ip_fw2.c revision 112250
198943Sluigi/* 298943Sluigi * Copyright (c) 2002 Luigi Rizzo, Universita` di Pisa 398943Sluigi * 498943Sluigi * Redistribution and use in source and binary forms, with or without 598943Sluigi * modification, are permitted provided that the following conditions 698943Sluigi * are met: 798943Sluigi * 1. Redistributions of source code must retain the above copyright 898943Sluigi * notice, this list of conditions and the following disclaimer. 9105775Smaxim * 2. Redistributions in binary form must reproduce the above copyright 1098943Sluigi * notice, this list of conditions and the following disclaimer in the 1198943Sluigi * documentation and/or other materials provided with the distribution. 12105775Smaxim * 1398943Sluigi * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 1498943Sluigi * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 1598943Sluigi * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 1698943Sluigi * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 1798943Sluigi * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 1898943Sluigi * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 1998943Sluigi * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 2098943Sluigi * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 2198943Sluigi * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 2298943Sluigi * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 2398943Sluigi * SUCH DAMAGE. 2498943Sluigi * 2598943Sluigi * $FreeBSD: head/sys/netinet/ip_fw2.c 112250 2003-03-15 01:13:00Z cjc $ 2698943Sluigi */ 2798943Sluigi 2898943Sluigi#define DEB(x) 2998943Sluigi#define DDB(x) x 3098943Sluigi 3198943Sluigi/* 3299622Sluigi * Implement IP packet firewall (new version) 3398943Sluigi */ 3498943Sluigi 3598943Sluigi#if !defined(KLD_MODULE) 3698943Sluigi#include "opt_ipfw.h" 3798943Sluigi#include "opt_ipdn.h" 3898943Sluigi#include "opt_ipdivert.h" 3998943Sluigi#include "opt_inet.h" 4098943Sluigi#ifndef INET 4198943Sluigi#error IPFIREWALL requires INET. 4298943Sluigi#endif /* INET */ 4398943Sluigi#endif 4498943Sluigi 4599622Sluigi#define IPFW2 1 4699622Sluigi#if IPFW2 4798943Sluigi#include <sys/param.h> 4898943Sluigi#include <sys/systm.h> 4998943Sluigi#include <sys/malloc.h> 5098943Sluigi#include <sys/mbuf.h> 5198943Sluigi#include <sys/kernel.h> 5298943Sluigi#include <sys/proc.h> 5398943Sluigi#include <sys/socket.h> 5498943Sluigi#include <sys/socketvar.h> 5598943Sluigi#include <sys/sysctl.h> 5698943Sluigi#include <sys/syslog.h> 5798943Sluigi#include <sys/ucred.h> 5898943Sluigi#include <net/if.h> 5998943Sluigi#include <net/route.h> 6098943Sluigi#include <netinet/in.h> 6198943Sluigi#include <netinet/in_systm.h> 6298943Sluigi#include <netinet/in_var.h> 6398943Sluigi#include <netinet/in_pcb.h> 6498943Sluigi#include <netinet/ip.h> 6598943Sluigi#include <netinet/ip_var.h> 6698943Sluigi#include <netinet/ip_icmp.h> 6798943Sluigi#include <netinet/ip_fw.h> 6898943Sluigi#include <netinet/ip_dummynet.h> 6998943Sluigi#include <netinet/tcp.h> 7098943Sluigi#include <netinet/tcp_timer.h> 7198943Sluigi#include <netinet/tcp_var.h> 7298943Sluigi#include <netinet/tcpip.h> 7398943Sluigi#include <netinet/udp.h> 7498943Sluigi#include <netinet/udp_var.h> 7598943Sluigi 7698943Sluigi#include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */ 7798943Sluigi 7899475Sluigi#include <machine/in_cksum.h> /* XXX for in_cksum */ 7999475Sluigi 80101628Sluigi/* 81101978Sluigi * XXX This one should go in sys/mbuf.h. It is used to avoid that 82101978Sluigi * a firewall-generated packet loops forever through the firewall. 83101978Sluigi */ 84101978Sluigi#ifndef M_SKIP_FIREWALL 85101978Sluigi#define M_SKIP_FIREWALL 0x4000 86101978Sluigi#endif 87101978Sluigi 88101978Sluigi/* 89101628Sluigi * set_disable contains one bit per set value (0..31). 90101628Sluigi * If the bit is set, all rules with the corresponding set 91101628Sluigi * are disabled. Set 31 is reserved for the default rule 92101628Sluigi * and CANNOT be disabled. 93101628Sluigi */ 94101628Sluigistatic u_int32_t set_disable; 95101628Sluigi 9699622Sluigistatic int fw_verbose; 9799622Sluigistatic int verbose_limit; 9898943Sluigi 99100004Sluigistatic struct callout_handle ipfw_timeout_h; 10098943Sluigi#define IPFW_DEFAULT_RULE 65535 10198943Sluigi 10298943Sluigi/* 10398943Sluigi * list of rules for layer 3 10498943Sluigi */ 10598943Sluigistatic struct ip_fw *layer3_chain; 10698943Sluigi 10798943SluigiMALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 10898943Sluigi 10998943Sluigistatic int fw_debug = 1; 11098943Sluigistatic int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 11198943Sluigi 11298943Sluigi#ifdef SYSCTL_NODE 11398943SluigiSYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 114102397ScjcSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, enable, 115109246Sdillon CTLFLAG_RW | CTLFLAG_SECURE3, 11698943Sluigi &fw_enable, 0, "Enable ipfw"); 11798943SluigiSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW, 11898943Sluigi &autoinc_step, 0, "Rule number autincrement step"); 119102397ScjcSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 120109246Sdillon CTLFLAG_RW | CTLFLAG_SECURE3, 121105775Smaxim &fw_one_pass, 0, 12298943Sluigi "Only do a single pass through ipfw when using dummynet(4)"); 123105775SmaximSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW, 12498943Sluigi &fw_debug, 0, "Enable printing of debug ip_fw statements"); 125102397ScjcSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, 126109246Sdillon CTLFLAG_RW | CTLFLAG_SECURE3, 12798943Sluigi &fw_verbose, 0, "Log matches to ipfw rules"); 128105775SmaximSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW, 12998943Sluigi &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged"); 13098943Sluigi 13198943Sluigi/* 13298943Sluigi * Description of dynamic rules. 13398943Sluigi * 13498943Sluigi * Dynamic rules are stored in lists accessed through a hash table 13598943Sluigi * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 13698943Sluigi * be modified through the sysctl variable dyn_buckets which is 13798943Sluigi * updated when the table becomes empty. 13898943Sluigi * 13998943Sluigi * XXX currently there is only one list, ipfw_dyn. 14098943Sluigi * 14198943Sluigi * When a packet is received, its address fields are first masked 14298943Sluigi * with the mask defined for the rule, then hashed, then matched 14398943Sluigi * against the entries in the corresponding list. 14498943Sluigi * Dynamic rules can be used for different purposes: 14598943Sluigi * + stateful rules; 14698943Sluigi * + enforcing limits on the number of sessions; 14798943Sluigi * + in-kernel NAT (not implemented yet) 14898943Sluigi * 14998943Sluigi * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 15098943Sluigi * measured in seconds and depending on the flags. 15198943Sluigi * 15298943Sluigi * The total number of dynamic rules is stored in dyn_count. 15398943Sluigi * The max number of dynamic rules is dyn_max. When we reach 15498943Sluigi * the maximum number of rules we do not create anymore. This is 15598943Sluigi * done to avoid consuming too much memory, but also too much 15698943Sluigi * time when searching on each packet (ideally, we should try instead 15798943Sluigi * to put a limit on the length of the list on each bucket...). 15898943Sluigi * 15998943Sluigi * Each dynamic rule holds a pointer to the parent ipfw rule so 16098943Sluigi * we know what action to perform. Dynamic rules are removed when 16198943Sluigi * the parent rule is deleted. XXX we should make them survive. 16298943Sluigi * 16398943Sluigi * There are some limitations with dynamic rules -- we do not 16498943Sluigi * obey the 'randomized match', and we do not do multiple 16598943Sluigi * passes through the firewall. XXX check the latter!!! 16698943Sluigi */ 16798943Sluigistatic ipfw_dyn_rule **ipfw_dyn_v = NULL; 16898943Sluigistatic u_int32_t dyn_buckets = 256; /* must be power of 2 */ 16998943Sluigistatic u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */ 17098943Sluigi 17198943Sluigi/* 17298943Sluigi * Timeouts for various events in handing dynamic rules. 17398943Sluigi */ 17498943Sluigistatic u_int32_t dyn_ack_lifetime = 300; 17598943Sluigistatic u_int32_t dyn_syn_lifetime = 20; 17698943Sluigistatic u_int32_t dyn_fin_lifetime = 1; 17798943Sluigistatic u_int32_t dyn_rst_lifetime = 1; 17898943Sluigistatic u_int32_t dyn_udp_lifetime = 10; 17998943Sluigistatic u_int32_t dyn_short_lifetime = 5; 18098943Sluigi 181101978Sluigi/* 182101978Sluigi * Keepalives are sent if dyn_keepalive is set. They are sent every 183101978Sluigi * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 184101978Sluigi * seconds of lifetime of a rule. 185101978Sluigi * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 186101978Sluigi * than dyn_keepalive_period. 187101978Sluigi */ 188105775Smaxim 189101978Sluigistatic u_int32_t dyn_keepalive_interval = 20; 190101978Sluigistatic u_int32_t dyn_keepalive_period = 5; 191100004Sluigistatic u_int32_t dyn_keepalive = 1; /* do send keepalives */ 19298943Sluigi 19399622Sluigistatic u_int32_t static_count; /* # of static rules */ 19499622Sluigistatic u_int32_t static_len; /* size in bytes of static rules */ 19599622Sluigistatic u_int32_t dyn_count; /* # of dynamic rules */ 196101978Sluigistatic u_int32_t dyn_max = 4096; /* max # of dynamic rules */ 19798943Sluigi 19898943SluigiSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW, 19998943Sluigi &dyn_buckets, 0, "Number of dyn. buckets"); 20098943SluigiSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD, 20198943Sluigi &curr_dyn_buckets, 0, "Current Number of dyn. buckets"); 20298943SluigiSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD, 20398943Sluigi &dyn_count, 0, "Number of dyn. rules"); 20498943SluigiSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW, 20598943Sluigi &dyn_max, 0, "Max number of dyn. rules"); 20698943SluigiSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD, 20798943Sluigi &static_count, 0, "Number of static rules"); 20898943SluigiSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW, 20998943Sluigi &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks"); 21098943SluigiSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW, 21198943Sluigi &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn"); 21298943SluigiSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW, 21398943Sluigi &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin"); 21498943SluigiSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW, 21598943Sluigi &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst"); 21698943SluigiSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW, 21798943Sluigi &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP"); 21898943SluigiSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW, 21998943Sluigi &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations"); 220100004SluigiSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW, 221100004Sluigi &dyn_keepalive, 0, "Enable keepalives for dyn. rules"); 22298943Sluigi 22398943Sluigi#endif /* SYSCTL_NODE */ 22498943Sluigi 22598943Sluigi 22698943Sluigistatic ip_fw_chk_t ipfw_chk; 22798943Sluigi 22898943Sluigiip_dn_ruledel_t *ip_dn_ruledel_ptr = NULL; /* hook into dummynet */ 22998943Sluigi 23098943Sluigi/* 23198943Sluigi * This macro maps an ip pointer into a layer3 header pointer of type T 23298943Sluigi */ 23398943Sluigi#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 23498943Sluigi 23599622Sluigistatic __inline int 23698943Sluigiicmptype_match(struct ip *ip, ipfw_insn_u32 *cmd) 23798943Sluigi{ 23898943Sluigi int type = L3HDR(struct icmp,ip)->icmp_type; 23998943Sluigi 24098943Sluigi return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 24198943Sluigi} 24298943Sluigi 24398943Sluigi#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 24498943Sluigi (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 24598943Sluigi 24698943Sluigistatic int 24798943Sluigiis_icmp_query(struct ip *ip) 24898943Sluigi{ 24998943Sluigi int type = L3HDR(struct icmp, ip)->icmp_type; 25098943Sluigi return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 25198943Sluigi} 25298943Sluigi#undef TT 25398943Sluigi 25498943Sluigi/* 25598943Sluigi * The following checks use two arrays of 8 or 16 bits to store the 25698943Sluigi * bits that we want set or clear, respectively. They are in the 25798943Sluigi * low and high half of cmd->arg1 or cmd->d[0]. 25898943Sluigi * 25998943Sluigi * We scan options and store the bits we find set. We succeed if 26098943Sluigi * 26198943Sluigi * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 26298943Sluigi * 26398943Sluigi * The code is sometimes optimized not to store additional variables. 