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>
66 #include <netinet/in.h>
67 #include <netinet/in_var.h>
68 #include <netinet/in_pcb.h>
69 #include <netinet/ip.h>
70 #include <netinet/ip_var.h>
71 #include <netinet/ip_icmp.h>
72 #include <netinet/ip_fw.h>
73 #include <netinet/ipfw/ip_fw_private.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/scope6_var.h>
85 #include <netinet6/ip6_var.h>
88 #include <machine/in_cksum.h> /* XXX for in_cksum */
91 #include <security/mac/mac_framework.h>
95 * static variables followed by global ones.
96 * All ipfw global variables are here.
99 /* ipfw_vnet_ready controls when we are open for business */
100 static VNET_DEFINE(int, ipfw_vnet_ready) = 0;
101 #define V_ipfw_vnet_ready VNET(ipfw_vnet_ready)
103 static VNET_DEFINE(int, fw_deny_unknown_exthdrs);
104 #define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs)
106 static VNET_DEFINE(int, fw_permit_single_frag6) = 1;
107 #define V_fw_permit_single_frag6 VNET(fw_permit_single_frag6)
109 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
110 static int default_to_accept = 1;
112 static int default_to_accept;
115 VNET_DEFINE(int, autoinc_step);
118 * Each rule belongs to one of 32 different sets (0..31).
119 * The variable set_disable contains one bit per set.
120 * If the bit is set, all rules in the corresponding set
121 * are disabled. Set RESVD_SET(31) is reserved for the default rule
122 * and rules that are not deleted by the flush command,
123 * and CANNOT be disabled.
124 * Rules in set RESVD_SET can only be deleted individually.
126 VNET_DEFINE(u_int32_t, set_disable);
127 #define V_set_disable VNET(set_disable)
129 VNET_DEFINE(int, fw_verbose);
130 /* counter for ipfw_log(NULL...) */
131 VNET_DEFINE(u_int64_t, norule_counter);
132 VNET_DEFINE(int, verbose_limit);
134 /* layer3_chain contains the list of rules for layer 3 */
135 VNET_DEFINE(struct ip_fw_chain, layer3_chain);
137 ipfw_nat_t *ipfw_nat_ptr = NULL;
138 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
139 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
140 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
141 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
142 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
145 uint32_t dummy_def = IPFW_DEFAULT_RULE;
146 uint32_t dummy_tables_max = IPFW_TABLES_MAX;
150 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
151 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
152 CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
153 "Only do a single pass through ipfw when using dummynet(4)");
154 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
155 CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
156 "Rule number auto-increment step");
157 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose,
158 CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
159 "Log matches to ipfw rules");
160 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
161 CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
162 "Set upper limit of matches of ipfw rules logged");
163 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
165 "The default/max possible rule number.");
166 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, tables_max, CTLFLAG_RD,
167 &dummy_tables_max, 0,
168 "The maximum number of tables.");
169 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
170 &default_to_accept, 0,
171 "Make the default rule accept all packets.");
172 TUNABLE_INT("net.inet.ip.fw.default_to_accept", &default_to_accept);
173 SYSCTL_VNET_INT(_net_inet_ip_fw, OID_AUTO, static_count,
174 CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
175 "Number of static rules");
178 SYSCTL_DECL(_net_inet6_ip6);
179 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
180 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
181 CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_deny_unknown_exthdrs), 0,
182 "Deny packets with unknown IPv6 Extension Headers");
183 SYSCTL_VNET_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
184 CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(fw_permit_single_frag6), 0,
185 "Permit single packet IPv6 fragments");
190 #endif /* SYSCTL_NODE */
194 * Some macros used in the various matching options.
195 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
196 * Other macros just cast void * into the appropriate type
198 #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
199 #define TCP(p) ((struct tcphdr *)(p))
200 #define SCTP(p) ((struct sctphdr *)(p))
201 #define UDP(p) ((struct udphdr *)(p))
202 #define ICMP(p) ((struct icmphdr *)(p))
203 #define ICMP6(p) ((struct icmp6_hdr *)(p))
206 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
208 int type = icmp->icmp_type;
210 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
213 #define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
214 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
217 is_icmp_query(struct icmphdr *icmp)
219 int type = icmp->icmp_type;
221 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
226 * The following checks use two arrays of 8 or 16 bits to store the
227 * bits that we want set or clear, respectively. They are in the
228 * low and high half of cmd->arg1 or cmd->d[0].
230 * We scan options and store the bits we find set. We succeed if
232 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
234 * The code is sometimes optimized not to store additional variables.
238 flags_match(ipfw_insn *cmd, u_int8_t bits)
243 if ( ((cmd->arg1 & 0xff) & bits) != 0)
244 return 0; /* some bits we want set were clear */
245 want_clear = (cmd->arg1 >> 8) & 0xff;
246 if ( (want_clear & bits) != want_clear)
247 return 0; /* some bits we want clear were set */
252 ipopts_match(struct ip *ip, ipfw_insn *cmd)
254 int optlen, bits = 0;
255 u_char *cp = (u_char *)(ip + 1);
256 int x = (ip->ip_hl << 2) - sizeof (struct ip);
258 for (; x > 0; x -= optlen, cp += optlen) {
259 int opt = cp[IPOPT_OPTVAL];
261 if (opt == IPOPT_EOL)
263 if (opt == IPOPT_NOP)
266 optlen = cp[IPOPT_OLEN];
267 if (optlen <= 0 || optlen > x)
268 return 0; /* invalid or truncated */
276 bits |= IP_FW_IPOPT_LSRR;
280 bits |= IP_FW_IPOPT_SSRR;
284 bits |= IP_FW_IPOPT_RR;
288 bits |= IP_FW_IPOPT_TS;
292 return (flags_match(cmd, bits));
296 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
298 int optlen, bits = 0;
299 u_char *cp = (u_char *)(tcp + 1);
300 int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
302 for (; x > 0; x -= optlen, cp += optlen) {
304 if (opt == TCPOPT_EOL)
306 if (opt == TCPOPT_NOP)
320 bits |= IP_FW_TCPOPT_MSS;
324 bits |= IP_FW_TCPOPT_WINDOW;
327 case TCPOPT_SACK_PERMITTED:
329 bits |= IP_FW_TCPOPT_SACK;
332 case TCPOPT_TIMESTAMP:
333 bits |= IP_FW_TCPOPT_TS;
338 return (flags_match(cmd, bits));
342 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd)
344 if (ifp == NULL) /* no iface with this packet, match fails */
346 /* Check by name or by IP address */
347 if (cmd->name[0] != '\0') { /* match by name */
350 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
353 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
357 #ifdef __FreeBSD__ /* and OSX too ? */
361 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
362 if (ia->ifa_addr->sa_family != AF_INET)
364 if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
365 (ia->ifa_addr))->sin_addr.s_addr) {
366 if_addr_runlock(ifp);
367 return(1); /* match */
370 if_addr_runlock(ifp);
371 #endif /* __FreeBSD__ */
373 return(0); /* no match, fail ... */
377 * The verify_path function checks if a route to the src exists and
378 * if it is reachable via ifp (when provided).
380 * The 'verrevpath' option checks that the interface that an IP packet
381 * arrives on is the same interface that traffic destined for the
382 * packet's source address would be routed out of.
383 * The 'versrcreach' option just checks that the source address is
384 * reachable via any route (except default) in the routing table.
385 * These two are a measure to block forged packets. This is also
386 * commonly known as "anti-spoofing" or Unicast Reverse Path
387 * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
388 * is purposely reminiscent of the Cisco IOS command,
390 * ip verify unicast reverse-path
391 * ip verify unicast source reachable-via any
393 * which implements the same functionality. But note that the syntax
394 * is misleading, and the check may be performed on all IP packets
395 * whether unicast, multicast, or broadcast.
