } 2649 break; 2650 2651 case 2: /* move rules with given number to new set */ 2652 rule = chain->rules; 2653 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 2654 if (rule->rulenum == rulenum) 2655 rule->set = new_set; 2656 break; 2657 2658 case 3: /* move rules with given set number to new set */ 2659 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 2660 if (rule->set == rulenum) 2661 rule->set = new_set; 2662 break; 2663 2664 case 4: /* swap two sets */ 2665 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 2666 if (rule->set == rulenum) 2667 rule->set = new_set; 2668 else if (rule->set == new_set) 2669 rule->set = rulenum; 2670 break; 2671 } 2672 /* 2673 * Look for rules to reclaim. We grab the list before 2674 * releasing the lock then reclaim them w/o the lock to 2675 * avoid a LOR with dummynet. 2676 */ 2677 rule = chain->reap; 2678 chain->reap = NULL; 2679 IPFW_UNLOCK(chain); 2680 if (rule) 2681 reap_rules(rule); 2682 return 0; 2683} 2684 2685/* 2686 * Clear counters for a specific rule. 2687 * The enclosing "table" is assumed locked. 2688 */ 2689static void 2690clear_counters(struct ip_fw *rule, int log_only) 2691{ 2692 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 2693 2694 if (log_only == 0) { 2695 rule->bcnt = rule->pcnt = 0; 2696 rule->timestamp = 0; 2697 } 2698 if (l->o.opcode == O_LOG) 2699 l->log_left = l->max_log; 2700} 2701 2702/** 2703 * Reset some or all counters on firewall rules. 2704 * @arg frwl is null to clear all entries, or contains a specific 2705 * rule number. 2706 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 2707 */ 2708static int 2709zero_entry(struct ip_fw_chain *chain, int rulenum, int log_only) 2710{ 2711 struct ip_fw *rule; 2712 char *msg; 2713 2714 IPFW_LOCK(chain); 2715 if (rulenum == 0) { 2716 norule_counter = 0; 2717 for (rule = chain->rules; rule; rule = rule->next) 2718 clear_counters(rule, log_only); 2719 msg = log_only ? "ipfw: All logging counts reset.\n" : 2720 "ipfw: Accounting cleared.\n"; 2721 } else { 2722 int cleared = 0; 2723 /* 2724 * We can have multiple rules with the same number, so we 2725 * need to clear them all. 2726 */ 2727 for (rule = chain->rules; rule; rule = rule->next) 2728 if (rule->rulenum == rulenum) { 2729 while (rule && rule->rulenum == rulenum) { 2730 clear_counters(rule, log_only); 2731 rule = rule->next; 2732 } 2733 cleared = 1; 2734 break; 2735 } 2736 if (!cleared) { /* we did not find any matching rules */ 2737 IPFW_UNLOCK(chain); 2738 return (EINVAL); 2739 } 2740 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 2741 "ipfw: Entry %d cleared.\n"; 2742 } 2743 IPFW_UNLOCK(chain); 2744 2745 if (fw_verbose) 2746 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 2747 return (0); 2748} 2749 2750/* 2751 * Check validity of the structure before insert. 2752 * Fortunately rules are simple, so this mostly need to check rule sizes. 2753 */ 2754static int 2755check_ipfw_struct(struct ip_fw *rule, int size) 2756{ 2757 int l, cmdlen = 0; 2758 int have_action=0; 2759 ipfw_insn *cmd; 2760 2761 if (size < sizeof(*rule)) { 2762 printf("ipfw: rule too short\n"); 2763 return (EINVAL); 2764 } 2765 /* first, check for valid size */ 2766 l = RULESIZE(rule); 2767 if (l != size) { 2768 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 2769 return (EINVAL); 2770 } 2771 /* 2772 * Now go for the individual checks. Very simple ones, basically only 2773 * instruction sizes. 