2 * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions
7 * 1. Redistributions of source code must retain the above copyright
8 * notice, this list of conditions and the following disclaimer.
9 * 2. Redistributions in binary form must reproduce the above copyright
10 * notice, this list of conditions and the following disclaimer in the
11 * documentation and/or other materials provided with the distribution.
13 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
14 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
15 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
16 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
17 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
18 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
19 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
20 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
21 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
22 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 #include <sys/cdefs.h>
27 __FBSDID("$FreeBSD$");
30 * The FreeBSD IP packet firewall, main file
34 #include "opt_ipdivert.h"
37 #error IPFIREWALL requires INET.
39 #include "opt_inet6.h"
40 #include "opt_ipsec.h"
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/condvar.h>
45 #include <sys/eventhandler.h>
46 #include <sys/malloc.h>
48 #include <sys/kernel.h>
51 #include <sys/module.h>
54 #include <sys/rwlock.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
57 #include <sys/sysctl.h>
58 #include <sys/syslog.h>
59 #include <sys/ucred.h>
60 #include <net/ethernet.h> /* for ETHERTYPE_IP */
62 #include <net/route.h>
63 #include <net/pf_mtag.h>
67 #include <netinet/in.h>
68 #include <netinet/in_var.h>
69 #include <netinet/in_pcb.h>
70 #include <netinet/ip.h>
71 #include <netinet/ip_var.h>
72 #include <netinet/ip_icmp.h>
73 #include <netinet/ip_fw.h>
74 #include <netinet/ip_carp.h>
75 #include <netinet/pim.h>
76 #include <netinet/tcp_var.h>
77 #include <netinet/udp.h>
78 #include <netinet/udp_var.h>
79 #include <netinet/sctp.h>
81 #include <netinet/ip6.h>
82 #include <netinet/icmp6.h>
84 #include <netinet6/in6_pcb.h>
85 #include <netinet6/scope6_var.h>
86 #include <netinet6/ip6_var.h>
89 #include <netpfil/ipfw/ip_fw_private.h>
91 #include <machine/in_cksum.h> /* XXX for in_cksum */
94 #include <security/mac/mac_framework.h>
98 * static variables followed by global ones.
99 * All ipfw global variables are here.
102 /* ipfw_vnet_ready controls when we are open for business */
103 static VNET_DEFINE(int, ipfw_vnet_ready) = 0;
104 #define V_ipfw_vnet_ready VNET(ipfw_vnet_ready)
106 static VNET_DEFINE(int, fw_deny_unknown_exthdrs);
107 #define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs)
109 static VNET_DEFINE(int, fw_permit_single_frag6) = 1;
110 #define V_fw_permit_single_frag6 VNET(fw_permit_single_frag6)
112 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
113 static int default_to_accept = 1;
115 static int default_to_accept;
118 VNET_DEFINE(int, autoinc_step);
119 VNET_DEFINE(int, fw_one_pass) = 1;
121 VNET_DEFINE(unsigned int, fw_tables_max);
122 /* Use 128 tables by default */
123 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT;
126 * Each rule belongs to one of 32 different sets (0..31).
127 * The variable set_disable contains one bit per set.
128 * If the bit is set, all rules in the corresponding set
129 * are disabled. Set RESVD_SET(31) is reserved for the default rule
130 * and rules that are not deleted by the flush command,
131 * and CANNOT be disabled.
132 * Rules in set RESVD_SET can only be deleted individually.
134 VNET_DEFINE(u_int32_t, set_disable);
135 #define V_set_disable VNET(set_disable)
137 VNET_DEFINE(int, fw_verbose);
138 /* counter for ipfw_log(NULL...) */
139 VNET_DEFINE(u_int64_t, norule_counter);
140 VNET_DEFINE(int, verbose_limit);
142 /* layer3_chain contains the list of rules for layer 3 */
143 VNET_DEFINE(struct ip_fw_chain, layer3_chain);
145 ipfw_nat_t *ipfw_nat_ptr = NULL;
146 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
147 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
148 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
149 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
150 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
153 uint32_t dummy_def = IPFW_DEFAULT_RULE;
154 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS);
158 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
159 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
160 CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
161 "Only do a single pass through ipfw when using dummynet(4)");
162 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
163 CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
164 "Rule number auto-increment step");
165 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose,
166 CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
167 "Log matches to ipfw rules");
168 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
169 CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
170 "Set upper limit of matches of ipfw rules logged");
171 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
173 "The default/max possible rule number.");
174 SYSCTL_VNET_PROC(_net_inet_ip_fw, OID_AUTO, tables_max,
175 CTLTYPE_UINT|CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU",
176 "Maximum number of tables");
177 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
178 &default_to_accept, 0,
179 "Make the default rule accept all packets.");
180 TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept);
181 TUNABLE_INT("net.inet.ip.fw.tables_max", &default_fw_tables);
182 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, static_count,
183 CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
184 "Number of static rules");
187 SYSCTL_DECL(_net_inet6_ip6);
188 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
189 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
190 CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0,
191 "Deny packets with unknown IPv6 Extension Headers");
192 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
193 CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_permit_single_frag6), 0,
194 "Permit single packet IPv6 fragments");
199 #endif /* SYSCTL_NODE */
203 * Some macros used in the various matching options.
204 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
205 * Other macros just cast void * into the appropriate type
207 #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
208 #define TCP(p) ((struct tcphdr *)(p))
209 #define SCTP(p) ((struct sctphdr *)(p))
210 #define UDP(p) ((struct udphdr *)(p))
211 #define ICMP(p) ((struct icmphdr *)(p))
212 #define ICMP6(p) ((struct icmp6_hdr *)(p))
215 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
217 int type = icmp->icmp_type;
219 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
222 #define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
223 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
226 is_icmp_query(struct icmphdr *icmp)
228 int type = icmp->icmp_type;
230 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
235 * The following checks use two arrays of 8 or 16 bits to store the
236 * bits that we want set or clear, respectively. They are in the
237 * low and high half of cmd->arg1 or cmd->d[0].
239 * We scan options and store the bits we find set. We succeed if
241 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
243 * The code is sometimes optimized not to store additional variables.
247 flags_match(ipfw_insn *cmd, u_int8_t bits)
252 if ( ((cmd->arg1 & 0xff) & bits) != 0)
253 return 0; /* some bits we want set were clear */
254 want_clear = (cmd->arg1 >> 8) & 0xff;
255 if ( (want_clear & bits) != want_clear)
256 return 0; /* some bits we want clear were set */
261 ipopts_match(struct ip *ip, ipfw_insn *cmd)
263 int optlen, bits = 0;
264 u_char *cp = (u_char *)(ip + 1);
265 int x = (ip->ip_hl << 2) - sizeof (struct ip);
267 for (; x > 0; x -= optlen, cp += optlen) {
268 int opt = cp[IPOPT_OPTVAL];
270 if (opt == IPOPT_EOL)
272 if (opt == IPOPT_NOP)
275 optlen = cp[IPOPT_OLEN];
276 if (optlen <= 0 || optlen > x)
277 return 0; /* invalid or truncated */
285 bits |= IP_FW_IPOPT_LSRR;
289 bits |= IP_FW_IPOPT_SSRR;
293 bits |= IP_FW_IPOPT_RR;
297 bits |= IP_FW_IPOPT_TS;
301 return (flags_match(cmd, bits));
305 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
307 int optlen, bits = 0;
308 u_char *cp = (u_char *)(tcp + 1);
309 int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
311 for (; x > 0; x -= optlen, cp += optlen) {
313 if (opt == TCPOPT_EOL)
315 if (opt == TCPOPT_NOP)
329 bits |= IP_FW_TCPOPT_MSS;
333 bits |= IP_FW_TCPOPT_WINDOW;
336 case TCPOPT_SACK_PERMITTED:
338 bits |= IP_FW_TCPOPT_SACK;
341 case TCPOPT_TIMESTAMP:
342 bits |= IP_FW_TCPOPT_TS;
347 return (flags_match(cmd, bits));
351 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain, uint32_t *tablearg)
353 if (ifp == NULL) /* no iface with this packet, match fails */
355 /* Check by name or by IP address */
356 if (cmd->name[0] != '\0') { /* match by name */
357 if (cmd->name[0] == '\1') /* use tablearg to match */
358 return ipfw_lookup_table_extended(chain, cmd->p.glob,
359 ifp->if_xname, tablearg, IPFW_TABLE_INTERFACE);
362 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
365 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
369 #ifdef __FreeBSD__ /* and OSX too ? */
373 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
374 if (ia->ifa_addr->sa_family != AF_INET)
376 if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
377 (ia->ifa_addr))->sin_addr.s_addr) {
378 if_addr_runlock(ifp);
379 return(1); /* match */
382 if_addr_runlock(ifp);
383 #endif /* __FreeBSD__ */
385 return(0); /* no match, fail ... */
389 * The verify_path function checks if a route to the src exists and
390 * if it is reachable via ifp (when provided).
