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 static VNET_DEFINE(int, fw_deny_unknown_exthdrs);
114 #define V_fw_deny_unknown_exthdrs VNET(fw_deny_unknown_exthdrs)
116 static VNET_DEFINE(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 TAILQ_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 TAILQ_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 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
830 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
832 tcp = (struct tcphdr *)((char *)ip6 + hlen);
834 if ((tcp->th_flags & TH_RST) == 0) {
836 m0 = ipfw_send_pkt(args->m, &(args->f_id),
837 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
838 tcp->th_flags | TH_RST);
840 ip6_output(m0, NULL, NULL, 0, NULL, NULL,
844 } else if (code == ICMP6_UNREACH_ABORT &&
845 args->f_id.proto == IPPROTO_SCTP) {
847 struct sctphdr *sctp;
851 sctp = (struct sctphdr *)((char *)ip6 + hlen);
853 v_tag = ntohl(sctp->v_tag);
854 /* Investigate the first chunk header if available */
855 if (m->m_len >= hlen + sizeof(struct sctphdr) +
856 sizeof(struct sctp_chunkhdr)) {
857 struct sctp_chunkhdr *chunk;
859 chunk = (struct sctp_chunkhdr *)(sctp + 1);
860 switch (chunk->chunk_type) {
861 case SCTP_INITIATION:
863 * Packets containing an INIT chunk MUST have
870 /* INIT chunk MUST NOT be bundled */
871 if (m->m_pkthdr.len >
872 hlen + sizeof(struct sctphdr) +
873 ntohs(chunk->chunk_length) + 3) {
876 /* Use the initiate tag if available */
877 if ((m->m_len >= hlen + sizeof(struct sctphdr) +
878 sizeof(struct sctp_chunkhdr) +
879 offsetof(struct sctp_init, a_rwnd))) {
880 struct sctp_init *init;
882 init = (struct sctp_init *)(chunk + 1);
883 v_tag = ntohl(init->initiate_tag);
887 case SCTP_ABORT_ASSOCIATION:
889 * If the packet contains an ABORT chunk, don't
891 * XXX: We should search through all chunks,
892 * but don't do to avoid attacks.
901 m0 = ipfw_send_abort(args->m, &(args->f_id), v_tag,
905 ip6_output(m0, NULL, NULL, 0, NULL, NULL, NULL);
907 } else if (code != ICMP6_UNREACH_RST && code != ICMP6_UNREACH_ABORT) {
908 /* Send an ICMPv6 unreach. */
911 * Unlike above, the mbufs need to line up with the ip6 hdr,
912 * as the contents are read. We need to m_adj() the
914 * The mbuf will however be thrown away so we can adjust it.
915 * Remember we did an m_pullup on it already so we
916 * can make some assumptions about contiguousness.
919 m_adj(m, args->L3offset);
921 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
932 * sends a reject message, consuming the mbuf passed as an argument.
935 send_reject(struct ip_fw_args *args, int code, int iplen, struct ip *ip)
939 /* XXX When ip is not guaranteed to be at mtod() we will
940 * need to account for this */
941 * The mbuf will however be thrown away so we can adjust it.
942 * Remember we did an m_pullup on it already so we
943 * can make some assumptions about contiguousness.
946 m_adj(m, args->L3offset);
948 if (code != ICMP_REJECT_RST && code != ICMP_REJECT_ABORT) {
949 /* Send an ICMP unreach */
950 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
951 } else if (code == ICMP_REJECT_RST && args->f_id.proto == IPPROTO_TCP) {
952 struct tcphdr *const tcp =
953 L3HDR(struct tcphdr, mtod(args->m, struct ip *));
954 if ( (tcp->th_flags & TH_RST) == 0) {
956 m = ipfw_send_pkt(args->m, &(args->f_id),
957 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
958 tcp->th_flags | TH_RST);
960 ip_output(m, NULL, NULL, 0, NULL, NULL);
963 } else if (code == ICMP_REJECT_ABORT &&
964 args->f_id.proto == IPPROTO_SCTP) {
966 struct sctphdr *sctp;
967 struct sctp_chunkhdr *chunk;
968 struct sctp_init *init;
972 sctp = L3HDR(struct sctphdr, mtod(args->m, struct ip *));
974 v_tag = ntohl(sctp->v_tag);
975 if (iplen >= (ip->ip_hl << 2) + sizeof(struct sctphdr) +
976 sizeof(struct sctp_chunkhdr)) {
977 /* Look at the first chunk header if available */
978 chunk = (struct sctp_chunkhdr *)(sctp + 1);
979 switch (chunk->chunk_type) {
980 case SCTP_INITIATION:
982 * Packets containing an INIT chunk MUST have
989 /* INIT chunk MUST NOT be bundled */
991 (ip->ip_hl << 2) + sizeof(struct sctphdr) +
992 ntohs(chunk->chunk_length) + 3) {
995 /* Use the initiate tag if available */
996 if ((iplen >= (ip->ip_hl << 2) +
997 sizeof(struct sctphdr) +
998 sizeof(struct sctp_chunkhdr) +
999 offsetof(struct sctp_init, a_rwnd))) {
1000 init = (struct sctp_init *)(chunk + 1);
1001 v_tag = ntohl(init->initiate_tag);
1005 case SCTP_ABORT_ASSOCIATION:
1007 * If the packet contains an ABORT chunk, don't
1009 * XXX: We should search through all chunks,
1010 * but don't do to avoid attacks.
1019 m = ipfw_send_abort(args->m, &(args->f_id), v_tag,
1023 ip_output(m, NULL, NULL, 0, NULL, NULL);
1031 * Support for uid/gid/jail lookup. These tests are expensive
1032 * (because we may need to look into the list of active sockets)
1033 * so we cache the results. ugid_lookupp is 0 if we have not
1034 * yet done a lookup, 1 if we succeeded, and -1 if we tried
1035 * and failed. The function always returns the match value.
1036 * We could actually spare the variable and use *uc, setting
1037 * it to '(void *)check_uidgid if we have no info, NULL if
1038 * we tried and failed, or any other value if successful.
