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 VNET_DEFINE(int, ipfw_nat_ready) = 0;
147 ipfw_nat_t *ipfw_nat_ptr = NULL;
148 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
149 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
150 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
151 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
152 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
155 uint32_t dummy_def = IPFW_DEFAULT_RULE;
156 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS);
160 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
161 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
162 CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
163 "Only do a single pass through ipfw when using dummynet(4)");
164 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
165 CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
166 "Rule number auto-increment step");
167 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose,
168 CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
169 "Log matches to ipfw rules");
170 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
171 CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
172 "Set upper limit of matches of ipfw rules logged");
173 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
175 "The default/max possible rule number.");
176 SYSCTL_VNET_PROC(_net_inet_ip_fw, OID_AUTO, tables_max,
177 CTLTYPE_UINT|CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU",
178 "Maximum number of tables");
179 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
180 &default_to_accept, 0,
181 "Make the default rule accept all packets.");
182 TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept);
183 TUNABLE_INT("net.inet.ip.fw.tables_max", &default_fw_tables);
184 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, static_count,
185 CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
186 "Number of static rules");
189 SYSCTL_DECL(_net_inet6_ip6);
190 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
191 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
192 CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0,
193 "Deny packets with unknown IPv6 Extension Headers");
194 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
195 CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_permit_single_frag6), 0,
196 "Permit single packet IPv6 fragments");
201 #endif /* SYSCTL_NODE */
205 * Some macros used in the various matching options.
206 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
207 * Other macros just cast void * into the appropriate type
209 #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
210 #define TCP(p) ((struct tcphdr *)(p))
211 #define SCTP(p) ((struct sctphdr *)(p))
212 #define UDP(p) ((struct udphdr *)(p))
213 #define ICMP(p) ((struct icmphdr *)(p))
214 #define ICMP6(p) ((struct icmp6_hdr *)(p))
217 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
219 int type = icmp->icmp_type;
221 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
224 #define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
225 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
228 is_icmp_query(struct icmphdr *icmp)
230 int type = icmp->icmp_type;
232 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
237 * The following checks use two arrays of 8 or 16 bits to store the
238 * bits that we want set or clear, respectively. They are in the
239 * low and high half of cmd->arg1 or cmd->d[0].
241 * We scan options and store the bits we find set. We succeed if
243 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
245 * The code is sometimes optimized not to store additional variables.
249 flags_match(ipfw_insn *cmd, u_int8_t bits)
254 if ( ((cmd->arg1 & 0xff) & bits) != 0)
255 return 0; /* some bits we want set were clear */
256 want_clear = (cmd->arg1 >> 8) & 0xff;
257 if ( (want_clear & bits) != want_clear)
258 return 0; /* some bits we want clear were set */
263 ipopts_match(struct ip *ip, ipfw_insn *cmd)
265 int optlen, bits = 0;
266 u_char *cp = (u_char *)(ip + 1);
267 int x = (ip->ip_hl << 2) - sizeof (struct ip);
269 for (; x > 0; x -= optlen, cp += optlen) {
270 int opt = cp[IPOPT_OPTVAL];
272 if (opt == IPOPT_EOL)
274 if (opt == IPOPT_NOP)
277 optlen = cp[IPOPT_OLEN];
278 if (optlen <= 0 || optlen > x)
279 return 0; /* invalid or truncated */
287 bits |= IP_FW_IPOPT_LSRR;
291 bits |= IP_FW_IPOPT_SSRR;
295 bits |= IP_FW_IPOPT_RR;
299 bits |= IP_FW_IPOPT_TS;
303 return (flags_match(cmd, bits));
307 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
309 int optlen, bits = 0;
310 u_char *cp = (u_char *)(tcp + 1);
311 int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
313 for (; x > 0; x -= optlen, cp += optlen) {
315 if (opt == TCPOPT_EOL)
317 if (opt == TCPOPT_NOP)
331 bits |= IP_FW_TCPOPT_MSS;
335 bits |= IP_FW_TCPOPT_WINDOW;
338 case TCPOPT_SACK_PERMITTED:
340 bits |= IP_FW_TCPOPT_SACK;
343 case TCPOPT_TIMESTAMP:
344 bits |= IP_FW_TCPOPT_TS;
349 return (flags_match(cmd, bits));
353 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain, uint32_t *tablearg)
355 if (ifp == NULL) /* no iface with this packet, match fails */
357 /* Check by name or by IP address */
358 if (cmd->name[0] != '\0') { /* match by name */
359 if (cmd->name[0] == '\1') /* use tablearg to match */
360 return ipfw_lookup_table_extended(chain, cmd->p.glob,
361 ifp->if_xname, tablearg, IPFW_TABLE_INTERFACE);
364 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
367 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
371 #ifdef __FreeBSD__ /* and OSX too ? */
375 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
376 if (ia->ifa_addr->sa_family != AF_INET)
378 if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
379 (ia->ifa_addr))->sin_addr.s_addr) {
380 if_addr_runlock(ifp);
381 return(1); /* match */
384 if_addr_runlock(ifp);
385 #endif /* __FreeBSD__ */
387 return(0); /* no match, fail ... */
391 * The verify_path function checks if a route to the src exists and
392 * if it is reachable via ifp (when provided).
394 * The 'verrevpath' option checks that the interface that an IP packet
395 * arrives on is the same interface that traffic destined for the
396 * packet's source address would be routed out of.
397 * The 'versrcreach' option just checks that the source address is
398 * reachable via any route (except default) in the routing table.
399 * These two are a measure to block forged packets. This is also
400 * commonly known as "anti-spoofing" or Unicast Reverse Path
401 * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
402 * is purposely reminiscent of the Cisco IOS command,
404 * ip verify unicast reverse-path
405 * ip verify unicast source reachable-via any
407 * which implements the same functionality. But note that the syntax
408 * is misleading, and the check may be performed on all IP packets
409 * whether unicast, multicast, or broadcast.
412 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
418 struct sockaddr_in *dst;
420 bzero(&ro, sizeof(ro));
422 dst = (struct sockaddr_in *)&(ro.ro_dst);
423 dst->sin_family = AF_INET;
424 dst->sin_len = sizeof(*dst);
426 in_rtalloc_ign(&ro, 0, fib);
428 if (ro.ro_rt == NULL)
432 * If ifp is provided, check for equality with rtentry.
433 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
434 * in order to pass packets injected back by if_simloop():
435 * if useloopback == 1 routing entry (via lo0) for our own address
436 * may exist, so we need to handle routing assymetry.
438 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
443 /* if no ifp provided, check if rtentry is not default route */
445 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) {
450 /* or if this is a blackhole/reject route */
451 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
456 /* found valid route */
459 #endif /* __FreeBSD__ */
464 * ipv6 specific rules here...
467 icmp6type_match (int type, ipfw_insn_u32 *cmd)
469 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
473 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
476 for (i=0; i <= cmd->o.arg1; ++i )
477 if (curr_flow == cmd->d[i] )
482 /* support for IP6_*_ME opcodes */
484 search_ip6_addr_net (struct in6_addr * ip6_addr)
488 struct in6_ifaddr *fdm;
489 struct in6_addr copia;
491 TAILQ_FOREACH(mdc, &V_ifnet, if_link) {
493 TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) {
494 if (mdc2->ifa_addr->sa_family == AF_INET6) {
495 fdm = (struct in6_ifaddr *)mdc2;
496 copia = fdm->ia_addr.sin6_addr;
497 /* need for leaving scope_id in the sock_addr */
498 in6_clearscope(&copia);
499 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) {
500 if_addr_runlock(mdc);
505 if_addr_runlock(mdc);
511 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
514 struct sockaddr_in6 *dst;
516 bzero(&ro, sizeof(ro));
518 dst = (struct sockaddr_in6 * )&(ro.ro_dst);
519 dst->sin6_family = AF_INET6;
520 dst->sin6_len = sizeof(*dst);
521 dst->sin6_addr = *src;
523 in6_rtalloc_ign(&ro, 0, fib);
524 if (ro.ro_rt == NULL)
528 * if ifp is provided, check for equality with rtentry
529 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
530 * to support the case of sending packets to an address of our own.