26498943Sluigi */ 26598943Sluigi 26698943Sluigistatic int 26798943Sluigiflags_match(ipfw_insn *cmd, u_int8_t bits) 26898943Sluigi{ 26998943Sluigi u_char want_clear; 27098943Sluigi bits = ~bits; 27198943Sluigi 27298943Sluigi if ( ((cmd->arg1 & 0xff) & bits) != 0) 27398943Sluigi return 0; /* some bits we want set were clear */ 27498943Sluigi want_clear = (cmd->arg1 >> 8) & 0xff; 27598943Sluigi if ( (want_clear & bits) != want_clear) 27698943Sluigi return 0; /* some bits we want clear were set */ 27798943Sluigi return 1; 27898943Sluigi} 27998943Sluigi 28098943Sluigistatic int 28198943Sluigiipopts_match(struct ip *ip, ipfw_insn *cmd) 28298943Sluigi{ 28398943Sluigi int optlen, bits = 0; 28498943Sluigi u_char *cp = (u_char *)(ip + 1); 28598943Sluigi int x = (ip->ip_hl << 2) - sizeof (struct ip); 28698943Sluigi 28798943Sluigi for (; x > 0; x -= optlen, cp += optlen) { 28898943Sluigi int opt = cp[IPOPT_OPTVAL]; 28998943Sluigi 29098943Sluigi if (opt == IPOPT_EOL) 29198943Sluigi break; 29298943Sluigi if (opt == IPOPT_NOP) 29398943Sluigi optlen = 1; 29498943Sluigi else { 29598943Sluigi optlen = cp[IPOPT_OLEN]; 29698943Sluigi if (optlen <= 0 || optlen > x) 29798943Sluigi return 0; /* invalid or truncated */ 29898943Sluigi } 29998943Sluigi switch (opt) { 30098943Sluigi 30198943Sluigi default: 30298943Sluigi break; 30398943Sluigi 30498943Sluigi case IPOPT_LSRR: 30598943Sluigi bits |= IP_FW_IPOPT_LSRR; 30698943Sluigi break; 30798943Sluigi 30898943Sluigi case IPOPT_SSRR: 30998943Sluigi bits |= IP_FW_IPOPT_SSRR; 31098943Sluigi break; 31198943Sluigi 31298943Sluigi case IPOPT_RR: 31398943Sluigi bits |= IP_FW_IPOPT_RR; 31498943Sluigi break; 31598943Sluigi 31698943Sluigi case IPOPT_TS: 31798943Sluigi bits |= IP_FW_IPOPT_TS; 31898943Sluigi break; 31998943Sluigi } 32098943Sluigi } 32198943Sluigi return (flags_match(cmd, bits)); 32298943Sluigi} 32398943Sluigi 32498943Sluigistatic int 32598943Sluigitcpopts_match(struct ip *ip, ipfw_insn *cmd) 32698943Sluigi{ 32798943Sluigi int optlen, bits = 0; 32898943Sluigi struct tcphdr *tcp = L3HDR(struct tcphdr,ip); 32998943Sluigi u_char *cp = (u_char *)(tcp + 1); 33098943Sluigi int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 33198943Sluigi 33298943Sluigi for (; x > 0; x -= optlen, cp += optlen) { 33398943Sluigi int opt = cp[0]; 33498943Sluigi if (opt == TCPOPT_EOL) 33598943Sluigi break; 33698943Sluigi if (opt == TCPOPT_NOP) 33798943Sluigi optlen = 1; 33898943Sluigi else { 33998943Sluigi optlen = cp[1]; 34098943Sluigi if (optlen <= 0) 34198943Sluigi break; 34298943Sluigi } 34398943Sluigi 34498943Sluigi switch (opt) { 34598943Sluigi 34698943Sluigi default: 34798943Sluigi break; 34898943Sluigi 34998943Sluigi case TCPOPT_MAXSEG: 35098943Sluigi bits |= IP_FW_TCPOPT_MSS; 35198943Sluigi break; 35298943Sluigi 35398943Sluigi case TCPOPT_WINDOW: 35498943Sluigi bits |= IP_FW_TCPOPT_WINDOW; 35598943Sluigi break; 35698943Sluigi 35798943Sluigi case TCPOPT_SACK_PERMITTED: 35898943Sluigi case TCPOPT_SACK: 35998943Sluigi bits |= IP_FW_TCPOPT_SACK; 36098943Sluigi break; 36198943Sluigi 36298943Sluigi case TCPOPT_TIMESTAMP: 36398943Sluigi bits |= IP_FW_TCPOPT_TS; 36498943Sluigi break; 36598943Sluigi 36698943Sluigi case TCPOPT_CC: 36798943Sluigi case TCPOPT_CCNEW: 36898943Sluigi case TCPOPT_CCECHO: 36998943Sluigi bits |= IP_FW_TCPOPT_CC; 37098943Sluigi break; 37198943Sluigi } 37298943Sluigi } 37398943Sluigi return (flags_match(cmd, bits)); 37498943Sluigi} 37598943Sluigi 37698943Sluigistatic int 37798943Sluigiiface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 37898943Sluigi{ 37998943Sluigi if (ifp == NULL) /* no iface with this packet, match fails */ 38098943Sluigi return 0; 38198943Sluigi /* Check by name or by IP address */ 38299475Sluigi if (cmd->name[0] != '\0') { /* match by name */ 38398943Sluigi /* Check unit number (-1 is wildcard) */ 38498943Sluigi if (cmd->p.unit != -1 && cmd->p.unit != ifp->if_unit) 38598943Sluigi return(0); 38698943Sluigi /* Check name */ 38798943Sluigi if (!strncmp(ifp->if_name, cmd->name, IFNAMSIZ)) 38898943Sluigi return(1); 38998943Sluigi } else { 39098943Sluigi struct ifaddr *ia; 39198943Sluigi 39298943Sluigi TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 39398943Sluigi if (ia->ifa_addr == NULL) 39498943Sluigi continue; 39598943Sluigi if (ia->ifa_addr->sa_family != AF_INET) 39698943Sluigi continue; 39798943Sluigi if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 39898943Sluigi (ia->ifa_addr))->sin_addr.s_addr) 39998943Sluigi return(1); /* match */ 40098943Sluigi } 40198943Sluigi } 40298943Sluigi return(0); /* no match, fail ... */ 40398943Sluigi} 40498943Sluigi 405112250Scjc/* 406112250Scjc * The 'verrevpath' option checks that the interface that an IP packet 407112250Scjc * arrives on is the same interface that traffic destined for the 408112250Scjc * packet's source address would be routed out of. This is a measure 409112250Scjc * to block forged packets. This is also commonly known as "anti-spoofing" 410112250Scjc * or Unicast Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The 411112250Scjc * name of the knob is purposely reminisent of the Cisco IOS command, 412112250Scjc * 413112250Scjc * ip verify unicast reverse-path 414112250Scjc * 415112250Scjc * which implements the same functionality. But note that syntax is 416112250Scjc * misleading. The check may be performed on all IP packets whether unicast, 417112250Scjc * multicast, or broadcast. 418112250Scjc */ 419112250Scjcstatic int 420112250Scjcverify_rev_path(struct in_addr src, struct ifnet *ifp) 421112250Scjc{ 422112250Scjc static struct route ro; 423112250Scjc struct sockaddr_in *dst; 424112250Scjc 425112250Scjc dst = (struct sockaddr_in *)&(ro.ro_dst); 426112250Scjc 427112250Scjc /* Check if we've cached the route from the previous call. */ 428112250Scjc if (src.s_addr != dst->sin_addr.s_addr) { 429112250Scjc ro.ro_rt = NULL; 430112250Scjc 431112250Scjc bzero(dst, sizeof(*dst)); 432112250Scjc dst->sin_family = AF_INET; 433112250Scjc dst->sin_len = sizeof(*dst); 434112250Scjc dst->sin_addr = src; 435112250Scjc 436112250Scjc rtalloc_ign(&ro, RTF_CLONING|RTF_PRCLONING); 437112250Scjc } 438112250Scjc 439112250Scjc if ((ro.ro_rt == NULL) || (ifp == NULL) || 440112250Scjc (ro.ro_rt->rt_ifp->if_index != ifp->if_index)) 441112250Scjc return 0; 442112250Scjc 443112250Scjc return 1; 444112250Scjc} 445112250Scjc 446112250Scjc 44798943Sluigistatic u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */ 44898943Sluigi 44998943Sluigi#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 45099622Sluigi#define SNP(buf) buf, sizeof(buf) 451100004Sluigi 45298943Sluigi/* 45398943Sluigi * We enter here when we have a rule with O_LOG. 45499622Sluigi * XXX this function alone takes about 2Kbytes of code! 45598943Sluigi */ 45698943Sluigistatic void 45798943Sluigiipfw_log(struct ip_fw *f, u_int hlen, struct ether_header *eh, 45898943Sluigi struct mbuf *m, struct ifnet *oif) 45998943Sluigi{ 46098943Sluigi char *action; 46198943Sluigi int limit_reached = 0; 46299622Sluigi char action2[40], proto[48], fragment[28]; 46398943Sluigi 46498943Sluigi fragment[0] = '\0'; 46598943Sluigi proto[0] = '\0'; 46698943Sluigi 46798943Sluigi if (f == NULL) { /* bogus pkt */ 46898943Sluigi if (verbose_limit != 0 && norule_counter >= verbose_limit) 46998943Sluigi return; 47098943Sluigi norule_counter++; 47198943Sluigi if (norule_counter == verbose_limit) 47298943Sluigi limit_reached = verbose_limit; 47398943Sluigi action = "Refuse"; 47498943Sluigi } else { /* O_LOG is the first action, find the real one */ 47598943Sluigi ipfw_insn *cmd = ACTION_PTR(f); 47698943Sluigi ipfw_insn_log *l = (ipfw_insn_log *)cmd; 47798943Sluigi 47898943Sluigi if (l->max_log != 0 && l->log_left == 0) 47998943Sluigi return; 48098943Sluigi l->log_left--; 48198943Sluigi if (l->log_left == 0) 48298943Sluigi limit_reached = l->max_log; 48398943Sluigi cmd += F_LEN(cmd); /* point to first action */ 48498943Sluigi if (cmd->opcode == O_PROB) 48598943Sluigi cmd += F_LEN(cmd); 48698943Sluigi 48798943Sluigi action = action2; 48898943Sluigi switch (cmd->opcode) { 48998943Sluigi case O_DENY: 49098943Sluigi action = "Deny"; 49198943Sluigi break; 49299475Sluigi 49398943Sluigi case O_REJECT: 49499475Sluigi if (cmd->arg1==ICMP_REJECT_RST) 49599475Sluigi action = "Reset"; 49699475Sluigi else if (cmd->arg1==ICMP_UNREACH_HOST) 49799475Sluigi action = "Reject"; 49899475Sluigi else 49999475Sluigi snprintf(SNPARGS(action2, 0), "Unreach %d", 50099475Sluigi cmd->arg1); 50198943Sluigi break; 50299475Sluigi 50398943Sluigi case O_ACCEPT: 50498943Sluigi action = "Accept"; 50598943Sluigi break; 50698943Sluigi case O_COUNT: 50798943Sluigi action = "Count"; 50898943Sluigi break; 50998943Sluigi case O_DIVERT: 51098943Sluigi snprintf(SNPARGS(action2, 0), "Divert %d", 51198943Sluigi cmd->arg1); 51298943Sluigi break; 51398943Sluigi case O_TEE: 51498943Sluigi snprintf(SNPARGS(action2, 0), "Tee %d", 51598943Sluigi cmd->arg1); 51698943Sluigi break; 51798943Sluigi case O_SKIPTO: 51898943Sluigi snprintf(SNPARGS(action2, 0), "SkipTo %d", 51998943Sluigi cmd->arg1); 52098943Sluigi break; 52198943Sluigi case O_PIPE: 52298943Sluigi snprintf(SNPARGS(action2, 0), "Pipe %d", 52398943Sluigi cmd->arg1); 52498943Sluigi break; 52598943Sluigi case O_QUEUE: 52698943Sluigi snprintf(SNPARGS(action2, 0), "Queue %d", 52798943Sluigi cmd->arg1); 52898943Sluigi break; 52998943Sluigi case O_FORWARD_IP: { 53098943Sluigi ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 531100004Sluigi int len; 53298943Sluigi 533100004Sluigi len = snprintf(SNPARGS(action2, 0), "Forward to %s", 534100004Sluigi inet_ntoa(sa->sa.sin_addr)); 53598943Sluigi if (sa->sa.sin_port) 536100004Sluigi snprintf(SNPARGS(action2, len), ":%d", 537107897Smaxim sa->sa.sin_port); 53898943Sluigi } 53998943Sluigi break; 540105775Smaxim default: 54198943Sluigi action = "UNKNOWN"; 54298943Sluigi break; 54398943Sluigi } 54498943Sluigi } 54598943Sluigi 54698943Sluigi if (hlen == 0) { /* non-ip */ 54798943Sluigi snprintf(SNPARGS(proto, 0), "MAC"); 54898943Sluigi } else { 54998943Sluigi struct ip *ip = mtod(m, struct ip *); 55098943Sluigi /* these three are all aliases to the same thing */ 55198943Sluigi struct icmp *const icmp = L3HDR(struct icmp, ip); 55298943Sluigi struct tcphdr *const tcp = (struct tcphdr *)icmp; 55398943Sluigi struct udphdr *const udp = (struct udphdr *)icmp; 55498943Sluigi 55598943Sluigi int ip_off, offset, ip_len; 55698943Sluigi 55798943Sluigi int len; 55898943Sluigi 55998943Sluigi if (eh != NULL) { /* layer 2 packets are as on the wire */ 56098943Sluigi ip_off = ntohs(ip->ip_off); 56198943Sluigi ip_len = ntohs(ip->ip_len); 56298943Sluigi } else { 56398943Sluigi ip_off = ip->ip_off; 56498943Sluigi ip_len = ip->ip_len; 56598943Sluigi } 56698943Sluigi offset = ip_off & IP_OFFMASK; 56798943Sluigi switch (ip->ip_p) { 56898943Sluigi case IPPROTO_TCP: 56998943Sluigi len = snprintf(SNPARGS(proto, 0), "TCP %s", 57098943Sluigi inet_ntoa(ip->ip_src)); 57198943Sluigi if (offset == 0) 572100004Sluigi snprintf(SNPARGS(proto, len), ":%d %s:%d", 573100004Sluigi ntohs(tcp->th_sport), 574100004Sluigi inet_ntoa(ip->ip_dst), 575100004Sluigi ntohs(tcp->th_dport)); 57698943Sluigi else 577100004Sluigi snprintf(SNPARGS(proto, len), " %s", 578100004Sluigi inet_ntoa(ip->ip_dst)); 57998943Sluigi break; 58098943Sluigi 58198943Sluigi case IPPROTO_UDP: 58298943Sluigi len = snprintf(SNPARGS(proto, 0), "UDP %s", 58398943Sluigi inet_ntoa(ip->ip_src)); 58498943Sluigi if (offset == 0) 585100004Sluigi snprintf(SNPARGS(proto, len), ":%d %s:%d", 586100004Sluigi ntohs(udp->uh_sport), 587100004Sluigi inet_ntoa(ip->ip_dst), 588100004Sluigi ntohs(udp->uh_dport)); 58998943Sluigi else 590100004Sluigi snprintf(SNPARGS(proto, len), " %s", 591100004Sluigi inet_ntoa(ip->ip_dst)); 59298943Sluigi break; 59398943Sluigi 59498943Sluigi case IPPROTO_ICMP: 59598943Sluigi if (offset == 0) 59698943Sluigi len = snprintf(SNPARGS(proto, 0), 59798943Sluigi "ICMP:%u.%u ", 59898943Sluigi icmp->icmp_type, icmp->icmp_code); 59998943Sluigi else 60098943Sluigi len = snprintf(SNPARGS(proto, 0), "ICMP "); 60198943Sluigi len += snprintf(SNPARGS(proto, len), "%s", 60298943Sluigi inet_ntoa(ip->ip_src)); 60398943Sluigi snprintf(SNPARGS(proto, len), " %s", 60498943Sluigi inet_ntoa(ip->ip_dst)); 60598943Sluigi break; 60698943Sluigi 60798943Sluigi default: 60898943Sluigi len = snprintf(SNPARGS(proto, 0), "P:%d %s", ip->ip_p, 60998943Sluigi inet_ntoa(ip->ip_src)); 61098943Sluigi snprintf(SNPARGS(proto, len), " %s", 61198943Sluigi inet_ntoa(ip->ip_dst)); 61298943Sluigi break; 61398943Sluigi } 61498943Sluigi 61598943Sluigi if (ip_off & (IP_MF | IP_OFFMASK)) 61698943Sluigi snprintf(SNPARGS(fragment, 0), " (frag %d:%d@%d%s)", 61798943Sluigi ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 61898943Sluigi offset << 3, 61998943Sluigi (ip_off & IP_MF) ? "+" : ""); 62098943Sluigi } 62198943Sluigi if (oif || m->m_pkthdr.rcvif) 62298943Sluigi log(LOG_SECURITY | LOG_INFO, 62398943Sluigi "ipfw: %d %s %s %s via %s%d%s\n", 62498943Sluigi f ? f->rulenum : -1, 62598943Sluigi action, proto, oif ? "out" : "in", 62698943Sluigi oif ? oif->if_name : m->m_pkthdr.rcvif->if_name, 62798943Sluigi oif ? oif->if_unit : m->m_pkthdr.rcvif->if_unit, 62898943Sluigi fragment); 62998943Sluigi else 63098943Sluigi log(LOG_SECURITY | LOG_INFO, 63198943Sluigi "ipfw: %d %s %s [no if info]%s\n", 63298943Sluigi f ? f->rulenum : -1, 63398943Sluigi action, proto, fragment); 63498943Sluigi if (limit_reached) 63598943Sluigi log(LOG_SECURITY | LOG_NOTICE, 63698943Sluigi "ipfw: limit %d reached on entry %d\n", 63798943Sluigi limit_reached, f ? f->rulenum : -1); 63898943Sluigi} 63998943Sluigi 64098943Sluigi/* 64198943Sluigi * IMPORTANT: the hash function for dynamic rules must be commutative 64299475Sluigi * in source and destination (ip,port), because rules are bidirectional 64398943Sluigi * and we want to find both in the same bucket. 64498943Sluigi */ 64598943Sluigistatic __inline int 64698943Sluigihash_packet(struct ipfw_flow_id *id) 64798943Sluigi{ 64898943Sluigi u_int32_t i; 64998943Sluigi 65098943Sluigi i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 65198943Sluigi i &= (curr_dyn_buckets - 1); 65298943Sluigi return i; 65398943Sluigi} 65498943Sluigi 65598943Sluigi/** 65698943Sluigi * unlink a dynamic rule from a chain. prev is a pointer to 65798943Sluigi * the previous one, q is a pointer to the rule to delete, 65898943Sluigi * head is a pointer to the head of the queue. 65998943Sluigi * Modifies q and potentially also head. 66098943Sluigi */ 66198943Sluigi#define UNLINK_DYN_RULE(prev, head, q) { \ 66298943Sluigi ipfw_dyn_rule *old_q = q; \ 66398943Sluigi \ 66498943Sluigi /* remove a refcount to the parent */ \ 66598943Sluigi if (q->dyn_type == O_LIMIT) \ 66698943Sluigi q->parent->count--; \ 667108258Smaxim DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 66898943Sluigi (q->id.src_ip), (q->id.src_port), \ 66998943Sluigi (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \ 67098943Sluigi if (prev != NULL) \ 67198943Sluigi prev->next = q = q->next; \ 67298943Sluigi else \ 67398943Sluigi head = q = q->next; \ 67498943Sluigi dyn_count--; \ 67598943Sluigi free(old_q, M_IPFW); } 67698943Sluigi 67798943Sluigi#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 67898943Sluigi 67998943Sluigi/** 68098943Sluigi * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 68198943Sluigi * 68298943Sluigi * If keep_me == NULL, rules are deleted even if not expired, 68398943Sluigi * otherwise only expired rules are removed. 