ip_fw2.c revision 115793
153913Sarchie/* 253913Sarchie * Copyright (c) 2002 Luigi Rizzo, Universita` di Pisa 353913Sarchie * 453913Sarchie * Redistribution and use in source and binary forms, with or without 553913Sarchie * modification, are permitted provided that the following conditions 653913Sarchie * are met: 753913Sarchie * 1. Redistributions of source code must retain the above copyright 853913Sarchie * notice, this list of conditions and the following disclaimer. 953913Sarchie * 2. Redistributions in binary form must reproduce the above copyright 1053913Sarchie * notice, this list of conditions and the following disclaimer in the 1153913Sarchie * documentation and/or other materials provided with the distribution. 1253913Sarchie * 1353913Sarchie * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 1453913Sarchie * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 1553913Sarchie * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 1653913Sarchie * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 1753913Sarchie * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 1853913Sarchie * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 1953913Sarchie * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 2053913Sarchie * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 2153913Sarchie * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 2253913Sarchie * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 2353913Sarchie * SUCH DAMAGE. 2453913Sarchie * 2553913Sarchie * $FreeBSD: head/sys/netinet/ip_fw2.c 115793 2003-06-04 01:17:37Z ticso $ 2653913Sarchie */ 2753913Sarchie 2853913Sarchie#define DEB(x) 2953913Sarchie#define DDB(x) x 3053913Sarchie 3153913Sarchie/* 3253913Sarchie * Implement IP packet firewall (new version) 3353913Sarchie */ 3453913Sarchie 3553913Sarchie#if !defined(KLD_MODULE) 3653913Sarchie#include "opt_ipfw.h" 3753913Sarchie#include "opt_ipdn.h" 3853913Sarchie#include "opt_ipdivert.h" 3953913Sarchie#include "opt_inet.h" 4053913Sarchie#ifndef INET 4153913Sarchie#error IPFIREWALL requires INET. 4253913Sarchie#endif /* INET */ 4353913Sarchie#endif 4453913Sarchie 4553913Sarchie#define IPFW2 1 4653913Sarchie#if IPFW2 4753913Sarchie#include <sys/param.h> 4853913Sarchie#include <sys/systm.h> 4953913Sarchie#include <sys/malloc.h> 5053998Sarchie#include <sys/mbuf.h> 5153998Sarchie#include <sys/kernel.h> 5253998Sarchie#include <sys/proc.h> 5353913Sarchie#include <sys/socket.h> 5453998Sarchie#include <sys/socketvar.h> 5553998Sarchie#include <sys/sysctl.h> 5653998Sarchie#include <sys/syslog.h> 5753998Sarchie#include <sys/ucred.h> 5853998Sarchie#include <net/if.h> 5953998Sarchie#include <net/route.h> 6053998Sarchie#include <netinet/in.h> 6153998Sarchie#include <netinet/in_systm.h> 6253998Sarchie#include <netinet/in_var.h> 6353913Sarchie#include <netinet/in_pcb.h> 6453913Sarchie#include <netinet/ip.h> 6553913Sarchie#include <netinet/ip_var.h> 6653913Sarchie#include <netinet/ip_icmp.h> 6753913Sarchie#include <netinet/ip_fw.h> 6853913Sarchie#include <netinet/ip_dummynet.h> 6953913Sarchie#include <netinet/tcp.h> 7053913Sarchie#include <netinet/tcp_timer.h> 7153913Sarchie#include <netinet/tcp_var.h> 7253913Sarchie#include <netinet/tcpip.h> 7353913Sarchie#include <netinet/udp.h> 7453913Sarchie#include <netinet/udp_var.h> 7553913Sarchie 7653913Sarchie#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 7753913Sarchie 7853913Sarchie#include <machine/in_cksum.h> /* XXX for in_cksum */ 7953913Sarchie 8053913Sarchie/* 8153913Sarchie * XXX This one should go in sys/mbuf.h. It is used to avoid that 8253913Sarchie * a firewall-generated packet loops forever through the firewall. 8353913Sarchie */ 8453913Sarchie#ifndef M_SKIP_FIREWALL 8553913Sarchie#define M_SKIP_FIREWALL 0x4000 8653913Sarchie#endif 8753913Sarchie 8853998Sarchie/* 8953913Sarchie * set_disable contains one bit per set value (0..31). 9053913Sarchie * If the bit is set, all rules with the corresponding set 9153913Sarchie * are disabled. Set 31 is reserved for the default rule 9253913Sarchie * and CANNOT be disabled. 9353998Sarchie */ 9453998Sarchiestatic u_int32_t set_disable; 9553998Sarchie 9653998Sarchiestatic int fw_verbose; 9753913Sarchiestatic int verbose_limit; 9853998Sarchie 9953998Sarchiestatic struct callout_handle ipfw_timeout_h; 10053913Sarchie#define IPFW_DEFAULT_RULE 65535 10153998Sarchie 10253998Sarchie/* 10353998Sarchie * list of rules for layer 3 10453998Sarchie */ 10553998Sarchiestatic struct ip_fw *layer3_chain; 10653998Sarchie 10753998SarchieMALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 10853913Sarchie 10953998Sarchiestatic int fw_debug = 1; 11053998Sarchiestatic int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 11153913Sarchie 11253998Sarchie#ifdef SYSCTL_NODE 11353998SarchieSYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 11453998SarchieSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, enable, 11553998Sarchie CTLFLAG_RW | CTLFLAG_SECURE3, 11653998Sarchie &fw_enable, 0, "Enable ipfw"); 11753998SarchieSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 11853998Sarchie &autoinc_step, 0, "Rule number autincrement step"); 11953998SarchieSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 12053998Sarchie CTLFLAG_RW | CTLFLAG_SECURE3, 12153998Sarchie &fw_one_pass, 0, 12253998Sarchie "Only do a single pass through ipfw when using dummynet(4)"); 12353998SarchieSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 12453998Sarchie &fw_debug, 0, "Enable printing of debug ip_fw statements"); 12553998SarchieSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 12653998Sarchie CTLFLAG_RW | CTLFLAG_SECURE3, 12753998Sarchie &fw_verbose, 0, "Log matches to ipfw rules"); 12853998SarchieSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 12953998Sarchie &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 13053998Sarchie 13153998Sarchie/* 13253998Sarchie * Description of dynamic rules. 13353998Sarchie * 13453998Sarchie * Dynamic rules are stored in lists accessed through a hash table 13553998Sarchie * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 13653998Sarchie * be modified through the sysctl variable dyn_buckets which is 13753998Sarchie * updated when the table becomes empty. 13853998Sarchie * 13953998Sarchie * XXX currently there is only one list, ipfw_dyn. 14053998Sarchie * 14153998Sarchie * When a packet is received, its address fields are first masked 14253998Sarchie * with the mask defined for the rule, then hashed, then matched 14353998Sarchie * against the entries in the corresponding list. 14453998Sarchie * Dynamic rules can be used for different purposes: 14553998Sarchie * + stateful rules; 14653998Sarchie * + enforcing limits on the number of sessions; 14753998Sarchie * + in-kernel NAT (not implemented yet) 14853998Sarchie * 14953998Sarchie * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 15053998Sarchie * measured in seconds and depending on the flags. 15153998Sarchie * 15253998Sarchie * The total number of dynamic rules is stored in dyn_count. 15353998Sarchie * The max number of dynamic rules is dyn_max. When we reach 15453998Sarchie * the maximum number of rules we do not create anymore. This is 15553998Sarchie * done to avoid consuming too much memory, but also too much 15653998Sarchie * time when searching on each packet (ideally, we should try instead 15753998Sarchie * to put a limit on the length of the list on each bucket...). 15853998Sarchie * 15953998Sarchie * Each dynamic rule holds a pointer to the parent ipfw rule so 16053998Sarchie * we know what action to perform. Dynamic rules are removed when 16153998Sarchie * the parent rule is deleted. XXX we should make them survive. 16253998Sarchie * 16353998Sarchie * There are some limitations with dynamic rules -- we do not 16453998Sarchie * obey the 'randomized match', and we do not do multiple 16553998Sarchie * passes through the firewall. XXX check the latter!!! 16653998Sarchie */ 16753998Sarchiestatic ipfw_dyn_rule **ipfw_dyn_v = NULL; 16853998Sarchiestatic u_int32_t dyn_buckets = 256; /* must be power of 2 */ 16953998Sarchiestatic u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 17053998Sarchie 17153998Sarchie/* 17253998Sarchie * Timeouts for various events in handing dynamic rules. 17353998Sarchie */ 17453998Sarchiestatic u_int32_t dyn_ack_lifetime = 300; 17553998Sarchiestatic u_int32_t dyn_syn_lifetime = 20; 17653998Sarchiestatic u_int32_t dyn_fin_lifetime = 1; 17753998Sarchiestatic u_int32_t dyn_rst_lifetime = 1; 17853998Sarchiestatic u_int32_t dyn_udp_lifetime = 10; 17953998Sarchiestatic u_int32_t dyn_short_lifetime = 5; 18053913Sarchie 18153998Sarchie/* 18253998Sarchie * Keepalives are sent if dyn_keepalive is set. They are sent every 18353998Sarchie * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 18453998Sarchie * seconds of lifetime of a rule. 18553913Sarchie * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 18653998Sarchie * than dyn_keepalive_period. 18753998Sarchie */ 18853998Sarchie 18953913Sarchiestatic u_int32_t dyn_keepalive_interval = 20; 19053913Sarchiestatic u_int32_t dyn_keepalive_period = 5; 19153913Sarchiestatic u_int32_t dyn_keepalive = 1; /* do send keepalives */ 19253913Sarchie 19353913Sarchiestatic u_int32_t static_count; /* # of static rules */ 19453913Sarchiestatic u_int32_t static_len; /* size in bytes of static rules */ 19553913Sarchiestatic u_int32_t dyn_count; /* # of dynamic rules */ 19653913Sarchiestatic u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 19753998Sarchie 19853998SarchieSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 19953913Sarchie &dyn_buckets, 0, "Number of dyn. buckets"); 20053913SarchieSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 20153913Sarchie &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 20253913SarchieSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 20353913Sarchie &dyn_count, 0, "Number of dyn. rules"); 20453913SarchieSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 20553913Sarchie &dyn_max, 0, "Max number of dyn. rules"); 20653913SarchieSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 20753913Sarchie &static_count, 0, "Number of static rules"); 20853913SarchieSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 20953913Sarchie &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 21053913SarchieSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 21153913Sarchie &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 21253913SarchieSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 21353913Sarchie &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 21453913SarchieSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 21553913Sarchie &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 