398 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
404 struct sockaddr_in *dst;
406 bzero(&ro, sizeof(ro));
408 dst = (struct sockaddr_in *)&(ro.ro_dst);
409 dst->sin_family = AF_INET;
410 dst->sin_len = sizeof(*dst);
412 in_rtalloc_ign(&ro, 0, fib);
414 if (ro.ro_rt == NULL)
418 * If ifp is provided, check for equality with rtentry.
419 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
420 * in order to pass packets injected back by if_simloop():
421 * if useloopback == 1 routing entry (via lo0) for our own address
422 * may exist, so we need to handle routing assymetry.
424 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
429 /* if no ifp provided, check if rtentry is not default route */
431 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) {
436 /* or if this is a blackhole/reject route */
437 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
442 /* found valid route */
445 #endif /* __FreeBSD__ */
450 * ipv6 specific rules here...
453 icmp6type_match (int type, ipfw_insn_u32 *cmd)
455 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
459 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
462 for (i=0; i <= cmd->o.arg1; ++i )
463 if (curr_flow == cmd->d[i] )
468 /* support for IP6_*_ME opcodes */
470 search_ip6_addr_net (struct in6_addr * ip6_addr)
474 struct in6_ifaddr *fdm;
475 struct in6_addr copia;
477 TAILQ_FOREACH(mdc, &V_ifnet, if_link) {
479 TAILQ_FOREACH(mdc2, &mdc->if_addrhead, ifa_link) {
480 if (mdc2->ifa_addr->sa_family == AF_INET6) {
481 fdm = (struct in6_ifaddr *)mdc2;
482 copia = fdm->ia_addr.sin6_addr;
483 /* need for leaving scope_id in the sock_addr */
484 in6_clearscope(&copia);
485 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia)) {
486 if_addr_runlock(mdc);
491 if_addr_runlock(mdc);
497 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
500 struct sockaddr_in6 *dst;
502 bzero(&ro, sizeof(ro));
504 dst = (struct sockaddr_in6 * )&(ro.ro_dst);
505 dst->sin6_family = AF_INET6;
506 dst->sin6_len = sizeof(*dst);
507 dst->sin6_addr = *src;
509 in6_rtalloc_ign(&ro, 0, fib);
510 if (ro.ro_rt == NULL)
514 * if ifp is provided, check for equality with rtentry
515 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
516 * to support the case of sending packets to an address of our own.
517 * (where the former interface is the first argument of if_simloop()
518 * (=ifp), the latter is lo0)
520 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
525 /* if no ifp provided, check if rtentry is not default route */
527 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) {
532 /* or if this is a blackhole/reject route */
533 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
538 /* found valid route */
545 is_icmp6_query(int icmp6_type)
547 if ((icmp6_type <= ICMP6_MAXTYPE) &&
548 (icmp6_type == ICMP6_ECHO_REQUEST ||
549 icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
550 icmp6_type == ICMP6_WRUREQUEST ||
551 icmp6_type == ICMP6_FQDN_QUERY ||
552 icmp6_type == ICMP6_NI_QUERY))
559 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
564 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
566 tcp = (struct tcphdr *)((char *)ip6 + hlen);
568 if ((tcp->th_flags & TH_RST) == 0) {
570 m0 = ipfw_send_pkt(args->m, &(args->f_id),
571 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
572 tcp->th_flags | TH_RST);
574 ip6_output(m0, NULL, NULL, 0, NULL, NULL,
578 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
581 * Unlike above, the mbufs need to line up with the ip6 hdr,
582 * as the contents are read. We need to m_adj() the
584 * The mbuf will however be thrown away so we can adjust it.
585 * Remember we did an m_pullup on it already so we
586 * can make some assumptions about contiguousness.
589 m_adj(m, args->L3offset);
591 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
602 * sends a reject message, consuming the mbuf passed as an argument.
605 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
609 /* XXX When ip is not guaranteed to be at mtod() we will
610 * need to account for this */
611 * The mbuf will however be thrown away so we can adjust it.
612 * Remember we did an m_pullup on it already so we
613 * can make some assumptions about contiguousness.
616 m_adj(m, args->L3offset);
618 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
619 /* We need the IP header in host order for icmp_error(). */
621 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
622 } else if (args->f_id.proto == IPPROTO_TCP) {
623 struct tcphdr *const tcp =
624 L3HDR(struct tcphdr, mtod(args->m, struct ip *));
625 if ( (tcp->th_flags & TH_RST) == 0) {
627 m = ipfw_send_pkt(args->m, &(args->f_id),
628 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
629 tcp->th_flags | TH_RST);
631 ip_output(m, NULL, NULL, 0, NULL, NULL);
640 * Support for uid/gid/jail lookup. These tests are expensive
641 * (because we may need to look into the list of active sockets)
642 * so we cache the results. ugid_lookupp is 0 if we have not
643 * yet done a lookup, 1 if we succeeded, and -1 if we tried
644 * and failed. The function always returns the match value.
645 * We could actually spare the variable and use *uc, setting
646 * it to '(void *)check_uidgid if we have no info, NULL if
647 * we tried and failed, or any other value if successful.
650 check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif,
651 struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip,
652 u_int16_t src_port, int *ugid_lookupp,
653 struct ucred **uc, struct inpcb *inp)
656 return cred_check(insn, proto, oif,
657 dst_ip, dst_port, src_ip, src_port,
658 (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
660 struct inpcbinfo *pi;
666 * Check to see if the UDP or TCP stack supplied us with
667 * the PCB. If so, rather then holding a lock and looking
668 * up the PCB, we can use the one that was supplied.
670 if (inp && *ugid_lookupp == 0) {
671 INP_LOCK_ASSERT(inp);
672 if (inp->inp_socket != NULL) {
673 *uc = crhold(inp->inp_cred);
679 * If we have already been here and the packet has no
680 * PCB entry associated with it, then we can safely
681 * assume that this is a no match.
683 if (*ugid_lookupp == -1)
685 if (proto == IPPROTO_TCP) {
688 } else if (proto == IPPROTO_UDP) {
689 wildcard = INPLOOKUP_WILDCARD;
694 if (*ugid_lookupp == 0) {
697 in_pcblookup_hash(pi,
698 dst_ip, htons(dst_port),
699 src_ip, htons(src_port),
701 in_pcblookup_hash(pi,
702 src_ip, htons(src_port),
703 dst_ip, htons(dst_port),
706 *uc = crhold(pcb->inp_cred);
709 INP_INFO_RUNLOCK(pi);
710 if (*ugid_lookupp == 0) {
712 * We tried and failed, set the variable to -1
713 * so we will not try again on this packet.
719 if (insn->o.opcode == O_UID)
720 match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
721 else if (insn->o.opcode == O_GID)
722 match = groupmember((gid_t)insn->d[0], *uc);
723 else if (insn->o.opcode == O_JAIL)
724 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
726 #endif /* __FreeBSD__ */
730 * Helper function to set args with info on the rule after the matching
731 * one. slot is precise, whereas we guess rule_id as they are
732 * assigned sequentially.
735 set_match(struct ip_fw_args *args, int slot,
736 struct ip_fw_chain *chain)
738 args->rule.chain_id = chain->id;
739 args->rule.slot = slot + 1; /* we use 0 as a marker */
740 args->rule.rule_id = 1 + chain->map[slot]->id;
741 args->rule.rulenum = chain->map[slot]->rulenum;
745 * The main check routine for the firewall.
747 * All arguments are in args so we can modify them and return them
748 * back to the caller.