392 * The 'verrevpath' option checks that the interface that an IP packet
393 * arrives on is the same interface that traffic destined for the
394 * packet's source address would be routed out of.
395 * The 'versrcreach' option just checks that the source address is
396 * reachable via any route (except default) in the routing table.
397 * These two are a measure to block forged packets. This is also
398 * commonly known as "anti-spoofing" or Unicast Reverse Path
399 * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
400 * is purposely reminiscent of the Cisco IOS command,
402 * ip verify unicast reverse-path
403 * ip verify unicast source reachable-via any
405 * which implements the same functionality. But note that the syntax
406 * is misleading, and the check may be performed on all IP packets
407 * whether unicast, multicast, or broadcast.
410 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
416 struct sockaddr_in *dst;
418 bzero(&ro, sizeof(ro));
420 dst = (struct sockaddr_in *)&(ro.ro_dst);
421 dst->sin_family = AF_INET;
422 dst->sin_len = sizeof(*dst);
424 in_rtalloc_ign(&ro, 0, fib);
426 if (ro.ro_rt == NULL)
430 * If ifp is provided, check for equality with rtentry.
431 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
432 * in order to pass packets injected back by if_simloop():
433 * if useloopback == 1 routing entry (via lo0) for our own address
434 * may exist, so we need to handle routing assymetry.
436 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
441 /* if no ifp provided, check if rtentry is not default route */
443 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) {
448 /* or if this is a blackhole/reject route */
449 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
454 /* found valid route */
457 #endif /* __FreeBSD__ */
462 * ipv6 specific rules here...
465 icmp6type_match (int type, ipfw_insn_u32 *cmd)
467 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
471 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
474 for (i=0; i <= cmd->o.arg1; ++i )
475 if (curr_flow == cmd->d[i] )
480 /* support for IP6_*_ME opcodes */
482 search_ip6_addr_net (struct in6_addr * ip6_addr)
486 struct in6_ifaddr *fdm;
487 struct in6_addr copia;
489 TAILQ_FOREACH(mdc, &V_ifnet, if_link) {
491 TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) {
492 if (mdc2->ifa_addr->sa_family == AF_INET6) {
493 fdm = (struct in6_ifaddr *)mdc2;
494 copia = fdm->ia_addr.sin6_addr;
495 /* need for leaving scope_id in the sock_addr */
496 in6_clearscope(&copia);
497 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) {
498 if_addr_runlock(mdc);
503 if_addr_runlock(mdc);
509 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
512 struct sockaddr_in6 *dst;
514 bzero(&ro, sizeof(ro));
516 dst = (struct sockaddr_in6 * )&(ro.ro_dst);
517 dst->sin6_family = AF_INET6;
518 dst->sin6_len = sizeof(*dst);
519 dst->sin6_addr = *src;
521 in6_rtalloc_ign(&ro, 0, fib);
522 if (ro.ro_rt == NULL)
526 * if ifp is provided, check for equality with rtentry
527 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
528 * to support the case of sending packets to an address of our own.
529 * (where the former interface is the first argument of if_simloop()
530 * (=ifp), the latter is lo0)
532 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
537 /* if no ifp provided, check if rtentry is not default route */
539 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) {
544 /* or if this is a blackhole/reject route */
545 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
550 /* found valid route */
557 is_icmp6_query(int icmp6_type)
559 if ((icmp6_type <= ICMP6_MAXTYPE) &&
560 (icmp6_type == ICMP6_ECHO_REQUEST ||
561 icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
562 icmp6_type == ICMP6_WRUREQUEST ||
563 icmp6_type == ICMP6_FQDN_QUERY ||
564 icmp6_type == ICMP6_NI_QUERY))
571 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
576 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
578 tcp = (struct tcphdr *)((char *)ip6 + hlen);
580 if ((tcp->th_flags & TH_RST) == 0) {
582 m0 = ipfw_send_pkt(args->m, &(args->f_id),
583 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
584 tcp->th_flags | TH_RST);
586 ip6_output(m0, NULL, NULL, 0, NULL, NULL,
590 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
593 * Unlike above, the mbufs need to line up with the ip6 hdr,
594 * as the contents are read. We need to m_adj() the
596 * The mbuf will however be thrown away so we can adjust it.
597 * Remember we did an m_pullup on it already so we
598 * can make some assumptions about contiguousness.
601 m_adj(m, args->L3offset);
603 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
614 * sends a reject message, consuming the mbuf passed as an argument.
617 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
621 /* XXX When ip is not guaranteed to be at mtod() we will
622 * need to account for this */
623 * The mbuf will however be thrown away so we can adjust it.
624 * Remember we did an m_pullup on it already so we
625 * can make some assumptions about contiguousness.
628 m_adj(m, args->L3offset);
630 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
631 /* We need the IP header in host order for icmp_error(). */
633 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
634 } else if (args->f_id.proto == IPPROTO_TCP) {
635 struct tcphdr *const tcp =
636 L3HDR(struct tcphdr, mtod(args->m, struct ip *));
637 if ( (tcp->th_flags & TH_RST) == 0) {
639 m = ipfw_send_pkt(args->m, &(args->f_id),
640 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
641 tcp->th_flags | TH_RST);
643 ip_output(m, NULL, NULL, 0, NULL, NULL);
652 * Support for uid/gid/jail lookup. These tests are expensive
653 * (because we may need to look into the list of active sockets)
654 * so we cache the results. ugid_lookupp is 0 if we have not
655 * yet done a lookup, 1 if we succeeded, and -1 if we tried
656 * and failed. The function always returns the match value.
657 * We could actually spare the variable and use *uc, setting
658 * it to '(void *)check_uidgid if we have no info, NULL if
659 * we tried and failed, or any other value if successful.
662 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
667 return cred_check(insn, proto, oif,
668 dst_ip, dst_port, src_ip, src_port,
669 (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
671 struct in_addr src_ip, dst_ip;
672 struct inpcbinfo *pi;
673 struct ipfw_flow_id *id;
674 struct inpcb *pcb, *inp;
684 * Check to see if the UDP or TCP stack supplied us with
685 * the PCB. If so, rather then holding a lock and looking
686 * up the PCB, we can use the one that was supplied.
688 if (inp && *ugid_lookupp == 0) {
689 INP_LOCK_ASSERT(inp);
690 if (inp->inp_socket != NULL) {
691 *uc = crhold(inp->inp_cred);
697 * If we have already been here and the packet has no
698 * PCB entry associated with it, then we can safely
699 * assume that this is a no match.
701 if (*ugid_lookupp == -1)
703 if (id->proto == IPPROTO_TCP) {
706 } else if (id->proto == IPPROTO_UDP) {
707 lookupflags = INPLOOKUP_WILDCARD;
711 lookupflags |= INPLOOKUP_RLOCKPCB;
713 if (*ugid_lookupp == 0) {
714 if (id->addr_type == 6) {
717 pcb = in6_pcblookup_mbuf(pi,
718 &id->src_ip6, htons(id->src_port),
719 &id->dst_ip6, htons(id->dst_port),
720 lookupflags, oif, args->m);
722 pcb = in6_pcblookup_mbuf(pi,
723 &id->dst_ip6, htons(id->dst_port),
724 &id->src_ip6, htons(id->src_port),
725 lookupflags, oif, args->m);
731 src_ip.s_addr = htonl(id->src_ip);
732 dst_ip.s_addr = htonl(id->dst_ip);
734 pcb = in_pcblookup_mbuf(pi,
735 src_ip, htons(id->src_port),
736 dst_ip, htons(id->dst_port),
737 lookupflags, oif, args->m);
739 pcb = in_pcblookup_mbuf(pi,
740 dst_ip, htons(id->dst_port),
741 src_ip, htons(id->src_port),
742 lookupflags, oif, args->m);
745 INP_RLOCK_ASSERT(pcb);
746 *uc = crhold(pcb->inp_cred);
750 if (*ugid_lookupp == 0) {
752 * We tried and failed, set the variable to -1
753 * so we will not try again on this packet.