531 * (where the former interface is the first argument of if_simloop()
532 * (=ifp), the latter is lo0)
534 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
539 /* if no ifp provided, check if rtentry is not default route */
541 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) {
546 /* or if this is a blackhole/reject route */
547 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
552 /* found valid route */
559 is_icmp6_query(int icmp6_type)
561 if ((icmp6_type <= ICMP6_MAXTYPE) &&
562 (icmp6_type == ICMP6_ECHO_REQUEST ||
563 icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
564 icmp6_type == ICMP6_WRUREQUEST ||
565 icmp6_type == ICMP6_FQDN_QUERY ||
566 icmp6_type == ICMP6_NI_QUERY))
573 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
578 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
580 tcp = (struct tcphdr *)((char *)ip6 + hlen);
582 if ((tcp->th_flags & TH_RST) == 0) {
584 m0 = ipfw_send_pkt(args->m, &(args->f_id),
585 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
586 tcp->th_flags | TH_RST);
588 ip6_output(m0, NULL, NULL, 0, NULL, NULL,
592 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
595 * Unlike above, the mbufs need to line up with the ip6 hdr,
596 * as the contents are read. We need to m_adj() the
598 * The mbuf will however be thrown away so we can adjust it.
599 * Remember we did an m_pullup on it already so we
600 * can make some assumptions about contiguousness.
603 m_adj(m, args->L3offset);
605 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
616 * sends a reject message, consuming the mbuf passed as an argument.
619 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
623 /* XXX When ip is not guaranteed to be at mtod() we will
624 * need to account for this */
625 * The mbuf will however be thrown away so we can adjust it.
626 * Remember we did an m_pullup on it already so we
627 * can make some assumptions about contiguousness.
630 m_adj(m, args->L3offset);
632 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
633 /* We need the IP header in host order for icmp_error(). */
635 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
636 } else if (args->f_id.proto == IPPROTO_TCP) {
637 struct tcphdr *const tcp =
638 L3HDR(struct tcphdr, mtod(args->m, struct ip *));
639 if ( (tcp->th_flags & TH_RST) == 0) {
641 m = ipfw_send_pkt(args->m, &(args->f_id),
642 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
643 tcp->th_flags | TH_RST);
645 ip_output(m, NULL, NULL, 0, NULL, NULL);
654 * Support for uid/gid/jail lookup. These tests are expensive
655 * (because we may need to look into the list of active sockets)
656 * so we cache the results. ugid_lookupp is 0 if we have not
657 * yet done a lookup, 1 if we succeeded, and -1 if we tried
658 * and failed. The function always returns the match value.
659 * We could actually spare the variable and use *uc, setting
660 * it to '(void *)check_uidgid if we have no info, NULL if
661 * we tried and failed, or any other value if successful.
664 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
669 return cred_check(insn, proto, oif,
670 dst_ip, dst_port, src_ip, src_port,
671 (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
673 struct in_addr src_ip, dst_ip;
674 struct inpcbinfo *pi;
675 struct ipfw_flow_id *id;
676 struct inpcb *pcb, *inp;
686 * Check to see if the UDP or TCP stack supplied us with
687 * the PCB. If so, rather then holding a lock and looking
688 * up the PCB, we can use the one that was supplied.
690 if (inp && *ugid_lookupp == 0) {
691 INP_LOCK_ASSERT(inp);
692 if (inp->inp_socket != NULL) {
693 *uc = crhold(inp->inp_cred);
699 * If we have already been here and the packet has no
700 * PCB entry associated with it, then we can safely
701 * assume that this is a no match.
703 if (*ugid_lookupp == -1)
705 if (id->proto == IPPROTO_TCP) {
708 } else if (id->proto == IPPROTO_UDP) {
709 lookupflags = INPLOOKUP_WILDCARD;
713 lookupflags |= INPLOOKUP_RLOCKPCB;
715 if (*ugid_lookupp == 0) {
716 if (id->addr_type == 6) {
719 pcb = in6_pcblookup_mbuf(pi,
720 &id->src_ip6, htons(id->src_port),
721 &id->dst_ip6, htons(id->dst_port),
722 lookupflags, oif, args->m);
724 pcb = in6_pcblookup_mbuf(pi,
725 &id->dst_ip6, htons(id->dst_port),
726 &id->src_ip6, htons(id->src_port),
727 lookupflags, oif, args->m);
733 src_ip.s_addr = htonl(id->src_ip);
734 dst_ip.s_addr = htonl(id->dst_ip);
736 pcb = in_pcblookup_mbuf(pi,
737 src_ip, htons(id->src_port),
738 dst_ip, htons(id->dst_port),
739 lookupflags, oif, args->m);
741 pcb = in_pcblookup_mbuf(pi,
742 dst_ip, htons(id->dst_port),
743 src_ip, htons(id->src_port),
744 lookupflags, oif, args->m);
747 INP_RLOCK_ASSERT(pcb);
748 *uc = crhold(pcb->inp_cred);
752 if (*ugid_lookupp == 0) {
754 * We tried and failed, set the variable to -1
755 * so we will not try again on this packet.
761 if (insn->o.opcode == O_UID)
762 match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
763 else if (insn->o.opcode == O_GID)
764 match = groupmember((gid_t)insn->d[0], *uc);
765 else if (insn->o.opcode == O_JAIL)
766 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
768 #endif /* __FreeBSD__ */
772 * Helper function to set args with info on the rule after the matching
773 * one. slot is precise, whereas we guess rule_id as they are
774 * assigned sequentially.
777 set_match(struct ip_fw_args *args, int slot,
778 struct ip_fw_chain *chain)
780 args->rule.chain_id = chain->id;
781 args->rule.slot = slot + 1; /* we use 0 as a marker */
782 args->rule.rule_id = 1 + chain->map[slot]->id;
783 args->rule.rulenum = chain->map[slot]->rulenum;
787 * Helper function to enable cached rule lookups using
788 * x_next and next_rule fields in ipfw rule.
791 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
792 int tablearg, int jump_backwards)
796 /* If possible use cached f_pos (in f->next_rule),
797 * whose version is written in f->next_rule
798 * (horrible hacks to avoid changing the ABI).
800 if (num != IP_FW_TABLEARG && (uintptr_t)f->x_next == chain->id)
801 f_pos = (uintptr_t)f->next_rule;
803 int i = IP_FW_ARG_TABLEARG(num);
804 /* make sure we do not jump backward */
805 if (jump_backwards == 0 && i <= f->rulenum)
807 f_pos = ipfw_find_rule(chain, i, 0);
808 /* update the cache */
809 if (num != IP_FW_TABLEARG) {
810 f->next_rule = (void *)(uintptr_t)f_pos;
811 f->x_next = (void *)(uintptr_t)chain->id;
819 * The main check routine for the firewall.
821 * All arguments are in args so we can modify them and return them
822 * back to the caller.
826 * args->m (in/out) The packet; we set to NULL when/if we nuke it.
827 * Starts with the IP header.
828 * args->eh (in) Mac header if present, NULL for layer3 packet.
829 * args->L3offset Number of bytes bypassed if we came from L2.
830 * e.g. often sizeof(eh) ** NOTYET **
831 * args->oif Outgoing interface, NULL if packet is incoming.
832 * The incoming interface is in the mbuf. (in)
833 * args->divert_rule (in/out)
834 * Skip up to the first rule past this rule number;
835 * upon return, non-zero port number for divert or tee.
837 * args->rule Pointer to the last matching rule (in/out)
838 * args->next_hop Socket we are forwarding to (out).
839 * args->next_hop6 IPv6 next hop we are forwarding to (out).