752 * args->m (in/out) The packet; we set to NULL when/if we nuke it.
753 * Starts with the IP header.
754 * args->eh (in) Mac header if present, NULL for layer3 packet.
755 * args->L3offset Number of bytes bypassed if we came from L2.
756 * e.g. often sizeof(eh) ** NOTYET **
757 * args->oif Outgoing interface, NULL if packet is incoming.
758 * The incoming interface is in the mbuf. (in)
759 * args->divert_rule (in/out)
760 * Skip up to the first rule past this rule number;
761 * upon return, non-zero port number for divert or tee.
763 * args->rule Pointer to the last matching rule (in/out)
764 * args->next_hop Socket we are forwarding to (out).
765 * args->f_id Addresses grabbed from the packet (out)
766 * args->rule.info a cookie depending on rule action
770 * IP_FW_PASS the packet must be accepted
771 * IP_FW_DENY the packet must be dropped
772 * IP_FW_DIVERT divert packet, port in m_tag
773 * IP_FW_TEE tee packet, port in m_tag
774 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie
775 * IP_FW_NETGRAPH into netgraph, cookie args->cookie
776 * args->rule contains the matching rule,
777 * args->rule.info has additional information.
781 ipfw_chk(struct ip_fw_args *args)
785 * Local variables holding state while processing a packet:
787 * IMPORTANT NOTE: to speed up the processing of rules, there
788 * are some assumption on the values of the variables, which
789 * are documented here. Should you change them, please check
790 * the implementation of the various instructions to make sure
791 * that they still work.
793 * args->eh The MAC header. It is non-null for a layer2
794 * packet, it is NULL for a layer-3 packet.
796 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
798 * m | args->m Pointer to the mbuf, as received from the caller.
799 * It may change if ipfw_chk() does an m_pullup, or if it
800 * consumes the packet because it calls send_reject().
801 * XXX This has to change, so that ipfw_chk() never modifies
802 * or consumes the buffer.
803 * ip is the beginning of the ip(4 or 6) header.
804 * Calculated by adding the L3offset to the start of data.
805 * (Until we start using L3offset, the packet is
806 * supposed to start with the ip header).
808 struct mbuf *m = args->m;
809 struct ip *ip = mtod(m, struct ip *);
812 * For rules which contain uid/gid or jail constraints, cache
813 * a copy of the users credentials after the pcb lookup has been
814 * executed. This will speed up the processing of rules with
815 * these types of constraints, as well as decrease contention
816 * on pcb related locks.
819 struct bsd_ucred ucred_cache;
821 struct ucred *ucred_cache = NULL;
823 int ucred_lookup = 0;
826 * oif | args->oif If NULL, ipfw_chk has been called on the
827 * inbound path (ether_input, ip_input).
828 * If non-NULL, ipfw_chk has been called on the outbound path
829 * (ether_output, ip_output).
831 struct ifnet *oif = args->oif;
833 int f_pos = 0; /* index of current rule in the array */
837 * hlen The length of the IP header.
839 u_int hlen = 0; /* hlen >0 means we have an IP pkt */
842 * offset The offset of a fragment. offset != 0 means that
843 * we have a fragment at this offset of an IPv4 packet.
844 * offset == 0 means that (if this is an IPv4 packet)
845 * this is the first or only fragment.
846 * For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
847 * or there is a single packet fragement (fragement header added
848 * without needed). We will treat a single packet fragment as if
849 * there was no fragment header (or log/block depending on the
850 * V_fw_permit_single_frag6 sysctl setting).
856 * Local copies of addresses. They are only valid if we have
859 * proto The protocol. Set to 0 for non-ip packets,
860 * or to the protocol read from the packet otherwise.
861 * proto != 0 means that we have an IPv4 packet.
863 * src_port, dst_port port numbers, in HOST format. Only
864 * valid for TCP and UDP packets.
866 * src_ip, dst_ip ip addresses, in NETWORK format.
867 * Only valid for IPv4 packets.
870 uint16_t src_port = 0, dst_port = 0; /* NOTE: host format */
871 struct in_addr src_ip, dst_ip; /* NOTE: network format */
874 uint16_t etype = 0; /* Host order stored ether type */
877 * dyn_dir = MATCH_UNKNOWN when rules unchecked,
878 * MATCH_NONE when checked and not matched (q = NULL),
879 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
881 int dyn_dir = MATCH_UNKNOWN;
882 ipfw_dyn_rule *q = NULL;
883 struct ip_fw_chain *chain = &V_layer3_chain;
886 * We store in ulp a pointer to the upper layer protocol header.
887 * In the ipv4 case this is easy to determine from the header,
888 * but for ipv6 we might have some additional headers in the middle.
889 * ulp is NULL if not found.
891 void *ulp = NULL; /* upper layer protocol pointer. */
893 /* XXX ipv6 variables */
895 uint8_t icmp6_type = 0;
896 uint16_t ext_hd = 0; /* bits vector for extension header filtering */
897 /* end of ipv6 variables */
901 int done = 0; /* flag to exit the outer loop */
903 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
904 return (IP_FW_PASS); /* accept */
906 dst_ip.s_addr = 0; /* make sure it is initialized */
907 src_ip.s_addr = 0; /* make sure it is initialized */
908 pktlen = m->m_pkthdr.len;
909 args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
910 proto = args->f_id.proto = 0; /* mark f_id invalid */
911 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
914 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
915 * then it sets p to point at the offset "len" in the mbuf. WARNING: the
916 * pointer might become stale after other pullups (but we never use it
919 #define PULLUP_TO(_len, p, T) PULLUP_LEN(_len, p, sizeof(T))
920 #define PULLUP_LEN(_len, p, T) \
922 int x = (_len) + T; \
923 if ((m)->m_len < x) { \
924 args->m = m = m_pullup(m, x); \
926 goto pullup_failed; \
928 p = (mtod(m, char *) + (_len)); \
932 * if we have an ether header,
935 etype = ntohs(args->eh->ether_type);
937 /* Identify IP packets and fill up variables. */
938 if (pktlen >= sizeof(struct ip6_hdr) &&
939 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
940 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
942 args->f_id.addr_type = 6;
943 hlen = sizeof(struct ip6_hdr);
944 proto = ip6->ip6_nxt;
946 /* Search extension headers to find upper layer protocols */
947 while (ulp == NULL && offset == 0) {
950 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
951 icmp6_type = ICMP6(ulp)->icmp6_type;
955 PULLUP_TO(hlen, ulp, struct tcphdr);
956 dst_port = TCP(ulp)->th_dport;
957 src_port = TCP(ulp)->th_sport;
958 /* save flags for dynamic rules */
959 args->f_id._flags = TCP(ulp)->th_flags;
963 PULLUP_TO(hlen, ulp, struct sctphdr);
964 src_port = SCTP(ulp)->src_port;
965 dst_port = SCTP(ulp)->dest_port;
969 PULLUP_TO(hlen, ulp, struct udphdr);
970 dst_port = UDP(ulp)->uh_dport;
971 src_port = UDP(ulp)->uh_sport;
974 case IPPROTO_HOPOPTS: /* RFC 2460 */
975 PULLUP_TO(hlen, ulp, struct ip6_hbh);
976 ext_hd |= EXT_HOPOPTS;
977 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
978 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
982 case IPPROTO_ROUTING: /* RFC 2460 */
983 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
984 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
986 ext_hd |= EXT_RTHDR0;
989 ext_hd |= EXT_RTHDR2;
993 printf("IPFW2: IPV6 - Unknown "
994 "Routing Header type(%d)\n",
995 ((struct ip6_rthdr *)
997 if (V_fw_deny_unknown_exthdrs)
1001 ext_hd |= EXT_ROUTING;
1002 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1003 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1007 case IPPROTO_FRAGMENT: /* RFC 2460 */
1008 PULLUP_TO(hlen, ulp, struct ip6_frag);
1009 ext_hd |= EXT_FRAGMENT;
1010 hlen += sizeof (struct ip6_frag);
1011 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1012 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1014 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1016 if (V_fw_permit_single_frag6 == 0 &&
1017 offset == 0 && ip6f_mf == 0) {
1019 printf("IPFW2: IPV6 - Invalid "
1020 "Fragment Header\n");
1021 if (V_fw_deny_unknown_exthdrs)
1022 return (IP_FW_DENY);
1026 ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1030 case IPPROTO_DSTOPTS: /* RFC 2460 */
1031 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1032 ext_hd |= EXT_DSTOPTS;
1033 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1034 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1038 case IPPROTO_AH: /* RFC 2402 */
1039 PULLUP_TO(hlen, ulp, struct ip6_ext);
1041 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1042 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1046 case IPPROTO_ESP: /* RFC 2406 */
1047 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
1048 /* Anything past Seq# is variable length and
1049 * data past this ext. header is encrypted. */
1053 case IPPROTO_NONE: /* RFC 2460 */
1055 * Packet ends here, and IPv6 header has
1056 * already been pulled up. If ip6e_len!=0
1057 * then octets must be ignored.