68498943Sluigi * 68598943Sluigi * The value of the second parameter is also used to point to identify 68698943Sluigi * a rule we absolutely do not want to remove (e.g. because we are 68798943Sluigi * holding a reference to it -- this is the case with O_LIMIT_PARENT 68898943Sluigi * rules). The pointer is only used for comparison, so any non-null 68998943Sluigi * value will do. 69098943Sluigi */ 69198943Sluigistatic void 69298943Sluigiremove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 69398943Sluigi{ 69498943Sluigi static u_int32_t last_remove = 0; 69598943Sluigi 69698943Sluigi#define FORCE (keep_me == NULL) 69798943Sluigi 69898943Sluigi ipfw_dyn_rule *prev, *q; 69998943Sluigi int i, pass = 0, max_pass = 0; 70098943Sluigi 70198943Sluigi if (ipfw_dyn_v == NULL || dyn_count == 0) 70298943Sluigi return; 70398943Sluigi /* do not expire more than once per second, it is useless */ 70498943Sluigi if (!FORCE && last_remove == time_second) 70598943Sluigi return; 70698943Sluigi last_remove = time_second; 70798943Sluigi 70898943Sluigi /* 70998943Sluigi * because O_LIMIT refer to parent rules, during the first pass only 71098943Sluigi * remove child and mark any pending LIMIT_PARENT, and remove 71198943Sluigi * them in a second pass. 71298943Sluigi */ 71398943Sluiginext_pass: 71498943Sluigi for (i = 0 ; i < curr_dyn_buckets ; i++) { 71598943Sluigi for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) { 71698943Sluigi /* 71798943Sluigi * Logic can become complex here, so we split tests. 71898943Sluigi */ 71998943Sluigi if (q == keep_me) 72098943Sluigi goto next; 72198943Sluigi if (rule != NULL && rule != q->rule) 72298943Sluigi goto next; /* not the one we are looking for */ 72398943Sluigi if (q->dyn_type == O_LIMIT_PARENT) { 72498943Sluigi /* 72598943Sluigi * handle parent in the second pass, 72698943Sluigi * record we need one. 72798943Sluigi */ 72898943Sluigi max_pass = 1; 72998943Sluigi if (pass == 0) 73098943Sluigi goto next; 73198943Sluigi if (FORCE && q->count != 0 ) { 73298943Sluigi /* XXX should not happen! */ 733108258Smaxim printf("ipfw: OUCH! cannot remove rule," 73498943Sluigi " count %d\n", q->count); 73598943Sluigi } 73698943Sluigi } else { 73798943Sluigi if (!FORCE && 73898943Sluigi !TIME_LEQ( q->expire, time_second )) 73998943Sluigi goto next; 74098943Sluigi } 74198943Sluigi UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 74298943Sluigi continue; 74398943Sluiginext: 74498943Sluigi prev=q; 74598943Sluigi q=q->next; 74698943Sluigi } 74798943Sluigi } 74898943Sluigi if (pass++ < max_pass) 74998943Sluigi goto next_pass; 75098943Sluigi} 75198943Sluigi 75298943Sluigi 75398943Sluigi/** 75498943Sluigi * lookup a dynamic rule. 75598943Sluigi */ 75698943Sluigistatic ipfw_dyn_rule * 757100004Sluigilookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 758100004Sluigi struct tcphdr *tcp) 75998943Sluigi{ 76098943Sluigi /* 76198943Sluigi * stateful ipfw extensions. 76298943Sluigi * Lookup into dynamic session queue 76398943Sluigi */ 76498943Sluigi#define MATCH_REVERSE 0 76598943Sluigi#define MATCH_FORWARD 1 76698943Sluigi#define MATCH_NONE 2 76798943Sluigi#define MATCH_UNKNOWN 3 76898943Sluigi int i, dir = MATCH_NONE; 76998943Sluigi ipfw_dyn_rule *prev, *q=NULL; 77098943Sluigi 77198943Sluigi if (ipfw_dyn_v == NULL) 77298943Sluigi goto done; /* not found */ 77398943Sluigi i = hash_packet( pkt ); 77498943Sluigi for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) { 77598943Sluigi if (q->dyn_type == O_LIMIT_PARENT) 77698943Sluigi goto next; 77798943Sluigi if (TIME_LEQ( q->expire, time_second)) { /* expire entry */ 77898943Sluigi UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q); 77998943Sluigi continue; 78098943Sluigi } 78198943Sluigi if ( pkt->proto == q->id.proto) { 78298943Sluigi if (pkt->src_ip == q->id.src_ip && 78398943Sluigi pkt->dst_ip == q->id.dst_ip && 78498943Sluigi pkt->src_port == q->id.src_port && 78598943Sluigi pkt->dst_port == q->id.dst_port ) { 78698943Sluigi dir = MATCH_FORWARD; 78798943Sluigi break; 78898943Sluigi } 78998943Sluigi if (pkt->src_ip == q->id.dst_ip && 79098943Sluigi pkt->dst_ip == q->id.src_ip && 79198943Sluigi pkt->src_port == q->id.dst_port && 79298943Sluigi pkt->dst_port == q->id.src_port ) { 79398943Sluigi dir = MATCH_REVERSE; 79498943Sluigi break; 79598943Sluigi } 79698943Sluigi } 79798943Sluiginext: 79898943Sluigi prev = q; 79998943Sluigi q = q->next; 80098943Sluigi } 80198943Sluigi if (q == NULL) 80298943Sluigi goto done; /* q = NULL, not found */ 80398943Sluigi 80498943Sluigi if ( prev != NULL) { /* found and not in front */ 80598943Sluigi prev->next = q->next; 80698943Sluigi q->next = ipfw_dyn_v[i]; 80798943Sluigi ipfw_dyn_v[i] = q; 80898943Sluigi } 80998943Sluigi if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 81098943Sluigi u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 81198943Sluigi 81298943Sluigi#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 81398943Sluigi#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 81498943Sluigi q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 81598943Sluigi switch (q->state) { 81698943Sluigi case TH_SYN: /* opening */ 81798943Sluigi q->expire = time_second + dyn_syn_lifetime; 81898943Sluigi break; 819101978Sluigi 82098943Sluigi case BOTH_SYN: /* move to established */ 82198943Sluigi case BOTH_SYN | TH_FIN : /* one side tries to close */ 82298943Sluigi case BOTH_SYN | (TH_FIN << 8) : 823100004Sluigi if (tcp) { 824100004Sluigi#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 825100004Sluigi u_int32_t ack = ntohl(tcp->th_ack); 826100004Sluigi if (dir == MATCH_FORWARD) { 827100004Sluigi if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 828100004Sluigi q->ack_fwd = ack; 829100004Sluigi else { /* ignore out-of-sequence */ 830100004Sluigi break; 831100004Sluigi } 832100004Sluigi } else { 833100004Sluigi if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 834100004Sluigi q->ack_rev = ack; 835100004Sluigi else { /* ignore out-of-sequence */ 836100004Sluigi break; 837100004Sluigi } 838100004Sluigi } 839100004Sluigi } 84098943Sluigi q->expire = time_second + dyn_ack_lifetime; 84198943Sluigi break; 842101978Sluigi 84398943Sluigi case BOTH_SYN | BOTH_FIN: /* both sides closed */ 844101978Sluigi if (dyn_fin_lifetime >= dyn_keepalive_period) 845101978Sluigi dyn_fin_lifetime = dyn_keepalive_period - 1; 84698943Sluigi q->expire = time_second + dyn_fin_lifetime; 84798943Sluigi break; 848101978Sluigi 84998943Sluigi default: 85098943Sluigi#if 0 85198943Sluigi /* 85298943Sluigi * reset or some invalid combination, but can also 85398943Sluigi * occur if we use keep-state the wrong way. 85498943Sluigi */ 85598943Sluigi if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 85698943Sluigi printf("invalid state: 0x%x\n", q->state); 85798943Sluigi#endif 858101978Sluigi if (dyn_rst_lifetime >= dyn_keepalive_period) 859101978Sluigi dyn_rst_lifetime = dyn_keepalive_period - 1; 86098943Sluigi q->expire = time_second + dyn_rst_lifetime; 86198943Sluigi break; 86298943Sluigi } 86398943Sluigi } else if (pkt->proto == IPPROTO_UDP) { 86498943Sluigi q->expire = time_second + dyn_udp_lifetime; 86598943Sluigi } else { 86698943Sluigi /* other protocols */ 86798943Sluigi q->expire = time_second + dyn_short_lifetime; 86898943Sluigi } 86998943Sluigidone: 87098943Sluigi if (match_direction) 87198943Sluigi *match_direction = dir; 87298943Sluigi return q; 87398943Sluigi} 87498943Sluigi 87598943Sluigistatic void 87698943Sluigirealloc_dynamic_table(void) 87798943Sluigi{ 878101978Sluigi /* 879101978Sluigi * Try reallocation, make sure we have a power of 2 and do 880101978Sluigi * not allow more than 64k entries. In case of overflow, 881101978Sluigi * default to 1024. 882101978Sluigi */ 88398943Sluigi 884101978Sluigi if (dyn_buckets > 65536) 885101978Sluigi dyn_buckets = 1024; 88698943Sluigi if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */ 88798943Sluigi dyn_buckets = curr_dyn_buckets; /* reset */ 88898943Sluigi return; 88998943Sluigi } 89098943Sluigi curr_dyn_buckets = dyn_buckets; 89198943Sluigi if (ipfw_dyn_v != NULL) 89298943Sluigi free(ipfw_dyn_v, M_IPFW); 893101978Sluigi for (;;) { 894101978Sluigi ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 895105440Smux M_IPFW, M_NOWAIT | M_ZERO); 896101978Sluigi if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2) 897101978Sluigi break; 898101978Sluigi curr_dyn_buckets /= 2; 899101978Sluigi } 90098943Sluigi} 90198943Sluigi 90298943Sluigi/** 90398943Sluigi * Install state of type 'type' for a dynamic session. 90498943Sluigi * The hash table contains two type of rules: 90598943Sluigi * - regular rules (O_KEEP_STATE) 90698943Sluigi * - rules for sessions with limited number of sess per user 90798943Sluigi * (O_LIMIT). When they are created, the parent is 90898943Sluigi * increased by 1, and decreased on delete. In this case, 90998943Sluigi * the third parameter is the parent rule and not the chain. 91098943Sluigi * - "parent" rules for the above (O_LIMIT_PARENT). 91198943Sluigi */ 91298943Sluigistatic ipfw_dyn_rule * 91398943Sluigiadd_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 91498943Sluigi{ 91598943Sluigi ipfw_dyn_rule *r; 91698943Sluigi int i; 91798943Sluigi 91898943Sluigi if (ipfw_dyn_v == NULL || 91998943Sluigi (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) { 92098943Sluigi realloc_dynamic_table(); 92198943Sluigi if (ipfw_dyn_v == NULL) 92298943Sluigi return NULL; /* failed ! */ 92398943Sluigi } 92498943Sluigi i = hash_packet(id); 92598943Sluigi 926105440Smux r = malloc(sizeof *r, M_IPFW, M_NOWAIT | M_ZERO); 92798943Sluigi if (r == NULL) { 928108258Smaxim printf ("ipfw: sorry cannot allocate state\n"); 92998943Sluigi return NULL; 93098943Sluigi } 93198943Sluigi 93298943Sluigi /* increase refcount on parent, and set pointer */ 93398943Sluigi if (dyn_type == O_LIMIT) { 93498943Sluigi ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 93598943Sluigi if ( parent->dyn_type != O_LIMIT_PARENT) 93698943Sluigi panic("invalid parent"); 93798943Sluigi parent->count++; 93898943Sluigi r->parent = parent; 93998943Sluigi rule = parent->rule; 94098943Sluigi } 94198943Sluigi 94298943Sluigi r->id = *id; 94398943Sluigi r->expire = time_second + dyn_syn_lifetime; 94498943Sluigi r->rule = rule; 94598943Sluigi r->dyn_type = dyn_type; 94698943Sluigi r->pcnt = r->bcnt = 0; 94798943Sluigi r->count = 0; 94898943Sluigi 94998943Sluigi r->bucket = i; 95098943Sluigi r->next = ipfw_dyn_v[i]; 95198943Sluigi ipfw_dyn_v[i] = r; 95298943Sluigi dyn_count++; 953108258Smaxim DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 95498943Sluigi dyn_type, 95598943Sluigi (r->id.src_ip), (r->id.src_port), 95698943Sluigi (r->id.dst_ip), (r->id.dst_port), 95798943Sluigi dyn_count ); ) 95898943Sluigi return r; 95998943Sluigi} 96098943Sluigi 96198943Sluigi/** 96298943Sluigi * lookup dynamic parent rule using pkt and rule as search keys. 96398943Sluigi * If the lookup fails, then install one. 96498943Sluigi */ 96598943Sluigistatic ipfw_dyn_rule * 96698943Sluigilookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 96798943Sluigi{ 96898943Sluigi ipfw_dyn_rule *q; 96998943Sluigi int i; 97098943Sluigi 97198943Sluigi if (ipfw_dyn_v) { 97298943Sluigi i = hash_packet( pkt ); 97398943Sluigi for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next) 97498943Sluigi if (q->dyn_type == O_LIMIT_PARENT && 97598943Sluigi rule== q->rule && 97698943Sluigi pkt->proto == q->id.proto && 97798943Sluigi pkt->src_ip == q->id.src_ip && 97898943Sluigi pkt->dst_ip == q->id.dst_ip && 97998943Sluigi pkt->src_port == q->id.src_port && 98098943Sluigi pkt->dst_port == q->id.dst_port) { 98198943Sluigi q->expire = time_second + dyn_short_lifetime; 982108258Smaxim DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 98398943Sluigi return q; 98498943Sluigi } 98598943Sluigi } 98698943Sluigi return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 98798943Sluigi} 98898943Sluigi 98998943Sluigi/** 99098943Sluigi * Install dynamic state for rule type cmd->o.opcode 99198943Sluigi * 99298943Sluigi * Returns 1 (failure) if state is not installed because of errors or because 99398943Sluigi * session limitations are enforced. 