21653913SarchieSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 21753913Sarchie &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 21853913SarchieSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 21953913Sarchie &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 22053913SarchieSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 22153913Sarchie &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 22253913Sarchie 22353913Sarchie#endif /* SYSCTL_NODE */ 22453913Sarchie 22553913Sarchie 22653913Sarchiestatic ip_fw_chk_t ipfw_chk; 22753913Sarchie 22853913Sarchieip_dn_ruledel_t *ip_dn_ruledel_ptr = NULL; /* hook into dummynet */ 22953913Sarchie 23053913Sarchie/* 23153913Sarchie * This macro maps an ip pointer into a layer3 header pointer of type T 23253913Sarchie */ 23353913Sarchie#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 23453913Sarchie 23553913Sarchiestatic __inline int 23653913Sarchieicmptype_match(struct ip *ip, ipfw_insn_u32 *cmd) 23753913Sarchie{ 23853913Sarchie int type = L3HDR(struct icmp,ip)->icmp_type; 23953913Sarchie 24053913Sarchie return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 24153913Sarchie} 24253913Sarchie 24353913Sarchie#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 24453913Sarchie (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 24553913Sarchie 24653913Sarchiestatic int 24753913Sarchieis_icmp_query(struct ip *ip) 24853913Sarchie{ 24953913Sarchie int type = L3HDR(struct icmp, ip)->icmp_type; 25053913Sarchie return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 25153913Sarchie} 25253913Sarchie#undef TT 25353913Sarchie 25453913Sarchie/* 25553913Sarchie * The following checks use two arrays of 8 or 16 bits to store the 25653913Sarchie * bits that we want set or clear, respectively. They are in the 25753913Sarchie * low and high half of cmd->arg1 or cmd->d[0]. 25853913Sarchie * 25953913Sarchie * We scan options and store the bits we find set. We succeed if 26053913Sarchie * 26153913Sarchie * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 26253913Sarchie * 26353913Sarchie * The code is sometimes optimized not to store additional variables. 26453913Sarchie */ 26553913Sarchie 26653913Sarchiestatic int 26753913Sarchieflags_match(ipfw_insn *cmd, u_int8_t bits) 26853998Sarchie{ 26953998Sarchie u_char want_clear; 27053998Sarchie bits = ~bits; 27153998Sarchie 27253998Sarchie if ( ((cmd->arg1 & 0xff) & bits) != 0) 27353998Sarchie return 0; /* some bits we want set were clear */ 27453998Sarchie want_clear = (cmd->arg1 >> 8) & 0xff; 27553998Sarchie if ( (want_clear & bits) != want_clear) 27653998Sarchie return 0; /* some bits we want clear were set */ 27753998Sarchie return 1; 27853913Sarchie} 27953913Sarchie 28053913Sarchiestatic int 28153913Sarchieipopts_match(struct ip *ip, ipfw_insn *cmd) 28253913Sarchie{ 28353913Sarchie int optlen, bits = 0; 28453913Sarchie u_char *cp = (u_char *)(ip + 1); 28553913Sarchie int x = (ip->ip_hl << 2) - sizeof (struct ip); 28653913Sarchie 28753913Sarchie for (; x > 0; x -= optlen, cp += optlen) { 28853913Sarchie int opt = cp[IPOPT_OPTVAL]; 28953913Sarchie 29053913Sarchie if (opt == IPOPT_EOL) 29153913Sarchie break; 29253913Sarchie if (opt == IPOPT_NOP) 29353913Sarchie optlen = 1; 29453998Sarchie else { 29553913Sarchie optlen = cp[IPOPT_OLEN]; 29653998Sarchie if (optlen <= 0 || optlen > x) 29753998Sarchie return 0; /* invalid or truncated */ 29853998Sarchie } 29953998Sarchie switch (opt) { 30053913Sarchie 30153913Sarchie default: 30253913Sarchie break; 30353913Sarchie 30453913Sarchie case IPOPT_LSRR: 30553913Sarchie bits |= IP_FW_IPOPT_LSRR; 30653913Sarchie break; 30753913Sarchie 30853913Sarchie case IPOPT_SSRR: 30953913Sarchie bits |= IP_FW_IPOPT_SSRR; 31053913Sarchie break; 31153913Sarchie 31253913Sarchie case IPOPT_RR: 31353913Sarchie bits |= IP_FW_IPOPT_RR; 31453913Sarchie break; 31553913Sarchie 31653913Sarchie case IPOPT_TS: 31753913Sarchie bits |= IP_FW_IPOPT_TS; 31853913Sarchie break; 31953998Sarchie } 32053913Sarchie } 32153998Sarchie return (flags_match(cmd, bits)); 32253998Sarchie} 32353998Sarchie 32453913Sarchiestatic int 32553913Sarchietcpopts_match(struct ip *ip, ipfw_insn *cmd) 32653913Sarchie{ 32753913Sarchie int optlen, bits = 0; 32853998Sarchie struct tcphdr *tcp = L3HDR(struct tcphdr,ip); 32953998Sarchie u_char *cp = (u_char *)(tcp + 1); 33053998Sarchie int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 33153998Sarchie 33253998Sarchie for (; x > 0; x -= optlen, cp += optlen) { 33353998Sarchie int opt = cp[0]; 33453998Sarchie if (opt == TCPOPT_EOL) 33553913Sarchie break; 33653913Sarchie if (opt == TCPOPT_NOP) 33753913Sarchie optlen = 1; 33853913Sarchie else { 33953913Sarchie optlen = cp[1]; 34053913Sarchie if (optlen <= 0) 34153913Sarchie break; 34253913Sarchie } 34353913Sarchie 34453913Sarchie switch (opt) { 34553913Sarchie 34653998Sarchie default: 34753913Sarchie break; 34853998Sarchie 34953998Sarchie case TCPOPT_MAXSEG: 35053998Sarchie bits |= IP_FW_TCPOPT_MSS; 35153913Sarchie break; 35253913Sarchie 35353913Sarchie case TCPOPT_WINDOW: 35453913Sarchie bits |= IP_FW_TCPOPT_WINDOW; 35553913Sarchie break; 35653998Sarchie 35753998Sarchie case TCPOPT_SACK_PERMITTED: 35853998Sarchie case TCPOPT_SACK: 35953998Sarchie bits |= IP_FW_TCPOPT_SACK; 36053998Sarchie break; 36153998Sarchie 36253998Sarchie case TCPOPT_TIMESTAMP: 36353998Sarchie bits |= IP_FW_TCPOPT_TS; 36453998Sarchie break; 36553913Sarchie 36653913Sarchie case TCPOPT_CC: 36753913Sarchie case TCPOPT_CCNEW: 36853913Sarchie case TCPOPT_CCECHO: 36953913Sarchie bits |= IP_FW_TCPOPT_CC; 37053913Sarchie break; 37153913Sarchie } 37253998Sarchie } 37353913Sarchie return (flags_match(cmd, bits)); 37453998Sarchie} 37553998Sarchie 37653913Sarchiestatic int 37753913Sarchieiface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 37853913Sarchie{ 37953913Sarchie if (ifp == NULL) /* no iface with this packet, match fails */ 38053913Sarchie return 0; 38153913Sarchie /* Check by name or by IP address */ 38253998Sarchie if (cmd->name[0] != '\0') { /* match by name */ 38353913Sarchie /* Check unit number (-1 is wildcard) */ 38453998Sarchie if (cmd->p.unit != -1 && cmd->p.unit != ifp->if_unit) 38553998Sarchie return(0); 38653998Sarchie /* Check name */ 38753913Sarchie if (!strncmp(ifp->if_name, cmd->name, IFNAMSIZ)) 38853913Sarchie return(1); 38953913Sarchie } else { 39053913Sarchie struct ifaddr *ia; 39153913Sarchie 39253913Sarchie TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 39353913Sarchie if (ia->ifa_addr == NULL) 39453913Sarchie continue; 39553913Sarchie if (ia->ifa_addr->sa_family != AF_INET) 39653913Sarchie continue; 39753998Sarchie if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 39853913Sarchie (ia->ifa_addr))->sin_addr.s_addr) 39953913Sarchie return(1); /* match */ 40053913Sarchie } 40153913Sarchie } 40253913Sarchie return(0); /* no match, fail ... */ 40353913Sarchie} 40453913Sarchie 40553913Sarchie/* 40653998Sarchie * The 'verrevpath' option checks that the interface that an IP packet 40753913Sarchie * arrives on is the same interface that traffic destined for the 40853913Sarchie * packet's source address would be routed out of. This is a measure 40953913Sarchie * to block forged packets. This is also commonly known as "anti-spoofing" 41053913Sarchie * or Unicast Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The 41153913Sarchie * name of the knob is purposely reminisent of the Cisco IOS command, 41253913Sarchie * 41353913Sarchie * ip verify unicast reverse-path 41453913Sarchie * 41553913Sarchie * which implements the same functionality. But note that syntax is 41653913Sarchie * misleading. The check may be performed on all IP packets whether unicast, 41753998Sarchie * multicast, or broadcast. 41853913Sarchie */ 41953913Sarchiestatic int 42053913Sarchieverify_rev_path(struct in_addr src, struct ifnet *ifp) 42153913Sarchie{ 42253913Sarchie static struct route ro; 42353913Sarchie struct sockaddr_in *dst; 42453913Sarchie 42553998Sarchie dst = (struct sockaddr_in *)&(ro.ro_dst); 42653913Sarchie 42753998Sarchie /* Check if we've cached the route from the previous call. */ 42853998Sarchie if (src.s_addr != dst->sin_addr.s_addr) { 42953998Sarchie ro.ro_rt = NULL; 43053998Sarchie 43153913Sarchie bzero(dst, sizeof(*dst)); 43253913Sarchie dst->sin_family = AF_INET; 43353913Sarchie dst->sin_len = sizeof(*dst); 43453913Sarchie dst->sin_addr = src; 43553913Sarchie 43653913Sarchie rtalloc_ign(&ro, RTF_CLONING|RTF_PRCLONING); 43753998Sarchie } 43853913Sarchie 43953998Sarchie if ((ro.ro_rt == NULL) || (ifp == NULL) || 44053998Sarchie (ro.ro_rt->rt_ifp->if_index != ifp->if_index)) 44153913Sarchie return 0; 44253913Sarchie 44353913Sarchie return 1; 44453913Sarchie} 44553913Sarchie 44653913Sarchie 44753913Sarchiestatic u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 44853913Sarchie 44953913Sarchie#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 45053913Sarchie#define SNP(buf) buf, sizeof(buf) 45153913Sarchie 45253913Sarchie/* 45353913Sarchie * We enter here when we have a rule with O_LOG. 45453913Sarchie * XXX this function alone takes about 2Kbytes of code! 45553913Sarchie */ 45653913Sarchiestatic void 45753913Sarchieipfw_log(struct ip_fw *f, u_int hlen, struct ether_header *eh, 45853913Sarchie struct mbuf *m, struct ifnet *oif) 45953913Sarchie{ 46053913Sarchie char *action; 46153913Sarchie int limit_reached = 0; 46253913Sarchie char action2[40], proto[48], fragment[28]; 46353913Sarchie 46453913Sarchie fragment[0] = '\0'; 46553913Sarchie proto[0] = '\0'; 46653913Sarchie 46753913Sarchie if (f == NULL) { /* bogus pkt */ 46853913Sarchie if (verbose_limit != 0 && norule_counter >= verbose_limit) 46953913Sarchie return; 47053913Sarchie norule_counter++; 47153913Sarchie if (norule_counter == verbose_limit) 47253913Sarchie limit_reached = verbose_limit; 47353913Sarchie action = "Refuse"; 47453913Sarchie } else { /* O_LOG is the first action, find the real one */ 47553913Sarchie ipfw_insn *cmd = ACTION_PTR(f); 47653913Sarchie ipfw_insn_log *l = (ipfw_insn_log *)cmd; 47753913Sarchie 47853913Sarchie if (l->max_log != 0 && l->log_left == 0) 47953913Sarchie return; 48053913Sarchie l->log_left--; 48153913Sarchie if (l->log_left == 0) 48253913Sarchie limit_reached = l->max_log; 48353913Sarchie cmd += F_LEN(cmd); /* point to first action */ 48453913Sarchie if (cmd->opcode == O_PROB) 48553913Sarchie cmd += F_LEN(cmd); 48653913Sarchie 48753913Sarchie action = action2; 48853913Sarchie switch (cmd->opcode) { 48953913Sarchie case O_DENY: 49053913Sarchie action = "Deny"; 49153913Sarchie break; 49253913Sarchie 49353913Sarchie case O_REJECT: 49453913Sarchie if (cmd->arg1==ICMP_REJECT_RST) 49553913Sarchie action = "Reset"; 49653913Sarchie else if (cmd->arg1==ICMP_UNREACH_HOST) 