ip_fw2.c revision 196451
162587Sitojun/*- 262587Sitojun * Copyright (c) 2002 Luigi Rizzo, Universita` di Pisa 362587Sitojun * 453541Sshin * Redistribution and use in source and binary forms, with or without 553541Sshin * modification, are permitted provided that the following conditions 653541Sshin * are met: 753541Sshin * 1. Redistributions of source code must retain the above copyright 853541Sshin * notice, this list of conditions and the following disclaimer. 953541Sshin * 2. Redistributions in binary form must reproduce the above copyright 1053541Sshin * notice, this list of conditions and the following disclaimer in the 1153541Sshin * documentation and/or other materials provided with the distribution. 1253541Sshin * 1353541Sshin * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 1453541Sshin * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 1553541Sshin * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 1653541Sshin * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 1753541Sshin * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 1853541Sshin * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 1953541Sshin * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 2053541Sshin * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 2153541Sshin * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 2253541Sshin * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 2353541Sshin * SUCH DAMAGE. 2453541Sshin */ 2553541Sshin 2653541Sshin#include <sys/cdefs.h> 2753541Sshin__FBSDID("$FreeBSD: head/sys/netinet/ipfw/ip_fw2.c 196451 2009-08-23 07:59:28Z julian $"); 2853541Sshin 2953541Sshin#define DEB(x) 3053541Sshin#define DDB(x) x 3153541Sshin 3253541Sshin/* 3353541Sshin * Implement IP packet firewall (new version) 3453541Sshin */ 3553541Sshin 3653541Sshin#if !defined(KLD_MODULE) 3753541Sshin#include "opt_ipfw.h" 3853541Sshin#include "opt_ipdivert.h" 3953541Sshin#include "opt_ipdn.h" 4053541Sshin#include "opt_inet.h" 4153541Sshin#ifndef INET 4253541Sshin#error IPFIREWALL requires INET. 4353541Sshin#endif /* INET */ 4453541Sshin#endif 4553541Sshin#include "opt_inet6.h" 4653541Sshin#include "opt_ipsec.h" 4753541Sshin 4853541Sshin#include <sys/param.h> 4953541Sshin#include <sys/systm.h> 5062587Sitojun#include <sys/condvar.h> 5153541Sshin#include <sys/eventhandler.h> 5262587Sitojun#include <sys/malloc.h> 5353541Sshin#include <sys/mbuf.h> 5453541Sshin#include <sys/kernel.h> 5553541Sshin#include <sys/lock.h> 5653541Sshin#include <sys/jail.h> 5753541Sshin#include <sys/module.h> 5853541Sshin#include <sys/priv.h> 5953541Sshin#include <sys/proc.h> 6053541Sshin#include <sys/rwlock.h> 6162587Sitojun#include <sys/socket.h> 6253541Sshin#include <sys/socketvar.h> 6362587Sitojun#include <sys/sysctl.h> 6462587Sitojun#include <sys/syslog.h> 6562587Sitojun#include <sys/ucred.h> 6662587Sitojun#include <net/ethernet.h> /* for ETHERTYPE_IP */ 6762587Sitojun#include <net/if.h> 6853541Sshin#include <net/radix.h> 6962587Sitojun#include <net/route.h> 7062587Sitojun#include <net/pf_mtag.h> 7162587Sitojun#include <net/vnet.h> 7253541Sshin 7362587Sitojun#define IPFW_INTERNAL /* Access to protected data structures in ip_fw.h. */ 7469774Sphk 7562587Sitojun#include <netinet/in.h> 7653541Sshin#include <netinet/in_var.h> 7753541Sshin#include <netinet/in_pcb.h> 7853541Sshin#include <netinet/ip.h> 7953541Sshin#include <netinet/ip_var.h> 8053541Sshin#include <netinet/ip_icmp.h> 8153541Sshin#include <netinet/ip_fw.h> 8253541Sshin#include <netinet/ip_divert.h> 8353541Sshin#include <netinet/ip_dummynet.h> 8453541Sshin#include <netinet/ip_carp.h> 8553541Sshin#include <netinet/pim.h> 8653541Sshin#include <netinet/tcp_var.h> 8753541Sshin#include <netinet/udp.h> 8853541Sshin#include <netinet/udp_var.h> 8962587Sitojun#include <netinet/sctp.h> 9053541Sshin 9153541Sshin#include <netgraph/ng_ipfw.h> 9253541Sshin 9353541Sshin#include <netinet/ip6.h> 9462587Sitojun#include <netinet/icmp6.h> 9562587Sitojun#ifdef INET6 9662587Sitojun#include <netinet6/scope6_var.h> 9762587Sitojun#endif 9862587Sitojun 9962587Sitojun#include <machine/in_cksum.h> /* XXX for in_cksum */ 10062587Sitojun 10162587Sitojun#ifdef MAC 10262587Sitojun#include <security/mac/mac_framework.h> 10362587Sitojun#endif 10462587Sitojun 10562587Sitojunstatic VNET_DEFINE(int, ipfw_vnet_ready) = 0; 10662587Sitojun#define V_ipfw_vnet_ready VNET(ipfw_vnet_ready) 10762587Sitojun/* 10862587Sitojun * set_disable contains one bit per set value (0..31). 10962587Sitojun * If the bit is set, all rules with the corresponding set 11062587Sitojun * are disabled. Set RESVD_SET(31) is reserved for the default rule 11162587Sitojun * and rules that are not deleted by the flush command, 11262587Sitojun * and CANNOT be disabled. 11362587Sitojun * Rules in set RESVD_SET can only be deleted explicitly. 11462587Sitojun */ 11562587Sitojunstatic VNET_DEFINE(u_int32_t, set_disable); 11662587Sitojunstatic VNET_DEFINE(int, fw_verbose); 11762587Sitojunstatic VNET_DEFINE(struct callout, ipfw_timeout); 11862587Sitojunstatic VNET_DEFINE(int, verbose_limit); 11962587Sitojun 12062587Sitojun#define V_set_disable VNET(set_disable) 12162587Sitojun#define V_fw_verbose VNET(fw_verbose) 12262587Sitojun#define V_ipfw_timeout VNET(ipfw_timeout) 12353541Sshin#define V_verbose_limit VNET(verbose_limit) 12453541Sshin 12553541Sshin#ifdef IPFIREWALL_DEFAULT_TO_ACCEPT 12653541Sshinstatic int default_to_accept = 1; 12753541Sshin#else 12853541Sshinstatic int default_to_accept; 12953541Sshin#endif 13053541Sshinstatic uma_zone_t ipfw_dyn_rule_zone; 13153541Sshin 13253541Sshinstruct ip_fw *ip_fw_default_rule; 13353541Sshin 13462587Sitojun/* 13553541Sshin * list of rules for layer 3 13653541Sshin */ 13762587SitojunVNET_DEFINE(struct ip_fw_chain, layer3_chain); 13853541Sshin 13953541SshinMALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's"); 14053541SshinMALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables"); 14153541Sshin#define IPFW_NAT_LOADED (ipfw_nat_ptr != NULL) 14253541Sshinipfw_nat_t *ipfw_nat_ptr = NULL; 14353541Sshinipfw_nat_cfg_t *ipfw_nat_cfg_ptr; 14462587Sitojunipfw_nat_cfg_t *ipfw_nat_del_ptr; 14562587Sitojunipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr; 14653541Sshinipfw_nat_cfg_t *ipfw_nat_get_log_ptr; 14753541Sshin 14862587Sitojunstruct table_entry { 14962587Sitojun struct radix_node rn[2]; 15062587Sitojun struct sockaddr_in addr, mask; 15162587Sitojun u_int32_t value; 15262587Sitojun}; 15353541Sshin 15453541Sshinstatic VNET_DEFINE(int, autoinc_step); 15553541Sshin#define V_autoinc_step VNET(autoinc_step) 15653541Sshinstatic VNET_DEFINE(int, fw_deny_unknown_exthdrs); 15753541Sshin#define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs) 15853541Sshin 15953541Sshinextern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS); 16053541Sshin 16153541Sshin#ifdef SYSCTL_NODE 16253541SshinSYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 16353541SshinSYSCTL_VNET_PROC(_net_inet_ip_fw, OID_AUTO, enable, 16453541Sshin CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_enable), 0, 16553541Sshin ipfw_chg_hook, "I", "Enable ipfw"); 16653541SshinSYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, 16753541Sshin CTLFLAG_RW, &VNET_NAME(autoinc_step), 0, 16853541Sshin "Rule number auto-increment step"); 16953541SshinSYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, one_pass, 17054350Sshin CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0, 17153541Sshin "Only do a single pass through ipfw when using dummynet(4)"); 17253541SshinSYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose, 17353541Sshin CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0, 17453541Sshin "Log matches to ipfw rules"); 17553541SshinSYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, 17653541Sshin CTLFLAG_RW, &VNET_NAME(verbose_limit), 0, 17753541Sshin "Set upper limit of matches of ipfw rules logged"); 17853541SshinSYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD, 17953541Sshin NULL, IPFW_DEFAULT_RULE, 18053541Sshin "The default/max possible rule number."); 18153541SshinSYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, tables_max, CTLFLAG_RD, 18253541Sshin NULL, IPFW_TABLES_MAX, 18353541Sshin "The maximum number of tables."); 18453541SshinSYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN, 18553541Sshin &default_to_accept, 0, 18653541Sshin "Make the default rule accept all packets."); 18753541SshinTUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept); 18853541Sshin#ifdef INET6 18953541SshinSYSCTL_DECL(_net_inet6_ip6); 19053541SshinSYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall"); 19153541SshinSYSCTL_VNET_PROC(_net_inet6_ip6_fw, OID_AUTO, enable, 19253541Sshin CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw6_enable), 0, 19353541Sshin ipfw_chg_hook, "I", "Enable ipfw+6"); 19453541SshinSYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs, 19553541Sshin CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0, 19662587Sitojun "Deny packets with unknown IPv6 Extension Headers"); 19753541Sshin#endif 19853541Sshin#endif 19953541Sshin 20053541Sshin/* 20153541Sshin * Description of dynamic rules. 20253541Sshin * 20353541Sshin * Dynamic rules are stored in lists accessed through a hash table 20462587Sitojun * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can 20553541Sshin * be modified through the sysctl variable dyn_buckets which is 20653541Sshin * updated when the table becomes empty. 20753541Sshin * 20853541Sshin * XXX currently there is only one list, ipfw_dyn. 20953541Sshin * 21053541Sshin * When a packet is received, its address fields are first masked 21153541Sshin * with the mask defined for the rule, then hashed, then matched 21253541Sshin * against the entries in the corresponding list. 21353541Sshin * Dynamic rules can be used for different purposes: 21453541Sshin * + stateful rules; 21553541Sshin * + enforcing limits on the number of sessions; 21653541Sshin * + in-kernel NAT (not implemented yet) 21753541Sshin * 21853541Sshin * The lifetime of dynamic rules is regulated by dyn_*_lifetime, 21953541Sshin * measured in seconds and depending on the flags. 22053541Sshin * 22153541Sshin * The total number of dynamic rules is stored in dyn_count. 22253541Sshin * The max number of dynamic rules is dyn_max. When we reach 22353541Sshin * the maximum number of rules we do not create anymore. This is 22453541Sshin * done to avoid consuming too much memory, but also too much 22553541Sshin * time when searching on each packet (ideally, we should try instead 22653541Sshin * to put a limit on the length of the list on each bucket...). 22753541Sshin * 22853541Sshin * Each dynamic rule holds a pointer to the parent ipfw rule so 22953541Sshin * we know what action to perform. Dynamic rules are removed when 23053541Sshin * the parent rule is deleted. XXX we should make them survive. 23153541Sshin * 23253541Sshin * There are some limitations with dynamic rules -- we do not 23353541Sshin * obey the 'randomized match', and we do not do multiple 23453541Sshin * passes through the firewall. XXX check the latter!!! 23562587Sitojun */ 23653541Sshinstatic VNET_DEFINE(ipfw_dyn_rule **, ipfw_dyn_v); 23753541Sshinstatic VNET_DEFINE(u_int32_t, dyn_buckets); 23853541Sshinstatic VNET_DEFINE(u_int32_t, curr_dyn_buckets); 23962587Sitojun 24053541Sshin#define V_ipfw_dyn_v VNET(ipfw_dyn_v) 24162587Sitojun#define V_dyn_buckets VNET(dyn_buckets) 24262587Sitojun#define V_curr_dyn_buckets VNET(curr_dyn_buckets) 24362587Sitojun 24453541Sshinstatic struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */ 24553541Sshin#define IPFW_DYN_LOCK_INIT() \ 24653541Sshin mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF) 24753541Sshin#define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx) 24853541Sshin#define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx) 24953541Sshin#define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx) 25053541Sshin#define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED) 25153541Sshin 25253541Sshin/* 25353541Sshin * Timeouts for various events in handing dynamic rules. 25453541Sshin */ 25553541Sshinstatic VNET_DEFINE(u_int32_t, dyn_ack_lifetime); 25653541Sshinstatic VNET_DEFINE(u_int32_t, dyn_syn_lifetime); 25753541Sshinstatic VNET_DEFINE(u_int32_t, dyn_fin_lifetime); 25853541Sshinstatic VNET_DEFINE(u_int32_t, dyn_rst_lifetime); 25953541Sshinstatic VNET_DEFINE(u_int32_t, dyn_udp_lifetime); 26053541Sshinstatic VNET_DEFINE(u_int32_t, dyn_short_lifetime); 26153541Sshin 26253541Sshin#define V_dyn_ack_lifetime VNET(dyn_ack_lifetime) 26353541Sshin#define V_dyn_syn_lifetime VNET(dyn_syn_lifetime) 26453541Sshin#define V_dyn_fin_lifetime VNET(dyn_fin_lifetime) 26553541Sshin#define V_dyn_rst_lifetime VNET(dyn_rst_lifetime) 26653541Sshin#define V_dyn_udp_lifetime VNET(dyn_udp_lifetime) 26753541Sshin#define V_dyn_short_lifetime VNET(dyn_short_lifetime) 26862587Sitojun 26953541Sshin/* 27053541Sshin * Keepalives are sent if dyn_keepalive is set. They are sent every 27153541Sshin * dyn_keepalive_period seconds, in the last dyn_keepalive_interval 27253541Sshin * seconds of lifetime of a rule. 27362587Sitojun * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower 27462587Sitojun * than dyn_keepalive_period. 27553541Sshin */ 27653541Sshin 27753541Sshinstatic VNET_DEFINE(u_int32_t, dyn_keepalive_interval); 27853541Sshinstatic VNET_DEFINE(u_int32_t, dyn_keepalive_period); 27953541Sshinstatic VNET_DEFINE(u_int32_t, dyn_keepalive); 28062587Sitojun 28162587Sitojun#define V_dyn_keepalive_interval VNET(dyn_keepalive_interval) 28253541Sshin#define V_dyn_keepalive_period VNET(dyn_keepalive_period) 28353541Sshin#define V_dyn_keepalive VNET(dyn_keepalive) 28453541Sshin 28553541Sshinstatic VNET_DEFINE(u_int32_t, static_count); /* # of static rules */ 28653541Sshinstatic VNET_DEFINE(u_int32_t, static_len); /* bytes of static rules */ 28753541Sshinstatic VNET_DEFINE(u_int32_t, dyn_count); /* # of dynamic rules */ 28853541Sshinstatic VNET_DEFINE(u_int32_t, dyn_max); /* max # of dynamic rules */ 28953541Sshin 29053541Sshin#define V_static_count VNET(static_count) 29153541Sshin#define V_static_len VNET(static_len) 29253541Sshin#define V_dyn_count VNET(dyn_count) 29353541Sshin#define V_dyn_max VNET(dyn_max) 29453541Sshin 29553541Sshin#ifdef SYSCTL_NODE 29653541SshinSYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, 29753541Sshin CTLFLAG_RW, &VNET_NAME(dyn_buckets), 0, 29853541Sshin "Number of dyn. buckets"); 29953541SshinSYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, 30053541Sshin CTLFLAG_RD, &VNET_NAME(curr_dyn_buckets), 0, 30162587Sitojun "Current Number of dyn. buckets"); 30253541SshinSYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, 30353541Sshin CTLFLAG_RD, &VNET_NAME(dyn_count), 0, 30453541Sshin "Number of dyn. rules"); 30553541SshinSYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, 30653541Sshin CTLFLAG_RW, &VNET_NAME(dyn_max), 0, 30753541Sshin "Max number of dyn. rules"); 30853541SshinSYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, static_count, 30953541Sshin CTLFLAG_RD, &VNET_NAME(static_count), 0, 31053541Sshin "Number of static rules"); 31153541SshinSYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, 31253541Sshin CTLFLAG_RW, &VNET_NAME(dyn_ack_lifetime), 0, 31353541Sshin "Lifetime of dyn. rules for acks"); 31453541SshinSYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, 31562587Sitojun CTLFLAG_RW, &VNET_NAME(dyn_syn_lifetime), 0, 31662587Sitojun "Lifetime of dyn. rules for syn"); 31753541SshinSYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, 31853541Sshin CTLFLAG_RW, &VNET_NAME(dyn_fin_lifetime), 0, 31953541Sshin "Lifetime of dyn. rules for fin"); 32053541SshinSYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, 32162587Sitojun CTLFLAG_RW, &VNET_NAME(dyn_rst_lifetime), 0, 32262587Sitojun "Lifetime of dyn. rules for rst"); 32362587SitojunSYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, 32462587Sitojun CTLFLAG_RW, &VNET_NAME(dyn_udp_lifetime), 0, 32562587Sitojun "Lifetime of dyn. rules for UDP"); 32662587SitojunSYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, 32762587Sitojun CTLFLAG_RW, &VNET_NAME(dyn_short_lifetime), 0, 32862587Sitojun "Lifetime of dyn. rules for other situations"); 32962587SitojunSYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, 33053541Sshin CTLFLAG_RW, &VNET_NAME(dyn_keepalive), 0, 33153541Sshin "Enable keepalives for dyn. rules"); 33253541Sshin#endif /* SYSCTL_NODE */ 33353541Sshin 33453541Sshin/* 33553541Sshin * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T 33653541Sshin * Other macros just cast void * into the appropriate type 33753541Sshin */ 33853541Sshin#define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl)) 33953541Sshin#define TCP(p) ((struct tcphdr *)(p)) 34053541Sshin#define SCTP(p) ((struct sctphdr *)(p)) 34153541Sshin#define UDP(p) ((struct udphdr *)(p)) 34253541Sshin#define ICMP(p) ((struct icmphdr *)(p)) 34353541Sshin#define ICMP6(p) ((struct icmp6_hdr *)(p)) 34453541Sshin 34553541Sshinstatic __inline int 34653541Sshinicmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd) 34753541Sshin{ 34853541Sshin int type = icmp->icmp_type; 34962587Sitojun 35053541Sshin return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) ); 35162587Sitojun} 35262587Sitojun 35362587Sitojun#define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \ 35462587Sitojun (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) ) 35562587Sitojun 35662587Sitojunstatic int 35762587Sitojunis_icmp_query(struct icmphdr *icmp) 35862587Sitojun{ 35962587Sitojun int type = icmp->icmp_type; 36062587Sitojun 36162587Sitojun return (type <= ICMP_MAXTYPE && (TT & (1<<type)) ); 36262587Sitojun} 36362587Sitojun#undef TT 36462587Sitojun 36562587Sitojun/* 36662587Sitojun * The following checks use two arrays of 8 or 16 bits to store the 36762587Sitojun * bits that we want set or clear, respectively. They are in the 36862587Sitojun * low and high half of cmd->arg1 or cmd->d[0]. 