1059 ulp = ip; /* non-NULL to get out of loop. */
1062 case IPPROTO_OSPFIGP:
1063 /* XXX OSPF header check? */
1064 PULLUP_TO(hlen, ulp, struct ip6_ext);
1068 /* XXX PIM header check? */
1069 PULLUP_TO(hlen, ulp, struct pim);
1073 PULLUP_TO(hlen, ulp, struct carp_header);
1074 if (((struct carp_header *)ulp)->carp_version !=
1076 return (IP_FW_DENY);
1077 if (((struct carp_header *)ulp)->carp_type !=
1079 return (IP_FW_DENY);
1082 case IPPROTO_IPV6: /* RFC 2893 */
1083 PULLUP_TO(hlen, ulp, struct ip6_hdr);
1086 case IPPROTO_IPV4: /* RFC 2893 */
1087 PULLUP_TO(hlen, ulp, struct ip);
1092 printf("IPFW2: IPV6 - Unknown "
1093 "Extension Header(%d), ext_hd=%x\n",
1095 if (V_fw_deny_unknown_exthdrs)
1096 return (IP_FW_DENY);
1097 PULLUP_TO(hlen, ulp, struct ip6_ext);
1101 ip = mtod(m, struct ip *);
1102 ip6 = (struct ip6_hdr *)ip;
1103 args->f_id.src_ip6 = ip6->ip6_src;
1104 args->f_id.dst_ip6 = ip6->ip6_dst;
1105 args->f_id.src_ip = 0;
1106 args->f_id.dst_ip = 0;
1107 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1108 } else if (pktlen >= sizeof(struct ip) &&
1109 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1111 hlen = ip->ip_hl << 2;
1112 args->f_id.addr_type = 4;
1115 * Collect parameters into local variables for faster matching.
1118 src_ip = ip->ip_src;
1119 dst_ip = ip->ip_dst;
1120 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1121 iplen = ntohs(ip->ip_len);
1122 pktlen = iplen < pktlen ? iplen : pktlen;
1127 PULLUP_TO(hlen, ulp, struct tcphdr);
1128 dst_port = TCP(ulp)->th_dport;
1129 src_port = TCP(ulp)->th_sport;
1130 /* save flags for dynamic rules */
1131 args->f_id._flags = TCP(ulp)->th_flags;
1135 PULLUP_TO(hlen, ulp, struct sctphdr);
1136 src_port = SCTP(ulp)->src_port;
1137 dst_port = SCTP(ulp)->dest_port;
1141 PULLUP_TO(hlen, ulp, struct udphdr);
1142 dst_port = UDP(ulp)->uh_dport;
1143 src_port = UDP(ulp)->uh_sport;
1147 PULLUP_TO(hlen, ulp, struct icmphdr);
1148 //args->f_id.flags = ICMP(ulp)->icmp_type;
1156 ip = mtod(m, struct ip *);
1157 args->f_id.src_ip = ntohl(src_ip.s_addr);
1158 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1161 if (proto) { /* we may have port numbers, store them */
1162 args->f_id.proto = proto;
1163 args->f_id.src_port = src_port = ntohs(src_port);
1164 args->f_id.dst_port = dst_port = ntohs(dst_port);
1168 if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1169 IPFW_RUNLOCK(chain);
1170 return (IP_FW_PASS); /* accept */
1172 if (args->rule.slot) {
1174 * Packet has already been tagged as a result of a previous
1175 * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1176 * REASS, NETGRAPH, DIVERT/TEE...)
1177 * Validate the slot and continue from the next one
1178 * if still present, otherwise do a lookup.
1180 f_pos = (args->rule.chain_id == chain->id) ?
1182 ipfw_find_rule(chain, args->rule.rulenum,
1183 args->rule.rule_id);
1189 * Now scan the rules, and parse microinstructions for each rule.
1190 * We have two nested loops and an inner switch. Sometimes we
1191 * need to break out of one or both loops, or re-enter one of
1192 * the loops with updated variables. Loop variables are:
1194 * f_pos (outer loop) points to the current rule.
1195 * On output it points to the matching rule.
1196 * done (outer loop) is used as a flag to break the loop.
1197 * l (inner loop) residual length of current rule.
1198 * cmd points to the current microinstruction.
1200 * We break the inner loop by setting l=0 and possibly
1201 * cmdlen=0 if we don't want to advance cmd.
1202 * We break the outer loop by setting done=1
1203 * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1206 for (; f_pos < chain->n_rules; f_pos++) {
1208 uint32_t tablearg = 0;
1209 int l, cmdlen, skip_or; /* skip rest of OR block */
1212 f = chain->map[f_pos];
1213 if (V_set_disable & (1 << f->set) )
1217 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1218 l -= cmdlen, cmd += cmdlen) {
1222 * check_body is a jump target used when we find a
1223 * CHECK_STATE, and need to jump to the body of
1228 cmdlen = F_LEN(cmd);
1230 * An OR block (insn_1 || .. || insn_n) has the
1231 * F_OR bit set in all but the last instruction.
1232 * The first match will set "skip_or", and cause
1233 * the following instructions to be skipped until
1234 * past the one with the F_OR bit clear.
1236 if (skip_or) { /* skip this instruction */
1237 if ((cmd->len & F_OR) == 0)
1238 skip_or = 0; /* next one is good */
1241 match = 0; /* set to 1 if we succeed */
1243 switch (cmd->opcode) {
1245 * The first set of opcodes compares the packet's
1246 * fields with some pattern, setting 'match' if a
1247 * match is found. At the end of the loop there is
1248 * logic to deal with F_NOT and F_OR flags associated
1256 printf("ipfw: opcode %d unimplemented\n",
1264 * We only check offset == 0 && proto != 0,
1265 * as this ensures that we have a
1266 * packet with the ports info.