49753913Sarchie action = "Reject"; 49853913Sarchie else 49953913Sarchie snprintf(SNPARGS(action2, 0), "Unreach %d", 50053913Sarchie cmd->arg1); 50153913Sarchie break; 50253913Sarchie 50353913Sarchie case O_ACCEPT: 50453913Sarchie action = "Accept"; 50553913Sarchie break; 50653913Sarchie case O_COUNT: 50753913Sarchie action = "Count"; 508 break; 509 case O_DIVERT: 510 snprintf(SNPARGS(action2, 0), "Divert %d", 511 cmd->arg1); 512 break; 513 case O_TEE: 514 snprintf(SNPARGS(action2, 0), "Tee %d", 515 cmd->arg1); 516 break; 517 case O_SKIPTO: 518 snprintf(SNPARGS(action2, 0), "SkipTo %d", 519 cmd->arg1); 520 break; 521 case O_PIPE: 522 snprintf(SNPARGS(action2, 0), "Pipe %d", 523 cmd->arg1); 524 break; 525 case O_QUEUE: 526 snprintf(SNPARGS(action2, 0), "Queue %d", 527 cmd->arg1); 528 break; 529 case O_FORWARD_IP: { 530 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 531 int len; 532 533 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 534 inet_ntoa(sa->sa.sin_addr)); 535 if (sa->sa.sin_port) 536 snprintf(SNPARGS(action2, len), ":%d", 537 sa->sa.sin_port); 538 } 539 break; 540 default: 541 action = "UNKNOWN"; 542 break; 543 } 544 } 545 546 if (hlen == 0) { /* non-ip */ 547 snprintf(SNPARGS(proto, 0), "MAC"); 548 } else { 549 struct ip *ip = mtod(m, struct ip *); 550 /* these three are all aliases to the same thing */ 551 struct icmp *const icmp = L3HDR(struct icmp, ip); 552 struct tcphdr *const tcp = (struct tcphdr *)icmp; 553 struct udphdr *const udp = (struct udphdr *)icmp; 554 555 int ip_off, offset, ip_len; 556 557 int len; 558 559 if (eh != NULL) { /* layer 2 packets are as on the wire */ 560 ip_off = ntohs(ip->ip_off); 561 ip_len = ntohs(ip->ip_len); 562 } else { 563 ip_off = ip->ip_off; 564 ip_len = ip->ip_len; 565 } 566 offset = ip_off & IP_OFFMASK; 567 switch (ip->ip_p) { 568 case IPPROTO_TCP: 569 len = snprintf(SNPARGS(proto, 0), "TCP %s", 570 inet_ntoa(ip->ip_src)); 571 if (offset == 0) 572 snprintf(SNPARGS(proto, len), ":%d %s:%d", 573 ntohs(tcp->th_sport), 574 inet_ntoa(ip->ip_dst), 575 ntohs(tcp->th_dport)); 576 else 577 snprintf(SNPARGS(proto, len), " %s", 578 inet_ntoa(ip->ip_dst)); 579 break; 580 581 case IPPROTO_UDP: 582 len = snprintf(SNPARGS(proto, 0), "UDP %s", 583 inet_ntoa(ip->ip_src)); 584 if (offset == 0) 585 snprintf(SNPARGS(proto, len), ":%d %s:%d", 586 ntohs(udp->uh_sport), 587 inet_ntoa(ip->ip_dst), 588 ntohs(udp->uh_dport)); 589 else 590 snprintf(SNPARGS(proto, len), " %s", 591 inet_ntoa(ip->ip_dst)); 592 break; 593 594 case IPPROTO_ICMP: 595 if (offset == 0) 596 len = snprintf(SNPARGS(proto, 0), 597 "ICMP:%u.%u ", 598 icmp->icmp_type, icmp->icmp_code); 599 else 600 len = snprintf(SNPARGS(proto, 0), "ICMP "); 601 len += snprintf(SNPARGS(proto, len), "%s", 602 inet_ntoa(ip->ip_src)); 603 snprintf(SNPARGS(proto, len), " %s", 604 inet_ntoa(ip->ip_dst)); 605 break; 606 607 default: 608 len = snprintf(SNPARGS(proto, 0), "P:%d %s", ip->ip_p, 609 inet_ntoa(ip->ip_src)); 610 snprintf(SNPARGS(proto, len), " %s", 611 inet_ntoa(ip->ip_dst)); 612 break; 613 } 614 615 if (ip_off & (IP_MF | IP_OFFMASK)) 616 snprintf(SNPARGS(fragment, 0), " (frag %d:%d@%d%s)", 617 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 618 offset << 3, 619 (ip_off & IP_MF) ? "+" : ""); 620 } 621 if (oif || m->m_pkthdr.rcvif) 622 log(LOG_SECURITY | LOG_INFO, 623 "ipfw: %d %s %s %s via %s%d%s\n", 624 f ? f->rulenum : -1, 625 action, proto, oif ? "out" : "in", 626 oif ? oif->if_name : m->m_pkthdr.rcvif->if_name, 627 oif ? oif->if_unit : m->m_pkthdr.rcvif->if_unit, 628 fragment); 629 else 630 log(LOG_SECURITY | LOG_INFO, 631 "ipfw: %d %s %s [no if info]%s\n", 632 f ? f->rulenum : -1, 633 action, proto, fragment); 634 if (limit_reached) 635 log(LOG_SECURITY | LOG_NOTICE, 636 "ipfw: limit %d reached on entry %d\n", 637 limit_reached, f ? f->rulenum : -1); 638} 639 640/* 641 * IMPORTANT: the hash function for dynamic rules must be commutative 642 * in source and destination (ip,port), because rules are bidirectional 643 * and we want to find both in the same bucket. 644 */ 645static __inline int 646hash_packet(struct ipfw_flow_id *id) 647{ 648 u_int32_t i; 649 650 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 651 i &= (curr_dyn_buckets - 1); 652 return i; 653} 654 655/** 656 * unlink a dynamic rule from a chain. prev is a pointer to 657 * the previous one, q is a pointer to the rule to delete, 658 * head is a pointer to the head of the queue. 659 * Modifies q and potentially also head. 660 */ 661#define UNLINK_DYN_RULE(prev, head, q) { \ 662 ipfw_dyn_rule *old_q = q; \ 663 \ 664 /* remove a refcount to the parent */ \ 665 if (q->dyn_type == O_LIMIT) \ 666 q->parent->count--; \ 667 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 668 (q->id.src_ip), (q->id.src_port), \ 669 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 670 if (prev != NULL) \ 671 prev->next = q = q->next; \ 672 else \ 673 head = q = q->next; \ 674 dyn_count--; \ 675 free(old_q, M_IPFW); } 676 677#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 678 679/** 680 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 681 * 682 * If keep_me == NULL, rules are deleted even if not expired, 683 * otherwise only expired rules are removed. 684 * 685 * The value of the second parameter is also used to point to identify 686 * a rule we absolutely do not want to remove (e.g. because we are 687 * holding a reference to it -- this is the case with O_LIMIT_PARENT 688 * rules). The pointer is only used for comparison, so any non-null 689 * value will do. 690 */ 691static void 692remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 693{ 694 static u_int32_t last_remove = 0; 695 696#define FORCE (keep_me == NULL) 697 698 ipfw_dyn_rule *prev, *q; 699 int i, pass = 0, max_pass = 0; 700 701 if (ipfw_dyn_v == NULL || dyn_count == 0) 702 return; 703 /* do not expire more than once per second, it is useless */ 704 if (!FORCE && last_remove == time_second) 705 return; 706 last_remove = time_second; 707 708 /* 709 * because O_LIMIT refer to parent rules, during the first pass only 710 * remove child and mark any pending LIMIT_PARENT, and remove 711 * them in a second pass. 712 */ 713next_pass: 714 for (i = 0 ; i < curr_dyn_buckets ; i++) { 715 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 716 /* 717 * Logic can become complex here, so we split tests. 718 */ 719 if (q == keep_me) 720 goto next; 721 if (rule != NULL && rule != q->rule) 722 goto next; /* not the one we are looking for */ 723 if (q->dyn_type == O_LIMIT_PARENT) { 724 /* 725 * handle parent in the second pass, 726 * record we need one. 727 */ 728 max_pass = 1; 729 if (pass == 0) 730 goto next; 731 if (FORCE && q->count != 0 ) { 732 /* XXX should not happen! */ 733 printf("ipfw: OUCH! cannot remove rule," 734 " count %d\n", q->count); 735 } 736 } else { 737 if (!FORCE && 738 !TIME_LEQ( q->expire, time_second )) 739 goto next; 740 } 741 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 742 continue; 743next: 744 prev=q; 745 q=q->next; 746 } 747 } 748 if (pass++ < max_pass) 749 goto next_pass; 750} 751 752 753/** 754 * lookup a dynamic rule. 755 */ 756static ipfw_dyn_rule * 757lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 758 struct tcphdr *tcp) 759{ 760 /* 761 * stateful ipfw extensions. 762 * Lookup into dynamic session queue 763 */ 764#define MATCH_REVERSE 0 765#define MATCH_FORWARD 1 766#define MATCH_NONE 2 767#define MATCH_UNKNOWN 3 768 int i, dir = MATCH_NONE; 769 ipfw_dyn_rule *prev, *q=NULL; 770 771 if (ipfw_dyn_v == NULL) 772 goto done; /* not found */ 773 i = hash_packet( pkt ); 774 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 775 if (q->dyn_type == O_LIMIT_PARENT) 776 goto next; 777 if (TIME_LEQ( q->expire, time_second)) { /* expire entry */ 778 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 779 continue; 780 } 781 if ( pkt->proto == q->id.proto) { 782 if (pkt->src_ip == q->id.src_ip && 783 pkt->dst_ip == q->id.dst_ip && 784 pkt->src_port == q->id.src_port && 785 pkt->dst_port == q->id.dst_port ) { 786 dir = MATCH_FORWARD; 787 break; 788 } 789 if (pkt->src_ip == q->id.dst_ip && 790 pkt->dst_ip == q->id.src_ip && 791 pkt->src_port == q->id.dst_port && 792 pkt->dst_port == q->id.src_port ) { 793 dir = MATCH_REVERSE; 794 break; 795 } 796 } 797next: 798 prev = q; 799 q = q->next; 800 } 801 if (q == NULL) 802 goto done; /* q = NULL, not found */ 803 804 if ( prev != NULL) { /* found and not in front */ 805 prev->next = q->next; 806 q->next = ipfw_dyn_v[i]; 807 ipfw_dyn_v[i] = q; 808 } 809 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 810 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 811 812#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 813#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 814 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 815 switch (q->state) { 816 case TH_SYN: /* opening */ 817 q->expire = time_second + dyn_syn_lifetime; 818 break; 819 820 case BOTH_SYN: /* move to established */ 821 case BOTH_SYN | TH_FIN : /* one side tries to close */ 822 case BOTH_SYN | (TH_FIN << 8) : 823 if (tcp) { 824#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 825 u_int32_t ack = ntohl(tcp->th_ack); 826 if (dir == MATCH_FORWARD) { 827 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 828 q->ack_fwd = ack; 829 else { /* ignore out-of-sequence */ 830 break; 831 } 832 } else { 833 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 834 q->ack_rev = ack; 835 else { /* ignore out-of-sequence */ 836 break; 837 } 838 } 839 } 840 q->expire = time_second + dyn_ack_lifetime; 841 break; 842 843 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 844 if (dyn_fin_lifetime >= dyn_keepalive_period) 845 dyn_fin_lifetime = dyn_keepalive_period - 1; 846 q->expire = time_second + dyn_fin_lifetime; 847 break; 848 849 default: 850#if 0 851 /* 852 * reset or some invalid combination, but can also 853 * occur if we use keep-state the wrong way. 854 */ 855 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 856 printf("invalid state: 0x%x\n", q->state); 857#endif 858 if (dyn_rst_lifetime >= dyn_keepalive_period) 859 dyn_rst_lifetime = dyn_keepalive_period - 1; 860 q->expire = time_second + dyn_rst_lifetime; 861 break; 862 } 863 } else if (pkt->proto == IPPROTO_UDP) { 864 q->expire = time_second + dyn_udp_lifetime; 865 } else { 866 /* other protocols */ 867 q->expire = time_second + dyn_short_lifetime; 868 } 869done: 870 if (match_direction) 871 *match_direction = dir; 872 return q; 873} 874 875static void 876realloc_dynamic_table(void) 877{ 878 /* 879 * Try reallocation, make sure we have a power of 2 and do 880 * not allow more than 64k entries. In case of overflow, 881 * default to 1024. 