1041 check_uidgid(ipfw_insn_u32 *insn, struct ip_fw_args *args, int *ugid_lookupp,
1044 #if defined(USERSPACE)
1045 return 0; // not supported in userspace
1049 return cred_check(insn, proto, oif,
1050 dst_ip, dst_port, src_ip, src_port,
1051 (struct bsd_ucred *)uc, ugid_lookupp, ((struct mbuf *)inp)->m_skb);
1053 struct in_addr src_ip, dst_ip;
1054 struct inpcbinfo *pi;
1055 struct ipfw_flow_id *id;
1056 struct inpcb *pcb, *inp;
1066 * Check to see if the UDP or TCP stack supplied us with
1067 * the PCB. If so, rather then holding a lock and looking
1068 * up the PCB, we can use the one that was supplied.
1070 if (inp && *ugid_lookupp == 0) {
1071 INP_LOCK_ASSERT(inp);
1072 if (inp->inp_socket != NULL) {
1073 *uc = crhold(inp->inp_cred);
1079 * If we have already been here and the packet has no
1080 * PCB entry associated with it, then we can safely
1081 * assume that this is a no match.
1083 if (*ugid_lookupp == -1)
1085 if (id->proto == IPPROTO_TCP) {
1088 } else if (id->proto == IPPROTO_UDP) {
1089 lookupflags = INPLOOKUP_WILDCARD;
1091 } else if (id->proto == IPPROTO_UDPLITE) {
1092 lookupflags = INPLOOKUP_WILDCARD;
1093 pi = &V_ulitecbinfo;
1096 lookupflags |= INPLOOKUP_RLOCKPCB;
1098 if (*ugid_lookupp == 0) {
1099 if (id->addr_type == 6) {
1102 pcb = in6_pcblookup_mbuf(pi,
1103 &id->src_ip6, htons(id->src_port),
1104 &id->dst_ip6, htons(id->dst_port),
1105 lookupflags, oif, args->m);
1107 pcb = in6_pcblookup_mbuf(pi,
1108 &id->dst_ip6, htons(id->dst_port),
1109 &id->src_ip6, htons(id->src_port),
1110 lookupflags, oif, args->m);
1116 src_ip.s_addr = htonl(id->src_ip);
1117 dst_ip.s_addr = htonl(id->dst_ip);
1119 pcb = in_pcblookup_mbuf(pi,
1120 src_ip, htons(id->src_port),
1121 dst_ip, htons(id->dst_port),
1122 lookupflags, oif, args->m);
1124 pcb = in_pcblookup_mbuf(pi,
1125 dst_ip, htons(id->dst_port),
1126 src_ip, htons(id->src_port),
1127 lookupflags, oif, args->m);
1130 INP_RLOCK_ASSERT(pcb);
1131 *uc = crhold(pcb->inp_cred);
1135 if (*ugid_lookupp == 0) {
1137 * We tried and failed, set the variable to -1
1138 * so we will not try again on this packet.
1144 if (insn->o.opcode == O_UID)
1145 match = ((*uc)->cr_uid == (uid_t)insn->d[0]);
1146 else if (insn->o.opcode == O_GID)
1147 match = groupmember((gid_t)insn->d[0], *uc);
1148 else if (insn->o.opcode == O_JAIL)
1149 match = ((*uc)->cr_prison->pr_id == (int)insn->d[0]);
1151 #endif /* __FreeBSD__ */
1152 #endif /* not supported in userspace */
1156 * Helper function to set args with info on the rule after the matching
1157 * one. slot is precise, whereas we guess rule_id as they are
1158 * assigned sequentially.
1161 set_match(struct ip_fw_args *args, int slot,
1162 struct ip_fw_chain *chain)
1164 args->rule.chain_id = chain->id;
1165 args->rule.slot = slot + 1; /* we use 0 as a marker */
1166 args->rule.rule_id = 1 + chain->map[slot]->id;
1167 args->rule.rulenum = chain->map[slot]->rulenum;
1170 #ifndef LINEAR_SKIPTO
1172 * Helper function to enable cached rule lookups using
1173 * cached_id and cached_pos fields in ipfw rule.
1176 jump_fast(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1177 int tablearg, int jump_backwards)
1181 /* If possible use cached f_pos (in f->cached_pos),
1182 * whose version is written in f->cached_id
1183 * (horrible hacks to avoid changing the ABI).
1185 if (num != IP_FW_TARG && f->cached_id == chain->id)
1186 f_pos = f->cached_pos;
1188 int i = IP_FW_ARG_TABLEARG(chain, num, skipto);
1189 /* make sure we do not jump backward */
1190 if (jump_backwards == 0 && i <= f->rulenum)
1192 if (chain->idxmap != NULL)
1193 f_pos = chain->idxmap[i];
1195 f_pos = ipfw_find_rule(chain, i, 0);
1196 /* update the cache */
1197 if (num != IP_FW_TARG) {
1198 f->cached_id = chain->id;
1199 f->cached_pos = f_pos;
1207 * Helper function to enable real fast rule lookups.
1210 jump_linear(struct ip_fw_chain *chain, struct ip_fw *f, int num,
1211 int tablearg, int jump_backwards)
1215 num = IP_FW_ARG_TABLEARG(chain, num, skipto);
1216 /* make sure we do not jump backward */
1217 if (jump_backwards == 0 && num <= f->rulenum)
1218 num = f->rulenum + 1;
1219 f_pos = chain->idxmap[num];
1225 #define TARG(k, f) IP_FW_ARG_TABLEARG(chain, k, f)
1227 * The main check routine for the firewall.
1229 * All arguments are in args so we can modify them and return them
1230 * back to the caller.
1234 * args->m (in/out) The packet; we set to NULL when/if we nuke it.
1235 * Starts with the IP header.
1236 * args->eh (in) Mac header if present, NULL for layer3 packet.
1237 * args->L3offset Number of bytes bypassed if we came from L2.
1238 * e.g. often sizeof(eh) ** NOTYET **
1239 * args->oif Outgoing interface, NULL if packet is incoming.
1240 * The incoming interface is in the mbuf. (in)
1241 * args->divert_rule (in/out)
1242 * Skip up to the first rule past this rule number;
1243 * upon return, non-zero port number for divert or tee.
1245 * args->rule Pointer to the last matching rule (in/out)
1246 * args->next_hop Socket we are forwarding to (out).
1247 * args->next_hop6 IPv6 next hop we are forwarding to (out).
1248 * args->f_id Addresses grabbed from the packet (out)
1249 * args->rule.info a cookie depending on rule action
1253 * IP_FW_PASS the packet must be accepted
1254 * IP_FW_DENY the packet must be dropped
1255 * IP_FW_DIVERT divert packet, port in m_tag
1256 * IP_FW_TEE tee packet, port in m_tag
1257 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie
1258 * IP_FW_NETGRAPH into netgraph, cookie args->cookie
1259 * args->rule contains the matching rule,
1260 * args->rule.info has additional information.