759 if (insn->o.opcode == O_UID)
760 match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
761 else if (insn->o.opcode == O_GID)
762 match = groupmember((gid_t)insn->d[0], *uc);
763 else if (insn->o.opcode == O_JAIL)
764 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
766 #endif /* __FreeBSD__ */
770 * Helper function to set args with info on the rule after the matching
771 * one. slot is precise, whereas we guess rule_id as they are
772 * assigned sequentially.
775 set_match(struct ip_fw_args *args, int slot,
776 struct ip_fw_chain *chain)
778 args->rule.chain_id = chain->id;
779 args->rule.slot = slot + 1; /* we use 0 as a marker */
780 args->rule.rule_id = 1 + chain->map[slot]->id;
781 args->rule.rulenum = chain->map[slot]->rulenum;
785 * Helper function to enable cached rule lookups using
786 * x_next and next_rule fields in ipfw rule.
789 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
790 int tablearg, int jump_backwards)
794 /* If possible use cached f_pos (in f->next_rule),
795 * whose version is written in f->next_rule
796 * (horrible hacks to avoid changing the ABI).
798 if (num != IP_FW_TABLEARG && (uintptr_t)f->x_next == chain->id)
799 f_pos = (uintptr_t)f->next_rule;
801 int i = IP_FW_ARG_TABLEARG(num);
802 /* make sure we do not jump backward */
803 if (jump_backwards == 0 && i <= f->rulenum)
805 f_pos = ipfw_find_rule(chain, i, 0);
806 /* update the cache */
807 if (num != IP_FW_TABLEARG) {
808 f->next_rule = (void *)(uintptr_t)f_pos;
809 f->x_next = (void *)(uintptr_t)chain->id;
817 * The main check routine for the firewall.
819 * All arguments are in args so we can modify them and return them
820 * back to the caller.
824 * args->m (in/out) The packet; we set to NULL when/if we nuke it.
825 * Starts with the IP header.
826 * args->eh (in) Mac header if present, NULL for layer3 packet.
827 * args->L3offset Number of bytes bypassed if we came from L2.
828 * e.g. often sizeof(eh) ** NOTYET **
829 * args->oif Outgoing interface, NULL if packet is incoming.
830 * The incoming interface is in the mbuf. (in)
831 * args->divert_rule (in/out)
832 * Skip up to the first rule past this rule number;
833 * upon return, non-zero port number for divert or tee.
835 * args->rule Pointer to the last matching rule (in/out)
836 * args->next_hop Socket we are forwarding to (out).
837 * args->next_hop6 IPv6 next hop we are forwarding to (out).
838 * args->f_id Addresses grabbed from the packet (out)
839 * args->rule.info a cookie depending on rule action
843 * IP_FW_PASS the packet must be accepted
844 * IP_FW_DENY the packet must be dropped
845 * IP_FW_DIVERT divert packet, port in m_tag
846 * IP_FW_TEE tee packet, port in m_tag
847 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie
848 * IP_FW_NETGRAPH into netgraph, cookie args->cookie
849 * args->rule contains the matching rule,
850 * args->rule.info has additional information.
854 ipfw_chk(struct ip_fw_args *args)
858 * Local variables holding state while processing a packet:
860 * IMPORTANT NOTE: to speed up the processing of rules, there
861 * are some assumption on the values of the variables, which
862 * are documented here. Should you change them, please check
863 * the implementation of the various instructions to make sure
864 * that they still work.
866 * args->eh The MAC header. It is non-null for a layer2
867 * packet, it is NULL for a layer-3 packet.
869 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
871 * m | args->m Pointer to the mbuf, as received from the caller.
872 * It may change if ipfw_chk() does an m_pullup, or if it
873 * consumes the packet because it calls send_reject().
874 * XXX This has to change, so that ipfw_chk() never modifies
875 * or consumes the buffer.
876 * ip is the beginning of the ip(4 or 6) header.
877 * Calculated by adding the L3offset to the start of data.
878 * (Until we start using L3offset, the packet is
879 * supposed to start with the ip header).
881 struct mbuf *m = args->m;
882 struct ip *ip = mtod(m, struct ip *);
885 * For rules which contain uid/gid or jail constraints, cache
886 * a copy of the users credentials after the pcb lookup has been
887 * executed. This will speed up the processing of rules with
888 * these types of constraints, as well as decrease contention
889 * on pcb related locks.
892 struct bsd_ucred ucred_cache;
894 struct ucred *ucred_cache = NULL;
896 int ucred_lookup = 0;
899 * oif | args->oif If NULL, ipfw_chk has been called on the
900 * inbound path (ether_input, ip_input).
901 * If non-NULL, ipfw_chk has been called on the outbound path
902 * (ether_output, ip_output).
904 struct ifnet *oif = args->oif;
906 int f_pos = 0; /* index of current rule in the array */
910 * hlen The length of the IP header.
912 u_int hlen = 0; /* hlen >0 means we have an IP pkt */
915 * offset The offset of a fragment. offset != 0 means that
916 * we have a fragment at this offset of an IPv4 packet.
917 * offset == 0 means that (if this is an IPv4 packet)
918 * this is the first or only fragment.
919 * For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
920 * or there is a single packet fragement (fragement header added
921 * without needed). We will treat a single packet fragment as if
922 * there was no fragment header (or log/block depending on the
923 * V_fw_permit_single_frag6 sysctl setting).
929 * Local copies of addresses. They are only valid if we have
932 * proto The protocol. Set to 0 for non-ip packets,
933 * or to the protocol read from the packet otherwise.
934 * proto != 0 means that we have an IPv4 packet.
936 * src_port, dst_port port numbers, in HOST format. Only
937 * valid for TCP and UDP packets.
939 * src_ip, dst_ip ip addresses, in NETWORK format.
940 * Only valid for IPv4 packets.
943 uint16_t src_port = 0, dst_port = 0; /* NOTE: host format */
944 struct in_addr src_ip, dst_ip; /* NOTE: network format */
947 uint16_t etype = 0; /* Host order stored ether type */
950 * dyn_dir = MATCH_UNKNOWN when rules unchecked,
951 * MATCH_NONE when checked and not matched (q = NULL),
952 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
954 int dyn_dir = MATCH_UNKNOWN;
955 ipfw_dyn_rule *q = NULL;
956 struct ip_fw_chain *chain = &V_layer3_chain;
959 * We store in ulp a pointer to the upper layer protocol header.
960 * In the ipv4 case this is easy to determine from the header,
961 * but for ipv6 we might have some additional headers in the middle.
962 * ulp is NULL if not found.
