2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2002-2009 Luigi Rizzo, Universita` di Pisa
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28 #include <sys/cdefs.h>
29 __FBSDID("$FreeBSD$");
32 * The FreeBSD IP packet firewall, main file
36 #include "opt_ipdivert.h"
39 #error "IPFIREWALL requires INET"
41 #include "opt_inet6.h"
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/condvar.h>
46 #include <sys/counter.h>
47 #include <sys/eventhandler.h>
48 #include <sys/malloc.h>
50 #include <sys/kernel.h>
53 #include <sys/module.h>
56 #include <sys/rwlock.h>
57 #include <sys/rmlock.h>
58 #include <sys/socket.h>
59 #include <sys/socketvar.h>
60 #include <sys/sysctl.h>
61 #include <sys/syslog.h>
62 #include <sys/ucred.h>
63 #include <net/ethernet.h> /* for ETHERTYPE_IP */
65 #include <net/if_var.h>
66 #include <net/route.h>
70 #include <netpfil/pf/pf_mtag.h>
72 #include <netinet/in.h>
73 #include <netinet/in_var.h>
74 #include <netinet/in_pcb.h>
75 #include <netinet/ip.h>
76 #include <netinet/ip_var.h>
77 #include <netinet/ip_icmp.h>
78 #include <netinet/ip_fw.h>
79 #include <netinet/ip_carp.h>
80 #include <netinet/pim.h>
81 #include <netinet/tcp_var.h>
82 #include <netinet/udp.h>
83 #include <netinet/udp_var.h>
84 #include <netinet/sctp.h>
85 #include <netinet/sctp_crc32.h>
86 #include <netinet/sctp_header.h>
88 #include <netinet/ip6.h>
89 #include <netinet/icmp6.h>
90 #include <netinet/in_fib.h>
92 #include <netinet6/in6_fib.h>
93 #include <netinet6/in6_pcb.h>
94 #include <netinet6/scope6_var.h>
95 #include <netinet6/ip6_var.h>
98 #include <net/if_gre.h> /* for struct grehdr */
100 #include <netpfil/ipfw/ip_fw_private.h>
102 #include <machine/in_cksum.h> /* XXX for in_cksum */
105 #include <security/mac/mac_framework.h>
109 * static variables followed by global ones.
110 * All ipfw global variables are here.
113 VNET_DEFINE_STATIC(int, fw_deny_unknown_exthdrs);
114 #define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs)
116 VNET_DEFINE_STATIC(int, fw_permit_single_frag6) = 1;
117 #define V_fw_permit_single_frag6 VNET(fw_permit_single_frag6)
119 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
120 static int default_to_accept = 1;
122 static int default_to_accept;
125 VNET_DEFINE(int, autoinc_step);
126 VNET_DEFINE(int, fw_one_pass) = 1;
128 VNET_DEFINE(unsigned int, fw_tables_max);
129 VNET_DEFINE(unsigned int, fw_tables_sets) = 0; /* Don't use set-aware tables */
130 /* Use 128 tables by default */
131 static unsigned int default_fw_tables = IPFW_TABLES_DEFAULT;
133 #ifndef LINEAR_SKIPTO
134 static int jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
135 int tablearg, int jump_backwards);
136 #define JUMP(ch, f, num, targ, back) jump_fast(ch, f, num, targ, back)
138 static int jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
139 int tablearg, int jump_backwards);
140 #define JUMP(ch, f, num, targ, back) jump_linear(ch, f, num, targ, back)
144 * Each rule belongs to one of 32 different sets (0..31).
145 * The variable set_disable contains one bit per set.
146 * If the bit is set, all rules in the corresponding set
147 * are disabled. Set RESVD_SET(31) is reserved for the default rule
148 * and rules that are not deleted by the flush command,
149 * and CANNOT be disabled.
150 * Rules in set RESVD_SET can only be deleted individually.
152 VNET_DEFINE(u_int32_t, set_disable);
153 #define V_set_disable VNET(set_disable)
155 VNET_DEFINE(int, fw_verbose);
156 /* counter for ipfw_log(NULL...) */
157 VNET_DEFINE(u_int64_t, norule_counter);
158 VNET_DEFINE(int, verbose_limit);
160 /* layer3_chain contains the list of rules for layer 3 */
161 VNET_DEFINE(struct ip_fw_chain, layer3_chain);
163 /* ipfw_vnet_ready controls when we are open for business */
164 VNET_DEFINE(int, ipfw_vnet_ready) = 0;
166 VNET_DEFINE(int, ipfw_nat_ready) = 0;
168 ipfw_nat_t *ipfw_nat_ptr = NULL;
169 struct cfg_nat *(*lookup_nat_ptr)(struct nat_list *, int);
170 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
171 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
172 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
173 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
176 uint32_t dummy_def = IPFW_DEFAULT_RULE;
177 static int sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS);
178 static int sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS);
182 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
183 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
184 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_one_pass), 0,
185 "Only do a single pass through ipfw when using dummynet(4)");
186 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step,
187 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(autoinc_step), 0,
188 "Rule number auto-increment step");
189 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose,
190 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE3, &VNET_NAME(fw_verbose), 0,
191 "Log matches to ipfw rules");
192 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit,
193 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(verbose_limit), 0,
194 "Set upper limit of matches of ipfw rules logged");
195 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
197 "The default/max possible rule number.");
198 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_max,
199 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW, 0, 0, sysctl_ipfw_table_num, "IU",
200 "Maximum number of concurrently used tables");
201 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, tables_sets,
202 CTLFLAG_VNET | CTLTYPE_UINT | CTLFLAG_RW,
203 0, 0, sysctl_ipfw_tables_sets, "IU",
204 "Use per-set namespace for tables");
205 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, default_to_accept, CTLFLAG_RDTUN,
206 &default_to_accept, 0,
207 "Make the default rule accept all packets.");
208 TUNABLE_INT("net.inet.ip.fw.tables_max", (int *)&default_fw_tables);
209 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count,
210 CTLFLAG_VNET | CTLFLAG_RD, &VNET_NAME(layer3_chain.n_rules), 0,
211 "Number of static rules");
214 SYSCTL_DECL(_net_inet6_ip6);
215 SYSCTL_NODE(_net_inet6_ip6, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
216 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, deny_unknown_exthdrs,
217 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
218 &VNET_NAME(fw_deny_unknown_exthdrs), 0,
219 "Deny packets with unknown IPv6 Extension Headers");
220 SYSCTL_INT(_net_inet6_ip6_fw, OID_AUTO, permit_single_frag6,
221 CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE,
222 &VNET_NAME(fw_permit_single_frag6), 0,
223 "Permit single packet IPv6 fragments");
228 #endif /* SYSCTL_NODE */
232 * Some macros used in the various matching options.
233 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
234 * Other macros just cast void * into the appropriate type
236 #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
237 #define TCP(p) ((struct tcphdr *)(p))
238 #define SCTP(p) ((struct sctphdr *)(p))
239 #define UDP(p) ((struct udphdr *)(p))
240 #define ICMP(p) ((struct icmphdr *)(p))
241 #define ICMP6(p) ((struct icmp6_hdr *)(p))
244 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
246 int type = icmp->icmp_type;
248 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
251 #define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
252 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
255 is_icmp_query(struct icmphdr *icmp)
257 int type = icmp->icmp_type;
259 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
264 * The following checks use two arrays of 8 or 16 bits to store the
265 * bits that we want set or clear, respectively. They are in the
266 * low and high half of cmd->arg1 or cmd->d[0].
268 * We scan options and store the bits we find set. We succeed if
270 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
272 * The code is sometimes optimized not to store additional variables.
276 flags_match(ipfw_insn *cmd, u_int8_t bits)
281 if ( ((cmd->arg1 & 0xff) & bits) != 0)
282 return 0; /* some bits we want set were clear */
283 want_clear = (cmd->arg1 >> 8) & 0xff;
284 if ( (want_clear & bits) != want_clear)
285 return 0; /* some bits we want clear were set */
290 ipopts_match(struct ip *ip, ipfw_insn *cmd)
292 int optlen, bits = 0;
293 u_char *cp = (u_char *)(ip + 1);
294 int x = (ip->ip_hl << 2) - sizeof (struct ip);
296 for (; x > 0; x -= optlen, cp += optlen) {
297 int opt = cp[IPOPT_OPTVAL];
299 if (opt == IPOPT_EOL)
301 if (opt == IPOPT_NOP)
304 optlen = cp[IPOPT_OLEN];
305 if (optlen <= 0 || optlen > x)
306 return 0; /* invalid or truncated */
314 bits |= IP_FW_IPOPT_LSRR;
318 bits |= IP_FW_IPOPT_SSRR;
322 bits |= IP_FW_IPOPT_RR;
326 bits |= IP_FW_IPOPT_TS;
330 return (flags_match(cmd, bits));
334 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
336 int optlen, bits = 0;
337 u_char *cp = (u_char *)(tcp + 1);
338 int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
340 for (; x > 0; x -= optlen, cp += optlen) {
342 if (opt == TCPOPT_EOL)
344 if (opt == TCPOPT_NOP)
358 bits |= IP_FW_TCPOPT_MSS;
362 bits |= IP_FW_TCPOPT_WINDOW;
365 case TCPOPT_SACK_PERMITTED:
367 bits |= IP_FW_TCPOPT_SACK;
370 case TCPOPT_TIMESTAMP:
371 bits |= IP_FW_TCPOPT_TS;
376 return (flags_match(cmd, bits));
380 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd, struct ip_fw_chain *chain,
384 if (ifp == NULL) /* no iface with this packet, match fails */
387 /* Check by name or by IP address */
388 if (cmd->name[0] != '\0') { /* match by name */
389 if (cmd->name[0] == '\1') /* use tablearg to match */
390 return ipfw_lookup_table(chain, cmd->p.kidx, 0,
391 &ifp->if_index, tablearg);
394 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
397 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
401 #if !defined(USERSPACE) && defined(__FreeBSD__) /* and OSX too ? */
405 CK_STAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
406 if (ia->ifa_addr->sa_family != AF_INET)
408 if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
409 (ia->ifa_addr))->sin_addr.s_addr) {
410 if_addr_runlock(ifp);
411 return(1); /* match */
414 if_addr_runlock(ifp);
415 #endif /* __FreeBSD__ */
417 return(0); /* no match, fail ... */
421 * The verify_path function checks if a route to the src exists and
422 * if it is reachable via ifp (when provided).
424 * The 'verrevpath' option checks that the interface that an IP packet
425 * arrives on is the same interface that traffic destined for the
426 * packet's source address would be routed out of.
427 * The 'versrcreach' option just checks that the source address is
428 * reachable via any route (except default) in the routing table.
429 * These two are a measure to block forged packets. This is also
430 * commonly known as "anti-spoofing" or Unicast Reverse Path
431 * Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
432 * is purposely reminiscent of the Cisco IOS command,
434 * ip verify unicast reverse-path
435 * ip verify unicast source reachable-via any
437 * which implements the same functionality. But note that the syntax
438 * is misleading, and the check may be performed on all IP packets
439 * whether unicast, multicast, or broadcast.