882 */ 883 884 if (dyn_buckets > 65536) 885 dyn_buckets = 1024; 886 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 887 dyn_buckets = curr_dyn_buckets; /* reset */ 888 return; 889 } 890 curr_dyn_buckets = dyn_buckets; 891 if (ipfw_dyn_v != NULL) 892 free(ipfw_dyn_v, M_IPFW); 893 for (;;) { 894 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 895 M_IPFW, M_NOWAIT | M_ZERO); 896 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 897 break; 898 curr_dyn_buckets /= 2; 899 } 900} 901 902/** 903 * Install state of type 'type' for a dynamic session. 904 * The hash table contains two type of rules: 905 * - regular rules (O_KEEP_STATE) 906 * - rules for sessions with limited number of sess per user 907 * (O_LIMIT). When they are created, the parent is 908 * increased by 1, and decreased on delete. In this case, 909 * the third parameter is the parent rule and not the chain. 910 * - "parent" rules for the above (O_LIMIT_PARENT). 911 */ 912static ipfw_dyn_rule * 913add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 914{ 915 ipfw_dyn_rule *r; 916 int i; 917 918 if (ipfw_dyn_v == NULL || 919 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 920 realloc_dynamic_table(); 921 if (ipfw_dyn_v == NULL) 922 return NULL; /* failed ! */ 923 } 924 i = hash_packet(id); 925 926 r = malloc(sizeof *r, M_IPFW, M_NOWAIT | M_ZERO); 927 if (r == NULL) { 928 printf ("ipfw: sorry cannot allocate state\n"); 929 return NULL; 930 } 931 932 /* increase refcount on parent, and set pointer */ 933 if (dyn_type == O_LIMIT) { 934 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 935 if ( parent->dyn_type != O_LIMIT_PARENT) 936 panic("invalid parent"); 937 parent->count++; 938 r->parent = parent; 939 rule = parent->rule; 940 } 941 942 r->id = *id; 943 r->expire = time_second + dyn_syn_lifetime; 944 r->rule = rule; 945 r->dyn_type = dyn_type; 946 r->pcnt = r->bcnt = 0; 947 r->count = 0; 948 949 r->bucket = i; 950 r->next = ipfw_dyn_v[i]; 951 ipfw_dyn_v[i] = r; 952 dyn_count++; 953 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 954 dyn_type, 955 (r->id.src_ip), (r->id.src_port), 956 (r->id.dst_ip), (r->id.dst_port), 957 dyn_count ); ) 958 return r; 959} 960 961/** 962 * lookup dynamic parent rule using pkt and rule as search keys. 963 * If the lookup fails, then install one. 964 */ 965static ipfw_dyn_rule * 966lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 967{ 968 ipfw_dyn_rule *q; 969 int i; 970 971 if (ipfw_dyn_v) { 972 i = hash_packet( pkt ); 973 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 974 if (q->dyn_type == O_LIMIT_PARENT && 975 rule== q->rule && 976 pkt->proto == q->id.proto && 977 pkt->src_ip == q->id.src_ip && 978 pkt->dst_ip == q->id.dst_ip && 979 pkt->src_port == q->id.src_port && 980 pkt->dst_port == q->id.dst_port) { 981 q->expire = time_second + dyn_short_lifetime; 982 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 983 return q; 984 } 985 } 986 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 987} 988 989/** 990 * Install dynamic state for rule type cmd->o.opcode 991 * 992 * Returns 1 (failure) if state is not installed because of errors or because 993 * session limitations are enforced. 994 */ 995static int 996install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 997 struct ip_fw_args *args) 998{ 999 static int last_log; 1000 1001 ipfw_dyn_rule *q; 1002 1003 DEB(printf("ipfw: install state type %d 0x%08x %u -> 0x%08x %u\n", 1004 cmd->o.opcode, 1005 (args->f_id.src_ip), (args->f_id.src_port), 1006 (args->f_id.dst_ip), (args->f_id.dst_port) );) 1007 1008 q = lookup_dyn_rule(&args->f_id, NULL, NULL); 1009 1010 if (q != NULL) { /* should never occur */ 1011 if (last_log != time_second) { 1012 last_log = time_second; 1013 printf("ipfw: install_state: entry already present, done\n"); 1014 } 1015 return 0; 1016 } 1017 1018 if (dyn_count >= dyn_max) 1019 /* 1020 * Run out of slots, try to remove any expired rule. 1021 */ 1022 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1023 1024 if (dyn_count >= dyn_max) { 1025 if (last_log != time_second) { 1026 last_log = time_second; 1027 printf("ipfw: install_state: Too many dynamic rules\n"); 1028 } 1029 return 1; /* cannot install, notify caller */ 1030 } 1031 1032 switch (cmd->o.opcode) { 1033 case O_KEEP_STATE: /* bidir rule */ 1034 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1035 break; 1036 1037 case O_LIMIT: /* limit number of sessions */ 1038 { 1039 u_int16_t limit_mask = cmd->limit_mask; 1040 struct ipfw_flow_id id; 1041 ipfw_dyn_rule *parent; 1042 1043 DEB(printf("ipfw: installing dyn-limit rule %d\n", 1044 cmd->conn_limit);) 1045 1046 id.dst_ip = id.src_ip = 0; 1047 id.dst_port = id.src_port = 0; 1048 id.proto = args->f_id.proto; 1049 1050 if (limit_mask & DYN_SRC_ADDR) 1051 id.src_ip = args->f_id.src_ip; 1052 if (limit_mask & DYN_DST_ADDR) 1053 id.dst_ip = args->f_id.dst_ip; 1054 if (limit_mask & DYN_SRC_PORT) 1055 id.src_port = args->f_id.src_port; 1056 if (limit_mask & DYN_DST_PORT) 1057 id.dst_port = args->f_id.dst_port; 1058 parent = lookup_dyn_parent(&id, rule); 1059 if (parent == NULL) { 1060 printf("ipfw: add parent failed\n"); 1061 return 1; 1062 } 1063 if (parent->count >= cmd->conn_limit) { 1064 /* 1065 * See if we can remove some expired rule. 1066 */ 1067 remove_dyn_rule(rule, parent); 1068 if (parent->count >= cmd->conn_limit) { 1069 if (fw_verbose && last_log != time_second) { 1070 last_log = time_second; 1071 log(LOG_SECURITY | LOG_DEBUG, 1072 "drop session, too many entries\n"); 1073 } 1074 return 1; 1075 } 1076 } 1077 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1078 } 1079 break; 1080 default: 1081 printf("ipfw: unknown dynamic rule type %u\n", cmd->o.opcode); 1082 return 1; 1083 } 1084 lookup_dyn_rule(&args->f_id, NULL, NULL); /* XXX just set lifetime */ 1085 return 0; 1086} 1087 1088/* 1089 * Transmit a TCP packet, containing either a RST or a keepalive. 1090 * When flags & TH_RST, we are sending a RST packet, because of a 1091 * "reset" action matched the packet. 1092 * Otherwise we are sending a keepalive, and flags & TH_ 1093 */ 1094static void 1095send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags) 1096{ 1097 struct mbuf *m; 1098 struct ip *ip; 1099 struct tcphdr *tcp; 1100 struct route sro; /* fake route */ 1101 1102 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1103 if (m == 0) 1104 return; 1105 m->m_pkthdr.rcvif = (struct ifnet *)0; 1106 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1107 m->m_data += max_linkhdr; 1108 1109 ip = mtod(m, struct ip *); 1110 bzero(ip, m->m_len); 1111 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1112 ip->ip_p = IPPROTO_TCP; 1113 tcp->th_off = 5; 1114 /* 1115 * Assume we are sending a RST (or a keepalive in the reverse 1116 * direction), swap src and destination addresses and ports. 1117 */ 1118 ip->ip_src.s_addr = htonl(id->dst_ip); 1119 ip->ip_dst.s_addr = htonl(id->src_ip); 1120 tcp->th_sport = htons(id->dst_port); 1121 tcp->th_dport = htons(id->src_port); 1122 if (flags & TH_RST) { /* we are sending a RST */ 1123 if (flags & TH_ACK) { 1124 tcp->th_seq = htonl(ack); 1125 tcp->th_ack = htonl(0); 1126 tcp->th_flags = TH_RST; 1127 } else { 1128 if (flags & TH_SYN) 1129 seq++; 1130 tcp->th_seq = htonl(0); 1131 tcp->th_ack = htonl(seq); 1132 tcp->th_flags = TH_RST | TH_ACK; 1133 } 1134 } else { 1135 /* 1136 * We are sending a keepalive. flags & TH_SYN determines 1137 * the direction, forward if set, reverse if clear. 1138 * NOTE: seq and ack are always assumed to be correct 1139 * as set by the caller. This may be confusing... 1140 */ 1141 if (flags & TH_SYN) { 1142 /* 1143 * we have to rewrite the correct addresses! 1144 */ 1145 ip->ip_dst.s_addr = htonl(id->dst_ip); 1146 ip->ip_src.s_addr = htonl(id->src_ip); 1147 tcp->th_dport = htons(id->dst_port); 1148 tcp->th_sport = htons(id->src_port); 1149 } 1150 tcp->th_seq = htonl(seq); 1151 tcp->th_ack = htonl(ack); 1152 tcp->th_flags = TH_ACK; 1153 } 1154 /* 1155 * set ip_len to the payload size so we can compute 1156 * the tcp checksum on the pseudoheader 1157 * XXX check this, could save a couple of words ? 1158 */ 1159 ip->ip_len = htons(sizeof(struct tcphdr)); 1160 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1161 /* 1162 * now fill fields left out earlier 1163 */ 1164 ip->ip_ttl = ip_defttl; 1165 ip->ip_len = m->m_pkthdr.len; 1166 bzero (&sro, sizeof (sro)); 1167 ip_rtaddr(ip->ip_dst, &sro); 1168 m->m_flags |= M_SKIP_FIREWALL; 1169 ip_output(m, NULL, &sro, 0, NULL, NULL); 1170 if (sro.ro_rt) 1171 RTFREE(sro.ro_rt); 1172} 1173 1174/* 1175 * sends a reject message, consuming the mbuf passed as an argument. 1176 */ 1177static void 1178send_reject(struct ip_fw_args *args, int code, int offset, int ip_len) 1179{ 1180 1181 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1182 /* We need the IP header in host order for icmp_error(). */ 1183 if (args->eh != NULL) { 1184 struct ip *ip = mtod(args->m, struct ip *); 1185 ip->ip_len = ntohs(ip->ip_len); 1186 ip->ip_off = ntohs(ip->ip_off); 1187 } 1188 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1189 } else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) { 1190 struct tcphdr *const tcp = 1191 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1192 if ( (tcp->th_flags & TH_RST) == 0) 1193 send_pkt(&(args->f_id), ntohl(tcp->th_seq), 1194 ntohl(tcp->th_ack), 1195 tcp->th_flags | TH_RST); 1196 m_freem(args->m); 1197 } else 1198 m_freem(args->m); 1199 args->m = NULL; 1200} 1201 1202/** 1203 * 1204 * Given an ip_fw *, lookup_next_rule will return a pointer 1205 * to the next rule, which can be either the jump 1206 * target (for skipto instructions) or the next one in the list (in 1207 * all other cases including a missing jump target). 1208 * The result is also written in the "next_rule" field of the rule. 1209 * Backward jumps are not allowed, so start looking from the next 1210 * rule... 1211 * 1212 * This never returns NULL -- in case we do not have an exact match, 1213 * the next rule is returned. When the ruleset is changed, 1214 * pointers are flushed so we are always correct. 1215 */ 1216 1217static struct ip_fw * 1218lookup_next_rule(struct ip_fw *me) 1219{ 1220 struct ip_fw *rule = NULL; 1221 ipfw_insn *cmd; 1222 1223 /* look for action, in case it is a skipto */ 1224 cmd = ACTION_PTR(me); 1225 if (cmd->opcode == O_LOG) 1226 cmd += F_LEN(cmd); 1227 if ( cmd->opcode == O_SKIPTO ) 1228 for (rule = me->next; rule ; rule = rule->next) 1229 if (rule->rulenum >= cmd->arg1) 1230 break; 1231 if (rule == NULL) /* failure or not a skipto */ 1232 rule = me->next; 1233 me->next_rule = rule; 1234 return rule; 1235} 1236 1237/* 1238 * The main check routine for the firewall. 1239 * 1240 * All arguments are in args so we can modify them and return them 1241 * back to the caller. 