964 void *ulp = NULL; /* upper layer protocol pointer. */
966 /* XXX ipv6 variables */
968 uint8_t icmp6_type = 0;
969 uint16_t ext_hd = 0; /* bits vector for extension header filtering */
970 /* end of ipv6 variables */
974 int done = 0; /* flag to exit the outer loop */
976 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
977 return (IP_FW_PASS); /* accept */
979 dst_ip.s_addr = 0; /* make sure it is initialized */
980 src_ip.s_addr = 0; /* make sure it is initialized */
981 pktlen = m->m_pkthdr.len;
982 args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
983 proto = args->f_id.proto = 0; /* mark f_id invalid */
984 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
987 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
988 * then it sets p to point at the offset "len" in the mbuf. WARNING: the
989 * pointer might become stale after other pullups (but we never use it
992 #define PULLUP_TO(_len, p, T) PULLUP_LEN(_len, p, sizeof(T))
993 #define PULLUP_LEN(_len, p, T) \
995 int x = (_len) + T; \
996 if ((m)->m_len < x) { \
997 args->m = m = m_pullup(m, x); \
999 goto pullup_failed; \
1001 p = (mtod(m, char *) + (_len)); \
1005 * if we have an ether header,
1008 etype = ntohs(args->eh->ether_type);
1010 /* Identify IP packets and fill up variables. */
1011 if (pktlen >= sizeof(struct ip6_hdr) &&
1012 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
1013 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1015 args->f_id.addr_type = 6;
1016 hlen = sizeof(struct ip6_hdr);
1017 proto = ip6->ip6_nxt;
1019 /* Search extension headers to find upper layer protocols */
1020 while (ulp == NULL && offset == 0) {
1022 case IPPROTO_ICMPV6:
1023 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1024 icmp6_type = ICMP6(ulp)->icmp6_type;
1028 PULLUP_TO(hlen, ulp, struct tcphdr);
1029 dst_port = TCP(ulp)->th_dport;
1030 src_port = TCP(ulp)->th_sport;
1031 /* save flags for dynamic rules */
1032 args->f_id._flags = TCP(ulp)->th_flags;
1036 PULLUP_TO(hlen, ulp, struct sctphdr);
1037 src_port = SCTP(ulp)->src_port;
1038 dst_port = SCTP(ulp)->dest_port;
1042 PULLUP_TO(hlen, ulp, struct udphdr);
1043 dst_port = UDP(ulp)->uh_dport;
1044 src_port = UDP(ulp)->uh_sport;
1047 case IPPROTO_HOPOPTS: /* RFC 2460 */
1048 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1049 ext_hd |= EXT_HOPOPTS;
1050 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1051 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1055 case IPPROTO_ROUTING: /* RFC 2460 */
1056 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1057 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1059 ext_hd |= EXT_RTHDR0;
1062 ext_hd |= EXT_RTHDR2;
1066 printf("IPFW2: IPV6 - Unknown "
1067 "Routing Header type(%d)\n",
1068 ((struct ip6_rthdr *)
1070 if (V_fw_deny_unknown_exthdrs)
1071 return (IP_FW_DENY);
1074 ext_hd |= EXT_ROUTING;
1075 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1076 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1080 case IPPROTO_FRAGMENT: /* RFC 2460 */
1081 PULLUP_TO(hlen, ulp, struct ip6_frag);
1082 ext_hd |= EXT_FRAGMENT;
1083 hlen += sizeof (struct ip6_frag);
1084 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1085 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1087 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1089 if (V_fw_permit_single_frag6 == 0 &&
1090 offset == 0 && ip6f_mf == 0) {
1092 printf("IPFW2: IPV6 - Invalid "
1093 "Fragment Header\n");
1094 if (V_fw_deny_unknown_exthdrs)
1095 return (IP_FW_DENY);
1099 ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1103 case IPPROTO_DSTOPTS: /* RFC 2460 */
1104 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1105 ext_hd |= EXT_DSTOPTS;
1106 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1107 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1111 case IPPROTO_AH: /* RFC 2402 */
1112 PULLUP_TO(hlen, ulp, struct ip6_ext);
1114 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1115 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1119 case IPPROTO_ESP: /* RFC 2406 */
1120 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
1121 /* Anything past Seq# is variable length and
1122 * data past this ext. header is encrypted. */
1126 case IPPROTO_NONE: /* RFC 2460 */
1128 * Packet ends here, and IPv6 header has
1129 * already been pulled up. If ip6e_len!=0
1130 * then octets must be ignored.
1132 ulp = ip; /* non-NULL to get out of loop. */
1135 case IPPROTO_OSPFIGP:
1136 /* XXX OSPF header check? */
1137 PULLUP_TO(hlen, ulp, struct ip6_ext);
1141 /* XXX PIM header check? */
1142 PULLUP_TO(hlen, ulp, struct pim);
1146 PULLUP_TO(hlen, ulp, struct carp_header);
1147 if (((struct carp_header *)ulp)->carp_version !=
1149 return (IP_FW_DENY);
1150 if (((struct carp_header *)ulp)->carp_type !=
1152 return (IP_FW_DENY);
1155 case IPPROTO_IPV6: /* RFC 2893 */
1156 PULLUP_TO(hlen, ulp, struct ip6_hdr);
1159 case IPPROTO_IPV4: /* RFC 2893 */
1160 PULLUP_TO(hlen, ulp, struct ip);
1165 printf("IPFW2: IPV6 - Unknown "
1166 "Extension Header(%d), ext_hd=%x\n",
1168 if (V_fw_deny_unknown_exthdrs)
1169 return (IP_FW_DENY);
1170 PULLUP_TO(hlen, ulp, struct ip6_ext);
1174 ip = mtod(m, struct ip *);
1175 ip6 = (struct ip6_hdr *)ip;
1176 args->f_id.src_ip6 = ip6->ip6_src;
1177 args->f_id.dst_ip6 = ip6->ip6_dst;
1178 args->f_id.src_ip = 0;
1179 args->f_id.dst_ip = 0;
1180 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1181 } else if (pktlen >= sizeof(struct ip) &&
1182 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1184 hlen = ip->ip_hl << 2;
1185 args->f_id.addr_type = 4;
1188 * Collect parameters into local variables for faster matching.
1191 src_ip = ip->ip_src;
1192 dst_ip = ip->ip_dst;
1193 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1194 iplen = ntohs(ip->ip_len);
1195 pktlen = iplen < pktlen ? iplen : pktlen;
1200 PULLUP_TO(hlen, ulp, struct tcphdr);
1201 dst_port = TCP(ulp)->th_dport;
1202 src_port = TCP(ulp)->th_sport;
1203 /* save flags for dynamic rules */
1204 args->f_id._flags = TCP(ulp)->th_flags;
1208 PULLUP_TO(hlen, ulp, struct sctphdr);
1209 src_port = SCTP(ulp)->src_port;
1210 dst_port = SCTP(ulp)->dest_port;
1214 PULLUP_TO(hlen, ulp, struct udphdr);
1215 dst_port = UDP(ulp)->uh_dport;
1216 src_port = UDP(ulp)->uh_sport;
1220 PULLUP_TO(hlen, ulp, struct icmphdr);
1221 //args->f_id.flags = ICMP(ulp)->icmp_type;
1229 ip = mtod(m, struct ip *);
1230 args->f_id.src_ip = ntohl(src_ip.s_addr);
1231 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1234 if (proto) { /* we may have port numbers, store them */
1235 args->f_id.proto = proto;
1236 args->f_id.src_port = src_port = ntohs(src_port);
1237 args->f_id.dst_port = dst_port = ntohs(dst_port);
1241 if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1242 IPFW_RUNLOCK(chain);
1243 return (IP_FW_PASS); /* accept */
1245 if (args->rule.slot) {
1247 * Packet has already been tagged as a result of a previous
1248 * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1249 * REASS, NETGRAPH, DIVERT/TEE...)
1250 * Validate the slot and continue from the next one
1251 * if still present, otherwise do a lookup.
1253 f_pos = (args->rule.chain_id == chain->id) ?
1255 ipfw_find_rule(chain, args->rule.rulenum,
1256 args->rule.rule_id);
1262 * Now scan the rules, and parse microinstructions for each rule.
1263 * We have two nested loops and an inner switch. Sometimes we
1264 * need to break out of one or both loops, or re-enter one of
1265 * the loops with updated variables. Loop variables are:
1267 * f_pos (outer loop) points to the current rule.
1268 * On output it points to the matching rule.
1269 * done (outer loop) is used as a flag to break the loop.
1270 * l (inner loop) residual length of current rule.
1271 * cmd points to the current microinstruction.
1273 * We break the inner loop by setting l=0 and possibly
1274 * cmdlen=0 if we don't want to advance cmd.
1275 * We break the outer loop by setting done=1
1276 * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1279 for (; f_pos < chain->n_rules; f_pos++) {
1281 uint32_t tablearg = 0;
1282 int l, cmdlen, skip_or; /* skip rest of OR block */
1285 f = chain->map[f_pos];
1286 if (V_set_disable & (1 << f->set) )
1290 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1291 l -= cmdlen, cmd += cmdlen) {
1295 * check_body is a jump target used when we find a
1296 * CHECK_STATE, and need to jump to the body of
1301 cmdlen = F_LEN(cmd);
1303 * An OR block (insn_1 || .. || insn_n) has the
1304 * F_OR bit set in all but the last instruction.
1305 * The first match will set "skip_or", and cause
1306 * the following instructions to be skipped until
1307 * past the one with the F_OR bit clear.
1309 if (skip_or) { /* skip this instruction */
1310 if ((cmd->len & F_OR) == 0)
1311 skip_or = 0; /* next one is good */
1314 match = 0; /* set to 1 if we succeed */
1316 switch (cmd->opcode) {
1318 * The first set of opcodes compares the packet's
1319 * fields with some pattern, setting 'match' if a
1320 * match is found. At the end of the loop there is
1321 * logic to deal with F_NOT and F_OR flags associated
1329 printf("ipfw: opcode %d unimplemented\n",
1337 * We only check offset == 0 && proto != 0,
1338 * as this ensures that we have a
1339 * packet with the ports info.