442 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
444 #if defined(USERSPACE) || !defined(__FreeBSD__)
447 struct nhop4_basic nh4;
449 if (fib4_lookup_nh_basic(fib, src, NHR_IFAIF, 0, &nh4) != 0)
453 * If ifp is provided, check for equality with rtentry.
454 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
455 * in order to pass packets injected back by if_simloop():
456 * routing entry (via lo0) for our own address
457 * may exist, so we need to handle routing assymetry.
459 if (ifp != NULL && ifp != nh4.nh_ifp)
462 /* if no ifp provided, check if rtentry is not default route */
463 if (ifp == NULL && (nh4.nh_flags & NHF_DEFAULT) != 0)
466 /* or if this is a blackhole/reject route */
467 if (ifp == NULL && (nh4.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
470 /* found valid route */
472 #endif /* __FreeBSD__ */
476 * Generate an SCTP packet containing an ABORT chunk. The verification tag
477 * is given by vtag. The T-bit is set in the ABORT chunk if and only if
478 * reflected is not 0.
482 ipfw_send_abort(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t vtag,
490 struct sctphdr *sctp;
491 struct sctp_chunkhdr *chunk;
492 u_int16_t hlen, plen, tlen;
494 MGETHDR(m, M_NOWAIT, MT_DATA);
498 M_SETFIB(m, id->fib);
501 mac_netinet_firewall_reply(replyto, m);
503 mac_netinet_firewall_send(m);
505 (void)replyto; /* don't warn about unused arg */
508 switch (id->addr_type) {
510 hlen = sizeof(struct ip);
514 hlen = sizeof(struct ip6_hdr);
522 plen = sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr);
524 m->m_data += max_linkhdr;
525 m->m_flags |= M_SKIP_FIREWALL;
526 m->m_pkthdr.len = m->m_len = tlen;
527 m->m_pkthdr.rcvif = NULL;
528 bzero(m->m_data, tlen);
530 switch (id->addr_type) {
532 ip = mtod(m, struct ip *);
535 ip->ip_hl = sizeof(struct ip) >> 2;
536 ip->ip_tos = IPTOS_LOWDELAY;
537 ip->ip_len = htons(tlen);
538 ip->ip_id = htons(0);
539 ip->ip_off = htons(0);
540 ip->ip_ttl = V_ip_defttl;
541 ip->ip_p = IPPROTO_SCTP;
543 ip->ip_src.s_addr = htonl(id->dst_ip);
544 ip->ip_dst.s_addr = htonl(id->src_ip);
546 sctp = (struct sctphdr *)(ip + 1);
550 ip6 = mtod(m, struct ip6_hdr *);
552 ip6->ip6_vfc = IPV6_VERSION;
553 ip6->ip6_plen = htons(plen);
554 ip6->ip6_nxt = IPPROTO_SCTP;
555 ip6->ip6_hlim = IPV6_DEFHLIM;
556 ip6->ip6_src = id->dst_ip6;
557 ip6->ip6_dst = id->src_ip6;
559 sctp = (struct sctphdr *)(ip6 + 1);
564 sctp->src_port = htons(id->dst_port);
565 sctp->dest_port = htons(id->src_port);
566 sctp->v_tag = htonl(vtag);
567 sctp->checksum = htonl(0);
569 chunk = (struct sctp_chunkhdr *)(sctp + 1);
570 chunk->chunk_type = SCTP_ABORT_ASSOCIATION;
571 chunk->chunk_flags = 0;
572 if (reflected != 0) {
573 chunk->chunk_flags |= SCTP_HAD_NO_TCB;
575 chunk->chunk_length = htons(sizeof(struct sctp_chunkhdr));
577 sctp->checksum = sctp_calculate_cksum(m, hlen);
583 * Generate a TCP packet, containing either a RST or a keepalive.
584 * When flags & TH_RST, we are sending a RST packet, because of a
585 * "reset" action matched the packet.
586 * Otherwise we are sending a keepalive, and flags & TH_
587 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required
588 * so that MAC can label the reply appropriately.
591 ipfw_send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq,
592 u_int32_t ack, int flags)
594 struct mbuf *m = NULL; /* stupid compiler */
595 struct ip *h = NULL; /* stupid compiler */
597 struct ip6_hdr *h6 = NULL;
599 struct tcphdr *th = NULL;
602 MGETHDR(m, M_NOWAIT, MT_DATA);
606 M_SETFIB(m, id->fib);
609 mac_netinet_firewall_reply(replyto, m);
611 mac_netinet_firewall_send(m);
613 (void)replyto; /* don't warn about unused arg */
616 switch (id->addr_type) {
618 len = sizeof(struct ip) + sizeof(struct tcphdr);
622 len = sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
630 dir = ((flags & (TH_SYN | TH_RST)) == TH_SYN);
632 m->m_data += max_linkhdr;
633 m->m_flags |= M_SKIP_FIREWALL;
634 m->m_pkthdr.len = m->m_len = len;
635 m->m_pkthdr.rcvif = NULL;
636 bzero(m->m_data, len);
638 switch (id->addr_type) {
640 h = mtod(m, struct ip *);
642 /* prepare for checksum */
643 h->ip_p = IPPROTO_TCP;
644 h->ip_len = htons(sizeof(struct tcphdr));
646 h->ip_src.s_addr = htonl(id->src_ip);
647 h->ip_dst.s_addr = htonl(id->dst_ip);
649 h->ip_src.s_addr = htonl(id->dst_ip);
650 h->ip_dst.s_addr = htonl(id->src_ip);
653 th = (struct tcphdr *)(h + 1);
657 h6 = mtod(m, struct ip6_hdr *);
659 /* prepare for checksum */
660 h6->ip6_nxt = IPPROTO_TCP;
661 h6->ip6_plen = htons(sizeof(struct tcphdr));
663 h6->ip6_src = id->src_ip6;
664 h6->ip6_dst = id->dst_ip6;
666 h6->ip6_src = id->dst_ip6;
667 h6->ip6_dst = id->src_ip6;
670 th = (struct tcphdr *)(h6 + 1);
676 th->th_sport = htons(id->src_port);
677 th->th_dport = htons(id->dst_port);
679 th->th_sport = htons(id->dst_port);
680 th->th_dport = htons(id->src_port);
682 th->th_off = sizeof(struct tcphdr) >> 2;
684 if (flags & TH_RST) {
685 if (flags & TH_ACK) {
686 th->th_seq = htonl(ack);
687 th->th_flags = TH_RST;
691 th->th_ack = htonl(seq);
692 th->th_flags = TH_RST | TH_ACK;
696 * Keepalive - use caller provided sequence numbers
698 th->th_seq = htonl(seq);
699 th->th_ack = htonl(ack);
700 th->th_flags = TH_ACK;
703 switch (id->addr_type) {
705 th->th_sum = in_cksum(m, len);
707 /* finish the ip header */
709 h->ip_hl = sizeof(*h) >> 2;
710 h->ip_tos = IPTOS_LOWDELAY;
711 h->ip_off = htons(0);
712 h->ip_len = htons(len);
713 h->ip_ttl = V_ip_defttl;
718 th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(*h6),
719 sizeof(struct tcphdr));
721 /* finish the ip6 header */
722 h6->ip6_vfc |= IPV6_VERSION;
723 h6->ip6_hlim = IPV6_DEFHLIM;
733 * ipv6 specific rules here...
736 icmp6type_match (int type, ipfw_insn_u32 *cmd)
738 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
742 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
745 for (i=0; i <= cmd->o.arg1; ++i )
746 if (curr_flow == cmd->d[i] )
751 /* support for IP6_*_ME opcodes */
752 static const struct in6_addr lla_mask = {{{
753 0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0xff, 0xff,
754 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
758 ipfw_localip6(struct in6_addr *in6)
760 struct rm_priotracker in6_ifa_tracker;
761 struct in6_ifaddr *ia;
763 if (IN6_IS_ADDR_MULTICAST(in6))
766 if (!IN6_IS_ADDR_LINKLOCAL(in6))
767 return (in6_localip(in6));
769 IN6_IFADDR_RLOCK(&in6_ifa_tracker);
770 CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
771 if (!IN6_IS_ADDR_LINKLOCAL(&ia->ia_addr.sin6_addr))
773 if (IN6_ARE_MASKED_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
775 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
779 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
784 verify_path6(struct in6_addr *src, struct ifnet *ifp, u_int fib)
786 struct nhop6_basic nh6;
788 if (IN6_IS_SCOPE_LINKLOCAL(src))
791 if (fib6_lookup_nh_basic(fib, src, 0, NHR_IFAIF, 0, &nh6) != 0)
794 /* If ifp is provided, check for equality with route table. */
795 if (ifp != NULL && ifp != nh6.nh_ifp)
798 /* if no ifp provided, check if rtentry is not default route */
799 if (ifp == NULL && (nh6.nh_flags & NHF_DEFAULT) != 0)
802 /* or if this is a blackhole/reject route */
803 if (ifp == NULL && (nh6.nh_flags & (NHF_REJECT|NHF_BLACKHOLE)) != 0)
806 /* found valid route */
811 is_icmp6_query(int icmp6_type)
813 if ((icmp6_type <= ICMP6_MAXTYPE) &&
814 (icmp6_type == ICMP6_ECHO_REQUEST ||
815 icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
816 icmp6_type == ICMP6_WRUREQUEST ||
817 icmp6_type == ICMP6_FQDN_QUERY ||
818 icmp6_type == ICMP6_NI_QUERY))
825 map_icmp_unreach(int code)
830 case ICMP_UNREACH_NET:
831 case ICMP_UNREACH_HOST:
832 case ICMP_UNREACH_SRCFAIL:
833 case ICMP_UNREACH_NET_UNKNOWN:
834 case ICMP_UNREACH_HOST_UNKNOWN:
835 case ICMP_UNREACH_TOSNET:
836 case ICMP_UNREACH_TOSHOST:
837 return (ICMP6_DST_UNREACH_NOROUTE);
838 case ICMP_UNREACH_PORT:
839 return (ICMP6_DST_UNREACH_NOPORT);
842 * Map the rest of codes into admit prohibited.
843 * XXX: unreach proto should be mapped into ICMPv6
844 * parameter problem, but we use only unreach type.