36962587Sitojun * 37062587Sitojun * We scan options and store the bits we find set. We succeed if 37162587Sitojun * 37262587Sitojun * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear 37362587Sitojun * 37462587Sitojun * The code is sometimes optimized not to store additional variables. 37562587Sitojun */ 37662587Sitojun 37762587Sitojunstatic int 37862587Sitojunflags_match(ipfw_insn *cmd, u_int8_t bits) 37962587Sitojun{ 38062587Sitojun u_char want_clear; 38162587Sitojun bits = ~bits; 38262587Sitojun 38362587Sitojun if ( ((cmd->arg1 & 0xff) & bits) != 0) 38462587Sitojun return 0; /* some bits we want set were clear */ 38562587Sitojun want_clear = (cmd->arg1 >> 8) & 0xff; 38653541Sshin if ( (want_clear & bits) != want_clear) 38753541Sshin return 0; /* some bits we want clear were set */ 38853541Sshin return 1; 38953541Sshin} 39053541Sshin 39153541Sshinstatic int 39253541Sshinipopts_match(struct ip *ip, ipfw_insn *cmd) 39353541Sshin{ 39476899Ssumikawa int optlen, bits = 0; 39553541Sshin u_char *cp = (u_char *)(ip + 1); 39653541Sshin int x = (ip->ip_hl << 2) - sizeof (struct ip); 39753541Sshin 39876899Ssumikawa for (; x > 0; x -= optlen, cp += optlen) { 39976899Ssumikawa int opt = cp[IPOPT_OPTVAL]; 40053541Sshin 40153541Sshin if (opt == IPOPT_EOL) 40253541Sshin break; 40353541Sshin if (opt == IPOPT_NOP) 40453541Sshin optlen = 1; 40553541Sshin else { 40676899Ssumikawa optlen = cp[IPOPT_OLEN]; 40753541Sshin if (optlen <= 0 || optlen > x) 40853541Sshin return 0; /* invalid or truncated */ 40953541Sshin } 41076899Ssumikawa switch (opt) { 41176899Ssumikawa 41253541Sshin default: 41353541Sshin break; 41453541Sshin 41562587Sitojun case IPOPT_LSRR: 41653541Sshin bits |= IP_FW_IPOPT_LSRR; 41753541Sshin break; 41853541Sshin 41953541Sshin case IPOPT_SSRR: 42053541Sshin bits |= IP_FW_IPOPT_SSRR; 42153541Sshin break; 42253541Sshin 42353541Sshin case IPOPT_RR: 42453541Sshin bits |= IP_FW_IPOPT_RR; 42553541Sshin break; 42662587Sitojun 42762587Sitojun case IPOPT_TS: 42862587Sitojun bits |= IP_FW_IPOPT_TS; 42962587Sitojun break; 43062587Sitojun } 43162587Sitojun } 43253541Sshin return (flags_match(cmd, bits)); 43353541Sshin} 43453541Sshin 43553541Sshinstatic int 43653541Sshintcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd) 43753541Sshin{ 43853541Sshin int optlen, bits = 0; 43953541Sshin u_char *cp = (u_char *)(tcp + 1); 44053541Sshin int x = (tcp->th_off << 2) - sizeof(struct tcphdr); 44153541Sshin 44253541Sshin for (; x > 0; x -= optlen, cp += optlen) { 44353541Sshin int opt = cp[0]; 44453541Sshin if (opt == TCPOPT_EOL) 44553541Sshin break; 44653541Sshin if (opt == TCPOPT_NOP) 44753541Sshin optlen = 1; 44853541Sshin else { 44953541Sshin optlen = cp[1]; 45053541Sshin if (optlen <= 0) 45153541Sshin break; 45262587Sitojun } 45353541Sshin 45462587Sitojun switch (opt) { 45562587Sitojun 45653541Sshin default: 45753541Sshin break; 45853541Sshin 45962587Sitojun case TCPOPT_MAXSEG: 46062587Sitojun bits |= IP_FW_TCPOPT_MSS; 46162587Sitojun break; 46253541Sshin 46353541Sshin case TCPOPT_WINDOW: 46453541Sshin bits |= IP_FW_TCPOPT_WINDOW; 46553541Sshin break; 46662587Sitojun 46762587Sitojun case TCPOPT_SACK_PERMITTED: 46853541Sshin case TCPOPT_SACK: 46953541Sshin bits |= IP_FW_TCPOPT_SACK; 47053541Sshin break; 47153541Sshin 47262587Sitojun case TCPOPT_TIMESTAMP: 47362587Sitojun bits |= IP_FW_TCPOPT_TS; 47462587Sitojun break; 47553541Sshin 47653541Sshin } 47753541Sshin } 47853541Sshin return (flags_match(cmd, bits)); 47962587Sitojun} 48053541Sshin 48153541Sshinstatic int 48262587Sitojuniface_match(struct ifnet *ifp, ipfw_insn_if *cmd) 48362587Sitojun{ 48462587Sitojun if (ifp == NULL) /* no iface with this packet, match fails */ 48562587Sitojun return 0; 48662587Sitojun /* Check by name or by IP address */ 48762587Sitojun if (cmd->name[0] != '\0') { /* match by name */ 48862587Sitojun /* Check name */ 48962587Sitojun if (cmd->p.glob) { 49062587Sitojun if (fnmatch(cmd->name, ifp->if_xname, 0) == 0) 49162587Sitojun return(1); 49262587Sitojun } else { 49362587Sitojun if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0) 49453541Sshin return(1); 49553541Sshin } 49653541Sshin } else { 49753541Sshin struct ifaddr *ia; 49853541Sshin 49953541Sshin if_addr_rlock(ifp); 50053541Sshin TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) { 50153541Sshin if (ia->ifa_addr->sa_family != AF_INET) 50253541Sshin continue; 50353541Sshin if (cmd->p.ip.s_addr == ((struct sockaddr_in *) 50453541Sshin (ia->ifa_addr))->sin_addr.s_addr) { 50553541Sshin if_addr_runlock(ifp); 50653541Sshin return(1); /* match */ 50753541Sshin } 50853541Sshin } 50953541Sshin if_addr_runlock(ifp); 51053541Sshin } 51153541Sshin return(0); /* no match, fail ... */ 51253541Sshin} 51353541Sshin 51453541Sshin/* 51553541Sshin * The verify_path function checks if a route to the src exists and 51653541Sshin * if it is reachable via ifp (when provided). 51753541Sshin * 51853541Sshin * The 'verrevpath' option checks that the interface that an IP packet 51953541Sshin * arrives on is the same interface that traffic destined for the 52053541Sshin * packet's source address would be routed out of. The 'versrcreach' 52153541Sshin * option just checks that the source address is reachable via any route 52253541Sshin * (except default) in the routing table. These two are a measure to block 52353541Sshin * forged packets. This is also commonly known as "anti-spoofing" or Unicast 52453541Sshin * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs 52553541Sshin * is purposely reminiscent of the Cisco IOS command, 52653541Sshin * 52753541Sshin * ip verify unicast reverse-path 52853541Sshin * ip verify unicast source reachable-via any 52953541Sshin * 53053541Sshin * which implements the same functionality. But note that syntax is 53153541Sshin * misleading. The check may be performed on all IP packets whether unicast, 53253541Sshin * multicast, or broadcast. 53353541Sshin */ 53453541Sshinstatic int 53553541Sshinverify_path(struct in_addr src, struct ifnet *ifp, u_int fib) 53653541Sshin{ 53753541Sshin struct route ro; 53853541Sshin struct sockaddr_in *dst; 53953541Sshin 54053541Sshin bzero(&ro, sizeof(ro)); 54153541Sshin 54253541Sshin dst = (struct sockaddr_in *)&(ro.ro_dst); 54353541Sshin dst->sin_family = AF_INET; 54453541Sshin dst->sin_len = sizeof(*dst); 54553541Sshin dst->sin_addr = src; 54653541Sshin in_rtalloc_ign(&ro, 0, fib); 54753541Sshin 54853541Sshin if (ro.ro_rt == NULL) 54953541Sshin return 0; 55053541Sshin 55153541Sshin /* 55253541Sshin * If ifp is provided, check for equality with rtentry. 55353541Sshin * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 55453541Sshin * in order to pass packets injected back by if_simloop(): 55553541Sshin * if useloopback == 1 routing entry (via lo0) for our own address 55653541Sshin * may exist, so we need to handle routing assymetry. 55753541Sshin */ 55853541Sshin if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 55953541Sshin RTFREE(ro.ro_rt); 56053541Sshin return 0; 56153541Sshin } 56253541Sshin 56353541Sshin /* if no ifp provided, check if rtentry is not default route */ 56453541Sshin if (ifp == NULL && 56553541Sshin satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) { 56653541Sshin RTFREE(ro.ro_rt); 56753541Sshin return 0; 56862587Sitojun } 56953541Sshin 57062587Sitojun /* or if this is a blackhole/reject route */ 57153541Sshin if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 57253541Sshin RTFREE(ro.ro_rt); 57353541Sshin return 0; 57453541Sshin } 57553541Sshin 57653541Sshin /* found valid route */ 57753541Sshin RTFREE(ro.ro_rt); 57853541Sshin return 1; 57953541Sshin} 58053541Sshin 58153541Sshin#ifdef INET6 58253541Sshin/* 58353541Sshin * ipv6 specific rules here... 58453541Sshin */ 58553541Sshinstatic __inline int 58653541Sshinicmp6type_match (int type, ipfw_insn_u32 *cmd) 58753541Sshin{ 58853541Sshin return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) ); 58953541Sshin} 59053541Sshin 59153541Sshinstatic int 59253541Sshinflow6id_match( int curr_flow, ipfw_insn_u32 *cmd ) 59353541Sshin{ 59453541Sshin int i; 59553541Sshin for (i=0; i <= cmd->o.arg1; ++i ) 59653541Sshin if (curr_flow == cmd->d[i] ) 59753541Sshin return 1; 59853541Sshin return 0; 59953541Sshin} 60053541Sshin 60153541Sshin/* support for IP6_*_ME opcodes */ 60253541Sshinstatic int 60353541Sshinsearch_ip6_addr_net (struct in6_addr * ip6_addr) 60453541Sshin{ 60553541Sshin struct ifnet *mdc; 60653541Sshin struct ifaddr *mdc2; 60753541Sshin struct in6_ifaddr *fdm; 60853541Sshin struct in6_addr copia; 60953541Sshin 61053541Sshin TAILQ_FOREACH(mdc, &V_ifnet, if_link) { 61153541Sshin if_addr_rlock(mdc); 61253541Sshin TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) { 61353541Sshin if (mdc2->ifa_addr->sa_family == AF_INET6) { 61453541Sshin fdm = (struct in6_ifaddr *)mdc2; 61553541Sshin copia = fdm->ia_addr.sin6_addr; 61653541Sshin /* need for leaving scope_id in the sock_addr */ 61753541Sshin in6_clearscope(&copia); 61853541Sshin if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) { 61953541Sshin if_addr_runlock(mdc); 62053541Sshin return 1; 62153541Sshin } 62253541Sshin } 62353541Sshin } 62453541Sshin if_addr_runlock(mdc); 62553541Sshin } 62653541Sshin return 0; 62753541Sshin} 62853541Sshin 62953541Sshinstatic int 63062587Sitojunverify_path6(struct in6_addr *src, struct ifnet *ifp) 63162587Sitojun{ 63262587Sitojun struct route_in6 ro; 63353541Sshin struct sockaddr_in6 *dst; 63453541Sshin 63553541Sshin bzero(&ro, sizeof(ro)); 63653541Sshin 63753541Sshin dst = (struct sockaddr_in6 * )&(ro.ro_dst); 63853541Sshin dst->sin6_family = AF_INET6; 63953541Sshin dst->sin6_len = sizeof(*dst); 64053541Sshin dst->sin6_addr = *src; 64153541Sshin /* XXX MRT 0 for ipv6 at this time */ 64253541Sshin rtalloc_ign((struct route *)&ro, 0); 64353541Sshin 64453541Sshin if (ro.ro_rt == NULL) 64553541Sshin return 0; 64653541Sshin 64753541Sshin /* 64853541Sshin * if ifp is provided, check for equality with rtentry 64953541Sshin * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp, 65053541Sshin * to support the case of sending packets to an address of our own. 65153541Sshin * (where the former interface is the first argument of if_simloop() 65253541Sshin * (=ifp), the latter is lo0) 65353541Sshin */ 65453541Sshin if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) { 65553541Sshin RTFREE(ro.ro_rt); 65653541Sshin return 0; 65762587Sitojun } 65862587Sitojun 65962587Sitojun /* if no ifp provided, check if rtentry is not default route */ 66062587Sitojun if (ifp == NULL && 66162587Sitojun IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) { 66262587Sitojun RTFREE(ro.ro_rt); 66362587Sitojun return 0; 66462587Sitojun } 66562587Sitojun 66662587Sitojun /* or if this is a blackhole/reject route */ 66762587Sitojun if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 66862587Sitojun RTFREE(ro.ro_rt); 66962587Sitojun return 0; 67062587Sitojun } 67162587Sitojun 67262587Sitojun /* found valid route */ 67362587Sitojun RTFREE(ro.ro_rt); 67453541Sshin return 1; 67553541Sshin 67653541Sshin} 67753541Sshinstatic __inline int 67853541Sshinhash_packet6(struct ipfw_flow_id *id) 67953541Sshin{ 68053541Sshin u_int32_t i; 68153541Sshin i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^ 68253541Sshin (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^ 68353541Sshin (id->src_ip6.__u6_addr.__u6_addr32[2]) ^ 68453541Sshin (id->src_ip6.__u6_addr.__u6_addr32[3]) ^ 68553541Sshin (id->dst_port) ^ (id->src_port); 68653541Sshin return i; 68753541Sshin} 68853541Sshin 68953541Sshinstatic int 69053541Sshinis_icmp6_query(int icmp6_type) 691{ 692 if ((icmp6_type <= ICMP6_MAXTYPE) && 693 (icmp6_type == ICMP6_ECHO_REQUEST || 694 icmp6_type == ICMP6_MEMBERSHIP_QUERY || 695 icmp6_type == ICMP6_WRUREQUEST || 696 icmp6_type == ICMP6_FQDN_QUERY || 697 icmp6_type == ICMP6_NI_QUERY)) 698 return (1); 699 700 return (0); 701} 702 703static void 704send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6) 705{ 706 struct mbuf *m; 707 708 m = args->m; 709 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) { 710 struct tcphdr *tcp; 711 tcp_seq ack, seq; 712 int flags; 713 struct { 714 struct ip6_hdr ip6; 715 struct tcphdr th; 716 } ti; 717 tcp = (struct tcphdr *)((char *)ip6 + hlen); 718 719 if ((tcp->th_flags & TH_RST) != 0) { 720 m_freem(m); 721 args->m = NULL; 722 return; 723 } 724 725 ti.ip6 = *ip6; 726 ti.th = *tcp; 727 ti.th.th_seq = ntohl(ti.th.th_seq); 728 ti.th.th_ack = ntohl(ti.th.th_ack); 729 ti.ip6.ip6_nxt = IPPROTO_TCP; 730 731 if (ti.th.th_flags & TH_ACK) { 732 ack = 0; 733 seq = ti.th.th_ack; 734 flags = TH_RST; 735 } else { 736 ack = ti.th.th_seq; 737 if ((m->m_flags & M_PKTHDR) != 0) { 738 /* 739 * total new data to ACK is: 740 * total packet length, 741 * minus the header length, 742 * minus the tcp header length. 743 */ 744 ack += m->m_pkthdr.len - hlen 745 - (ti.th.th_off << 2); 746 } else if (ip6->ip6_plen) { 747 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) - 748 hlen - (ti.th.th_off << 2); 749 } else { 750 m_freem(m); 751 return; 752 } 753 if (tcp->th_flags & TH_SYN) 754 ack++; 755 seq = 0; 756 flags = TH_RST|TH_ACK; 757 } 758 bcopy(&ti, ip6, sizeof(ti)); 759 /* 760 * m is only used to recycle the mbuf 761 * The data in it is never read so we don't need 762 * to correct the offsets or anything 763 */ 764 tcp_respond(NULL, ip6, tcp, m, ack, seq, flags); 765 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */ 766#if 0 767 /* 768 * Unlike above, the mbufs need to line up with the ip6 hdr, 769 * as the contents are read. We need to m_adj() the 770 * needed amount. 771 * The mbuf will however be thrown away so we can adjust it. 772 * Remember we did an m_pullup on it already so we 773 * can make some assumptions about contiguousness. 774 */ 775 if (args->L3offset) 776 m_adj(m, args->L3offset); 777#endif 778 icmp6_error(m, ICMP6_DST_UNREACH, code, 0); 779 } else 780 m_freem(m); 781 782 args->m = NULL; 783} 784 785#endif /* INET6 */ 786 787/* counter for ipfw_log(NULL...) */ 788static VNET_DEFINE(u_int64_t, norule_counter); 789#define V_norule_counter VNET(norule_counter) 790 791#define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0 792#define SNP(buf) buf, sizeof(buf) 793 794/* 795 * We enter here when we have a rule with O_LOG. 796 * XXX this function alone takes about 2Kbytes of code! 797 */ 798static void 799ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args, 800 struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg, 801 struct ip *ip) 802{ 803 struct ether_header *eh = args->eh; 804 char *action; 805 int limit_reached = 0; 806 char action2[40], proto[128], fragment[32]; 807 808 fragment[0] = '\0'; 809 proto[0] = '\0'; 810 811 if (f == NULL) { /* bogus pkt */ 812 if (V_verbose_limit != 0 && V_norule_counter >= V_verbose_limit) 813 return; 814 V_norule_counter++; 815 if (V_norule_counter == V_verbose_limit) 816 limit_reached = V_verbose_limit; 817 action = "Refuse"; 818 } else { /* O_LOG is the first action, find the real one */ 819 ipfw_insn *cmd = ACTION_PTR(f); 820 ipfw_insn_log *l = (ipfw_insn_log *)cmd; 821 822 if (l->max_log != 0 && l->log_left == 0) 823 return; 824 l->log_left--; 825 if (l->log_left == 0) 826 limit_reached = l->max_log; 827 cmd += F_LEN(cmd); /* point to first action */ 828 if (cmd->opcode == O_ALTQ) { 829 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 830 831 snprintf(SNPARGS(action2, 0), "Altq %d", 832 altq->qid); 833 cmd += F_LEN(cmd); 834 } 835 if (cmd->opcode == O_PROB) 836 cmd += F_LEN(cmd); 837 838 if (cmd->opcode == O_TAG) 839 cmd += F_LEN(cmd); 840 841 action = action2; 842 switch (cmd->opcode) { 843 case O_DENY: 844 action = "Deny"; 845 break; 846 847 case O_REJECT: 848 if (cmd->arg1==ICMP_REJECT_RST) 849 action = "Reset"; 850 else if (cmd->arg1==ICMP_UNREACH_HOST) 851 action = "Reject"; 852 else 853 snprintf(SNPARGS(action2, 0), "Unreach %d", 854 cmd->arg1); 855 break; 856 857 case O_UNREACH6: 858 if (cmd->arg1==ICMP6_UNREACH_RST) 859 action = "Reset"; 860 else 861 snprintf(SNPARGS(action2, 0), "Unreach %d", 862 cmd->arg1); 863 break; 864 865 case O_ACCEPT: 866 action = "Accept"; 867 break; 868 case O_COUNT: 869 action = "Count"; 870 break; 871 case O_DIVERT: 872 snprintf(SNPARGS(action2, 0), "Divert %d", 873 cmd->arg1); 874 break; 875 case O_TEE: 876 snprintf(SNPARGS(action2, 0), "Tee %d", 877 cmd->arg1); 878 break; 879 case O_SETFIB: 880 snprintf(SNPARGS(action2, 0), "SetFib %d", 881 cmd->arg1); 882 break; 883 case O_SKIPTO: 884 snprintf(SNPARGS(action2, 0), "SkipTo %d", 885 cmd->arg1); 886 break; 887 case O_PIPE: 888 snprintf(SNPARGS(action2, 0), "Pipe %d", 889 cmd->arg1); 890 break; 891 case O_QUEUE: 892 snprintf(SNPARGS(action2, 0), "Queue %d", 893 cmd->arg1); 894 break; 895 case O_FORWARD_IP: { 896 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd; 897 int len; 898 struct in_addr dummyaddr; 899 if (sa->sa.sin_addr.s_addr == INADDR_ANY) 900 dummyaddr.s_addr = htonl(tablearg); 901 else 902 dummyaddr.s_addr = sa->sa.sin_addr.s_addr; 903 904 len = snprintf(SNPARGS(action2, 0), "Forward to %s", 905 inet_ntoa(dummyaddr)); 906 907 if (sa->sa.sin_port) 908 snprintf(SNPARGS(action2, len), ":%d", 909 sa->sa.sin_port); 910 } 911 break; 912 case O_NETGRAPH: 913 snprintf(SNPARGS(action2, 0), "Netgraph %d", 914 cmd->arg1); 915 break; 916 case O_NGTEE: 917 snprintf(SNPARGS(action2, 0), "Ngtee %d", 918 cmd->arg1); 919 break; 920 case O_NAT: 921 action = "Nat"; 922 break; 923 case O_REASS: 924 action = "Reass"; 925 break; 926 default: 927 action = "UNKNOWN"; 928 break; 929 } 930 } 931 932 if (hlen == 0) { /* non-ip */ 933 snprintf(SNPARGS(proto, 0), "MAC"); 934 935 } else { 936 int len; 937 char src[48], dst[48]; 938 struct icmphdr *icmp; 939 struct tcphdr *tcp; 940 struct udphdr *udp; 941#ifdef INET6 942 struct ip6_hdr *ip6 = NULL; 943 struct icmp6_hdr *icmp6; 944#endif 945 src[0] = '\0'; 946 dst[0] = '\0'; 947#ifdef INET6 948 if (IS_IP6_FLOW_ID(&(args->f_id))) { 949 char ip6buf[INET6_ADDRSTRLEN]; 950 snprintf(src, sizeof(src), "[%s]", 951 ip6_sprintf(ip6buf, &args->f_id.src_ip6)); 952 snprintf(dst, sizeof(dst), "[%s]", 953 ip6_sprintf(ip6buf, &args->f_id.dst_ip6)); 954 955 ip6 = (struct ip6_hdr *)ip; 956 tcp = (struct tcphdr *)(((char *)ip) + hlen); 957 udp = (struct udphdr *)(((char *)ip) + hlen); 958 } else 959#endif 960 { 961 tcp = L3HDR(struct tcphdr, ip); 962 udp = L3HDR(struct udphdr, ip); 963 964 inet_ntoa_r(ip->ip_src, src); 965 inet_ntoa_r(ip->ip_dst, dst); 966 } 967 968 switch (args->f_id.proto) { 969 case IPPROTO_TCP: 970 len = snprintf(SNPARGS(proto, 0), "TCP %s", src); 971 if (offset == 0) 972 snprintf(SNPARGS(proto, len), ":%d %s:%d", 973 ntohs(tcp->th_sport), 974 dst, 975 ntohs(tcp->th_dport)); 976 else 977 snprintf(SNPARGS(proto, len), " %s", dst); 978 break; 979 980 case IPPROTO_UDP: 981 len = snprintf(SNPARGS(proto, 0), "UDP %s", src); 982 if (offset == 0) 983 snprintf(SNPARGS(proto, len), ":%d %s:%d", 984 ntohs(udp->uh_sport), 985 dst, 986 ntohs(udp->uh_dport)); 987 else 988 snprintf(SNPARGS(proto, len), " %s", dst); 989 break; 990 991 case IPPROTO_ICMP: 992 icmp = L3HDR(struct icmphdr, ip); 993 if (offset == 0) 994 len = snprintf(SNPARGS(proto, 0), 995 "ICMP:%u.