1343 if (proto == IPPROTO_TCP ||
1344 proto == IPPROTO_UDP)
1345 match = check_uidgid(
1346 (ipfw_insn_u32 *)cmd,
1347 args, &ucred_lookup,
1351 (void *)&ucred_cache);
1356 match = iface_match(m->m_pkthdr.rcvif,
1357 (ipfw_insn_if *)cmd, chain, &tablearg);
1361 match = iface_match(oif, (ipfw_insn_if *)cmd,
1366 match = iface_match(oif ? oif :
1367 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1372 if (args->eh != NULL) { /* have MAC header */
1373 u_int32_t *want = (u_int32_t *)
1374 ((ipfw_insn_mac *)cmd)->addr;
1375 u_int32_t *mask = (u_int32_t *)
1376 ((ipfw_insn_mac *)cmd)->mask;
1377 u_int32_t *hdr = (u_int32_t *)args->eh;
1380 ( want[0] == (hdr[0] & mask[0]) &&
1381 want[1] == (hdr[1] & mask[1]) &&
1382 want[2] == (hdr[2] & mask[2]) );
1387 if (args->eh != NULL) {
1389 ((ipfw_insn_u16 *)cmd)->ports;
1392 for (i = cmdlen - 1; !match && i>0;
1394 match = (etype >= p[0] &&
1400 match = (offset != 0);
1403 case O_IN: /* "out" is "not in" */
1404 match = (oif == NULL);
1408 match = (args->eh != NULL);
1413 /* For diverted packets, args->rule.info
1414 * contains the divert port (in host format)
1415 * reason and direction.
1417 uint32_t i = args->rule.info;
1418 match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1419 cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1425 * We do not allow an arg of 0 so the
1426 * check of "proto" only suffices.
1428 match = (proto == cmd->arg1);
1433 (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1437 case O_IP_SRC_LOOKUP:
1438 case O_IP_DST_LOOKUP:
1441 (cmd->opcode == O_IP_DST_LOOKUP) ?
1442 dst_ip.s_addr : src_ip.s_addr;
1445 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1446 /* generic lookup. The key must be
1447 * in 32bit big-endian format.
1449 v = ((ipfw_insn_u32 *)cmd)->d[1];
1451 key = dst_ip.s_addr;
1453 key = src_ip.s_addr;
1454 else if (v == 6) /* dscp */
1455 key = (ip->ip_tos >> 2) & 0x3f;
1456 else if (offset != 0)
1458 else if (proto != IPPROTO_TCP &&
1459 proto != IPPROTO_UDP)
1462 key = htonl(dst_port);
1464 key = htonl(src_port);
1465 else if (v == 4 || v == 5) {
1467 (ipfw_insn_u32 *)cmd,
1468 args, &ucred_lookup,
1471 if (v == 4 /* O_UID */)
1472 key = ucred_cache->cr_uid;
1473 else if (v == 5 /* O_JAIL */)
1474 key = ucred_cache->cr_prison->pr_id;
1475 #else /* !__FreeBSD__ */
1476 (void *)&ucred_cache);
1477 if (v ==4 /* O_UID */)
1478 key = ucred_cache.uid;
1479 else if (v == 5 /* O_JAIL */)
1480 key = ucred_cache.xid;
1481 #endif /* !__FreeBSD__ */
1486 match = ipfw_lookup_table(chain,
1487 cmd->arg1, key, &v);
1490 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1492 ((ipfw_insn_u32 *)cmd)->d[0] == v;
1495 } else if (is_ipv6) {
1497 void *pkey = (cmd->opcode == O_IP_DST_LOOKUP) ?
1498 &args->f_id.dst_ip6: &args->f_id.src_ip6;
1499 match = ipfw_lookup_table_extended(chain,
1500 cmd->arg1, pkey, &v,
1502 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1503 match = ((ipfw_insn_u32 *)cmd)->d[0] == v;
1513 (cmd->opcode == O_IP_DST_MASK) ?
1514 dst_ip.s_addr : src_ip.s_addr;
1515 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
1518 for (; !match && i>0; i-= 2, p+= 2)
1519 match = (p[0] == (a & p[1]));
1527 INADDR_TO_IFP(src_ip, tif);
1528 match = (tif != NULL);
1534 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6);
1541 u_int32_t *d = (u_int32_t *)(cmd+1);
1543 cmd->opcode == O_IP_DST_SET ?
1549 addr -= d[0]; /* subtract base */
1550 match = (addr < cmd->arg1) &&
1551 ( d[ 1 + (addr>>5)] &
1552 (1<<(addr & 0x1f)) );
1558 (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1566 INADDR_TO_IFP(dst_ip, tif);
1567 match = (tif != NULL);
1573 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6);
1581 * offset == 0 && proto != 0 is enough
1582 * to guarantee that we have a
1583 * packet with port info.
1585 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
1588 (cmd->opcode == O_IP_SRCPORT) ?
1589 src_port : dst_port ;
1591 ((ipfw_insn_u16 *)cmd)->ports;
1594 for (i = cmdlen - 1; !match && i>0;
1596 match = (x>=p[0] && x<=p[1]);
1601 match = (offset == 0 && proto==IPPROTO_ICMP &&
1602 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
1607 match = is_ipv6 && offset == 0 &&
1608 proto==IPPROTO_ICMPV6 &&
1610 ICMP6(ulp)->icmp6_type,
1611 (ipfw_insn_u32 *)cmd);
1617 ipopts_match(ip, cmd) );
1622 cmd->arg1 == ip->ip_v);
1628 if (is_ipv4) { /* only for IP packets */
1633 if (cmd->opcode == O_IPLEN)
1635 else if (cmd->opcode == O_IPTTL)
1637 else /* must be IPID */
1638 x = ntohs(ip->ip_id);
1640 match = (cmd->arg1 == x);
1643 /* otherwise we have ranges */
1644 p = ((ipfw_insn_u16 *)cmd)->ports;
1646 for (; !match && i>0; i--, p += 2)
1647 match = (x >= p[0] && x <= p[1]);
1651 case O_IPPRECEDENCE:
1653 (cmd->arg1 == (ip->ip_tos & 0xe0)) );
1658 flags_match(cmd, ip->ip_tos));
1666 p = ((ipfw_insn_u32 *)cmd)->d;
1669 x = ip->ip_tos >> 2;
1672 v = &((struct ip6_hdr *)ip)->ip6_vfc;
1673 x = (*v & 0x0F) << 2;
1679 /* DSCP bitmask is stored as low_u32 high_u32 */
1681 match = *(p + 1) & (1 << (x - 32));
1683 match = *p & (1 << x);
1688 if (proto == IPPROTO_TCP && offset == 0) {
1696 ((ip->ip_hl + tcp->th_off) << 2);
1698 match = (cmd->arg1 == x);
1701 /* otherwise we have ranges */
1702 p = ((ipfw_insn_u16 *)cmd)->ports;
1704 for (; !match && i>0; i--, p += 2)
1705 match = (x >= p[0] && x <= p[1]);
1710 match = (proto == IPPROTO_TCP && offset == 0 &&
1711 flags_match(cmd, TCP(ulp)->th_flags));
1715 PULLUP_LEN(hlen, ulp, (TCP(ulp)->th_off << 2));
1716 match = (proto == IPPROTO_TCP && offset == 0 &&
1717 tcpopts_match(TCP(ulp), cmd));
1721 match = (proto == IPPROTO_TCP && offset == 0 &&
1722 ((ipfw_insn_u32 *)cmd)->d[0] ==
1727 match = (proto == IPPROTO_TCP && offset == 0 &&
1728 ((ipfw_insn_u32 *)cmd)->d[0] ==
1733 if (proto == IPPROTO_TCP && offset == 0) {
1738 x = ntohs(TCP(ulp)->th_win);
1740 match = (cmd->arg1 == x);
1743 /* Otherwise we have ranges. */
1744 p = ((ipfw_insn_u16 *)cmd)->ports;
1746 for (; !match && i > 0; i--, p += 2)
1747 match = (x >= p[0] && x <= p[1]);
1752 /* reject packets which have SYN only */
1753 /* XXX should i also check for TH_ACK ? */
1754 match = (proto == IPPROTO_TCP && offset == 0 &&
1755 (TCP(ulp)->th_flags &
1756 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
1761 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
1764 at = pf_find_mtag(m);
1765 if (at != NULL && at->qid != 0)
1767 at = pf_get_mtag(m);
1770 * Let the packet fall back to the
1775 at->qid = altq->qid;
1781 ipfw_log(f, hlen, args, m,