846 return (ICMP6_DST_UNREACH_ADMIN);
851 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
856 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
858 tcp = (struct tcphdr *)((char *)ip6 + hlen);
860 if ((tcp->th_flags & TH_RST) == 0) {
862 m0 = ipfw_send_pkt(args->m, &(args->f_id),
863 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
864 tcp->th_flags | TH_RST);
866 ip6_output(m0, NULL, NULL, 0, NULL, NULL,
870 } else if (code == ICMP6_UNREACH_ABORT &&
871 args->f_id.proto == IPPROTO_SCTP) {
873 struct sctphdr *sctp;
877 sctp = (struct sctphdr *)((char *)ip6 + hlen);
879 v_tag = ntohl(sctp->v_tag);
880 /* Investigate the first chunk header if available */
881 if (m->m_len >= hlen + sizeof(struct sctphdr) +
882 sizeof(struct sctp_chunkhdr)) {
883 struct sctp_chunkhdr *chunk;
885 chunk = (struct sctp_chunkhdr *)(sctp + 1);
886 switch (chunk->chunk_type) {
887 case SCTP_INITIATION:
889 * Packets containing an INIT chunk MUST have
896 /* INIT chunk MUST NOT be bundled */
897 if (m->m_pkthdr.len >
898 hlen + sizeof(struct sctphdr) +
899 ntohs(chunk->chunk_length) + 3) {
902 /* Use the initiate tag if available */
903 if ((m->m_len >= hlen + sizeof(struct sctphdr) +
904 sizeof(struct sctp_chunkhdr) +
905 offsetof(struct sctp_init, a_rwnd))) {
906 struct sctp_init *init;
908 init = (struct sctp_init *)(chunk + 1);
909 v_tag = ntohl(init->initiate_tag);
913 case SCTP_ABORT_ASSOCIATION:
915 * If the packet contains an ABORT chunk, don't
917 * XXX: We should search through all chunks,
918 * but don't do to avoid attacks.
927 m0 = ipfw_send_abort(args->m, &(args->f_id), v_tag,
931 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
933 } else if (code != ICMP6_UNREACH_RST && code != ICMP6_UNREACH_ABORT) {
934 /* Send an ICMPv6 unreach. */
937 * Unlike above, the mbufs need to line up with the ip6 hdr,
938 * as the contents are read. We need to m_adj() the
940 * The mbuf will however be thrown away so we can adjust it.
941 * Remember we did an m_pullup on it already so we
942 * can make some assumptions about contiguousness.
945 m_adj(m, args->L3offset);
947 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
958 * sends a reject message, consuming the mbuf passed as an argument.
961 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
965 /* XXX When ip is not guaranteed to be at mtod() we will
966 * need to account for this */
967 * The mbuf will however be thrown away so we can adjust it.
968 * Remember we did an m_pullup on it already so we
969 * can make some assumptions about contiguousness.
972 m_adj(m, args->L3offset);
974 if (code != ICMP_REJECT_RST && code != ICMP_REJECT_ABORT) {
975 /* Send an ICMP unreach */
976 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
977 } else if (code == ICMP_REJECT_RST && args->f_id.proto == IPPROTO_TCP) {
978 struct tcphdr *const tcp =
979 L3HDR(struct tcphdr, mtod(args->m, struct ip *));
980 if ( (tcp->th_flags & TH_RST) == 0) {
982 m = ipfw_send_pkt(args->m, &(args->f_id),
983 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
984 tcp->th_flags | TH_RST);
986 ip_output(m, NULL, NULL, 0, NULL, NULL);
989 } else if (code == ICMP_REJECT_ABORT &&
990 args->f_id.proto == IPPROTO_SCTP) {
992 struct sctphdr *sctp;
993 struct sctp_chunkhdr *chunk;
994 struct sctp_init *init;
998 sctp = L3HDR(struct sctphdr, mtod(args->m, struct ip *));
1000 v_tag = ntohl(sctp->v_tag);
1001 if (iplen >= (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1002 sizeof(struct sctp_chunkhdr)) {
1003 /* Look at the first chunk header if available */
1004 chunk = (struct sctp_chunkhdr *)(sctp + 1);
1005 switch (chunk->chunk_type) {
1006 case SCTP_INITIATION:
1008 * Packets containing an INIT chunk MUST have
1015 /* INIT chunk MUST NOT be bundled */
1017 (ip->ip_hl << 2) + sizeof(struct sctphdr) +
1018 ntohs(chunk->chunk_length) + 3) {
1021 /* Use the initiate tag if available */
1022 if ((iplen >= (ip->ip_hl << 2) +
1023 sizeof(struct sctphdr) +
1024 sizeof(struct sctp_chunkhdr) +
1025 offsetof(struct sctp_init, a_rwnd))) {
1026 init = (struct sctp_init *)(chunk + 1);
1027 v_tag = ntohl(init->initiate_tag);
1031 case SCTP_ABORT_ASSOCIATION:
1033 * If the packet contains an ABORT chunk, don't
1035 * XXX: We should search through all chunks,
1036 * but don't do to avoid attacks.
1045 m = ipfw_send_abort(args->m, &(args->f_id), v_tag,
1049 ip_output(m, NULL, NULL, 0, NULL, NULL);
1057 * Support for uid/gid/jail lookup. These tests are expensive
1058 * (because we may need to look into the list of active sockets)
1059 * so we cache the results. ugid_lookupp is 0 if we have not
1060 * yet done a lookup, 1 if we succeeded, and -1 if we tried
1061 * and failed. The function always returns the match value.
1062 * We could actually spare the variable and use *uc, setting
1063 * it to '(void *)check_uidgid if we have no info, NULL if
1064 * we tried and failed, or any other value if successful.
1067 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
1070 #if defined(USERSPACE)
1071 return 0; // not supported in userspace
1075 return cred_check(insn, proto, oif,
1076 dst_ip, dst_port, src_ip, src_port,
1077 (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
1079 struct in_addr src_ip, dst_ip;
1080 struct inpcbinfo *pi;
1081 struct ipfw_flow_id *id;
1082 struct inpcb *pcb, *inp;
1090 * Check to see if the UDP or TCP stack supplied us with
1091 * the PCB. If so, rather then holding a lock and looking
1092 * up the PCB, we can use the one that was supplied.
1094 if (inp && *ugid_lookupp == 0) {
1095 INP_LOCK_ASSERT(inp);
1096 if (inp->inp_socket != NULL) {
1097 *uc = crhold(inp->inp_cred);
1103 * If we have already been here and the packet has no
1104 * PCB entry associated with it, then we can safely
1105 * assume that this is a no match.
1107 if (*ugid_lookupp == -1)
1109 if (id->proto == IPPROTO_TCP) {
1112 } else if (id->proto == IPPROTO_UDP) {
1113 lookupflags = INPLOOKUP_WILDCARD;
1115 } else if (id->proto == IPPROTO_UDPLITE) {
1116 lookupflags = INPLOOKUP_WILDCARD;
1117 pi = &V_ulitecbinfo;
1120 lookupflags |= INPLOOKUP_RLOCKPCB;
1122 if (*ugid_lookupp == 0) {
1123 if (id->addr_type == 6) {
1125 if (args->flags & IPFW_ARGS_IN)
1126 pcb = in6_pcblookup_mbuf(pi,
1127 &id->src_ip6, htons(id->src_port),
1128 &id->dst_ip6, htons(id->dst_port),
1129 lookupflags, NULL, args->m);
1131 pcb = in6_pcblookup_mbuf(pi,
1132 &id->dst_ip6, htons(id->dst_port),
1133 &id->src_ip6, htons(id->src_port),
1134 lookupflags, args->ifp, args->m);
1140 src_ip.s_addr = htonl(id->src_ip);
1141 dst_ip.s_addr = htonl(id->dst_ip);
1142 if (args->flags & IPFW_ARGS_IN)
1143 pcb = in_pcblookup_mbuf(pi,
1144 src_ip, htons(id->src_port),
1145 dst_ip, htons(id->dst_port),
1146 lookupflags, NULL, args->m);
1148 pcb = in_pcblookup_mbuf(pi,
1149 dst_ip, htons(id->dst_port),
1150 src_ip, htons(id->src_port),
1151 lookupflags, args->ifp, args->m);
1154 INP_RLOCK_ASSERT(pcb);
1155 *uc = crhold(pcb->inp_cred);
1159 if (*ugid_lookupp == 0) {
1161 * We tried and failed, set the variable to -1
1162 * so we will not try again on this packet.
1168 if (insn->o.opcode == O_UID)
1169 match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
1170 else if (insn->o.opcode == O_GID)
1171 match = groupmember((gid_t)insn->d[0], *uc);
1172 else if (insn->o.opcode == O_JAIL)
1173 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
1175 #endif /* __FreeBSD__ */
1176 #endif /* not supported in userspace */
1180 * Helper function to set args with info on the rule after the matching
1181 * one. slot is precise, whereas we guess rule_id as they are
1182 * assigned sequentially.
1185 set_match(struct ip_fw_args *args, int slot,
1186 struct ip_fw_chain *chain)
1188 args->rule.chain_id = chain->id;
1189 args->rule.slot = slot + 1; /* we use 0 as a marker */
1190 args->rule.rule_id = 1 + chain->map[slot]->id;
1191 args->rule.rulenum = chain->map[slot]->rulenum;
1192 args->flags |= IPFW_ARGS_REF;
1195 #ifndef LINEAR_SKIPTO
1197 * Helper function to enable cached rule lookups using
1198 * cached_id and cached_pos fields in ipfw rule.
1201 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1202 int tablearg, int jump_backwards)
1206 /* If possible use cached f_pos (in f->cached_pos),
1207 * whose version is written in f->cached_id
1208 * (horrible hacks to avoid changing the ABI).
1210 if (num != IP_FW_TARG && f->cached_id == chain->id)
1211 f_pos = f->cached_pos;
1213 int i = IP_FW_ARG_TABLEARG(chain, num, skipto);
1214 /* make sure we do not jump backward */
1215 if (jump_backwards == 0 && i <= f->rulenum)
1217 if (chain->idxmap != NULL)
1218 f_pos = chain->idxmap[i];
1220 f_pos = ipfw_find_rule(chain, i, 0);
1221 /* update the cache */
1222 if (num != IP_FW_TARG) {
1223 f->cached_id = chain->id;
1224 f->cached_pos = f_pos;
1232 * Helper function to enable real fast rule lookups.
1235 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1236 int tablearg, int jump_backwards)
1240 num = IP_FW_ARG_TABLEARG(chain, num, skipto);
1241 /* make sure we do not jump backward */
1242 if (jump_backwards == 0 && num <= f->rulenum)
1243 num = f->rulenum + 1;
1244 f_pos = chain->idxmap[num];
1250 #define TARG(k, f) IP_FW_ARG_TABLEARG(chain, k, f)
1252 * The main check routine for the firewall.
1254 * All arguments are in args so we can modify them and return them
1255 * back to the caller.
1259 * args->m (in/out) The packet; we set to NULL when/if we nuke it.
1260 * Starts with the IP header.
1261 * args->L3offset Number of bytes bypassed if we came from L2.
1262 * e.g. often sizeof(eh) ** NOTYET **
1263 * args->ifp Incoming or outgoing interface.
1264 * args->divert_rule (in/out)
1265 * Skip up to the first rule past this rule number;
1266 * upon return, non-zero port number for divert or tee.
1268 * args->rule Pointer to the last matching rule (in/out)
1269 * args->next_hop Socket we are forwarding to (out).
1270 * args->next_hop6 IPv6 next hop we are forwarding to (out).