1264 ipfw_chk(struct ip_fw_args *args)
1268 * Local variables holding state while processing a packet:
1270 * IMPORTANT NOTE: to speed up the processing of rules, there
1271 * are some assumption on the values of the variables, which
1272 * are documented here. Should you change them, please check
1273 * the implementation of the various instructions to make sure
1274 * that they still work.
1276 * args->eh The MAC header. It is non-null for a layer2
1277 * packet, it is NULL for a layer-3 packet.
1279 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
1281 * m | args->m Pointer to the mbuf, as received from the caller.
1282 * It may change if ipfw_chk() does an m_pullup, or if it
1283 * consumes the packet because it calls send_reject().
1284 * XXX This has to change, so that ipfw_chk() never modifies
1285 * or consumes the buffer.
1286 * ip is the beginning of the ip(4 or 6) header.
1287 * Calculated by adding the L3offset to the start of data.
1288 * (Until we start using L3offset, the packet is
1289 * supposed to start with the ip header).
1291 struct mbuf *m = args->m;
1292 struct ip *ip = mtod(m, struct ip *);
1295 * For rules which contain uid/gid or jail constraints, cache
1296 * a copy of the users credentials after the pcb lookup has been
1297 * executed. This will speed up the processing of rules with
1298 * these types of constraints, as well as decrease contention
1299 * on pcb related locks.
1302 struct bsd_ucred ucred_cache;
1304 struct ucred *ucred_cache = NULL;
1306 int ucred_lookup = 0;
1309 * oif | args->oif If NULL, ipfw_chk has been called on the
1310 * inbound path (ether_input, ip_input).
1311 * If non-NULL, ipfw_chk has been called on the outbound path
1312 * (ether_output, ip_output).
1314 struct ifnet *oif = args->oif;
1316 int f_pos = 0; /* index of current rule in the array */
1320 * hlen The length of the IP header.
1322 u_int hlen = 0; /* hlen >0 means we have an IP pkt */
1325 * offset The offset of a fragment. offset != 0 means that
1326 * we have a fragment at this offset of an IPv4 packet.
1327 * offset == 0 means that (if this is an IPv4 packet)
1328 * this is the first or only fragment.
1329 * For IPv6 offset|ip6f_mf == 0 means there is no Fragment Header
1330 * or there is a single packet fragment (fragment header added
1331 * without needed). We will treat a single packet fragment as if
1332 * there was no fragment header (or log/block depending on the
1333 * V_fw_permit_single_frag6 sysctl setting).
1336 u_short ip6f_mf = 0;
1339 * Local copies of addresses. They are only valid if we have
1342 * proto The protocol. Set to 0 for non-ip packets,
1343 * or to the protocol read from the packet otherwise.
1344 * proto != 0 means that we have an IPv4 packet.
1346 * src_port, dst_port port numbers, in HOST format. Only
1347 * valid for TCP and UDP packets.
1349 * src_ip, dst_ip ip addresses, in NETWORK format.
1350 * Only valid for IPv4 packets.
1353 uint16_t src_port = 0, dst_port = 0; /* NOTE: host format */
1354 struct in_addr src_ip, dst_ip; /* NOTE: network format */
1357 uint16_t etype = 0; /* Host order stored ether type */
1360 * dyn_dir = MATCH_UNKNOWN when rules unchecked,
1361 * MATCH_NONE when checked and not matched (q = NULL),
1362 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
1364 int dyn_dir = MATCH_UNKNOWN;
1365 uint16_t dyn_name = 0;
1366 struct ip_fw *q = NULL;
1367 struct ip_fw_chain *chain = &V_layer3_chain;
1370 * We store in ulp a pointer to the upper layer protocol header.
1371 * In the ipv4 case this is easy to determine from the header,
1372 * but for ipv6 we might have some additional headers in the middle.
1373 * ulp is NULL if not found.
1375 void *ulp = NULL; /* upper layer protocol pointer. */
1377 /* XXX ipv6 variables */
1379 uint8_t icmp6_type = 0;
1380 uint16_t ext_hd = 0; /* bits vector for extension header filtering */
1381 /* end of ipv6 variables */
1385 int done = 0; /* flag to exit the outer loop */
1387 if (m->m_flags & M_SKIP_FIREWALL || (! V_ipfw_vnet_ready))
1388 return (IP_FW_PASS); /* accept */
1390 dst_ip.s_addr = 0; /* make sure it is initialized */
1391 src_ip.s_addr = 0; /* make sure it is initialized */
1392 pktlen = m->m_pkthdr.len;
1393 args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
1394 proto = args->f_id.proto = 0; /* mark f_id invalid */
1395 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
1398 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
1399 * then it sets p to point at the offset "len" in the mbuf. WARNING: the
1400 * pointer might become stale after other pullups (but we never use it
1403 #define PULLUP_TO(_len, p, T) PULLUP_LEN(_len, p, sizeof(T))
1404 #define PULLUP_LEN(_len, p, T) \
1406 int x = (_len) + T; \
1407 if ((m)->m_len < x) { \
1408 args->m = m = m_pullup(m, x); \
1410 goto pullup_failed; \
1412 p = (mtod(m, char *) + (_len)); \
1416 * if we have an ether header,
1419 etype = ntohs(args->eh->ether_type);
1421 /* Identify IP packets and fill up variables. */
1422 if (pktlen >= sizeof(struct ip6_hdr) &&
1423 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
1424 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
1426 args->f_id.addr_type = 6;
1427 hlen = sizeof(struct ip6_hdr);
1428 proto = ip6->ip6_nxt;
1430 /* Search extension headers to find upper layer protocols */
1431 while (ulp == NULL && offset == 0) {
1433 case IPPROTO_ICMPV6:
1434 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
1435 icmp6_type = ICMP6(ulp)->icmp6_type;
1439 PULLUP_TO(hlen, ulp, struct tcphdr);
1440 dst_port = TCP(ulp)->th_dport;
1441 src_port = TCP(ulp)->th_sport;
1442 /* save flags for dynamic rules */
1443 args->f_id._flags = TCP(ulp)->th_flags;
1447 if (pktlen >= hlen + sizeof(struct sctphdr) +
1448 sizeof(struct sctp_chunkhdr) +
1449 offsetof(struct sctp_init, a_rwnd))
1450 PULLUP_LEN(hlen, ulp,