%u ", 996 icmp->icmp_type, icmp->icmp_code); 997 else 998 len = snprintf(SNPARGS(proto, 0), "ICMP "); 999 len += snprintf(SNPARGS(proto, len), "%s", src); 1000 snprintf(SNPARGS(proto, len), " %s", dst); 1001 break; 1002#ifdef INET6 1003 case IPPROTO_ICMPV6: 1004 icmp6 = (struct icmp6_hdr *)(((char *)ip) + hlen); 1005 if (offset == 0) 1006 len = snprintf(SNPARGS(proto, 0), 1007 "ICMPv6:%u.%u ", 1008 icmp6->icmp6_type, icmp6->icmp6_code); 1009 else 1010 len = snprintf(SNPARGS(proto, 0), "ICMPv6 "); 1011 len += snprintf(SNPARGS(proto, len), "%s", src); 1012 snprintf(SNPARGS(proto, len), " %s", dst); 1013 break; 1014#endif 1015 default: 1016 len = snprintf(SNPARGS(proto, 0), "P:%d %s", 1017 args->f_id.proto, src); 1018 snprintf(SNPARGS(proto, len), " %s", dst); 1019 break; 1020 } 1021 1022#ifdef INET6 1023 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1024 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG)) 1025 snprintf(SNPARGS(fragment, 0), 1026 " (frag %08x:%d@%d%s)", 1027 args->f_id.frag_id6, 1028 ntohs(ip6->ip6_plen) - hlen, 1029 ntohs(offset & IP6F_OFF_MASK) << 3, 1030 (offset & IP6F_MORE_FRAG) ? "+" : ""); 1031 } else 1032#endif 1033 { 1034 int ip_off, ip_len; 1035 if (eh != NULL) { /* layer 2 packets are as on the wire */ 1036 ip_off = ntohs(ip->ip_off); 1037 ip_len = ntohs(ip->ip_len); 1038 } else { 1039 ip_off = ip->ip_off; 1040 ip_len = ip->ip_len; 1041 } 1042 if (ip_off & (IP_MF | IP_OFFMASK)) 1043 snprintf(SNPARGS(fragment, 0), 1044 " (frag %d:%d@%d%s)", 1045 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2), 1046 offset << 3, 1047 (ip_off & IP_MF) ? "+" : ""); 1048 } 1049 } 1050 if (oif || m->m_pkthdr.rcvif) 1051 log(LOG_SECURITY | LOG_INFO, 1052 "ipfw: %d %s %s %s via %s%s\n", 1053 f ? f->rulenum : -1, 1054 action, proto, oif ? "out" : "in", 1055 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname, 1056 fragment); 1057 else 1058 log(LOG_SECURITY | LOG_INFO, 1059 "ipfw: %d %s %s [no if info]%s\n", 1060 f ? f->rulenum : -1, 1061 action, proto, fragment); 1062 if (limit_reached) 1063 log(LOG_SECURITY | LOG_NOTICE, 1064 "ipfw: limit %d reached on entry %d\n", 1065 limit_reached, f ? f->rulenum : -1); 1066} 1067 1068/* 1069 * IMPORTANT: the hash function for dynamic rules must be commutative 1070 * in source and destination (ip,port), because rules are bidirectional 1071 * and we want to find both in the same bucket. 1072 */ 1073static __inline int 1074hash_packet(struct ipfw_flow_id *id) 1075{ 1076 u_int32_t i; 1077 1078#ifdef INET6 1079 if (IS_IP6_FLOW_ID(id)) 1080 i = hash_packet6(id); 1081 else 1082#endif /* INET6 */ 1083 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port); 1084 i &= (V_curr_dyn_buckets - 1); 1085 return i; 1086} 1087 1088/** 1089 * unlink a dynamic rule from a chain. prev is a pointer to 1090 * the previous one, q is a pointer to the rule to delete, 1091 * head is a pointer to the head of the queue. 1092 * Modifies q and potentially also head. 1093 */ 1094#define UNLINK_DYN_RULE(prev, head, q) { \ 1095 ipfw_dyn_rule *old_q = q; \ 1096 \ 1097 /* remove a refcount to the parent */ \ 1098 if (q->dyn_type == O_LIMIT) \ 1099 q->parent->count--; \ 1100 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\ 1101 (q->id.src_ip), (q->id.src_port), \ 1102 (q->id.dst_ip), (q->id.dst_port), V_dyn_count-1 ); ) \ 1103 if (prev != NULL) \ 1104 prev->next = q = q->next; \ 1105 else \ 1106 head = q = q->next; \ 1107 V_dyn_count--; \ 1108 uma_zfree(ipfw_dyn_rule_zone, old_q); } 1109 1110#define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0) 1111 1112/** 1113 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL. 1114 * 1115 * If keep_me == NULL, rules are deleted even if not expired, 1116 * otherwise only expired rules are removed. 1117 * 1118 * The value of the second parameter is also used to point to identify 1119 * a rule we absolutely do not want to remove (e.g. because we are 1120 * holding a reference to it -- this is the case with O_LIMIT_PARENT 1121 * rules). The pointer is only used for comparison, so any non-null 1122 * value will do. 1123 */ 1124static void 1125remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me) 1126{ 1127 static u_int32_t last_remove = 0; 1128 1129#define FORCE (keep_me == NULL) 1130 1131 ipfw_dyn_rule *prev, *q; 1132 int i, pass = 0, max_pass = 0; 1133 1134 IPFW_DYN_LOCK_ASSERT(); 1135 1136 if (V_ipfw_dyn_v == NULL || V_dyn_count == 0) 1137 return; 1138 /* do not expire more than once per second, it is useless */ 1139 if (!FORCE && last_remove == time_uptime) 1140 return; 1141 last_remove = time_uptime; 1142 1143 /* 1144 * because O_LIMIT refer to parent rules, during the first pass only 1145 * remove child and mark any pending LIMIT_PARENT, and remove 1146 * them in a second pass. 1147 */ 1148next_pass: 1149 for (i = 0 ; i < V_curr_dyn_buckets ; i++) { 1150 for (prev=NULL, q = V_ipfw_dyn_v[i] ; q ; ) { 1151 /* 1152 * Logic can become complex here, so we split tests. 1153 */ 1154 if (q == keep_me) 1155 goto next; 1156 if (rule != NULL && rule != q->rule) 1157 goto next; /* not the one we are looking for */ 1158 if (q->dyn_type == O_LIMIT_PARENT) { 1159 /* 1160 * handle parent in the second pass, 1161 * record we need one. 1162 */ 1163 max_pass = 1; 1164 if (pass == 0) 1165 goto next; 1166 if (FORCE && q->count != 0 ) { 1167 /* XXX should not happen! */ 1168 printf("ipfw: OUCH! cannot remove rule," 1169 " count %d\n", q->count); 1170 } 1171 } else { 1172 if (!FORCE && 1173 !TIME_LEQ( q->expire, time_uptime )) 1174 goto next; 1175 } 1176 if (q->dyn_type != O_LIMIT_PARENT || !q->count) { 1177 UNLINK_DYN_RULE(prev, V_ipfw_dyn_v[i], q); 1178 continue; 1179 } 1180next: 1181 prev=q; 1182 q=q->next; 1183 } 1184 } 1185 if (pass++ < max_pass) 1186 goto next_pass; 1187} 1188 1189 1190/** 1191 * lookup a dynamic rule. 1192 */ 1193static ipfw_dyn_rule * 1194lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction, 1195 struct tcphdr *tcp) 1196{ 1197 /* 1198 * stateful ipfw extensions. 1199 * Lookup into dynamic session queue 1200 */ 1201#define MATCH_REVERSE 0 1202#define MATCH_FORWARD 1 1203#define MATCH_NONE 2 1204#define MATCH_UNKNOWN 3 1205 int i, dir = MATCH_NONE; 1206 ipfw_dyn_rule *prev, *q=NULL; 1207 1208 IPFW_DYN_LOCK_ASSERT(); 1209 1210 if (V_ipfw_dyn_v == NULL) 1211 goto done; /* not found */ 1212 i = hash_packet( pkt ); 1213 for (prev=NULL, q = V_ipfw_dyn_v[i] ; q != NULL ; ) { 1214 if (q->dyn_type == O_LIMIT_PARENT && q->count) 1215 goto next; 1216 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */ 1217 UNLINK_DYN_RULE(prev, V_ipfw_dyn_v[i], q); 1218 continue; 1219 } 1220 if (pkt->proto == q->id.proto && 1221 q->dyn_type != O_LIMIT_PARENT) { 1222 if (IS_IP6_FLOW_ID(pkt)) { 1223 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1224 &(q->id.src_ip6)) && 1225 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1226 &(q->id.dst_ip6)) && 1227 pkt->src_port == q->id.src_port && 1228 pkt->dst_port == q->id.dst_port ) { 1229 dir = MATCH_FORWARD; 1230 break; 1231 } 1232 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1233 &(q->id.dst_ip6)) && 1234 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1235 &(q->id.src_ip6)) && 1236 pkt->src_port == q->id.dst_port && 1237 pkt->dst_port == q->id.src_port ) { 1238 dir = MATCH_REVERSE; 1239 break; 1240 } 1241 } else { 1242 if (pkt->src_ip == q->id.src_ip && 1243 pkt->dst_ip == q->id.dst_ip && 1244 pkt->src_port == q->id.src_port && 1245 pkt->dst_port == q->id.dst_port ) { 1246 dir = MATCH_FORWARD; 1247 break; 1248 } 1249 if (pkt->src_ip == q->id.dst_ip && 1250 pkt->dst_ip == q->id.src_ip && 1251 pkt->src_port == q->id.dst_port && 1252 pkt->dst_port == q->id.src_port ) { 1253 dir = MATCH_REVERSE; 1254 break; 1255 } 1256 } 1257 } 1258next: 1259 prev = q; 1260 q = q->next; 1261 } 1262 if (q == NULL) 1263 goto done; /* q = NULL, not found */ 1264 1265 if ( prev != NULL) { /* found and not in front */ 1266 prev->next = q->next; 1267 q->next = V_ipfw_dyn_v[i]; 1268 V_ipfw_dyn_v[i] = q; 1269 } 1270 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */ 1271 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST); 1272 1273#define BOTH_SYN (TH_SYN | (TH_SYN << 8)) 1274#define BOTH_FIN (TH_FIN | (TH_FIN << 8)) 1275 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8); 1276 switch (q->state) { 1277 case TH_SYN: /* opening */ 1278 q->expire = time_uptime + V_dyn_syn_lifetime; 1279 break; 1280 1281 case BOTH_SYN: /* move to established */ 1282 case BOTH_SYN | TH_FIN : /* one side tries to close */ 1283 case BOTH_SYN | (TH_FIN << 8) : 1284 if (tcp) { 1285#define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0) 1286 u_int32_t ack = ntohl(tcp->th_ack); 1287 if (dir == MATCH_FORWARD) { 1288 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd)) 1289 q->ack_fwd = ack; 1290 else { /* ignore out-of-sequence */ 1291 break; 1292 } 1293 } else { 1294 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev)) 1295 q->ack_rev = ack; 1296 else { /* ignore out-of-sequence */ 1297 break; 1298 } 1299 } 1300 } 1301 q->expire = time_uptime + V_dyn_ack_lifetime; 1302 break; 1303 1304 case BOTH_SYN | BOTH_FIN: /* both sides closed */ 1305 if (V_dyn_fin_lifetime >= V_dyn_keepalive_period) 1306 V_dyn_fin_lifetime = V_dyn_keepalive_period - 1; 1307 q->expire = time_uptime + V_dyn_fin_lifetime; 1308 break; 1309 1310 default: 1311#if 0 1312 /* 1313 * reset or some invalid combination, but can also 1314 * occur if we use keep-state the wrong way. 1315 */ 1316 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0) 1317 printf("invalid state: 0x%x\n", q->state); 1318#endif 1319 if (V_dyn_rst_lifetime >= V_dyn_keepalive_period) 1320 V_dyn_rst_lifetime = V_dyn_keepalive_period - 1; 1321 q->expire = time_uptime + V_dyn_rst_lifetime; 1322 break; 1323 } 1324 } else if (pkt->proto == IPPROTO_UDP) { 1325 q->expire = time_uptime + V_dyn_udp_lifetime; 1326 } else { 1327 /* other protocols */ 1328 q->expire = time_uptime + V_dyn_short_lifetime; 1329 } 1330done: 1331 if (match_direction) 1332 *match_direction = dir; 1333 return q; 1334} 1335 1336static ipfw_dyn_rule * 1337lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction, 1338 struct tcphdr *tcp) 1339{ 1340 ipfw_dyn_rule *q; 1341 1342 IPFW_DYN_LOCK(); 1343 q = lookup_dyn_rule_locked(pkt, match_direction, tcp); 1344 if (q == NULL) 1345 IPFW_DYN_UNLOCK(); 1346 /* NB: return table locked when q is not NULL */ 1347 return q; 1348} 1349 1350static void 1351realloc_dynamic_table(void) 1352{ 1353 IPFW_DYN_LOCK_ASSERT(); 1354 1355 /* 1356 * Try reallocation, make sure we have a power of 2 and do 1357 * not allow more than 64k entries. In case of overflow, 1358 * default to 1024. 1359 */ 1360 1361 if (V_dyn_buckets > 65536) 1362 V_dyn_buckets = 1024; 1363 if ((V_dyn_buckets & (V_dyn_buckets-1)) != 0) { /* not a power of 2 */ 1364 V_dyn_buckets = V_curr_dyn_buckets; /* reset */ 1365 return; 1366 } 1367 V_curr_dyn_buckets = V_dyn_buckets; 1368 if (V_ipfw_dyn_v != NULL) 1369 free(V_ipfw_dyn_v, M_IPFW); 1370 for (;;) { 1371 V_ipfw_dyn_v = malloc(V_curr_dyn_buckets * sizeof(ipfw_dyn_rule *), 1372 M_IPFW, M_NOWAIT | M_ZERO); 1373 if (V_ipfw_dyn_v != NULL || V_curr_dyn_buckets <= 2) 1374 break; 1375 V_curr_dyn_buckets /= 2; 1376 } 1377} 1378 1379/** 1380 * Install state of type 'type' for a dynamic session. 1381 * The hash table contains two type of rules: 1382 * - regular rules (O_KEEP_STATE) 1383 * - rules for sessions with limited number of sess per user 1384 * (O_LIMIT). When they are created, the parent is 1385 * increased by 1, and decreased on delete. In this case, 1386 * the third parameter is the parent rule and not the chain. 1387 * - "parent" rules for the above (O_LIMIT_PARENT). 1388 */ 1389static ipfw_dyn_rule * 1390add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule) 1391{ 1392 ipfw_dyn_rule *r; 1393 int i; 1394 1395 IPFW_DYN_LOCK_ASSERT(); 1396 1397 if (V_ipfw_dyn_v == NULL || 1398 (V_dyn_count == 0 && V_dyn_buckets != V_curr_dyn_buckets)) { 1399 realloc_dynamic_table(); 1400 if (V_ipfw_dyn_v == NULL) 1401 return NULL; /* failed ! */ 1402 } 1403 i = hash_packet(id); 1404 1405 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO); 1406 if (r == NULL) { 1407 printf ("ipfw: sorry cannot allocate state\n"); 1408 return NULL; 1409 } 1410 1411 /* increase refcount on parent, and set pointer */ 1412 if (dyn_type == O_LIMIT) { 1413 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule; 1414 if ( parent->dyn_type != O_LIMIT_PARENT) 1415 panic("invalid parent"); 1416 parent->count++; 1417 r->parent = parent; 1418 rule = parent->rule; 1419 } 1420 1421 r->id = *id; 1422 r->expire = time_uptime + V_dyn_syn_lifetime; 1423 r->rule = rule; 1424 r->dyn_type = dyn_type; 1425 r->pcnt = r->bcnt = 0; 1426 r->count = 0; 1427 1428 r->bucket = i; 1429 r->next = V_ipfw_dyn_v[i]; 1430 V_ipfw_dyn_v[i] = r; 1431 V_dyn_count++; 1432 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n", 1433 dyn_type, 1434 (r->id.src_ip), (r->id.src_port), 1435 (r->id.dst_ip), (r->id.dst_port), 1436 V_dyn_count ); ) 1437 return r; 1438} 1439 1440/** 1441 * lookup dynamic parent rule using pkt and rule as search keys. 1442 * If the lookup fails, then install one. 1443 */ 1444static ipfw_dyn_rule * 1445lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule) 1446{ 1447 ipfw_dyn_rule *q; 1448 int i; 1449 1450 IPFW_DYN_LOCK_ASSERT(); 1451 1452 if (V_ipfw_dyn_v) { 1453 int is_v6 = IS_IP6_FLOW_ID(pkt); 1454 i = hash_packet( pkt ); 1455 for (q = V_ipfw_dyn_v[i] ; q != NULL ; q=q->next) 1456 if (q->dyn_type == O_LIMIT_PARENT && 1457 rule== q->rule && 1458 pkt->proto == q->id.proto && 1459 pkt->src_port == q->id.src_port && 1460 pkt->dst_port == q->id.dst_port && 1461 ( 1462 (is_v6 && 1463 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6), 1464 &(q->id.src_ip6)) && 1465 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6), 1466 &(q->id.dst_ip6))) || 1467 (!is_v6 && 1468 pkt->src_ip == q->id.src_ip && 1469 pkt->dst_ip == q->id.dst_ip) 1470 ) 1471 ) { 1472 q->expire = time_uptime + V_dyn_short_lifetime; 1473 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);) 1474 return q; 1475 } 1476 } 1477 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule); 1478} 1479 1480/** 1481 * Install dynamic state for rule type cmd->o.opcode 1482 * 1483 * Returns 1 (failure) if state is not installed because of errors or because 1484 * session limitations are enforced. 1485 */ 1486static int 1487install_state(struct ip_fw *rule, ipfw_insn_limit *cmd, 1488 struct ip_fw_args *args, uint32_t tablearg) 1489{ 1490 static int last_log; 1491 ipfw_dyn_rule *q; 1492 struct in_addr da; 1493 char src[48], dst[48]; 1494 1495 src[0] = '\0'; 1496 dst[0] = '\0'; 1497 1498 DEB( 1499 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n", 1500 __func__, cmd->o.opcode, 1501 (args->f_id.src_ip), (args->f_id.src_port), 1502 (args->f_id.dst_ip), (args->f_id.dst_port)); 1503 ) 1504 1505 IPFW_DYN_LOCK(); 1506 1507 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1508 1509 if (q != NULL) { /* should never occur */ 1510 if (last_log != time_uptime) { 1511 last_log = time_uptime; 1512 printf("ipfw: %s: entry already present, done\n", 1513 __func__); 1514 } 1515 IPFW_DYN_UNLOCK(); 1516 return (0); 1517 } 1518 1519 if (V_dyn_count >= V_dyn_max) 1520 /* Run out of slots, try to remove any expired rule. */ 1521 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1); 1522 1523 if (V_dyn_count >= V_dyn_max) { 1524 if (last_log != time_uptime) { 1525 last_log = time_uptime; 1526 printf("ipfw: %s: Too many dynamic rules\n", __func__); 1527 } 1528 IPFW_DYN_UNLOCK(); 1529 return (1); /* cannot install, notify caller */ 1530 } 1531 1532 switch (cmd->o.opcode) { 1533 case O_KEEP_STATE: /* bidir rule */ 1534 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule); 1535 break; 1536 1537 case O_LIMIT: { /* limit number of sessions */ 1538 struct ipfw_flow_id id; 1539 ipfw_dyn_rule *parent; 1540 uint32_t conn_limit; 1541 uint16_t limit_mask = cmd->limit_mask; 1542 1543 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ? 