1270 if (is_ipv6) /* XXX to be fixed later */
1272 if (proto == IPPROTO_TCP ||
1273 proto == IPPROTO_UDP)
1274 match = check_uidgid(
1275 (ipfw_insn_u32 *)cmd,
1278 src_ip, src_port, &ucred_lookup,
1280 &ucred_cache, args->inp);
1282 (void *)&ucred_cache,
1283 (struct inpcb *)args->m);
1288 match = iface_match(m->m_pkthdr.rcvif,
1289 (ipfw_insn_if *)cmd);
1293 match = iface_match(oif, (ipfw_insn_if *)cmd);
1297 match = iface_match(oif ? oif :
1298 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd);
1302 if (args->eh != NULL) { /* have MAC header */
1303 u_int32_t *want = (u_int32_t *)
1304 ((ipfw_insn_mac *)cmd)->addr;
1305 u_int32_t *mask = (u_int32_t *)
1306 ((ipfw_insn_mac *)cmd)->mask;
1307 u_int32_t *hdr = (u_int32_t *)args->eh;
1310 ( want[0] == (hdr[0] & mask[0]) &&
1311 want[1] == (hdr[1] & mask[1]) &&
1312 want[2] == (hdr[2] & mask[2]) );
1317 if (args->eh != NULL) {
1319 ((ipfw_insn_u16 *)cmd)->ports;
1322 for (i = cmdlen - 1; !match && i>0;
1324 match = (etype >= p[0] &&
1330 match = (offset != 0);
1333 case O_IN: /* "out" is "not in" */
1334 match = (oif == NULL);
1338 match = (args->eh != NULL);
1343 /* For diverted packets, args->rule.info
1344 * contains the divert port (in host format)
1345 * reason and direction.
1347 uint32_t i = args->rule.info;
1348 match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1349 cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1355 * We do not allow an arg of 0 so the
1356 * check of "proto" only suffices.
1358 match = (proto == cmd->arg1);
1363 (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1367 case O_IP_SRC_LOOKUP:
1368 case O_IP_DST_LOOKUP:
1371 (cmd->opcode == O_IP_DST_LOOKUP) ?
1372 dst_ip.s_addr : src_ip.s_addr;
1375 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1376 /* generic lookup. The key must be
1377 * in 32bit big-endian format.
1379 v = ((ipfw_insn_u32 *)cmd)->d[1];
1381 key = dst_ip.s_addr;
1383 key = src_ip.s_addr;
1384 else if (v == 6) /* dscp */
1385 key = (ip->ip_tos >> 2) & 0x3f;
1386 else if (offset != 0)
1388 else if (proto != IPPROTO_TCP &&
1389 proto != IPPROTO_UDP)
1392 key = htonl(dst_port);
1394 key = htonl(src_port);
1395 else if (v == 4 || v == 5) {
1397 (ipfw_insn_u32 *)cmd,
1400 src_ip, src_port, &ucred_lookup,
1402 &ucred_cache, args->inp);
1403 if (v == 4 /* O_UID */)
1404 key = ucred_cache->cr_uid;
1405 else if (v == 5 /* O_JAIL */)
1406 key = ucred_cache->cr_prison->pr_id;
1407 #else /* !__FreeBSD__ */
1408 (void *)&ucred_cache,
1409 (struct inpcb *)args->m);
1410 if (v ==4 /* O_UID */)
1411 key = ucred_cache.uid;
1412 else if (v == 5 /* O_JAIL */)
1413 key = ucred_cache.xid;
1414 #endif /* !__FreeBSD__ */
1419 match = ipfw_lookup_table(chain,
1420 cmd->arg1, key, &v);
1423 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
1425 ((ipfw_insn_u32 *)cmd)->d[0] == v;
1435 (cmd->opcode == O_IP_DST_MASK) ?
1436 dst_ip.s_addr : src_ip.s_addr;
1437 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
1440 for (; !match && i>0; i-= 2, p+= 2)
1441 match = (p[0] == (a & p[1]));
1449 INADDR_TO_IFP(src_ip, tif);
1450 match = (tif != NULL);
1456 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6);
1463 u_int32_t *d = (u_int32_t *)(cmd+1);
1465 cmd->opcode == O_IP_DST_SET ?
1471 addr -= d[0]; /* subtract base */
1472 match = (addr < cmd->arg1) &&
1473 ( d[ 1 + (addr>>5)] &
1474 (1<<(addr & 0x1f)) );
1480 (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1488 INADDR_TO_IFP(dst_ip, tif);
1489 match = (tif != NULL);
1495 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6);
1503 * offset == 0 && proto != 0 is enough
1504 * to guarantee that we have a
1505 * packet with port info.
1507 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
1510 (cmd->opcode == O_IP_SRCPORT) ?
1511 src_port : dst_port ;
1513 ((ipfw_insn_u16 *)cmd)->ports;
1516 for (i = cmdlen - 1; !match && i>0;
1518 match = (x>=p[0] && x<=p[1]);
1523 match = (offset == 0 && proto==IPPROTO_ICMP &&
1524 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
1529 match = is_ipv6 && offset == 0 &&
1530 proto==IPPROTO_ICMPV6 &&
1532 ICMP6(ulp)->icmp6_type,
1533 (ipfw_insn_u32 *)cmd);
1539 ipopts_match(ip, cmd) );
1544 cmd->arg1 == ip->ip_v);
1550 if (is_ipv4) { /* only for IP packets */
1555 if (cmd->opcode == O_IPLEN)
1557 else if (cmd->opcode == O_IPTTL)
1559 else /* must be IPID */
1560 x = ntohs(ip->ip_id);
1562 match = (cmd->arg1 == x);
1565 /* otherwise we have ranges */
1566 p = ((ipfw_insn_u16 *)cmd)->ports;
1568 for (; !match && i>0; i--, p += 2)
1569 match = (x >= p[0] && x <= p[1]);
1573 case O_IPPRECEDENCE:
1575 (cmd->arg1 == (ip->ip_tos & 0xe0)) );
1580 flags_match(cmd, ip->ip_tos));
1584 if (proto == IPPROTO_TCP && offset == 0) {
1592 ((ip->ip_hl + tcp->th_off) << 2);
1594 match = (cmd->arg1 == x);
1597 /* otherwise we have ranges */
1598 p = ((ipfw_insn_u16 *)cmd)->ports;
1600 for (; !match && i>0; i--, p += 2)
1601 match = (x >= p[0] && x <= p[1]);
1606 match = (proto == IPPROTO_TCP && offset == 0 &&
1607 flags_match(cmd, TCP(ulp)->th_flags));
1611 PULLUP_LEN(hlen, ulp, (TCP(ulp)->th_off << 2));
1612 match = (proto == IPPROTO_TCP && offset == 0 &&
1613 tcpopts_match(TCP(ulp), cmd));
1617 match = (proto == IPPROTO_TCP && offset == 0 &&
1618 ((ipfw_insn_u32 *)cmd)->d[0] ==
1623 match = (proto == IPPROTO_TCP && offset == 0 &&
1624 ((ipfw_insn_u32 *)cmd)->d[0] ==
1629 match = (proto == IPPROTO_TCP && offset == 0 &&
1630 cmd->arg1 == TCP(ulp)->th_win);
1634 /* reject packets which have SYN only */
1635 /* XXX should i also check for TH_ACK ? */
1636 match = (proto == IPPROTO_TCP && offset == 0 &&
1637 (TCP(ulp)->th_flags &
1638 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
1643 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
1646 at = pf_find_mtag(m);
1647 if (at != NULL && at->qid != 0)
1649 at = pf_get_mtag(m);
1652 * Let the packet fall back to the
1657 at->qid = altq->qid;
1667 ipfw_log(f, hlen, args, m,
1668 oif, offset | ip6f_mf, tablearg, ip);
1673 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
1677 /* Outgoing packets automatically pass/match */
1678 match = ((oif != NULL) ||
1679 (m->m_pkthdr.rcvif == NULL) ||
1683 verify_path6(&(args->f_id.src_ip6),
1684 m->m_pkthdr.rcvif, args->f_id.fib) :
1686 verify_path(src_ip, m->m_pkthdr.rcvif,
1691 /* Outgoing packets automatically pass/match */
1692 match = (hlen > 0 && ((oif != NULL) ||
1695 verify_path6(&(args->f_id.src_ip6),
1696 NULL, args->f_id.fib) :
1698 verify_path(src_ip, NULL, args->f_id.fib)));
1702 /* Outgoing packets automatically pass/match */
1703 if (oif == NULL && hlen > 0 &&
1704 ( (is_ipv4 && in_localaddr(src_ip))
1707 in6_localaddr(&(args->f_id.src_ip6)))
1712 is_ipv6 ? verify_path6(
1713 &(args->f_id.src_ip6),
1726 match = (m_tag_find(m,
1727 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
1729 /* otherwise no match */
1735 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
1736 &((ipfw_insn_ip6 *)cmd)->addr6);
1741 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
1742 &((ipfw_insn_ip6 *)cmd)->addr6);
1744 case O_IP6_SRC_MASK:
1745 case O_IP6_DST_MASK:
1749 struct in6_addr *d =
1750 &((ipfw_insn_ip6 *)cmd)->addr6;
1752 for (; !match && i > 0; d += 2,
1753 i -= F_INSN_SIZE(struct in6_addr)
1759 APPLY_MASK(&p, &d[1]);
1761 IN6_ARE_ADDR_EQUAL(&d[0],
1769 flow6id_match(args->f_id.flow_id6,
1770 (ipfw_insn_u32 *) cmd);
1775 (ext_hd & ((ipfw_insn *) cmd)->arg1);
1789 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
1790 tablearg : cmd->arg1;
1792 /* Packet is already tagged with this tag? */
1793 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
1795 /* We have `untag' action when F_NOT flag is
1796 * present. And we must remove this mtag from
1797 * mbuf and reset `match' to zero (`match' will
1798 * be inversed later).