1451 sizeof(struct sctphdr) +
1452 sizeof(struct sctp_chunkhdr) +
1453 offsetof(struct sctp_init, a_rwnd));
1454 else if (pktlen >= hlen + sizeof(struct sctphdr))
1455 PULLUP_LEN(hlen, ulp, pktlen - hlen);
1457 PULLUP_LEN(hlen, ulp,
1458 sizeof(struct sctphdr));
1459 src_port = SCTP(ulp)->src_port;
1460 dst_port = SCTP(ulp)->dest_port;
1464 case IPPROTO_UDPLITE:
1465 PULLUP_TO(hlen, ulp, struct udphdr);
1466 dst_port = UDP(ulp)->uh_dport;
1467 src_port = UDP(ulp)->uh_sport;
1470 case IPPROTO_HOPOPTS: /* RFC 2460 */
1471 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1472 ext_hd |= EXT_HOPOPTS;
1473 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1474 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1478 case IPPROTO_ROUTING: /* RFC 2460 */
1479 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
1480 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
1482 ext_hd |= EXT_RTHDR0;
1485 ext_hd |= EXT_RTHDR2;
1489 printf("IPFW2: IPV6 - Unknown "
1490 "Routing Header type(%d)\n",
1491 ((struct ip6_rthdr *)
1493 if (V_fw_deny_unknown_exthdrs)
1494 return (IP_FW_DENY);
1497 ext_hd |= EXT_ROUTING;
1498 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
1499 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
1503 case IPPROTO_FRAGMENT: /* RFC 2460 */
1504 PULLUP_TO(hlen, ulp, struct ip6_frag);
1505 ext_hd |= EXT_FRAGMENT;
1506 hlen += sizeof (struct ip6_frag);
1507 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
1508 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
1510 ip6f_mf = ((struct ip6_frag *)ulp)->ip6f_offlg &
1512 if (V_fw_permit_single_frag6 == 0 &&
1513 offset == 0 && ip6f_mf == 0) {
1515 printf("IPFW2: IPV6 - Invalid "
1516 "Fragment Header\n");
1517 if (V_fw_deny_unknown_exthdrs)
1518 return (IP_FW_DENY);
1522 ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
1526 case IPPROTO_DSTOPTS: /* RFC 2460 */
1527 PULLUP_TO(hlen, ulp, struct ip6_hbh);
1528 ext_hd |= EXT_DSTOPTS;
1529 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
1530 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
1534 case IPPROTO_AH: /* RFC 2402 */
1535 PULLUP_TO(hlen, ulp, struct ip6_ext);
1537 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
1538 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
1542 case IPPROTO_ESP: /* RFC 2406 */
1543 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
1544 /* Anything past Seq# is variable length and
1545 * data past this ext. header is encrypted. */
1549 case IPPROTO_NONE: /* RFC 2460 */
1551 * Packet ends here, and IPv6 header has
1552 * already been pulled up. If ip6e_len!=0
1553 * then octets must be ignored.
1555 ulp = ip; /* non-NULL to get out of loop. */
1558 case IPPROTO_OSPFIGP:
1559 /* XXX OSPF header check? */
1560 PULLUP_TO(hlen, ulp, struct ip6_ext);
1564 /* XXX PIM header check? */
1565 PULLUP_TO(hlen, ulp, struct pim);
1568 case IPPROTO_GRE: /* RFC 1701 */
1569 /* XXX GRE header check? */
1570 PULLUP_TO(hlen, ulp, struct grehdr);
1574 PULLUP_TO(hlen, ulp, struct carp_header);
1575 if (((struct carp_header *)ulp)->carp_version !=
1577 return (IP_FW_DENY);
1578 if (((struct carp_header *)ulp)->carp_type !=
1580 return (IP_FW_DENY);
1583 case IPPROTO_IPV6: /* RFC 2893 */
1584 PULLUP_TO(hlen, ulp, struct ip6_hdr);
1587 case IPPROTO_IPV4: /* RFC 2893 */
1588 PULLUP_TO(hlen, ulp, struct ip);
1593 printf("IPFW2: IPV6 - Unknown "
1594 "Extension Header(%d), ext_hd=%x\n",
1596 if (V_fw_deny_unknown_exthdrs)
1597 return (IP_FW_DENY);
1598 PULLUP_TO(hlen, ulp, struct ip6_ext);
1602 ip = mtod(m, struct ip *);
1603 ip6 = (struct ip6_hdr *)ip;
1604 args->f_id.src_ip6 = ip6->ip6_src;
1605 args->f_id.dst_ip6 = ip6->ip6_dst;
1606 args->f_id.src_ip = 0;
1607 args->f_id.dst_ip = 0;
1608 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
1609 iplen = ntohs(ip6->ip6_plen) + sizeof(*ip6);
1610 } else if (pktlen >= sizeof(struct ip) &&
1611 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
1613 hlen = ip->ip_hl << 2;
1614 args->f_id.addr_type = 4;
1617 * Collect parameters into local variables for faster matching.
1620 src_ip = ip->ip_src;
1621 dst_ip = ip->ip_dst;
1622 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1623 iplen = ntohs(ip->ip_len);
1628 PULLUP_TO(hlen, ulp, struct tcphdr);
1629 dst_port = TCP(ulp)->th_dport;
1630 src_port = TCP(ulp)->th_sport;
1631 /* save flags for dynamic rules */
1632 args->f_id._flags = TCP(ulp)->th_flags;
1636 if (pktlen >= hlen + sizeof(struct sctphdr) +
1637 sizeof(struct sctp_chunkhdr) +
1638 offsetof(struct sctp_init, a_rwnd))
1639 PULLUP_LEN(hlen, ulp,
1640 sizeof(struct sctphdr) +
1641 sizeof(struct sctp_chunkhdr) +
1642 offsetof(struct sctp_init, a_rwnd));
1643 else if (pktlen >= hlen + sizeof(struct sctphdr))
1644 PULLUP_LEN(hlen, ulp, pktlen - hlen);
1646 PULLUP_LEN(hlen, ulp,
1647 sizeof(struct sctphdr));
1648 src_port = SCTP(ulp)->src_port;
1649 dst_port = SCTP(ulp)->dest_port;
1653 case IPPROTO_UDPLITE:
1654 PULLUP_TO(hlen, ulp, struct udphdr);
1655 dst_port = UDP(ulp)->uh_dport;
1656 src_port = UDP(ulp)->uh_sport;
1660 PULLUP_TO(hlen, ulp, struct icmphdr);
1661 //args->f_id.flags = ICMP(ulp)->icmp_type;
1669 ip = mtod(m, struct ip *);
1670 args->f_id.src_ip = ntohl(src_ip.s_addr);
1671 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1674 pktlen = iplen < pktlen ? iplen: pktlen;
1675 if (proto) { /* we may have port numbers, store them */
1676 args->f_id.proto = proto;
1677 args->f_id.src_port = src_port = ntohs(src_port);
1678 args->f_id.dst_port = dst_port = ntohs(dst_port);
1681 IPFW_PF_RLOCK(chain);
1682 if (! V_ipfw_vnet_ready) { /* shutting down, leave NOW. */
1683 IPFW_PF_RUNLOCK(chain);
1684 return (IP_FW_PASS); /* accept */
1686 if (args->rule.slot) {
1688 * Packet has already been tagged as a result of a previous
1689 * match on rule args->rule aka args->rule_id (PIPE, QUEUE,
1690 * REASS, NETGRAPH, DIVERT/TEE...)