1271 * args->f_id Addresses grabbed from the packet (out)
1272 * args->rule.info a cookie depending on rule action
1276 * IP_FW_PASS the packet must be accepted
1277 * IP_FW_DENY the packet must be dropped
1278 * IP_FW_DIVERT divert packet, port in m_tag
1279 * IP_FW_TEE tee packet, port in m_tag
1280 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie
1281 * IP_FW_NETGRAPH into netgraph, cookie args->cookie
1282 * args->rule contains the matching rule,
1283 * args->rule.info has additional information.
1287 ipfw_chk(struct ip_fw_args *args)
1291 * Local variables holding state while processing a packet:
1293 * IMPORTANT NOTE: to speed up the processing of rules, there
1294 * are some assumption on the values of the variables, which
1295 * are documented here. Should you change them, please check
1296 * the implementation of the various instructions to make sure
1297 * that they still work.
1299 * m | args->m Pointer to the mbuf, as received from the caller.
1300 * It may change if ipfw_chk() does an m_pullup, or if it
1301 * consumes the packet because it calls send_reject().
1302 * XXX This has to change, so that ipfw_chk() never modifies
1303 * or consumes the buffer.
1305 * args->mem Pointer to contigous memory chunk.
1306 * ip Is the beginning of the ip(4 or 6) header.
1307 * eh Ethernet header in case if input is Layer2.
1311 struct ether_header *eh;
1314 * For rules which contain uid/gid or jail constraints, cache
1315 * a copy of the users credentials after the pcb lookup has been
1316 * executed. This will speed up the processing of rules with
1317 * these types of constraints, as well as decrease contention
1318 * on pcb related locks.
1321 struct bsd_ucred ucred_cache;
1323 struct ucred *ucred_cache = NULL;
1325 int ucred_lookup = 0;
1326 int f_pos = 0; /* index of current rule in the array */
1328 struct ifnet *oif, *iif;
1331 * hlen The length of the IP header.
1333 u_int hlen = 0; /* hlen >0 means we have an IP pkt */
1336 * offset The offset of a fragment. offset != 0 means that
1337 * we have a fragment at this offset of an IPv4 packet.
1338 * offset == 0 means that (if this is an IPv4 packet)
1339 * this is the first or only fragment.
1340 * For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
1341 * or there is a single packet fragment (fragment header added
1342 * without needed). We will treat a single packet fragment as if
1343 * there was no fragment header (or log/block depending on the
1344 * V_fw_permit_single_frag6 sysctl setting).
1347 u_short ip6f_mf = 0;
1350 * Local copies of addresses. They are only valid if we have
1353 * proto The protocol. Set to 0 for non-ip packets,
1354 * or to the protocol read from the packet otherwise.
1355 * proto != 0 means that we have an IPv4 packet.
1357 * src_port, dst_port port numbers, in HOST format. Only
1358 * valid for TCP and UDP packets.
1360 * src_ip, dst_ip ip addresses, in NETWORK format.
1361 * Only valid for IPv4 packets.
1364 uint16_t src_port, dst_port; /* NOTE: host format */
1365 struct in_addr src_ip, dst_ip; /* NOTE: network format */
1369 struct ipfw_dyn_info dyn_info;
1370 struct ip_fw *q = NULL;
1371 struct ip_fw_chain *chain = &V_layer3_chain;
1374 * We store in ulp a pointer to the upper layer protocol header.
1375 * In the ipv4 case this is easy to determine from the header,
1376 * but for ipv6 we might have some additional headers in the middle.
1377 * ulp is NULL if not found.
1379 void *ulp = NULL; /* upper layer protocol pointer. */
1381 /* XXX ipv6 variables */
1383 uint8_t icmp6_type = 0;
1384 uint16_t ext_hd = 0; /* bits vector for extension header filtering */
1385 /* end of ipv6 variables */
1389 int done = 0; /* flag to exit the outer loop */
1393 if ((mem = (args->flags & IPFW_ARGS_LENMASK))) {
1394 if (args->flags & IPFW_ARGS_ETHER) {
1395 eh = (struct ether_header *)args->mem;
1396 if (eh->ether_type == htons(ETHERTYPE_VLAN))
1398 ((struct ether_vlan_header *)eh + 1);
1400 ip = (struct ip *)(eh + 1);
1403 ip = (struct ip *)args->mem;
1405 pktlen = IPFW_ARGS_LENGTH(args->flags);
1406 args->f_id.fib = args->ifp->if_fib; /* best guess */
1409 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
1410 return (IP_FW_PASS); /* accept */
1411 if (args->flags & IPFW_ARGS_ETHER) {
1412 /* We need some amount of data to be contiguous. */
1413 if (m->m_len < min(m->m_pkthdr.len, max_protohdr) &&
1414 (args->m = m = m_pullup(m, min(m->m_pkthdr.len,
1415 max_protohdr))) == NULL)
1417 eh = mtod(m, struct ether_header *);
1418 ip = (struct ip *)(eh + 1);
1421 ip = mtod(m, struct ip *);
1423 pktlen = m->m_pkthdr.len;
1424 args->f_id.fib = M_GETFIB(m); /* mbuf not altered */
1427 dst_ip.s_addr = 0; /* make sure it is initialized */
1428 src_ip.s_addr = 0; /* make sure it is initialized */
1429 src_port = dst_port = 0;
1431 DYN_INFO_INIT(&dyn_info);
1433 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1434 * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1435 * pointer might become stale after other pullups (but we never use it
1438 #define PULLUP_TO(_len, p, T) PULLUP_LEN(_len, p, sizeof(T))
1439 #define EHLEN (eh != NULL ? ((char *)ip - (char *)eh) : 0)
1440 #define PULLUP_LEN(_len, p, T) \
1442 int x = (_len) + T + EHLEN; \
1444 MPASS(pktlen >= x); \
1445 p = (char *)args->mem + (_len) + EHLEN; \
1447 if (__predict_false((m)->m_len < x)) { \
1448 args->m = m = m_pullup(m, x); \
1450 goto pullup_failed; \
1452 p = mtod(m, char *) + (_len) + EHLEN; \
1456 * In case pointers got stale after pullups, update them.
1458 #define UPDATE_POINTERS() \
1462 eh = mtod(m, struct ether_header *); \
1463 ip = (struct ip *)(eh + 1); \
1465 ip = mtod(m, struct ip *); \
1470 /* Identify IP packets and fill up variables. */
1471 if (pktlen >= sizeof(struct ip6_hdr) &&
1472 (eh == NULL || eh->ether_type == htons(ETHERTYPE_IPV6)) &&
1474 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1477 args->flags |= IPFW_ARGS_IP6;
1478 hlen = sizeof(struct ip6_hdr);
1479 proto = ip6->ip6_nxt;
1480 /* Search extension headers to find upper layer protocols */
1481 while (ulp == NULL && offset == 0) {
1483 case IPPROTO_ICMPV6:
1484 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1485 icmp6_type = ICMP6(ulp)->icmp6_type;
1489 PULLUP_TO(hlen, ulp, struct tcphdr);
1490 dst_port = TCP(ulp)->th_dport;
1491 src_port = TCP(ulp)->th_sport;
1492 /* save flags for dynamic rules */
1493 args->f_id._flags = TCP(ulp)->th_flags;
1497 if (pktlen >= hlen + sizeof(struct sctphdr) +
1498 sizeof(struct sctp_chunkhdr) +
1499 offsetof(struct sctp_init, a_rwnd))
1500 PULLUP_LEN(hlen, ulp,
1501 sizeof(struct sctphdr) +
1502 sizeof(struct sctp_chunkhdr) +
1503 offsetof(struct sctp_init, a_rwnd));
1504 else if (pktlen >= hlen + sizeof(struct sctphdr))
1505 PULLUP_LEN(hlen, ulp, pktlen - hlen);
1507 PULLUP_LEN(hlen, ulp,
1508 sizeof(struct sctphdr));
1509 src_port = SCTP(ulp)->src_port;
1510 dst_port = SCTP(ulp)->dest_port;
1514 case IPPROTO_UDPLITE:
1515 PULLUP_TO(hlen, ulp, struct udphdr);
1516 dst_port = UDP(ulp)->uh_dport;
1517 src_port = UDP(ulp)->uh_sport;
1520 case IPPROTO_HOPOPTS: /* RFC 2460 */
1521 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1522 ext_hd |= EXT_HOPOPTS;
1523 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1524 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1528 case IPPROTO_ROUTING: /* RFC 2460 */
1529 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1530 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1532 ext_hd |= EXT_RTHDR0;
1535 ext_hd |= EXT_RTHDR2;
1539 printf("IPFW2: IPV6 - Unknown "
1540 "Routing Header type(%d)\n",
1541 ((struct ip6_rthdr *)
1543 if (V_fw_deny_unknown_exthdrs)
1544 return (IP_FW_DENY);
1547 ext_hd |= EXT_ROUTING;
1548 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1549 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1553 case IPPROTO_FRAGMENT: /* RFC 2460 */
1554 PULLUP_TO(hlen, ulp, struct ip6_frag);
1555 ext_hd |= EXT_FRAGMENT;
1556 hlen += sizeof (struct ip6_frag);
1557 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1558 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1560 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1562 if (V_fw_permit_single_frag6 == 0 &&
1563 offset == 0 && ip6f_mf == 0) {
1565 printf("IPFW2: IPV6 - Invalid "
1566 "Fragment Header\n");
1567 if (V_fw_deny_unknown_exthdrs)
1568 return (IP_FW_DENY);
1572 ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1576 case IPPROTO_DSTOPTS: /* RFC 2460 */
1577 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1578 ext_hd |= EXT_DSTOPTS;
1579 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1580 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1584 case IPPROTO_AH: /* RFC 2402 */
1585 PULLUP_TO(hlen, ulp, struct ip6_ext);
1587 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1588 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1592 case IPPROTO_ESP: /* RFC 2406 */
1593 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
1594 /* Anything past Seq# is variable length and
1595 * data past this ext. header is encrypted. */
1599 case IPPROTO_NONE: /* RFC 2460 */
1601 * Packet ends here, and IPv6 header has
1602 * already been pulled up. If ip6e_len!=0
1603 * then octets must be ignored.