1799 * Otherwise we should allocate new mtag and
1800 * push it into mbuf.
1802 if (cmd->len & F_NOT) { /* `untag' action */
1804 m_tag_delete(m, mtag);
1808 mtag = m_tag_alloc( MTAG_IPFW,
1811 m_tag_prepend(m, mtag);
1818 case O_FIB: /* try match the specified fib */
1819 if (args->f_id.fib == cmd->arg1)
1825 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
1826 tablearg : cmd->arg1;
1829 match = m_tag_locate(m, MTAG_IPFW,
1834 /* we have ranges */
1835 for (mtag = m_tag_first(m);
1836 mtag != NULL && !match;
1837 mtag = m_tag_next(m, mtag)) {
1841 if (mtag->m_tag_cookie != MTAG_IPFW)
1844 p = ((ipfw_insn_u16 *)cmd)->ports;
1846 for(; !match && i > 0; i--, p += 2)
1848 mtag->m_tag_id >= p[0] &&
1849 mtag->m_tag_id <= p[1];
1855 * The second set of opcodes represents 'actions',
1856 * i.e. the terminal part of a rule once the packet
1857 * matches all previous patterns.
1858 * Typically there is only one action for each rule,
1859 * and the opcode is stored at the end of the rule
1860 * (but there are exceptions -- see below).
1862 * In general, here we set retval and terminate the
1863 * outer loop (would be a 'break 3' in some language,
1864 * but we need to set l=0, done=1)
1867 * O_COUNT and O_SKIPTO actions:
1868 * instead of terminating, we jump to the next rule
1869 * (setting l=0), or to the SKIPTO target (setting
1870 * f/f_len, cmd and l as needed), respectively.
1872 * O_TAG, O_LOG and O_ALTQ action parameters:
1873 * perform some action and set match = 1;
1875 * O_LIMIT and O_KEEP_STATE: these opcodes are
1876 * not real 'actions', and are stored right
1877 * before the 'action' part of the rule.
1878 * These opcodes try to install an entry in the
1879 * state tables; if successful, we continue with
1880 * the next opcode (match=1; break;), otherwise
1881 * the packet must be dropped (set retval,
1882 * break loops with l=0, done=1)
1884 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
1885 * cause a lookup of the state table, and a jump
1886 * to the 'action' part of the parent rule
1887 * if an entry is found, or
1888 * (CHECK_STATE only) a jump to the next rule if
1889 * the entry is not found.
1890 * The result of the lookup is cached so that
1891 * further instances of these opcodes become NOPs.
1892 * The jump to the next rule is done by setting
1897 if (ipfw_install_state(f,
1898 (ipfw_insn_limit *)cmd, args, tablearg)) {
1899 /* error or limit violation */
1900 retval = IP_FW_DENY;
1901 l = 0; /* exit inner loop */
1902 done = 1; /* exit outer loop */
1910 * dynamic rules are checked at the first
1911 * keep-state or check-state occurrence,
1912 * with the result being stored in dyn_dir.
1913 * The compiler introduces a PROBE_STATE
1914 * instruction for us when we have a
1915 * KEEP_STATE (because PROBE_STATE needs
1918 if (dyn_dir == MATCH_UNKNOWN &&
1919 (q = ipfw_lookup_dyn_rule(&args->f_id,
1920 &dyn_dir, proto == IPPROTO_TCP ?
1924 * Found dynamic entry, update stats
1925 * and jump to the 'action' part of
1926 * the parent rule by setting
1927 * f, cmd, l and clearing cmdlen.
1931 /* XXX we would like to have f_pos
1932 * readily accessible in the dynamic
1933 * rule, instead of having to
1937 f_pos = ipfw_find_rule(chain,
1939 cmd = ACTION_PTR(f);
1940 l = f->cmd_len - f->act_ofs;
1947 * Dynamic entry not found. If CHECK_STATE,
1948 * skip to next rule, if PROBE_STATE just
1949 * ignore and continue with next opcode.
1951 if (cmd->opcode == O_CHECK_STATE)
1952 l = 0; /* exit inner loop */
1957 retval = 0; /* accept */
1958 l = 0; /* exit inner loop */
1959 done = 1; /* exit outer loop */
1964 set_match(args, f_pos, chain);
1965 args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ?
1966 tablearg : cmd->arg1;
1967 if (cmd->opcode == O_PIPE)
1968 args->rule.info |= IPFW_IS_PIPE;
1970 args->rule.info |= IPFW_ONEPASS;
1971 retval = IP_FW_DUMMYNET;
1972 l = 0; /* exit inner loop */
1973 done = 1; /* exit outer loop */
1978 if (args->eh) /* not on layer 2 */
1980 /* otherwise this is terminal */
1981 l = 0; /* exit inner loop */
1982 done = 1; /* exit outer loop */
1983 retval = (cmd->opcode == O_DIVERT) ?
1984 IP_FW_DIVERT : IP_FW_TEE;
1985 set_match(args, f_pos, chain);
1986 args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ?
1987 tablearg : cmd->arg1;
1991 f->pcnt++; /* update stats */
1993 f->timestamp = time_uptime;
1994 l = 0; /* exit inner loop */
1998 f->pcnt++; /* update stats */
2000 f->timestamp = time_uptime;
2001 /* If possible use cached f_pos (in f->next_rule),
2002 * whose version is written in f->next_rule
2003 * (horrible hacks to avoid changing the ABI).
2005 if (cmd->arg1 != IP_FW_TABLEARG &&
2006 (uintptr_t)f->x_next == chain->id) {
2007 f_pos = (uintptr_t)f->next_rule;
2009 int i = (cmd->arg1 == IP_FW_TABLEARG) ?