1691 * Validate the slot and continue from the next one
1692 * if still present, otherwise do a lookup.
1694 f_pos = (args->rule.chain_id == chain->id) ?
1696 ipfw_find_rule(chain, args->rule.rulenum,
1697 args->rule.rule_id);
1703 * Now scan the rules, and parse microinstructions for each rule.
1704 * We have two nested loops and an inner switch. Sometimes we
1705 * need to break out of one or both loops, or re-enter one of
1706 * the loops with updated variables. Loop variables are:
1708 * f_pos (outer loop) points to the current rule.
1709 * On output it points to the matching rule.
1710 * done (outer loop) is used as a flag to break the loop.
1711 * l (inner loop) residual length of current rule.
1712 * cmd points to the current microinstruction.
1714 * We break the inner loop by setting l=0 and possibly
1715 * cmdlen=0 if we don't want to advance cmd.
1716 * We break the outer loop by setting done=1
1717 * We can restart the inner loop by setting l>0 and f_pos, f, cmd
1720 for (; f_pos < chain->n_rules; f_pos++) {
1722 uint32_t tablearg = 0;
1723 int l, cmdlen, skip_or; /* skip rest of OR block */
1726 f = chain->map[f_pos];
1727 if (V_set_disable & (1 << f->set) )
1731 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1732 l -= cmdlen, cmd += cmdlen) {
1736 * check_body is a jump target used when we find a
1737 * CHECK_STATE, and need to jump to the body of
1742 cmdlen = F_LEN(cmd);
1744 * An OR block (insn_1 || .. || insn_n) has the
1745 * F_OR bit set in all but the last instruction.
1746 * The first match will set "skip_or", and cause
1747 * the following instructions to be skipped until
1748 * past the one with the F_OR bit clear.
1750 if (skip_or) { /* skip this instruction */
1751 if ((cmd->len & F_OR) == 0)
1752 skip_or = 0; /* next one is good */
1755 match = 0; /* set to 1 if we succeed */
1757 switch (cmd->opcode) {
1759 * The first set of opcodes compares the packet's
1760 * fields with some pattern, setting 'match' if a
1761 * match is found. At the end of the loop there is
1762 * logic to deal with F_NOT and F_OR flags associated
1770 printf("ipfw: opcode %d unimplemented\n",
1778 * We only check offset == 0 && proto != 0,
1779 * as this ensures that we have a
1780 * packet with the ports info.
1784 if (proto == IPPROTO_TCP ||
1785 proto == IPPROTO_UDP ||
1786 proto == IPPROTO_UDPLITE)
1787 match = check_uidgid(
1788 (ipfw_insn_u32 *)cmd,
1789 args, &ucred_lookup,
1793 (void *)&ucred_cache);
1798 match = iface_match(m->m_pkthdr.rcvif,
1799 (ipfw_insn_if *)cmd, chain, &tablearg);
1803 match = iface_match(oif, (ipfw_insn_if *)cmd,
1808 match = iface_match(oif ? oif :
1809 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd,
1814 if (args->eh != NULL) { /* have MAC header */
1815 u_int32_t *want = (u_int32_t *)
1816 ((ipfw_insn_mac *)cmd)->addr;
1817 u_int32_t *mask = (u_int32_t *)
1818 ((ipfw_insn_mac *)cmd)->mask;
1819 u_int32_t *hdr = (u_int32_t *)args->eh;
1822 ( want[0] == (hdr[0] & mask[0]) &&
1823 want[1] == (hdr[1] & mask[1]) &&
1824 want[2] == (hdr[2] & mask[2]) );
1829 if (args->eh != NULL) {
1831 ((ipfw_insn_u16 *)cmd)->ports;
1834 for (i = cmdlen - 1; !match && i>0;
1836 match = (etype >= p[0] &&
1842 match = (offset != 0);
1845 case O_IN: /* "out" is "not in" */
1846 match = (oif == NULL);
1850 match = (args->eh != NULL);
1855 /* For diverted packets, args->rule.info
1856 * contains the divert port (in host format)
1857 * reason and direction.
1859 uint32_t i = args->rule.info;
1860 match = (i&IPFW_IS_MASK) == IPFW_IS_DIVERT &&
1861 cmd->arg1 & ((i & IPFW_INFO_IN) ? 1 : 2);
1867 * We do not allow an arg of 0 so the
1868 * check of "proto" only suffices.