2774 */ 2775 for (l = rule->cmd_len, cmd = rule->cmd ; 2776 l > 0 ; l -= cmdlen, cmd += cmdlen) { 2777 cmdlen = F_LEN(cmd); 2778 if (cmdlen > l) { 2779 printf("ipfw: opcode %d size truncated\n", 2780 cmd->opcode); 2781 return EINVAL; 2782 } 2783 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 2784 switch (cmd->opcode) { 2785 case O_PROBE_STATE: 2786 case O_KEEP_STATE: 2787 case O_PROTO: 2788 case O_IP_SRC_ME: 2789 case O_IP_DST_ME: 2790 case O_LAYER2: 2791 case O_IN: 2792 case O_FRAG: 2793 case O_IPOPT: 2794 case O_IPTOS: 2795 case O_IPPRECEDENCE: 2796 case O_IPVER: 2797 case O_TCPWIN: 2798 case O_TCPFLAGS: 2799 case O_TCPOPTS: 2800 case O_ESTAB: 2801 case O_VERREVPATH: 2802 case O_VERSRCREACH: 2803 case O_IPSEC: 2804 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 2805 goto bad_size; 2806 break; 2807 2808 case O_UID: 2809 case O_GID: 2810 case O_IP_SRC: 2811 case O_IP_DST: 2812 case O_TCPSEQ: 2813 case O_TCPACK: 2814 case O_PROB: 2815 case O_ICMPTYPE: 2816 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 2817 goto bad_size; 2818 break; 2819 2820 case O_LIMIT: 2821 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 2822 goto bad_size; 2823 break; 2824 2825 case O_LOG: 2826 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 2827 goto bad_size; 2828 2829 ((ipfw_insn_log *)cmd)->log_left = 2830 ((ipfw_insn_log *)cmd)->max_log; 2831 2832 break; 2833 2834 case O_IP_SRC_MASK: 2835 case O_IP_DST_MASK: 2836 /* only odd command lengths */ 2837 if ( !(cmdlen & 1) || cmdlen > 31) 2838 goto bad_size; 2839 break; 2840 2841 case O_IP_SRC_SET: 2842 case O_IP_DST_SET: 2843 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 2844 printf("ipfw: invalid set size %d\n", 2845 cmd->arg1); 2846 return EINVAL; 2847 } 2848 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 2849 (cmd->arg1+31)/32 ) 2850 goto bad_size; 2851 break; 2852 2853 case O_IP_SRC_LOOKUP: 2854 case O_IP_DST_LOOKUP: 2855 if (cmd->arg1 >= IPFW_TABLES_MAX) { 2856 printf("ipfw: invalid table number %d\n", 2857 cmd->arg1); 2858 return (EINVAL); 2859 } 2860 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 2861 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 2862 goto bad_size; 2863 break; 2864 2865 case O_MACADDR2: 2866 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 2867 goto bad_size; 2868 break; 2869 2870 case O_NOP: 2871 case O_IPID: 2872 case O_IPTTL: 2873 case O_IPLEN: 2874 if (cmdlen < 1 || cmdlen > 31) 2875 goto bad_size; 2876 break; 2877 2878 case O_MAC_TYPE: 2879 case O_IP_SRCPORT: 2880 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 2881 if (cmdlen < 2 || cmdlen > 31) 2882 goto bad_size; 2883 break; 2884 2885 case O_RECV: 2886 case O_XMIT: 2887 case O_VIA: 2888 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 2889 goto bad_size; 2890 break; 2891 2892 case O_PIPE: 2893 case O_QUEUE: 2894 if (cmdlen != F_INSN_SIZE(ipfw_insn_pipe)) 2895 goto bad_size; 2896 goto check_action; 2897 2898 case O_FORWARD_IP: 2899 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 2900 goto bad_size; 2901 goto check_action; 2902 2903 case O_FORWARD_MAC: /* XXX not implemented yet */ 2904 case O_CHECK_STATE: 2905 case O_COUNT: 2906 case O_ACCEPT: 2907 case O_DENY: 2908 case O_REJECT: 2909 case O_SKIPTO: 2910 case O_DIVERT: 2911 case O_TEE: 2912 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 