840 * args->f_id Addresses grabbed from the packet (out)
841 * args->rule.info a cookie depending on rule action
845 * IP_FW_PASS the packet must be accepted
846 * IP_FW_DENY the packet must be dropped
847 * IP_FW_DIVERT divert packet, port in m_tag
848 * IP_FW_TEE tee packet, port in m_tag
849 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie
850 * IP_FW_NETGRAPH into netgraph, cookie args->cookie
851 * args->rule contains the matching rule,
852 * args->rule.info has additional information.
856 ipfw_chk(struct ip_fw_args *args)
860 * Local variables holding state while processing a packet:
862 * IMPORTANT NOTE: to speed up the processing of rules, there
863 * are some assumption on the values of the variables, which
864 * are documented here. Should you change them, please check
865 * the implementation of the various instructions to make sure
866 * that they still work.
868 * args->eh The MAC header. It is non-null for a layer2
869 * packet, it is NULL for a layer-3 packet.
871 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
873 * m | args->m Pointer to the mbuf, as received from the caller.
874 * It may change if ipfw_chk() does an m_pullup, or if it
875 * consumes the packet because it calls send_reject().
876 * XXX This has to change, so that ipfw_chk() never modifies
877 * or consumes the buffer.
878 * ip is the beginning of the ip(4 or 6) header.
879 * Calculated by adding the L3offset to the start of data.
880 * (Until we start using L3offset, the packet is
881 * supposed to start with the ip header).
883 struct mbuf *m = args->m;
884 struct ip *ip = mtod(m, struct ip *);
887 * For rules which contain uid/gid or jail constraints, cache
888 * a copy of the users credentials after the pcb lookup has been
889 * executed. This will speed up the processing of rules with
890 * these types of constraints, as well as decrease contention
891 * on pcb related locks.
894 struct bsd_ucred ucred_cache;
896 struct ucred *ucred_cache = NULL;
898 int ucred_lookup = 0;
901 * oif | args->oif If NULL, ipfw_chk has been called on the
902 * inbound path (ether_input, ip_input).
903 * If non-NULL, ipfw_chk has been called on the outbound path
904 * (ether_output, ip_output).
906 struct ifnet *oif = args->oif;
908 int f_pos = 0; /* index of current rule in the array */
912 * hlen The length of the IP header.
914 u_int hlen = 0; /* hlen >0 means we have an IP pkt */
917 * offset The offset of a fragment. offset != 0 means that
918 * we have a fragment at this offset of an IPv4 packet.
919 * offset == 0 means that (if this is an IPv4 packet)
920 * this is the first or only fragment.
921 * For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
922 * or there is a single packet fragement (fragement header added
923 * without needed). We will treat a single packet fragment as if
924 * there was no fragment header (or log/block depending on the
925 * V_fw_permit_single_frag6 sysctl setting).
931 * Local copies of addresses. They are only valid if we have
934 * proto The protocol. Set to 0 for non-ip packets,
935 * or to the protocol read from the packet otherwise.
936 * proto != 0 means that we have an IPv4 packet.
938 * src_port, dst_port port numbers, in HOST format. Only
939 * valid for TCP and UDP packets.
941 * src_ip, dst_ip ip addresses, in NETWORK format.
942 * Only valid for IPv4 packets.
945 uint16_t src_port = 0, dst_port = 0; /* NOTE: host format */
946 struct in_addr src_ip, dst_ip; /* NOTE: network format */
949 uint16_t etype = 0; /* Host order stored ether type */
952 * dyn_dir = MATCH_UNKNOWN when rules unchecked,
953 * MATCH_NONE when checked and not matched (q = NULL),
954 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
956 int dyn_dir = MATCH_UNKNOWN;
957 ipfw_dyn_rule *q = NULL;
958 struct ip_fw_chain *chain = &V_layer3_chain;
961 * We store in ulp a pointer to the upper layer protocol header.
962 * In the ipv4 case this is easy to determine from the header,
963 * but for ipv6 we might have some additional headers in the middle.
964 * ulp is NULL if not found.
966 void *ulp = NULL; /* upper layer protocol pointer. */
968 /* XXX ipv6 variables */
970 uint8_t icmp6_type = 0;
971 uint16_t ext_hd = 0; /* bits vector for extension header filtering */
972 /* end of ipv6 variables */
976 int done = 0; /* flag to exit the outer loop */
978 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
979 return (IP_FW_PASS); /* accept */
981 dst_ip.s_addr = 0; /* make sure it is initialized */
982 src_ip.s_addr = 0; /* make sure it is initialized */
983 pktlen = m->m_pkthdr.len;
984 args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
985 proto = args->f_id.proto = 0; /* mark f_id invalid */
986 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
989 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
990 * then it sets p to point at the offset "len" in the mbuf. WARNING: the
991 * pointer might become stale after other pullups (but we never use it
994 #define PULLUP_TO(_len, p, T) PULLUP_LEN(_len, p, sizeof(T))
995 #define PULLUP_LEN(_len, p, T) \
997 int x = (_len) + T; \
998 if ((m)->m_len < x) { \
999 args->m = m = m_pullup(m, x); \
1001 goto pullup_failed; \
1003 p = (mtod(m, char *) + (_len)); \
1007 * if we have an ether header,
1010 etype = ntohs(args->eh->ether_type);
1012 /* Identify IP packets and fill up variables. */
1013 if (pktlen >= sizeof(struct ip6_hdr) &&
1014 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
1015 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1017 args->f_id.addr_type = 6;
1018 hlen = sizeof(struct ip6_hdr);
1019 proto = ip6->ip6_nxt;
1021 /* Search extension headers to find upper layer protocols */
1022 while (ulp == NULL && offset == 0) {
1024 case IPPROTO_ICMPV6:
1025 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1026 icmp6_type = ICMP6(ulp)->icmp6_type;
1030 PULLUP_TO(hlen, ulp, struct tcphdr);
1031 dst_port = TCP(ulp)->th_dport;
1032 src_port = TCP(ulp)->th_sport;
1033 /* save flags for dynamic rules */
1034 args->f_id._flags = TCP(ulp)->th_flags;
1038 PULLUP_TO(hlen, ulp, struct sctphdr);
1039 src_port = SCTP(ulp)->src_port;
1040 dst_port = SCTP(ulp)->dest_port;
1044 PULLUP_TO(hlen, ulp, struct udphdr);
1045 dst_port = UDP(ulp)->uh_dport;
1046 src_port = UDP(ulp)->uh_sport;
1049 case IPPROTO_HOPOPTS: /* RFC 2460 */
1050 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1051 ext_hd |= EXT_HOPOPTS;
1052 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1053 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1057 case IPPROTO_ROUTING: /* RFC 2460 */
1058 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1059 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1061 ext_hd |= EXT_RTHDR0;
1064 ext_hd |= EXT_RTHDR2;
1068 printf("IPFW2: IPV6 - Unknown "
1069 "Routing Header type(%d)\n",
1070 ((struct ip6_rthdr *)
1072 if (V_fw_deny_unknown_exthdrs)
1073 return (IP_FW_DENY);
1076 ext_hd |= EXT_ROUTING;
1077 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1078 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1082 case IPPROTO_FRAGMENT: /* RFC 2460 */
1083 PULLUP_TO(hlen, ulp, struct ip6_frag);
1084 ext_hd |= EXT_FRAGMENT;
1085 hlen += sizeof (struct ip6_frag);
1086 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1087 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1089 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1091 if (V_fw_permit_single_frag6 == 0 &&
1092 offset == 0 && ip6f_mf == 0) {
1094 printf("IPFW2: IPV6 - Invalid "
1095 "Fragment Header\n");
1096 if (V_fw_deny_unknown_exthdrs)
1097 return (IP_FW_DENY);
1101 ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1105 case IPPROTO_DSTOPTS: /* RFC 2460 */
1106 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1107 ext_hd |= EXT_DSTOPTS;
1108 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1109 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1113 case IPPROTO_AH: /* RFC 2402 */
1114 PULLUP_TO(hlen, ulp, struct ip6_ext);
1116 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1117 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1121 case IPPROTO_ESP: /* RFC 2406 */
1122 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
1123 /* Anything past Seq# is variable length and
1124 * data past this ext. header is encrypted. */
1128 case IPPROTO_NONE: /* RFC 2460 */
1130 * Packet ends here, and IPv6 header has
1131 * already been pulled up. If ip6e_len!=0
1132 * then octets must be ignored.