99498943Sluigi */ 99598943Sluigistatic int 99698943Sluigiinstall_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 99798943Sluigi struct ip_fw_args *args) 99898943Sluigi{ 99998943Sluigi static int last_log; 100098943Sluigi 100198943Sluigi ipfw_dyn_rule *q; 100298943Sluigi 1003108258Smaxim DEB(printf("ipfw: install state type %d 0x%08x %u -> 0x%08x %u\n", 100498943Sluigi cmd->o.opcode, 100598943Sluigi (args->f_id.src_ip), (args->f_id.src_port), 100698943Sluigi (args->f_id.dst_ip), (args->f_id.dst_port) );) 100798943Sluigi 1008100004Sluigi q = lookup_dyn_rule(&args->f_id, NULL, NULL); 100998943Sluigi 101098943Sluigi if (q != NULL) { /* should never occur */ 101198943Sluigi if (last_log != time_second) { 101298943Sluigi last_log = time_second; 1013108258Smaxim printf("ipfw: install_state: entry already present, done\n"); 101498943Sluigi } 101598943Sluigi return 0; 101698943Sluigi } 101798943Sluigi 101898943Sluigi if (dyn_count >= dyn_max) 101998943Sluigi /* 102098943Sluigi * Run out of slots, try to remove any expired rule. 102198943Sluigi */ 102298943Sluigi remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 102398943Sluigi 102498943Sluigi if (dyn_count >= dyn_max) { 102598943Sluigi if (last_log != time_second) { 102698943Sluigi last_log = time_second; 1027108258Smaxim printf("ipfw: install_state: Too many dynamic rules\n"); 102898943Sluigi } 102998943Sluigi return 1; /* cannot install, notify caller */ 103098943Sluigi } 103198943Sluigi 103298943Sluigi switch (cmd->o.opcode) { 103398943Sluigi case O_KEEP_STATE: /* bidir rule */ 103498943Sluigi add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 103598943Sluigi break; 103698943Sluigi 103798943Sluigi case O_LIMIT: /* limit number of sessions */ 103898943Sluigi { 103998943Sluigi u_int16_t limit_mask = cmd->limit_mask; 104098943Sluigi struct ipfw_flow_id id; 104198943Sluigi ipfw_dyn_rule *parent; 104298943Sluigi 1043108258Smaxim DEB(printf("ipfw: installing dyn-limit rule %d\n", 1044108258Smaxim cmd->conn_limit);) 104598943Sluigi 104698943Sluigi id.dst_ip = id.src_ip = 0; 104798943Sluigi id.dst_port = id.src_port = 0; 104898943Sluigi id.proto = args->f_id.proto; 104998943Sluigi 105098943Sluigi if (limit_mask & DYN_SRC_ADDR) 105198943Sluigi id.src_ip = args->f_id.src_ip; 105298943Sluigi if (limit_mask & DYN_DST_ADDR) 105398943Sluigi id.dst_ip = args->f_id.dst_ip; 105498943Sluigi if (limit_mask & DYN_SRC_PORT) 105598943Sluigi id.src_port = args->f_id.src_port; 105698943Sluigi if (limit_mask & DYN_DST_PORT) 105798943Sluigi id.dst_port = args->f_id.dst_port; 105898943Sluigi parent = lookup_dyn_parent(&id, rule); 105998943Sluigi if (parent == NULL) { 1060108258Smaxim printf("ipfw: add parent failed\n"); 106198943Sluigi return 1; 106298943Sluigi } 106398943Sluigi if (parent->count >= cmd->conn_limit) { 106498943Sluigi /* 106598943Sluigi * See if we can remove some expired rule. 106698943Sluigi */ 106798943Sluigi remove_dyn_rule(rule, parent); 106898943Sluigi if (parent->count >= cmd->conn_limit) { 106998943Sluigi if (fw_verbose && last_log != time_second) { 107098943Sluigi last_log = time_second; 1071106118Smaxim log(LOG_SECURITY | LOG_DEBUG, 107298943Sluigi "drop session, too many entries\n"); 107398943Sluigi } 107498943Sluigi return 1; 107598943Sluigi } 107698943Sluigi } 107798943Sluigi add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 107898943Sluigi } 107998943Sluigi break; 108098943Sluigi default: 1081108258Smaxim printf("ipfw: unknown dynamic rule type %u\n", cmd->o.opcode); 108298943Sluigi return 1; 108398943Sluigi } 1084100004Sluigi lookup_dyn_rule(&args->f_id, NULL, NULL); /* XXX just set lifetime */ 108598943Sluigi return 0; 108698943Sluigi} 108798943Sluigi 1088100004Sluigi/* 1089100004Sluigi * Transmit a TCP packet, containing either a RST or a keepalive. 1090100004Sluigi * When flags & TH_RST, we are sending a RST packet, because of a 1091100004Sluigi * "reset" action matched the packet. 1092100004Sluigi * Otherwise we are sending a keepalive, and flags & TH_ 1093100004Sluigi */ 109499475Sluigistatic void 1095100004Sluigisend_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags) 109699475Sluigi{ 109799475Sluigi struct mbuf *m; 1098100004Sluigi struct ip *ip; 109999475Sluigi struct tcphdr *tcp; 110099475Sluigi struct route sro; /* fake route */ 110199475Sluigi 1102111119Simp MGETHDR(m, M_DONTWAIT, MT_HEADER); 1103105775Smaxim if (m == 0) 110499475Sluigi return; 110599475Sluigi m->m_pkthdr.rcvif = (struct ifnet *)0; 110699475Sluigi m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 110799475Sluigi m->m_data += max_linkhdr; 110899475Sluigi 1109100004Sluigi ip = mtod(m, struct ip *); 1110100004Sluigi bzero(ip, m->m_len); 1111100004Sluigi tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 111299475Sluigi ip->ip_p = IPPROTO_TCP; 1113100004Sluigi tcp->th_off = 5; 1114100004Sluigi /* 1115100004Sluigi * Assume we are sending a RST (or a keepalive in the reverse 1116100004Sluigi * direction), swap src and destination addresses and ports. 1117100004Sluigi */ 1118100004Sluigi ip->ip_src.s_addr = htonl(id->dst_ip); 1119100004Sluigi ip->ip_dst.s_addr = htonl(id->src_ip); 1120100004Sluigi tcp->th_sport = htons(id->dst_port); 1121100004Sluigi tcp->th_dport = htons(id->src_port); 1122100004Sluigi if (flags & TH_RST) { /* we are sending a RST */ 1123100004Sluigi if (flags & TH_ACK) { 1124100004Sluigi tcp->th_seq = htonl(ack); 1125100004Sluigi tcp->th_ack = htonl(0); 1126100004Sluigi tcp->th_flags = TH_RST; 1127100004Sluigi } else { 1128100004Sluigi if (flags & TH_SYN) 1129100004Sluigi seq++; 1130100004Sluigi tcp->th_seq = htonl(0); 1131100004Sluigi tcp->th_ack = htonl(seq); 1132100004Sluigi tcp->th_flags = TH_RST | TH_ACK; 1133100004Sluigi } 113499475Sluigi } else { 1135100004Sluigi /* 1136100004Sluigi * We are sending a keepalive. flags & TH_SYN determines 1137100004Sluigi * the direction, forward if set, reverse if clear. 1138100004Sluigi * NOTE: seq and ack are always assumed to be correct 1139100004Sluigi * as set by the caller. This may be confusing... 1140100004Sluigi */ 1141100004Sluigi if (flags & TH_SYN) { 1142100004Sluigi /* 1143100004Sluigi * we have to rewrite the correct addresses! 1144100004Sluigi */ 1145100004Sluigi ip->ip_dst.s_addr = htonl(id->dst_ip); 1146100004Sluigi ip->ip_src.s_addr = htonl(id->src_ip); 1147100004Sluigi tcp->th_dport = htons(id->dst_port); 1148100004Sluigi tcp->th_sport = htons(id->src_port); 1149100004Sluigi } 1150100004Sluigi tcp->th_seq = htonl(seq); 1151100004Sluigi tcp->th_ack = htonl(ack); 1152100004Sluigi tcp->th_flags = TH_ACK; 115399475Sluigi } 1154100004Sluigi /* 1155100004Sluigi * set ip_len to the payload size so we can compute 1156100004Sluigi * the tcp checksum on the pseudoheader 1157100004Sluigi * XXX check this, could save a couple of words ? 1158100004Sluigi */ 1159100004Sluigi ip->ip_len = htons(sizeof(struct tcphdr)); 116099475Sluigi tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1161100004Sluigi /* 1162100004Sluigi * now fill fields left out earlier 1163100004Sluigi */ 116499475Sluigi ip->ip_ttl = ip_defttl; 116599475Sluigi ip->ip_len = m->m_pkthdr.len; 116699475Sluigi bzero (&sro, sizeof (sro)); 116799475Sluigi ip_rtaddr(ip->ip_dst, &sro); 1168101978Sluigi m->m_flags |= M_SKIP_FIREWALL; 1169105194Ssam ip_output(m, NULL, &sro, 0, NULL, NULL); 117099475Sluigi if (sro.ro_rt) 117199475Sluigi RTFREE(sro.ro_rt); 117299475Sluigi} 117399475Sluigi 117498943Sluigi/* 117598943Sluigi * sends a reject message, consuming the mbuf passed as an argument. 117698943Sluigi */ 117798943Sluigistatic void 117899475Sluigisend_reject(struct ip_fw_args *args, int code, int offset, int ip_len) 117998943Sluigi{ 1180101843Sphk 1181108327Siedowse if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1182108327Siedowse /* We need the IP header in host order for icmp_error(). */ 1183108327Siedowse if (args->eh != NULL) { 1184108327Siedowse struct ip *ip = mtod(args->m, struct ip *); 1185108327Siedowse ip->ip_len = ntohs(ip->ip_len); 1186108327Siedowse ip->ip_off = ntohs(ip->ip_off); 1187108327Siedowse } 118899475Sluigi icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1189108327Siedowse } else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) { 119099475Sluigi struct tcphdr *const tcp = 119199475Sluigi L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 119299475Sluigi if ( (tcp->th_flags & TH_RST) == 0) 1193100004Sluigi send_pkt(&(args->f_id), ntohl(tcp->th_seq), 1194100004Sluigi ntohl(tcp->th_ack), 1195100004Sluigi tcp->th_flags | TH_RST); 119699475Sluigi m_freem(args->m); 119799475Sluigi } else 119899475Sluigi m_freem(args->m); 119999475Sluigi args->m = NULL; 120098943Sluigi} 120198943Sluigi 120298943Sluigi/** 120398943Sluigi * 120498943Sluigi * Given an ip_fw *, lookup_next_rule will return a pointer 120598943Sluigi * to the next rule, which can be either the jump 120698943Sluigi * target (for skipto instructions) or the next one in the list (in 120798943Sluigi * all other cases including a missing jump target). 120898943Sluigi * The result is also written in the "next_rule" field of the rule. 120998943Sluigi * Backward jumps are not allowed, so start looking from the next 121098943Sluigi * rule... 121198943Sluigi * 121298943Sluigi * This never returns NULL -- in case we do not have an exact match, 121398943Sluigi * the next rule is returned. When the ruleset is changed, 121498943Sluigi * pointers are flushed so we are always correct. 1215105775Smaxim */ 121698943Sluigi 121798943Sluigistatic struct ip_fw * 121898943Sluigilookup_next_rule(struct ip_fw *me) 121998943Sluigi{ 122098943Sluigi struct ip_fw *rule = NULL; 122198943Sluigi ipfw_insn *cmd; 122298943Sluigi 122398943Sluigi /* look for action, in case it is a skipto */ 122498943Sluigi cmd = ACTION_PTR(me); 1225109566Smaxim if (cmd->opcode == O_LOG) 1226109566Smaxim cmd += F_LEN(cmd); 122798943Sluigi if ( cmd->opcode == O_SKIPTO ) 122898943Sluigi for (rule = me->next; rule ; rule = rule->next) 122998943Sluigi if (rule->rulenum >= cmd->arg1) 123098943Sluigi break; 123198943Sluigi if (rule == NULL) /* failure or not a skipto */ 123298943Sluigi rule = me->next; 123398943Sluigi me->next_rule = rule; 123498943Sluigi return rule; 123598943Sluigi} 123698943Sluigi 123798943Sluigi/* 123898943Sluigi * The main check routine for the firewall. 123998943Sluigi * 124098943Sluigi * All arguments are in args so we can modify them and return them 124198943Sluigi * back to the caller. 124298943Sluigi * 124398943Sluigi * Parameters: 124498943Sluigi * 124598943Sluigi * args->m (in/out) The packet; we set to NULL when/if we nuke it. 124698943Sluigi * Starts with the IP header. 124798943Sluigi * args->eh (in) Mac header if present, or NULL for layer3 packet. 124898943Sluigi * args->oif Outgoing interface, or NULL if packet is incoming. 124998943Sluigi * The incoming interface is in the mbuf. (in) 125098943Sluigi * args->divert_rule (in/out) 125198943Sluigi * Skip up to the first rule past this rule number; 125298943Sluigi * upon return, non-zero port number for divert or tee. 125398943Sluigi * 125498943Sluigi * args->rule Pointer to the last matching rule (in/out) 125598943Sluigi * args->next_hop Socket we are forwarding to (out). 125698943Sluigi * args->f_id Addresses grabbed from the packet (out) 125798943Sluigi * 125898943Sluigi * Return value: 125998943Sluigi * 126098943Sluigi * IP_FW_PORT_DENY_FLAG the packet must be dropped. 126198943Sluigi * 0 The packet is to be accepted and routed normally OR 126298943Sluigi * the packet was denied/rejected and has been dropped; 126398943Sluigi * in the latter case, *m is equal to NULL upon return. 126498943Sluigi * port Divert the packet to port, with these caveats: 126598943Sluigi * 126698943Sluigi * - If IP_FW_PORT_TEE_FLAG is set, tee the packet instead 126798943Sluigi * of diverting it (ie, 'ipfw tee'). 126898943Sluigi * 126998943Sluigi * - If IP_FW_PORT_DYNT_FLAG is set, interpret the lower 127098943Sluigi * 16 bits as a dummynet pipe number instead of diverting 127198943Sluigi */ 127298943Sluigi 1273105775Smaximstatic int 127498943Sluigiipfw_chk(struct ip_fw_args *args) 127598943Sluigi{ 127698943Sluigi /* 127798943Sluigi * Local variables hold state during the processing of a packet. 127898943Sluigi * 127998943Sluigi * IMPORTANT NOTE: to speed up the processing of rules, there 128098943Sluigi * are some assumption on the values of the variables, which 128198943Sluigi * are documented here. Should you change them, please check 128298943Sluigi * the implementation of the various instructions to make sure 128398943Sluigi * that they still work. 1284101978Sluigi * 128598943Sluigi * args->eh The MAC header. It is non-null for a layer2 128698943Sluigi * packet, it is NULL for a layer-3 packet. 128798943Sluigi * 128898943Sluigi * m | args->m Pointer to the mbuf, as received from the caller. 128998943Sluigi * It may change if ipfw_chk() does an m_pullup, or if it 129098943Sluigi * consumes the packet because it calls send_reject(). 129198943Sluigi * XXX This has to change, so that ipfw_chk() never modifies 129298943Sluigi * or consumes the buffer. 129398943Sluigi * ip is simply an alias of the value of m, and it is kept 129498943Sluigi * in sync with it (the packet is supposed to start with 129598943Sluigi * the ip header). 129698943Sluigi */ 129798943Sluigi struct mbuf *m = args->m; 129898943Sluigi struct ip *ip = mtod(m, struct ip *); 129998943Sluigi 130098943Sluigi /* 130198943Sluigi * oif | args->oif If NULL, ipfw_chk has been called on the 130298943Sluigi * inbound path (ether_input, bdg_forward, ip_input). 130398943Sluigi * If non-NULL, ipfw_chk has been called on the outbound path 130498943Sluigi * (ether_output, ip_output). 130598943Sluigi */ 130698943Sluigi struct ifnet *oif = args->oif; 130798943Sluigi 130898943Sluigi struct ip_fw *f = NULL; /* matching rule */ 130998943Sluigi int retval = 0; 131098943Sluigi 131198943Sluigi /* 131298943Sluigi * hlen The length of the IPv4 header. 131398943Sluigi * hlen >0 means we have an IPv4 packet. 131498943Sluigi */ 131598943Sluigi u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 131698943Sluigi 131798943Sluigi /* 131898943Sluigi * offset The offset of a fragment. offset != 0 means that 131998943Sluigi * we have a fragment at this offset of an IPv4 packet. 132098943Sluigi * offset == 0 means that (if this is an IPv4 packet) 132198943Sluigi * this is the first or only fragment. 132298943Sluigi */ 132398943Sluigi u_short offset = 0; 132498943Sluigi 132598943Sluigi /* 132698943Sluigi * Local copies of addresses. They are only valid if we have 132798943Sluigi * an IP packet. 132898943Sluigi * 132998943Sluigi * proto The protocol. Set to 0 for non-ip packets, 133098943Sluigi * or to the protocol read from the packet otherwise. 133198943Sluigi * proto != 0 means that we have an IPv4 packet. 133298943Sluigi * 133398943Sluigi * src_port, dst_port port numbers, in HOST format. Only 133498943Sluigi * valid for TCP and UDP packets. 133598943Sluigi * 133698943Sluigi * src_ip, dst_ip ip addresses, in NETWORK format. 133798943Sluigi * Only valid for IPv4 packets. 