1544 tablearg : cmd->conn_limit; 1545 1546 DEB( 1547 if (cmd->conn_limit == IP_FW_TABLEARG) 1548 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u " 1549 "(tablearg)\n", __func__, conn_limit); 1550 else 1551 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n", 1552 __func__, conn_limit); 1553 ) 1554 1555 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0; 1556 id.proto = args->f_id.proto; 1557 id.addr_type = args->f_id.addr_type; 1558 id.fib = M_GETFIB(args->m); 1559 1560 if (IS_IP6_FLOW_ID (&(args->f_id))) { 1561 if (limit_mask & DYN_SRC_ADDR) 1562 id.src_ip6 = args->f_id.src_ip6; 1563 if (limit_mask & DYN_DST_ADDR) 1564 id.dst_ip6 = args->f_id.dst_ip6; 1565 } else { 1566 if (limit_mask & DYN_SRC_ADDR) 1567 id.src_ip = args->f_id.src_ip; 1568 if (limit_mask & DYN_DST_ADDR) 1569 id.dst_ip = args->f_id.dst_ip; 1570 } 1571 if (limit_mask & DYN_SRC_PORT) 1572 id.src_port = args->f_id.src_port; 1573 if (limit_mask & DYN_DST_PORT) 1574 id.dst_port = args->f_id.dst_port; 1575 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) { 1576 printf("ipfw: %s: add parent failed\n", __func__); 1577 IPFW_DYN_UNLOCK(); 1578 return (1); 1579 } 1580 1581 if (parent->count >= conn_limit) { 1582 /* See if we can remove some expired rule. */ 1583 remove_dyn_rule(rule, parent); 1584 if (parent->count >= conn_limit) { 1585 if (V_fw_verbose && last_log != time_uptime) { 1586 last_log = time_uptime; 1587#ifdef INET6 1588 /* 1589 * XXX IPv6 flows are not 1590 * supported yet. 1591 */ 1592 if (IS_IP6_FLOW_ID(&(args->f_id))) { 1593 char ip6buf[INET6_ADDRSTRLEN]; 1594 snprintf(src, sizeof(src), 1595 "[%s]", ip6_sprintf(ip6buf, 1596 &args->f_id.src_ip6)); 1597 snprintf(dst, sizeof(dst), 1598 "[%s]", ip6_sprintf(ip6buf, 1599 &args->f_id.dst_ip6)); 1600 } else 1601#endif 1602 { 1603 da.s_addr = 1604 htonl(args->f_id.src_ip); 1605 inet_ntoa_r(da, src); 1606 da.s_addr = 1607 htonl(args->f_id.dst_ip); 1608 inet_ntoa_r(da, dst); 1609 } 1610 log(LOG_SECURITY | LOG_DEBUG, 1611 "ipfw: %d %s %s:%u -> %s:%u, %s\n", 1612 parent->rule->rulenum, 1613 "drop session", 1614 src, (args->f_id.src_port), 1615 dst, (args->f_id.dst_port), 1616 "too many entries"); 1617 } 1618 IPFW_DYN_UNLOCK(); 1619 return (1); 1620 } 1621 } 1622 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent); 1623 break; 1624 } 1625 default: 1626 printf("ipfw: %s: unknown dynamic rule type %u\n", 1627 __func__, cmd->o.opcode); 1628 IPFW_DYN_UNLOCK(); 1629 return (1); 1630 } 1631 1632 /* XXX just set lifetime */ 1633 lookup_dyn_rule_locked(&args->f_id, NULL, NULL); 1634 1635 IPFW_DYN_UNLOCK(); 1636 return (0); 1637} 1638 1639/* 1640 * Generate a TCP packet, containing either a RST or a keepalive. 1641 * When flags & TH_RST, we are sending a RST packet, because of a 1642 * "reset" action matched the packet. 1643 * Otherwise we are sending a keepalive, and flags & TH_ 1644 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required 1645 * so that MAC can label the reply appropriately. 1646 */ 1647static struct mbuf * 1648send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq, 1649 u_int32_t ack, int flags) 1650{ 1651 struct mbuf *m; 1652 struct ip *ip; 1653 struct tcphdr *tcp; 1654 1655 MGETHDR(m, M_DONTWAIT, MT_DATA); 1656 if (m == 0) 1657 return (NULL); 1658 m->m_pkthdr.rcvif = (struct ifnet *)0; 1659 1660 M_SETFIB(m, id->fib); 1661#ifdef MAC 1662 if (replyto != NULL) 1663 mac_netinet_firewall_reply(replyto, m); 1664 else 1665 mac_netinet_firewall_send(m); 1666#else 1667 (void)replyto; /* don't warn about unused arg */ 1668#endif 1669 1670 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr); 1671 m->m_data += max_linkhdr; 1672 1673 ip = mtod(m, struct ip *); 1674 bzero(ip, m->m_len); 1675 tcp = (struct tcphdr *)(ip + 1); /* no IP options */ 1676 ip->ip_p = IPPROTO_TCP; 1677 tcp->th_off = 5; 1678 /* 1679 * Assume we are sending a RST (or a keepalive in the reverse 1680 * direction), swap src and destination addresses and ports. 1681 */ 1682 ip->ip_src.s_addr = htonl(id->dst_ip); 1683 ip->ip_dst.s_addr = htonl(id->src_ip); 1684 tcp->th_sport = htons(id->dst_port); 1685 tcp->th_dport = htons(id->src_port); 1686 if (flags & TH_RST) { /* we are sending a RST */ 1687 if (flags & TH_ACK) { 1688 tcp->th_seq = htonl(ack); 1689 tcp->th_ack = htonl(0); 1690 tcp->th_flags = TH_RST; 1691 } else { 1692 if (flags & TH_SYN) 1693 seq++; 1694 tcp->th_seq = htonl(0); 1695 tcp->th_ack = htonl(seq); 1696 tcp->th_flags = TH_RST | TH_ACK; 1697 } 1698 } else { 1699 /* 1700 * We are sending a keepalive. flags & TH_SYN determines 1701 * the direction, forward if set, reverse if clear. 1702 * NOTE: seq and ack are always assumed to be correct 1703 * as set by the caller. This may be confusing... 1704 */ 1705 if (flags & TH_SYN) { 1706 /* 1707 * we have to rewrite the correct addresses! 1708 */ 1709 ip->ip_dst.s_addr = htonl(id->dst_ip); 1710 ip->ip_src.s_addr = htonl(id->src_ip); 1711 tcp->th_dport = htons(id->dst_port); 1712 tcp->th_sport = htons(id->src_port); 1713 } 1714 tcp->th_seq = htonl(seq); 1715 tcp->th_ack = htonl(ack); 1716 tcp->th_flags = TH_ACK; 1717 } 1718 /* 1719 * set ip_len to the payload size so we can compute 1720 * the tcp checksum on the pseudoheader 1721 * XXX check this, could save a couple of words ? 1722 */ 1723 ip->ip_len = htons(sizeof(struct tcphdr)); 1724 tcp->th_sum = in_cksum(m, m->m_pkthdr.len); 1725 /* 1726 * now fill fields left out earlier 1727 */ 1728 ip->ip_ttl = V_ip_defttl; 1729 ip->ip_len = m->m_pkthdr.len; 1730 m->m_flags |= M_SKIP_FIREWALL; 1731 return (m); 1732} 1733 1734/* 1735 * sends a reject message, consuming the mbuf passed as an argument. 1736 */ 1737static void 1738send_reject(struct ip_fw_args *args, int code, int ip_len, struct ip *ip) 1739{ 1740 1741#if 0 1742 /* XXX When ip is not guaranteed to be at mtod() we will 1743 * need to account for this */ 1744 * The mbuf will however be thrown away so we can adjust it. 1745 * Remember we did an m_pullup on it already so we 1746 * can make some assumptions about contiguousness. 1747 */ 1748 if (args->L3offset) 1749 m_adj(m, args->L3offset); 1750#endif 1751 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */ 1752 /* We need the IP header in host order for icmp_error(). */ 1753 if (args->eh != NULL) { 1754 ip->ip_len = ntohs(ip->ip_len); 1755 ip->ip_off = ntohs(ip->ip_off); 1756 } 1757 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0); 1758 } else if (args->f_id.proto == IPPROTO_TCP) { 1759 struct tcphdr *const tcp = 1760 L3HDR(struct tcphdr, mtod(args->m, struct ip *)); 1761 if ( (tcp->th_flags & TH_RST) == 0) { 1762 struct mbuf *m; 1763 m = send_pkt(args->m, &(args->f_id), 1764 ntohl(tcp->th_seq), ntohl(tcp->th_ack), 1765 tcp->th_flags | TH_RST); 1766 if (m != NULL) 1767 ip_output(m, NULL, NULL, 0, NULL, NULL); 1768 } 1769 m_freem(args->m); 1770 } else 1771 m_freem(args->m); 1772 args->m = NULL; 1773} 1774 1775/** 1776 * 1777 * Given an ip_fw *, lookup_next_rule will return a pointer 1778 * to the next rule, which can be either the jump 1779 * target (for skipto instructions) or the next one in the list (in 1780 * all other cases including a missing jump target). 1781 * The result is also written in the "next_rule" field of the rule. 1782 * Backward jumps are not allowed, so start looking from the next 1783 * rule... 1784 * 1785 * This never returns NULL -- in case we do not have an exact match, 1786 * the next rule is returned. When the ruleset is changed, 1787 * pointers are flushed so we are always correct. 1788 */ 1789 1790static struct ip_fw * 1791lookup_next_rule(struct ip_fw *me, u_int32_t tablearg) 1792{ 1793 struct ip_fw *rule = NULL; 1794 ipfw_insn *cmd; 1795 u_int16_t rulenum; 1796 1797 /* look for action, in case it is a skipto */ 1798 cmd = ACTION_PTR(me); 1799 if (cmd->opcode == O_LOG) 1800 cmd += F_LEN(cmd); 1801 if (cmd->opcode == O_ALTQ) 1802 cmd += F_LEN(cmd); 1803 if (cmd->opcode == O_TAG) 1804 cmd += F_LEN(cmd); 1805 if (cmd->opcode == O_SKIPTO ) { 1806 if (tablearg != 0) { 1807 rulenum = (u_int16_t)tablearg; 1808 } else { 1809 rulenum = cmd->arg1; 1810 } 1811 for (rule = me->next; rule ; rule = rule->next) { 1812 if (rule->rulenum >= rulenum) { 1813 break; 1814 } 1815 } 1816 } 1817 if (rule == NULL) /* failure or not a skipto */ 1818 rule = me->next; 1819 me->next_rule = rule; 1820 return rule; 1821} 1822 1823static int 1824add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1825 uint8_t mlen, uint32_t value) 1826{ 1827 struct radix_node_head *rnh; 1828 struct table_entry *ent; 1829 struct radix_node *rn; 1830 1831 if (tbl >= IPFW_TABLES_MAX) 1832 return (EINVAL); 1833 rnh = ch->tables[tbl]; 1834 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO); 1835 if (ent == NULL) 1836 return (ENOMEM); 1837 ent->value = value; 1838 ent->addr.sin_len = ent->mask.sin_len = 8; 1839 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1840 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr; 1841 IPFW_WLOCK(ch); 1842 rn = rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent); 1843 if (rn == NULL) { 1844 IPFW_WUNLOCK(ch); 1845 free(ent, M_IPFW_TBL); 1846 return (EEXIST); 1847 } 1848 IPFW_WUNLOCK(ch); 1849 return (0); 1850} 1851 1852static int 1853del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1854 uint8_t mlen) 1855{ 1856 struct radix_node_head *rnh; 1857 struct table_entry *ent; 1858 struct sockaddr_in sa, mask; 1859 1860 if (tbl >= IPFW_TABLES_MAX) 1861 return (EINVAL); 1862 rnh = ch->tables[tbl]; 1863 sa.sin_len = mask.sin_len = 8; 1864 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0); 1865 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr; 1866 IPFW_WLOCK(ch); 1867 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh); 1868 if (ent == NULL) { 1869 IPFW_WUNLOCK(ch); 1870 return (ESRCH); 1871 } 1872 IPFW_WUNLOCK(ch); 1873 free(ent, M_IPFW_TBL); 1874 return (0); 1875} 1876 1877static int 1878flush_table_entry(struct radix_node *rn, void *arg) 1879{ 1880 struct radix_node_head * const rnh = arg; 1881 struct table_entry *ent; 1882 1883 ent = (struct table_entry *) 1884 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh); 1885 if (ent != NULL) 1886 free(ent, M_IPFW_TBL); 1887 return (0); 1888} 1889 1890static int 1891flush_table(struct ip_fw_chain *ch, uint16_t tbl) 1892{ 1893 struct radix_node_head *rnh; 1894 1895 IPFW_WLOCK_ASSERT(ch); 1896 1897 if (tbl >= IPFW_TABLES_MAX) 1898 return (EINVAL); 1899 rnh = ch->tables[tbl]; 1900 KASSERT(rnh != NULL, ("NULL IPFW table")); 1901 rnh->rnh_walktree(rnh, flush_table_entry, rnh); 1902 return (0); 1903} 1904 1905static void 1906flush_tables(struct ip_fw_chain *ch) 1907{ 1908 uint16_t tbl; 1909 1910 IPFW_WLOCK_ASSERT(ch); 1911 1912 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++) 1913 flush_table(ch, tbl); 1914} 1915 1916static int 1917init_tables(struct ip_fw_chain *ch) 1918{ 1919 int i; 1920 uint16_t j; 1921 1922 for (i = 0; i < IPFW_TABLES_MAX; i++) { 1923 if (!rn_inithead((void **)&ch->tables[i], 32)) { 1924 for (j = 0; j < i; j++) { 1925 (void) flush_table(ch, j); 1926 } 1927 return (ENOMEM); 1928 } 1929 } 1930 return (0); 1931} 1932 1933static int 1934lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr, 1935 uint32_t *val) 1936{ 1937 struct radix_node_head *rnh; 1938 struct table_entry *ent; 1939 struct sockaddr_in sa; 1940 1941 if (tbl >= IPFW_TABLES_MAX) 1942 return (0); 1943 rnh = ch->tables[tbl]; 1944 sa.sin_len = 8; 1945 sa.sin_addr.s_addr = addr; 1946 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh)); 1947 if (ent != NULL) { 1948 *val = ent->value; 1949 return (1); 1950 } 1951 return (0); 1952} 1953 1954static int 1955count_table_entry(struct radix_node *rn, void *arg) 1956{ 1957 u_int32_t * const cnt = arg; 1958 1959 (*cnt)++; 1960 return (0); 1961} 1962 1963static int 1964count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt) 1965{ 1966 struct radix_node_head *rnh; 1967 1968 if (tbl >= IPFW_TABLES_MAX) 1969 return (EINVAL); 1970 rnh = ch->tables[tbl]; 1971 *cnt = 0; 1972 rnh->rnh_walktree(rnh, count_table_entry, cnt); 1973 return (0); 1974} 1975 1976static int 1977dump_table_entry(struct radix_node *rn, void *arg) 1978{ 1979 struct table_entry * const n = (struct table_entry *)rn; 1980 ipfw_table * const tbl = arg; 1981 ipfw_table_entry *ent; 1982 1983 if (tbl->cnt == tbl->size) 1984 return (1); 1985 ent = &tbl->ent[tbl->cnt]; 1986 ent->tbl = tbl->tbl; 1987 if (in_nullhost(n->mask.sin_addr)) 1988 ent->masklen = 0; 1989 else 1990 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr)); 1991 ent->addr = n->addr.sin_addr.s_addr; 1992 ent->value = n->value; 1993 tbl->cnt++; 1994 return (0); 1995} 1996 1997static int 1998dump_table(struct ip_fw_chain *ch, ipfw_table *tbl) 1999{ 2000 struct radix_node_head *rnh; 2001 2002 if (tbl->tbl >= IPFW_TABLES_MAX) 2003 return (EINVAL); 2004 rnh = ch->tables[tbl->tbl]; 2005 tbl->cnt = 0; 2006 rnh->rnh_walktree(rnh, dump_table_entry, tbl); 2007 return (0); 2008} 2009 2010static int 2011check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif, 2012 struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip, 2013 u_int16_t src_port, struct ucred **uc, int *ugid_lookupp, 2014 struct inpcb *inp) 2015{ 2016 struct inpcbinfo *pi; 2017 int wildcard; 2018 struct inpcb *pcb; 2019 int match; 2020 2021 /* 2022 * Check to see if the UDP or TCP stack supplied us with 2023 * the PCB. If so, rather then holding a lock and looking 2024 * up the PCB, we can use the one that was supplied. 2025 */ 2026 if (inp && *ugid_lookupp == 0) { 2027 INP_LOCK_ASSERT(inp); 2028 if (inp->inp_socket != NULL) { 2029 *uc = crhold(inp->inp_cred); 2030 *ugid_lookupp = 1; 2031 } else 2032 *ugid_lookupp = -1; 2033 } 2034 /* 2035 * If we have already been here and the packet has no 2036 * PCB entry associated with it, then we can safely 2037 * assume that this is a no match. 2038 */ 2039 if (*ugid_lookupp == -1) 2040 return (0); 2041 if (proto == IPPROTO_TCP) { 2042 wildcard = 0; 2043 pi = &V_tcbinfo; 2044 } else if (proto == IPPROTO_UDP) { 2045 wildcard = INPLOOKUP_WILDCARD; 2046 pi = &V_udbinfo; 2047 } else 2048 return 0; 2049 match = 0; 2050 if (*ugid_lookupp == 0) { 2051 INP_INFO_RLOCK(pi); 2052 pcb = (oif) ? 2053 in_pcblookup_hash(pi, 2054 dst_ip, htons(dst_port), 2055 src_ip, htons(src_port), 2056 wildcard, oif) : 2057 in_pcblookup_hash(pi, 2058 src_ip, htons(src_port), 2059 dst_ip, htons(dst_port), 2060 wildcard, NULL); 2061 if (pcb != NULL) { 2062 *uc = crhold(pcb->inp_cred); 2063 *ugid_lookupp = 1; 2064 } 2065 INP_INFO_RUNLOCK(pi); 2066 if (*ugid_lookupp == 0) { 2067 /* 2068 * If the lookup did not yield any results, there 2069 * is no sense in coming back and trying again. So 2070 * we can set lookup to -1 and ensure that we wont 2071 * bother the pcb system again. 2072 */ 2073 *ugid_lookupp = -1; 2074 return (0); 2075 } 2076 } 2077 if (insn->o.opcode == O_UID) 2078 match = ((*uc)->cr_uid == (uid_t)insn->d[0]); 2079 else if (insn->o.opcode == O_GID) 2080 match = groupmember((gid_t)insn->d[0], *uc); 2081 else if (insn->o.opcode == O_JAIL) 2082 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]); 2083 return match; 2084} 2085 2086/* 2087 * The main check routine for the firewall. 2088 * 2089 * All arguments are in args so we can modify them and return them 2090 * back to the caller. 2091 * 2092 * Parameters: 2093 * 2094 * args->m (in/out) The packet; we set to NULL when/if we nuke it. 2095 * Starts with the IP header. 2096 * args->eh (in) Mac header if present, or NULL for layer3 packet. 2097 * args->L3offset Number of bytes bypassed if we came from L2. 2098 * e.g. often sizeof(eh) ** NOTYET ** 2099 * args->oif Outgoing interface, or NULL if packet is incoming. 2100 * The incoming interface is in the mbuf. (in) 2101 * args->divert_rule (in/out) 2102 * Skip up to the first rule past this rule number; 2103 * upon return, non-zero port number for divert or tee. 2104 * 2105 * args->rule Pointer to the last matching rule (in/out) 2106 * args->next_hop Socket we are forwarding to (out). 2107 * args->f_id Addresses grabbed from the packet (out) 2108 * args->cookie a cookie depending on rule action 2109 * 2110 * Return value: 2111 * 2112 * IP_FW_PASS the packet must be accepted 2113 * IP_FW_DENY the packet must be dropped 2114 * IP_FW_DIVERT divert packet, port in m_tag 2115 * IP_FW_TEE tee packet, port in m_tag 2116 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie 2117 * IP_FW_NETGRAPH into netgraph, cookie args->cookie 2118 * 2119 */ 2120int 2121ipfw_chk(struct ip_fw_args *args) 2122{ 2123 2124 /* 2125 * Local variables holding state during the processing of a packet: 2126 * 2127 * IMPORTANT NOTE: to speed up the processing of rules, there 2128 * are some assumption on the values of the variables, which 2129 * are documented here. Should you change them, please check 2130 * the implementation of the various instructions to make sure 2131 * that they still work. 2132 * 2133 * args->eh The MAC header. It is non-null for a layer2 2134 * packet, it is NULL for a layer-3 packet. 2135 * **notyet** 2136 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header. 2137 * 2138 * m | args->m Pointer to the mbuf, as received from the caller. 2139 * It may change if ipfw_chk() does an m_pullup, or if it 2140 * consumes the packet because it calls send_reject(). 2141 * XXX This has to change, so that ipfw_chk() never modifies 2142 * or consumes the buffer. 2143 * ip is the beginning of the ip(4 or 6) header. 2144 * Calculated by adding the L3offset to the start of data. 2145 * (Until we start using L3offset, the packet is 2146 * supposed to start with the ip header). 2147 */ 2148 struct mbuf *m = args->m; 2149 struct ip *ip = mtod(m, struct ip *); 2150 2151 /* 2152 * For rules which contain uid/gid or jail constraints, cache 2153 * a copy of the users credentials after the pcb lookup has been 2154 * executed. This will speed up the processing of rules with 2155 * these types of constraints, as well as decrease contention 2156 * on pcb related locks. 2157 */ 2158 struct ucred *ucred_cache = NULL; 2159 int ucred_lookup = 0; 2160 2161 /* 2162 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG 2163 * associated with a packet input on a divert socket. This 2164 * will allow to distinguish traffic and its direction when 2165 * it originates from a divert socket. 2166 */ 2167 u_int divinput_flags = 0; 2168 2169 /* 2170 * oif | args->oif If NULL, ipfw_chk has been called on the 2171 * inbound path (ether_input, ip_input). 2172 * If non-NULL, ipfw_chk has been called on the outbound path 2173 * (ether_output, ip_output). 2174 */ 2175 struct ifnet *oif = args->oif; 2176 2177 struct ip_fw *f = NULL; /* matching rule */ 2178 int retval = 0; 2179 2180 /* 2181 * hlen The length of the IP header. 2182 */ 2183 u_int hlen = 0; /* hlen >0 means we have an IP pkt */ 2184 2185 /* 2186 * offset The offset of a fragment. offset != 0 means that 2187 * we have a fragment at this offset of an IPv4 packet. 