1134 ulp = ip; /* non-NULL to get out of loop. */
1137 case IPPROTO_OSPFIGP:
1138 /* XXX OSPF header check? */
1139 PULLUP_TO(hlen, ulp, struct ip6_ext);
1143 /* XXX PIM header check? */
1144 PULLUP_TO(hlen, ulp, struct pim);
1148 PULLUP_TO(hlen, ulp, struct carp_header);
1149 if (((struct carp_header *)ulp)->carp_version !=
1151 return (IP_FW_DENY);
1152 if (((struct carp_header *)ulp)->carp_type !=
1154 return (IP_FW_DENY);
1157 case IPPROTO_IPV6: /* RFC 2893 */
1158 PULLUP_TO(hlen, ulp, struct ip6_hdr);
1161 case IPPROTO_IPV4: /* RFC 2893 */
1162 PULLUP_TO(hlen, ulp, struct ip);
1167 printf("IPFW2: IPV6 - Unknown "
1168 "Extension Header(%d), ext_hd=%x\n",
1170 if (V_fw_deny_unknown_exthdrs)
1171 return (IP_FW_DENY);
1172 PULLUP_TO(hlen, ulp, struct ip6_ext);
1176 ip = mtod(m, struct ip *);
1177 ip6 = (struct ip6_hdr *)ip;
1178 args->f_id.src_ip6 = ip6->ip6_src;
1179 args->f_id.dst_ip6 = ip6->ip6_dst;
1180 args->f_id.src_ip = 0;
1181 args->f_id.dst_ip = 0;
1182 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1183 } else if (pktlen >= sizeof(struct ip) &&
1184 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1186 hlen = ip->ip_hl << 2;
1187 args->f_id.addr_type = 4;
1190 * Collect parameters into local variables for faster matching.
1193 src_ip = ip->ip_src;
1194 dst_ip = ip->ip_dst;
1195 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1196 iplen = ntohs(ip->ip_len);
1197 pktlen = iplen < pktlen ? iplen : pktlen;
1202 PULLUP_TO(hlen, ulp, struct tcphdr);
1203 dst_port = TCP(ulp)->th_dport;
1204 src_port = TCP(ulp)->th_sport;
1205 /* save flags for dynamic rules */
1206 args->f_id._flags = TCP(ulp)->th_flags;
1210 PULLUP_TO(hlen, ulp, struct sctphdr);
1211 src_port = SCTP(ulp)->src_port;
1212 dst_port = SCTP(ulp)->dest_port;
1216 PULLUP_TO(hlen, ulp, struct udphdr);
1217 dst_port = UDP(ulp)->uh_dport;
1218 src_port = UDP(ulp)->uh_sport;
1222 PULLUP_TO(hlen, ulp, struct icmphdr);
1223 //args->f_id.flags = ICMP(ulp)->icmp_type;
1231 ip = mtod(m, struct ip *);
1232 args->f_id.src_ip = ntohl(src_ip.s_addr);
1233 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1236 if (proto) { /* we may have port numbers, store them */
1237 args->f_id.proto = proto;
1238 args->f_id.src_port = src_port = ntohs(src_port);
1239 args->f_id.dst_port = dst_port = ntohs(dst_port);
1243 if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1244 IPFW_RUNLOCK(chain);
1245 return (IP_FW_PASS); /* accept */
1247 if (args->rule.slot) {
1249 * Packet has already been tagged as a result of a previous
1250 * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1251 * REASS, NETGRAPH, DIVERT/TEE...)
1252 * Validate the slot and continue from the next one
1253 * if still present, otherwise do a lookup.
1255 f_pos = (args->rule.chain_id == chain->id) ?
1257 ipfw_find_rule(chain, args->rule.rulenum,
1258 args->rule.rule_id);
1264 * Now scan the rules, and parse microinstructions for each rule.
1265 * We have two nested loops and an inner switch. Sometimes we
1266 * need to break out of one or both loops, or re-enter one of
1267 * the loops with updated variables. Loop variables are:
1269 * f_pos (outer loop) points to the current rule.
1270 * On output it points to the matching rule.
1271 * done (outer loop) is used as a flag to break the loop.
1272 * l (inner loop) residual length of current rule.
1273 * cmd points to the current microinstruction.
1275 * We break the inner loop by setting l=0 and possibly
1276 * cmdlen=0 if we don't want to advance cmd.
1277 * We break the outer loop by setting done=1
1278 * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1281 for (; f_pos < chain->n_rules; f_pos++) {
1283 uint32_t tablearg = 0;
1284 int l, cmdlen, skip_or; /* skip rest of OR block */
1287 f = chain->map[f_pos];
1288 if (V_set_disable & (1 << f->set) )
1292 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1293 l -= cmdlen, cmd += cmdlen) {
1297 * check_body is a jump target used when we find a
1298 * CHECK_STATE, and need to jump to the body of
1303 cmdlen = F_LEN(cmd);
1305 * An OR block (insn_1 || .. || insn_n) has the
1306 * F_OR bit set in all but the last instruction.
1307 * The first match will set "skip_or", and cause
1308 * the following instructions to be skipped until
1309 * past the one with the F_OR bit clear.
1311 if (skip_or) { /* skip this instruction */
1312 if ((cmd->len & F_OR) == 0)
1313 skip_or = 0; /* next one is good */
1316 match = 0; /* set to 1 if we succeed */
1318 switch (cmd->opcode) {
1320 * The first set of opcodes compares the packet's
1321 * fields with some pattern, setting 'match' if a
1322 * match is found. At the end of the loop there is
1323 * logic to deal with F_NOT and F_OR flags associated
1331 printf("ipfw: opcode %d unimplemented\n",
1339 * We only check offset == 0 && proto != 0,
1340 * as this ensures that we have a
1341 * packet with the ports info.
1345 if (proto == IPPROTO_TCP ||
1346 proto == IPPROTO_UDP)
1347 match = check_uidgid(
1348 (ipfw_insn_u32 *)cmd,
1349 args, &ucred_lookup,
1353 (void *)&ucred_cache);
1358 match = iface_match(m->m_pkthdr.rcvif,
1359 (ipfw_insn_if *)cmd, chain, &tablearg);
1363 match = iface_match(oif, (ipfw_insn_if *)cmd,
1368 match = iface_match(oif ? oif :
1369 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1374 if (args->eh != NULL) { /* have MAC header */
1375 u_int32_t *want = (u_int32_t *)
1376 ((ipfw_insn_mac *)cmd)->addr;
1377 u_int32_t *mask = (u_int32_t *)
1378 ((ipfw_insn_mac *)cmd)->mask;
1379 u_int32_t *hdr = (u_int32_t *)args->eh;
1382 ( want[0] == (hdr[0] & mask[0]) &&
1383 want[1] == (hdr[1] & mask[1]) &&
1384 want[2] == (hdr[2] & mask[2]) );
1389 if (args->eh != NULL) {
1391 ((ipfw_insn_u16 *)cmd)->ports;
1394 for (i = cmdlen - 1; !match && i>0;
1396 match = (etype >= p[0] &&
1402 match = (offset != 0);
1405 case O_IN: /* "out" is "not in" */
1406 match = (oif == NULL);
1410 match = (args->eh != NULL);
1415 /* For diverted packets, args->rule.info
1416 * contains the divert port (in host format)
1417 * reason and direction.