1605 ulp = ip; /* non-NULL to get out of loop. */
1608 case IPPROTO_OSPFIGP:
1609 /* XXX OSPF header check? */
1610 PULLUP_TO(hlen, ulp, struct ip6_ext);
1614 /* XXX PIM header check? */
1615 PULLUP_TO(hlen, ulp, struct pim);
1618 case IPPROTO_GRE: /* RFC 1701 */
1619 /* XXX GRE header check? */
1620 PULLUP_TO(hlen, ulp, struct grehdr);
1624 PULLUP_TO(hlen, ulp, offsetof(
1625 struct carp_header, carp_counter));
1626 if (CARP_ADVERTISEMENT !=
1627 ((struct carp_header *)ulp)->carp_type)
1628 return (IP_FW_DENY);
1631 case IPPROTO_IPV6: /* RFC 2893 */
1632 PULLUP_TO(hlen, ulp, struct ip6_hdr);
1635 case IPPROTO_IPV4: /* RFC 2893 */
1636 PULLUP_TO(hlen, ulp, struct ip);
1641 printf("IPFW2: IPV6 - Unknown "
1642 "Extension Header(%d), ext_hd=%x\n",
1644 if (V_fw_deny_unknown_exthdrs)
1645 return (IP_FW_DENY);
1646 PULLUP_TO(hlen, ulp, struct ip6_ext);
1651 ip6 = (struct ip6_hdr *)ip;
1652 args->f_id.addr_type = 6;
1653 args->f_id.src_ip6 = ip6->ip6_src;
1654 args->f_id.dst_ip6 = ip6->ip6_dst;
1655 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1656 iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6);
1657 } else if (pktlen >= sizeof(struct ip) &&
1658 (eh == NULL || eh->ether_type == htons(ETHERTYPE_IP)) &&
1661 args->flags |= IPFW_ARGS_IP4;
1662 hlen = ip->ip_hl << 2;
1664 * Collect parameters into local variables for faster
1668 src_ip = ip->ip_src;
1669 dst_ip = ip->ip_dst;
1670 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1671 iplen = ntohs(ip->ip_len);
1676 PULLUP_TO(hlen, ulp, struct tcphdr);
1677 dst_port = TCP(ulp)->th_dport;
1678 src_port = TCP(ulp)->th_sport;
1679 /* save flags for dynamic rules */
1680 args->f_id._flags = TCP(ulp)->th_flags;
1684 if (pktlen >= hlen + sizeof(struct sctphdr) +
1685 sizeof(struct sctp_chunkhdr) +
1686 offsetof(struct sctp_init, a_rwnd))
1687 PULLUP_LEN(hlen, ulp,
1688 sizeof(struct sctphdr) +
1689 sizeof(struct sctp_chunkhdr) +
1690 offsetof(struct sctp_init, a_rwnd));
1691 else if (pktlen >= hlen + sizeof(struct sctphdr))
1692 PULLUP_LEN(hlen, ulp, pktlen - hlen);
1694 PULLUP_LEN(hlen, ulp,
1695 sizeof(struct sctphdr));
1696 src_port = SCTP(ulp)->src_port;
1697 dst_port = SCTP(ulp)->dest_port;
1701 case IPPROTO_UDPLITE:
1702 PULLUP_TO(hlen, ulp, struct udphdr);
1703 dst_port = UDP(ulp)->uh_dport;
1704 src_port = UDP(ulp)->uh_sport;
1708 PULLUP_TO(hlen, ulp, struct icmphdr);
1709 //args->f_id.flags = ICMP(ulp)->icmp_type;
1718 args->f_id.addr_type = 4;
1719 args->f_id.src_ip = ntohl(src_ip.s_addr);
1720 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1723 dst_ip.s_addr = src_ip.s_addr = 0;
1725 args->f_id.addr_type = 1; /* XXX */
1728 pktlen = iplen < pktlen ? iplen: pktlen;
1730 /* Properly initialize the rest of f_id */
1731 args->f_id.proto = proto;
1732 args->f_id.src_port = src_port = ntohs(src_port);
1733 args->f_id.dst_port = dst_port = ntohs(dst_port);
1735 IPFW_PF_RLOCK(chain);
1736 if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1737 IPFW_PF_RUNLOCK(chain);
1738 return (IP_FW_PASS); /* accept */
1740 if (args->flags & IPFW_ARGS_REF) {
1742 * Packet has already been tagged as a result of a previous
1743 * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1744 * REASS, NETGRAPH, DIVERT/TEE...)
1745 * Validate the slot and continue from the next one
1746 * if still present, otherwise do a lookup.
1748 f_pos = (args->rule.chain_id == chain->id) ?
1750 ipfw_find_rule(chain, args->rule.rulenum,
1751 args->rule.rule_id);
1756 if (args->flags & IPFW_ARGS_IN) {
1760 MPASS(args->flags & IPFW_ARGS_OUT);
1761 iif = mem ? NULL : m->m_pkthdr.rcvif;
1766 * Now scan the rules, and parse microinstructions for each rule.
1767 * We have two nested loops and an inner switch. Sometimes we
1768 * need to break out of one or both loops, or re-enter one of
1769 * the loops with updated variables. Loop variables are:
1771 * f_pos (outer loop) points to the current rule.
1772 * On output it points to the matching rule.
1773 * done (outer loop) is used as a flag to break the loop.
1774 * l (inner loop) residual length of current rule.
1775 * cmd points to the current microinstruction.
1777 * We break the inner loop by setting l=0 and possibly
1778 * cmdlen=0 if we don't want to advance cmd.
1779 * We break the outer loop by setting done=1
1780 * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1783 for (; f_pos < chain->n_rules; f_pos++) {
1785 uint32_t tablearg = 0;
1786 int l, cmdlen, skip_or; /* skip rest of OR block */
1789 f = chain->map[f_pos];
1790 if (V_set_disable & (1 << f->set) )
1794 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1795 l -= cmdlen, cmd += cmdlen) {
1799 * check_body is a jump target used when we find a
1800 * CHECK_STATE, and need to jump to the body of
1805 cmdlen = F_LEN(cmd);
1807 * An OR block (insn_1 || .. || insn_n) has the
1808 * F_OR bit set in all but the last instruction.
1809 * The first match will set "skip_or", and cause
1810 * the following instructions to be skipped until
1811 * past the one with the F_OR bit clear.
1813 if (skip_or) { /* skip this instruction */
1814 if ((cmd->len & F_OR) == 0)
1815 skip_or = 0; /* next one is good */
1818 match = 0; /* set to 1 if we succeed */
1820 switch (cmd->opcode) {
1822 * The first set of opcodes compares the packet's
1823 * fields with some pattern, setting 'match' if a
1824 * match is found. At the end of the loop there is
1825 * logic to deal with F_NOT and F_OR flags associated
1833 printf("ipfw: opcode %d unimplemented\n",
1841 * We only check offset == 0 && proto != 0,
1842 * as this ensures that we have a
1843 * packet with the ports info.
1847 if (proto == IPPROTO_TCP ||
1848 proto == IPPROTO_UDP ||
1849 proto == IPPROTO_UDPLITE)
1850 match = check_uidgid(
1851 (ipfw_insn_u32 *)cmd,
1852 args, &ucred_lookup,
1856 (void *)&ucred_cache);
1861 match = iface_match(iif, (ipfw_insn_if *)cmd,
1866 match = iface_match(oif, (ipfw_insn_if *)cmd,
1871 match = iface_match(args->ifp,
1872 (ipfw_insn_if *)cmd, chain, &tablearg);
1876 if (args->flags & IPFW_ARGS_ETHER) {
1877 u_int32_t *want = (u_int32_t *)
1878 ((ipfw_insn_mac *)cmd)->addr;
1879 u_int32_t *mask = (u_int32_t *)
1880 ((ipfw_insn_mac *)cmd)->mask;
1881 u_int32_t *hdr = (u_int32_t *)eh;
1884 ( want[0] == (hdr[0] & mask[0]) &&
1885 want[1] == (hdr[1] & mask[1]) &&
1886 want[2] == (hdr[2] & mask[2]) );
1891 if (args->flags & IPFW_ARGS_ETHER) {
1893 ((ipfw_insn_u16 *)cmd)->ports;
1896 for (i = cmdlen - 1; !match && i>0;
1899 (ntohs(eh->ether_type) >=
1901 ntohs(eh->ether_type) <=
1907 match = (offset != 0);
1910 case O_IN: /* "out" is "not in" */
1911 match = (oif == NULL);
1915 match = (args->flags & IPFW_ARGS_ETHER);
1919 if ((args->flags & IPFW_ARGS_REF) == 0)
1922 * For diverted packets, args->rule.info
1923 * contains the divert port (in host format)
1924 * reason and direction.
1926 match = ((args->rule.info & IPFW_IS_MASK) ==
1927 IPFW_IS_DIVERT) && (
1928 ((args->rule.info & IPFW_INFO_IN) ?
1934 * We do not allow an arg of 0 so the
1935 * check of "proto" only suffices.
1937 match = (proto == cmd->arg1);
1942 (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1946 case O_IP_DST_LOOKUP:
1952 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1953 /* Determine lookup key type */
1954 vidx = ((ipfw_insn_u32 *)cmd)->d[1];
1955 if (vidx != 4 /* uid */ &&
1956 vidx != 5 /* jail */ &&
1957 is_ipv6 == 0 && is_ipv4 == 0)
1959 /* Determine key length */
1960 if (vidx == 0 /* dst-ip */ ||
1961 vidx == 1 /* src-ip */)
1963 sizeof(struct in6_addr):
1966 keylen = sizeof(key);
1969 if (vidx == 0 /* dst-ip */)
1970 pkey = is_ipv4 ? (void *)&dst_ip:
1971 (void *)&args->f_id.dst_ip6;
1972 else if (vidx == 1 /* src-ip */)
1973 pkey = is_ipv4 ? (void *)&src_ip:
1974 (void *)&args->f_id.src_ip6;
1975 else if (vidx == 6 /* dscp */) {
1977 key = ip->ip_tos >> 2;
1979 key = args->f_id.flow_id6;
1980 key = (key & 0x0f) << 2 |
1981 (key & 0xf000) >> 14;
1984 } else if (vidx == 2 /* dst-port */ ||
1985 vidx == 3 /* src-port */) {
1986 /* Skip fragments */
1989 /* Skip proto without ports */
1990 if (proto != IPPROTO_TCP &&
1991 proto != IPPROTO_UDP &&
1992 proto != IPPROTO_UDPLITE &&
1993 proto != IPPROTO_SCTP)
1995 if (vidx == 2 /* dst-port */)
2001 else if (vidx == 4 /* uid */ ||
2002 vidx == 5 /* jail */) {
2004 (ipfw_insn_u32 *)cmd,
2005 args, &ucred_lookup,
2008 if (vidx == 4 /* uid */)
2009 key = ucred_cache->cr_uid;
2010 else if (vidx == 5 /* jail */)
2011 key = ucred_cache->cr_prison->pr_id;
2012 #else /* !__FreeBSD__ */
2013 (void *)&ucred_cache);
2014 if (vidx == 4 /* uid */)
2015 key = ucred_cache.uid;
2016 else if (vidx == 5 /* jail */)
2017 key = ucred_cache.xid;
2018 #endif /* !__FreeBSD__ */
2020 #endif /* !USERSPACE */
2023 match = ipfw_lookup_table(chain,
2024 cmd->arg1, keylen, pkey, &vidx);
2030 /* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */
2033 case O_IP_SRC_LOOKUP:
2040 keylen = sizeof(in_addr_t);
2041 if (cmd->opcode == O_IP_DST_LOOKUP)
2045 } else if (is_ipv6) {
2046 keylen = sizeof(struct in6_addr);
2047 if (cmd->opcode == O_IP_DST_LOOKUP)
2048 pkey = &args->f_id.dst_ip6;
2050 pkey = &args->f_id.src_ip6;
2053 match = ipfw_lookup_table(chain, cmd->arg1,
2054 keylen, pkey, &vidx);
2057 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) {
2058 match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2059 TARG_VAL(chain, vidx, tag);
2067 case O_IP_FLOW_LOOKUP:
2070 match = ipfw_lookup_table(chain,
2071 cmd->arg1, 0, &args->f_id, &v);
2072 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
2073 match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2074 TARG_VAL(chain, v, tag);
2083 (cmd->opcode == O_IP_DST_MASK) ?