1782 oif, offset | ip6f_mf, tablearg, ip);
1787 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
1791 /* Outgoing packets automatically pass/match */
1792 match = ((oif != NULL) ||
1793 (m->m_pkthdr.rcvif == NULL) ||
1797 verify_path6(&(args->f_id.src_ip6),
1798 m->m_pkthdr.rcvif, args->f_id.fib) :
1800 verify_path(src_ip, m->m_pkthdr.rcvif,
1805 /* Outgoing packets automatically pass/match */
1806 match = (hlen > 0 && ((oif != NULL) ||
1809 verify_path6(&(args->f_id.src_ip6),
1810 NULL, args->f_id.fib) :
1812 verify_path(src_ip, NULL, args->f_id.fib)));
1816 /* Outgoing packets automatically pass/match */
1817 if (oif == NULL && hlen > 0 &&
1818 ( (is_ipv4 && in_localaddr(src_ip))
1821 in6_localaddr(&(args->f_id.src_ip6)))
1826 is_ipv6 ? verify_path6(
1827 &(args->f_id.src_ip6),
1840 match = (m_tag_find(m,
1841 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
1843 /* otherwise no match */
1849 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
1850 &((ipfw_insn_ip6 *)cmd)->addr6);
1855 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
1856 &((ipfw_insn_ip6 *)cmd)->addr6);
1858 case O_IP6_SRC_MASK:
1859 case O_IP6_DST_MASK:
1863 struct in6_addr *d =
1864 &((ipfw_insn_ip6 *)cmd)->addr6;
1866 for (; !match && i > 0; d += 2,
1867 i -= F_INSN_SIZE(struct in6_addr)
1873 APPLY_MASK(&p, &d[1]);
1875 IN6_ARE_ADDR_EQUAL(&d[0],
1883 flow6id_match(args->f_id.flow_id6,
1884 (ipfw_insn_u32 *) cmd);
1889 (ext_hd & ((ipfw_insn *) cmd)->arg1);
1903 uint32_t tag = IP_FW_ARG_TABLEARG(cmd->arg1);
1905 /* Packet is already tagged with this tag? */
1906 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
1908 /* We have `untag' action when F_NOT flag is
1909 * present. And we must remove this mtag from
1910 * mbuf and reset `match' to zero (`match' will
1911 * be inversed later).
1912 * Otherwise we should allocate new mtag and
1913 * push it into mbuf.
1915 if (cmd->len & F_NOT) { /* `untag' action */
1917 m_tag_delete(m, mtag);
1921 mtag = m_tag_alloc( MTAG_IPFW,
1924 m_tag_prepend(m, mtag);
1931 case O_FIB: /* try match the specified fib */
1932 if (args->f_id.fib == cmd->arg1)
1937 struct inpcb *inp = args->inp;
1938 struct inpcbinfo *pi;
1940 if (is_ipv6) /* XXX can we remove this ? */
1943 if (proto == IPPROTO_TCP)
1945 else if (proto == IPPROTO_UDP)
1951 * XXXRW: so_user_cookie should almost
1952 * certainly be inp_user_cookie?
1955 /* For incomming packet, lookup up the
1956 inpcb using the src/dest ip/port tuple */
1958 inp = in_pcblookup(pi,
1959 src_ip, htons(src_port),
1960 dst_ip, htons(dst_port),
1961 INPLOOKUP_RLOCKPCB, NULL);
1964 inp->inp_socket->so_user_cookie;
1970 if (inp->inp_socket) {
1972 inp->inp_socket->so_user_cookie;
1982 uint32_t tag = IP_FW_ARG_TABLEARG(cmd->arg1);
1985 match = m_tag_locate(m, MTAG_IPFW,
1990 /* we have ranges */
1991 for (mtag = m_tag_first(m);
1992 mtag != NULL && !match;
1993 mtag = m_tag_next(m, mtag)) {
1997 if (mtag->m_tag_cookie != MTAG_IPFW)
2000 p = ((ipfw_insn_u16 *)cmd)->ports;
2002 for(; !match && i > 0; i--, p += 2)
2004 mtag->m_tag_id >= p[0] &&
2005 mtag->m_tag_id <= p[1];
2011 * The second set of opcodes represents 'actions',
2012 * i.e. the terminal part of a rule once the packet
2013 * matches all previous patterns.
2014 * Typically there is only one action for each rule,
2015 * and the opcode is stored at the end of the rule
2016 * (but there are exceptions -- see below).
2018 * In general, here we set retval and terminate the
2019 * outer loop (would be a 'break 3' in some language,
2020 * but we need to set l=0, done=1)
2023 * O_COUNT and O_SKIPTO actions:
2024 * instead of terminating, we jump to the next rule
2025 * (setting l=0), or to the SKIPTO target (setting
2026 * f/f_len, cmd and l as needed), respectively.
2028 * O_TAG, O_LOG and O_ALTQ action parameters:
2029 * perform some action and set match = 1;
2031 * O_LIMIT and O_KEEP_STATE: these opcodes are
2032 * not real 'actions', and are stored right
2033 * before the 'action' part of the rule.
2034 * These opcodes try to install an entry in the
2035 * state tables; if successful, we continue with
2036 * the next opcode (match=1; break;), otherwise
2037 * the packet must be dropped (set retval,
2038 * break loops with l=0, done=1)
2040 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2041 * cause a lookup of the state table, and a jump
2042 * to the 'action' part of the parent rule
2043 * if an entry is found, or
2044 * (CHECK_STATE only) a jump to the next rule if
2045 * the entry is not found.
2046 * The result of the lookup is cached so that
2047 * further instances of these opcodes become NOPs.
2048 * The jump to the next rule is done by setting
2053 if (ipfw_install_state(f,
2054 (ipfw_insn_limit *)cmd, args, tablearg)) {
2055 /* error or limit violation */
2056 retval = IP_FW_DENY;
2057 l = 0; /* exit inner loop */
2058 done = 1; /* exit outer loop */
2066 * dynamic rules are checked at the first
2067 * keep-state or check-state occurrence,
2068 * with the result being stored in dyn_dir.
2069 * The compiler introduces a PROBE_STATE
2070 * instruction for us when we have a
2071 * KEEP_STATE (because PROBE_STATE needs
2074 if (dyn_dir == MATCH_UNKNOWN &&
2075 (q = ipfw_lookup_dyn_rule(&args->f_id,
2076 &dyn_dir, proto == IPPROTO_TCP ?
2080 * Found dynamic entry, update stats
2081 * and jump to the 'action' part of
2082 * the parent rule by setting
2083 * f, cmd, l and clearing cmdlen.
2085 IPFW_INC_DYN_COUNTER(q, pktlen);
2086 /* XXX we would like to have f_pos
2087 * readily accessible in the dynamic
2088 * rule, instead of having to
2092 f_pos = ipfw_find_rule(chain,
2094 cmd = ACTION_PTR(f);
2095 l = f->cmd_len - f->act_ofs;
2102 * Dynamic entry not found. If CHECK_STATE,
2103 * skip to next rule, if PROBE_STATE just
2104 * ignore and continue with next opcode.
2106 if (cmd->opcode == O_CHECK_STATE)
2107 l = 0; /* exit inner loop */
2112 retval = 0; /* accept */
2113 l = 0; /* exit inner loop */
2114 done = 1; /* exit outer loop */
2119 set_match(args, f_pos, chain);
2120 args->rule.info = IP_FW_ARG_TABLEARG(cmd->arg1);
2121 if (cmd->opcode == O_PIPE)
2122 args->rule.info |= IPFW_IS_PIPE;
2124 args->rule.info |= IPFW_ONEPASS;
2125 retval = IP_FW_DUMMYNET;
2126 l = 0; /* exit inner loop */
2127 done = 1; /* exit outer loop */
2132 if (args->eh) /* not on layer 2 */
2134 /* otherwise this is terminal */
2135 l = 0; /* exit inner loop */
2136 done = 1; /* exit outer loop */
2137 retval = (cmd->opcode == O_DIVERT) ?