1419 uint32_t i = args->rule.info;
1420 match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1421 cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1427 * We do not allow an arg of 0 so the
1428 * check of "proto" only suffices.
1430 match = (proto == cmd->arg1);
1435 (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1439 case O_IP_SRC_LOOKUP:
1440 case O_IP_DST_LOOKUP:
1443 (cmd->opcode == O_IP_DST_LOOKUP) ?
1444 dst_ip.s_addr : src_ip.s_addr;
1447 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1448 /* generic lookup. The key must be
1449 * in 32bit big-endian format.
1451 v = ((ipfw_insn_u32 *)cmd)->d[1];
1453 key = dst_ip.s_addr;
1455 key = src_ip.s_addr;
1456 else if (v == 6) /* dscp */
1457 key = (ip->ip_tos >> 2) & 0x3f;
1458 else if (offset != 0)
1460 else if (proto != IPPROTO_TCP &&
1461 proto != IPPROTO_UDP)
1464 key = htonl(dst_port);
1466 key = htonl(src_port);
1467 else if (v == 4 || v == 5) {
1469 (ipfw_insn_u32 *)cmd,
1470 args, &ucred_lookup,
1473 if (v == 4 /* O_UID */)
1474 key = ucred_cache->cr_uid;
1475 else if (v == 5 /* O_JAIL */)
1476 key = ucred_cache->cr_prison->pr_id;
1477 #else /* !__FreeBSD__ */
1478 (void *)&ucred_cache);
1479 if (v ==4 /* O_UID */)
1480 key = ucred_cache.uid;
1481 else if (v == 5 /* O_JAIL */)
1482 key = ucred_cache.xid;
1483 #endif /* !__FreeBSD__ */
1488 match = ipfw_lookup_table(chain,
1489 cmd->arg1, key, &v);
1492 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1494 ((ipfw_insn_u32 *)cmd)->d[0] == v;
1497 } else if (is_ipv6) {
1499 void *pkey = (cmd->opcode == O_IP_DST_LOOKUP) ?
1500 &args->f_id.dst_ip6: &args->f_id.src_ip6;
1501 match = ipfw_lookup_table_extended(chain,
1502 cmd->arg1, pkey, &v,
1504 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1505 match = ((ipfw_insn_u32 *)cmd)->d[0] == v;
1515 (cmd->opcode == O_IP_DST_MASK) ?
1516 dst_ip.s_addr : src_ip.s_addr;
1517 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
1520 for (; !match && i>0; i-= 2, p+= 2)
1521 match = (p[0] == (a & p[1]));
1529 INADDR_TO_IFP(src_ip, tif);
1530 match = (tif != NULL);
1536 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6);
1543 u_int32_t *d = (u_int32_t *)(cmd+1);
1545 cmd->opcode == O_IP_DST_SET ?
1551 addr -= d[0]; /* subtract base */
1552 match = (addr < cmd->arg1) &&
1553 ( d[ 1 + (addr>>5)] &
1554 (1<<(addr & 0x1f)) );
1560 (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1568 INADDR_TO_IFP(dst_ip, tif);
1569 match = (tif != NULL);
1575 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6);
1583 * offset == 0 && proto != 0 is enough
1584 * to guarantee that we have a
1585 * packet with port info.
1587 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
1590 (cmd->opcode == O_IP_SRCPORT) ?
1591 src_port : dst_port ;
1593 ((ipfw_insn_u16 *)cmd)->ports;
1596 for (i = cmdlen - 1; !match && i>0;
1598 match = (x>=p[0] && x<=p[1]);
1603 match = (offset == 0 && proto==IPPROTO_ICMP &&
1604 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
1609 match = is_ipv6 && offset == 0 &&
1610 proto==IPPROTO_ICMPV6 &&
1612 ICMP6(ulp)->icmp6_type,
1613 (ipfw_insn_u32 *)cmd);
1619 ipopts_match(ip, cmd) );
1624 cmd->arg1 == ip->ip_v);
1630 if (is_ipv4) { /* only for IP packets */
1635 if (cmd->opcode == O_IPLEN)
1637 else if (cmd->opcode == O_IPTTL)
1639 else /* must be IPID */
1640 x = ntohs(ip->ip_id);
1642 match = (cmd->arg1 == x);
1645 /* otherwise we have ranges */
1646 p = ((ipfw_insn_u16 *)cmd)->ports;
1648 for (; !match && i>0; i--, p += 2)
1649 match = (x >= p[0] && x <= p[1]);
1653 case O_IPPRECEDENCE:
1655 (cmd->arg1 == (ip->ip_tos & 0xe0)) );
1660 flags_match(cmd, ip->ip_tos));
1668 p = ((ipfw_insn_u32 *)cmd)->d;
1671 x = ip->ip_tos >> 2;
1674 v = &((struct ip6_hdr *)ip)->ip6_vfc;
1675 x = (*v & 0x0F) << 2;
1681 /* DSCP bitmask is stored as low_u32 high_u32 */
1683 match = *(p + 1) & (1 << (x - 32));
1685 match = *p & (1 << x);
1690 if (proto == IPPROTO_TCP && offset == 0) {
1698 ((ip->ip_hl + tcp->th_off) << 2);
1700 match = (cmd->arg1 == x);
1703 /* otherwise we have ranges */
1704 p = ((ipfw_insn_u16 *)cmd)->ports;
1706 for (; !match && i>0; i--, p += 2)
1707 match = (x >= p[0] && x <= p[1]);
1712 match = (proto == IPPROTO_TCP && offset == 0 &&
1713 flags_match(cmd, TCP(ulp)->th_flags));
1717 PULLUP_LEN(hlen, ulp, (TCP(ulp)->th_off << 2));
1718 match = (proto == IPPROTO_TCP && offset == 0 &&
1719 tcpopts_match(TCP(ulp), cmd));
1723 match = (proto == IPPROTO_TCP && offset == 0 &&
1724 ((ipfw_insn_u32 *)cmd)->d[0] ==
1729 match = (proto == IPPROTO_TCP && offset == 0 &&
1730 ((ipfw_insn_u32 *)cmd)->d[0] ==
1735 if (proto == IPPROTO_TCP && offset == 0) {
1740 x = ntohs(TCP(ulp)->th_win);
1742 match = (cmd->arg1 == x);
1745 /* Otherwise we have ranges. */
1746 p = ((ipfw_insn_u16 *)cmd)->ports;
1748 for (; !match && i > 0; i--, p += 2)
1749 match = (x >= p[0] && x <= p[1]);
1754 /* reject packets which have SYN only */
1755 /* XXX should i also check for TH_ACK ? */
1756 match = (proto == IPPROTO_TCP && offset == 0 &&
1757 (TCP(ulp)->th_flags &
1758 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
1763 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
1766 at = pf_find_mtag(m);
1767 if (at != NULL && at->qid != 0)
1769 at = pf_get_mtag(m);
1772 * Let the packet fall back to the
1777 at->qid = altq->qid;
1783 ipfw_log(f, hlen, args, m,
1784 oif, offset | ip6f_mf, tablearg, ip);
1789 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
1793 /* Outgoing packets automatically pass/match */
1794 match = ((oif != NULL) ||
1795 (m->m_pkthdr.rcvif == NULL) ||
1799 verify_path6(&(args->f_id.src_ip6),
1800 m->m_pkthdr.rcvif, args->f_id.fib) :
1802 verify_path(src_ip, m->m_pkthdr.rcvif,
1807 /* Outgoing packets automatically pass/match */
1808 match = (hlen > 0 && ((oif != NULL) ||
1811 verify_path6(&(args->f_id.src_ip6),
1812 NULL, args->f_id.fib) :
1814 verify_path(src_ip, NULL, args->f_id.fib)));
1818 /* Outgoing packets automatically pass/match */
1819 if (oif == NULL && hlen > 0 &&
1820 ( (is_ipv4 && in_localaddr(src_ip))
1823 in6_localaddr(&(args->f_id.src_ip6)))
1828 is_ipv6 ? verify_path6(
1829 &(args->f_id.src_ip6),
1842 match = (m_tag_find(m,
1843 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
1845 /* otherwise no match */
1851 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
1852 &((ipfw_insn_ip6 *)cmd)->addr6);
1857 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
1858 &((ipfw_insn_ip6 *)cmd)->addr6);
1860 case O_IP6_SRC_MASK:
1861 case O_IP6_DST_MASK:
1865 struct in6_addr *d =
1866 &((ipfw_insn_ip6 *)cmd)->addr6;
1868 for (; !match && i > 0; d += 2,
1869 i -= F_INSN_SIZE(struct in6_addr)
1875 APPLY_MASK(&p, &d[1]);
1877 IN6_ARE_ADDR_EQUAL(&d[0],
1885 flow6id_match(args->f_id.flow_id6,
1886 (ipfw_insn_u32 *) cmd);
1891 (ext_hd & ((ipfw_insn *) cmd)->arg1);
1905 uint32_t tag = IP_FW_ARG_TABLEARG(cmd->arg1);
1907 /* Packet is already tagged with this tag? */
1908 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
1910 /* We have `untag' action when F_NOT flag is
1911 * present. And we must remove this mtag from
1912 * mbuf and reset `match' to zero (`match' will
1913 * be inversed later).