1242 * 1243 * Parameters: 1244 * 1245 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 1246 * Starts with the IP header. 1247 * args->eh (in) Mac header if present, or NULL for layer3 packet. 1248 * args->oif Outgoing interface, or NULL if packet is incoming. 1249 * The incoming interface is in the mbuf. (in) 1250 * args->divert_rule (in/out) 1251 * Skip up to the first rule past this rule number; 1252 * upon return, non-zero port number for divert or tee. 1253 * 1254 * args->rule Pointer to the last matching rule (in/out) 1255 * args->next_hop Socket we are forwarding to (out). 1256 * args->f_id Addresses grabbed from the packet (out) 1257 * 1258 * Return value: 1259 * 1260 * IP_FW_PORT_DENY_FLAG the packet must be dropped. 1261 * 0 The packet is to be accepted and routed normally OR 1262 * the packet was denied/rejected and has been dropped; 1263 * in the latter case, *m is equal to NULL upon return. 1264 * port Divert the packet to port, with these caveats: 1265 * 1266 * - If IP_FW_PORT_TEE_FLAG is set, tee the packet instead 1267 * of diverting it (ie, 'ipfw tee'). 1268 * 1269 * - If IP_FW_PORT_DYNT_FLAG is set, interpret the lower 1270 * 16 bits as a dummynet pipe number instead of diverting 1271 */ 1272 1273static int 1274ipfw_chk(struct ip_fw_args *args) 1275{ 1276 /* 1277 * Local variables hold state during the processing of a packet. 1278 * 1279 * IMPORTANT NOTE: to speed up the processing of rules, there 1280 * are some assumption on the values of the variables, which 1281 * are documented here. Should you change them, please check 1282 * the implementation of the various instructions to make sure 1283 * that they still work. 1284 * 1285 * args->eh The MAC header. It is non-null for a layer2 1286 * packet, it is NULL for a layer-3 packet. 1287 * 1288 * m | args->m Pointer to the mbuf, as received from the caller. 1289 * It may change if ipfw_chk() does an m_pullup, or if it 1290 * consumes the packet because it calls send_reject(). 1291 * XXX This has to change, so that ipfw_chk() never modifies 1292 * or consumes the buffer. 1293 * ip is simply an alias of the value of m, and it is kept 1294 * in sync with it (the packet is supposed to start with 1295 * the ip header). 1296 */ 1297 struct mbuf *m = args->m; 1298 struct ip *ip = mtod(m, struct ip *); 1299 1300 /* 1301 * oif | args->oif If NULL, ipfw_chk has been called on the 1302 * inbound path (ether_input, bdg_forward, ip_input). 1303 * If non-NULL, ipfw_chk has been called on the outbound path 1304 * (ether_output, ip_output). 1305 */ 1306 struct ifnet *oif = args->oif; 1307 1308 struct ip_fw *f = NULL; /* matching rule */ 1309 int retval = 0; 1310 1311 /* 1312 * hlen The length of the IPv4 header. 1313 * hlen >0 means we have an IPv4 packet. 1314 */ 1315 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 1316 1317 /* 1318 * offset The offset of a fragment. offset != 0 means that 1319 * we have a fragment at this offset of an IPv4 packet. 1320 * offset == 0 means that (if this is an IPv4 packet) 1321 * this is the first or only fragment. 1322 */ 1323 u_short offset = 0; 1324 1325 /* 1326 * Local copies of addresses. They are only valid if we have 1327 * an IP packet. 1328 * 1329 * proto The protocol. Set to 0 for non-ip packets, 1330 * or to the protocol read from the packet otherwise. 1331 * proto != 0 means that we have an IPv4 packet. 1332 * 1333 * src_port, dst_port port numbers, in HOST format. Only 1334 * valid for TCP and UDP packets. 1335 * 1336 * src_ip, dst_ip ip addresses, in NETWORK format. 1337 * Only valid for IPv4 packets. 1338 */ 1339 u_int8_t proto; 1340 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 1341 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 1342 u_int16_t ip_len=0; 1343 int pktlen; 1344 int dyn_dir = MATCH_UNKNOWN; 1345 ipfw_dyn_rule *q = NULL; 1346 1347 if (m->m_flags & M_SKIP_FIREWALL) 1348 return 0; /* accept */ 1349 /* 1350 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 1351 * MATCH_NONE when checked and not matched (q = NULL), 1352 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 1353 */ 1354 1355 pktlen = m->m_pkthdr.len; 1356 if (args->eh == NULL || /* layer 3 packet */ 1357 ( m->m_pkthdr.len >= sizeof(struct ip) && 1358 ntohs(args->eh->ether_type) == ETHERTYPE_IP)) 1359 hlen = ip->ip_hl << 2; 1360 1361 /* 1362 * Collect parameters into local variables for faster matching. 1363 */ 1364 if (hlen == 0) { /* do not grab addresses for non-ip pkts */ 1365 proto = args->f_id.proto = 0; /* mark f_id invalid */ 1366 goto after_ip_checks; 1367 } 1368 1369 proto = args->f_id.proto = ip->ip_p; 1370 src_ip = ip->ip_src; 1371 dst_ip = ip->ip_dst; 1372 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 1373 offset = ntohs(ip->ip_off) & IP_OFFMASK; 1374 ip_len = ntohs(ip->ip_len); 1375 } else { 1376 offset = ip->ip_off & IP_OFFMASK; 1377 ip_len = ip->ip_len; 1378 } 1379 pktlen = ip_len < pktlen ? ip_len : pktlen; 1380 1381#define PULLUP_TO(len) \ 1382 do { \ 1383 if ((m)->m_len < (len)) { \ 1384 args->m = m = m_pullup(m, (len)); \ 1385 if (m == 0) \ 1386 goto pullup_failed; \ 1387 ip = mtod(m, struct ip *); \ 1388 } \ 1389 } while (0) 1390 1391 if (offset == 0) { 1392 switch (proto) { 1393 case IPPROTO_TCP: 1394 { 1395 struct tcphdr *tcp; 1396 1397 PULLUP_TO(hlen + sizeof(struct tcphdr)); 1398 tcp = L3HDR(struct tcphdr, ip); 1399 dst_port = tcp->th_dport; 1400 src_port = tcp->th_sport; 1401 args->f_id.flags = tcp->th_flags; 1402 } 1403 break; 1404 1405 case IPPROTO_UDP: 1406 { 1407 struct udphdr *udp; 1408 1409 PULLUP_TO(hlen + sizeof(struct udphdr)); 1410 udp = L3HDR(struct udphdr, ip); 1411 dst_port = udp->uh_dport; 1412 src_port = udp->uh_sport; 1413 } 1414 break; 1415 1416 case IPPROTO_ICMP: 1417 PULLUP_TO(hlen + 4); /* type, code and checksum. */ 1418 args->f_id.flags = L3HDR(struct icmp, ip)->icmp_type; 1419 break; 1420 1421 default: 1422 break; 1423 } 1424#undef PULLUP_TO 1425 } 1426 1427 args->f_id.src_ip = ntohl(src_ip.s_addr); 1428 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 1429 args->f_id.src_port = src_port = ntohs(src_port); 1430 args->f_id.dst_port = dst_port = ntohs(dst_port); 1431 1432after_ip_checks: 1433 if (args->rule) { 1434 /* 1435 * Packet has already been tagged. Look for the next rule 1436 * to restart processing. 1437 * 1438 * If fw_one_pass != 0 then just accept it. 1439 * XXX should not happen here, but optimized out in 1440 * the caller. 1441 */ 1442 if (fw_one_pass) 1443 return 0; 1444 1445 f = args->rule->next_rule; 1446 if (f == NULL) 1447 f = lookup_next_rule(args->rule); 1448 } else { 1449 /* 1450 * Find the starting rule. It can be either the first 1451 * one, or the one after divert_rule if asked so. 1452 */ 1453 int skipto = args->divert_rule; 1454 1455 f = layer3_chain; 1456 if (args->eh == NULL && skipto != 0) { 1457 if (skipto >= IPFW_DEFAULT_RULE) 1458 return(IP_FW_PORT_DENY_FLAG); /* invalid */ 1459 while (f && f->rulenum <= skipto) 1460 f = f->next; 1461 if (f == NULL) /* drop packet */ 1462 return(IP_FW_PORT_DENY_FLAG); 1463 } 1464 } 1465 args->divert_rule = 0; /* reset to avoid confusion later */ 1466 1467 /* 1468 * Now scan the rules, and parse microinstructions for each rule. 1469 */ 1470 for (; f; f = f->next) { 1471 int l, cmdlen; 1472 ipfw_insn *cmd; 1473 int skip_or; /* skip rest of OR block */ 1474 1475again: 1476 if (set_disable & (1 << f->set) ) 1477 continue; 1478 1479 skip_or = 0; 1480 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 1481 l -= cmdlen, cmd += cmdlen) { 1482 int match; 1483 1484 /* 1485 * check_body is a jump target used when we find a 1486 * CHECK_STATE, and need to jump to the body of 1487 * the target rule. 1488 */ 1489 1490check_body: 1491 cmdlen = F_LEN(cmd); 1492 /* 1493 * An OR block (insn_1 || .. || insn_n) has the 1494 * F_OR bit set in all but the last instruction. 1495 * The first match will set "skip_or", and cause 1496 * the following instructions to be skipped until 1497 * past the one with the F_OR bit clear. 1498 */ 1499 if (skip_or) { /* skip this instruction */ 1500 if ((cmd->len & F_OR) == 0) 1501 skip_or = 0; /* next one is good */ 1502 continue; 1503 } 1504 match = 0; /* set to 1 if we succeed */ 1505 1506 switch (cmd->opcode) { 1507 /* 1508 * The first set of opcodes compares the packet's 1509 * fields with some pattern, setting 'match' if a 1510 * match is found. At the end of the loop there is 1511 * logic to deal with F_NOT and F_OR flags associated 1512 * with the opcode. 1513 */ 1514 case O_NOP: 1515 match = 1; 1516 break; 1517 1518 case O_FORWARD_MAC: 1519 printf("ipfw: opcode %d unimplemented\n", 1520 cmd->opcode); 1521 break; 1522 1523 case O_GID: 1524 case O_UID: 1525 /* 1526 * We only check offset == 0 && proto != 0, 1527 * as this ensures that we have an IPv4 1528 * packet with the ports info. 1529 */ 1530 if (offset!=0) 1531 break; 1532 { 1533 struct inpcbinfo *pi; 1534 int wildcard; 1535 struct inpcb *pcb; 1536 1537 if (proto == IPPROTO_TCP) { 1538 wildcard = 0; 1539 pi = &tcbinfo; 1540 } else if (proto == IPPROTO_UDP) { 1541 wildcard = 1; 1542 pi = &udbinfo; 1543 } else 1544 break; 1545 1546 pcb = (oif) ? 1547 in_pcblookup_hash(pi, 1548 dst_ip, htons(dst_port), 1549 src_ip, htons(src_port), 1550 wildcard, oif) : 1551 in_pcblookup_hash(pi, 1552 src_ip, htons(src_port), 1553 dst_ip, htons(dst_port), 1554 wildcard, NULL); 1555 1556 if (pcb == NULL || pcb->inp_socket == NULL) 1557 break; 1558#if __FreeBSD_version < 500034 1559#define socheckuid(a,b) ((a)->so_cred->cr_uid != (b)) 1560#endif 1561 if (cmd->opcode == O_UID) { 1562 match = 1563 !socheckuid(pcb->inp_socket, 1564 (uid_t)((ipfw_insn_u32 *)cmd)->d[0]); 1565 } else { 1566 match = groupmember( 1567 (uid_t)((ipfw_insn_u32 *)cmd)->d[0], 1568 pcb->inp_socket->so_cred); 1569 } 1570 } 1571 break; 1572 1573 case O_RECV: 1574 match = iface_match(m->m_pkthdr.rcvif, 1575 (ipfw_insn_if *)cmd); 1576 break; 1577 1578 case O_XMIT: 1579 match = iface_match(oif, (ipfw_insn_if *)cmd); 1580 break; 1581 1582 case O_VIA: 1583 match = iface_match(oif ? oif : 1584 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 1585 break; 1586 1587 case O_MACADDR2: 1588 if (args->eh != NULL) { /* have MAC header */ 1589 u_int32_t *want = (u_int32_t *) 1590 ((ipfw_insn_mac *)cmd)->addr; 1591 u_int32_t *mask = (u_int32_t *) 1592 ((ipfw_insn_mac *)cmd)->mask; 1593 u_int32_t *hdr = (u_int32_t *)args->eh; 1594 1595 match = 1596 ( want[0] == (hdr[0] & mask[0]) && 1597 want[1] == (hdr[1] & mask[1]) && 1598 want[2] == (hdr[2] & mask[2]) ); 1599 } 1600 break; 1601 1602 case O_MAC_TYPE: 1603 if (args->eh != NULL) { 1604 u_int16_t t = 1605 ntohs(args->eh->ether_type); 1606 u_int16_t *p = 1607 ((ipfw_insn_u16 *)cmd)->ports; 1608 int i; 1609 1610 for (i = cmdlen - 1; !match && i>0; 1611 i--, p += 2) 1612 match = (t>=p[0] && t<=p[1]); 1613 } 1614 break; 1615 1616 case O_FRAG: 1617 match = (hlen > 0 && offset != 0); 1618 break; 1619 1620 case O_IN: /* "out" is "not in" */ 1621 match = (oif == NULL); 1622 break; 1623 1624 case O_LAYER2: 1625 match = (args->eh != NULL); 1626 break; 1627 1628 case O_PROTO: 1629 /* 1630 * We do not allow an arg of 0 so the 1631 * check of "proto" only suffices. 