2188 * offset == 0 means that (if this is an IPv4 packet) 2189 * this is the first or only fragment. 2190 * For IPv6 offset == 0 means there is no Fragment Header. 2191 * If offset != 0 for IPv6 always use correct mask to 2192 * get the correct offset because we add IP6F_MORE_FRAG 2193 * to be able to dectect the first fragment which would 2194 * otherwise have offset = 0. 2195 */ 2196 u_short offset = 0; 2197 2198 /* 2199 * Local copies of addresses. They are only valid if we have 2200 * an IP packet. 2201 * 2202 * proto The protocol. Set to 0 for non-ip packets, 2203 * or to the protocol read from the packet otherwise. 2204 * proto != 0 means that we have an IPv4 packet. 2205 * 2206 * src_port, dst_port port numbers, in HOST format. Only 2207 * valid for TCP and UDP packets. 2208 * 2209 * src_ip, dst_ip ip addresses, in NETWORK format. 2210 * Only valid for IPv4 packets. 2211 */ 2212 u_int8_t proto; 2213 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */ 2214 struct in_addr src_ip, dst_ip; /* NOTE: network format */ 2215 u_int16_t ip_len=0; 2216 int pktlen; 2217 u_int16_t etype = 0; /* Host order stored ether type */ 2218 2219 /* 2220 * dyn_dir = MATCH_UNKNOWN when rules unchecked, 2221 * MATCH_NONE when checked and not matched (q = NULL), 2222 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL) 2223 */ 2224 int dyn_dir = MATCH_UNKNOWN; 2225 ipfw_dyn_rule *q = NULL; 2226 struct ip_fw_chain *chain = &V_layer3_chain; 2227 struct m_tag *mtag; 2228 2229 /* 2230 * We store in ulp a pointer to the upper layer protocol header. 2231 * In the ipv4 case this is easy to determine from the header, 2232 * but for ipv6 we might have some additional headers in the middle. 2233 * ulp is NULL if not found. 2234 */ 2235 void *ulp = NULL; /* upper layer protocol pointer. */ 2236 /* XXX ipv6 variables */ 2237 int is_ipv6 = 0; 2238 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */ 2239 /* end of ipv6 variables */ 2240 int is_ipv4 = 0; 2241 2242 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready)) 2243 return (IP_FW_PASS); /* accept */ 2244 2245 dst_ip.s_addr = 0; /* make sure it is initialized */ 2246 pktlen = m->m_pkthdr.len; 2247 args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */ 2248 proto = args->f_id.proto = 0; /* mark f_id invalid */ 2249 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */ 2250 2251/* 2252 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous, 2253 * then it sets p to point at the offset "len" in the mbuf. WARNING: the 2254 * pointer might become stale after other pullups (but we never use it 2255 * this way). 2256 */ 2257#define PULLUP_TO(len, p, T) \ 2258do { \ 2259 int x = (len) + sizeof(T); \ 2260 if ((m)->m_len < x) { \ 2261 args->m = m = m_pullup(m, x); \ 2262 if (m == NULL) \ 2263 goto pullup_failed; \ 2264 } \ 2265 p = (mtod(m, char *) + (len)); \ 2266} while (0) 2267 2268 /* 2269 * if we have an ether header, 2270 */ 2271 if (args->eh) 2272 etype = ntohs(args->eh->ether_type); 2273 2274 /* Identify IP packets and fill up variables. */ 2275 if (pktlen >= sizeof(struct ip6_hdr) && 2276 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) { 2277 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip; 2278 is_ipv6 = 1; 2279 args->f_id.addr_type = 6; 2280 hlen = sizeof(struct ip6_hdr); 2281 proto = ip6->ip6_nxt; 2282 2283 /* Search extension headers to find upper layer protocols */ 2284 while (ulp == NULL) { 2285 switch (proto) { 2286 case IPPROTO_ICMPV6: 2287 PULLUP_TO(hlen, ulp, struct icmp6_hdr); 2288 args->f_id.flags = ICMP6(ulp)->icmp6_type; 2289 break; 2290 2291 case IPPROTO_TCP: 2292 PULLUP_TO(hlen, ulp, struct tcphdr); 2293 dst_port = TCP(ulp)->th_dport; 2294 src_port = TCP(ulp)->th_sport; 2295 args->f_id.flags = TCP(ulp)->th_flags; 2296 break; 2297 2298 case IPPROTO_SCTP: 2299 PULLUP_TO(hlen, ulp, struct sctphdr); 2300 src_port = SCTP(ulp)->src_port; 2301 dst_port = SCTP(ulp)->dest_port; 2302 break; 2303 2304 case IPPROTO_UDP: 2305 PULLUP_TO(hlen, ulp, struct udphdr); 2306 dst_port = UDP(ulp)->uh_dport; 2307 src_port = UDP(ulp)->uh_sport; 2308 break; 2309 2310 case IPPROTO_HOPOPTS: /* RFC 2460 */ 2311 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2312 ext_hd |= EXT_HOPOPTS; 2313 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2314 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2315 ulp = NULL; 2316 break; 2317 2318 case IPPROTO_ROUTING: /* RFC 2460 */ 2319 PULLUP_TO(hlen, ulp, struct ip6_rthdr); 2320 switch (((struct ip6_rthdr *)ulp)->ip6r_type) { 2321 case 0: 2322 ext_hd |= EXT_RTHDR0; 2323 break; 2324 case 2: 2325 ext_hd |= EXT_RTHDR2; 2326 break; 2327 default: 2328 printf("IPFW2: IPV6 - Unknown Routing " 2329 "Header type(%d)\n", 2330 ((struct ip6_rthdr *)ulp)->ip6r_type); 2331 if (V_fw_deny_unknown_exthdrs) 2332 return (IP_FW_DENY); 2333 break; 2334 } 2335 ext_hd |= EXT_ROUTING; 2336 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3; 2337 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt; 2338 ulp = NULL; 2339 break; 2340 2341 case IPPROTO_FRAGMENT: /* RFC 2460 */ 2342 PULLUP_TO(hlen, ulp, struct ip6_frag); 2343 ext_hd |= EXT_FRAGMENT; 2344 hlen += sizeof (struct ip6_frag); 2345 proto = ((struct ip6_frag *)ulp)->ip6f_nxt; 2346 offset = ((struct ip6_frag *)ulp)->ip6f_offlg & 2347 IP6F_OFF_MASK; 2348 /* Add IP6F_MORE_FRAG for offset of first 2349 * fragment to be != 0. */ 2350 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg & 2351 IP6F_MORE_FRAG; 2352 if (offset == 0) { 2353 printf("IPFW2: IPV6 - Invalid Fragment " 2354 "Header\n"); 2355 if (V_fw_deny_unknown_exthdrs) 2356 return (IP_FW_DENY); 2357 break; 2358 } 2359 args->f_id.frag_id6 = 2360 ntohl(((struct ip6_frag *)ulp)->ip6f_ident); 2361 ulp = NULL; 2362 break; 2363 2364 case IPPROTO_DSTOPTS: /* RFC 2460 */ 2365 PULLUP_TO(hlen, ulp, struct ip6_hbh); 2366 ext_hd |= EXT_DSTOPTS; 2367 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3; 2368 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt; 2369 ulp = NULL; 2370 break; 2371 2372 case IPPROTO_AH: /* RFC 2402 */ 2373 PULLUP_TO(hlen, ulp, struct ip6_ext); 2374 ext_hd |= EXT_AH; 2375 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2; 2376 proto = ((struct ip6_ext *)ulp)->ip6e_nxt; 2377 ulp = NULL; 2378 break; 2379 2380 case IPPROTO_ESP: /* RFC 2406 */ 2381 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */ 2382 /* Anything past Seq# is variable length and 2383 * data past this ext. header is encrypted. */ 2384 ext_hd |= EXT_ESP; 2385 break; 2386 2387 case IPPROTO_NONE: /* RFC 2460 */ 2388 /* 2389 * Packet ends here, and IPv6 header has 2390 * already been pulled up. If ip6e_len!=0 2391 * then octets must be ignored. 2392 */ 2393 ulp = ip; /* non-NULL to get out of loop. */ 2394 break; 2395 2396 case IPPROTO_OSPFIGP: 2397 /* XXX OSPF header check? */ 2398 PULLUP_TO(hlen, ulp, struct ip6_ext); 2399 break; 2400 2401 case IPPROTO_PIM: 2402 /* XXX PIM header check? */ 2403 PULLUP_TO(hlen, ulp, struct pim); 2404 break; 2405 2406 case IPPROTO_CARP: 2407 PULLUP_TO(hlen, ulp, struct carp_header); 2408 if (((struct carp_header *)ulp)->carp_version != 2409 CARP_VERSION) 2410 return (IP_FW_DENY); 2411 if (((struct carp_header *)ulp)->carp_type != 2412 CARP_ADVERTISEMENT) 2413 return (IP_FW_DENY); 2414 break; 2415 2416 case IPPROTO_IPV6: /* RFC 2893 */ 2417 PULLUP_TO(hlen, ulp, struct ip6_hdr); 2418 break; 2419 2420 case IPPROTO_IPV4: /* RFC 2893 */ 2421 PULLUP_TO(hlen, ulp, struct ip); 2422 break; 2423 2424 default: 2425 printf("IPFW2: IPV6 - Unknown Extension " 2426 "Header(%d), ext_hd=%x\n", proto, ext_hd); 2427 if (V_fw_deny_unknown_exthdrs) 2428 return (IP_FW_DENY); 2429 PULLUP_TO(hlen, ulp, struct ip6_ext); 2430 break; 2431 } /*switch */ 2432 } 2433 ip = mtod(m, struct ip *); 2434 ip6 = (struct ip6_hdr *)ip; 2435 args->f_id.src_ip6 = ip6->ip6_src; 2436 args->f_id.dst_ip6 = ip6->ip6_dst; 2437 args->f_id.src_ip = 0; 2438 args->f_id.dst_ip = 0; 2439 args->f_id.flow_id6 = ntohl(ip6->ip6_flow); 2440 } else if (pktlen >= sizeof(struct ip) && 2441 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) { 2442 is_ipv4 = 1; 2443 hlen = ip->ip_hl << 2; 2444 args->f_id.addr_type = 4; 2445 2446 /* 2447 * Collect parameters into local variables for faster matching. 2448 */ 2449 proto = ip->ip_p; 2450 src_ip = ip->ip_src; 2451 dst_ip = ip->ip_dst; 2452 if (args->eh != NULL) { /* layer 2 packets are as on the wire */ 2453 offset = ntohs(ip->ip_off) & IP_OFFMASK; 2454 ip_len = ntohs(ip->ip_len); 2455 } else { 2456 offset = ip->ip_off & IP_OFFMASK; 2457 ip_len = ip->ip_len; 2458 } 2459 pktlen = ip_len < pktlen ? ip_len : pktlen; 2460 2461 if (offset == 0) { 2462 switch (proto) { 2463 case IPPROTO_TCP: 2464 PULLUP_TO(hlen, ulp, struct tcphdr); 2465 dst_port = TCP(ulp)->th_dport; 2466 src_port = TCP(ulp)->th_sport; 2467 args->f_id.flags = TCP(ulp)->th_flags; 2468 break; 2469 2470 case IPPROTO_UDP: 2471 PULLUP_TO(hlen, ulp, struct udphdr); 2472 dst_port = UDP(ulp)->uh_dport; 2473 src_port = UDP(ulp)->uh_sport; 2474 break; 2475 2476 case IPPROTO_ICMP: 2477 PULLUP_TO(hlen, ulp, struct icmphdr); 2478 args->f_id.flags = ICMP(ulp)->icmp_type; 2479 break; 2480 2481 default: 2482 break; 2483 } 2484 } 2485 2486 ip = mtod(m, struct ip *); 2487 args->f_id.src_ip = ntohl(src_ip.s_addr); 2488 args->f_id.dst_ip = ntohl(dst_ip.s_addr); 2489 } 2490#undef PULLUP_TO 2491 if (proto) { /* we may have port numbers, store them */ 2492 args->f_id.proto = proto; 2493 args->f_id.src_port = src_port = ntohs(src_port); 2494 args->f_id.dst_port = dst_port = ntohs(dst_port); 2495 } 2496 2497 IPFW_RLOCK(chain); 2498 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL); 2499 if (args->rule) { 2500 /* 2501 * Packet has already been tagged. Look for the next rule 2502 * to restart processing. Make sure that args->rule still 2503 * exists and not changed. 2504 */ 2505 if (chain->id != args->chain_id) { 2506 for (f = chain->rules; f != NULL; f = f->next) 2507 if (f == args->rule && f->id == args->rule_id) 2508 break; 2509 2510 if (f != NULL) 2511 f = f->next_rule; 2512 else 2513 f = ip_fw_default_rule; 2514 } else 2515 f = args->rule->next_rule; 2516 2517 if (f == NULL) 2518 f = lookup_next_rule(args->rule, 0); 2519 } else { 2520 /* 2521 * Find the starting rule. It can be either the first 2522 * one, or the one after divert_rule if asked so. 2523 */ 2524 int skipto = mtag ? divert_cookie(mtag) : 0; 2525 2526 f = chain->rules; 2527 if (args->eh == NULL && skipto != 0) { 2528 if (skipto >= IPFW_DEFAULT_RULE) { 2529 IPFW_RUNLOCK(chain); 2530 return (IP_FW_DENY); /* invalid */ 2531 } 2532 while (f && f->rulenum <= skipto) 2533 f = f->next; 2534 if (f == NULL) { /* drop packet */ 2535 IPFW_RUNLOCK(chain); 2536 return (IP_FW_DENY); 2537 } 2538 } 2539 } 2540 /* reset divert rule to avoid confusion later */ 2541 if (mtag) { 2542 divinput_flags = divert_info(mtag) & 2543 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG); 2544 m_tag_delete(m, mtag); 2545 } 2546 2547 /* 2548 * Now scan the rules, and parse microinstructions for each rule. 2549 */ 2550 for (; f; f = f->next) { 2551 ipfw_insn *cmd; 2552 uint32_t tablearg = 0; 2553 int l, cmdlen, skip_or; /* skip rest of OR block */ 2554 2555again: 2556 if (V_set_disable & (1 << f->set) ) 2557 continue; 2558 2559 skip_or = 0; 2560 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ; 2561 l -= cmdlen, cmd += cmdlen) { 2562 int match; 2563 2564 /* 2565 * check_body is a jump target used when we find a 2566 * CHECK_STATE, and need to jump to the body of 2567 * the target rule. 2568 */ 2569 2570check_body: 2571 cmdlen = F_LEN(cmd); 2572 /* 2573 * An OR block (insn_1 || .. || insn_n) has the 2574 * F_OR bit set in all but the last instruction. 2575 * The first match will set "skip_or", and cause 2576 * the following instructions to be skipped until 2577 * past the one with the F_OR bit clear. 2578 */ 2579 if (skip_or) { /* skip this instruction */ 2580 if ((cmd->len & F_OR) == 0) 2581 skip_or = 0; /* next one is good */ 2582 continue; 2583 } 2584 match = 0; /* set to 1 if we succeed */ 2585 2586 switch (cmd->opcode) { 2587 /* 2588 * The first set of opcodes compares the packet's 2589 * fields with some pattern, setting 'match' if a 2590 * match is found. At the end of the loop there is 2591 * logic to deal with F_NOT and F_OR flags associated 2592 * with the opcode. 2593 */ 2594 case O_NOP: 2595 match = 1; 2596 break; 2597 2598 case O_FORWARD_MAC: 2599 printf("ipfw: opcode %d unimplemented\n", 2600 cmd->opcode); 2601 break; 2602 2603 case O_GID: 2604 case O_UID: 2605 case O_JAIL: 2606 /* 2607 * We only check offset == 0 && proto != 0, 2608 * as this ensures that we have a 2609 * packet with the ports info. 2610 */ 2611 if (offset!=0) 2612 break; 2613 if (is_ipv6) /* XXX to be fixed later */ 2614 break; 2615 if (proto == IPPROTO_TCP || 2616 proto == IPPROTO_UDP) 2617 match = check_uidgid( 2618 (ipfw_insn_u32 *)cmd, 2619 proto, oif, 2620 dst_ip, dst_port, 2621 src_ip, src_port, &ucred_cache, 2622 &ucred_lookup, args->inp); 2623 break; 2624 2625 case O_RECV: 2626 match = iface_match(m->m_pkthdr.rcvif, 2627 (ipfw_insn_if *)cmd); 2628 break; 2629 2630 case O_XMIT: 2631 match = iface_match(oif, (ipfw_insn_if *)cmd); 2632 break; 2633 2634 case O_VIA: 2635 match = iface_match(oif ? oif : 2636 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd); 2637 break; 2638 2639 case O_MACADDR2: 2640 if (args->eh != NULL) { /* have MAC header */ 2641 u_int32_t *want = (u_int32_t *) 2642 ((ipfw_insn_mac *)cmd)->addr; 2643 u_int32_t *mask = (u_int32_t *) 2644 ((ipfw_insn_mac *)cmd)->mask; 2645 u_int32_t *hdr = (u_int32_t *)args->eh; 2646 2647 match = 2648 ( want[0] == (hdr[0] & mask[0]) && 2649 want[1] == (hdr[1] & mask[1]) && 2650 want[2] == (hdr[2] & mask[2]) ); 2651 } 2652 break; 2653 2654 case O_MAC_TYPE: 2655 if (args->eh != NULL) { 2656 u_int16_t *p = 2657 ((ipfw_insn_u16 *)cmd)->ports; 2658 int i; 2659 2660 for (i = cmdlen - 1; !match && i>0; 2661 i--, p += 2) 2662 match = (etype >= p[0] && 2663 etype <= p[1]); 2664 } 2665 break; 2666 2667 case O_FRAG: 2668 match = (offset != 0); 2669 break; 2670 2671 case O_IN: /* "out" is "not in" */ 2672 match = (oif == NULL); 2673 break; 2674 2675 case O_LAYER2: 2676 match = (args->eh != NULL); 2677 break; 2678 2679 case O_DIVERTED: 2680 match = (cmd->arg1 & 1 && divinput_flags & 2681 IP_FW_DIVERT_LOOPBACK_FLAG) || 2682 (cmd->arg1 & 2 && divinput_flags & 2683 IP_FW_DIVERT_OUTPUT_FLAG); 2684 break; 2685 2686 case O_PROTO: 2687 /* 2688 * We do not allow an arg of 0 so the 2689 * check of "proto" only suffices. 2690 */ 2691 match = (proto == cmd->arg1); 2692 break; 2693 2694 case O_IP_SRC: 2695 match = is_ipv4 && 2696 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2697 src_ip.s_addr); 2698 break; 2699 2700 case O_IP_SRC_LOOKUP: 2701 case O_IP_DST_LOOKUP: 2702 if (is_ipv4) { 2703 uint32_t a = 2704 (cmd->opcode == O_IP_DST_LOOKUP) ? 2705 dst_ip.s_addr : src_ip.s_addr; 2706 uint32_t v = 0; 2707 2708 match = lookup_table(chain, cmd->arg1, a, 2709 &v); 2710 if (!match) 2711 break; 2712 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) 2713 match = 2714 ((ipfw_insn_u32 *)cmd)->d[0] == v; 2715 else 2716 tablearg = v; 2717 } 2718 break; 2719 2720 case O_IP_SRC_MASK: 2721 case O_IP_DST_MASK: 2722 if (is_ipv4) { 2723 uint32_t a = 2724 (cmd->opcode == O_IP_DST_MASK) ? 2725 dst_ip.s_addr : src_ip.s_addr; 2726 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d; 2727 int i = cmdlen-1; 2728 2729 for (; !match && i>0; i-= 2, p+= 2) 2730 match = (p[0] == (a & p[1])); 2731 } 2732 break; 2733 2734 case O_IP_SRC_ME: 2735 if (is_ipv4) { 2736 struct ifnet *tif; 2737 2738 INADDR_TO_IFP(src_ip, tif); 2739 match = (tif != NULL); 2740 } 2741 break; 2742 2743 case O_IP_DST_SET: 2744 case O_IP_SRC_SET: 2745 if (is_ipv4) { 2746 u_int32_t *d = (u_int32_t *)(cmd+1); 2747 u_int32_t addr = 2748 cmd->opcode == O_IP_DST_SET ? 2749 args->f_id.dst_ip : 2750 args->f_id.src_ip; 2751 2752 if (addr < d[0]) 2753 break; 2754 addr -= d[0]; /* subtract base */ 2755 match = (addr < cmd->arg1) && 2756 ( d[ 1 + (addr>>5)] & 2757 (1<<(addr & 0x1f)) ); 2758 } 2759 break; 2760 2761 case O_IP_DST: 2762 match = is_ipv4 && 2763 (((ipfw_insn_ip *)cmd)->addr.s_addr == 2764 dst_ip.s_addr); 2765 break; 2766 2767 case O_IP_DST_ME: 2768 if (is_ipv4) { 2769 struct ifnet *tif; 2770 2771 INADDR_TO_IFP(dst_ip, tif); 2772 match = (tif != NULL); 2773 } 2774 break; 2775 2776 case O_IP_SRCPORT: 2777 case O_IP_DSTPORT: 2778 /* 2779 * offset == 0 && proto != 0 is enough 2780 * to guarantee that we have a 2781 * packet with port info. 2782 */ 2783 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP) 2784 && offset == 0) { 2785 u_int16_t x = 2786 (cmd->opcode == O_IP_SRCPORT) ? 