1914 * Otherwise we should allocate new mtag and
1915 * push it into mbuf.
1917 if (cmd->len & F_NOT) { /* `untag' action */
1919 m_tag_delete(m, mtag);
1923 mtag = m_tag_alloc( MTAG_IPFW,
1926 m_tag_prepend(m, mtag);
1933 case O_FIB: /* try match the specified fib */
1934 if (args->f_id.fib == cmd->arg1)
1939 struct inpcb *inp = args->inp;
1940 struct inpcbinfo *pi;
1942 if (is_ipv6) /* XXX can we remove this ? */
1945 if (proto == IPPROTO_TCP)
1947 else if (proto == IPPROTO_UDP)
1953 * XXXRW: so_user_cookie should almost
1954 * certainly be inp_user_cookie?
1957 /* For incomming packet, lookup up the
1958 inpcb using the src/dest ip/port tuple */
1960 inp = in_pcblookup(pi,
1961 src_ip, htons(src_port),
1962 dst_ip, htons(dst_port),
1963 INPLOOKUP_RLOCKPCB, NULL);
1966 inp->inp_socket->so_user_cookie;
1972 if (inp->inp_socket) {
1974 inp->inp_socket->so_user_cookie;
1984 uint32_t tag = IP_FW_ARG_TABLEARG(cmd->arg1);
1987 match = m_tag_locate(m, MTAG_IPFW,
1992 /* we have ranges */
1993 for (mtag = m_tag_first(m);
1994 mtag != NULL && !match;
1995 mtag = m_tag_next(m, mtag)) {
1999 if (mtag->m_tag_cookie != MTAG_IPFW)
2002 p = ((ipfw_insn_u16 *)cmd)->ports;
2004 for(; !match && i > 0; i--, p += 2)
2006 mtag->m_tag_id >= p[0] &&
2007 mtag->m_tag_id <= p[1];
2013 * The second set of opcodes represents 'actions',
2014 * i.e. the terminal part of a rule once the packet
2015 * matches all previous patterns.
2016 * Typically there is only one action for each rule,
2017 * and the opcode is stored at the end of the rule
2018 * (but there are exceptions -- see below).
2020 * In general, here we set retval and terminate the
2021 * outer loop (would be a 'break 3' in some language,
2022 * but we need to set l=0, done=1)
2025 * O_COUNT and O_SKIPTO actions:
2026 * instead of terminating, we jump to the next rule
2027 * (setting l=0), or to the SKIPTO target (setting
2028 * f/f_len, cmd and l as needed), respectively.
2030 * O_TAG, O_LOG and O_ALTQ action parameters:
2031 * perform some action and set match = 1;
2033 * O_LIMIT and O_KEEP_STATE: these opcodes are
2034 * not real 'actions', and are stored right
2035 * before the 'action' part of the rule.
2036 * These opcodes try to install an entry in the
2037 * state tables; if successful, we continue with
2038 * the next opcode (match=1; break;), otherwise
2039 * the packet must be dropped (set retval,
2040 * break loops with l=0, done=1)
2042 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2043 * cause a lookup of the state table, and a jump
2044 * to the 'action' part of the parent rule
2045 * if an entry is found, or
2046 * (CHECK_STATE only) a jump to the next rule if
2047 * the entry is not found.
2048 * The result of the lookup is cached so that
2049 * further instances of these opcodes become NOPs.
2050 * The jump to the next rule is done by setting
2055 if (ipfw_install_state(f,
2056 (ipfw_insn_limit *)cmd, args, tablearg)) {
2057 /* error or limit violation */
2058 retval = IP_FW_DENY;
2059 l = 0; /* exit inner loop */
2060 done = 1; /* exit outer loop */
2068 * dynamic rules are checked at the first
2069 * keep-state or check-state occurrence,
2070 * with the result being stored in dyn_dir.
2071 * The compiler introduces a PROBE_STATE
2072 * instruction for us when we have a
2073 * KEEP_STATE (because PROBE_STATE needs
2076 if (dyn_dir == MATCH_UNKNOWN &&
2077 (q = ipfw_lookup_dyn_rule(&args->f_id,
2078 &dyn_dir, proto == IPPROTO_TCP ?
2082 * Found dynamic entry, update stats
2083 * and jump to the 'action' part of
2084 * the parent rule by setting
2085 * f, cmd, l and clearing cmdlen.
2087 IPFW_INC_DYN_COUNTER(q, pktlen);
2088 /* XXX we would like to have f_pos
2089 * readily accessible in the dynamic
2090 * rule, instead of having to
2094 f_pos = ipfw_find_rule(chain,
2096 cmd = ACTION_PTR(f);
2097 l = f->cmd_len - f->act_ofs;
2104 * Dynamic entry not found. If CHECK_STATE,
2105 * skip to next rule, if PROBE_STATE just
2106 * ignore and continue with next opcode.
2108 if (cmd->opcode == O_CHECK_STATE)
2109 l = 0; /* exit inner loop */
2114 retval = 0; /* accept */
2115 l = 0; /* exit inner loop */
2116 done = 1; /* exit outer loop */
2121 set_match(args, f_pos, chain);
2122 args->rule.info = IP_FW_ARG_TABLEARG(cmd->arg1);
2123 if (cmd->opcode == O_PIPE)
2124 args->rule.info |= IPFW_IS_PIPE;
2126 args->rule.info |= IPFW_ONEPASS;
2127 retval = IP_FW_DUMMYNET;
2128 l = 0; /* exit inner loop */
2129 done = 1; /* exit outer loop */
2134 if (args->eh) /* not on layer 2 */
2136 /* otherwise this is terminal */
2137 l = 0; /* exit inner loop */
2138 done = 1; /* exit outer loop */
2139 retval = (cmd->opcode == O_DIVERT) ?
2140 IP_FW_DIVERT : IP_FW_TEE;
2141 set_match(args, f_pos, chain);
2142 args->rule.info = IP_FW_ARG_TABLEARG(cmd->arg1);
2146 IPFW_INC_RULE_COUNTER(f, pktlen);
2147 l = 0; /* exit inner loop */
2151 IPFW_INC_RULE_COUNTER(f, pktlen);
2152 f_pos = jump_fast(chain, f, cmd->arg1, tablearg, 0);
2154 * Skip disabled rules, and re-enter
2155 * the inner loop with the correct
2156 * f_pos, f, l and cmd.