1632 */ 1633 match = (proto == cmd->arg1); 1634 break; 1635 1636 case O_IP_SRC: 1637 match = (hlen > 0 && 1638 ((ipfw_insn_ip *)cmd)->addr.s_addr == 1639 src_ip.s_addr); 1640 break; 1641 1642 case O_IP_SRC_MASK: 1643 match = (hlen > 0 && 1644 ((ipfw_insn_ip *)cmd)->addr.s_addr == 1645 (src_ip.s_addr & 1646 ((ipfw_insn_ip *)cmd)->mask.s_addr)); 1647 break; 1648 1649 case O_IP_SRC_ME: 1650 if (hlen > 0) { 1651 struct ifnet *tif; 1652 1653 INADDR_TO_IFP(src_ip, tif); 1654 match = (tif != NULL); 1655 } 1656 break; 1657 1658 case O_IP_DST_SET: 1659 case O_IP_SRC_SET: 1660 if (hlen > 0) { 1661 u_int32_t *d = (u_int32_t *)(cmd+1); 1662 u_int32_t addr = 1663 cmd->opcode == O_IP_DST_SET ? 1664 args->f_id.dst_ip : 1665 args->f_id.src_ip; 1666 1667 if (addr < d[0]) 1668 break; 1669 addr -= d[0]; /* subtract base */ 1670 match = (addr < cmd->arg1) && 1671 ( d[ 1 + (addr>>5)] & 1672 (1<<(addr & 0x1f)) ); 1673 } 1674 break; 1675 1676 case O_IP_DST: 1677 match = (hlen > 0 && 1678 ((ipfw_insn_ip *)cmd)->addr.s_addr == 1679 dst_ip.s_addr); 1680 break; 1681 1682 case O_IP_DST_MASK: 1683 match = (hlen > 0) && 1684 (((ipfw_insn_ip *)cmd)->addr.s_addr == 1685 (dst_ip.s_addr & 1686 ((ipfw_insn_ip *)cmd)->mask.s_addr)); 1687 break; 1688 1689 case O_IP_DST_ME: 1690 if (hlen > 0) { 1691 struct ifnet *tif; 1692 1693 INADDR_TO_IFP(dst_ip, tif); 1694 match = (tif != NULL); 1695 } 1696 break; 1697 1698 case O_IP_SRCPORT: 1699 case O_IP_DSTPORT: 1700 /* 1701 * offset == 0 && proto != 0 is enough 1702 * to guarantee that we have an IPv4 1703 * packet with port info. 1704 */ 1705 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 1706 && offset == 0) { 1707 u_int16_t x = 1708 (cmd->opcode == O_IP_SRCPORT) ? 1709 src_port : dst_port ; 1710 u_int16_t *p = 1711 ((ipfw_insn_u16 *)cmd)->ports; 1712 int i; 1713 1714 for (i = cmdlen - 1; !match && i>0; 1715 i--, p += 2) 1716 match = (x>=p[0] && x<=p[1]); 1717 } 1718 break; 1719 1720 case O_ICMPTYPE: 1721 match = (offset == 0 && proto==IPPROTO_ICMP && 1722 icmptype_match(ip, (ipfw_insn_u32 *)cmd) ); 1723 break; 1724 1725 case O_IPOPT: 1726 match = (hlen > 0 && ipopts_match(ip, cmd) ); 1727 break; 1728 1729 case O_IPVER: 1730 match = (hlen > 0 && cmd->arg1 == ip->ip_v); 1731 break; 1732 1733 case O_IPTTL: 1734 match = (hlen > 0 && cmd->arg1 == ip->ip_ttl); 1735 break; 1736 1737 case O_IPID: 1738 match = (hlen > 0 && 1739 cmd->arg1 == ntohs(ip->ip_id)); 1740 break; 1741 1742 case O_IPLEN: 1743 match = (hlen > 0 && cmd->arg1 == ip_len); 1744 break; 1745 1746 case O_IPPRECEDENCE: 1747 match = (hlen > 0 && 1748 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 1749 break; 1750 1751 case O_IPTOS: 1752 match = (hlen > 0 && 1753 flags_match(cmd, ip->ip_tos)); 1754 break; 1755 1756 case O_TCPFLAGS: 1757 match = (proto == IPPROTO_TCP && offset == 0 && 1758 flags_match(cmd, 1759 L3HDR(struct tcphdr,ip)->th_flags)); 1760 break; 1761 1762 case O_TCPOPTS: 1763 match = (proto == IPPROTO_TCP && offset == 0 && 1764 tcpopts_match(ip, cmd)); 1765 break; 1766 1767 case O_TCPSEQ: 1768 match = (proto == IPPROTO_TCP && offset == 0 && 1769 ((ipfw_insn_u32 *)cmd)->d[0] == 1770 L3HDR(struct tcphdr,ip)->th_seq); 1771 break; 1772 1773 case O_TCPACK: 1774 match = (proto == IPPROTO_TCP && offset == 0 && 1775 ((ipfw_insn_u32 *)cmd)->d[0] == 1776 L3HDR(struct tcphdr,ip)->th_ack); 1777 break; 1778 1779 case O_TCPWIN: 1780 match = (proto == IPPROTO_TCP && offset == 0 && 1781 cmd->arg1 == 1782 L3HDR(struct tcphdr,ip)->th_win); 1783 break; 1784 1785 case O_ESTAB: 1786 /* reject packets which have SYN only */ 1787 /* XXX should i also check for TH_ACK ? */ 1788 match = (proto == IPPROTO_TCP && offset == 0 && 1789 (L3HDR(struct tcphdr,ip)->th_flags & 1790 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 1791 break; 1792 1793 case O_LOG: 1794 if (fw_verbose) 1795 ipfw_log(f, hlen, args->eh, m, oif); 1796 match = 1; 1797 break; 1798 1799 case O_PROB: 1800 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 1801 break; 1802 1803 case O_VERREVPATH: 1804 /* Outgoing packets automatically pass/match */ 1805 match = ((oif != NULL) || 1806 (m->m_pkthdr.rcvif == NULL) || 1807 verify_rev_path(src_ip, m->m_pkthdr.rcvif)); 1808 break; 1809 1810 /* 1811 * The second set of opcodes represents 'actions', 1812 * i.e. the terminal part of a rule once the packet 1813 * matches all previous patterns. 1814 * Typically there is only one action for each rule, 1815 * and the opcode is stored at the end of the rule 1816 * (but there are exceptions -- see below). 1817 * 1818 * In general, here we set retval and terminate the 1819 * outer loop (would be a 'break 3' in some language, 1820 * but we need to do a 'goto done'). 1821 * 1822 * Exceptions: 1823 * O_COUNT and O_SKIPTO actions: 1824 * instead of terminating, we jump to the next rule 1825 * ('goto next_rule', equivalent to a 'break 2'), 1826 * or to the SKIPTO target ('goto again' after 1827 * having set f, cmd and l), respectively. 1828 * 1829 * O_LIMIT and O_KEEP_STATE: these opcodes are 1830 * not real 'actions', and are stored right 1831 * before the 'action' part of the rule. 1832 * These opcodes try to install an entry in the 1833 * state tables; if successful, we continue with 1834 * the next opcode (match=1; break;), otherwise 1835 * the packet * must be dropped 1836 * ('goto done' after setting retval); 1837 * 1838 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 1839 * cause a lookup of the state table, and a jump 1840 * to the 'action' part of the parent rule 1841 * ('goto check_body') if an entry is found, or 1842 * (CHECK_STATE only) a jump to the next rule if 1843 * the entry is not found ('goto next_rule'). 1844 * The result of the lookup is cached to make 1845 * further instances of these opcodes are 1846 * effectively NOPs. 1847 */ 1848 case O_LIMIT: 1849 case O_KEEP_STATE: 1850 if (install_state(f, 1851 (ipfw_insn_limit *)cmd, args)) { 1852 retval = IP_FW_PORT_DENY_FLAG; 1853 goto done; /* error/limit violation */ 1854 } 1855 match = 1; 1856 break; 1857 1858 case O_PROBE_STATE: 1859 case O_CHECK_STATE: 1860 /* 1861 * dynamic rules are checked at the first 1862 * keep-state or check-state occurrence, 1863 * with the result being stored in dyn_dir. 1864 * The compiler introduces a PROBE_STATE 1865 * instruction for us when we have a 1866 * KEEP_STATE (because PROBE_STATE needs 1867 * to be run first). 1868 */ 1869 if (dyn_dir == MATCH_UNKNOWN && 1870 (q = lookup_dyn_rule(&args->f_id, 1871 &dyn_dir, proto == IPPROTO_TCP ? 1872 L3HDR(struct tcphdr, ip) : NULL)) 1873 != NULL) { 1874 /* 1875 * Found dynamic entry, update stats 1876 * and jump to the 'action' part of 1877 * the parent rule. 1878 */ 1879 q->pcnt++; 1880 q->bcnt += pktlen; 1881 f = q->rule; 1882 cmd = ACTION_PTR(f); 1883 l = f->cmd_len - f->act_ofs; 1884 goto check_body; 1885 } 1886 /* 1887 * Dynamic entry not found. If CHECK_STATE, 1888 * skip to next rule, if PROBE_STATE just 1889 * ignore and continue with next opcode. 1890 */ 1891 if (cmd->opcode == O_CHECK_STATE) 1892 goto next_rule; 1893 match = 1; 1894 break; 1895 1896 case O_ACCEPT: 1897 retval = 0; /* accept */ 1898 goto done; 1899 1900 case O_PIPE: 1901 case O_QUEUE: 1902 args->rule = f; /* report matching rule */ 1903 retval = cmd->arg1 | IP_FW_PORT_DYNT_FLAG; 1904 goto done; 1905 1906 case O_DIVERT: 1907 case O_TEE: 1908 if (args->eh) /* not on layer 2 */ 1909 break; 1910 args->divert_rule = f->rulenum; 1911 retval = (cmd->opcode == O_DIVERT) ? 1912 cmd->arg1 : 1913 cmd->arg1 | IP_FW_PORT_TEE_FLAG; 1914 goto done; 1915 1916 case O_COUNT: 1917 case O_SKIPTO: 1918 f->pcnt++; /* update stats */ 1919 f->bcnt += pktlen; 1920 f->timestamp = time_second; 1921 if (cmd->opcode == O_COUNT) 1922 goto next_rule; 1923 /* handle skipto */ 1924 if (f->next_rule == NULL) 1925 lookup_next_rule(f); 1926 f = f->next_rule; 1927 goto again; 1928 1929 case O_REJECT: 1930 /* 1931 * Drop the packet and send a reject notice 1932 * if the packet is not ICMP (or is an ICMP 1933 * query), and it is not multicast/broadcast. 1934 */ 1935 if (hlen > 0 && 1936 (proto != IPPROTO_ICMP || 1937 is_icmp_query(ip)) && 1938 !(m->m_flags & (M_BCAST|M_MCAST)) && 1939 !IN_MULTICAST(dst_ip.s_addr)) { 1940 send_reject(args, cmd->arg1, 1941 offset,ip_len); 1942 m = args->m; 1943 } 1944 /* FALLTHROUGH */ 1945 case O_DENY: 1946 retval = IP_FW_PORT_DENY_FLAG; 1947 goto done; 1948 1949 case O_FORWARD_IP: 1950 if (args->eh) /* not valid on layer2 pkts */ 1951 break; 1952 if (!q || dyn_dir == MATCH_FORWARD) 1953 args->next_hop = 1954 &((ipfw_insn_sa *)cmd)->sa; 1955 retval = 0; 1956 goto done; 1957 1958 default: 1959 panic("-- unknown opcode %d\n", cmd->opcode); 1960 } /* end of switch() on opcodes */ 1961 1962 if (cmd->len & F_NOT) 1963 match = !match; 1964 1965 if (match) { 1966 if (cmd->len & F_OR) 1967 skip_or = 1; 1968 } else { 1969 if (!(cmd->len & F_OR)) /* not an OR block, */ 1970 break; /* try next rule */ 1971 } 1972 1973 } /* end of inner for, scan opcodes */ 1974 1975next_rule:; /* try next rule */ 1976 1977 } /* end of outer for, scan rules */ 1978 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 1979 return(IP_FW_PORT_DENY_FLAG); 1980 1981done: 1982 /* Update statistics */ 1983 f->pcnt++; 1984 f->bcnt += pktlen; 1985 f->timestamp = time_second; 1986 return retval; 1987 1988pullup_failed: 1989 if (fw_verbose) 1990 printf("ipfw: pullup failed\n"); 1991 return(IP_FW_PORT_DENY_FLAG); 1992} 1993 1994/* 1995 * When a rule is added/deleted, clear the next_rule pointers in all rules. 1996 * These will be reconstructed on the fly as packets are matched. 1997 * Must be called at splimp(). 1998 */ 1999static void 2000flush_rule_ptrs(void) 2001{ 2002 struct ip_fw *rule; 2003 2004 for (rule = layer3_chain; rule; rule = rule->next) 2005 rule->next_rule = NULL; 2006} 2007 2008/* 2009 * When pipes/queues are deleted, clear the "pipe_ptr" pointer to a given 2010 * pipe/queue, or to all of them (match == NULL). 2011 * Must be called at splimp(). 2012 */ 2013void 2014flush_pipe_ptrs(struct dn_flow_set *match) 2015{ 2016 struct ip_fw *rule; 2017 2018 for (rule = layer3_chain; rule; rule = rule->next) { 2019 ipfw_insn_pipe *cmd = (ipfw_insn_pipe *)ACTION_PTR(rule); 2020 2021 if (cmd->o.opcode != O_PIPE && cmd->o.opcode != O_QUEUE) 2022 continue; 2023 /* 2024 * XXX Use bcmp/bzero to handle pipe_ptr to overcome 2025 * possible alignment problems on 64-bit architectures. 