2138 IP_FW_DIVERT : IP_FW_TEE;
2139 set_match(args, f_pos, chain);
2140 args->rule.info = IP_FW_ARG_TABLEARG(cmd->arg1);
2144 IPFW_INC_RULE_COUNTER(f, pktlen);
2145 l = 0; /* exit inner loop */
2149 IPFW_INC_RULE_COUNTER(f, pktlen);
2150 f_pos = jump_fast(chain, f, cmd->arg1, tablearg, 0);
2152 * Skip disabled rules, and re-enter
2153 * the inner loop with the correct
2154 * f_pos, f, l and cmd.
2155 * Also clear cmdlen and skip_or
2157 for (; f_pos < chain->n_rules - 1 &&
2159 (1 << chain->map[f_pos]->set));
2162 /* Re-enter the inner loop at the skipto rule. */
2163 f = chain->map[f_pos];
2170 break; /* not reached */
2172 case O_CALLRETURN: {
2174 * Implementation of `subroutine' call/return,
2175 * in the stack carried in an mbuf tag. This
2176 * is different from `skipto' in that any call
2177 * address is possible (`skipto' must prevent
2178 * backward jumps to avoid endless loops).
2179 * We have `return' action when F_NOT flag is
2180 * present. The `m_tag_id' field is used as
2184 uint16_t jmpto, *stack;
2186 #define IS_CALL ((cmd->len & F_NOT) == 0)
2187 #define IS_RETURN ((cmd->len & F_NOT) != 0)
2189 * Hand-rolled version of m_tag_locate() with
2191 * If not already tagged, allocate new tag.
2193 mtag = m_tag_first(m);
2194 while (mtag != NULL) {
2195 if (mtag->m_tag_cookie ==
2198 mtag = m_tag_next(m, mtag);
2200 if (mtag == NULL && IS_CALL) {
2201 mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2202 IPFW_CALLSTACK_SIZE *
2203 sizeof(uint16_t), M_NOWAIT);
2205 m_tag_prepend(m, mtag);
2209 * On error both `call' and `return' just
2210 * continue with next rule.
2212 if (IS_RETURN && (mtag == NULL ||
2213 mtag->m_tag_id == 0)) {
2214 l = 0; /* exit inner loop */
2217 if (IS_CALL && (mtag == NULL ||
2218 mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2219 printf("ipfw: call stack error, "
2220 "go to next rule\n");
2221 l = 0; /* exit inner loop */
2225 IPFW_INC_RULE_COUNTER(f, pktlen);
2226 stack = (uint16_t *)(mtag + 1);
2229 * The `call' action may use cached f_pos
2230 * (in f->next_rule), whose version is written
2232 * The `return' action, however, doesn't have
2233 * fixed jump address in cmd->arg1 and can't use
2237 stack[mtag->m_tag_id] = f->rulenum;
2239 f_pos = jump_fast(chain, f, cmd->arg1,
2241 } else { /* `return' action */
2243 jmpto = stack[mtag->m_tag_id] + 1;
2244 f_pos = ipfw_find_rule(chain, jmpto, 0);
2248 * Skip disabled rules, and re-enter
2249 * the inner loop with the correct
2250 * f_pos, f, l and cmd.
2251 * Also clear cmdlen and skip_or
2253 for (; f_pos < chain->n_rules - 1 &&
2255 (1 << chain->map[f_pos]->set)); f_pos++)
2257 /* Re-enter the inner loop at the dest rule. */
2258 f = chain->map[f_pos];
2264 break; /* NOTREACHED */
2271 * Drop the packet and send a reject notice
2272 * if the packet is not ICMP (or is an ICMP
2273 * query), and it is not multicast/broadcast.
2275 if (hlen > 0 && is_ipv4 && offset == 0 &&
2276 (proto != IPPROTO_ICMP ||
2277 is_icmp_query(ICMP(ulp))) &&
2278 !(m->m_flags & (M_BCAST|M_MCAST)) &&
2279 !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2280 send_reject(args, cmd->arg1, iplen, ip);
2286 if (hlen > 0 && is_ipv6 &&
2287 ((offset & IP6F_OFF_MASK) == 0) &&
2288 (proto != IPPROTO_ICMPV6 ||
2289 (is_icmp6_query(icmp6_type) == 1)) &&
2290 !(m->m_flags & (M_BCAST|M_MCAST)) &&
2291 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
2293 args, cmd->arg1, hlen,
2294 (struct ip6_hdr *)ip);
2300 retval = IP_FW_DENY;
2301 l = 0; /* exit inner loop */
2302 done = 1; /* exit outer loop */
2306 if (args->eh) /* not valid on layer2 pkts */
2308 if (q == NULL || q->rule != f ||
2309 dyn_dir == MATCH_FORWARD) {
2310 struct sockaddr_in *sa;
2311 sa = &(((ipfw_insn_sa *)cmd)->sa);
2312 if (sa->sin_addr.s_addr == INADDR_ANY) {
2313 bcopy(sa, &args->hopstore,
2315 args->hopstore.sin_addr.s_addr =
2317 args->next_hop = &args->hopstore;
2319 args->next_hop = sa;
2322 retval = IP_FW_PASS;
2323 l = 0; /* exit inner loop */
2324 done = 1; /* exit outer loop */
2329 if (args->eh) /* not valid on layer2 pkts */
2331 if (q == NULL || q->rule != f ||
2332 dyn_dir == MATCH_FORWARD) {
2333 struct sockaddr_in6 *sin6;
2335 sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2336 args->next_hop6 = sin6;
2338 retval = IP_FW_PASS;
2339 l = 0; /* exit inner loop */
2340 done = 1; /* exit outer loop */
2346 set_match(args, f_pos, chain);
2347 args->rule.info = IP_FW_ARG_TABLEARG(cmd->arg1);
2349 args->rule.info |= IPFW_ONEPASS;
2350 retval = (cmd->opcode == O_NETGRAPH) ?
2351 IP_FW_NETGRAPH : IP_FW_NGTEE;
2352 l = 0; /* exit inner loop */
2353 done = 1; /* exit outer loop */
2359 IPFW_INC_RULE_COUNTER(f, pktlen);
2360 fib = IP_FW_ARG_TABLEARG(cmd->arg1);
2361 if (fib >= rt_numfibs)
2364 args->f_id.fib = fib;
2365 l = 0; /* exit inner loop */
2372 code = IP_FW_ARG_TABLEARG(cmd->arg1) & 0x3F;
2373 l = 0; /* exit inner loop */
2378 ip->ip_tos = (code << 2) | (ip->ip_tos & 0x03);
2379 a += ntohs(ip->ip_sum) - ip->ip_tos;
2380 ip->ip_sum = htons(a);
2381 } else if (is_ipv6) {
2384 v = &((struct ip6_hdr *)ip)->ip6_vfc;
2385 *v = (*v & 0xF0) | (code >> 2);
2387 *v = (*v & 0x3F) | ((code & 0x03) << 6);
2391 IPFW_INC_RULE_COUNTER(f, pktlen);
2396 if (!IPFW_NAT_LOADED) {
2397 retval = IP_FW_DENY;
2402 set_match(args, f_pos, chain);
2403 /* Check if this is 'global' nat rule */
2404 if (cmd->arg1 == 0) {
2405 retval = ipfw_nat_ptr(args, NULL, m);
2410 t = ((ipfw_insn_nat *)cmd)->nat;
2412 nat_id = IP_FW_ARG_TABLEARG(cmd->arg1);
2413 t = (*lookup_nat_ptr)(&chain->nat, nat_id);
2416 retval = IP_FW_DENY;
2417 l = 0; /* exit inner loop */
2418 done = 1; /* exit outer loop */
2421 if (cmd->arg1 != IP_FW_TABLEARG)
2422 ((ipfw_insn_nat *)cmd)->nat = t;
2424 retval = ipfw_nat_ptr(args, t, m);
2426 l = 0; /* exit inner loop */
2427 done = 1; /* exit outer loop */
2433 IPFW_INC_RULE_COUNTER(f, pktlen);
2434 l = 0; /* in any case exit inner loop */
2435 ip_off = ntohs(ip->ip_off);
2437 /* if not fragmented, go to next rule */
2438 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
2441 * ip_reass() expects len & off in host
2446 args->m = m = ip_reass(m);
2449 * do IP header checksum fixup.