2913 goto bad_size; 2914check_action: 2915 if (have_action) { 2916 printf("ipfw: opcode %d, multiple actions" 2917 " not allowed\n", 2918 cmd->opcode); 2919 return EINVAL; 2920 } 2921 have_action = 1; 2922 if (l != cmdlen) { 2923 printf("ipfw: opcode %d, action must be" 2924 " last opcode\n", 2925 cmd->opcode); 2926 return EINVAL; 2927 } 2928 break; 2929 default: 2930 printf("ipfw: opcode %d, unknown opcode\n", 2931 cmd->opcode); 2932 return EINVAL; 2933 } 2934 } 2935 if (have_action == 0) { 2936 printf("ipfw: missing action\n"); 2937 return EINVAL; 2938 } 2939 return 0; 2940 2941bad_size: 2942 printf("ipfw: opcode %d size %d wrong\n", 2943 cmd->opcode, cmdlen); 2944 return EINVAL; 2945} 2946 2947/* 2948 * Copy the static and dynamic rules to the supplied buffer 2949 * and return the amount of space actually used. 2950 */ 2951static size_t 2952ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 2953{ 2954 char *bp = buf; 2955 char *ep = bp + space; 2956 struct ip_fw *rule; 2957 int i; 2958 2959 /* XXX this can take a long time and locking will block packet flow */ 2960 IPFW_LOCK(chain); 2961 for (rule = chain->rules; rule ; rule = rule->next) { 2962 /* 2963 * Verify the entry fits in the buffer in case the 2964 * rules changed between calculating buffer space and 2965 * now. This would be better done using a generation 2966 * number but should suffice for now. 2967 */ 2968 i = RULESIZE(rule); 2969 if (bp + i <= ep) { 2970 bcopy(rule, bp, i); 2971 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule), 2972 sizeof(set_disable)); 2973 bp += i; 2974 } 2975 } 2976 IPFW_UNLOCK(chain); 2977 if (ipfw_dyn_v) { 2978 ipfw_dyn_rule *p, *last = NULL; 2979 2980 IPFW_DYN_LOCK(); 2981 for (i = 0 ; i < curr_dyn_buckets; i++) 2982 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) { 2983 if (bp + sizeof *p <= ep) { 2984 ipfw_dyn_rule *dst = 2985 (ipfw_dyn_rule *)bp; 2986 bcopy(p, dst, sizeof *p); 2987 bcopy(&(p->rule->rulenum), &(dst->rule), 2988 sizeof(p->rule->rulenum)); 2989 /* 2990 * store a non-null value in "next". 2991 * The userland code will interpret a 2992 * NULL here as a marker 2993 * for the last dynamic rule. 2994 */ 2995 bcopy(&dst, &dst->next, sizeof(dst)); 2996 last = dst; 2997 dst->expire = 2998 TIME_LEQ(dst->expire, time_second) ? 2999 0 : dst->expire - time_second ; 3000 bp += sizeof(ipfw_dyn_rule); 3001 } 3002 } 3003 IPFW_DYN_UNLOCK(); 3004 if (last != NULL) /* mark last dynamic rule */ 3005 bzero(&last->next, sizeof(last)); 3006 } 3007 return (bp - (char *)buf); 3008} 3009 3010 3011/** 3012 * {set|get}sockopt parser. 3013 */ 3014static int 3015ipfw_ctl(struct sockopt *sopt) 3016{ 3017#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 3018 int error, rule_num; 3019 size_t size; 3020 struct ip_fw *buf, *rule; 3021 u_int32_t rulenum[2]; 3022 3023 error = suser(sopt->sopt_td); 3024 if (error) 3025 return (error); 3026 3027 /* 3028 * Disallow modifications in really-really secure mode, but still allow 3029 * the logging counters to be reset. 3030 */ 3031 if (sopt->sopt_name == IP_FW_ADD || 3032 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 3033#if __FreeBSD_version >= 500034 3034 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 3035 if (error) 3036 return (error); 3037#else /* FreeBSD 4.x */ 3038 if (securelevel >= 3) 3039 return (EPERM); 3040#endif 3041 } 3042 3043 error = 0; 3044 3045 switch (sopt->sopt_name) { 3046 case IP_FW_GET: 3047 /* 3048 * pass up a copy of the current rules. Static rules 3049 * come first (the last of which has number IPFW_DEFAULT_RULE), 3050 * followed by a possibly empty list of dynamic rule. 3051 * The last dynamic rule has NULL in the "next" field. 