2084 dst_ip.s_addr : src_ip.s_addr;
2085 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2088 for (; !match && i>0; i-= 2, p+= 2)
2089 match = (p[0] == (a & p[1]));
2095 match = in_localip(src_ip);
2102 ipfw_localip6(&args->f_id.src_ip6);
2109 u_int32_t *d = (u_int32_t *)(cmd+1);
2111 cmd->opcode == O_IP_DST_SET ?
2117 addr -= d[0]; /* subtract base */
2118 match = (addr < cmd->arg1) &&
2119 ( d[ 1 + (addr>>5)] &
2120 (1<<(addr & 0x1f)) );
2126 (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2132 match = in_localip(dst_ip);
2139 ipfw_localip6(&args->f_id.dst_ip6);
2147 * offset == 0 && proto != 0 is enough
2148 * to guarantee that we have a
2149 * packet with port info.
2151 if ((proto == IPPROTO_UDP ||
2152 proto == IPPROTO_UDPLITE ||
2153 proto == IPPROTO_TCP ||
2154 proto == IPPROTO_SCTP) && offset == 0) {
2156 (cmd->opcode == O_IP_SRCPORT) ?
2157 src_port : dst_port ;
2159 ((ipfw_insn_u16 *)cmd)->ports;
2162 for (i = cmdlen - 1; !match && i>0;
2164 match = (x>=p[0] && x<=p[1]);
2169 match = (offset == 0 && proto==IPPROTO_ICMP &&
2170 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
2175 match = is_ipv6 && offset == 0 &&
2176 proto==IPPROTO_ICMPV6 &&
2178 ICMP6(ulp)->icmp6_type,
2179 (ipfw_insn_u32 *)cmd);
2185 ipopts_match(ip, cmd) );
2190 cmd->arg1 == ip->ip_v);
2196 if (is_ipv4) { /* only for IP packets */
2201 if (cmd->opcode == O_IPLEN)
2203 else if (cmd->opcode == O_IPTTL)
2205 else /* must be IPID */
2206 x = ntohs(ip->ip_id);
2208 match = (cmd->arg1 == x);
2211 /* otherwise we have ranges */
2212 p = ((ipfw_insn_u16 *)cmd)->ports;
2214 for (; !match && i>0; i--, p += 2)
2215 match = (x >= p[0] && x <= p[1]);
2219 case O_IPPRECEDENCE:
2221 (cmd->arg1 == (ip->ip_tos & 0xe0)) );
2226 flags_match(cmd, ip->ip_tos));
2234 p = ((ipfw_insn_u32 *)cmd)->d;
2237 x = ip->ip_tos >> 2;
2240 v = &((struct ip6_hdr *)ip)->ip6_vfc;
2241 x = (*v & 0x0F) << 2;
2247 /* DSCP bitmask is stored as low_u32 high_u32 */
2249 match = *(p + 1) & (1 << (x - 32));
2251 match = *p & (1 << x);
2256 if (proto == IPPROTO_TCP && offset == 0) {
2263 struct ip6_hdr *ip6;
2265 ip6 = (struct ip6_hdr *)ip;
2266 if (ip6->ip6_plen == 0) {
2268 * Jumbo payload is not
2277 x = iplen - (ip->ip_hl << 2);
2279 x -= tcp->th_off << 2;
2281 match = (cmd->arg1 == x);
2284 /* otherwise we have ranges */
2285 p = ((ipfw_insn_u16 *)cmd)->ports;
2287 for (; !match && i>0; i--, p += 2)
2288 match = (x >= p[0] && x <= p[1]);
2293 match = (proto == IPPROTO_TCP && offset == 0 &&
2294 flags_match(cmd, TCP(ulp)->th_flags));
2298 if (proto == IPPROTO_TCP && offset == 0 && ulp){
2299 PULLUP_LEN(hlen, ulp,
2300 (TCP(ulp)->th_off << 2));
2301 match = tcpopts_match(TCP(ulp), cmd);
2306 match = (proto == IPPROTO_TCP && offset == 0 &&
2307 ((ipfw_insn_u32 *)cmd)->d[0] ==
2312 match = (proto == IPPROTO_TCP && offset == 0 &&
2313 ((ipfw_insn_u32 *)cmd)->d[0] ==
2318 if (proto == IPPROTO_TCP && offset == 0) {
2323 x = ntohs(TCP(ulp)->th_win);
2325 match = (cmd->arg1 == x);
2328 /* Otherwise we have ranges. */
2329 p = ((ipfw_insn_u16 *)cmd)->ports;
2331 for (; !match && i > 0; i--, p += 2)
2332 match = (x >= p[0] && x <= p[1]);
2337 /* reject packets which have SYN only */
2338 /* XXX should i also check for TH_ACK ? */
2339 match = (proto == IPPROTO_TCP && offset == 0 &&
2340 (TCP(ulp)->th_flags &
2341 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
2347 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
2350 * ALTQ uses mbuf tags from another
2351 * packet filtering system - pf(4).
2352 * We allocate a tag in its format
2353 * and fill it in, pretending to be pf(4).
2356 at = pf_find_mtag(m);
2357 if (at != NULL && at->qid != 0)
2359 mtag = m_tag_get(PACKET_TAG_PF,
2360 sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
2363 * Let the packet fall back to the
2368 m_tag_prepend(m, mtag);
2369 at = (struct pf_mtag *)(mtag + 1);
2370 at->qid = altq->qid;
2376 ipfw_log(chain, f, hlen, args,
2377 offset | ip6f_mf, tablearg, ip);
2382 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
2386 /* Outgoing packets automatically pass/match */
2387 match = (args->flags & IPFW_ARGS_OUT ||
2391 verify_path6(&(args->f_id.src_ip6),
2392 iif, args->f_id.fib) :
2394 verify_path(src_ip, iif, args->f_id.fib)));
2398 /* Outgoing packets automatically pass/match */
2399 match = (hlen > 0 && ((oif != NULL) || (
2402 verify_path6(&(args->f_id.src_ip6),
2403 NULL, args->f_id.fib) :
2405 verify_path(src_ip, NULL, args->f_id.fib))));
2409 /* Outgoing packets automatically pass/match */
2410 if (oif == NULL && hlen > 0 &&
2411 ( (is_ipv4 && in_localaddr(src_ip))
2414 in6_localaddr(&(args->f_id.src_ip6)))
2419 is_ipv6 ? verify_path6(
2420 &(args->f_id.src_ip6), iif,
2423 verify_path(src_ip, iif,
2430 match = (m_tag_find(m,
2431 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
2432 /* otherwise no match */
2438 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
2439 &((ipfw_insn_ip6 *)cmd)->addr6);
2444 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
2445 &((ipfw_insn_ip6 *)cmd)->addr6);
2447 case O_IP6_SRC_MASK:
2448 case O_IP6_DST_MASK:
2452 struct in6_addr *d =
2453 &((ipfw_insn_ip6 *)cmd)->addr6;
2455 for (; !match && i > 0; d += 2,
2456 i -= F_INSN_SIZE(struct in6_addr)
2462 APPLY_MASK(&p, &d[1]);
2464 IN6_ARE_ADDR_EQUAL(&d[0],
2472 flow6id_match(args->f_id.flow_id6,
2473 (ipfw_insn_u32 *) cmd);
2478 (ext_hd & ((ipfw_insn *) cmd)->arg1);
2492 uint32_t tag = TARG(cmd->arg1, tag);
2494 /* Packet is already tagged with this tag? */
2495 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
2497 /* We have `untag' action when F_NOT flag is
2498 * present. And we must remove this mtag from
2499 * mbuf and reset `match' to zero (`match' will
2500 * be inversed later).
2501 * Otherwise we should allocate new mtag and
2502 * push it into mbuf.
2504 if (cmd->len & F_NOT) { /* `untag' action */
2506 m_tag_delete(m, mtag);
2510 mtag = m_tag_alloc( MTAG_IPFW,
2513 m_tag_prepend(m, mtag);
2520 case O_FIB: /* try match the specified fib */
2521 if (args->f_id.fib == cmd->arg1)
2526 #ifndef USERSPACE /* not supported in userspace */
2527 struct inpcb *inp = args->inp;
2528 struct inpcbinfo *pi;
2530 if (is_ipv6) /* XXX can we remove this ? */
2533 if (proto == IPPROTO_TCP)
2535 else if (proto == IPPROTO_UDP)
2537 else if (proto == IPPROTO_UDPLITE)
2538 pi = &V_ulitecbinfo;
2543 * XXXRW: so_user_cookie should almost
2544 * certainly be inp_user_cookie?
2547 /* For incoming packet, lookup up the
2548 inpcb using the src/dest ip/port tuple */
2550 inp = in_pcblookup(pi,
2551 src_ip, htons(src_port),
2552 dst_ip, htons(dst_port),
2553 INPLOOKUP_RLOCKPCB, NULL);
2556 inp->inp_socket->so_user_cookie;
2562 if (inp->inp_socket) {
2564 inp->inp_socket->so_user_cookie;
2569 #endif /* !USERSPACE */
2575 uint32_t tag = TARG(cmd->arg1, tag);
2578 match = m_tag_locate(m, MTAG_IPFW,
2583 /* we have ranges */
2584 for (mtag = m_tag_first(m);
2585 mtag != NULL && !match;
2586 mtag = m_tag_next(m, mtag)) {
2590 if (mtag->m_tag_cookie != MTAG_IPFW)
2593 p = ((ipfw_insn_u16 *)cmd)->ports;
2595 for(; !match && i > 0; i--, p += 2)
2597 mtag->m_tag_id >= p[0] &&
2598 mtag->m_tag_id <= p[1];
2604 * The second set of opcodes represents 'actions',
2605 * i.e. the terminal part of a rule once the packet
2606 * matches all previous patterns.
2607 * Typically there is only one action for each rule,
2608 * and the opcode is stored at the end of the rule
2609 * (but there are exceptions -- see below).
2611 * In general, here we set retval and terminate the
2612 * outer loop (would be a 'break 3' in some language,
2613 * but we need to set l=0, done=1)
2616 * O_COUNT and O_SKIPTO actions:
2617 * instead of terminating, we jump to the next rule
2618 * (setting l=0), or to the SKIPTO target (setting
2619 * f/f_len, cmd and l as needed), respectively.
2621 * O_TAG, O_LOG and O_ALTQ action parameters:
2622 * perform some action and set match = 1;
2624 * O_LIMIT and O_KEEP_STATE: these opcodes are
2625 * not real 'actions', and are stored right
2626 * before the 'action' part of the rule (one
2627 * exception is O_SKIP_ACTION which could be
2628 * between these opcodes and 'action' one).