2451 if (m == NULL) { /* fragment got swallowed */
2452 retval = IP_FW_DENY;
2453 } else { /* good, packet complete */
2456 ip = mtod(m, struct ip *);
2457 hlen = ip->ip_hl << 2;
2460 if (hlen == sizeof(struct ip))
2461 ip->ip_sum = in_cksum_hdr(ip);
2463 ip->ip_sum = in_cksum(m, hlen);
2464 retval = IP_FW_REASS;
2465 set_match(args, f_pos, chain);
2467 done = 1; /* exit outer loop */
2472 panic("-- unknown opcode %d\n", cmd->opcode);
2473 } /* end of switch() on opcodes */
2475 * if we get here with l=0, then match is irrelevant.
2478 if (cmd->len & F_NOT)
2482 if (cmd->len & F_OR)
2485 if (!(cmd->len & F_OR)) /* not an OR block, */
2486 break; /* try next rule */
2489 } /* end of inner loop, scan opcodes */
2495 /* next_rule:; */ /* try next rule */
2497 } /* end of outer for, scan rules */
2500 struct ip_fw *rule = chain->map[f_pos];
2501 /* Update statistics */
2502 IPFW_INC_RULE_COUNTER(rule, pktlen);
2504 retval = IP_FW_DENY;
2505 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
2507 IPFW_RUNLOCK(chain);
2509 if (ucred_cache != NULL)
2510 crfree(ucred_cache);
2516 printf("ipfw: pullup failed\n");
2517 return (IP_FW_DENY);
2521 * Set maximum number of tables that can be used in given VNET ipfw instance.
2525 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
2528 unsigned int ntables;
2530 ntables = V_fw_tables_max;
2532 error = sysctl_handle_int(oidp, &ntables, 0, req);
2533 /* Read operation or some error */
2534 if ((error != 0) || (req->newptr == NULL))
2537 return (ipfw_resize_tables(&V_layer3_chain, ntables));
2541 * Module and VNET glue
2545 * Stuff that must be initialised only on boot or module load
2553 * Only print out this stuff the first time around,
2554 * when called from the sysinit code.
2560 "initialized, divert %s, nat %s, "
2561 "default to %s, logging ",
2567 #ifdef IPFIREWALL_NAT
2572 default_to_accept ? "accept" : "deny");
2575 * Note: V_xxx variables can be accessed here but the vnet specific
2576 * initializer may not have been called yet for the VIMAGE case.
2577 * Tuneables will have been processed. We will print out values for
2579 * XXX This should all be rationalized AFTER 8.0
2581 if (V_fw_verbose == 0)
2582 printf("disabled\n");
2583 else if (V_verbose_limit == 0)
2584 printf("unlimited\n");
2586 printf("limited to %d packets/entry by default\n",
2589 /* Check user-supplied table count for validness */
2590 if (default_fw_tables > IPFW_TABLES_MAX)
2591 default_fw_tables = IPFW_TABLES_MAX;
2593 ipfw_log_bpf(1); /* init */
2598 * Called for the removal of the last instance only on module unload.
2604 ipfw_log_bpf(0); /* uninit */
2605 printf("IP firewall unloaded\n");
2609 * Stuff that must be initialized for every instance
2610 * (including the first of course).
2613 vnet_ipfw_init(const void *unused)
2616 struct ip_fw *rule = NULL;
2617 struct ip_fw_chain *chain;
2619 chain = &V_layer3_chain;
2621 /* First set up some values that are compile time options */
2622 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */
2623 V_fw_deny_unknown_exthdrs = 1;
2624 #ifdef IPFIREWALL_VERBOSE
2627 #ifdef IPFIREWALL_VERBOSE_LIMIT
2628 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
2630 #ifdef IPFIREWALL_NAT
2631 LIST_INIT(&chain->nat);
2634 /* insert the default rule and create the initial map */
2636 chain->static_len = sizeof(struct ip_fw);
2637 chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_WAITOK | M_ZERO);
2639 rule = malloc(chain->static_len, M_IPFW, M_WAITOK | M_ZERO);
2641 /* Set initial number of tables */
2642 V_fw_tables_max = default_fw_tables;
2643 error = ipfw_init_tables(chain);
2645 printf("ipfw2: setting up tables failed\n");
2646 free(chain->map, M_IPFW);
2651 /* fill and insert the default rule */
2653 rule->rulenum = IPFW_DEFAULT_RULE;
2655 rule->set = RESVD_SET;
2656 rule->cmd[0].len = 1;
2657 rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
2658 chain->rules = chain->default_rule = chain->map[0] = rule;
2659 chain->id = rule->id = 1;
2661 IPFW_LOCK_INIT(chain);
2662 ipfw_dyn_init(chain);
2664 /* First set up some values that are compile time options */
2665 V_ipfw_vnet_ready = 1; /* Open for business */
2668 * Hook the sockopt handler, and the layer2 (V_ip_fw_chk_ptr)
2669 * and pfil hooks for ipv4 and ipv6. Even if the latter two fail
2670 * we still keep the module alive because the sockopt and
2671 * layer2 paths are still useful.
2672 * ipfw[6]_hook return 0 on success, ENOENT on failure,
2673 * so we can ignore the exact return value and just set a flag.
2675 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
2676 * changes in the underlying (per-vnet) variables trigger
2677 * immediate hook()/unhook() calls.
2678 * In layer2 we have the same behaviour, except that V_ether_ipfw
2679 * is checked on each packet because there are no pfil hooks.
2681 V_ip_fw_ctl_ptr = ipfw_ctl;
2682 V_ip_fw_chk_ptr = ipfw_chk;
2683 error = ipfw_attach_hooks(1);
2688 * Called for the removal of each instance.
2691 vnet_ipfw_uninit(const void *unused)
2693 struct ip_fw *reap, *rule;
2694 struct ip_fw_chain *chain = &V_layer3_chain;
2697 V_ipfw_vnet_ready = 0; /* tell new callers to go away */
2699 * disconnect from ipv4, ipv6, layer2 and sockopt.
2700 * Then grab, release and grab again the WLOCK so we make
2701 * sure the update is propagated and nobody will be in.
2703 (void)ipfw_attach_hooks(0 /* detach */);
2704 V_ip_fw_chk_ptr = NULL;
2705 V_ip_fw_ctl_ptr = NULL;
2706 IPFW_UH_WLOCK(chain);
2707 IPFW_UH_WUNLOCK(chain);
2708 IPFW_UH_WLOCK(chain);
2711 ipfw_dyn_uninit(0); /* run the callout_drain */
2712 IPFW_WUNLOCK(chain);
2714 ipfw_destroy_tables(chain);
2717 for (i = 0; i < chain->n_rules; i++) {
2718 rule = chain->map[i];
2719 rule->x_next = reap;
2723 free(chain->map, M_IPFW);
2724 IPFW_WUNLOCK(chain);
2725 IPFW_UH_WUNLOCK(chain);
2727 ipfw_reap_rules(reap);
2728 IPFW_LOCK_DESTROY(chain);
2729 ipfw_dyn_uninit(1); /* free the remaining parts */
2734 * Module event handler.
2735 * In general we have the choice of handling most of these events by the
2736 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
2737 * use the SYSINIT handlers as they are more capable of expressing the
2738 * flow of control during module and vnet operations, so this is just
2739 * a skeleton. Note there is no SYSINIT equivalent of the module
2740 * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
2743 ipfw_modevent(module_t mod, int type, void *unused)
2749 /* Called once at module load or
2750 * system boot if compiled in. */
2753 /* Called before unload. May veto unloading. */
2756 /* Called during unload. */
2759 /* Called during system shutdown. */
2768 static moduledata_t ipfwmod = {
2774 /* Define startup order. */
2775 #define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_IFATTACHDOMAIN
2776 #define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */
2777 #define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */
2778 #define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */
2780 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
2781 MODULE_VERSION(ipfw, 2);
2782 /* should declare some dependencies here */
2785 * Starting up. Done in order after ipfwmod() has been called.
2786 * VNET_SYSINIT is also called for each existing vnet and each new vnet.
2788 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2790 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2791 vnet_ipfw_init, NULL);
2794 * Closing up shop. These are done in REVERSE ORDER, but still
2795 * after ipfwmod() has been called. Not called on reboot.
2796 * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
2797 * or when the module is unloaded.
2799 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2800 ipfw_destroy, NULL);
2801 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2802 vnet_ipfw_uninit, NULL);