2787 src_port : dst_port ; 2788 u_int16_t *p = 2789 ((ipfw_insn_u16 *)cmd)->ports; 2790 int i; 2791 2792 for (i = cmdlen - 1; !match && i>0; 2793 i--, p += 2) 2794 match = (x>=p[0] && x<=p[1]); 2795 } 2796 break; 2797 2798 case O_ICMPTYPE: 2799 match = (offset == 0 && proto==IPPROTO_ICMP && 2800 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) ); 2801 break; 2802 2803#ifdef INET6 2804 case O_ICMP6TYPE: 2805 match = is_ipv6 && offset == 0 && 2806 proto==IPPROTO_ICMPV6 && 2807 icmp6type_match( 2808 ICMP6(ulp)->icmp6_type, 2809 (ipfw_insn_u32 *)cmd); 2810 break; 2811#endif /* INET6 */ 2812 2813 case O_IPOPT: 2814 match = (is_ipv4 && 2815 ipopts_match(ip, cmd) ); 2816 break; 2817 2818 case O_IPVER: 2819 match = (is_ipv4 && 2820 cmd->arg1 == ip->ip_v); 2821 break; 2822 2823 case O_IPID: 2824 case O_IPLEN: 2825 case O_IPTTL: 2826 if (is_ipv4) { /* only for IP packets */ 2827 uint16_t x; 2828 uint16_t *p; 2829 int i; 2830 2831 if (cmd->opcode == O_IPLEN) 2832 x = ip_len; 2833 else if (cmd->opcode == O_IPTTL) 2834 x = ip->ip_ttl; 2835 else /* must be IPID */ 2836 x = ntohs(ip->ip_id); 2837 if (cmdlen == 1) { 2838 match = (cmd->arg1 == x); 2839 break; 2840 } 2841 /* otherwise we have ranges */ 2842 p = ((ipfw_insn_u16 *)cmd)->ports; 2843 i = cmdlen - 1; 2844 for (; !match && i>0; i--, p += 2) 2845 match = (x >= p[0] && x <= p[1]); 2846 } 2847 break; 2848 2849 case O_IPPRECEDENCE: 2850 match = (is_ipv4 && 2851 (cmd->arg1 == (ip->ip_tos & 0xe0)) ); 2852 break; 2853 2854 case O_IPTOS: 2855 match = (is_ipv4 && 2856 flags_match(cmd, ip->ip_tos)); 2857 break; 2858 2859 case O_TCPDATALEN: 2860 if (proto == IPPROTO_TCP && offset == 0) { 2861 struct tcphdr *tcp; 2862 uint16_t x; 2863 uint16_t *p; 2864 int i; 2865 2866 tcp = TCP(ulp); 2867 x = ip_len - 2868 ((ip->ip_hl + tcp->th_off) << 2); 2869 if (cmdlen == 1) { 2870 match = (cmd->arg1 == x); 2871 break; 2872 } 2873 /* otherwise we have ranges */ 2874 p = ((ipfw_insn_u16 *)cmd)->ports; 2875 i = cmdlen - 1; 2876 for (; !match && i>0; i--, p += 2) 2877 match = (x >= p[0] && x <= p[1]); 2878 } 2879 break; 2880 2881 case O_TCPFLAGS: 2882 match = (proto == IPPROTO_TCP && offset == 0 && 2883 flags_match(cmd, TCP(ulp)->th_flags)); 2884 break; 2885 2886 case O_TCPOPTS: 2887 match = (proto == IPPROTO_TCP && offset == 0 && 2888 tcpopts_match(TCP(ulp), cmd)); 2889 break; 2890 2891 case O_TCPSEQ: 2892 match = (proto == IPPROTO_TCP && offset == 0 && 2893 ((ipfw_insn_u32 *)cmd)->d[0] == 2894 TCP(ulp)->th_seq); 2895 break; 2896 2897 case O_TCPACK: 2898 match = (proto == IPPROTO_TCP && offset == 0 && 2899 ((ipfw_insn_u32 *)cmd)->d[0] == 2900 TCP(ulp)->th_ack); 2901 break; 2902 2903 case O_TCPWIN: 2904 match = (proto == IPPROTO_TCP && offset == 0 && 2905 cmd->arg1 == TCP(ulp)->th_win); 2906 break; 2907 2908 case O_ESTAB: 2909 /* reject packets which have SYN only */ 2910 /* XXX should i also check for TH_ACK ? */ 2911 match = (proto == IPPROTO_TCP && offset == 0 && 2912 (TCP(ulp)->th_flags & 2913 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN); 2914 break; 2915 2916 case O_ALTQ: { 2917 struct pf_mtag *at; 2918 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd; 2919 2920 match = 1; 2921 at = pf_find_mtag(m); 2922 if (at != NULL && at->qid != 0) 2923 break; 2924 at = pf_get_mtag(m); 2925 if (at == NULL) { 2926 /* 2927 * Let the packet fall back to the 2928 * default ALTQ. 2929 */ 2930 break; 2931 } 2932 at->qid = altq->qid; 2933 if (is_ipv4) 2934 at->af = AF_INET; 2935 else 2936 at->af = AF_LINK; 2937 at->hdr = ip; 2938 break; 2939 } 2940 2941 case O_LOG: 2942 if (V_fw_verbose) 2943 ipfw_log(f, hlen, args, m, 2944 oif, offset, tablearg, ip); 2945 match = 1; 2946 break; 2947 2948 case O_PROB: 2949 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]); 2950 break; 2951 2952 case O_VERREVPATH: 2953 /* Outgoing packets automatically pass/match */ 2954 match = ((oif != NULL) || 2955 (m->m_pkthdr.rcvif == NULL) || 2956 ( 2957#ifdef INET6 2958 is_ipv6 ? 2959 verify_path6(&(args->f_id.src_ip6), 2960 m->m_pkthdr.rcvif) : 2961#endif 2962 verify_path(src_ip, m->m_pkthdr.rcvif, 2963 args->f_id.fib))); 2964 break; 2965 2966 case O_VERSRCREACH: 2967 /* Outgoing packets automatically pass/match */ 2968 match = (hlen > 0 && ((oif != NULL) || 2969#ifdef INET6 2970 is_ipv6 ? 2971 verify_path6(&(args->f_id.src_ip6), 2972 NULL) : 2973#endif 2974 verify_path(src_ip, NULL, args->f_id.fib))); 2975 break; 2976 2977 case O_ANTISPOOF: 2978 /* Outgoing packets automatically pass/match */ 2979 if (oif == NULL && hlen > 0 && 2980 ( (is_ipv4 && in_localaddr(src_ip)) 2981#ifdef INET6 2982 || (is_ipv6 && 2983 in6_localaddr(&(args->f_id.src_ip6))) 2984#endif 2985 )) 2986 match = 2987#ifdef INET6 2988 is_ipv6 ? verify_path6( 2989 &(args->f_id.src_ip6), 2990 m->m_pkthdr.rcvif) : 2991#endif 2992 verify_path(src_ip, 2993 m->m_pkthdr.rcvif, 2994 args->f_id.fib); 2995 else 2996 match = 1; 2997 break; 2998 2999 case O_IPSEC: 3000#ifdef IPSEC 3001 match = (m_tag_find(m, 3002 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL); 3003#endif 3004 /* otherwise no match */ 3005 break; 3006 3007#ifdef INET6 3008 case O_IP6_SRC: 3009 match = is_ipv6 && 3010 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6, 3011 &((ipfw_insn_ip6 *)cmd)->addr6); 3012 break; 3013 3014 case O_IP6_DST: 3015 match = is_ipv6 && 3016 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6, 3017 &((ipfw_insn_ip6 *)cmd)->addr6); 3018 break; 3019 case O_IP6_SRC_MASK: 3020 case O_IP6_DST_MASK: 3021 if (is_ipv6) { 3022 int i = cmdlen - 1; 3023 struct in6_addr p; 3024 struct in6_addr *d = 3025 &((ipfw_insn_ip6 *)cmd)->addr6; 3026 3027 for (; !match && i > 0; d += 2, 3028 i -= F_INSN_SIZE(struct in6_addr) 3029 * 2) { 3030 p = (cmd->opcode == 3031 O_IP6_SRC_MASK) ? 3032 args->f_id.src_ip6: 3033 args->f_id.dst_ip6; 3034 APPLY_MASK(&p, &d[1]); 3035 match = 3036 IN6_ARE_ADDR_EQUAL(&d[0], 3037 &p); 3038 } 3039 } 3040 break; 3041 3042 case O_IP6_SRC_ME: 3043 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6); 3044 break; 3045 3046 case O_IP6_DST_ME: 3047 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6); 3048 break; 3049 3050 case O_FLOW6ID: 3051 match = is_ipv6 && 3052 flow6id_match(args->f_id.flow_id6, 3053 (ipfw_insn_u32 *) cmd); 3054 break; 3055 3056 case O_EXT_HDR: 3057 match = is_ipv6 && 3058 (ext_hd & ((ipfw_insn *) cmd)->arg1); 3059 break; 3060 3061 case O_IP6: 3062 match = is_ipv6; 3063 break; 3064#endif 3065 3066 case O_IP4: 3067 match = is_ipv4; 3068 break; 3069 3070 case O_TAG: { 3071 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3072 tablearg : cmd->arg1; 3073 3074 /* Packet is already tagged with this tag? */ 3075 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL); 3076 3077 /* We have `untag' action when F_NOT flag is 3078 * present. And we must remove this mtag from 3079 * mbuf and reset `match' to zero (`match' will 3080 * be inversed later). 3081 * Otherwise we should allocate new mtag and 3082 * push it into mbuf. 3083 */ 3084 if (cmd->len & F_NOT) { /* `untag' action */ 3085 if (mtag != NULL) 3086 m_tag_delete(m, mtag); 3087 } else if (mtag == NULL) { 3088 if ((mtag = m_tag_alloc(MTAG_IPFW, 3089 tag, 0, M_NOWAIT)) != NULL) 3090 m_tag_prepend(m, mtag); 3091 } 3092 match = (cmd->len & F_NOT) ? 0: 1; 3093 break; 3094 } 3095 3096 case O_FIB: /* try match the specified fib */ 3097 if (args->f_id.fib == cmd->arg1) 3098 match = 1; 3099 break; 3100 3101 case O_TAGGED: { 3102 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ? 3103 tablearg : cmd->arg1; 3104 3105 if (cmdlen == 1) { 3106 match = m_tag_locate(m, MTAG_IPFW, 3107 tag, NULL) != NULL; 3108 break; 3109 } 3110 3111 /* we have ranges */ 3112 for (mtag = m_tag_first(m); 3113 mtag != NULL && !match; 3114 mtag = m_tag_next(m, mtag)) { 3115 uint16_t *p; 3116 int i; 3117 3118 if (mtag->m_tag_cookie != MTAG_IPFW) 3119 continue; 3120 3121 p = ((ipfw_insn_u16 *)cmd)->ports; 3122 i = cmdlen - 1; 3123 for(; !match && i > 0; i--, p += 2) 3124 match = 3125 mtag->m_tag_id >= p[0] && 3126 mtag->m_tag_id <= p[1]; 3127 } 3128 break; 3129 } 3130 3131 /* 3132 * The second set of opcodes represents 'actions', 3133 * i.e. the terminal part of a rule once the packet 3134 * matches all previous patterns. 3135 * Typically there is only one action for each rule, 3136 * and the opcode is stored at the end of the rule 3137 * (but there are exceptions -- see below). 3138 * 3139 * In general, here we set retval and terminate the 3140 * outer loop (would be a 'break 3' in some language, 3141 * but we need to do a 'goto done'). 3142 * 3143 * Exceptions: 3144 * O_COUNT and O_SKIPTO actions: 3145 * instead of terminating, we jump to the next rule 3146 * ('goto next_rule', equivalent to a 'break 2'), 3147 * or to the SKIPTO target ('goto again' after 3148 * having set f, cmd and l), respectively. 3149 * 3150 * O_TAG, O_LOG and O_ALTQ action parameters: 3151 * perform some action and set match = 1; 3152 * 3153 * O_LIMIT and O_KEEP_STATE: these opcodes are 3154 * not real 'actions', and are stored right 3155 * before the 'action' part of the rule. 3156 * These opcodes try to install an entry in the 3157 * state tables; if successful, we continue with 3158 * the next opcode (match=1; break;), otherwise 3159 * the packet * must be dropped 3160 * ('goto done' after setting retval); 3161 * 3162 * O_PROBE_STATE and O_CHECK_STATE: these opcodes 3163 * cause a lookup of the state table, and a jump 3164 * to the 'action' part of the parent rule 3165 * ('goto check_body') if an entry is found, or 3166 * (CHECK_STATE only) a jump to the next rule if 3167 * the entry is not found ('goto next_rule'). 3168 * The result of the lookup is cached to make 3169 * further instances of these opcodes are 3170 * effectively NOPs. 3171 */ 3172 case O_LIMIT: 3173 case O_KEEP_STATE: 3174 if (install_state(f, 3175 (ipfw_insn_limit *)cmd, args, tablearg)) { 3176 retval = IP_FW_DENY; 3177 goto done; /* error/limit violation */ 3178 } 3179 match = 1; 3180 break; 3181 3182 case O_PROBE_STATE: 3183 case O_CHECK_STATE: 3184 /* 3185 * dynamic rules are checked at the first 3186 * keep-state or check-state occurrence, 3187 * with the result being stored in dyn_dir. 3188 * The compiler introduces a PROBE_STATE 3189 * instruction for us when we have a 3190 * KEEP_STATE (because PROBE_STATE needs 3191 * to be run first). 3192 */ 3193 if (dyn_dir == MATCH_UNKNOWN && 3194 (q = lookup_dyn_rule(&args->f_id, 3195 &dyn_dir, proto == IPPROTO_TCP ? 3196 TCP(ulp) : NULL)) 3197 != NULL) { 3198 /* 3199 * Found dynamic entry, update stats 3200 * and jump to the 'action' part of 3201 * the parent rule. 3202 */ 3203 q->pcnt++; 3204 q->bcnt += pktlen; 3205 f = q->rule; 3206 cmd = ACTION_PTR(f); 3207 l = f->cmd_len - f->act_ofs; 3208 IPFW_DYN_UNLOCK(); 3209 goto check_body; 3210 } 3211 /* 3212 * Dynamic entry not found. If CHECK_STATE, 3213 * skip to next rule, if PROBE_STATE just 3214 * ignore and continue with next opcode. 3215 */ 3216 if (cmd->opcode == O_CHECK_STATE) 3217 goto next_rule; 3218 match = 1; 3219 break; 3220 3221 case O_ACCEPT: 3222 retval = 0; /* accept */ 3223 goto done; 3224 3225 case O_PIPE: 3226 case O_QUEUE: 3227 args->rule = f; /* report matching rule */ 3228 args->rule_id = f->id; 3229 args->chain_id = chain->id; 3230 if (cmd->arg1 == IP_FW_TABLEARG) 3231 args->cookie = tablearg; 3232 else 3233 args->cookie = cmd->arg1; 3234 retval = IP_FW_DUMMYNET; 3235 goto done; 3236 3237 case O_DIVERT: 3238 case O_TEE: { 3239 struct divert_tag *dt; 3240 3241 if (args->eh) /* not on layer 2 */ 3242 break; 3243 mtag = m_tag_get(PACKET_TAG_DIVERT, 3244 sizeof(struct divert_tag), 3245 M_NOWAIT); 3246 if (mtag == NULL) { 3247 /* XXX statistic */ 3248 /* drop packet */ 3249 IPFW_RUNLOCK(chain); 3250 if (ucred_cache != NULL) 3251 crfree(ucred_cache); 3252 return (IP_FW_DENY); 3253 } 3254 dt = (struct divert_tag *)(mtag+1); 3255 dt->cookie = f->rulenum; 3256 if (cmd->arg1 == IP_FW_TABLEARG) 3257 dt->info = tablearg; 3258 else 3259 dt->info = cmd->arg1; 3260 m_tag_prepend(m, mtag); 3261 retval = (cmd->opcode == O_DIVERT) ? 3262 IP_FW_DIVERT : IP_FW_TEE; 3263 goto done; 3264 } 3265 case O_COUNT: 3266 case O_SKIPTO: 3267 f->pcnt++; /* update stats */ 3268 f->bcnt += pktlen; 3269 f->timestamp = time_uptime; 3270 if (cmd->opcode == O_COUNT) 3271 goto next_rule; 3272 /* handle skipto */ 3273 if (cmd->arg1 == IP_FW_TABLEARG) { 3274 f = lookup_next_rule(f, tablearg); 3275 } else { 3276 if (f->next_rule == NULL) 3277 lookup_next_rule(f, 0); 3278 f = f->next_rule; 3279 } 3280 goto again; 3281 3282 case O_REJECT: 3283 /* 3284 * Drop the packet and send a reject notice 3285 * if the packet is not ICMP (or is an ICMP 3286 * query), and it is not multicast/broadcast. 3287 */ 3288 if (hlen > 0 && is_ipv4 && offset == 0 && 3289 (proto != IPPROTO_ICMP || 3290 is_icmp_query(ICMP(ulp))) && 3291 !(m->m_flags & (M_BCAST|M_MCAST)) && 3292 !IN_MULTICAST(ntohl(dst_ip.s_addr))) { 3293 send_reject(args, cmd->arg1, ip_len, ip); 3294 m = args->m; 3295 } 3296 /* FALLTHROUGH */ 3297#ifdef INET6 3298 case O_UNREACH6: 3299 if (hlen > 0 && is_ipv6 && 3300 ((offset & IP6F_OFF_MASK) == 0) && 3301 (proto != IPPROTO_ICMPV6 || 3302 (is_icmp6_query(args->f_id.flags) == 1)) && 3303 !(m->m_flags & (M_BCAST|M_MCAST)) && 3304 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) { 3305 send_reject6( 3306 args, cmd->arg1, hlen, 3307 (struct ip6_hdr *)ip); 3308 m = args->m; 3309 } 3310 /* FALLTHROUGH */ 3311#endif 3312 case O_DENY: 3313 retval = IP_FW_DENY; 3314 goto done; 3315 3316 case O_FORWARD_IP: { 3317 struct sockaddr_in *sa; 3318 sa = &(((ipfw_insn_sa *)cmd)->sa); 3319 if (args->eh) /* not valid on layer2 pkts */ 3320 break; 3321 if (!q || dyn_dir == MATCH_FORWARD) { 3322 if (sa->sin_addr.s_addr == INADDR_ANY) { 3323 bcopy(sa, &args->hopstore, 3324 sizeof(*sa)); 3325 args->hopstore.sin_addr.s_addr = 3326 htonl(tablearg); 3327 args->next_hop = 3328 &args->hopstore; 3329 } else { 3330 args->next_hop = sa; 3331 } 3332 } 3333 retval = IP_FW_PASS; 3334 } 3335 goto done; 3336 3337 case O_NETGRAPH: 3338 case O_NGTEE: 3339 args->rule = f; /* report matching rule */ 3340 args->rule_id = f->id; 3341 args->chain_id = chain->id; 3342 if (cmd->arg1 == IP_FW_TABLEARG) 3343 args->cookie = tablearg; 3344 else 3345 args->cookie = cmd->arg1; 3346 retval = (cmd->opcode == O_NETGRAPH) ? 3347 IP_FW_NETGRAPH : IP_FW_NGTEE; 3348 goto done; 3349 3350 case O_SETFIB: 3351 f->pcnt++; /* update stats */ 3352 f->bcnt += pktlen; 3353 f->timestamp = time_uptime; 3354 M_SETFIB(m, cmd->arg1); 3355 args->f_id.fib = cmd->arg1; 3356 goto next_rule; 3357 3358 case O_NAT: { 3359 struct cfg_nat *t; 3360 int nat_id; 3361 3362 if (IPFW_NAT_LOADED) { 3363 args->rule = f; /* Report matching rule. */ 3364 args->rule_id = f->id; 3365 args->chain_id = chain->id; 3366 t = ((ipfw_insn_nat *)cmd)->nat; 3367 if (t == NULL) { 3368 nat_id = (cmd->arg1 == IP_FW_TABLEARG) ? 3369 tablearg : cmd->arg1; 3370 LOOKUP_NAT(V_layer3_chain, nat_id, t); 3371 if (t == NULL) { 3372 retval = IP_FW_DENY; 3373 goto done; 3374 } 3375 if (cmd->arg1 != IP_FW_TABLEARG) 3376 ((ipfw_insn_nat *)cmd)->nat = t; 3377 } 3378 retval = ipfw_nat_ptr(args, t, m); 3379 } else 3380 retval = IP_FW_DENY; 3381 goto done; 3382 } 3383 3384 case O_REASS: { 3385 int ip_off; 3386 3387 f->pcnt++; 3388 f->bcnt += pktlen; 3389 ip_off = (args->eh != NULL) ? ntohs(ip->ip_off) : ip->ip_off; 3390 if (ip_off & (IP_MF | IP_OFFMASK)) { 3391 /* 3392 * ip_reass() expects len & off in host 3393 * byte order: fix them in case we come 3394 * from layer2. 3395 */ 3396 if (args->eh != NULL) { 3397 ip->ip_len = ntohs(ip->ip_len); 3398 ip->ip_off = ntohs(ip->ip_off); 3399 } 3400 3401 m = ip_reass(m); 3402 args->m = m; 3403 3404 /* 3405 * IP header checksum fixup after 3406 * reassembly and leave header 3407 * in network byte order. 3408 */ 3409 if (m != NULL) { 3410 int hlen; 3411 3412 ip = mtod(m, struct ip *); 3413 hlen = ip->ip_hl << 2; 3414 /* revert len & off for layer2 pkts */ 3415 if (args->eh != NULL) 3416 ip->ip_len = htons(ip->ip_len); 3417 ip->ip_sum = 0; 3418 if (hlen == sizeof(struct ip)) 3419 ip->ip_sum = in_cksum_hdr(ip); 3420 else 3421 ip->ip_sum = in_cksum(m, hlen); 3422 retval = IP_FW_REASS; 3423 args->rule = f; 3424 args->rule_id = f->id; 3425 args->chain_id = chain->id; 3426 goto done; 3427 } else { 3428 retval = IP_FW_DENY; 3429 goto done; 3430 } 3431 } 3432 goto next_rule; 3433 } 3434 3435 default: 3436 panic("-- unknown opcode %d\n", cmd->opcode); 3437 } /* end of switch() on opcodes */ 3438 3439 if (cmd->len & F_NOT) 3440 match = !match; 3441 3442 if (match) { 3443 if (cmd->len & F_OR) 3444 skip_or = 1; 3445 } else { 3446 if (!(cmd->len & F_OR)) /* not an OR block, */ 3447 break; /* try next rule */ 3448 } 3449 3450 } /* end of inner for, scan opcodes */ 3451 3452next_rule:; /* try next rule */ 3453 3454 } /* end of outer for, scan rules */ 3455 printf("ipfw: ouch!, skip past end of rules, denying packet\n"); 3456 IPFW_RUNLOCK(chain); 3457 if (ucred_cache != NULL) 3458 crfree(ucred_cache); 3459 return (IP_FW_DENY); 3460 3461done: 3462 /* Update statistics */ 3463 f->pcnt++; 3464 f->bcnt += pktlen; 3465 f->timestamp = time_uptime; 3466 IPFW_RUNLOCK(chain); 3467 if (ucred_cache != NULL) 3468 crfree(ucred_cache); 3469 return (retval); 3470 3471pullup_failed: 3472 if (V_fw_verbose) 3473 printf("ipfw: pullup failed\n"); 3474 return (IP_FW_DENY); 3475} 3476 3477/* 3478 * When a rule is added/deleted, clear the next_rule pointers in all rules. 3479 * These will be reconstructed on the fly as packets are matched. 3480 */ 3481static void 3482flush_rule_ptrs(struct ip_fw_chain *chain) 3483{ 3484 struct ip_fw *rule; 3485 3486 IPFW_WLOCK_ASSERT(chain); 3487 3488 chain->id++; 3489 3490 for (rule = chain->rules; rule; rule = rule->next) 3491 rule->next_rule = NULL; 3492} 3493 3494/* 3495 * Add a new rule to the list. Copy the rule into a malloc'ed area, then 3496 * possibly create a rule number and add the rule to the list. 3497 * Update the rule_number in the input struct so the caller knows it as well. 3498 */ 3499static int 3500add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule) 3501{ 3502 struct ip_fw *rule, *f, *prev; 3503 int l = RULESIZE(input_rule); 3504 3505 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE) 3506 return (EINVAL); 3507 3508 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO); 3509 if (rule == NULL) 3510 return (ENOSPC); 3511 3512 bcopy(input_rule, rule, l); 3513 3514 rule->next = NULL; 3515 rule->next_rule = NULL; 3516 3517 rule->pcnt = 0; 3518 rule->bcnt = 0; 3519 rule->timestamp = 0; 3520 3521 IPFW_WLOCK(chain); 3522 3523 if (chain->rules == NULL) { /* default rule */ 3524 chain->rules = rule; 3525 rule->id = ++chain->id; 3526 goto done; 3527 } 3528 3529 /* 3530 * If rulenum is 0, find highest numbered rule before the 3531 * default rule, and add autoinc_step 3532 */ 3533 if (V_autoinc_step < 1) 3534 V_autoinc_step = 1; 3535 else if (V_autoinc_step > 1000) 3536 V_autoinc_step = 1000; 3537 if (rule->rulenum == 0) { 3538 /* 3539 * locate the highest numbered rule before default 3540 */ 3541 for (f = chain->rules; f; f = f->next) { 3542 if (f->rulenum == IPFW_DEFAULT_RULE) 3543 break; 3544 rule->rulenum = f->rulenum; 3545 } 3546 if (rule->rulenum < IPFW_DEFAULT_RULE - V_autoinc_step) 3547 rule->rulenum += V_autoinc_step; 3548 input_rule->rulenum = rule->rulenum; 3549 } 3550 3551 /* 3552 * Now insert the new rule in the right place in the sorted list. 3553 */ 3554 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) { 3555 if (f->rulenum > rule->rulenum) { /* found the location */ 3556 if (prev) { 3557 rule->next = f; 3558 prev->next = rule; 3559 } else { /* head insert */ 3560 rule->next = chain->rules; 3561 chain->rules = rule; 3562 } 3563 break; 3564 } 3565 } 3566 flush_rule_ptrs(chain); 3567 /* chain->id incremented inside flush_rule_ptrs() */ 3568 rule->id = chain->id; 3569done: 3570 V_static_count++; 3571 V_static_len += l; 3572 IPFW_WUNLOCK(chain); 3573 DEB(printf("ipfw: installed rule %d, static count now %d\n", 3574 rule->rulenum, V_static_count);) 3575 return (0); 3576} 3577 3578/** 3579 * Remove a static rule (including derived * dynamic rules) 3580 * and place it on the ``reap list'' for later reclamation. 