133898943Sluigi */ 133998943Sluigi u_int8_t proto; 134098943Sluigi u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 134198943Sluigi struct in_addr src_ip, dst_ip; /* NOTE: network format */ 134298943Sluigi u_int16_t ip_len=0; 134398943Sluigi int dyn_dir = MATCH_UNKNOWN; 134498943Sluigi ipfw_dyn_rule *q = NULL; 134598943Sluigi 1346101978Sluigi if (m->m_flags & M_SKIP_FIREWALL) 1347101978Sluigi return 0; /* accept */ 134898943Sluigi /* 134998943Sluigi * dyn_dir = MATCH_UNKNOWN when rules unchecked, 135098943Sluigi * MATCH_NONE when checked and not matched (q = NULL), 135198943Sluigi * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 135298943Sluigi */ 135398943Sluigi 135498943Sluigi if (args->eh == NULL || /* layer 3 packet */ 135598943Sluigi ( m->m_pkthdr.len >= sizeof(struct ip) && 135698943Sluigi ntohs(args->eh->ether_type) == ETHERTYPE_IP)) 135798943Sluigi hlen = ip->ip_hl << 2; 135898943Sluigi 135998943Sluigi /* 136098943Sluigi * Collect parameters into local variables for faster matching. 136198943Sluigi */ 136298943Sluigi if (hlen == 0) { /* do not grab addresses for non-ip pkts */ 136398943Sluigi proto = args->f_id.proto = 0; /* mark f_id invalid */ 136498943Sluigi goto after_ip_checks; 136598943Sluigi } 136698943Sluigi 136798943Sluigi proto = args->f_id.proto = ip->ip_p; 136898943Sluigi src_ip = ip->ip_src; 136998943Sluigi dst_ip = ip->ip_dst; 137098943Sluigi if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 137198943Sluigi offset = ntohs(ip->ip_off) & IP_OFFMASK; 137298943Sluigi ip_len = ntohs(ip->ip_len); 137398943Sluigi } else { 137498943Sluigi offset = ip->ip_off & IP_OFFMASK; 137598943Sluigi ip_len = ip->ip_len; 137698943Sluigi } 137798943Sluigi 137898943Sluigi#define PULLUP_TO(len) \ 137998943Sluigi do { \ 138098943Sluigi if ((m)->m_len < (len)) { \ 138198943Sluigi args->m = m = m_pullup(m, (len)); \ 138298943Sluigi if (m == 0) \ 138398943Sluigi goto pullup_failed; \ 138498943Sluigi ip = mtod(m, struct ip *); \ 138598943Sluigi } \ 138698943Sluigi } while (0) 138798943Sluigi 138898943Sluigi if (offset == 0) { 138998943Sluigi switch (proto) { 139098943Sluigi case IPPROTO_TCP: 139198943Sluigi { 139298943Sluigi struct tcphdr *tcp; 139398943Sluigi 139498943Sluigi PULLUP_TO(hlen + sizeof(struct tcphdr)); 139598943Sluigi tcp = L3HDR(struct tcphdr, ip); 139698943Sluigi dst_port = tcp->th_dport; 139798943Sluigi src_port = tcp->th_sport; 139898943Sluigi args->f_id.flags = tcp->th_flags; 139998943Sluigi } 140098943Sluigi break; 140198943Sluigi 140298943Sluigi case IPPROTO_UDP: 140398943Sluigi { 140498943Sluigi struct udphdr *udp; 140598943Sluigi 140698943Sluigi PULLUP_TO(hlen + sizeof(struct udphdr)); 140798943Sluigi udp = L3HDR(struct udphdr, ip); 140898943Sluigi dst_port = udp->uh_dport; 140998943Sluigi src_port = udp->uh_sport; 141098943Sluigi } 141198943Sluigi break; 141298943Sluigi 141398943Sluigi case IPPROTO_ICMP: 141498943Sluigi PULLUP_TO(hlen + 4); /* type, code and checksum. */ 141598943Sluigi args->f_id.flags = L3HDR(struct icmp, ip)->icmp_type; 141698943Sluigi break; 141798943Sluigi 141898943Sluigi default: 141998943Sluigi break; 142098943Sluigi } 142198943Sluigi#undef PULLUP_TO 142298943Sluigi } 142398943Sluigi 142498943Sluigi args->f_id.src_ip = ntohl(src_ip.s_addr); 142598943Sluigi args->f_id.dst_ip = ntohl(dst_ip.s_addr); 142698943Sluigi args->f_id.src_port = src_port = ntohs(src_port); 142798943Sluigi args->f_id.dst_port = dst_port = ntohs(dst_port); 142898943Sluigi 142998943Sluigiafter_ip_checks: 143098943Sluigi if (args->rule) { 143198943Sluigi /* 143298943Sluigi * Packet has already been tagged. Look for the next rule 143398943Sluigi * to restart processing. 143498943Sluigi * 143598943Sluigi * If fw_one_pass != 0 then just accept it. 143698943Sluigi * XXX should not happen here, but optimized out in 143798943Sluigi * the caller. 143898943Sluigi */ 143998943Sluigi if (fw_one_pass) 144098943Sluigi return 0; 144198943Sluigi 144298943Sluigi f = args->rule->next_rule; 144398943Sluigi if (f == NULL) 144498943Sluigi f = lookup_next_rule(args->rule); 144598943Sluigi } else { 144698943Sluigi /* 144798943Sluigi * Find the starting rule. It can be either the first 144898943Sluigi * one, or the one after divert_rule if asked so. 144998943Sluigi */ 145098943Sluigi int skipto = args->divert_rule; 145198943Sluigi 145298943Sluigi f = layer3_chain; 145398943Sluigi if (args->eh == NULL && skipto != 0) { 145498943Sluigi if (skipto >= IPFW_DEFAULT_RULE) 145598943Sluigi return(IP_FW_PORT_DENY_FLAG); /* invalid */ 145698943Sluigi while (f && f->rulenum <= skipto) 145798943Sluigi f = f->next; 145898943Sluigi if (f == NULL) /* drop packet */ 145998943Sluigi return(IP_FW_PORT_DENY_FLAG); 146098943Sluigi } 146198943Sluigi } 146298943Sluigi args->divert_rule = 0; /* reset to avoid confusion later */ 146398943Sluigi 146498943Sluigi /* 146598943Sluigi * Now scan the rules, and parse microinstructions for each rule. 146698943Sluigi */ 146798943Sluigi for (; f; f = f->next) { 146898943Sluigi int l, cmdlen; 146998943Sluigi ipfw_insn *cmd; 147098943Sluigi int skip_or; /* skip rest of OR block */ 147198943Sluigi 147298943Sluigiagain: 1473101628Sluigi if (set_disable & (1 << f->set) ) 1474101628Sluigi continue; 1475101628Sluigi 147698943Sluigi skip_or = 0; 147798943Sluigi for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 147898943Sluigi l -= cmdlen, cmd += cmdlen) { 147999622Sluigi int match; 148098943Sluigi 148198943Sluigi /* 148298943Sluigi * check_body is a jump target used when we find a 148398943Sluigi * CHECK_STATE, and need to jump to the body of 148498943Sluigi * the target rule. 148598943Sluigi */ 148698943Sluigi 148798943Sluigicheck_body: 148898943Sluigi cmdlen = F_LEN(cmd); 148998943Sluigi /* 149098943Sluigi * An OR block (insn_1 || .. || insn_n) has the 149198943Sluigi * F_OR bit set in all but the last instruction. 149298943Sluigi * The first match will set "skip_or", and cause 149398943Sluigi * the following instructions to be skipped until 149498943Sluigi * past the one with the F_OR bit clear. 149598943Sluigi */ 149698943Sluigi if (skip_or) { /* skip this instruction */ 149798943Sluigi if ((cmd->len & F_OR) == 0) 149898943Sluigi skip_or = 0; /* next one is good */ 149998943Sluigi continue; 150098943Sluigi } 150199622Sluigi match = 0; /* set to 1 if we succeed */ 150299622Sluigi 150398943Sluigi switch (cmd->opcode) { 150499622Sluigi /* 150599622Sluigi * The first set of opcodes compares the packet's 150699622Sluigi * fields with some pattern, setting 'match' if a 150799622Sluigi * match is found. At the end of the loop there is 150899622Sluigi * logic to deal with F_NOT and F_OR flags associated 150999622Sluigi * with the opcode. 151099622Sluigi */ 151198943Sluigi case O_NOP: 151299622Sluigi match = 1; 151399622Sluigi break; 151498943Sluigi 151598943Sluigi case O_FORWARD_MAC: 151698943Sluigi printf("ipfw: opcode %d unimplemented\n", 151798943Sluigi cmd->opcode); 151899622Sluigi break; 151998943Sluigi 152098943Sluigi case O_GID: 152198943Sluigi case O_UID: 152298943Sluigi /* 152398943Sluigi * We only check offset == 0 && proto != 0, 152498943Sluigi * as this ensures that we have an IPv4 152598943Sluigi * packet with the ports info. 152698943Sluigi */ 152798943Sluigi if (offset!=0) 152899622Sluigi break; 152998943Sluigi { 153098943Sluigi struct inpcbinfo *pi; 153198943Sluigi int wildcard; 153298943Sluigi struct inpcb *pcb; 153398943Sluigi 153498943Sluigi if (proto == IPPROTO_TCP) { 153598943Sluigi wildcard = 0; 153698943Sluigi pi = &tcbinfo; 153798943Sluigi } else if (proto == IPPROTO_UDP) { 153898943Sluigi wildcard = 1; 153998943Sluigi pi = &udbinfo; 154098943Sluigi } else 154199622Sluigi break; 154298943Sluigi 154398943Sluigi pcb = (oif) ? 154498943Sluigi in_pcblookup_hash(pi, 154598943Sluigi dst_ip, htons(dst_port), 154698943Sluigi src_ip, htons(src_port), 154798943Sluigi wildcard, oif) : 154898943Sluigi in_pcblookup_hash(pi, 154998943Sluigi src_ip, htons(src_port), 155098943Sluigi dst_ip, htons(dst_port), 155198943Sluigi wildcard, NULL); 155298943Sluigi 155398943Sluigi if (pcb == NULL || pcb->inp_socket == NULL) 155499622Sluigi break; 155599622Sluigi#if __FreeBSD_version < 500034 1556111037Smaxim#define socheckuid(a,b) ((a)->so_cred->cr_uid != (b)) 155799622Sluigi#endif 155898943Sluigi if (cmd->opcode == O_UID) { 155999622Sluigi match = 1560111037Smaxim !socheckuid(pcb->inp_socket, 156199622Sluigi (uid_t)((ipfw_insn_u32 *)cmd)->d[0]); 156298943Sluigi } else { 156399622Sluigi match = groupmember( 156498943Sluigi (uid_t)((ipfw_insn_u32 *)cmd)->d[0], 156599622Sluigi pcb->inp_socket->so_cred); 156698943Sluigi } 156798943Sluigi } 156899622Sluigi break; 156998943Sluigi 157098943Sluigi case O_RECV: 157199622Sluigi match = iface_match(m->m_pkthdr.rcvif, 157299622Sluigi (ipfw_insn_if *)cmd); 157399622Sluigi break; 157498943Sluigi 157598943Sluigi case O_XMIT: 157699622Sluigi match = iface_match(oif, (ipfw_insn_if *)cmd); 157799622Sluigi break; 157898943Sluigi 157998943Sluigi case O_VIA: 158099622Sluigi match = iface_match(oif ? oif : 158199622Sluigi m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 158299622Sluigi break; 158398943Sluigi 158498943Sluigi case O_MACADDR2: 158598943Sluigi if (args->eh != NULL) { /* have MAC header */ 158698943Sluigi u_int32_t *want = (u_int32_t *) 158798943Sluigi ((ipfw_insn_mac *)cmd)->addr; 158898943Sluigi u_int32_t *mask = (u_int32_t *) 158998943Sluigi ((ipfw_insn_mac *)cmd)->mask; 159098943Sluigi u_int32_t *hdr = (u_int32_t *)args->eh; 159198943Sluigi 159299622Sluigi match = 159399622Sluigi ( want[0] == (hdr[0] & mask[0]) && 159499622Sluigi want[1] == (hdr[1] & mask[1]) && 159599622Sluigi want[2] == (hdr[2] & mask[2]) ); 159698943Sluigi } 159799622Sluigi break; 159898943Sluigi 159998943Sluigi case O_MAC_TYPE: 160098943Sluigi if (args->eh != NULL) { 160199622Sluigi u_int16_t t = 160298943Sluigi ntohs(args->eh->ether_type); 160398943Sluigi u_int16_t *p = 160498943Sluigi ((ipfw_insn_u16 *)cmd)->ports; 160598943Sluigi int i; 160698943Sluigi 160799622Sluigi for (i = cmdlen - 1; !match && i>0; 160899622Sluigi i--, p += 2) 160999622Sluigi match = (t>=p[0] && t<=p[1]); 161098943Sluigi } 161199622Sluigi break; 161298943Sluigi 161398943Sluigi case O_FRAG: 161499622Sluigi match = (hlen > 0 && offset != 0); 161599622Sluigi break; 161698943Sluigi 161798943Sluigi case O_IN: /* "out" is "not in" */ 161899622Sluigi match = (oif == NULL); 161999622Sluigi break; 162098943Sluigi 162198943Sluigi case O_LAYER2: 162299622Sluigi match = (args->eh != NULL); 162399622Sluigi break; 162498943Sluigi 162598943Sluigi case O_PROTO: 162698943Sluigi /* 162798943Sluigi * We do not allow an arg of 0 so the 162898943Sluigi * check of "proto" only suffices. 162998943Sluigi */ 163099622Sluigi match = (proto == cmd->arg1); 163199622Sluigi break; 163298943Sluigi 163398943Sluigi case O_IP_SRC: 163499622Sluigi match = (hlen > 0 && 163598943Sluigi ((ipfw_insn_ip *)cmd)->addr.s_addr == 163699622Sluigi src_ip.s_addr); 163799622Sluigi break; 1638105775Smaxim 163998943Sluigi case O_IP_SRC_MASK: 164099622Sluigi match = (hlen > 0 && 164198943Sluigi ((ipfw_insn_ip *)cmd)->addr.s_addr == 164299622Sluigi (src_ip.s_addr & 164399622Sluigi ((ipfw_insn_ip *)cmd)->mask.s_addr)); 164499622Sluigi break; 164598943Sluigi 164698943Sluigi case O_IP_SRC_ME: 164799622Sluigi if (hlen > 0) { 164899622Sluigi struct ifnet *tif; 164998943Sluigi 165099622Sluigi INADDR_TO_IFP(src_ip, tif); 165199622Sluigi match = (tif != NULL); 165299622Sluigi } 165399622Sluigi break; 1654105775Smaxim 165598943Sluigi case O_IP_DST_SET: 165698943Sluigi case O_IP_SRC_SET: 165799622Sluigi if (hlen > 0) { 165899622Sluigi u_int32_t *d = (u_int32_t *)(cmd+1); 165999622Sluigi u_int32_t addr = 166099622Sluigi cmd->opcode == O_IP_DST_SET ? 1661105886Sluigi args->f_id.dst_ip : 1662105886Sluigi args->f_id.src_ip; 166398943Sluigi 166499622Sluigi if (addr < d[0]) 166599622Sluigi break; 166699622Sluigi addr -= d[0]; /* subtract base */ 166799622Sluigi match = (addr < cmd->arg1) && 166899622Sluigi ( d[ 1 + (addr>>5)] & 166999622Sluigi (1<<(addr & 0x1f)) ); 167099622Sluigi } 167199622Sluigi break; 167298943Sluigi 167398943Sluigi case O_IP_DST: 167499622Sluigi match = (hlen > 0 && 167598943Sluigi ((ipfw_insn_ip *)cmd)->addr.s_addr == 167699622Sluigi dst_ip.s_addr); 167799622Sluigi break; 167898943Sluigi 167998943Sluigi case O_IP_DST_MASK: 168099622Sluigi match = (hlen > 0) && 168199622Sluigi (((ipfw_insn_ip *)cmd)->addr.s_addr == 168299622Sluigi (dst_ip.s_addr & 168399622Sluigi ((ipfw_insn_ip *)cmd)->mask.s_addr)); 168499622Sluigi break; 168598943Sluigi 168698943Sluigi case O_IP_DST_ME: 168799622Sluigi if (hlen > 0) { 168899622Sluigi struct ifnet *tif; 168999622Sluigi 169099622Sluigi INADDR_TO_IFP(dst_ip, tif); 169199622Sluigi match = (tif != NULL); 169299622Sluigi } 169399622Sluigi break; 1694105775Smaxim 169598943Sluigi case O_IP_SRCPORT: 169698943Sluigi case O_IP_DSTPORT: 169798943Sluigi /* 169898943Sluigi * offset == 0 && proto != 0 is enough 169998943Sluigi * to guarantee that we have an IPv4 170098943Sluigi * packet with port info. 170198943Sluigi */ 170299622Sluigi if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 170399622Sluigi && offset == 0) { 170499622Sluigi u_int16_t x = 170598943Sluigi (cmd->opcode == O_IP_SRCPORT) ? 