2629 * These opcodes try to install an entry in the
2630 * state tables; if successful, we continue with
2631 * the next opcode (match=1; break;), otherwise
2632 * the packet must be dropped (set retval,
2633 * break loops with l=0, done=1)
2635 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2636 * cause a lookup of the state table, and a jump
2637 * to the 'action' part of the parent rule
2638 * if an entry is found, or
2639 * (CHECK_STATE only) a jump to the next rule if
2640 * the entry is not found.
2641 * The result of the lookup is cached so that
2642 * further instances of these opcodes become NOPs.
2643 * The jump to the next rule is done by setting
2646 * O_SKIP_ACTION: this opcode is not a real 'action'
2647 * either, and is stored right before the 'action'
2648 * part of the rule, right after the O_KEEP_STATE
2649 * opcode. It causes match failure so the real
2650 * 'action' could be executed only if the rule
2651 * is checked via dynamic rule from the state
2652 * table, as in such case execution starts
2653 * from the true 'action' opcode directly.
2658 if (ipfw_dyn_install_state(chain, f,
2659 (ipfw_insn_limit *)cmd, args, ulp,
2660 pktlen, &dyn_info, tablearg)) {
2661 /* error or limit violation */
2662 retval = IP_FW_DENY;
2663 l = 0; /* exit inner loop */
2664 done = 1; /* exit outer loop */
2672 * dynamic rules are checked at the first
2673 * keep-state or check-state occurrence,
2674 * with the result being stored in dyn_info.
2675 * The compiler introduces a PROBE_STATE
2676 * instruction for us when we have a
2677 * KEEP_STATE (because PROBE_STATE needs
2680 if (DYN_LOOKUP_NEEDED(&dyn_info, cmd) &&
2681 (q = ipfw_dyn_lookup_state(args, ulp,
2682 pktlen, cmd, &dyn_info)) != NULL) {
2684 * Found dynamic entry, jump to the
2685 * 'action' part of the parent rule
2686 * by setting f, cmd, l and clearing
2690 f_pos = dyn_info.f_pos;
2691 cmd = ACTION_PTR(f);
2692 l = f->cmd_len - f->act_ofs;
2698 * Dynamic entry not found. If CHECK_STATE,
2699 * skip to next rule, if PROBE_STATE just
2700 * ignore and continue with next opcode.
2702 if (cmd->opcode == O_CHECK_STATE)
2703 l = 0; /* exit inner loop */
2708 match = 0; /* skip to the next rule */
2709 l = 0; /* exit inner loop */
2713 retval = 0; /* accept */
2714 l = 0; /* exit inner loop */
2715 done = 1; /* exit outer loop */
2720 set_match(args, f_pos, chain);
2721 args->rule.info = TARG(cmd->arg1, pipe);
2722 if (cmd->opcode == O_PIPE)
2723 args->rule.info |= IPFW_IS_PIPE;
2725 args->rule.info |= IPFW_ONEPASS;
2726 retval = IP_FW_DUMMYNET;
2727 l = 0; /* exit inner loop */
2728 done = 1; /* exit outer loop */
2733 if (args->flags & IPFW_ARGS_ETHER)
2734 break; /* not on layer 2 */
2735 /* otherwise this is terminal */
2736 l = 0; /* exit inner loop */
2737 done = 1; /* exit outer loop */
2738 retval = (cmd->opcode == O_DIVERT) ?
2739 IP_FW_DIVERT : IP_FW_TEE;
2740 set_match(args, f_pos, chain);
2741 args->rule.info = TARG(cmd->arg1, divert);
2745 IPFW_INC_RULE_COUNTER(f, pktlen);
2746 l = 0; /* exit inner loop */
2750 IPFW_INC_RULE_COUNTER(f, pktlen);
2751 f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2753 * Skip disabled rules, and re-enter
2754 * the inner loop with the correct
2755 * f_pos, f, l and cmd.
2756 * Also clear cmdlen and skip_or
2758 for (; f_pos < chain->n_rules - 1 &&
2760 (1 << chain->map[f_pos]->set));
2763 /* Re-enter the inner loop at the skipto rule. */
2764 f = chain->map[f_pos];
2771 break; /* not reached */
2773 case O_CALLRETURN: {
2775 * Implementation of `subroutine' call/return,
2776 * in the stack carried in an mbuf tag. This
2777 * is different from `skipto' in that any call
2778 * address is possible (`skipto' must prevent
2779 * backward jumps to avoid endless loops).
2780 * We have `return' action when F_NOT flag is
2781 * present. The `m_tag_id' field is used as
2785 uint16_t jmpto, *stack;
2787 #define IS_CALL ((cmd->len & F_NOT) == 0)
2788 #define IS_RETURN ((cmd->len & F_NOT) != 0)
2790 * Hand-rolled version of m_tag_locate() with
2792 * If not already tagged, allocate new tag.
2794 mtag = m_tag_first(m);
2795 while (mtag != NULL) {
2796 if (mtag->m_tag_cookie ==
2799 mtag = m_tag_next(m, mtag);
2801 if (mtag == NULL && IS_CALL) {
2802 mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2803 IPFW_CALLSTACK_SIZE *
2804 sizeof(uint16_t), M_NOWAIT);
2806 m_tag_prepend(m, mtag);
2810 * On error both `call' and `return' just
2811 * continue with next rule.
2813 if (IS_RETURN && (mtag == NULL ||
2814 mtag->m_tag_id == 0)) {
2815 l = 0; /* exit inner loop */
2818 if (IS_CALL && (mtag == NULL ||
2819 mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2820 printf("ipfw: call stack error, "
2821 "go to next rule\n");
2822 l = 0; /* exit inner loop */
2826 IPFW_INC_RULE_COUNTER(f, pktlen);
2827 stack = (uint16_t *)(mtag + 1);
2830 * The `call' action may use cached f_pos
2831 * (in f->next_rule), whose version is written
2833 * The `return' action, however, doesn't have
2834 * fixed jump address in cmd->arg1 and can't use
2838 stack[mtag->m_tag_id] = f->rulenum;
2840 f_pos = JUMP(chain, f, cmd->arg1,
2842 } else { /* `return' action */
2844 jmpto = stack[mtag->m_tag_id] + 1;
2845 f_pos = ipfw_find_rule(chain, jmpto, 0);
2849 * Skip disabled rules, and re-enter
2850 * the inner loop with the correct
2851 * f_pos, f, l and cmd.
2852 * Also clear cmdlen and skip_or
2854 for (; f_pos < chain->n_rules - 1 &&
2856 (1 << chain->map[f_pos]->set)); f_pos++)
2858 /* Re-enter the inner loop at the dest rule. */
2859 f = chain->map[f_pos];
2865 break; /* NOTREACHED */
2872 * Drop the packet and send a reject notice
2873 * if the packet is not ICMP (or is an ICMP
2874 * query), and it is not multicast/broadcast.
2876 if (hlen > 0 && is_ipv4 && offset == 0 &&
2877 (proto != IPPROTO_ICMP ||
2878 is_icmp_query(ICMP(ulp))) &&
2879 !(m->m_flags & (M_BCAST|M_MCAST)) &&
2880 !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2881 send_reject(args, cmd->arg1, iplen, ip);
2887 if (hlen > 0 && is_ipv6 &&
2888 ((offset & IP6F_OFF_MASK) == 0) &&
2889 (proto != IPPROTO_ICMPV6 ||
2890 (is_icmp6_query(icmp6_type) == 1)) &&
2891 !(m->m_flags & (M_BCAST|M_MCAST)) &&
2892 !IN6_IS_ADDR_MULTICAST(
2893 &args->f_id.dst_ip6)) {
2895 cmd->opcode == O_REJECT ?
2896 map_icmp_unreach(cmd->arg1):
2898 (struct ip6_hdr *)ip);
2904 retval = IP_FW_DENY;
2905 l = 0; /* exit inner loop */
2906 done = 1; /* exit outer loop */
2910 if (args->flags & IPFW_ARGS_ETHER)
2911 break; /* not valid on layer2 pkts */
2913 dyn_info.direction == MATCH_FORWARD) {
2914 struct sockaddr_in *sa;
2916 sa = &(((ipfw_insn_sa *)cmd)->sa);
2917 if (sa->sin_addr.s_addr == INADDR_ANY) {
2920 * We use O_FORWARD_IP opcode for
2921 * fwd rule with tablearg, but tables
2922 * now support IPv6 addresses. And
2923 * when we are inspecting IPv6 packet,
2924 * we can use nh6 field from
2925 * table_value as next_hop6 address.
2928 struct ip_fw_nh6 *nh6;
2930 args->flags |= IPFW_ARGS_NH6;
2931 nh6 = &args->hopstore6;
2932 nh6->sin6_addr = TARG_VAL(
2933 chain, tablearg, nh6);
2934 nh6->sin6_port = sa->sin_port;
2935 nh6->sin6_scope_id = TARG_VAL(
2936 chain, tablearg, zoneid);
2940 args->flags |= IPFW_ARGS_NH4;
2941 args->hopstore.sin_port =
2943 sa = &args->hopstore;
2944 sa->sin_family = AF_INET;
2945 sa->sin_len = sizeof(*sa);
2946 sa->sin_addr.s_addr = htonl(
2947 TARG_VAL(chain, tablearg,
2951 args->flags |= IPFW_ARGS_NH4PTR;
2952 args->next_hop = sa;
2955 retval = IP_FW_PASS;
2956 l = 0; /* exit inner loop */
2957 done = 1; /* exit outer loop */
2962 if (args->flags & IPFW_ARGS_ETHER)
2963 break; /* not valid on layer2 pkts */
2965 dyn_info.direction == MATCH_FORWARD) {
2966 struct sockaddr_in6 *sin6;
2968 sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2969 args->flags |= IPFW_ARGS_NH6PTR;
2970 args->next_hop6 = sin6;
2972 retval = IP_FW_PASS;
2973 l = 0; /* exit inner loop */
2974 done = 1; /* exit outer loop */
2980 set_match(args, f_pos, chain);
2981 args->rule.info = TARG(cmd->arg1, netgraph);
2983 args->rule.info |= IPFW_ONEPASS;
2984 retval = (cmd->opcode == O_NETGRAPH) ?
2985 IP_FW_NETGRAPH : IP_FW_NGTEE;
2986 l = 0; /* exit inner loop */
2987 done = 1; /* exit outer loop */
2993 IPFW_INC_RULE_COUNTER(f, pktlen);
2994 fib = TARG(cmd->arg1, fib) & 0x7FFF;
2995 if (fib >= rt_numfibs)
2998 args->f_id.fib = fib; /* XXX */
2999 l = 0; /* exit inner loop */
3006 code = TARG(cmd->arg1, dscp) & 0x3F;
3007 l = 0; /* exit inner loop */
3011 old = *(uint16_t *)ip;
3012 ip->ip_tos = (code << 2) |
3013 (ip->ip_tos & 0x03);
3014 ip->ip_sum = cksum_adjust(ip->ip_sum,
3015 old, *(uint16_t *)ip);
3016 } else if (is_ipv6) {
3019 v = &((struct ip6_hdr *)ip)->ip6_vfc;
3020 *v = (*v & 0xF0) | (code >> 2);
3022 *v = (*v & 0x3F) | ((code & 0x03) << 6);
3026 IPFW_INC_RULE_COUNTER(f, pktlen);
3031 l = 0; /* exit inner loop */
3032 done = 1; /* exit outer loop */
3034 * Ensure that we do not invoke NAT handler for
3035 * non IPv4 packets. Libalias expects only IPv4.