3581 * The caller is in charge of clearing rule pointers to avoid 3582 * dangling pointers. 3583 * @return a pointer to the next entry. 3584 * Arguments are not checked, so they better be correct. 3585 */ 3586static struct ip_fw * 3587remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule, 3588 struct ip_fw *prev) 3589{ 3590 struct ip_fw *n; 3591 int l = RULESIZE(rule); 3592 3593 IPFW_WLOCK_ASSERT(chain); 3594 3595 n = rule->next; 3596 IPFW_DYN_LOCK(); 3597 remove_dyn_rule(rule, NULL /* force removal */); 3598 IPFW_DYN_UNLOCK(); 3599 if (prev == NULL) 3600 chain->rules = n; 3601 else 3602 prev->next = n; 3603 V_static_count--; 3604 V_static_len -= l; 3605 3606 rule->next = chain->reap; 3607 chain->reap = rule; 3608 3609 return n; 3610} 3611 3612/* 3613 * Reclaim storage associated with a list of rules. This is 3614 * typically the list created using remove_rule. 3615 * A NULL pointer on input is handled correctly. 3616 */ 3617static void 3618reap_rules(struct ip_fw *head) 3619{ 3620 struct ip_fw *rule; 3621 3622 while ((rule = head) != NULL) { 3623 head = head->next; 3624 free(rule, M_IPFW); 3625 } 3626} 3627 3628/* 3629 * Remove all rules from a chain (except rules in set RESVD_SET 3630 * unless kill_default = 1). The caller is responsible for 3631 * reclaiming storage for the rules left in chain->reap. 3632 */ 3633static void 3634free_chain(struct ip_fw_chain *chain, int kill_default) 3635{ 3636 struct ip_fw *prev, *rule; 3637 3638 IPFW_WLOCK_ASSERT(chain); 3639 3640 chain->reap = NULL; 3641 flush_rule_ptrs(chain); /* more efficient to do outside the loop */ 3642 for (prev = NULL, rule = chain->rules; rule ; ) 3643 if (kill_default || rule->set != RESVD_SET) 3644 rule = remove_rule(chain, rule, prev); 3645 else { 3646 prev = rule; 3647 rule = rule->next; 3648 } 3649} 3650 3651/** 3652 * Remove all rules with given number, and also do set manipulation. 3653 * Assumes chain != NULL && *chain != NULL. 3654 * 3655 * The argument is an u_int32_t. The low 16 bit are the rule or set number, 3656 * the next 8 bits are the new set, the top 8 bits are the command: 3657 * 3658 * 0 delete rules with given number 3659 * 1 delete rules with given set number 3660 * 2 move rules with given number to new set 3661 * 3 move rules with given set number to new set 3662 * 4 swap sets with given numbers 3663 * 5 delete rules with given number and with given set number 3664 */ 3665static int 3666del_entry(struct ip_fw_chain *chain, u_int32_t arg) 3667{ 3668 struct ip_fw *prev = NULL, *rule; 3669 u_int16_t rulenum; /* rule or old_set */ 3670 u_int8_t cmd, new_set; 3671 3672 rulenum = arg & 0xffff; 3673 cmd = (arg >> 24) & 0xff; 3674 new_set = (arg >> 16) & 0xff; 3675 3676 if (cmd > 5 || new_set > RESVD_SET) 3677 return EINVAL; 3678 if (cmd == 0 || cmd == 2 || cmd == 5) { 3679 if (rulenum >= IPFW_DEFAULT_RULE) 3680 return EINVAL; 3681 } else { 3682 if (rulenum > RESVD_SET) /* old_set */ 3683 return EINVAL; 3684 } 3685 3686 IPFW_WLOCK(chain); 3687 rule = chain->rules; /* common starting point */ 3688 chain->reap = NULL; /* prepare for deletions */ 3689 switch (cmd) { 3690 case 0: /* delete rules with given number */ 3691 /* 3692 * locate first rule to delete 3693 */ 3694 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3695 ; 3696 if (rule->rulenum != rulenum) { 3697 IPFW_WUNLOCK(chain); 3698 return EINVAL; 3699 } 3700 3701 /* 3702 * flush pointers outside the loop, then delete all matching 3703 * rules. prev remains the same throughout the cycle. 3704 */ 3705 flush_rule_ptrs(chain); 3706 while (rule->rulenum == rulenum) 3707 rule = remove_rule(chain, rule, prev); 3708 break; 3709 3710 case 1: /* delete all rules with given set number */ 3711 flush_rule_ptrs(chain); 3712 while (rule->rulenum < IPFW_DEFAULT_RULE) { 3713 if (rule->set == rulenum) 3714 rule = remove_rule(chain, rule, prev); 3715 else { 3716 prev = rule; 3717 rule = rule->next; 3718 } 3719 } 3720 break; 3721 3722 case 2: /* move rules with given number to new set */ 3723 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3724 if (rule->rulenum == rulenum) 3725 rule->set = new_set; 3726 break; 3727 3728 case 3: /* move rules with given set number to new set */ 3729 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3730 if (rule->set == rulenum) 3731 rule->set = new_set; 3732 break; 3733 3734 case 4: /* swap two sets */ 3735 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next) 3736 if (rule->set == rulenum) 3737 rule->set = new_set; 3738 else if (rule->set == new_set) 3739 rule->set = rulenum; 3740 break; 3741 3742 case 5: /* delete rules with given number and with given set number. 3743 * rulenum - given rule number; 3744 * new_set - given set number. 3745 */ 3746 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next) 3747 ; 3748 if (rule->rulenum != rulenum) { 3749 IPFW_WUNLOCK(chain); 3750 return (EINVAL); 3751 } 3752 flush_rule_ptrs(chain); 3753 while (rule->rulenum == rulenum) { 3754 if (rule->set == new_set) 3755 rule = remove_rule(chain, rule, prev); 3756 else { 3757 prev = rule; 3758 rule = rule->next; 3759 } 3760 } 3761 } 3762 /* 3763 * Look for rules to reclaim. We grab the list before 3764 * releasing the lock then reclaim them w/o the lock to 3765 * avoid a LOR with dummynet. 3766 */ 3767 rule = chain->reap; 3768 IPFW_WUNLOCK(chain); 3769 reap_rules(rule); 3770 return 0; 3771} 3772 3773/* 3774 * Clear counters for a specific rule. 3775 * The enclosing "table" is assumed locked. 3776 */ 3777static void 3778clear_counters(struct ip_fw *rule, int log_only) 3779{ 3780 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule); 3781 3782 if (log_only == 0) { 3783 rule->bcnt = rule->pcnt = 0; 3784 rule->timestamp = 0; 3785 } 3786 if (l->o.opcode == O_LOG) 3787 l->log_left = l->max_log; 3788} 3789 3790/** 3791 * Reset some or all counters on firewall rules. 3792 * The argument `arg' is an u_int32_t. The low 16 bit are the rule number, 3793 * the next 8 bits are the set number, the top 8 bits are the command: 3794 * 0 work with rules from all set's; 3795 * 1 work with rules only from specified set. 3796 * Specified rule number is zero if we want to clear all entries. 3797 * log_only is 1 if we only want to reset logs, zero otherwise. 3798 */ 3799static int 3800zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only) 3801{ 3802 struct ip_fw *rule; 3803 char *msg; 3804 3805 uint16_t rulenum = arg & 0xffff; 3806 uint8_t set = (arg >> 16) & 0xff; 3807 uint8_t cmd = (arg >> 24) & 0xff; 3808 3809 if (cmd > 1) 3810 return (EINVAL); 3811 if (cmd == 1 && set > RESVD_SET) 3812 return (EINVAL); 3813 3814 IPFW_WLOCK(chain); 3815 if (rulenum == 0) { 3816 V_norule_counter = 0; 3817 for (rule = chain->rules; rule; rule = rule->next) { 3818 /* Skip rules from another set. */ 3819 if (cmd == 1 && rule->set != set) 3820 continue; 3821 clear_counters(rule, log_only); 3822 } 3823 msg = log_only ? "All logging counts reset" : 3824 "Accounting cleared"; 3825 } else { 3826 int cleared = 0; 3827 /* 3828 * We can have multiple rules with the same number, so we 3829 * need to clear them all. 3830 */ 3831 for (rule = chain->rules; rule; rule = rule->next) 3832 if (rule->rulenum == rulenum) { 3833 while (rule && rule->rulenum == rulenum) { 3834 if (cmd == 0 || rule->set == set) 3835 clear_counters(rule, log_only); 3836 rule = rule->next; 3837 } 3838 cleared = 1; 3839 break; 3840 } 3841 if (!cleared) { /* we did not find any matching rules */ 3842 IPFW_WUNLOCK(chain); 3843 return (EINVAL); 3844 } 3845 msg = log_only ? "logging count reset" : "cleared"; 3846 } 3847 IPFW_WUNLOCK(chain); 3848 3849 if (V_fw_verbose) { 3850 int lev = LOG_SECURITY | LOG_NOTICE; 3851 3852 if (rulenum) 3853 log(lev, "ipfw: Entry %d %s.\n", rulenum, msg); 3854 else 3855 log(lev, "ipfw: %s.\n", msg); 3856 } 3857 return (0); 3858} 3859 3860/* 3861 * Check validity of the structure before insert. 3862 * Fortunately rules are simple, so this mostly need to check rule sizes. 3863 */ 3864static int 3865check_ipfw_struct(struct ip_fw *rule, int size) 3866{ 3867 int l, cmdlen = 0; 3868 int have_action=0; 3869 ipfw_insn *cmd; 3870 3871 if (size < sizeof(*rule)) { 3872 printf("ipfw: rule too short\n"); 3873 return (EINVAL); 3874 } 3875 /* first, check for valid size */ 3876 l = RULESIZE(rule); 3877 if (l != size) { 3878 printf("ipfw: size mismatch (have %d want %d)\n", size, l); 3879 return (EINVAL); 3880 } 3881 if (rule->act_ofs >= rule->cmd_len) { 3882 printf("ipfw: bogus action offset (%u > %u)\n", 3883 rule->act_ofs, rule->cmd_len - 1); 3884 return (EINVAL); 3885 } 3886 /* 3887 * Now go for the individual checks. Very simple ones, basically only 3888 * instruction sizes. 3889 */ 3890 for (l = rule->cmd_len, cmd = rule->cmd ; 3891 l > 0 ; l -= cmdlen, cmd += cmdlen) { 3892 cmdlen = F_LEN(cmd); 3893 if (cmdlen > l) { 3894 printf("ipfw: opcode %d size truncated\n", 3895 cmd->opcode); 3896 return EINVAL; 3897 } 3898 DEB(printf("ipfw: opcode %d\n", cmd->opcode);) 3899 switch (cmd->opcode) { 3900 case O_PROBE_STATE: 3901 case O_KEEP_STATE: 3902 case O_PROTO: 3903 case O_IP_SRC_ME: 3904 case O_IP_DST_ME: 3905 case O_LAYER2: 3906 case O_IN: 3907 case O_FRAG: 3908 case O_DIVERTED: 3909 case O_IPOPT: 3910 case O_IPTOS: 3911 case O_IPPRECEDENCE: 3912 case O_IPVER: 3913 case O_TCPWIN: 3914 case O_TCPFLAGS: 3915 case O_TCPOPTS: 3916 case O_ESTAB: 3917 case O_VERREVPATH: 3918 case O_VERSRCREACH: 3919 case O_ANTISPOOF: 3920 case O_IPSEC: 3921#ifdef INET6 3922 case O_IP6_SRC_ME: 3923 case O_IP6_DST_ME: 3924 case O_EXT_HDR: 3925 case O_IP6: 3926#endif 3927 case O_IP4: 3928 case O_TAG: 3929 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3930 goto bad_size; 3931 break; 3932 3933 case O_FIB: 3934 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3935 goto bad_size; 3936 if (cmd->arg1 >= rt_numfibs) { 3937 printf("ipfw: invalid fib number %d\n", 3938 cmd->arg1); 3939 return EINVAL; 3940 } 3941 break; 3942 3943 case O_SETFIB: 3944 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 3945 goto bad_size; 3946 if (cmd->arg1 >= rt_numfibs) { 3947 printf("ipfw: invalid fib number %d\n", 3948 cmd->arg1); 3949 return EINVAL; 3950 } 3951 goto check_action; 3952 3953 case O_UID: 3954 case O_GID: 3955 case O_JAIL: 3956 case O_IP_SRC: 3957 case O_IP_DST: 3958 case O_TCPSEQ: 3959 case O_TCPACK: 3960 case O_PROB: 3961 case O_ICMPTYPE: 3962 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 3963 goto bad_size; 3964 break; 3965 3966 case O_LIMIT: 3967 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit)) 3968 goto bad_size; 3969 break; 3970 3971 case O_LOG: 3972 if (cmdlen != F_INSN_SIZE(ipfw_insn_log)) 3973 goto bad_size; 3974 3975 ((ipfw_insn_log *)cmd)->log_left = 3976 ((ipfw_insn_log *)cmd)->max_log; 3977 3978 break; 3979 3980 case O_IP_SRC_MASK: 3981 case O_IP_DST_MASK: 3982 /* only odd command lengths */ 3983 if ( !(cmdlen & 1) || cmdlen > 31) 3984 goto bad_size; 3985 break; 3986 3987 case O_IP_SRC_SET: 3988 case O_IP_DST_SET: 3989 if (cmd->arg1 == 0 || cmd->arg1 > 256) { 3990 printf("ipfw: invalid set size %d\n", 3991 cmd->arg1); 3992 return EINVAL; 3993 } 3994 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 3995 (cmd->arg1+31)/32 ) 3996 goto bad_size; 3997 break; 3998 3999 case O_IP_SRC_LOOKUP: 4000 case O_IP_DST_LOOKUP: 4001 if (cmd->arg1 >= IPFW_TABLES_MAX) { 4002 printf("ipfw: invalid table number %d\n", 4003 cmd->arg1); 4004 return (EINVAL); 4005 } 4006 if (cmdlen != F_INSN_SIZE(ipfw_insn) && 4007 cmdlen != F_INSN_SIZE(ipfw_insn_u32)) 4008 goto bad_size; 4009 break; 4010 4011 case O_MACADDR2: 4012 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac)) 4013 goto bad_size; 4014 break; 4015 4016 case O_NOP: 4017 case O_IPID: 4018 case O_IPTTL: 4019 case O_IPLEN: 4020 case O_TCPDATALEN: 4021 case O_TAGGED: 4022 if (cmdlen < 1 || cmdlen > 31) 4023 goto bad_size; 4024 break; 4025 4026 case O_MAC_TYPE: 4027 case O_IP_SRCPORT: 4028 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */ 4029 if (cmdlen < 2 || cmdlen > 31) 4030 goto bad_size; 4031 break; 4032 4033 case O_RECV: 4034 case O_XMIT: 4035 case O_VIA: 4036 if (cmdlen != F_INSN_SIZE(ipfw_insn_if)) 4037 goto bad_size; 4038 break; 4039 4040 case O_ALTQ: 4041 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq)) 4042 goto bad_size; 4043 break; 4044 4045 case O_PIPE: 4046 case O_QUEUE: 4047 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 4048 goto bad_size; 4049 goto check_action; 4050 4051 case O_FORWARD_IP: 4052#ifdef IPFIREWALL_FORWARD 4053 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa)) 4054 goto bad_size; 4055 goto check_action; 4056#else 4057 return EINVAL; 4058#endif 4059 4060 case O_DIVERT: 4061 case O_TEE: 4062 if (ip_divert_ptr == NULL) 4063 return EINVAL; 4064 else 4065 goto check_size; 4066 case O_NETGRAPH: 4067 case O_NGTEE: 4068 if (!NG_IPFW_LOADED) 4069 return EINVAL; 4070 else 4071 goto check_size; 4072 case O_NAT: 4073 if (!IPFW_NAT_LOADED) 4074 return EINVAL; 4075 if (cmdlen != F_INSN_SIZE(ipfw_insn_nat)) 4076 goto bad_size; 4077 goto check_action; 4078 case O_FORWARD_MAC: /* XXX not implemented yet */ 4079 case O_CHECK_STATE: 4080 case O_COUNT: 4081 case O_ACCEPT: 4082 case O_DENY: 4083 case O_REJECT: 4084#ifdef INET6 4085 case O_UNREACH6: 4086#endif 4087 case O_SKIPTO: 4088 case O_REASS: 4089check_size: 4090 if (cmdlen != F_INSN_SIZE(ipfw_insn)) 4091 goto bad_size; 4092check_action: 4093 if (have_action) { 4094 printf("ipfw: opcode %d, multiple actions" 4095 " not allowed\n", 4096 cmd->opcode); 4097 return EINVAL; 4098 } 4099 have_action = 1; 4100 if (l != cmdlen) { 4101 printf("ipfw: opcode %d, action must be" 4102 " last opcode\n", 4103 cmd->opcode); 4104 return EINVAL; 4105 } 4106 break; 4107#ifdef INET6 4108 case O_IP6_SRC: 4109 case O_IP6_DST: 4110 if (cmdlen != F_INSN_SIZE(struct in6_addr) + 4111 F_INSN_SIZE(ipfw_insn)) 4112 goto bad_size; 4113 break; 4114 4115 case O_FLOW6ID: 4116 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) + 4117 ((ipfw_insn_u32 *)cmd)->o.arg1) 4118 goto bad_size; 4119 break; 4120 4121 case O_IP6_SRC_MASK: 4122 case O_IP6_DST_MASK: 4123 if ( !(cmdlen & 1) || cmdlen > 127) 4124 goto bad_size; 4125 break; 4126 case O_ICMP6TYPE: 4127 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) ) 4128 goto bad_size; 4129 break; 4130#endif 4131 4132 default: 4133 switch (cmd->opcode) { 4134#ifndef INET6 4135 case O_IP6_SRC_ME: 4136 case O_IP6_DST_ME: 4137 case O_EXT_HDR: 4138 case O_IP6: 4139 case O_UNREACH6: 4140 case O_IP6_SRC: 4141 case O_IP6_DST: 4142 case O_FLOW6ID: 4143 case O_IP6_SRC_MASK: 4144 case O_IP6_DST_MASK: 4145 case O_ICMP6TYPE: 4146 printf("ipfw: no IPv6 support in kernel\n"); 4147 return EPROTONOSUPPORT; 4148#endif 4149 default: 4150 printf("ipfw: opcode %d, unknown opcode\n", 4151 cmd->opcode); 4152 return EINVAL; 4153 } 4154 } 4155 } 4156 if (have_action == 0) { 4157 printf("ipfw: missing action\n"); 4158 return EINVAL; 4159 } 4160 return 0; 4161 4162bad_size: 4163 printf("ipfw: opcode %d size %d wrong\n", 4164 cmd->opcode, cmdlen); 4165 return EINVAL; 4166} 4167 4168/* 4169 * Copy the static and dynamic rules to the supplied buffer 4170 * and return the amount of space actually used. 4171 */ 4172static size_t 4173ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space) 4174{ 4175 char *bp = buf; 4176 char *ep = bp + space; 4177 struct ip_fw *rule; 4178 int i; 4179 time_t boot_seconds; 4180 4181 boot_seconds = boottime.tv_sec; 4182 /* XXX this can take a long time and locking will block packet flow */ 4183 IPFW_RLOCK(chain); 4184 for (rule = chain->rules; rule ; rule = rule->next) { 4185 /* 4186 * Verify the entry fits in the buffer in case the 4187 * rules changed between calculating buffer space and 4188 * now. This would be better done using a generation 4189 * number but should suffice for now. 4190 */ 4191 i = RULESIZE(rule); 4192 if (bp + i <= ep) { 4193 bcopy(rule, bp, i); 4194 /* 4195 * XXX HACK. Store the disable mask in the "next" 4196 * pointer in a wild attempt to keep the ABI the same. 4197 * Why do we do this on EVERY rule? 4198 */ 4199 bcopy(&V_set_disable, 4200 &(((struct ip_fw *)bp)->next_rule), 4201 sizeof(V_set_disable)); 4202 if (((struct ip_fw *)bp)->timestamp) 4203 ((struct ip_fw *)bp)->timestamp += boot_seconds; 4204 bp += i; 4205 } 4206 } 4207 IPFW_RUNLOCK(chain); 4208 if (V_ipfw_dyn_v) { 4209 ipfw_dyn_rule *p, *last = NULL; 4210 4211 IPFW_DYN_LOCK(); 4212 for (i = 0 ; i < V_curr_dyn_buckets; i++) 4213 for (p = V_ipfw_dyn_v[i] ; p != NULL; p = p->next) { 4214 if (bp + sizeof *p <= ep) { 4215 ipfw_dyn_rule *dst = 4216 (ipfw_dyn_rule *)bp; 4217 bcopy(p, dst, sizeof *p); 4218 bcopy(&(p->rule->rulenum), &(dst->rule), 4219 sizeof(p->rule->rulenum)); 4220 /* 4221 * store set number into high word of 4222 * dst->rule pointer. 4223 */ 4224 bcopy(&(p->rule->set), 4225 (char *)&dst->rule + 4226 sizeof(p->rule->rulenum), 4227 sizeof(p->rule->set)); 4228 /* 4229 * store a non-null value in "next". 4230 * The userland code will interpret a 4231 * NULL here as a marker 4232 * for the last dynamic rule. 4233 */ 4234 bcopy(&dst, &dst->next, sizeof(dst)); 4235 last = dst; 4236 dst->expire = 4237 TIME_LEQ(dst->expire, time_uptime) ? 