2157 * Also clear cmdlen and skip_or
2159 for (; f_pos < chain->n_rules - 1 &&
2161 (1 << chain->map[f_pos]->set));
2164 /* Re-enter the inner loop at the skipto rule. */
2165 f = chain->map[f_pos];
2172 break; /* not reached */
2174 case O_CALLRETURN: {
2176 * Implementation of `subroutine' call/return,
2177 * in the stack carried in an mbuf tag. This
2178 * is different from `skipto' in that any call
2179 * address is possible (`skipto' must prevent
2180 * backward jumps to avoid endless loops).
2181 * We have `return' action when F_NOT flag is
2182 * present. The `m_tag_id' field is used as
2186 uint16_t jmpto, *stack;
2188 #define IS_CALL ((cmd->len & F_NOT) == 0)
2189 #define IS_RETURN ((cmd->len & F_NOT) != 0)
2191 * Hand-rolled version of m_tag_locate() with
2193 * If not already tagged, allocate new tag.
2195 mtag = m_tag_first(m);
2196 while (mtag != NULL) {
2197 if (mtag->m_tag_cookie ==
2200 mtag = m_tag_next(m, mtag);
2202 if (mtag == NULL && IS_CALL) {
2203 mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2204 IPFW_CALLSTACK_SIZE *
2205 sizeof(uint16_t), M_NOWAIT);
2207 m_tag_prepend(m, mtag);
2211 * On error both `call' and `return' just
2212 * continue with next rule.
2214 if (IS_RETURN && (mtag == NULL ||
2215 mtag->m_tag_id == 0)) {
2216 l = 0; /* exit inner loop */
2219 if (IS_CALL && (mtag == NULL ||
2220 mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2221 printf("ipfw: call stack error, "
2222 "go to next rule\n");
2223 l = 0; /* exit inner loop */
2227 IPFW_INC_RULE_COUNTER(f, pktlen);
2228 stack = (uint16_t *)(mtag + 1);
2231 * The `call' action may use cached f_pos
2232 * (in f->next_rule), whose version is written
2234 * The `return' action, however, doesn't have
2235 * fixed jump address in cmd->arg1 and can't use
2239 stack[mtag->m_tag_id] = f->rulenum;
2241 f_pos = jump_fast(chain, f, cmd->arg1,
2243 } else { /* `return' action */
2245 jmpto = stack[mtag->m_tag_id] + 1;
2246 f_pos = ipfw_find_rule(chain, jmpto, 0);
2250 * Skip disabled rules, and re-enter
2251 * the inner loop with the correct
2252 * f_pos, f, l and cmd.
2253 * Also clear cmdlen and skip_or
2255 for (; f_pos < chain->n_rules - 1 &&
2257 (1 << chain->map[f_pos]->set)); f_pos++)
2259 /* Re-enter the inner loop at the dest rule. */
2260 f = chain->map[f_pos];
2266 break; /* NOTREACHED */
2273 * Drop the packet and send a reject notice
2274 * if the packet is not ICMP (or is an ICMP
2275 * query), and it is not multicast/broadcast.
2277 if (hlen > 0 && is_ipv4 && offset == 0 &&
2278 (proto != IPPROTO_ICMP ||
2279 is_icmp_query(ICMP(ulp))) &&
2280 !(m->m_flags & (M_BCAST|M_MCAST)) &&
2281 !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2282 send_reject(args, cmd->arg1, iplen, ip);
2288 if (hlen > 0 && is_ipv6 &&
2289 ((offset & IP6F_OFF_MASK) == 0) &&
2290 (proto != IPPROTO_ICMPV6 ||
2291 (is_icmp6_query(icmp6_type) == 1)) &&
2292 !(m->m_flags & (M_BCAST|M_MCAST)) &&
2293 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
2295 args, cmd->arg1, hlen,
2296 (struct ip6_hdr *)ip);
2302 retval = IP_FW_DENY;
2303 l = 0; /* exit inner loop */
2304 done = 1; /* exit outer loop */
2308 if (args->eh) /* not valid on layer2 pkts */
2310 if (q == NULL || q->rule != f ||
2311 dyn_dir == MATCH_FORWARD) {
2312 struct sockaddr_in *sa;
2313 sa = &(((ipfw_insn_sa *)cmd)->sa);
2314 if (sa->sin_addr.s_addr == INADDR_ANY) {
2315 bcopy(sa, &args->hopstore,
2317 args->hopstore.sin_addr.s_addr =
2319 args->next_hop = &args->hopstore;
2321 args->next_hop = sa;
2324 retval = IP_FW_PASS;
2325 l = 0; /* exit inner loop */
2326 done = 1; /* exit outer loop */
2331 if (args->eh) /* not valid on layer2 pkts */
2333 if (q == NULL || q->rule != f ||
2334 dyn_dir == MATCH_FORWARD) {
2335 struct sockaddr_in6 *sin6;
2337 sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2338 args->next_hop6 = sin6;
2340 retval = IP_FW_PASS;
2341 l = 0; /* exit inner loop */
2342 done = 1; /* exit outer loop */
2348 set_match(args, f_pos, chain);
2349 args->rule.info = IP_FW_ARG_TABLEARG(cmd->arg1);
2351 args->rule.info |= IPFW_ONEPASS;
2352 retval = (cmd->opcode == O_NETGRAPH) ?
2353 IP_FW_NETGRAPH : IP_FW_NGTEE;
2354 l = 0; /* exit inner loop */
2355 done = 1; /* exit outer loop */
2361 IPFW_INC_RULE_COUNTER(f, pktlen);
2362 fib = IP_FW_ARG_TABLEARG(cmd->arg1);
2363 if (fib >= rt_numfibs)
2366 args->f_id.fib = fib;
2367 l = 0; /* exit inner loop */
2374 code = IP_FW_ARG_TABLEARG(cmd->arg1) & 0x3F;
2375 l = 0; /* exit inner loop */
2380 ip->ip_tos = (code << 2) | (ip->ip_tos & 0x03);
2381 a += ntohs(ip->ip_sum) - ip->ip_tos;
2382 ip->ip_sum = htons(a);
2383 } else if (is_ipv6) {
2386 v = &((struct ip6_hdr *)ip)->ip6_vfc;
2387 *v = (*v & 0xF0) | (code >> 2);
2389 *v = (*v & 0x3F) | ((code & 0x03) << 6);
2393 IPFW_INC_RULE_COUNTER(f, pktlen);
2398 l = 0; /* exit inner loop */
2399 done = 1; /* exit outer loop */
2400 if (!IPFW_NAT_LOADED) {
2401 retval = IP_FW_DENY;
2408 set_match(args, f_pos, chain);
2409 /* Check if this is 'global' nat rule */
2410 if (cmd->arg1 == 0) {
2411 retval = ipfw_nat_ptr(args, NULL, m);
2414 t = ((ipfw_insn_nat *)cmd)->nat;
2416 nat_id = IP_FW_ARG_TABLEARG(cmd->arg1);
2417 t = (*lookup_nat_ptr)(&chain->nat, nat_id);
2420 retval = IP_FW_DENY;
2423 if (cmd->arg1 != IP_FW_TABLEARG)
2424 ((ipfw_insn_nat *)cmd)->nat = t;
2426 retval = ipfw_nat_ptr(args, t, m);
2432 IPFW_INC_RULE_COUNTER(f, pktlen);
2433 l = 0; /* in any case exit inner loop */
2434 ip_off = ntohs(ip->ip_off);
2436 /* if not fragmented, go to next rule */
2437 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
2440 * ip_reass() expects len & off in host
2445 args->m = m = ip_reass(m);
2448 * do IP header checksum fixup.
2450 if (m == NULL) { /* fragment got swallowed */
2451 retval = IP_FW_DENY;
2452 } else { /* good, packet complete */
2455 ip = mtod(m, struct ip *);
2456 hlen = ip->ip_hl << 2;
2459 if (hlen == sizeof(struct ip))
2460 ip->ip_sum = in_cksum_hdr(ip);
2462 ip->ip_sum = in_cksum(m, hlen);
2463 retval = IP_FW_REASS;
2464 set_match(args, f_pos, chain);
2466 done = 1; /* exit outer loop */
2471 panic("-- unknown opcode %d\n", cmd->opcode);
2472 } /* end of switch() on opcodes */
2474 * if we get here with l=0, then match is irrelevant.