170699622Sluigi src_port : dst_port ; 170798943Sluigi u_int16_t *p = 170898943Sluigi ((ipfw_insn_u16 *)cmd)->ports; 170998943Sluigi int i; 171098943Sluigi 171199622Sluigi for (i = cmdlen - 1; !match && i>0; 171299622Sluigi i--, p += 2) 171399622Sluigi match = (x>=p[0] && x<=p[1]); 171498943Sluigi } 171599622Sluigi break; 171698943Sluigi 171798943Sluigi case O_ICMPTYPE: 171899622Sluigi match = (offset == 0 && proto==IPPROTO_ICMP && 171999622Sluigi icmptype_match(ip, (ipfw_insn_u32 *)cmd) ); 172099622Sluigi break; 172198943Sluigi 172298943Sluigi case O_IPOPT: 172399622Sluigi match = (hlen > 0 && ipopts_match(ip, cmd) ); 172499622Sluigi break; 172598943Sluigi 172698943Sluigi case O_IPVER: 172799622Sluigi match = (hlen > 0 && cmd->arg1 == ip->ip_v); 172899622Sluigi break; 172998943Sluigi 173098943Sluigi case O_IPTTL: 173199622Sluigi match = (hlen > 0 && cmd->arg1 == ip->ip_ttl); 173299622Sluigi break; 173398943Sluigi 173498943Sluigi case O_IPID: 173599622Sluigi match = (hlen > 0 && 173699622Sluigi cmd->arg1 == ntohs(ip->ip_id)); 173799622Sluigi break; 173898943Sluigi 173998943Sluigi case O_IPLEN: 174099622Sluigi match = (hlen > 0 && cmd->arg1 == ip_len); 174199622Sluigi break; 174298943Sluigi 174399475Sluigi case O_IPPRECEDENCE: 174499622Sluigi match = (hlen > 0 && 174599622Sluigi (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 174699622Sluigi break; 174799475Sluigi 174898943Sluigi case O_IPTOS: 174999622Sluigi match = (hlen > 0 && 175099622Sluigi flags_match(cmd, ip->ip_tos)); 175199622Sluigi break; 175298943Sluigi 175398943Sluigi case O_TCPFLAGS: 175499622Sluigi match = (proto == IPPROTO_TCP && offset == 0 && 175599622Sluigi flags_match(cmd, 175699622Sluigi L3HDR(struct tcphdr,ip)->th_flags)); 175799622Sluigi break; 175898943Sluigi 175998943Sluigi case O_TCPOPTS: 176099622Sluigi match = (proto == IPPROTO_TCP && offset == 0 && 176199622Sluigi tcpopts_match(ip, cmd)); 176299622Sluigi break; 176398943Sluigi 176498943Sluigi case O_TCPSEQ: 176599622Sluigi match = (proto == IPPROTO_TCP && offset == 0 && 176699622Sluigi ((ipfw_insn_u32 *)cmd)->d[0] == 176799622Sluigi L3HDR(struct tcphdr,ip)->th_seq); 176899622Sluigi break; 176998943Sluigi 177098943Sluigi case O_TCPACK: 177199622Sluigi match = (proto == IPPROTO_TCP && offset == 0 && 177299622Sluigi ((ipfw_insn_u32 *)cmd)->d[0] == 177399622Sluigi L3HDR(struct tcphdr,ip)->th_ack); 177499622Sluigi break; 177598943Sluigi 177698943Sluigi case O_TCPWIN: 177799622Sluigi match = (proto == IPPROTO_TCP && offset == 0 && 177899622Sluigi cmd->arg1 == 177999622Sluigi L3HDR(struct tcphdr,ip)->th_win); 178099622Sluigi break; 178198943Sluigi 178298943Sluigi case O_ESTAB: 178398943Sluigi /* reject packets which have SYN only */ 178499622Sluigi /* XXX should i also check for TH_ACK ? */ 178599622Sluigi match = (proto == IPPROTO_TCP && offset == 0 && 178699622Sluigi (L3HDR(struct tcphdr,ip)->th_flags & 178799622Sluigi (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 178899622Sluigi break; 178998943Sluigi 179098943Sluigi case O_LOG: 1791100589Sluigi if (fw_verbose) 1792100589Sluigi ipfw_log(f, hlen, args->eh, m, oif); 179399622Sluigi match = 1; 179499622Sluigi break; 179598943Sluigi 179699475Sluigi case O_PROB: 179799622Sluigi match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 179899622Sluigi break; 179998943Sluigi 1800112250Scjc case O_VERREVPATH: 1801112250Scjc /* Outgoing packets automatically pass/match */ 1802112250Scjc match = ((oif != NULL) || 1803112250Scjc (m->m_pkthdr.rcvif == NULL) || 1804112250Scjc verify_rev_path(src_ip, m->m_pkthdr.rcvif)); 1805112250Scjc break; 1806112250Scjc 180799622Sluigi /* 180899622Sluigi * The second set of opcodes represents 'actions', 180999622Sluigi * i.e. the terminal part of a rule once the packet 181099622Sluigi * matches all previous patterns. 181199622Sluigi * Typically there is only one action for each rule, 181299622Sluigi * and the opcode is stored at the end of the rule 181399622Sluigi * (but there are exceptions -- see below). 181499622Sluigi * 181599622Sluigi * In general, here we set retval and terminate the 181699622Sluigi * outer loop (would be a 'break 3' in some language, 181799622Sluigi * but we need to do a 'goto done'). 181899622Sluigi * 181999622Sluigi * Exceptions: 182099622Sluigi * O_COUNT and O_SKIPTO actions: 182199622Sluigi * instead of terminating, we jump to the next rule 182299622Sluigi * ('goto next_rule', equivalent to a 'break 2'), 182399622Sluigi * or to the SKIPTO target ('goto again' after 182499622Sluigi * having set f, cmd and l), respectively. 182599622Sluigi * 182699622Sluigi * O_LIMIT and O_KEEP_STATE: these opcodes are 182799622Sluigi * not real 'actions', and are stored right 182899622Sluigi * before the 'action' part of the rule. 182999622Sluigi * These opcodes try to install an entry in the 183099622Sluigi * state tables; if successful, we continue with 183199622Sluigi * the next opcode (match=1; break;), otherwise 183299622Sluigi * the packet * must be dropped 183399622Sluigi * ('goto done' after setting retval); 183499622Sluigi * 183599622Sluigi * O_PROBE_STATE and O_CHECK_STATE: these opcodes 183699622Sluigi * cause a lookup of the state table, and a jump 183799622Sluigi * to the 'action' part of the parent rule 183899622Sluigi * ('goto check_body') if an entry is found, or 183999622Sluigi * (CHECK_STATE only) a jump to the next rule if 184099622Sluigi * the entry is not found ('goto next_rule'). 184199622Sluigi * The result of the lookup is cached to make 184299622Sluigi * further instances of these opcodes are 184399622Sluigi * effectively NOPs. 184499622Sluigi */ 184598943Sluigi case O_LIMIT: 184698943Sluigi case O_KEEP_STATE: 184798943Sluigi if (install_state(f, 184899622Sluigi (ipfw_insn_limit *)cmd, args)) { 184999622Sluigi retval = IP_FW_PORT_DENY_FLAG; 185099622Sluigi goto done; /* error/limit violation */ 185199622Sluigi } 185299622Sluigi match = 1; 185399622Sluigi break; 185498943Sluigi 185598943Sluigi case O_PROBE_STATE: 185698943Sluigi case O_CHECK_STATE: 185798943Sluigi /* 185898943Sluigi * dynamic rules are checked at the first 185999622Sluigi * keep-state or check-state occurrence, 186099622Sluigi * with the result being stored in dyn_dir. 186199622Sluigi * The compiler introduces a PROBE_STATE 186298943Sluigi * instruction for us when we have a 186399622Sluigi * KEEP_STATE (because PROBE_STATE needs 186498943Sluigi * to be run first). 186598943Sluigi */ 186699622Sluigi if (dyn_dir == MATCH_UNKNOWN && 186799622Sluigi (q = lookup_dyn_rule(&args->f_id, 1868100004Sluigi &dyn_dir, proto == IPPROTO_TCP ? 1869100004Sluigi L3HDR(struct tcphdr, ip) : NULL)) 1870100004Sluigi != NULL) { 187199622Sluigi /* 187299622Sluigi * Found dynamic entry, update stats 187399622Sluigi * and jump to the 'action' part of 187499622Sluigi * the parent rule. 187599622Sluigi */ 187699622Sluigi q->pcnt++; 187799622Sluigi q->bcnt += ip_len; 187899622Sluigi f = q->rule; 187999622Sluigi cmd = ACTION_PTR(f); 188099622Sluigi l = f->cmd_len - f->act_ofs; 188199622Sluigi goto check_body; 188298943Sluigi } 188399622Sluigi /* 188499622Sluigi * Dynamic entry not found. If CHECK_STATE, 188599622Sluigi * skip to next rule, if PROBE_STATE just 188699622Sluigi * ignore and continue with next opcode. 188799622Sluigi */ 188898943Sluigi if (cmd->opcode == O_CHECK_STATE) 188998943Sluigi goto next_rule; 189099622Sluigi match = 1; 189199622Sluigi break; 189298943Sluigi 189398943Sluigi case O_ACCEPT: 189498943Sluigi retval = 0; /* accept */ 189599622Sluigi goto done; 189698943Sluigi 189798943Sluigi case O_PIPE: 189898943Sluigi case O_QUEUE: 189998943Sluigi args->rule = f; /* report matching rule */ 190098943Sluigi retval = cmd->arg1 | IP_FW_PORT_DYNT_FLAG; 190199622Sluigi goto done; 1902105775Smaxim 190398943Sluigi case O_DIVERT: 190498943Sluigi case O_TEE: 190598943Sluigi if (args->eh) /* not on layer 2 */ 190699622Sluigi break; 190798943Sluigi args->divert_rule = f->rulenum; 190899622Sluigi retval = (cmd->opcode == O_DIVERT) ? 190999622Sluigi cmd->arg1 : 191099622Sluigi cmd->arg1 | IP_FW_PORT_TEE_FLAG; 191199622Sluigi goto done; 191298943Sluigi 191398943Sluigi case O_COUNT: 191498943Sluigi case O_SKIPTO: 191598943Sluigi f->pcnt++; /* update stats */ 191698943Sluigi f->bcnt += ip_len; 191798943Sluigi f->timestamp = time_second; 191898943Sluigi if (cmd->opcode == O_COUNT) 191998943Sluigi goto next_rule; 192098943Sluigi /* handle skipto */ 192198943Sluigi if (f->next_rule == NULL) 192298943Sluigi lookup_next_rule(f); 192398943Sluigi f = f->next_rule; 192498943Sluigi goto again; 192598943Sluigi 192698943Sluigi case O_REJECT: 192798943Sluigi /* 192898943Sluigi * Drop the packet and send a reject notice 192998943Sluigi * if the packet is not ICMP (or is an ICMP 193098943Sluigi * query), and it is not multicast/broadcast. 193198943Sluigi */ 193298943Sluigi if (hlen > 0 && 193398943Sluigi (proto != IPPROTO_ICMP || 193499475Sluigi is_icmp_query(ip)) && 193598943Sluigi !(m->m_flags & (M_BCAST|M_MCAST)) && 193698943Sluigi !IN_MULTICAST(dst_ip.s_addr)) { 193799475Sluigi send_reject(args, cmd->arg1, 193899475Sluigi offset,ip_len); 193999475Sluigi m = args->m; 194098943Sluigi } 194199622Sluigi /* FALLTHROUGH */ 194299622Sluigi case O_DENY: 194399622Sluigi retval = IP_FW_PORT_DENY_FLAG; 194499622Sluigi goto done; 194598943Sluigi 194698943Sluigi case O_FORWARD_IP: 194798943Sluigi if (args->eh) /* not valid on layer2 pkts */ 194899622Sluigi break; 194998943Sluigi if (!q || dyn_dir == MATCH_FORWARD) 195098943Sluigi args->next_hop = 195198943Sluigi &((ipfw_insn_sa *)cmd)->sa; 195298943Sluigi retval = 0; 195399622Sluigi goto done; 195498943Sluigi 195598943Sluigi default: 195698943Sluigi panic("-- unknown opcode %d\n", cmd->opcode); 195799622Sluigi } /* end of switch() on opcodes */ 195898943Sluigi 195999622Sluigi if (cmd->len & F_NOT) 196099622Sluigi match = !match; 196198943Sluigi 196299622Sluigi if (match) { 196399622Sluigi if (cmd->len & F_OR) 196499622Sluigi skip_or = 1; 196599622Sluigi } else { 196699622Sluigi if (!(cmd->len & F_OR)) /* not an OR block, */ 196799622Sluigi break; /* try next rule */ 196898943Sluigi } 196998943Sluigi 197098943Sluigi } /* end of inner for, scan opcodes */ 197198943Sluigi 197298965Sdfrnext_rule:; /* try next rule */ 1973105775Smaxim 197498943Sluigi } /* end of outer for, scan rules */ 1975108258Smaxim printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 1976101628Sluigi return(IP_FW_PORT_DENY_FLAG); 197798943Sluigi 197899622Sluigidone: 197998943Sluigi /* Update statistics */ 198098943Sluigi f->pcnt++; 198198943Sluigi f->bcnt += ip_len; 198298943Sluigi f->timestamp = time_second; 198398943Sluigi return retval; 198498943Sluigi 198598943Sluigipullup_failed: 198698943Sluigi if (fw_verbose) 1987108258Smaxim printf("ipfw: pullup failed\n"); 198898943Sluigi return(IP_FW_PORT_DENY_FLAG); 198998943Sluigi} 199098943Sluigi 199198943Sluigi/* 199298943Sluigi * When a rule is added/deleted, clear the next_rule pointers in all rules. 199398943Sluigi * These will be reconstructed on the fly as packets are matched. 199498943Sluigi * Must be called at splimp(). 199598943Sluigi */ 199698943Sluigistatic void 199798943Sluigiflush_rule_ptrs(void) 199898943Sluigi{ 199998943Sluigi struct ip_fw *rule; 200098943Sluigi 200198943Sluigi for (rule = layer3_chain; rule; rule = rule->next) 200298943Sluigi rule->next_rule = NULL; 200398943Sluigi} 200498943Sluigi 200598943Sluigi/* 200698943Sluigi * When pipes/queues are deleted, clear the "pipe_ptr" pointer to a given 200798943Sluigi * pipe/queue, or to all of them (match == NULL). 200898943Sluigi * Must be called at splimp(). 200998943Sluigi */ 201098943Sluigivoid 201198943Sluigiflush_pipe_ptrs(struct dn_flow_set *match) 201298943Sluigi{ 201398943Sluigi struct ip_fw *rule; 201498943Sluigi 201598943Sluigi for (rule = layer3_chain; rule; rule = rule->next) { 201698943Sluigi ipfw_insn_pipe *cmd = (ipfw_insn_pipe *)ACTION_PTR(rule); 2017105775Smaxim 201898943Sluigi if (cmd->o.opcode != O_PIPE && cmd->o.opcode != O_QUEUE) 201998943Sluigi continue; 202098943Sluigi if (match == NULL || cmd->pipe_ptr == match) 202198943Sluigi cmd->pipe_ptr = NULL; 202298943Sluigi } 202398943Sluigi} 202498943Sluigi 202598943Sluigi/* 202698943Sluigi * Add a new rule to the list. Copy the rule into a malloc'ed area, then 202798943Sluigi * possibly create a rule number and add the rule to the list. 202898943Sluigi * Update the rule_number in the input struct so the caller knows it as well. 202998943Sluigi */ 203098943Sluigistatic int 203198943Sluigiadd_rule(struct ip_fw **head, struct ip_fw *input_rule) 203298943Sluigi{ 203398943Sluigi struct ip_fw *rule, *f, *prev; 203498943Sluigi int s; 203598943Sluigi int l = RULESIZE(input_rule); 203698943Sluigi 203798943Sluigi if (*head == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 203898943Sluigi return (EINVAL); 203998943Sluigi 2040105440Smux rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 204198943Sluigi if (rule == NULL) 204298943Sluigi return (ENOSPC); 204398943Sluigi 204498943Sluigi bcopy(input_rule, rule, l); 204598943Sluigi 204698943Sluigi rule->next = NULL; 204798943Sluigi rule->next_rule = NULL; 204898943Sluigi 204998943Sluigi rule->pcnt = 0; 205098943Sluigi rule->bcnt = 0; 205198943Sluigi rule->timestamp = 0; 205298943Sluigi 205398943Sluigi s = splimp(); 205498943Sluigi 205598943Sluigi if (*head == NULL) { /* default rule */ 205698943Sluigi *head = rule; 205798943Sluigi goto done; 205898943Sluigi } 205998943Sluigi 206098943Sluigi /* 206198943Sluigi * If rulenum is 0, find highest numbered rule before the 206298943Sluigi * default rule, and add autoinc_step 206398943Sluigi */ 206498943Sluigi if (autoinc_step < 1) 206598943Sluigi autoinc_step = 1; 206698943Sluigi else if (autoinc_step > 1000) 206798943Sluigi autoinc_step = 1000; 206898943Sluigi if (rule->rulenum == 0) { 206998943Sluigi /* 207098943Sluigi * locate the highest numbered rule before default 207198943Sluigi */ 207298943Sluigi for (f = *head; f; f = f->next) { 207398943Sluigi if (f->rulenum == IPFW_DEFAULT_RULE) 207498943Sluigi break; 207598943Sluigi rule->rulenum = f->rulenum; 207698943Sluigi } 207798943Sluigi if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step) 207898943Sluigi rule->rulenum += autoinc_step; 207998943Sluigi input_rule->rulenum = rule->rulenum; 208098943Sluigi } 208198943Sluigi 208298943Sluigi /* 208398943Sluigi * Now insert the new rule in the right place in the sorted list. 208498943Sluigi */ 208598943Sluigi for (prev = NULL, f = *head; f; prev = f, f = f->next) { 208698943Sluigi if (f->rulenum > rule->rulenum) { /* found the location */ 208798943Sluigi if (prev) { 208898943Sluigi rule->next = f; 208998943Sluigi prev->next = rule; 209098943Sluigi } else { /* head insert */ 209198943Sluigi rule->next = *head; 209298943Sluigi *head = rule; 209398943Sluigi } 209498943Sluigi break; 209598943Sluigi } 209698943Sluigi } 209798943Sluigi flush_rule_ptrs(); 209898943Sluigidone: 209998943Sluigi static_count++; 210098943Sluigi static_len += l; 210198943Sluigi splx(s); 2102108258Smaxim DEB(printf("ipfw: installed rule %d, static count now %d\n", 210398943Sluigi rule->rulenum, static_count);) 210498943Sluigi return (0); 210598943Sluigi} 210698943Sluigi 210798943Sluigi/** 210898943Sluigi * Free storage associated with a static rule (including derived 210998943Sluigi * dynamic rules). 