2026 * This code is seldom used so we do not worry too 2027 * much about efficiency. 2028 */ 2029 if (match == NULL || 2030 !bcmp(&cmd->pipe_ptr, &match, sizeof(match)) ) 2031 bzero(&cmd->pipe_ptr, sizeof(cmd->pipe_ptr)); 2032 } 2033} 2034 2035/* 2036 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 2037 * possibly create a rule number and add the rule to the list. 2038 * Update the rule_number in the input struct so the caller knows it as well. 2039 */ 2040static int 2041add_rule(struct ip_fw **head, struct ip_fw *input_rule) 2042{ 2043 struct ip_fw *rule, *f, *prev; 2044 int s; 2045 int l = RULESIZE(input_rule); 2046 2047 if (*head == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 2048 return (EINVAL); 2049 2050 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 2051 if (rule == NULL) 2052 return (ENOSPC); 2053 2054 bcopy(input_rule, rule, l); 2055 2056 rule->next = NULL; 2057 rule->next_rule = NULL; 2058 2059 rule->pcnt = 0; 2060 rule->bcnt = 0; 2061 rule->timestamp = 0; 2062 2063 s = splimp(); 2064 2065 if (*head == NULL) { /* default rule */ 2066 *head = rule; 2067 goto done; 2068 } 2069 2070 /* 2071 * If rulenum is 0, find highest numbered rule before the 2072 * default rule, and add autoinc_step 2073 */ 2074 if (autoinc_step < 1) 2075 autoinc_step = 1; 2076 else if (autoinc_step > 1000) 2077 autoinc_step = 1000; 2078 if (rule->rulenum == 0) { 2079 /* 2080 * locate the highest numbered rule before default 2081 */ 2082 for (f = *head; f; f = f->next) { 2083 if (f->rulenum == IPFW_DEFAULT_RULE) 2084 break; 2085 rule->rulenum = f->rulenum; 2086 } 2087 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 2088 rule->rulenum += autoinc_step; 2089 input_rule->rulenum = rule->rulenum; 2090 } 2091 2092 /* 2093 * Now insert the new rule in the right place in the sorted list. 2094 */ 2095 for (prev = NULL, f = *head; f; prev = f, f = f->next) { 2096 if (f->rulenum > rule->rulenum) { /* found the location */ 2097 if (prev) { 2098 rule->next = f; 2099 prev->next = rule; 2100 } else { /* head insert */ 2101 rule->next = *head; 2102 *head = rule; 2103 } 2104 break; 2105 } 2106 } 2107 flush_rule_ptrs(); 2108done: 2109 static_count++; 2110 static_len += l; 2111 splx(s); 2112 DEB(printf("ipfw: installed rule %d, static count now %d\n", 2113 rule->rulenum, static_count);) 2114 return (0); 2115} 2116 2117/** 2118 * Free storage associated with a static rule (including derived 2119 * dynamic rules). 2120 * The caller is in charge of clearing rule pointers to avoid 2121 * dangling pointers. 2122 * @return a pointer to the next entry. 2123 * Arguments are not checked, so they better be correct. 2124 * Must be called at splimp(). 2125 */ 2126static struct ip_fw * 2127delete_rule(struct ip_fw **head, struct ip_fw *prev, struct ip_fw *rule) 2128{ 2129 struct ip_fw *n; 2130 int l = RULESIZE(rule); 2131 2132 n = rule->next; 2133 remove_dyn_rule(rule, NULL /* force removal */); 2134 if (prev == NULL) 2135 *head = n; 2136 else 2137 prev->next = n; 2138 static_count--; 2139 static_len -= l; 2140 2141 if (DUMMYNET_LOADED) 2142 ip_dn_ruledel_ptr(rule); 2143 free(rule, M_IPFW); 2144 return n; 2145} 2146 2147/* 2148 * Deletes all rules from a chain (including the default rule 2149 * if the second argument is set). 2150 * Must be called at splimp(). 2151 */ 2152static void 2153free_chain(struct ip_fw **chain, int kill_default) 2154{ 2155 struct ip_fw *rule; 2156 2157 flush_rule_ptrs(); /* more efficient to do outside the loop */ 2158 2159 while ( (rule = *chain) != NULL && 2160 (kill_default || rule->rulenum != IPFW_DEFAULT_RULE) ) 2161 delete_rule(chain, NULL, rule); 2162} 2163 2164/** 2165 * Remove all rules with given number, and also do set manipulation. 2166 * 2167 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 2168 * the next 8 bits are the new set, the top 8 bits are the command: 2169 * 2170 * 0 delete rules with given number 2171 * 1 delete rules with given set number 2172 * 2 move rules with given number to new set 2173 * 3 move rules with given set number to new set 2174 * 4 swap sets with given numbers 2175 */ 2176static int 2177del_entry(struct ip_fw **chain, u_int32_t arg) 2178{ 2179 struct ip_fw *prev, *rule; 2180 int s; 2181 u_int16_t rulenum; 2182 u_int8_t cmd, new_set; 2183 2184 rulenum = arg & 0xffff; 2185 cmd = (arg >> 24) & 0xff; 2186 new_set = (arg >> 16) & 0xff; 2187 2188 if (cmd > 4) 2189 return EINVAL; 2190 if (new_set > 30) 2191 return EINVAL; 2192 if (cmd == 0 || cmd == 2) { 2193 if (rulenum == IPFW_DEFAULT_RULE) 2194 return EINVAL; 2195 } else { 2196 if (rulenum > 30) 2197 return EINVAL; 2198 } 2199 2200 switch (cmd) { 2201 case 0: /* delete rules with given number */ 2202 /* 2203 * locate first rule to delete 2204 */ 2205 for (prev = NULL, rule = *chain; 2206 rule && rule->rulenum < rulenum; 2207 prev = rule, rule = rule->next) 2208 ; 2209 if (rule->rulenum != rulenum) 2210 return EINVAL; 2211 2212 s = splimp(); /* no access to rules while removing */ 2213 /* 2214 * flush pointers outside the loop, then delete all matching 2215 * rules. prev remains the same throughout the cycle. 2216 */ 2217 flush_rule_ptrs(); 2218 while (rule && rule->rulenum == rulenum) 2219 rule = delete_rule(chain, prev, rule); 2220 splx(s); 2221 break; 2222 2223 case 1: /* delete all rules with given set number */ 2224 s = splimp(); 2225 flush_rule_ptrs(); 2226 for (prev = NULL, rule = *chain; rule ; ) 2227 if (rule->set == rulenum) 2228 rule = delete_rule(chain, prev, rule); 2229 else { 2230 prev = rule; 2231 rule = rule->next; 2232 } 2233 splx(s); 2234 break; 2235 2236 case 2: /* move rules with given number to new set */ 2237 s = splimp(); 2238 for (rule = *chain; rule ; rule = rule->next) 2239 if (rule->rulenum == rulenum) 2240 rule->set = new_set; 2241 splx(s); 2242 break; 2243 2244 case 3: /* move rules with given set number to new set */ 2245 s = splimp(); 2246 for (rule = *chain; rule ; rule = rule->next) 2247 if (rule->set == rulenum) 2248 rule->set = new_set; 2249 splx(s); 2250 break; 2251 2252 case 4: /* swap two sets */ 2253 s = splimp(); 2254 for (rule = *chain; rule ; rule = rule->next) 2255 if (rule->set == rulenum) 2256 rule->set = new_set; 2257 else if (rule->set == new_set) 2258 rule->set = rulenum; 2259 splx(s); 2260 break; 2261 } 2262 return 0; 2263} 2264 2265/* 2266 * Clear counters for a specific rule. 2267 */ 2268static void 2269clear_counters(struct ip_fw *rule, int log_only) 2270{ 2271 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 2272 2273 if (log_only == 0) { 2274 rule->bcnt = rule->pcnt = 0; 2275 rule->timestamp = 0; 2276 } 2277 if (l->o.opcode == O_LOG) 2278 l->log_left = l->max_log; 2279} 2280 2281/** 2282 * Reset some or all counters on firewall rules. 2283 * @arg frwl is null to clear all entries, or contains a specific 2284 * rule number. 2285 * @arg log_only is 1 if we only want to reset logs, zero otherwise. 2286 */ 2287static int 2288zero_entry(int rulenum, int log_only) 2289{ 2290 struct ip_fw *rule; 2291 int s; 2292 char *msg; 2293 2294 if (rulenum == 0) { 2295 s = splimp(); 2296 norule_counter = 0; 2297 for (rule = layer3_chain; rule; rule = rule->next) 2298 clear_counters(rule, log_only); 2299 splx(s); 2300 msg = log_only ? "ipfw: All logging counts reset.\n" : 2301 "ipfw: Accounting cleared.\n"; 2302 } else { 2303 int cleared = 0; 2304 /* 2305 * We can have multiple rules with the same number, so we 2306 * need to clear them all. 2307 */ 2308 for (rule = layer3_chain; rule; rule = rule->next) 2309 if (rule->rulenum == rulenum) { 2310 s = splimp(); 2311 while (rule && rule->rulenum == rulenum) { 2312 clear_counters(rule, log_only); 2313 rule = rule->next; 2314 } 2315 splx(s); 2316 cleared = 1; 2317 break; 2318 } 2319 if (!cleared) /* we did not find any matching rules */ 2320 return (EINVAL); 2321 msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 2322 "ipfw: Entry %d cleared.\n"; 2323 } 2324 if (fw_verbose) 2325 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 2326 return (0); 2327} 2328 2329/* 2330 * Check validity of the structure before insert. 2331 * Fortunately rules are simple, so this mostly need to check rule sizes. 2332 */ 2333static int 2334check_ipfw_struct(struct ip_fw *rule, int size) 2335{ 2336 int l, cmdlen = 0; 2337 int have_action=0; 2338 ipfw_insn *cmd; 2339 2340 if (size < sizeof(*rule)) { 2341 printf("ipfw: rule too short\n"); 2342 return (EINVAL); 2343 } 2344 /* first, check for valid size */ 2345 l = RULESIZE(rule); 2346 if (l != size) { 2347 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 2348 return (EINVAL); 2349 } 2350 /* 2351 * Now go for the individual checks. Very simple ones, basically only 2352 * instruction sizes. 2353 */ 2354 for (l = rule->cmd_len, cmd = rule->cmd ; 2355 l > 0 ; l -= cmdlen, cmd += cmdlen) { 2356 cmdlen = F_LEN(cmd); 2357 if (cmdlen > l) { 2358 printf("ipfw: opcode %d size truncated\n", 2359 cmd->opcode); 2360 return EINVAL; 2361 } 2362 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 2363 switch (cmd->opcode) { 2364 case O_NOP: 2365 case O_PROBE_STATE: 2366 case O_KEEP_STATE: 2367 case O_PROTO: 2368 case O_IP_SRC_ME: 2369 case O_IP_DST_ME: 2370 case O_LAYER2: 2371 case O_IN: 2372 case O_FRAG: 2373 case O_IPOPT: 2374 case O_IPLEN: 2375 case O_IPID: 2376 case O_IPTOS: 2377 case O_IPPRECEDENCE: 2378 case O_IPTTL: 2379 case O_IPVER: 2380 case O_TCPWIN: 2381 case O_TCPFLAGS: 2382 case O_TCPOPTS: 2383 case O_ESTAB: 2384 case O_VERREVPATH: 2385 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 2386 goto bad_size; 2387 break; 2388 2389 case O_UID: 2390 case O_GID: 2391 case O_IP_SRC: 2392 case O_IP_DST: 2393 case O_TCPSEQ: 2394 case O_TCPACK: 2395 case O_PROB: 2396 case O_ICMPTYPE: 2397 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 2398 goto bad_size; 2399 break; 2400 2401 case O_LIMIT: 2402 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 2403 goto bad_size; 2404 break; 2405 2406 case O_LOG: 2407 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 2408 goto bad_size; 2409 2410 ((ipfw_insn_log *)cmd)->log_left = 2411 ((ipfw_insn_log *)cmd)->max_log; 2412 2413 break; 2414 2415 case O_IP_SRC_MASK: 2416 case O_IP_DST_MASK: 2417 if (cmdlen != F_INSN_SIZE(ipfw_insn_ip)) 2418 goto bad_size; 2419 if (((ipfw_insn_ip *)cmd)->mask.s_addr == 0) { 2420 printf("ipfw: opcode %d, useless rule\n", 2421 cmd->opcode); 2422 return EINVAL; 2423 } 2424 break; 2425 2426 case O_IP_SRC_SET: 2427 case O_IP_DST_SET: 2428 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 2429 printf("ipfw: invalid set size %d\n", 2430 cmd->arg1); 2431 return EINVAL; 2432 } 2433 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 2434 (cmd->arg1+31)/32 ) 2435 goto bad_size; 2436 break; 2437 2438 case O_MACADDR2: 2439 