3037 if (!is_ipv4 || !IPFW_NAT_LOADED) {
3038 retval = IP_FW_DENY;
3045 args->rule.info = 0;
3046 set_match(args, f_pos, chain);
3047 /* Check if this is 'global' nat rule */
3048 if (cmd->arg1 == IP_FW_NAT44_GLOBAL) {
3049 retval = ipfw_nat_ptr(args, NULL, m);
3052 t = ((ipfw_insn_nat *)cmd)->nat;
3054 nat_id = TARG(cmd->arg1, nat);
3055 t = (*lookup_nat_ptr)(&chain->nat, nat_id);
3058 retval = IP_FW_DENY;
3061 if (cmd->arg1 != IP_FW_TARG)
3062 ((ipfw_insn_nat *)cmd)->nat = t;
3064 retval = ipfw_nat_ptr(args, t, m);
3070 l = 0; /* in any case exit inner loop */
3071 if (is_ipv6) /* IPv6 is not supported yet */
3073 IPFW_INC_RULE_COUNTER(f, pktlen);
3074 ip_off = ntohs(ip->ip_off);
3076 /* if not fragmented, go to next rule */
3077 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
3080 args->m = m = ip_reass(m);
3083 * do IP header checksum fixup.
3085 if (m == NULL) { /* fragment got swallowed */
3086 retval = IP_FW_DENY;
3087 } else { /* good, packet complete */
3090 ip = mtod(m, struct ip *);
3091 hlen = ip->ip_hl << 2;
3093 if (hlen == sizeof(struct ip))
3094 ip->ip_sum = in_cksum_hdr(ip);
3096 ip->ip_sum = in_cksum(m, hlen);
3097 retval = IP_FW_REASS;
3098 args->rule.info = 0;
3099 set_match(args, f_pos, chain);
3101 done = 1; /* exit outer loop */
3104 case O_EXTERNAL_ACTION:
3105 l = 0; /* in any case exit inner loop */
3106 retval = ipfw_run_eaction(chain, args,
3109 * If both @retval and @done are zero,
3110 * consider this as rule matching and
3113 if (retval == 0 && done == 0) {
3114 IPFW_INC_RULE_COUNTER(f, pktlen);
3116 * Reset the result of the last
3117 * dynamic state lookup.
3118 * External action can change
3119 * @args content, and it may be
3120 * used for new state lookup later.
3122 DYN_INFO_INIT(&dyn_info);
3127 panic("-- unknown opcode %d\n", cmd->opcode);
3128 } /* end of switch() on opcodes */
3130 * if we get here with l=0, then match is irrelevant.
3133 if (cmd->len & F_NOT)
3137 if (cmd->len & F_OR)
3140 if (!(cmd->len & F_OR)) /* not an OR block, */
3141 break; /* try next rule */
3144 } /* end of inner loop, scan opcodes */
3150 /* next_rule:; */ /* try next rule */
3152 } /* end of outer for, scan rules */
3155 struct ip_fw *rule = chain->map[f_pos];
3156 /* Update statistics */
3157 IPFW_INC_RULE_COUNTER(rule, pktlen);
3159 retval = IP_FW_DENY;
3160 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
3162 IPFW_PF_RUNLOCK(chain);
3164 if (ucred_cache != NULL)
3165 crfree(ucred_cache);
3171 printf("ipfw: pullup failed\n");
3172 return (IP_FW_DENY);
3176 * Set maximum number of tables that can be used in given VNET ipfw instance.
3180 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
3183 unsigned int ntables;
3185 ntables = V_fw_tables_max;
3187 error = sysctl_handle_int(oidp, &ntables, 0, req);
3188 /* Read operation or some error */
3189 if ((error != 0) || (req->newptr == NULL))
3192 return (ipfw_resize_tables(&V_layer3_chain, ntables));
3196 * Switches table namespace between global and per-set.
3199 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
3204 sets = V_fw_tables_sets;
3206 error = sysctl_handle_int(oidp, &sets, 0, req);
3207 /* Read operation or some error */
3208 if ((error != 0) || (req->newptr == NULL))
3211 return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
3216 * Module and VNET glue
3220 * Stuff that must be initialised only on boot or module load
3228 * Only print out this stuff the first time around,
3229 * when called from the sysinit code.
3235 "initialized, divert %s, nat %s, "
3236 "default to %s, logging ",
3242 #ifdef IPFIREWALL_NAT
3247 default_to_accept ? "accept" : "deny");
3250 * Note: V_xxx variables can be accessed here but the vnet specific
3251 * initializer may not have been called yet for the VIMAGE case.
3252 * Tuneables will have been processed. We will print out values for
3254 * XXX This should all be rationalized AFTER 8.0
3256 if (V_fw_verbose == 0)
3257 printf("disabled\n");
3258 else if (V_verbose_limit == 0)
3259 printf("unlimited\n");
3261 printf("limited to %d packets/entry by default\n",
3264 /* Check user-supplied table count for validness */
3265 if (default_fw_tables > IPFW_TABLES_MAX)
3266 default_fw_tables = IPFW_TABLES_MAX;
3268 ipfw_init_sopt_handler();
3269 ipfw_init_obj_rewriter();
3275 * Called for the removal of the last instance only on module unload.
3281 ipfw_iface_destroy();
3282 ipfw_destroy_sopt_handler();
3283 ipfw_destroy_obj_rewriter();
3284 printf("IP firewall unloaded\n");
3288 * Stuff that must be initialized for every instance
3289 * (including the first of course).
3292 vnet_ipfw_init(const void *unused)
3295 struct ip_fw *rule = NULL;
3296 struct ip_fw_chain *chain;
3298 chain = &V_layer3_chain;
3300 first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3302 /* First set up some values that are compile time options */
3303 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */
3304 V_fw_deny_unknown_exthdrs = 1;
3305 #ifdef IPFIREWALL_VERBOSE
3308 #ifdef IPFIREWALL_VERBOSE_LIMIT
3309 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
3311 #ifdef IPFIREWALL_NAT
3312 LIST_INIT(&chain->nat);
3315 /* Init shared services hash table */
3316 ipfw_init_srv(chain);
3318 ipfw_init_counters();
3319 /* Set initial number of tables */
3320 V_fw_tables_max = default_fw_tables;
3321 error = ipfw_init_tables(chain, first);
3323 printf("ipfw2: setting up tables failed\n");
3324 free(chain->map, M_IPFW);
3329 IPFW_LOCK_INIT(chain);
3331 /* fill and insert the default rule */
3332 rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
3334 rule->cmd[0].len = 1;
3335 rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
3336 chain->default_rule = rule;
3337 ipfw_add_protected_rule(chain, rule, 0);
3339 ipfw_dyn_init(chain);
3340 ipfw_eaction_init(chain, first);
3341 #ifdef LINEAR_SKIPTO
3342 ipfw_init_skipto_cache(chain);
3344 ipfw_bpf_init(first);
3346 /* First set up some values that are compile time options */
3347 V_ipfw_vnet_ready = 1; /* Open for business */
3350 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
3351 * Even if the latter two fail we still keep the module alive
3352 * because the sockopt and layer2 paths are still useful.
3353 * ipfw[6]_hook return 0 on success, ENOENT on failure,
3354 * so we can ignore the exact return value and just set a flag.
3356 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
3357 * changes in the underlying (per-vnet) variables trigger
3358 * immediate hook()/unhook() calls.
3359 * In layer2 we have the same behaviour, except that V_ether_ipfw
3360 * is checked on each packet because there are no pfil hooks.
3362 V_ip_fw_ctl_ptr = ipfw_ctl3;
3363 error = ipfw_attach_hooks(1);
3368 * Called for the removal of each instance.
3371 vnet_ipfw_uninit(const void *unused)
3374 struct ip_fw_chain *chain = &V_layer3_chain;
3377 V_ipfw_vnet_ready = 0; /* tell new callers to go away */
3379 * disconnect from ipv4, ipv6, layer2 and sockopt.
3380 * Then grab, release and grab again the WLOCK so we make
3381 * sure the update is propagated and nobody will be in.
3383 (void)ipfw_attach_hooks(0 /* detach */);
3384 V_ip_fw_ctl_ptr = NULL;
3386 last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3388 IPFW_UH_WLOCK(chain);
3389 IPFW_UH_WUNLOCK(chain);
3391 ipfw_dyn_uninit(0); /* run the callout_drain */
3393 IPFW_UH_WLOCK(chain);
3397 for (i = 0; i < chain->n_rules; i++)
3398 ipfw_reap_add(chain, &reap, chain->map[i]);
3399 free(chain->map, M_IPFW);
3400 #ifdef LINEAR_SKIPTO
3401 ipfw_destroy_skipto_cache(chain);
3403 IPFW_WUNLOCK(chain);
3404 IPFW_UH_WUNLOCK(chain);
3405 ipfw_destroy_tables(chain, last);
3406 ipfw_eaction_uninit(chain, last);
3408 ipfw_reap_rules(reap);
3409 vnet_ipfw_iface_destroy(chain);
3410 ipfw_destroy_srv(chain);
3411 IPFW_LOCK_DESTROY(chain);
3412 ipfw_dyn_uninit(1); /* free the remaining parts */
3413 ipfw_destroy_counters();
3414 ipfw_bpf_uninit(last);
3419 * Module event handler.
3420 * In general we have the choice of handling most of these events by the
3421 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
3422 * use the SYSINIT handlers as they are more capable of expressing the
3423 * flow of control during module and vnet operations, so this is just
3424 * a skeleton. Note there is no SYSINIT equivalent of the module
3425 * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
3428 ipfw_modevent(module_t mod, int type, void *unused)
3434 /* Called once at module load or
3435 * system boot if compiled in. */
3438 /* Called before unload. May veto unloading. */
3441 /* Called during unload. */
3444 /* Called during system shutdown. */
3453 static moduledata_t ipfwmod = {
3459 /* Define startup order. */
3460 #define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_FIREWALL
3461 #define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */
3462 #define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */
3463 #define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */
3465 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
3466 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
3467 MODULE_VERSION(ipfw, 3);
3468 /* should declare some dependencies here */
3471 * Starting up. Done in order after ipfwmod() has been called.
3472 * VNET_SYSINIT is also called for each existing vnet and each new vnet.
3474 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3476 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3477 vnet_ipfw_init, NULL);
3480 * Closing up shop. These are done in REVERSE ORDER, but still
3481 * after ipfwmod() has been called. Not called on reboot.
3482 * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
3483 * or when the module is unloaded.
3485 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3486 ipfw_destroy, NULL);
3487 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3488 vnet_ipfw_uninit, NULL);