1870 match = (proto == cmd->arg1);
1875 (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1879 case O_IP_DST_LOOKUP:
1885 if (cmdlen > F_INSN_SIZE(ipfw_insn_u32)) {
1886 /* Determine lookup key type */
1887 vidx = ((ipfw_insn_u32 *)cmd)->d[1];
1888 if (vidx != 4 /* uid */ &&
1889 vidx != 5 /* jail */ &&
1890 is_ipv6 == 0 && is_ipv4 == 0)
1892 /* Determine key length */
1893 if (vidx == 0 /* dst-ip */ ||
1894 vidx == 1 /* src-ip */)
1896 sizeof(struct in6_addr):
1899 keylen = sizeof(key);
1902 if (vidx == 0 /* dst-ip */)
1903 pkey = is_ipv4 ? (void *)&dst_ip:
1904 (void *)&args->f_id.dst_ip6;
1905 else if (vidx == 1 /* src-ip */)
1906 pkey = is_ipv4 ? (void *)&src_ip:
1907 (void *)&args->f_id.src_ip6;
1908 else if (vidx == 6 /* dscp */) {
1910 key = ip->ip_tos >> 2;
1912 key = args->f_id.flow_id6;
1913 key = (key & 0x0f) << 2 |
1914 (key & 0xf000) >> 14;
1917 } else if (vidx == 2 /* dst-port */ ||
1918 vidx == 3 /* src-port */) {
1919 /* Skip fragments */
1922 /* Skip proto without ports */
1923 if (proto != IPPROTO_TCP &&
1924 proto != IPPROTO_UDP &&
1925 proto != IPPROTO_UDPLITE &&
1926 proto != IPPROTO_SCTP)
1928 if (vidx == 2 /* dst-port */)
1934 else if (vidx == 4 /* uid */ ||
1935 vidx == 5 /* jail */) {
1937 (ipfw_insn_u32 *)cmd,
1938 args, &ucred_lookup,
1941 if (vidx == 4 /* uid */)
1942 key = ucred_cache->cr_uid;
1943 else if (vidx == 5 /* jail */)
1944 key = ucred_cache->cr_prison->pr_id;
1945 #else /* !__FreeBSD__ */
1946 (void *)&ucred_cache);
1947 if (vidx == 4 /* uid */)
1948 key = ucred_cache.uid;
1949 else if (vidx == 5 /* jail */)
1950 key = ucred_cache.xid;
1951 #endif /* !__FreeBSD__ */
1953 #endif /* !USERSPACE */
1956 match = ipfw_lookup_table(chain,
1957 cmd->arg1, keylen, pkey, &vidx);
1963 /* cmdlen =< F_INSN_SIZE(ipfw_insn_u32) */
1966 case O_IP_SRC_LOOKUP:
1973 keylen = sizeof(in_addr_t);
1974 if (cmd->opcode == O_IP_DST_LOOKUP)
1978 } else if (is_ipv6) {
1979 keylen = sizeof(struct in6_addr);
1980 if (cmd->opcode == O_IP_DST_LOOKUP)
1981 pkey = &args->f_id.dst_ip6;
1983 pkey = &args->f_id.src_ip6;
1986 match = ipfw_lookup_table(chain, cmd->arg1,
1987 keylen, pkey, &vidx);
1990 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32)) {
1991 match = ((ipfw_insn_u32 *)cmd)->d[0] ==
1992 TARG_VAL(chain, vidx, tag);
2000 case O_IP_FLOW_LOOKUP:
2003 match = ipfw_lookup_table(chain,
2004 cmd->arg1, 0, &args->f_id, &v);
2005 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
2006 match = ((ipfw_insn_u32 *)cmd)->d[0] ==
2007 TARG_VAL(chain, v, tag);
2016 (cmd->opcode == O_IP_DST_MASK) ?
2017 dst_ip.s_addr : src_ip.s_addr;
2018 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2021 for (; !match && i>0; i-= 2, p+= 2)
2022 match = (p[0] == (a & p[1]));
2028 match = in_localip(src_ip);
2034 match= is_ipv6 && ipfw_localip6(&args->f_id.src_ip6);
2041 u_int32_t *d = (u_int32_t *)(cmd+1);
2043 cmd->opcode == O_IP_DST_SET ?
2049 addr -= d[0]; /* subtract base */
2050 match = (addr < cmd->arg1) &&
2051 ( d[ 1 + (addr>>5)] &
2052 (1<<(addr & 0x1f)) );
2058 (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2064 match = in_localip(dst_ip);
2070 match= is_ipv6 && ipfw_localip6(&args->f_id.dst_ip6);
2078 * offset == 0 && proto != 0 is enough
2079 * to guarantee that we have a
2080 * packet with port info.
2082 if ((proto == IPPROTO_UDP ||
2083 proto == IPPROTO_UDPLITE ||
2084 proto == IPPROTO_TCP ||
2085 proto == IPPROTO_SCTP) && offset == 0) {
2087 (cmd->opcode == O_IP_SRCPORT) ?
2088 src_port : dst_port ;
2090 ((ipfw_insn_u16 *)cmd)->ports;
2093 for (i = cmdlen - 1; !match && i>0;
2095 match = (x>=p[0] && x<=p[1]);
2100 match = (offset == 0 && proto==IPPROTO_ICMP &&
2101 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
2106 match = is_ipv6 && offset == 0 &&
2107 proto==IPPROTO_ICMPV6 &&
2109 ICMP6(ulp)->icmp6_type,
2110 (ipfw_insn_u32 *)cmd);
2116 ipopts_match(ip, cmd) );
2121 cmd->arg1 == ip->ip_v);
2127 if (is_ipv4) { /* only for IP packets */
2132 if (cmd->opcode == O_IPLEN)
2134 else if (cmd->opcode == O_IPTTL)
2136 else /* must be IPID */
2137 x = ntohs(ip->ip_id);
2139 match = (cmd->arg1 == x);
2142 /* otherwise we have ranges */
2143 p = ((ipfw_insn_u16 *)cmd)->ports;
2145 for (; !match && i>0; i--, p += 2)
2146 match = (x >= p[0] && x <= p[1]);
2150 case O_IPPRECEDENCE:
2152 (cmd->arg1 == (ip->ip_tos & 0xe0)) );
2157 flags_match(cmd, ip->ip_tos));
2165 p = ((ipfw_insn_u32 *)cmd)->d;
2168 x = ip->ip_tos >> 2;
2171 v = &((struct ip6_hdr *)ip)->ip6_vfc;
2172 x = (*v & 0x0F) << 2;
2178 /* DSCP bitmask is stored as low_u32 high_u32 */
2180 match = *(p + 1) & (1 << (x - 32));
2182 match = *p & (1 << x);
2187 if (proto == IPPROTO_TCP && offset == 0) {
2194 struct ip6_hdr *ip6;
2196 ip6 = (struct ip6_hdr *)ip;
2197 if (ip6->ip6_plen == 0) {
2199 * Jumbo payload is not
2208 x = iplen - (ip->ip_hl << 2);
2210 x -= tcp->th_off << 2;
2212 match = (cmd->arg1 == x);
2215 /* otherwise we have ranges */
2216 p = ((ipfw_insn_u16 *)cmd)->ports;
2218 for (; !match && i>0; i--, p += 2)
2219 match = (x >= p[0] && x <= p[1]);
2224 match = (proto == IPPROTO_TCP && offset == 0 &&
2225 flags_match(cmd, TCP(ulp)->th_flags));
2229 if (proto == IPPROTO_TCP && offset == 0 && ulp){
2230 PULLUP_LEN(hlen, ulp,
2231 (TCP(ulp)->th_off << 2));
2232 match = tcpopts_match(TCP(ulp), cmd);
2237 match = (proto == IPPROTO_TCP && offset == 0 &&
2238 ((ipfw_insn_u32 *)cmd)->d[0] ==
2243 match = (proto == IPPROTO_TCP && offset == 0 &&
2244 ((ipfw_insn_u32 *)cmd)->d[0] ==
2249 if (proto == IPPROTO_TCP && offset == 0) {
2254 x = ntohs(TCP(ulp)->th_win);
2256 match = (cmd->arg1 == x);
2259 /* Otherwise we have ranges. */
2260 p = ((ipfw_insn_u16 *)cmd)->ports;
2262 for (; !match && i > 0; i--, p += 2)
2263 match = (x >= p[0] && x <= p[1]);
2268 /* reject packets which have SYN only */
2269 /* XXX should i also check for TH_ACK ? */
2270 match = (proto == IPPROTO_TCP && offset == 0 &&
2271 (TCP(ulp)->th_flags &
2272 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
2278 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
2281 * ALTQ uses mbuf tags from another
2282 * packet filtering system - pf(4).