211098943Sluigi * The caller is in charge of clearing rule pointers to avoid 211198943Sluigi * dangling pointers. 211298943Sluigi * @return a pointer to the next entry. 211398943Sluigi * Arguments are not checked, so they better be correct. 211498943Sluigi * Must be called at splimp(). 211598943Sluigi */ 211698943Sluigistatic struct ip_fw * 211798943Sluigidelete_rule(struct ip_fw **head, struct ip_fw *prev, struct ip_fw *rule) 211898943Sluigi{ 211998943Sluigi struct ip_fw *n; 212098943Sluigi int l = RULESIZE(rule); 212198943Sluigi 212298943Sluigi n = rule->next; 212398943Sluigi remove_dyn_rule(rule, NULL /* force removal */); 212498943Sluigi if (prev == NULL) 212598943Sluigi *head = n; 212698943Sluigi else 212798943Sluigi prev->next = n; 212898943Sluigi static_count--; 212998943Sluigi static_len -= l; 213098943Sluigi 213198943Sluigi if (DUMMYNET_LOADED) 213298943Sluigi ip_dn_ruledel_ptr(rule); 213398943Sluigi free(rule, M_IPFW); 213498943Sluigi return n; 213598943Sluigi} 213698943Sluigi 213798943Sluigi/* 213898943Sluigi * Deletes all rules from a chain (including the default rule 213998943Sluigi * if the second argument is set). 214098943Sluigi * Must be called at splimp(). 214198943Sluigi */ 214298943Sluigistatic void 214398943Sluigifree_chain(struct ip_fw **chain, int kill_default) 214498943Sluigi{ 214598943Sluigi struct ip_fw *rule; 214698943Sluigi 214798943Sluigi flush_rule_ptrs(); /* more efficient to do outside the loop */ 214898943Sluigi 214998943Sluigi while ( (rule = *chain) != NULL && 215098943Sluigi (kill_default || rule->rulenum != IPFW_DEFAULT_RULE) ) 215198943Sluigi delete_rule(chain, NULL, rule); 215298943Sluigi} 215398943Sluigi 215498943Sluigi/** 2155101628Sluigi * Remove all rules with given number, and also do set manipulation. 2156101628Sluigi * 2157101978Sluigi * The argument is an u_int32_t. The low 16 bit are the rule or set number, 2158101978Sluigi * the next 8 bits are the new set, the top 8 bits are the command: 2159105775Smaxim * 2160101978Sluigi * 0 delete rules with given number 2161101978Sluigi * 1 delete rules with given set number 2162101978Sluigi * 2 move rules with given number to new set 2163101978Sluigi * 3 move rules with given set number to new set 2164101978Sluigi * 4 swap sets with given numbers 216598943Sluigi */ 216698943Sluigistatic int 2167101628Sluigidel_entry(struct ip_fw **chain, u_int32_t arg) 216898943Sluigi{ 216998943Sluigi struct ip_fw *prev, *rule; 217098943Sluigi int s; 2171101978Sluigi u_int16_t rulenum; 2172101978Sluigi u_int8_t cmd, new_set; 217398943Sluigi 2174101628Sluigi rulenum = arg & 0xffff; 2175101978Sluigi cmd = (arg >> 24) & 0xff; 2176101978Sluigi new_set = (arg >> 16) & 0xff; 2177101628Sluigi 2178101978Sluigi if (cmd > 4) 217998943Sluigi return EINVAL; 2180101978Sluigi if (new_set > 30) 2181101628Sluigi return EINVAL; 2182101978Sluigi if (cmd == 0 || cmd == 2) { 2183101978Sluigi if (rulenum == IPFW_DEFAULT_RULE) 2184101978Sluigi return EINVAL; 2185101978Sluigi } else { 2186101978Sluigi if (rulenum > 30) 2187101978Sluigi return EINVAL; 2188101978Sluigi } 2189105775Smaxim 2190101628Sluigi switch (cmd) { 2191101978Sluigi case 0: /* delete rules with given number */ 2192101628Sluigi /* 2193101628Sluigi * locate first rule to delete 2194101628Sluigi */ 2195101628Sluigi for (prev = NULL, rule = *chain; 2196101628Sluigi rule && rule->rulenum < rulenum; 2197101628Sluigi prev = rule, rule = rule->next) 2198101628Sluigi ; 2199101628Sluigi if (rule->rulenum != rulenum) 2200101628Sluigi return EINVAL; 2201101628Sluigi 2202101628Sluigi s = splimp(); /* no access to rules while removing */ 2203101628Sluigi /* 2204101978Sluigi * flush pointers outside the loop, then delete all matching 2205101978Sluigi * rules. prev remains the same throughout the cycle. 2206101628Sluigi */ 2207101978Sluigi flush_rule_ptrs(); 2208101628Sluigi while (rule && rule->rulenum == rulenum) 2209101628Sluigi rule = delete_rule(chain, prev, rule); 2210101628Sluigi splx(s); 2211101628Sluigi break; 2212101628Sluigi 2213101978Sluigi case 1: /* delete all rules with given set number */ 2214101978Sluigi s = splimp(); 2215101978Sluigi flush_rule_ptrs(); 2216101628Sluigi for (prev = NULL, rule = *chain; rule ; ) 2217101628Sluigi if (rule->set == rulenum) 2218101628Sluigi rule = delete_rule(chain, prev, rule); 2219101628Sluigi else { 2220101628Sluigi prev = rule; 2221101628Sluigi rule = rule->next; 2222101628Sluigi } 2223101628Sluigi splx(s); 2224101628Sluigi break; 2225101628Sluigi 2226101978Sluigi case 2: /* move rules with given number to new set */ 2227101628Sluigi s = splimp(); 2228101978Sluigi for (rule = *chain; rule ; rule = rule->next) 2229101978Sluigi if (rule->rulenum == rulenum) 2230101978Sluigi rule->set = new_set; 2231101628Sluigi splx(s); 2232101628Sluigi break; 2233101628Sluigi 2234101978Sluigi case 3: /* move rules with given set number to new set */ 2235101628Sluigi s = splimp(); 2236101978Sluigi for (rule = *chain; rule ; rule = rule->next) 2237101978Sluigi if (rule->set == rulenum) 2238101978Sluigi rule->set = new_set; 2239101628Sluigi splx(s); 2240101628Sluigi break; 2241101978Sluigi 2242101978Sluigi case 4: /* swap two sets */ 2243101978Sluigi s = splimp(); 2244101978Sluigi for (rule = *chain; rule ; rule = rule->next) 2245101978Sluigi if (rule->set == rulenum) 2246101978Sluigi rule->set = new_set; 2247101978Sluigi else if (rule->set == new_set) 2248101978Sluigi rule->set = rulenum; 2249101978Sluigi splx(s); 2250101978Sluigi break; 2251101628Sluigi } 225298943Sluigi return 0; 225398943Sluigi} 225498943Sluigi 225598943Sluigi/* 225698943Sluigi * Clear counters for a specific rule. 225798943Sluigi */ 225898943Sluigistatic void 225998943Sluigiclear_counters(struct ip_fw *rule, int log_only) 226098943Sluigi{ 226198943Sluigi ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 226298943Sluigi 226398943Sluigi if (log_only == 0) { 226498943Sluigi rule->bcnt = rule->pcnt = 0; 226598943Sluigi rule->timestamp = 0; 226698943Sluigi } 226798943Sluigi if (l->o.opcode == O_LOG) 226898943Sluigi l->log_left = l->max_log; 226998943Sluigi} 227098943Sluigi 227198943Sluigi/** 227298943Sluigi * Reset some or all counters on firewall rules. 227398943Sluigi * @arg frwl is null to clear all entries, or contains a specific 227498943Sluigi * rule number. 227598943Sluigi * @arg log_only is 1 if we only want to reset logs, zero otherwise. 227698943Sluigi */ 227798943Sluigistatic int 227898943Sluigizero_entry(int rulenum, int log_only) 227998943Sluigi{ 228098943Sluigi struct ip_fw *rule; 228198943Sluigi int s; 228298943Sluigi char *msg; 228398943Sluigi 228498943Sluigi if (rulenum == 0) { 228598943Sluigi s = splimp(); 228698943Sluigi norule_counter = 0; 228798943Sluigi for (rule = layer3_chain; rule; rule = rule->next) 228898943Sluigi clear_counters(rule, log_only); 228998943Sluigi splx(s); 229098943Sluigi msg = log_only ? "ipfw: All logging counts reset.\n" : 229198943Sluigi "ipfw: Accounting cleared.\n"; 229298943Sluigi } else { 229398943Sluigi int cleared = 0; 229498943Sluigi /* 229598943Sluigi * We can have multiple rules with the same number, so we 229698943Sluigi * need to clear them all. 229798943Sluigi */ 229898943Sluigi for (rule = layer3_chain; rule; rule = rule->next) 229998943Sluigi if (rule->rulenum == rulenum) { 230098943Sluigi s = splimp(); 230198943Sluigi while (rule && rule->rulenum == rulenum) { 230298943Sluigi clear_counters(rule, log_only); 230398943Sluigi rule = rule->next; 230498943Sluigi } 230598943Sluigi splx(s); 230698943Sluigi cleared = 1; 230798943Sluigi break; 230898943Sluigi } 230998943Sluigi if (!cleared) /* we did not find any matching rules */ 231098943Sluigi return (EINVAL); 231198943Sluigi msg = log_only ? "ipfw: Entry %d logging count reset.\n" : 231298943Sluigi "ipfw: Entry %d cleared.\n"; 231398943Sluigi } 231498943Sluigi if (fw_verbose) 231598943Sluigi log(LOG_SECURITY | LOG_NOTICE, msg, rulenum); 231698943Sluigi return (0); 231798943Sluigi} 231898943Sluigi 231998943Sluigi/* 232098943Sluigi * Check validity of the structure before insert. 232198943Sluigi * Fortunately rules are simple, so this mostly need to check rule sizes. 232298943Sluigi */ 232398943Sluigistatic int 232498943Sluigicheck_ipfw_struct(struct ip_fw *rule, int size) 232598943Sluigi{ 232698943Sluigi int l, cmdlen = 0; 232798943Sluigi int have_action=0; 232898943Sluigi ipfw_insn *cmd; 232998943Sluigi 233098943Sluigi if (size < sizeof(*rule)) { 233199622Sluigi printf("ipfw: rule too short\n"); 233298943Sluigi return (EINVAL); 233398943Sluigi } 233498943Sluigi /* first, check for valid size */ 233598943Sluigi l = RULESIZE(rule); 233698943Sluigi if (l != size) { 233799622Sluigi printf("ipfw: size mismatch (have %d want %d)\n", size, l); 233898943Sluigi return (EINVAL); 233998943Sluigi } 234098943Sluigi /* 234198943Sluigi * Now go for the individual checks. Very simple ones, basically only 234298943Sluigi * instruction sizes. 234398943Sluigi */ 234498943Sluigi for (l = rule->cmd_len, cmd = rule->cmd ; 234598943Sluigi l > 0 ; l -= cmdlen, cmd += cmdlen) { 234698943Sluigi cmdlen = F_LEN(cmd); 234798943Sluigi if (cmdlen > l) { 234899622Sluigi printf("ipfw: opcode %d size truncated\n", 234998943Sluigi cmd->opcode); 235098943Sluigi return EINVAL; 235198943Sluigi } 235299622Sluigi DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 235398943Sluigi switch (cmd->opcode) { 235498943Sluigi case O_NOP: 235598943Sluigi case O_PROBE_STATE: 235698943Sluigi case O_KEEP_STATE: 235798943Sluigi case O_PROTO: 235898943Sluigi case O_IP_SRC_ME: 235998943Sluigi case O_IP_DST_ME: 236098943Sluigi case O_LAYER2: 236198943Sluigi case O_IN: 236298943Sluigi case O_FRAG: 236398943Sluigi case O_IPOPT: 236498943Sluigi case O_IPLEN: 236598943Sluigi case O_IPID: 236698943Sluigi case O_IPTOS: 236799475Sluigi case O_IPPRECEDENCE: 236898943Sluigi case O_IPTTL: 236998943Sluigi case O_IPVER: 237098943Sluigi case O_TCPWIN: 237198943Sluigi case O_TCPFLAGS: 237298943Sluigi case O_TCPOPTS: 237398943Sluigi case O_ESTAB: 2374112250Scjc case O_VERREVPATH: 237598943Sluigi if (cmdlen != F_INSN_SIZE(ipfw_insn)) 237698943Sluigi goto bad_size; 237798943Sluigi break; 237898943Sluigi 237998943Sluigi case O_UID: 238098943Sluigi case O_GID: 238198943Sluigi case O_IP_SRC: 238298943Sluigi case O_IP_DST: 238398943Sluigi case O_TCPSEQ: 238498943Sluigi case O_TCPACK: 238598943Sluigi case O_PROB: 238698943Sluigi case O_ICMPTYPE: 238798943Sluigi if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 238898943Sluigi goto bad_size; 238998943Sluigi break; 239098943Sluigi 239198943Sluigi case O_LIMIT: 239298943Sluigi if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 239398943Sluigi goto bad_size; 239498943Sluigi break; 239598943Sluigi 239698943Sluigi case O_LOG: 239798943Sluigi if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 239898943Sluigi goto bad_size; 2399105775Smaxim 240098943Sluigi ((ipfw_insn_log *)cmd)->log_left = 240198943Sluigi ((ipfw_insn_log *)cmd)->max_log; 240298943Sluigi 240398943Sluigi break; 240498943Sluigi 240598943Sluigi case O_IP_SRC_MASK: 240698943Sluigi case O_IP_DST_MASK: 240798943Sluigi if (cmdlen != F_INSN_SIZE(ipfw_insn_ip)) 240898943Sluigi goto bad_size; 240998943Sluigi if (((ipfw_insn_ip *)cmd)->mask.s_addr == 0) { 241099622Sluigi printf("ipfw: opcode %d, useless rule\n", 241198943Sluigi cmd->opcode); 241298943Sluigi return EINVAL; 241398943Sluigi } 241498943Sluigi break; 241598943Sluigi 241698943Sluigi case O_IP_SRC_SET: 241798943Sluigi case O_IP_DST_SET: 241898943Sluigi if (cmd->arg1 == 0 || cmd->arg1 > 256) { 241999622Sluigi printf("ipfw: invalid set size %d\n", 242098943Sluigi cmd->arg1); 242198943Sluigi return EINVAL; 242298943Sluigi } 242398943Sluigi if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 242498943Sluigi (cmd->arg1+31)/32 ) 242598943Sluigi goto bad_size; 242698943Sluigi break; 242798943Sluigi 242898943Sluigi case O_MACADDR2: 242998943Sluigi if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 243098943Sluigi goto bad_size; 243198943Sluigi break; 243298943Sluigi 243398943Sluigi case O_MAC_TYPE: 243498943Sluigi case O_IP_SRCPORT: 2435102086Sluigi case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 2436102086Sluigi if (cmdlen < 2 || cmdlen > 31) 243798943Sluigi goto bad_size; 243898943Sluigi break; 243998943Sluigi 244098943Sluigi case O_RECV: 244198943Sluigi case O_XMIT: 244298943Sluigi case O_VIA: 244398943Sluigi if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 244498943Sluigi goto bad_size; 244598943Sluigi break; 244698943Sluigi 244798943Sluigi case O_PIPE: 244898943Sluigi case O_QUEUE: 244998943Sluigi if (cmdlen != F_INSN_SIZE(ipfw_insn_pipe)) 245098943Sluigi goto bad_size; 245198943Sluigi goto check_action; 245298943Sluigi 245399475Sluigi case O_FORWARD_IP: 245498943Sluigi if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 245598943Sluigi goto bad_size; 245698943Sluigi goto check_action; 245798943Sluigi 245899475Sluigi case O_FORWARD_MAC: /* XXX not implemented yet */ 245998943Sluigi case O_CHECK_STATE: 246098943Sluigi case O_COUNT: 246198943Sluigi case O_ACCEPT: 246298943Sluigi case O_DENY: 246398943Sluigi case O_REJECT: 246498943Sluigi case O_SKIPTO: 246598943Sluigi case O_DIVERT: 246698943Sluigi case O_TEE: 246798943Sluigi if (cmdlen != F_INSN_SIZE(ipfw_insn)) 246898943Sluigi goto bad_size; 246998943Sluigicheck_action: 247098943Sluigi if (have_action) { 247199622Sluigi printf("ipfw: opcode %d, multiple actions" 247298943Sluigi " not allowed\n", 247398943Sluigi cmd->opcode); 247498943Sluigi return EINVAL; 247598943Sluigi } 247698943Sluigi have_action = 1; 247798943Sluigi if (l != cmdlen) { 247899622Sluigi printf("ipfw: opcode %d, action must be" 247998943Sluigi " last opcode\n", 248098943Sluigi cmd->opcode); 248198943Sluigi return EINVAL; 248298943Sluigi } 248398943Sluigi break; 248498943Sluigi default: 248599622Sluigi printf("ipfw: opcode %d, unknown opcode\n", 248698943Sluigi cmd->opcode); 248798943Sluigi return EINVAL; 248898943Sluigi } 248998943Sluigi } 249098943Sluigi if (have_action == 0) { 249199622Sluigi printf("ipfw: missing action\n"); 249298943Sluigi return EINVAL; 249398943Sluigi } 249498943Sluigi return 0; 249598943Sluigi 249698943Sluigibad_size: 249799622Sluigi printf("ipfw: opcode %d size %d wrong\n", 249898943Sluigi cmd->opcode, cmdlen); 249998943Sluigi return EINVAL; 250098943Sluigi} 250198943Sluigi 250298943Sluigi 250398943Sluigi/** 250498943Sluigi * {set|get}sockopt parser. 