3052 * 3053 * Note that the calculated size is used to bound the 3054 * amount of data returned to the user. The rule set may 3055 * change between calculating the size and returning the 3056 * data in which case we'll just return what fits. 3057 */ 3058 size = static_len; /* size of static rules */ 3059 if (ipfw_dyn_v) /* add size of dyn.rules */ 3060 size += (dyn_count * sizeof(ipfw_dyn_rule)); 3061 3062 /* 3063 * XXX todo: if the user passes a short length just to know 3064 * how much room is needed, do not bother filling up the 3065 * buffer, just jump to the sooptcopyout. 3066 */ 3067 buf = malloc(size, M_TEMP, M_WAITOK); 3068 error = sooptcopyout(sopt, buf, 3069 ipfw_getrules(&layer3_chain, buf, size)); 3070 free(buf, M_TEMP); 3071 break; 3072 3073 case IP_FW_FLUSH: 3074 /* 3075 * Normally we cannot release the lock on each iteration. 3076 * We could do it here only because we start from the head all 3077 * the times so there is no risk of missing some entries. 3078 * On the other hand, the risk is that we end up with 3079 * a very inconsistent ruleset, so better keep the lock 3080 * around the whole cycle. 3081 * 3082 * XXX this code can be improved by resetting the head of 3083 * the list to point to the default rule, and then freeing 3084 * the old list without the need for a lock. 3085 */ 3086 3087 IPFW_LOCK(&layer3_chain); 3088 layer3_chain.reap = NULL; 3089 free_chain(&layer3_chain, 0 /* keep default rule */); 3090 rule = layer3_chain.reap, layer3_chain.reap = NULL; 3091 IPFW_UNLOCK(&layer3_chain); 3092 if (layer3_chain.reap != NULL) 3093 reap_rules(rule); 3094 break; 3095 3096 case IP_FW_ADD: 3097 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 3098 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 3099 sizeof(struct ip_fw) ); 3100 if (error == 0) 3101 error = check_ipfw_struct(rule, sopt->sopt_valsize); 3102 if (error == 0) { 3103 error = add_rule(&layer3_chain, rule); 3104 size = RULESIZE(rule); 3105 if (!error && sopt->sopt_dir == SOPT_GET) 3106 error = sooptcopyout(sopt, rule, size); 3107 } 3108 free(rule, M_TEMP); 3109 break; 3110 3111 case IP_FW_DEL: 3112 /* 3113 * IP_FW_DEL is used for deleting single rules or sets, 3114 * and (ab)used to atomically manipulate sets. Argument size 3115 * is used to distinguish between the two: 3116 * sizeof(u_int32_t) 3117 * delete single rule or set of rules, 3118 * or reassign rules (or sets) to a different set. 3119 * 2*sizeof(u_int32_t) 3120 * atomic disable/enable sets. 3121 * first u_int32_t contains sets to be disabled, 3122 * second u_int32_t contains sets to be enabled. 