2477 if (cmd->len & F_NOT)
2481 if (cmd->len & F_OR)
2484 if (!(cmd->len & F_OR)) /* not an OR block, */
2485 break; /* try next rule */
2488 } /* end of inner loop, scan opcodes */
2494 /* next_rule:; */ /* try next rule */
2496 } /* end of outer for, scan rules */
2499 struct ip_fw *rule = chain->map[f_pos];
2500 /* Update statistics */
2501 IPFW_INC_RULE_COUNTER(rule, pktlen);
2503 retval = IP_FW_DENY;
2504 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
2506 IPFW_RUNLOCK(chain);
2508 if (ucred_cache != NULL)
2509 crfree(ucred_cache);
2515 printf("ipfw: pullup failed\n");
2516 return (IP_FW_DENY);
2520 * Set maximum number of tables that can be used in given VNET ipfw instance.
2524 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
2527 unsigned int ntables;
2529 ntables = V_fw_tables_max;
2531 error = sysctl_handle_int(oidp, &ntables, 0, req);
2532 /* Read operation or some error */
2533 if ((error != 0) || (req->newptr == NULL))
2536 return (ipfw_resize_tables(&V_layer3_chain, ntables));
2540 * Module and VNET glue
2544 * Stuff that must be initialised only on boot or module load
2552 * Only print out this stuff the first time around,
2553 * when called from the sysinit code.
2559 "initialized, divert %s, nat %s, "
2560 "default to %s, logging ",
2566 #ifdef IPFIREWALL_NAT
2571 default_to_accept ? "accept" : "deny");
2574 * Note: V_xxx variables can be accessed here but the vnet specific
2575 * initializer may not have been called yet for the VIMAGE case.
2576 * Tuneables will have been processed. We will print out values for
2578 * XXX This should all be rationalized AFTER 8.0
2580 if (V_fw_verbose == 0)
2581 printf("disabled\n");
2582 else if (V_verbose_limit == 0)
2583 printf("unlimited\n");
2585 printf("limited to %d packets/entry by default\n",
2588 /* Check user-supplied table count for validness */
2589 if (default_fw_tables > IPFW_TABLES_MAX)
2590 default_fw_tables = IPFW_TABLES_MAX;
2592 ipfw_log_bpf(1); /* init */
2597 * Called for the removal of the last instance only on module unload.
2603 ipfw_log_bpf(0); /* uninit */
2604 printf("IP firewall unloaded\n");
2608 * Stuff that must be initialized for every instance
2609 * (including the first of course).
2612 vnet_ipfw_init(const void *unused)
2615 struct ip_fw *rule = NULL;
2616 struct ip_fw_chain *chain;
2618 chain = &V_layer3_chain;
2620 /* First set up some values that are compile time options */
2621 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */
2622 V_fw_deny_unknown_exthdrs = 1;
2623 #ifdef IPFIREWALL_VERBOSE
2626 #ifdef IPFIREWALL_VERBOSE_LIMIT
2627 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
2629 #ifdef IPFIREWALL_NAT
2630 LIST_INIT(&chain->nat);
2633 /* insert the default rule and create the initial map */
2635 chain->static_len = sizeof(struct ip_fw);
2636 chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_WAITOK | M_ZERO);
2638 rule = malloc(chain->static_len, M_IPFW, M_WAITOK | M_ZERO);
2640 /* Set initial number of tables */
2641 V_fw_tables_max = default_fw_tables;
2642 error = ipfw_init_tables(chain);
2644 printf("ipfw2: setting up tables failed\n");
2645 free(chain->map, M_IPFW);
2650 /* fill and insert the default rule */
2652 rule->rulenum = IPFW_DEFAULT_RULE;
2654 rule->set = RESVD_SET;
2655 rule->cmd[0].len = 1;
2656 rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
2657 chain->default_rule = chain->map[0] = rule;
2658 chain->id = rule->id = 1;
2660 IPFW_LOCK_INIT(chain);
2661 ipfw_dyn_init(chain);
2663 /* First set up some values that are compile time options */
2664 V_ipfw_vnet_ready = 1; /* Open for business */
2667 * Hook the sockopt handler, and the layer2 (V_ip_fw_chk_ptr)
2668 * and pfil hooks for ipv4 and ipv6. Even if the latter two fail
2669 * we still keep the module alive because the sockopt and
2670 * layer2 paths are still useful.
2671 * ipfw[6]_hook return 0 on success, ENOENT on failure,
2672 * so we can ignore the exact return value and just set a flag.
2674 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
2675 * changes in the underlying (per-vnet) variables trigger
2676 * immediate hook()/unhook() calls.
2677 * In layer2 we have the same behaviour, except that V_ether_ipfw
2678 * is checked on each packet because there are no pfil hooks.
2680 V_ip_fw_ctl_ptr = ipfw_ctl;
2681 V_ip_fw_chk_ptr = ipfw_chk;
2682 error = ipfw_attach_hooks(1);
2687 * Called for the removal of each instance.
2690 vnet_ipfw_uninit(const void *unused)
2692 struct ip_fw *reap, *rule;
2693 struct ip_fw_chain *chain = &V_layer3_chain;
2696 V_ipfw_vnet_ready = 0; /* tell new callers to go away */
2698 * disconnect from ipv4, ipv6, layer2 and sockopt.
2699 * Then grab, release and grab again the WLOCK so we make
2700 * sure the update is propagated and nobody will be in.
2702 (void)ipfw_attach_hooks(0 /* detach */);
2703 V_ip_fw_chk_ptr = NULL;
2704 V_ip_fw_ctl_ptr = NULL;
2705 IPFW_UH_WLOCK(chain);
2706 IPFW_UH_WUNLOCK(chain);
2707 IPFW_UH_WLOCK(chain);
2710 ipfw_dyn_uninit(0); /* run the callout_drain */
2711 IPFW_WUNLOCK(chain);
2713 ipfw_destroy_tables(chain);
2716 for (i = 0; i < chain->n_rules; i++) {
2717 rule = chain->map[i];
2718 rule->x_next = reap;
2722 free(chain->map, M_IPFW);
2723 IPFW_WUNLOCK(chain);
2724 IPFW_UH_WUNLOCK(chain);
2726 ipfw_reap_rules(reap);
2727 IPFW_LOCK_DESTROY(chain);
2728 ipfw_dyn_uninit(1); /* free the remaining parts */
2733 * Module event handler.
2734 * In general we have the choice of handling most of these events by the
2735 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
2736 * use the SYSINIT handlers as they are more capable of expressing the
2737 * flow of control during module and vnet operations, so this is just
2738 * a skeleton. Note there is no SYSINIT equivalent of the module
2739 * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
2742 ipfw_modevent(module_t mod, int type, void *unused)
2748 /* Called once at module load or
2749 * system boot if compiled in. */
2752 /* Called before unload. May veto unloading. */
2755 /* Called during unload. */
2758 /* Called during system shutdown. */
2767 static moduledata_t ipfwmod = {
2773 /* Define startup order. */
2774 #define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_IFATTACHDOMAIN
2775 #define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */
2776 #define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */
2777 #define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */
2779 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
2780 MODULE_VERSION(ipfw, 2);
2781 /* should declare some dependencies here */
2784 * Starting up. Done in order after ipfwmod() has been called.
2785 * VNET_SYSINIT is also called for each existing vnet and each new vnet.
2787 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2789 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2790 vnet_ipfw_init, NULL);
2793 * Closing up shop. These are done in REVERSE ORDER, but still
2794 * after ipfwmod() has been called. Not called on reboot.
2795 * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
2796 * or when the module is unloaded.
2798 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2799 ipfw_destroy, NULL);
2800 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2801 vnet_ipfw_uninit, NULL);