4238 0 : dst->expire - time_uptime ; 4239 bp += sizeof(ipfw_dyn_rule); 4240 } 4241 } 4242 IPFW_DYN_UNLOCK(); 4243 if (last != NULL) /* mark last dynamic rule */ 4244 bzero(&last->next, sizeof(last)); 4245 } 4246 return (bp - (char *)buf); 4247} 4248 4249 4250/** 4251 * {set|get}sockopt parser. 4252 */ 4253static int 4254ipfw_ctl(struct sockopt *sopt) 4255{ 4256#define RULE_MAXSIZE (256*sizeof(u_int32_t)) 4257 int error; 4258 size_t size; 4259 struct ip_fw *buf, *rule; 4260 u_int32_t rulenum[2]; 4261 4262 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW); 4263 if (error) 4264 return (error); 4265 4266 /* 4267 * Disallow modifications in really-really secure mode, but still allow 4268 * the logging counters to be reset. 4269 */ 4270 if (sopt->sopt_name == IP_FW_ADD || 4271 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) { 4272 error = securelevel_ge(sopt->sopt_td->td_ucred, 3); 4273 if (error) 4274 return (error); 4275 } 4276 4277 error = 0; 4278 4279 switch (sopt->sopt_name) { 4280 case IP_FW_GET: 4281 /* 4282 * pass up a copy of the current rules. Static rules 4283 * come first (the last of which has number IPFW_DEFAULT_RULE), 4284 * followed by a possibly empty list of dynamic rule. 4285 * The last dynamic rule has NULL in the "next" field. 4286 * 4287 * Note that the calculated size is used to bound the 4288 * amount of data returned to the user. The rule set may 4289 * change between calculating the size and returning the 4290 * data in which case we'll just return what fits. 4291 */ 4292 size = V_static_len; /* size of static rules */ 4293 if (V_ipfw_dyn_v) /* add size of dyn.rules */ 4294 size += (V_dyn_count * sizeof(ipfw_dyn_rule)); 4295 4296 if (size >= sopt->sopt_valsize) 4297 break; 4298 /* 4299 * XXX todo: if the user passes a short length just to know 4300 * how much room is needed, do not bother filling up the 4301 * buffer, just jump to the sooptcopyout. 4302 */ 4303 buf = malloc(size, M_TEMP, M_WAITOK); 4304 error = sooptcopyout(sopt, buf, 4305 ipfw_getrules(&V_layer3_chain, buf, size)); 4306 free(buf, M_TEMP); 4307 break; 4308 4309 case IP_FW_FLUSH: 4310 /* 4311 * Normally we cannot release the lock on each iteration. 4312 * We could do it here only because we start from the head all 4313 * the times so there is no risk of missing some entries. 4314 * On the other hand, the risk is that we end up with 4315 * a very inconsistent ruleset, so better keep the lock 4316 * around the whole cycle. 4317 * 4318 * XXX this code can be improved by resetting the head of 4319 * the list to point to the default rule, and then freeing 4320 * the old list without the need for a lock. 4321 */ 4322 4323 IPFW_WLOCK(&V_layer3_chain); 4324 free_chain(&V_layer3_chain, 0 /* keep default rule */); 4325 rule = V_layer3_chain.reap; 4326 IPFW_WUNLOCK(&V_layer3_chain); 4327 reap_rules(rule); 4328 break; 4329 4330 case IP_FW_ADD: 4331 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK); 4332 error = sooptcopyin(sopt, rule, RULE_MAXSIZE, 4333 sizeof(struct ip_fw) ); 4334 if (error == 0) 4335 error = check_ipfw_struct(rule, sopt->sopt_valsize); 4336 if (error == 0) { 4337 error = add_rule(&V_layer3_chain, rule); 4338 size = RULESIZE(rule); 4339 if (!error && sopt->sopt_dir == SOPT_GET) 4340 error = sooptcopyout(sopt, rule, size); 4341 } 4342 free(rule, M_TEMP); 4343 break; 4344 4345 case IP_FW_DEL: 4346 /* 4347 * IP_FW_DEL is used for deleting single rules or sets, 4348 * and (ab)used to atomically manipulate sets. Argument size 4349 * is used to distinguish between the two: 4350 * sizeof(u_int32_t) 4351 * delete single rule or set of rules, 4352 * or reassign rules (or sets) to a different set. 4353 * 2*sizeof(u_int32_t) 4354 * atomic disable/enable sets. 4355 * first u_int32_t contains sets to be disabled, 4356 * second u_int32_t contains sets to be enabled. 4357 */ 4358 error = sooptcopyin(sopt, rulenum, 4359 2*sizeof(u_int32_t), sizeof(u_int32_t)); 4360 if (error) 4361 break; 4362 size = sopt->sopt_valsize; 4363 if (size == sizeof(u_int32_t)) /* delete or reassign */ 4364 error = del_entry(&V_layer3_chain, rulenum[0]); 4365 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */ 4366 V_set_disable = 4367 (V_set_disable | rulenum[0]) & ~rulenum[1] & 4368 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */ 4369 else 4370 error = EINVAL; 4371 break; 4372 4373 case IP_FW_ZERO: 4374 case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */ 4375 rulenum[0] = 0; 4376 if (sopt->sopt_val != 0) { 4377 error = sooptcopyin(sopt, rulenum, 4378 sizeof(u_int32_t), sizeof(u_int32_t)); 4379 if (error) 4380 break; 4381 } 4382 error = zero_entry(&V_layer3_chain, rulenum[0], 4383 sopt->sopt_name == IP_FW_RESETLOG); 4384 break; 4385 4386 case IP_FW_TABLE_ADD: 4387 { 4388 ipfw_table_entry ent; 4389 4390 error = sooptcopyin(sopt, &ent, 4391 sizeof(ent), sizeof(ent)); 4392 if (error) 4393 break; 4394 error = add_table_entry(&V_layer3_chain, ent.tbl, 4395 ent.addr, ent.masklen, ent.value); 4396 } 4397 break; 4398 4399 case IP_FW_TABLE_DEL: 4400 { 4401 ipfw_table_entry ent; 4402 4403 error = sooptcopyin(sopt, &ent, 4404 sizeof(ent), sizeof(ent)); 4405 if (error) 4406 break; 4407 error = del_table_entry(&V_layer3_chain, ent.tbl, 4408 ent.addr, ent.masklen); 4409 } 4410 break; 4411 4412 case IP_FW_TABLE_FLUSH: 4413 { 4414 u_int16_t tbl; 4415 4416 error = sooptcopyin(sopt, &tbl, 4417 sizeof(tbl), sizeof(tbl)); 4418 if (error) 4419 break; 4420 IPFW_WLOCK(&V_layer3_chain); 4421 error = flush_table(&V_layer3_chain, tbl); 4422 IPFW_WUNLOCK(&V_layer3_chain); 4423 } 4424 break; 4425 4426 case IP_FW_TABLE_GETSIZE: 4427 { 4428 u_int32_t tbl, cnt; 4429 4430 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl), 4431 sizeof(tbl)))) 4432 break; 4433 IPFW_RLOCK(&V_layer3_chain); 4434 error = count_table(&V_layer3_chain, tbl, &cnt); 4435 IPFW_RUNLOCK(&V_layer3_chain); 4436 if (error) 4437 break; 4438 error = sooptcopyout(sopt, &cnt, sizeof(cnt)); 4439 } 4440 break; 4441 4442 case IP_FW_TABLE_LIST: 4443 { 4444 ipfw_table *tbl; 4445 4446 if (sopt->sopt_valsize < sizeof(*tbl)) { 4447 error = EINVAL; 4448 break; 4449 } 4450 size = sopt->sopt_valsize; 4451 tbl = malloc(size, M_TEMP, M_WAITOK); 4452 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl)); 4453 if (error) { 4454 free(tbl, M_TEMP); 4455 break; 4456 } 4457 tbl->size = (size - sizeof(*tbl)) / 4458 sizeof(ipfw_table_entry); 4459 IPFW_RLOCK(&V_layer3_chain); 4460 error = dump_table(&V_layer3_chain, tbl); 4461 IPFW_RUNLOCK(&V_layer3_chain); 4462 if (error) { 4463 free(tbl, M_TEMP); 4464 break; 4465 } 4466 error = sooptcopyout(sopt, tbl, size); 4467 free(tbl, M_TEMP); 4468 } 4469 break; 4470 4471 case IP_FW_NAT_CFG: 4472 if (IPFW_NAT_LOADED) 4473 error = ipfw_nat_cfg_ptr(sopt); 4474 else { 4475 printf("IP_FW_NAT_CFG: %s\n", 4476 "ipfw_nat not present, please load it"); 4477 error = EINVAL; 4478 } 4479 break; 4480 4481 case IP_FW_NAT_DEL: 4482 if (IPFW_NAT_LOADED) 4483 error = ipfw_nat_del_ptr(sopt); 4484 else { 4485 printf("IP_FW_NAT_DEL: %s\n", 4486 "ipfw_nat not present, please load it"); 4487 error = EINVAL; 4488 } 4489 break; 4490 4491 case IP_FW_NAT_GET_CONFIG: 4492 if (IPFW_NAT_LOADED) 4493 error = ipfw_nat_get_cfg_ptr(sopt); 4494 else { 4495 printf("IP_FW_NAT_GET_CFG: %s\n", 4496 "ipfw_nat not present, please load it"); 4497 error = EINVAL; 4498 } 4499 break; 4500 4501 case IP_FW_NAT_GET_LOG: 4502 if (IPFW_NAT_LOADED) 4503 error = ipfw_nat_get_log_ptr(sopt); 4504 else { 4505 printf("IP_FW_NAT_GET_LOG: %s\n", 4506 "ipfw_nat not present, please load it"); 4507 error = EINVAL; 4508 } 4509 break; 4510 4511 default: 4512 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name); 4513 error = EINVAL; 4514 } 4515 4516 return (error); 4517#undef RULE_MAXSIZE 4518} 4519 4520 4521/* 4522 * This procedure is only used to handle keepalives. It is invoked 4523 * every dyn_keepalive_period 4524 */ 4525static void 4526ipfw_tick(void * vnetx) 4527{ 4528 struct mbuf *m0, *m, *mnext, **mtailp; 4529 int i; 4530 ipfw_dyn_rule *q; 4531#ifdef VIMAGE 4532 struct vnet *vp = vnetx; 4533#endif 4534 4535 CURVNET_SET(vp); 4536 if (V_dyn_keepalive == 0 || V_ipfw_dyn_v == NULL || V_dyn_count == 0) 4537 goto done; 4538 4539 /* 4540 * We make a chain of packets to go out here -- not deferring 4541 * until after we drop the IPFW dynamic rule lock would result 4542 * in a lock order reversal with the normal packet input -> ipfw 4543 * call stack. 4544 */ 4545 m0 = NULL; 4546 mtailp = &m0; 4547 IPFW_DYN_LOCK(); 4548 for (i = 0 ; i < V_curr_dyn_buckets ; i++) { 4549 for (q = V_ipfw_dyn_v[i] ; q ; q = q->next ) { 4550 if (q->dyn_type == O_LIMIT_PARENT) 4551 continue; 4552 if (q->id.proto != IPPROTO_TCP) 4553 continue; 4554 if ( (q->state & BOTH_SYN) != BOTH_SYN) 4555 continue; 4556 if (TIME_LEQ(time_uptime + V_dyn_keepalive_interval, 4557 q->expire)) 4558 continue; /* too early */ 4559 if (TIME_LEQ(q->expire, time_uptime)) 4560 continue; /* too late, rule expired */ 4561 4562 *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1, 4563 q->ack_fwd, TH_SYN); 4564 if (*mtailp != NULL) 4565 mtailp = &(*mtailp)->m_nextpkt; 4566 *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1, 4567 q->ack_rev, 0); 4568 if (*mtailp != NULL) 4569 mtailp = &(*mtailp)->m_nextpkt; 4570 } 4571 } 4572 IPFW_DYN_UNLOCK(); 4573 for (m = mnext = m0; m != NULL; m = mnext) { 4574 mnext = m->m_nextpkt; 4575 m->m_nextpkt = NULL; 4576 ip_output(m, NULL, NULL, 0, NULL, NULL); 4577 } 4578done: 4579 callout_reset(&V_ipfw_timeout, V_dyn_keepalive_period * hz, 4580 ipfw_tick, vnetx); 4581 CURVNET_RESTORE(); 4582} 4583 4584/**************** 4585 * Stuff that must be initialised only on boot or module load 4586 */ 4587static int 4588ipfw_init(void) 4589{ 4590 int error = 0; 4591 4592 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule", 4593 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL, 4594 UMA_ALIGN_PTR, 0); 4595 4596 IPFW_DYN_LOCK_INIT(); 4597 /* 4598 * Only print out this stuff the first time around, 4599 * when called from the sysinit code. 4600 */ 4601 printf("ipfw2 " 4602#ifdef INET6 4603 "(+ipv6) " 4604#endif 4605 "initialized, divert %s, nat %s, " 4606 "rule-based forwarding " 4607#ifdef IPFIREWALL_FORWARD 4608 "enabled, " 4609#else 4610 "disabled, " 4611#endif 4612 "default to %s, logging ", 4613#ifdef IPDIVERT 4614 "enabled", 4615#else 4616 "loadable", 4617#endif 4618#ifdef IPFIREWALL_NAT 4619 "enabled", 4620#else 4621 "loadable", 4622#endif 4623 default_to_accept ? "accept" : "deny"); 4624 4625 /* 4626 * Note: V_xxx variables can be accessed here but the vnet specific 4627 * initializer may not have been called yet for the VIMAGE case. 4628 * Tuneables will have been processed. We will print out values for 4629 * the default vnet. 4630 * XXX This should all be rationalized AFTER 8.0 4631 */ 4632 if (V_fw_verbose == 0) 4633 printf("disabled\n"); 4634 else if (V_verbose_limit == 0) 4635 printf("unlimited\n"); 4636 else 4637 printf("limited to %d packets/entry by default\n", 4638 V_verbose_limit); 4639 4640 /* 4641 * Hook us up to pfil. 4642 * Eventually pfil will be per vnet. 4643 */ 4644 if ((error = ipfw_hook()) != 0) { 4645 printf("ipfw_hook() error\n"); 4646 return (error); 4647 } 4648#ifdef INET6 4649 if ((error = ipfw6_hook()) != 0) { 4650 printf("ipfw6_hook() error\n"); 4651 return (error); 4652 } 4653#endif 4654 /* 4655 * Other things that are only done the first time. 4656 * (now that we a re guaranteed of success). 4657 */ 4658 ip_fw_ctl_ptr = ipfw_ctl; 4659 ip_fw_chk_ptr = ipfw_chk; 4660 return (error); 4661} 4662 4663/**************** 4664 * Stuff that must be initialized for every instance 4665 * (including the first of course). 4666 */ 4667static int 4668vnet_ipfw_init(const void *unused) 4669{ 4670 int error; 4671 struct ip_fw default_rule; 4672 4673 /* First set up some values that are compile time options */ 4674#ifdef IPFIREWALL_VERBOSE 4675 V_fw_verbose = 1; 4676#endif 4677#ifdef IPFIREWALL_VERBOSE_LIMIT 4678 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT; 4679#endif 4680 4681 error = init_tables(&V_layer3_chain); 4682 if (error) { 4683 panic("init_tables"); /* XXX Marko fix this ! */ 4684 } 4685#ifdef IPFIREWALL_NAT 4686 LIST_INIT(&V_layer3_chain.nat); 4687#endif 4688 4689 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */ 4690 4691 V_ipfw_dyn_v = NULL; 4692 V_dyn_buckets = 256; /* must be power of 2 */ 4693 V_curr_dyn_buckets = 256; /* must be power of 2 */ 4694 4695 V_dyn_ack_lifetime = 300; 4696 V_dyn_syn_lifetime = 20; 4697 V_dyn_fin_lifetime = 1; 4698 V_dyn_rst_lifetime = 1; 4699 V_dyn_udp_lifetime = 10; 4700 V_dyn_short_lifetime = 5; 4701 4702 V_dyn_keepalive_interval = 20; 4703 V_dyn_keepalive_period = 5; 4704 V_dyn_keepalive = 1; /* do send keepalives */ 4705 4706 V_dyn_max = 4096; /* max # of dynamic rules */ 4707 4708 V_fw_deny_unknown_exthdrs = 1; 4709 4710 V_layer3_chain.rules = NULL; 4711 IPFW_LOCK_INIT(&V_layer3_chain); 4712 callout_init(&V_ipfw_timeout, CALLOUT_MPSAFE); 4713 4714 bzero(&default_rule, sizeof default_rule); 4715 default_rule.act_ofs = 0; 4716 default_rule.rulenum = IPFW_DEFAULT_RULE; 4717 default_rule.cmd_len = 1; 4718 default_rule.set = RESVD_SET; 4719 default_rule.cmd[0].len = 1; 4720 default_rule.cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY; 4721 error = add_rule(&V_layer3_chain, &default_rule); 4722 4723 if (error != 0) { 4724 printf("ipfw2: error %u initializing default rule " 4725 "(support disabled)\n", error); 4726 IPFW_LOCK_DESTROY(&V_layer3_chain); 4727 printf("leaving ipfw_iattach (1) with error %d\n", error); 4728 return (error); 4729 } 4730 4731 ip_fw_default_rule = V_layer3_chain.rules; 4732 4733 if (error) { 4734 IPFW_LOCK_DESTROY(&V_layer3_chain); 4735 printf("leaving ipfw_iattach (2) with error %d\n", error); 4736 return (error); 4737 } 4738#ifdef VIMAGE /* want a better way to do this */ 4739 callout_reset(&V_ipfw_timeout, hz, ipfw_tick, curvnet); 4740#else 4741 callout_reset(&V_ipfw_timeout, hz, ipfw_tick, NULL); 4742#endif 4743 4744 /* First set up some values that are compile time options */ 4745 V_ipfw_vnet_ready = 1; /* Open for business */ 4746 return (0); 4747} 4748 4749/********************** 4750 * Called for the removal of the last instance only on module unload. 4751 */ 4752static void 4753ipfw_destroy(void) 4754{ 4755 4756 uma_zdestroy(ipfw_dyn_rule_zone); 4757 IPFW_DYN_LOCK_DESTROY(); 4758 printf("IP firewall unloaded\n"); 4759} 4760 4761/*********************** 4762 * Called for the removal of each instance. 4763 */ 4764static int 4765vnet_ipfw_uninit(const void *unused) 4766{ 4767 struct ip_fw *reap; 4768 4769 V_ipfw_vnet_ready = 0; /* tell new callers to go away */ 4770 callout_drain(&V_ipfw_timeout); 4771 /* We wait on the wlock here until the last user leaves */ 4772 IPFW_WLOCK(&V_layer3_chain); 4773 flush_tables(&V_layer3_chain); 4774 V_layer3_chain.reap = NULL; 4775 free_chain(&V_layer3_chain, 1 /* kill default rule */); 4776 reap = V_layer3_chain.reap; 4777 V_layer3_chain.reap = NULL; 4778 IPFW_WUNLOCK(&V_layer3_chain); 4779 if (reap != NULL) 4780 reap_rules(reap); 4781 IPFW_LOCK_DESTROY(&V_layer3_chain); 4782 if (V_ipfw_dyn_v != NULL) 4783 free(V_ipfw_dyn_v, M_IPFW); 4784 return 0; 4785} 4786 4787/* 4788 * Module event handler. 4789 * In general we have the choice of handling most of these events by the 4790 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to 4791 * use the SYSINIT handlers as they are more capable of expressing the 4792 * flow of control during module and vnet operations, so this is just 4793 * a skeleton. Note there is no SYSINIT equivalent of the module 4794 * SHUTDOWN handler, but we don't have anything to do in that case anyhow. 4795 */ 4796static int 4797ipfw_modevent(module_t mod, int type, void *unused) 4798{ 4799 int err = 0; 4800 4801 switch (type) { 4802 case MOD_LOAD: 4803 /* Called once at module load or 4804 * system boot if compiled in. */ 4805 break; 4806 case MOD_UNLOAD: 4807 break; 4808 case MOD_QUIESCE: 4809 /* Yes, the unhooks can return errors, we can safely ignore 4810 * them. Eventually these will be done per jail as they 4811 * shut down. We will wait on each vnet's l3 lock as existing 4812 * callers go away. 4813 */ 4814 ipfw_unhook(); 4815#ifdef INET6 4816 ipfw6_unhook(); 4817#endif 4818 /* layer2 and other entrypoints still come in this way. */ 4819 ip_fw_chk_ptr = NULL; 4820 ip_fw_ctl_ptr = NULL; 4821 /* Called during unload. */ 4822 break; 4823 case MOD_SHUTDOWN: 4824 /* Called during system shutdown. */ 4825 break; 4826 default: 4827 err = EOPNOTSUPP; 4828 break; 4829 } 4830 return err; 4831} 4832 4833static moduledata_t ipfwmod = { 4834 "ipfw", 4835 ipfw_modevent, 4836 0 4837}; 4838 4839/* Define startup order. */ 4840#define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_IFATTACHDOMAIN 4841#define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */ 4842#define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */ 4843#define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */ 4844 4845DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER); 4846MODULE_VERSION(ipfw, 2); 4847/* should declare some dependencies here */ 4848 4849/* 4850 * Starting up. Done in order after ipfwmod() has been called. 4851 * VNET_SYSINIT is also called for each existing vnet and each new vnet. 4852 */ 4853SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 4854 ipfw_init, NULL); 4855VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 4856 vnet_ipfw_init, NULL); 4857 4858/* 4859 * Closing up shop. These are done in REVERSE ORDER, but still 4860 * after ipfwmod() has been called. Not called on reboot. 4861 * VNET_SYSUNINIT is also called for each exiting vnet as it exits. 4862 * or when the module is unloaded. 4863 */ 4864SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER, 4865 ipfw_destroy, NULL); 4866VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER, 4867 vnet_ipfw_uninit, NULL); 4868 4869 4870