2010 tablearg : cmd->arg1;
2011 /* make sure we do not jump backward */
2012 if (i <= f->rulenum)
2014 f_pos = ipfw_find_rule(chain, i, 0);
2015 /* update the cache */
2016 if (cmd->arg1 != IP_FW_TABLEARG) {
2018 (void *)(uintptr_t)f_pos;
2020 (void *)(uintptr_t)chain->id;
2024 * Skip disabled rules, and re-enter
2025 * the inner loop with the correct
2026 * f_pos, f, l and cmd.
2027 * Also clear cmdlen and skip_or
2029 for (; f_pos < chain->n_rules - 1 &&
2031 (1 << chain->map[f_pos]->set));
2034 /* Re-enter the inner loop at the skipto rule. */
2035 f = chain->map[f_pos];
2042 break; /* not reached */
2044 case O_CALLRETURN: {
2046 * Implementation of `subroutine' call/return,
2047 * in the stack carried in an mbuf tag. This
2048 * is different from `skipto' in that any call
2049 * address is possible (`skipto' must prevent
2050 * backward jumps to avoid endless loops).
2051 * We have `return' action when F_NOT flag is
2052 * present. The `m_tag_id' field is used as
2056 uint16_t jmpto, *stack;
2058 #define IS_CALL ((cmd->len & F_NOT) == 0)
2059 #define IS_RETURN ((cmd->len & F_NOT) != 0)
2061 * Hand-rolled version of m_tag_locate() with
2063 * If not already tagged, allocate new tag.
2065 mtag = m_tag_first(m);
2066 while (mtag != NULL) {
2067 if (mtag->m_tag_cookie ==
2070 mtag = m_tag_next(m, mtag);
2072 if (mtag == NULL && IS_CALL) {
2073 mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2074 IPFW_CALLSTACK_SIZE *
2075 sizeof(uint16_t), M_NOWAIT);
2077 m_tag_prepend(m, mtag);
2081 * On error both `call' and `return' just
2082 * continue with next rule.
2084 if (IS_RETURN && (mtag == NULL ||
2085 mtag->m_tag_id == 0)) {
2086 l = 0; /* exit inner loop */
2089 if (IS_CALL && (mtag == NULL ||
2090 mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2091 printf("ipfw: call stack error, "
2092 "go to next rule\n");
2093 l = 0; /* exit inner loop */
2097 f->pcnt++; /* update stats */
2099 f->timestamp = time_uptime;
2100 stack = (uint16_t *)(mtag + 1);
2103 * The `call' action may use cached f_pos
2104 * (in f->next_rule), whose version is written
2106 * The `return' action, however, doesn't have
2107 * fixed jump address in cmd->arg1 and can't use
2111 stack[mtag->m_tag_id] = f->rulenum;
2113 if (cmd->arg1 != IP_FW_TABLEARG &&
2114 (uintptr_t)f->x_next == chain->id) {
2115 f_pos = (uintptr_t)f->next_rule;
2117 jmpto = (cmd->arg1 ==
2118 IP_FW_TABLEARG) ? tablearg:
2120 f_pos = ipfw_find_rule(chain,
2122 /* update the cache */
2133 } else { /* `return' action */
2135 jmpto = stack[mtag->m_tag_id] + 1;
2136 f_pos = ipfw_find_rule(chain, jmpto, 0);
2140 * Skip disabled rules, and re-enter
2141 * the inner loop with the correct
2142 * f_pos, f, l and cmd.
2143 * Also clear cmdlen and skip_or
2145 for (; f_pos < chain->n_rules - 1 &&
2147 (1 << chain->map[f_pos]->set)); f_pos++)
2149 /* Re-enter the inner loop at the dest rule. */
2150 f = chain->map[f_pos];
2156 break; /* NOTREACHED */
2163 * Drop the packet and send a reject notice
2164 * if the packet is not ICMP (or is an ICMP
2165 * query), and it is not multicast/broadcast.
2167 if (hlen > 0 && is_ipv4 && offset == 0 &&
2168 (proto != IPPROTO_ICMP ||
2169 is_icmp_query(ICMP(ulp))) &&
2170 !(m->m_flags & (M_BCAST|M_MCAST)) &&
2171 !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2172 send_reject(args, cmd->arg1, iplen, ip);
2178 if (hlen > 0 && is_ipv6 &&
2179 ((offset & IP6F_OFF_MASK) == 0) &&
2180 (proto != IPPROTO_ICMPV6 ||
2181 (is_icmp6_query(icmp6_type) == 1)) &&
2182 !(m->m_flags & (M_BCAST|M_MCAST)) &&
2183 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
2185 args, cmd->arg1, hlen,
2186 (struct ip6_hdr *)ip);
2192 retval = IP_FW_DENY;
2193 l = 0; /* exit inner loop */
2194 done = 1; /* exit outer loop */
2198 if (args->eh) /* not valid on layer2 pkts */
2200 if (q == NULL || q->rule != f ||
2201 dyn_dir == MATCH_FORWARD) {
2202 struct sockaddr_in *sa;
2203 sa = &(((ipfw_insn_sa *)cmd)->sa);
2204 if (sa->sin_addr.s_addr == INADDR_ANY) {
2205 bcopy(sa, &args->hopstore,
2207 args->hopstore.sin_addr.s_addr =
2209 args->next_hop = &args->hopstore;
2211 args->next_hop = sa;
2214 retval = IP_FW_PASS;
2215 l = 0; /* exit inner loop */
2216 done = 1; /* exit outer loop */
2221 set_match(args, f_pos, chain);
2222 args->rule.info = (cmd->arg1 == IP_FW_TABLEARG) ?
2223 tablearg : cmd->arg1;
2225 args->rule.info |= IPFW_ONEPASS;
2226 retval = (cmd->opcode == O_NETGRAPH) ?
2227 IP_FW_NETGRAPH : IP_FW_NGTEE;
2228 l = 0; /* exit inner loop */
2229 done = 1; /* exit outer loop */
2235 f->pcnt++; /* update stats */
2237 f->timestamp = time_uptime;
2238 fib = (cmd->arg1 == IP_FW_TABLEARG) ? tablearg:
2240 if (fib >= rt_numfibs)
2243 args->f_id.fib = fib;
2244 l = 0; /* exit inner loop */
2249 if (!IPFW_NAT_LOADED) {
2250 retval = IP_FW_DENY;
2255 set_match(args, f_pos, chain);
2256 /* Check if this is 'global' nat rule */
2257 if (cmd->arg1 == 0) {
2258 retval = ipfw_nat_ptr(args, NULL, m);
2263 t = ((ipfw_insn_nat *)cmd)->nat;
2265 nat_id = (cmd->arg1 == IP_FW_TABLEARG) ?
2266 tablearg : cmd->arg1;
2267 t = (*lookup_nat_ptr)(&chain->nat, nat_id);
2270 retval = IP_FW_DENY;
2271 l = 0; /* exit inner loop */
2272 done = 1; /* exit outer loop */
2275 if (cmd->arg1 != IP_FW_TABLEARG)
2276 ((ipfw_insn_nat *)cmd)->nat = t;
2278 retval = ipfw_nat_ptr(args, t, m);
2280 l = 0; /* exit inner loop */
2281 done = 1; /* exit outer loop */
2289 l = 0; /* in any case exit inner loop */
2290 ip_off = ntohs(ip->ip_off);
2292 /* if not fragmented, go to next rule */
2293 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
2296 * ip_reass() expects len & off in host
2301 args->m = m = ip_reass(m);
2304 * do IP header checksum fixup.