3123 */ 3124 error = sooptcopyin(sopt, rulenum, 3125 2*sizeof(u_int32_t), sizeof(u_int32_t)); 3126 if (error) 3127 break; 3128 size = sopt->sopt_valsize; 3129 if (size == sizeof(u_int32_t)) /* delete or reassign */ 3130 error = del_entry(&layer3_chain, rulenum[0]); 3131 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 3132 set_disable = 3133 (set_disable | rulenum[0]) & ~rulenum[1] & 3134 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 3135 else 3136 error = EINVAL; 3137 break; 3138 3139 case IP_FW_ZERO: 3140 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 3141 rule_num = 0; 3142 if (sopt->sopt_val != 0) { 3143 error = sooptcopyin(sopt, &rule_num, 3144 sizeof(int), sizeof(int)); 3145 if (error) 3146 break; 3147 } 3148 error = zero_entry(&layer3_chain, rule_num, 3149 sopt->sopt_name == IP_FW_RESETLOG); 3150 break; 3151 3152 case IP_FW_TABLE_ADD: 3153 { 3154 ipfw_table_entry ent; 3155 3156 error = sooptcopyin(sopt, &ent, 3157 sizeof(ent), sizeof(ent)); 3158 if (error) 3159 break; 3160 error = add_table_entry(ent.tbl, ent.addr, 3161 ent.masklen, ent.value); 3162 } 3163 break; 3164 3165 case IP_FW_TABLE_DEL: 3166 { 3167 ipfw_table_entry ent; 3168 3169 error = sooptcopyin(sopt, &ent, 3170 sizeof(ent), sizeof(ent)); 3171 if (error) 3172 break; 3173 error = del_table_entry(ent.tbl, ent.addr, ent.masklen); 3174 } 3175 break; 3176 3177 case IP_FW_TABLE_FLUSH: 3178 { 3179 u_int16_t tbl; 3180 3181 error = sooptcopyin(sopt, &tbl, 3182 sizeof(tbl), sizeof(tbl)); 3183 if (error) 3184 break; 3185 error = flush_table(tbl); 3186 } 3187 break; 3188 3189 case IP_FW_TABLE_GETSIZE: 3190 { 3191 u_int32_t tbl, cnt; 3192 3193 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 3194 sizeof(tbl)))) 3195 break; 3196 if ((error = count_table(tbl, &cnt))) 3197 break; 3198 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 3199 } 3200 break; 3201 3202 case IP_FW_TABLE_LIST: 3203 { 3204 ipfw_table *tbl; 3205 3206 if (sopt->sopt_valsize < sizeof(*tbl)) { 3207 error = EINVAL; 3208 break; 3209 } 3210 size = sopt->sopt_valsize; 3211 tbl = malloc(size, M_TEMP, M_WAITOK); 3212 if (tbl == NULL) { 3213 error = ENOMEM; 3214 break; 3215 } 3216 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 3217 if (error) { 3218 free(tbl, M_TEMP); 3219 break; 3220 } 3221 tbl->size = (size - sizeof(*tbl)) / 3222 sizeof(ipfw_table_entry); 3223 error = dump_table(tbl); 3224 if (error) { 3225 free(tbl, M_TEMP); 3226 break; 3227 } 3228 error = sooptcopyout(sopt, tbl, size); 3229 free(tbl, M_TEMP); 3230 } 3231 break; 3232 3233 default: 3234 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 3235 error = EINVAL; 3236 } 3237 3238 return (error); 3239#undef RULE_MAXSIZE 3240} 3241 3242/** 3243 * dummynet needs a reference to the default rule, because rules can be 3244 * deleted while packets hold a reference to them. When this happens, 3245 * dummynet changes the reference to the default rule (it could well be a 3246 * NULL pointer, but this way we do not need to check for the special 3247 * case, plus here he have info on the default behaviour). 3248 */ 3249struct ip_fw *ip_fw_default_rule; 3250 3251/* 3252 * This procedure is only used to handle keepalives. It is invoked 3253 * every dyn_keepalive_period 3254 */ 3255static void 3256ipfw_tick(void * __unused unused) 3257{ 3258 int i; 3259 ipfw_dyn_rule *q; 3260 3261 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 3262 goto done; 3263 3264 IPFW_DYN_LOCK(); 3265 for (i = 0 ; i < curr_dyn_buckets ; i++) { 3266 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 3267 if (q->dyn_type == O_LIMIT_PARENT) 3268 continue; 3269 if (q->id.proto != IPPROTO_TCP) 3270 continue; 3271 if ( (q->state & BOTH_SYN) != BOTH_SYN) 3272 continue; 3273 if (TIME_LEQ( time_second+dyn_keepalive_interval, 3274 q->expire)) 3275 continue; /* too early */ 3276 if (TIME_LEQ(q->expire, time_second)) 3277 continue; /* too late, rule expired */ 3278 3279 send_pkt(&(q->id), q->ack_rev - 1, q->ack_fwd, TH_SYN); 3280 send_pkt(&(q->id), q->ack_fwd - 1, q->ack_rev, 0); 3281 } 3282 } 3283 IPFW_DYN_UNLOCK(); 3284done: 3285 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL); 3286} 3287 3288static int 3289ipfw_init(void) 3290{ 3291 struct ip_fw default_rule; 3292 int error; 3293 3294 layer3_chain.rules = NULL; 3295 IPFW_LOCK_INIT(&layer3_chain); 3296 IPFW_DYN_LOCK_INIT(); 3297 callout_init(&ipfw_timeout, debug_mpsafenet ? CALLOUT_MPSAFE : 0); 3298 3299 bzero(&default_rule, sizeof default_rule); 3300 3301 default_rule.act_ofs = 0; 3302 default_rule.rulenum = IPFW_DEFAULT_RULE; 3303 default_rule.cmd_len = 1; 3304 default_rule.set = RESVD_SET; 3305 3306 default_rule.cmd[0].len = 1; 3307 default_rule.cmd[0].opcode = 3308#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 3309 1 ? O_ACCEPT : 3310#endif 3311 O_DENY; 3312 3313 error = add_rule(&layer3_chain, &default_rule); 3314 if (error != 0) { 3315 printf("ipfw2: error %u initializing default rule " 3316 "(support disabled)\n", error); 3317 IPFW_DYN_LOCK_DESTROY(); 3318 IPFW_LOCK_DESTROY(&layer3_chain); 3319 return (error); 3320 } 3321 3322 ip_fw_default_rule = layer3_chain.rules; 3323 printf("ipfw2 initialized, divert %s, " 3324 "rule-based forwarding enabled, default to %s, logging ", 3325#ifdef IPDIVERT 3326 "enabled", 3327#else 3328 "disabled", 3329#endif 3330 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 3331 3332#ifdef IPFIREWALL_VERBOSE 3333 fw_verbose = 1; 3334#endif 3335#ifdef IPFIREWALL_VERBOSE_LIMIT 3336 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 3337#endif 3338 if (fw_verbose == 0) 3339 printf("disabled\n"); 3340 else if (verbose_limit == 0) 3341 printf("unlimited\n"); 3342 else 3343 printf("limited to %d packets/entry by default\n", 3344 verbose_limit); 3345 3346 ip_fw_chk_ptr = ipfw_chk; 3347 ip_fw_ctl_ptr = ipfw_ctl; 3348 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL); 3349 3350 return (0); 3351} 3352 3353static void 3354ipfw_destroy(void) 3355{ 3356 struct ip_fw *reap; 3357 3358 IPFW_LOCK(&layer3_chain); 3359 callout_stop(&ipfw_timeout); 3360 ip_fw_chk_ptr = NULL; 3361 ip_fw_ctl_ptr = NULL; 3362 layer3_chain.reap = NULL; 3363 free_chain(&layer3_chain, 1 /* kill default rule */); 3364 reap = layer3_chain.reap, layer3_chain.reap = NULL; 3365 IPFW_UNLOCK(&layer3_chain); 3366 if (reap != NULL) 3367 reap_rules(reap); 3368 flush_tables(); 3369 IPFW_DYN_LOCK_DESTROY(); 3370 IPFW_LOCK_DESTROY(&layer3_chain); 3371 printf("IP firewall unloaded\n"); 3372} 3373 3374static int 3375ipfw_modevent(module_t mod, int type, void *unused) 3376{ 3377 int err = 0; 3378 3379 switch (type) { 3380 case MOD_LOAD: 3381 if (IPFW_LOADED) { 3382 printf("IP firewall already loaded\n"); 3383 err = EEXIST; 3384 } else { 3385 err = ipfw_init(); 3386 } 3387 break; 3388 3389 case MOD_UNLOAD: 3390 ipfw_destroy(); 3391 err = 0; 3392 break; 3393 default: 3394 break; 3395 } 3396 return err; 3397} 3398 3399static moduledata_t ipfwmod = { 3400 "ipfw", 3401 ipfw_modevent, 3402 0 3403}; 3404DECLARE_MODULE(ipfw, ipfwmod, SI_SUB_PSEUDO, SI_ORDER_ANY); 3405MODULE_VERSION(ipfw, 1); 3406 3407/* Must be run after route_init(). */ 3408SYSINIT(ipfw, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY, init_tables, 0) 3409 3410#endif /* IPFW2 */ 3411