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 2440 goto bad_size; 2441 break; 2442 2443 case O_MAC_TYPE: 2444 case O_IP_SRCPORT: 2445 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 2446 if (cmdlen < 2 || cmdlen > 31) 2447 goto bad_size; 2448 break; 2449 2450 case O_RECV: 2451 case O_XMIT: 2452 case O_VIA: 2453 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 2454 goto bad_size; 2455 break; 2456 2457 case O_PIPE: 2458 case O_QUEUE: 2459 if (cmdlen != F_INSN_SIZE(ipfw_insn_pipe)) 2460 goto bad_size; 2461 goto check_action; 2462 2463 case O_FORWARD_IP: 2464 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 2465 goto bad_size; 2466 goto check_action; 2467 2468 case O_FORWARD_MAC: /* XXX not implemented yet */ 2469 case O_CHECK_STATE: 2470 case O_COUNT: 2471 case O_ACCEPT: 2472 case O_DENY: 2473 case O_REJECT: 2474 case O_SKIPTO: 2475 case O_DIVERT: 2476 case O_TEE: 2477 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 2478 goto bad_size; 2479check_action: 2480 if (have_action) { 2481 printf("ipfw: opcode %d, multiple actions" 2482 " not allowed\n", 2483 cmd->opcode); 2484 return EINVAL; 2485 } 2486 have_action = 1; 2487 if (l != cmdlen) { 2488 printf("ipfw: opcode %d, action must be" 2489 " last opcode\n", 2490 cmd->opcode); 2491 return EINVAL; 2492 } 2493 break; 2494 default: 2495 printf("ipfw: opcode %d, unknown opcode\n", 2496 cmd->opcode); 2497 return EINVAL; 2498 } 2499 } 2500 if (have_action == 0) { 2501 printf("ipfw: missing action\n"); 2502 return EINVAL; 2503 } 2504 return 0; 2505 2506bad_size: 2507 printf("ipfw: opcode %d size %d wrong\n", 2508 cmd->opcode, cmdlen); 2509 return EINVAL; 2510} 2511 2512 2513/** 2514 * {set|get}sockopt parser. 2515 */ 2516static int 2517ipfw_ctl(struct sockopt *sopt) 2518{ 2519 int error, s, rulenum; 2520 size_t size; 2521 struct ip_fw *bp , *buf, *rule; 2522 2523 static u_int32_t rule_buf[255]; /* we copy the data here */ 2524 2525 /* 2526 * Disallow modifications in really-really secure mode, but still allow 2527 * the logging counters to be reset. 2528 */ 2529 if (sopt->sopt_name == IP_FW_ADD || 2530 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 2531#if __FreeBSD_version >= 500034 2532 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 2533 if (error) 2534 return (error); 2535#else /* FreeBSD 4.x */ 2536 if (securelevel >= 3) 2537 return (EPERM); 2538#endif 2539 } 2540 2541 error = 0; 2542 2543 switch (sopt->sopt_name) { 2544 case IP_FW_GET: 2545 /* 2546 * pass up a copy of the current rules. Static rules 2547 * come first (the last of which has number IPFW_DEFAULT_RULE), 2548 * followed by a possibly empty list of dynamic rule. 2549 * The last dynamic rule has NULL in the "next" field. 2550 */ 2551 s = splimp(); 2552 size = static_len; /* size of static rules */ 2553 if (ipfw_dyn_v) /* add size of dyn.rules */ 2554 size += (dyn_count * sizeof(ipfw_dyn_rule)); 2555 2556 /* 2557 * XXX todo: if the user passes a short length just to know 2558 * how much room is needed, do not bother filling up the 2559 * buffer, just jump to the sooptcopyout. 2560 */ 2561 buf = malloc(size, M_TEMP, M_WAITOK); 2562 if (buf == 0) { 2563 splx(s); 2564 error = ENOBUFS; 2565 break; 2566 } 2567 2568 bp = buf; 2569 for (rule = layer3_chain; rule ; rule = rule->next) { 2570 int i = RULESIZE(rule); 2571 bcopy(rule, bp, i); 2572 bcopy(&set_disable, &(bp->next_rule), 2573 sizeof(set_disable)); 2574 bp = (struct ip_fw *)((char *)bp + i); 2575 } 2576 if (ipfw_dyn_v) { 2577 int i; 2578 ipfw_dyn_rule *p, *dst, *last = NULL; 2579 2580 dst = (ipfw_dyn_rule *)bp; 2581 for (i = 0 ; i < curr_dyn_buckets ; i++ ) 2582 for ( p = ipfw_dyn_v[i] ; p != NULL ; 2583 p = p->next, dst++ ) { 2584 bcopy(p, dst, sizeof *p); 2585 bcopy(&(p->rule->rulenum), &(dst->rule), 2586 sizeof(p->rule->rulenum)); 2587 /* 2588 * store a non-null value in "next". 2589 * The userland code will interpret a 2590 * NULL here as a marker 2591 * for the last dynamic rule. 2592 */ 2593 bcopy(&dst, &dst->next, sizeof(dst)); 2594 last = dst ; 2595 dst->expire = 2596 TIME_LEQ(dst->expire, time_second) ? 2597 0 : dst->expire - time_second ; 2598 } 2599 if (last != NULL) /* mark last dynamic rule */ 2600 bzero(&last->next, sizeof(last)); 2601 } 2602 splx(s); 2603 2604 error = sooptcopyout(sopt, buf, size); 2605 free(buf, M_TEMP); 2606 break; 2607 2608 case IP_FW_FLUSH: 2609 /* 2610 * Normally we cannot release the lock on each iteration. 2611 * We could do it here only because we start from the head all 2612 * the times so there is no risk of missing some entries. 2613 * On the other hand, the risk is that we end up with 2614 * a very inconsistent ruleset, so better keep the lock 2615 * around the whole cycle. 2616 * 2617 * XXX this code can be improved by resetting the head of 2618 * the list to point to the default rule, and then freeing 2619 * the old list without the need for a lock. 2620 */ 2621 2622 s = splimp(); 2623 free_chain(&layer3_chain, 0 /* keep default rule */); 2624 splx(s); 2625 break; 2626 2627 case IP_FW_ADD: 2628 rule = (struct ip_fw *)rule_buf; /* XXX do a malloc */ 2629 error = sooptcopyin(sopt, rule, sizeof(rule_buf), 2630 sizeof(struct ip_fw) ); 2631 size = sopt->sopt_valsize; 2632 if (error || (error = check_ipfw_struct(rule, size))) 2633 break; 2634 2635 error = add_rule(&layer3_chain, rule); 2636 size = RULESIZE(rule); 2637 if (!error && sopt->sopt_dir == SOPT_GET) 2638 error = sooptcopyout(sopt, rule, size); 2639 break; 2640 2641 case IP_FW_DEL: 2642 /* 2643 * IP_FW_DEL is used for deleting single rules or sets, 2644 * and (ab)used to atomically manipulate sets. Argument size 2645 * is used to distinguish between the two: 2646 * sizeof(u_int32_t) 2647 * delete single rule or set of rules, 2648 * or reassign rules (or sets) to a different set. 2649 * 2*sizeof(u_int32_t) 2650 * atomic disable/enable sets. 2651 * first u_int32_t contains sets to be disabled, 2652 * second u_int32_t contains sets to be enabled. 2653 */ 2654 error = sooptcopyin(sopt, rule_buf, 2655 2*sizeof(u_int32_t), sizeof(u_int32_t)); 2656 if (error) 2657 break; 2658 size = sopt->sopt_valsize; 2659 if (size == sizeof(u_int32_t)) /* delete or reassign */ 2660 error = del_entry(&layer3_chain, rule_buf[0]); 2661 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 2662 set_disable = 2663 (set_disable | rule_buf[0]) & ~rule_buf[1] & 2664 ~(1<<31); /* set 31 always enabled */ 2665 else 2666 error = EINVAL; 2667 break; 2668 2669 case IP_FW_ZERO: 2670 case IP_FW_RESETLOG: /* argument is an int, the rule number */ 2671 rulenum=0; 2672 2673 if (sopt->sopt_val != 0) { 2674 error = sooptcopyin(sopt, &rulenum, 2675 sizeof(int), sizeof(int)); 2676 if (error) 2677 break; 2678 } 2679 error = zero_entry(rulenum, sopt->sopt_name == IP_FW_RESETLOG); 2680 break; 2681 2682 default: 2683 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 2684 error = EINVAL; 2685 } 2686 2687 return (error); 2688} 2689 2690/** 2691 * dummynet needs a reference to the default rule, because rules can be 2692 * deleted while packets hold a reference to them. When this happens, 2693 * dummynet changes the reference to the default rule (it could well be a 2694 * NULL pointer, but this way we do not need to check for the special 2695 * case, plus here he have info on the default behaviour). 2696 */ 2697struct ip_fw *ip_fw_default_rule; 2698 2699/* 2700 * This procedure is only used to handle keepalives. It is invoked 2701 * every dyn_keepalive_period 2702 */ 2703static void 2704ipfw_tick(void * __unused unused) 2705{ 2706 int i; 2707 int s; 2708 ipfw_dyn_rule *q; 2709 2710 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 2711 goto done; 2712 2713 s = splimp(); 2714 for (i = 0 ; i < curr_dyn_buckets ; i++) { 2715 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 2716 if (q->dyn_type == O_LIMIT_PARENT) 2717 continue; 2718 if (q->id.proto != IPPROTO_TCP) 2719 continue; 2720 if ( (q->state & BOTH_SYN) != BOTH_SYN) 2721 continue; 2722 if (TIME_LEQ( time_second+dyn_keepalive_interval, 2723 q->expire)) 2724 continue; /* too early */ 2725 if (TIME_LEQ(q->expire, time_second)) 2726 continue; /* too late, rule expired */ 2727 2728 send_pkt(&(q->id), q->ack_rev - 1, q->ack_fwd, TH_SYN); 2729 send_pkt(&(q->id), q->ack_fwd - 1, q->ack_rev, 0); 2730 } 2731 } 2732 splx(s); 2733done: 2734 ipfw_timeout_h = timeout(ipfw_tick, NULL, dyn_keepalive_period*hz); 2735} 2736 2737static void 2738ipfw_init(void) 2739{ 2740 struct ip_fw default_rule; 2741 2742 ip_fw_chk_ptr = ipfw_chk; 2743 ip_fw_ctl_ptr = ipfw_ctl; 2744 layer3_chain = NULL; 2745 2746 bzero(&default_rule, sizeof default_rule); 2747 2748 default_rule.act_ofs = 0; 2749 default_rule.rulenum = IPFW_DEFAULT_RULE; 2750 default_rule.cmd_len = 1; 2751 default_rule.set = 31; 2752 2753 default_rule.cmd[0].len = 1; 2754 default_rule.cmd[0].opcode = 2755#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 2756 1 ? O_ACCEPT : 2757#endif 2758 O_DENY; 2759 2760 add_rule(&layer3_chain, &default_rule); 2761 2762 ip_fw_default_rule = layer3_chain; 2763 printf("ipfw2 initialized, divert %s, " 2764 "rule-based forwarding enabled, default to %s, logging ", 2765#ifdef IPDIVERT 2766 "enabled", 2767#else 2768 "disabled", 2769#endif 2770 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 2771 2772#ifdef IPFIREWALL_VERBOSE 2773 fw_verbose = 1; 2774#endif 2775#ifdef IPFIREWALL_VERBOSE_LIMIT 2776 verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 2777#endif 2778 if (fw_verbose == 0) 2779 printf("disabled\n"); 2780 else if (verbose_limit == 0) 2781 printf("unlimited\n"); 2782 else 2783 printf("limited to %d packets/entry by default\n", 2784 verbose_limit); 2785 bzero(&ipfw_timeout_h, sizeof(struct callout_handle)); 2786 ipfw_timeout_h = timeout(ipfw_tick, NULL, hz); 2787} 2788 2789static int 2790ipfw_modevent(module_t mod, int type, void *unused) 2791{ 2792 int s; 2793 int err = 0; 2794 2795 switch (type) { 2796 case MOD_LOAD: 2797 s = splimp(); 2798 if (IPFW_LOADED) { 2799 splx(s); 2800 printf("IP firewall already loaded\n"); 2801 err = EEXIST; 2802 } else { 2803 ipfw_init(); 2804 splx(s); 2805 } 2806 break; 2807 2808 case MOD_UNLOAD: 2809#if !defined(KLD_MODULE) 2810 printf("ipfw statically compiled, cannot unload\n"); 2811 err = EBUSY; 2812#else 2813 s = splimp(); 2814 untimeout(ipfw_tick, NULL, ipfw_timeout_h); 2815 ip_fw_chk_ptr = NULL; 2816 ip_fw_ctl_ptr = NULL; 2817 free_chain(&layer3_chain, 1 /* kill default rule */); 2818 splx(s); 2819 printf("IP firewall unloaded\n"); 2820#endif 2821 break; 2822 default: 2823 break; 2824 } 2825 return err; 2826} 2827 2828static moduledata_t ipfwmod = { 2829 "ipfw", 2830 ipfw_modevent, 2831 0 2832}; 2833DECLARE_MODULE(ipfw, ipfwmod, SI_SUB_PSEUDO, SI_ORDER_ANY); 2834MODULE_VERSION(ipfw, 1); 2835#endif /* IPFW2 */ 2836