2283 * We allocate a tag in its format
2284 * and fill it in, pretending to be pf(4).
2287 at = pf_find_mtag(m);
2288 if (at != NULL && at->qid != 0)
2290 mtag = m_tag_get(PACKET_TAG_PF,
2291 sizeof(struct pf_mtag), M_NOWAIT | M_ZERO);
2294 * Let the packet fall back to the
2299 m_tag_prepend(m, mtag);
2300 at = (struct pf_mtag *)(mtag + 1);
2301 at->qid = altq->qid;
2307 ipfw_log(chain, f, hlen, args, m,
2308 oif, offset | ip6f_mf, tablearg, ip);
2313 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
2317 /* Outgoing packets automatically pass/match */
2318 match = ((oif != NULL) ||
2319 (m->m_pkthdr.rcvif == NULL) ||
2323 verify_path6(&(args->f_id.src_ip6),
2324 m->m_pkthdr.rcvif, args->f_id.fib) :
2326 verify_path(src_ip, m->m_pkthdr.rcvif,
2331 /* Outgoing packets automatically pass/match */
2332 match = (hlen > 0 && ((oif != NULL) ||
2335 verify_path6(&(args->f_id.src_ip6),
2336 NULL, args->f_id.fib) :
2338 verify_path(src_ip, NULL, args->f_id.fib)));
2342 /* Outgoing packets automatically pass/match */
2343 if (oif == NULL && hlen > 0 &&
2344 ( (is_ipv4 && in_localaddr(src_ip))
2347 in6_localaddr(&(args->f_id.src_ip6)))
2352 is_ipv6 ? verify_path6(
2353 &(args->f_id.src_ip6),
2365 match = (m_tag_find(m,
2366 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
2367 /* otherwise no match */
2373 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
2374 &((ipfw_insn_ip6 *)cmd)->addr6);
2379 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
2380 &((ipfw_insn_ip6 *)cmd)->addr6);
2382 case O_IP6_SRC_MASK:
2383 case O_IP6_DST_MASK:
2387 struct in6_addr *d =
2388 &((ipfw_insn_ip6 *)cmd)->addr6;
2390 for (; !match && i > 0; d += 2,
2391 i -= F_INSN_SIZE(struct in6_addr)
2397 APPLY_MASK(&p, &d[1]);
2399 IN6_ARE_ADDR_EQUAL(&d[0],
2407 flow6id_match(args->f_id.flow_id6,
2408 (ipfw_insn_u32 *) cmd);
2413 (ext_hd & ((ipfw_insn *) cmd)->arg1);
2427 uint32_t tag = TARG(cmd->arg1, tag);
2429 /* Packet is already tagged with this tag? */
2430 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
2432 /* We have `untag' action when F_NOT flag is
2433 * present. And we must remove this mtag from
2434 * mbuf and reset `match' to zero (`match' will
2435 * be inversed later).
2436 * Otherwise we should allocate new mtag and
2437 * push it into mbuf.
2439 if (cmd->len & F_NOT) { /* `untag' action */
2441 m_tag_delete(m, mtag);
2445 mtag = m_tag_alloc( MTAG_IPFW,
2448 m_tag_prepend(m, mtag);
2455 case O_FIB: /* try match the specified fib */
2456 if (args->f_id.fib == cmd->arg1)
2461 #ifndef USERSPACE /* not supported in userspace */
2462 struct inpcb *inp = args->inp;
2463 struct inpcbinfo *pi;
2465 if (is_ipv6) /* XXX can we remove this ? */
2468 if (proto == IPPROTO_TCP)
2470 else if (proto == IPPROTO_UDP)
2472 else if (proto == IPPROTO_UDPLITE)
2473 pi = &V_ulitecbinfo;
2478 * XXXRW: so_user_cookie should almost
2479 * certainly be inp_user_cookie?
2482 /* For incoming packet, lookup up the
2483 inpcb using the src/dest ip/port tuple */
2485 inp = in_pcblookup(pi,
2486 src_ip, htons(src_port),
2487 dst_ip, htons(dst_port),
2488 INPLOOKUP_RLOCKPCB, NULL);
2491 inp->inp_socket->so_user_cookie;
2497 if (inp->inp_socket) {
2499 inp->inp_socket->so_user_cookie;
2504 #endif /* !USERSPACE */
2510 uint32_t tag = TARG(cmd->arg1, tag);
2513 match = m_tag_locate(m, MTAG_IPFW,
2518 /* we have ranges */
2519 for (mtag = m_tag_first(m);
2520 mtag != NULL && !match;
2521 mtag = m_tag_next(m, mtag)) {
2525 if (mtag->m_tag_cookie != MTAG_IPFW)
2528 p = ((ipfw_insn_u16 *)cmd)->ports;
2530 for(; !match && i > 0; i--, p += 2)
2532 mtag->m_tag_id >= p[0] &&
2533 mtag->m_tag_id <= p[1];
2539 * The second set of opcodes represents 'actions',
2540 * i.e. the terminal part of a rule once the packet
2541 * matches all previous patterns.
2542 * Typically there is only one action for each rule,
2543 * and the opcode is stored at the end of the rule
2544 * (but there are exceptions -- see below).