250598943Sluigi */ 250698943Sluigistatic int 250798943Sluigiipfw_ctl(struct sockopt *sopt) 250898943Sluigi{ 250998943Sluigi int error, s, rulenum; 251098943Sluigi size_t size; 251198943Sluigi struct ip_fw *bp , *buf, *rule; 251298943Sluigi 251398943Sluigi static u_int32_t rule_buf[255]; /* we copy the data here */ 251498943Sluigi 251598943Sluigi /* 251698943Sluigi * Disallow modifications in really-really secure mode, but still allow 251798943Sluigi * the logging counters to be reset. 251898943Sluigi */ 251998943Sluigi if (sopt->sopt_name == IP_FW_ADD || 252098943Sluigi (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 252198943Sluigi#if __FreeBSD_version >= 500034 252298943Sluigi error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 252398943Sluigi if (error) 252498943Sluigi return (error); 252598943Sluigi#else /* FreeBSD 4.x */ 252698943Sluigi if (securelevel >= 3) 252798943Sluigi return (EPERM); 252898943Sluigi#endif 252998943Sluigi } 253098943Sluigi 253198943Sluigi error = 0; 253298943Sluigi 253398943Sluigi switch (sopt->sopt_name) { 253498943Sluigi case IP_FW_GET: 253598943Sluigi /* 253698943Sluigi * pass up a copy of the current rules. Static rules 253798943Sluigi * come first (the last of which has number IPFW_DEFAULT_RULE), 253898943Sluigi * followed by a possibly empty list of dynamic rule. 253998943Sluigi * The last dynamic rule has NULL in the "next" field. 254098943Sluigi */ 254198943Sluigi s = splimp(); 254298943Sluigi size = static_len; /* size of static rules */ 254398943Sluigi if (ipfw_dyn_v) /* add size of dyn.rules */ 254498943Sluigi size += (dyn_count * sizeof(ipfw_dyn_rule)); 254598943Sluigi 254698943Sluigi /* 254798943Sluigi * XXX todo: if the user passes a short length just to know 254898943Sluigi * how much room is needed, do not bother filling up the 254998943Sluigi * buffer, just jump to the sooptcopyout. 255098943Sluigi */ 2551111119Simp buf = malloc(size, M_TEMP, M_WAITOK); 255298943Sluigi if (buf == 0) { 255398943Sluigi splx(s); 255498943Sluigi error = ENOBUFS; 255598943Sluigi break; 255698943Sluigi } 255798943Sluigi 255898943Sluigi bp = buf; 255998943Sluigi for (rule = layer3_chain; rule ; rule = rule->next) { 256098943Sluigi int i = RULESIZE(rule); 256198943Sluigi bcopy(rule, bp, i); 2562105887Smux ((struct ip_fw *)bp)->set_disable = set_disable; 256398943Sluigi bp = (struct ip_fw *)((char *)bp + i); 256498943Sluigi } 256598943Sluigi if (ipfw_dyn_v) { 256698943Sluigi int i; 256798943Sluigi ipfw_dyn_rule *p, *dst, *last = NULL; 256898943Sluigi 256998943Sluigi dst = (ipfw_dyn_rule *)bp; 257098943Sluigi for (i = 0 ; i < curr_dyn_buckets ; i++ ) 257198943Sluigi for ( p = ipfw_dyn_v[i] ; p != NULL ; 257298943Sluigi p = p->next, dst++ ) { 257398943Sluigi bcopy(p, dst, sizeof *p); 2574105887Smux dst->rulenum = p->rule->rulenum; 257598943Sluigi /* 257698943Sluigi * store a non-null value in "next". 257798943Sluigi * The userland code will interpret a 257898943Sluigi * NULL here as a marker 257998943Sluigi * for the last dynamic rule. 258098943Sluigi */ 258198943Sluigi dst->next = dst ; 258298943Sluigi last = dst ; 258398943Sluigi dst->expire = 258498943Sluigi TIME_LEQ(dst->expire, time_second) ? 258598943Sluigi 0 : dst->expire - time_second ; 258698943Sluigi } 258798943Sluigi if (last != NULL) /* mark last dynamic rule */ 258898943Sluigi last->next = NULL; 258998943Sluigi } 259098943Sluigi splx(s); 259198943Sluigi 259298943Sluigi error = sooptcopyout(sopt, buf, size); 259398943Sluigi free(buf, M_TEMP); 259498943Sluigi break; 259598943Sluigi 259698943Sluigi case IP_FW_FLUSH: 259798943Sluigi /* 259898943Sluigi * Normally we cannot release the lock on each iteration. 259998943Sluigi * We could do it here only because we start from the head all 260098943Sluigi * the times so there is no risk of missing some entries. 260198943Sluigi * On the other hand, the risk is that we end up with 260298943Sluigi * a very inconsistent ruleset, so better keep the lock 260398943Sluigi * around the whole cycle. 2604105775Smaxim * 260598943Sluigi * XXX this code can be improved by resetting the head of 260698943Sluigi * the list to point to the default rule, and then freeing 260798943Sluigi * the old list without the need for a lock. 260898943Sluigi */ 260998943Sluigi 261098943Sluigi s = splimp(); 261198943Sluigi free_chain(&layer3_chain, 0 /* keep default rule */); 261298943Sluigi splx(s); 261398943Sluigi break; 261498943Sluigi 261598943Sluigi case IP_FW_ADD: 261698943Sluigi rule = (struct ip_fw *)rule_buf; /* XXX do a malloc */ 261798943Sluigi error = sooptcopyin(sopt, rule, sizeof(rule_buf), 261898943Sluigi sizeof(struct ip_fw) ); 261998943Sluigi size = sopt->sopt_valsize; 262098943Sluigi if (error || (error = check_ipfw_struct(rule, size))) 262198943Sluigi break; 262298943Sluigi 262398943Sluigi error = add_rule(&layer3_chain, rule); 262498943Sluigi size = RULESIZE(rule); 262598943Sluigi if (!error && sopt->sopt_dir == SOPT_GET) 262698943Sluigi error = sooptcopyout(sopt, rule, size); 262798943Sluigi break; 262898943Sluigi 2629101978Sluigi case IP_FW_DEL: 2630101628Sluigi /* 2631101978Sluigi * IP_FW_DEL is used for deleting single rules or sets, 2632101978Sluigi * and (ab)used to atomically manipulate sets. Argument size 2633101978Sluigi * is used to distinguish between the two: 2634101978Sluigi * sizeof(u_int32_t) 2635101978Sluigi * delete single rule or set of rules, 2636101978Sluigi * or reassign rules (or sets) to a different set. 2637101978Sluigi * 2*sizeof(u_int32_t) 2638101978Sluigi * atomic disable/enable sets. 2639101978Sluigi * first u_int32_t contains sets to be disabled, 2640101978Sluigi * second u_int32_t contains sets to be enabled. 2641101628Sluigi */ 2642101978Sluigi error = sooptcopyin(sopt, rule_buf, 2643101978Sluigi 2*sizeof(u_int32_t), sizeof(u_int32_t)); 264498943Sluigi if (error) 264598943Sluigi break; 2646101978Sluigi size = sopt->sopt_valsize; 2647101978Sluigi if (size == sizeof(u_int32_t)) /* delete or reassign */ 2648101978Sluigi error = del_entry(&layer3_chain, rule_buf[0]); 2649101978Sluigi else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 2650101978Sluigi set_disable = 2651101978Sluigi (set_disable | rule_buf[0]) & ~rule_buf[1] & 2652101978Sluigi ~(1<<31); /* set 31 always enabled */ 2653101978Sluigi else 265498943Sluigi error = EINVAL; 265598943Sluigi break; 265698943Sluigi 265798943Sluigi case IP_FW_ZERO: 265898943Sluigi case IP_FW_RESETLOG: /* argument is an int, the rule number */ 265998943Sluigi rulenum=0; 266098943Sluigi 266198943Sluigi if (sopt->sopt_val != 0) { 266298943Sluigi error = sooptcopyin(sopt, &rulenum, 266398943Sluigi sizeof(int), sizeof(int)); 266498943Sluigi if (error) 266598943Sluigi break; 266698943Sluigi } 266798943Sluigi error = zero_entry(rulenum, sopt->sopt_name == IP_FW_RESETLOG); 266898943Sluigi break; 266998943Sluigi 267098943Sluigi default: 2671108258Smaxim printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 267298943Sluigi error = EINVAL; 267398943Sluigi } 267498943Sluigi 267598943Sluigi return (error); 267698943Sluigi} 267798943Sluigi 267898943Sluigi/** 267998943Sluigi * dummynet needs a reference to the default rule, because rules can be 268098943Sluigi * deleted while packets hold a reference to them. When this happens, 268198943Sluigi * dummynet changes the reference to the default rule (it could well be a 268298943Sluigi * NULL pointer, but this way we do not need to check for the special 268398943Sluigi * case, plus here he have info on the default behaviour). 268498943Sluigi */ 268598943Sluigistruct ip_fw *ip_fw_default_rule; 268698943Sluigi 2687101978Sluigi/* 2688101978Sluigi * This procedure is only used to handle keepalives. It is invoked 2689101978Sluigi * every dyn_keepalive_period 2690101978Sluigi */ 269198943Sluigistatic void 2692100004Sluigiipfw_tick(void * __unused unused) 2693100004Sluigi{ 2694100004Sluigi int i; 2695100004Sluigi int s; 2696100004Sluigi ipfw_dyn_rule *q; 2697100004Sluigi 2698100004Sluigi if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0) 2699100004Sluigi goto done; 2700100004Sluigi 2701100004Sluigi s = splimp(); 2702100004Sluigi for (i = 0 ; i < curr_dyn_buckets ; i++) { 2703100004Sluigi for (q = ipfw_dyn_v[i] ; q ; q = q->next ) { 2704100004Sluigi if (q->dyn_type == O_LIMIT_PARENT) 2705100004Sluigi continue; 2706100004Sluigi if (q->id.proto != IPPROTO_TCP) 2707100004Sluigi continue; 2708100004Sluigi if ( (q->state & BOTH_SYN) != BOTH_SYN) 2709100004Sluigi continue; 2710101978Sluigi if (TIME_LEQ( time_second+dyn_keepalive_interval, 2711101978Sluigi q->expire)) 2712100004Sluigi continue; /* too early */ 2713100004Sluigi if (TIME_LEQ(q->expire, time_second)) 2714100004Sluigi continue; /* too late, rule expired */ 2715100004Sluigi 2716100004Sluigi send_pkt(&(q->id), q->ack_rev - 1, q->ack_fwd, TH_SYN); 2717100004Sluigi send_pkt(&(q->id), q->ack_fwd - 1, q->ack_rev, 0); 2718100004Sluigi } 2719100004Sluigi } 2720100004Sluigi splx(s); 2721100004Sluigidone: 2722101978Sluigi ipfw_timeout_h = timeout(ipfw_tick, NULL, dyn_keepalive_period*hz); 2723100004Sluigi} 2724100004Sluigi 2725100004Sluigistatic void 272698943Sluigiipfw_init(void) 272798943Sluigi{ 272898943Sluigi struct ip_fw default_rule; 272998943Sluigi 273098943Sluigi ip_fw_chk_ptr = ipfw_chk; 273198943Sluigi ip_fw_ctl_ptr = ipfw_ctl; 273298943Sluigi layer3_chain = NULL; 273398943Sluigi 273498943Sluigi bzero(&default_rule, sizeof default_rule); 273598943Sluigi 273698943Sluigi default_rule.act_ofs = 0; 273798943Sluigi default_rule.rulenum = IPFW_DEFAULT_RULE; 273898943Sluigi default_rule.cmd_len = 1; 2739101628Sluigi default_rule.set = 31; 274098943Sluigi 274198943Sluigi default_rule.cmd[0].len = 1; 274298943Sluigi default_rule.cmd[0].opcode = 274398943Sluigi#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 274498943Sluigi 1 ? O_ACCEPT : 274598943Sluigi#endif 274698943Sluigi O_DENY; 274798943Sluigi 274898943Sluigi add_rule(&layer3_chain, &default_rule); 274998943Sluigi 275098943Sluigi ip_fw_default_rule = layer3_chain; 2751102086Sluigi printf("ipfw2 initialized, divert %s, " 2752102086Sluigi "rule-based forwarding enabled, default to %s, logging ", 275398943Sluigi#ifdef IPDIVERT 275498943Sluigi "enabled", 275598943Sluigi#else 275698943Sluigi "disabled", 275798943Sluigi#endif 275898943Sluigi default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny"); 275998943Sluigi 276098943Sluigi#ifdef IPFIREWALL_VERBOSE 276198943Sluigi fw_verbose = 1; 276298943Sluigi#endif 276398943Sluigi#ifdef IPFIREWALL_VERBOSE_LIMIT 276498943Sluigi verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 276598943Sluigi#endif 276698943Sluigi if (fw_verbose == 0) 276798943Sluigi printf("disabled\n"); 276898943Sluigi else if (verbose_limit == 0) 276998943Sluigi printf("unlimited\n"); 277098943Sluigi else 277198943Sluigi printf("limited to %d packets/entry by default\n", 277298943Sluigi verbose_limit); 2773100004Sluigi bzero(&ipfw_timeout_h, sizeof(struct callout_handle)); 2774100004Sluigi ipfw_timeout_h = timeout(ipfw_tick, NULL, hz); 277598943Sluigi} 277698943Sluigi 277798943Sluigistatic int 277898943Sluigiipfw_modevent(module_t mod, int type, void *unused) 277998943Sluigi{ 278098943Sluigi int s; 278198943Sluigi int err = 0; 2782105775Smaxim 278398943Sluigi switch (type) { 278498943Sluigi case MOD_LOAD: 278598943Sluigi s = splimp(); 278698943Sluigi if (IPFW_LOADED) { 278798943Sluigi splx(s); 278898943Sluigi printf("IP firewall already loaded\n"); 278998943Sluigi err = EEXIST; 279098943Sluigi } else { 279198943Sluigi ipfw_init(); 279298943Sluigi splx(s); 279398943Sluigi } 279498943Sluigi break; 279598943Sluigi 279698943Sluigi case MOD_UNLOAD: 279798943Sluigi#if !defined(KLD_MODULE) 279898943Sluigi printf("ipfw statically compiled, cannot unload\n"); 279998943Sluigi err = EBUSY; 280098943Sluigi#else 280198943Sluigi s = splimp(); 2802100004Sluigi untimeout(ipfw_tick, NULL, ipfw_timeout_h); 280398943Sluigi ip_fw_chk_ptr = NULL; 280498943Sluigi ip_fw_ctl_ptr = NULL; 280598943Sluigi free_chain(&layer3_chain, 1 /* kill default rule */); 280698943Sluigi splx(s); 280798943Sluigi printf("IP firewall unloaded\n"); 280898943Sluigi#endif 280998943Sluigi break; 281098943Sluigi default: 281198943Sluigi break; 281298943Sluigi } 281398943Sluigi return err; 281498943Sluigi} 281598943Sluigi 281698943Sluigistatic moduledata_t ipfwmod = { 281798943Sluigi "ipfw", 281898943Sluigi ipfw_modevent, 281998943Sluigi 0 282098943Sluigi}; 282198943SluigiDECLARE_MODULE(ipfw, ipfwmod, SI_SUB_PSEUDO, SI_ORDER_ANY); 282298943SluigiMODULE_VERSION(ipfw, 1); 282399622Sluigi#endif /* IPFW2 */ 2824