2306 if (m == NULL) { /* fragment got swallowed */
2307 retval = IP_FW_DENY;
2308 } else { /* good, packet complete */
2311 ip = mtod(m, struct ip *);
2312 hlen = ip->ip_hl << 2;
2315 if (hlen == sizeof(struct ip))
2316 ip->ip_sum = in_cksum_hdr(ip);
2318 ip->ip_sum = in_cksum(m, hlen);
2319 retval = IP_FW_REASS;
2320 set_match(args, f_pos, chain);
2322 done = 1; /* exit outer loop */
2327 panic("-- unknown opcode %d\n", cmd->opcode);
2328 } /* end of switch() on opcodes */
2330 * if we get here with l=0, then match is irrelevant.
2333 if (cmd->len & F_NOT)
2337 if (cmd->len & F_OR)
2340 if (!(cmd->len & F_OR)) /* not an OR block, */
2341 break; /* try next rule */
2344 } /* end of inner loop, scan opcodes */
2350 /* next_rule:; */ /* try next rule */
2352 } /* end of outer for, scan rules */
2355 struct ip_fw *rule = chain->map[f_pos];
2356 /* Update statistics */
2358 rule->bcnt += pktlen;
2359 rule->timestamp = time_uptime;
2361 retval = IP_FW_DENY;
2362 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
2364 IPFW_RUNLOCK(chain);
2366 if (ucred_cache != NULL)
2367 crfree(ucred_cache);
2373 printf("ipfw: pullup failed\n");
2374 return (IP_FW_DENY);
2378 * Module and VNET glue
2382 * Stuff that must be initialised only on boot or module load
2391 * Only print out this stuff the first time around,
2392 * when called from the sysinit code.
2398 "initialized, divert %s, nat %s, "
2399 "rule-based forwarding "
2400 #ifdef IPFIREWALL_FORWARD
2405 "default to %s, logging ",
2411 #ifdef IPFIREWALL_NAT
2416 default_to_accept ? "accept" : "deny");
2419 * Note: V_xxx variables can be accessed here but the vnet specific
2420 * initializer may not have been called yet for the VIMAGE case.
2421 * Tuneables will have been processed. We will print out values for
2423 * XXX This should all be rationalized AFTER 8.0
2425 if (V_fw_verbose == 0)
2426 printf("disabled\n");
2427 else if (V_verbose_limit == 0)
2428 printf("unlimited\n");
2430 printf("limited to %d packets/entry by default\n",
2433 ipfw_log_bpf(1); /* init */
2438 * Called for the removal of the last instance only on module unload.
2444 ipfw_log_bpf(0); /* uninit */
2446 printf("IP firewall unloaded\n");
2450 * Stuff that must be initialized for every instance
2451 * (including the first of course).
2454 vnet_ipfw_init(const void *unused)
2457 struct ip_fw *rule = NULL;
2458 struct ip_fw_chain *chain;
2460 chain = &V_layer3_chain;
2462 /* First set up some values that are compile time options */
2463 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */
2464 V_fw_deny_unknown_exthdrs = 1;
2465 #ifdef IPFIREWALL_VERBOSE
2468 #ifdef IPFIREWALL_VERBOSE_LIMIT
2469 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
2471 #ifdef IPFIREWALL_NAT
2472 LIST_INIT(&chain->nat);
2475 /* insert the default rule and create the initial map */
2477 chain->static_len = sizeof(struct ip_fw);
2478 chain->map = malloc(sizeof(struct ip_fw *), M_IPFW, M_NOWAIT | M_ZERO);
2480 rule = malloc(chain->static_len, M_IPFW, M_NOWAIT | M_ZERO);
2483 free(chain->map, M_IPFW);
2484 printf("ipfw2: ENOSPC initializing default rule "
2485 "(support disabled)\n");
2488 error = ipfw_init_tables(chain);
2490 panic("init_tables"); /* XXX Marko fix this ! */
2493 /* fill and insert the default rule */
2495 rule->rulenum = IPFW_DEFAULT_RULE;
2497 rule->set = RESVD_SET;
2498 rule->cmd[0].len = 1;
2499 rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
2500 chain->rules = chain->default_rule = chain->map[0] = rule;
2501 chain->id = rule->id = 1;
2503 IPFW_LOCK_INIT(chain);
2506 /* First set up some values that are compile time options */
2507 V_ipfw_vnet_ready = 1; /* Open for business */
2510 * Hook the sockopt handler, and the layer2 (V_ip_fw_chk_ptr)
2511 * and pfil hooks for ipv4 and ipv6. Even if the latter two fail
2512 * we still keep the module alive because the sockopt and
2513 * layer2 paths are still useful.
2514 * ipfw[6]_hook return 0 on success, ENOENT on failure,
2515 * so we can ignore the exact return value and just set a flag.
2517 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
2518 * changes in the underlying (per-vnet) variables trigger
2519 * immediate hook()/unhook() calls.
2520 * In layer2 we have the same behaviour, except that V_ether_ipfw
2521 * is checked on each packet because there are no pfil hooks.
2523 V_ip_fw_ctl_ptr = ipfw_ctl;
2524 V_ip_fw_chk_ptr = ipfw_chk;
2525 error = ipfw_attach_hooks(1);
2530 * Called for the removal of each instance.
2533 vnet_ipfw_uninit(const void *unused)
2535 struct ip_fw *reap, *rule;
2536 struct ip_fw_chain *chain = &V_layer3_chain;
2539 V_ipfw_vnet_ready = 0; /* tell new callers to go away */
2541 * disconnect from ipv4, ipv6, layer2 and sockopt.
2542 * Then grab, release and grab again the WLOCK so we make
2543 * sure the update is propagated and nobody will be in.
2545 (void)ipfw_attach_hooks(0 /* detach */);
2546 V_ip_fw_chk_ptr = NULL;
2547 V_ip_fw_ctl_ptr = NULL;
2548 IPFW_UH_WLOCK(chain);
2549 IPFW_UH_WUNLOCK(chain);
2550 IPFW_UH_WLOCK(chain);
2553 IPFW_WUNLOCK(chain);
2556 ipfw_dyn_uninit(0); /* run the callout_drain */
2557 ipfw_destroy_tables(chain);
2559 for (i = 0; i < chain->n_rules; i++) {
2560 rule = chain->map[i];
2561 rule->x_next = reap;
2565 free(chain->map, M_IPFW);
2566 IPFW_WUNLOCK(chain);
2567 IPFW_UH_WUNLOCK(chain);
2569 ipfw_reap_rules(reap);
2570 IPFW_LOCK_DESTROY(chain);
2571 ipfw_dyn_uninit(1); /* free the remaining parts */
2576 * Module event handler.
2577 * In general we have the choice of handling most of these events by the
2578 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
2579 * use the SYSINIT handlers as they are more capable of expressing the
2580 * flow of control during module and vnet operations, so this is just
2581 * a skeleton. Note there is no SYSINIT equivalent of the module
2582 * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
2585 ipfw_modevent(module_t mod, int type, void *unused)
2591 /* Called once at module load or
2592 * system boot if compiled in. */
2595 /* Called before unload. May veto unloading. */
2598 /* Called during unload. */
2601 /* Called during system shutdown. */
2610 static moduledata_t ipfwmod = {
2616 /* Define startup order. */
2617 #define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_IFATTACHDOMAIN
2618 #define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */
2619 #define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */
2620 #define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */
2622 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
2623 MODULE_VERSION(ipfw, 2);
2624 /* should declare some dependencies here */
2627 * Starting up. Done in order after ipfwmod() has been called.
2628 * VNET_SYSINIT is also called for each existing vnet and each new vnet.
2630 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2632 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2633 vnet_ipfw_init, NULL);
2636 * Closing up shop. These are done in REVERSE ORDER, but still
2637 * after ipfwmod() has been called. Not called on reboot.
2638 * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
2639 * or when the module is unloaded.
2641 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
2642 ipfw_destroy, NULL);
2643 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
2644 vnet_ipfw_uninit, NULL);