2546 * In general, here we set retval and terminate the
2547 * outer loop (would be a 'break 3' in some language,
2548 * but we need to set l=0, done=1)
2551 * O_COUNT and O_SKIPTO actions:
2552 * instead of terminating, we jump to the next rule
2553 * (setting l=0), or to the SKIPTO target (setting
2554 * f/f_len, cmd and l as needed), respectively.
2556 * O_TAG, O_LOG and O_ALTQ action parameters:
2557 * perform some action and set match = 1;
2559 * O_LIMIT and O_KEEP_STATE: these opcodes are
2560 * not real 'actions', and are stored right
2561 * before the 'action' part of the rule.
2562 * These opcodes try to install an entry in the
2563 * state tables; if successful, we continue with
2564 * the next opcode (match=1; break;), otherwise
2565 * the packet must be dropped (set retval,
2566 * break loops with l=0, done=1)
2568 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2569 * cause a lookup of the state table, and a jump
2570 * to the 'action' part of the parent rule
2571 * if an entry is found, or
2572 * (CHECK_STATE only) a jump to the next rule if
2573 * the entry is not found.
2574 * The result of the lookup is cached so that
2575 * further instances of these opcodes become NOPs.
2576 * The jump to the next rule is done by setting
2581 if (ipfw_dyn_install_state(chain, f,
2582 (ipfw_insn_limit *)cmd, args, tablearg)) {
2583 /* error or limit violation */
2584 retval = IP_FW_DENY;
2585 l = 0; /* exit inner loop */
2586 done = 1; /* exit outer loop */
2594 * dynamic rules are checked at the first
2595 * keep-state or check-state occurrence,
2596 * with the result being stored in dyn_dir
2598 * The compiler introduces a PROBE_STATE
2599 * instruction for us when we have a
2600 * KEEP_STATE (because PROBE_STATE needs
2603 * (dyn_dir == MATCH_UNKNOWN) means this is
2604 * first lookup for such f_id. Do lookup.
2606 * (dyn_dir != MATCH_UNKNOWN &&
2607 * dyn_name != 0 && dyn_name != cmd->arg1)
2608 * means previous lookup didn't find dynamic
2609 * rule for specific state name and current
2610 * lookup will search rule with another state
2611 * name. Redo lookup.
2613 * (dyn_dir != MATCH_UNKNOWN && dyn_name == 0)
2614 * means previous lookup was for `any' name
2615 * and it didn't find rule. No need to do
2618 if ((dyn_dir == MATCH_UNKNOWN ||
2620 dyn_name != cmd->arg1)) &&
2621 (q = ipfw_dyn_lookup_state(&args->f_id,
2622 ulp, pktlen, &dyn_dir,
2623 (dyn_name = cmd->arg1))) != NULL) {
2625 * Found dynamic entry, jump to the
2626 * 'action' part of the parent rule
2627 * by setting f, cmd, l and clearing
2631 /* XXX we would like to have f_pos
2632 * readily accessible in the dynamic
2633 * rule, instead of having to
2636 f_pos = ipfw_find_rule(chain,
2638 cmd = ACTION_PTR(f);
2639 l = f->cmd_len - f->act_ofs;
2645 * Dynamic entry not found. If CHECK_STATE,
2646 * skip to next rule, if PROBE_STATE just
2647 * ignore and continue with next opcode.
2649 if (cmd->opcode == O_CHECK_STATE)
2650 l = 0; /* exit inner loop */
2655 retval = 0; /* accept */
2656 l = 0; /* exit inner loop */
2657 done = 1; /* exit outer loop */
2662 set_match(args, f_pos, chain);
2663 args->rule.info = TARG(cmd->arg1, pipe);
2664 if (cmd->opcode == O_PIPE)
2665 args->rule.info |= IPFW_IS_PIPE;
2667 args->rule.info |= IPFW_ONEPASS;
2668 retval = IP_FW_DUMMYNET;
2669 l = 0; /* exit inner loop */
2670 done = 1; /* exit outer loop */
2675 if (args->eh) /* not on layer 2 */
2677 /* otherwise this is terminal */
2678 l = 0; /* exit inner loop */
2679 done = 1; /* exit outer loop */
2680 retval = (cmd->opcode == O_DIVERT) ?
2681 IP_FW_DIVERT : IP_FW_TEE;
2682 set_match(args, f_pos, chain);
2683 args->rule.info = TARG(cmd->arg1, divert);
2687 IPFW_INC_RULE_COUNTER(f, pktlen);
2688 l = 0; /* exit inner loop */
2692 IPFW_INC_RULE_COUNTER(f, pktlen);
2693 f_pos = JUMP(chain, f, cmd->arg1, tablearg, 0);
2695 * Skip disabled rules, and re-enter
2696 * the inner loop with the correct
2697 * f_pos, f, l and cmd.
2698 * Also clear cmdlen and skip_or
2700 for (; f_pos < chain->n_rules - 1 &&
2702 (1 << chain->map[f_pos]->set));
2705 /* Re-enter the inner loop at the skipto rule. */
2706 f = chain->map[f_pos];
2713 break; /* not reached */
2715 case O_CALLRETURN: {
2717 * Implementation of `subroutine' call/return,
2718 * in the stack carried in an mbuf tag. This
2719 * is different from `skipto' in that any call
2720 * address is possible (`skipto' must prevent
2721 * backward jumps to avoid endless loops).
2722 * We have `return' action when F_NOT flag is
2723 * present. The `m_tag_id' field is used as
2727 uint16_t jmpto, *stack;
2729 #define IS_CALL ((cmd->len & F_NOT) == 0)
2730 #define IS_RETURN ((cmd->len & F_NOT) != 0)
2732 * Hand-rolled version of m_tag_locate() with
2734 * If not already tagged, allocate new tag.
2736 mtag = m_tag_first(m);
2737 while (mtag != NULL) {
2738 if (mtag->m_tag_cookie ==
2741 mtag = m_tag_next(m, mtag);
2743 if (mtag == NULL && IS_CALL) {
2744 mtag = m_tag_alloc(MTAG_IPFW_CALL, 0,
2745 IPFW_CALLSTACK_SIZE *
2746 sizeof(uint16_t), M_NOWAIT);
2748 m_tag_prepend(m, mtag);
2752 * On error both `call' and `return' just
2753 * continue with next rule.
2755 if (IS_RETURN && (mtag == NULL ||
2756 mtag->m_tag_id == 0)) {
2757 l = 0; /* exit inner loop */
2760 if (IS_CALL && (mtag == NULL ||
2761 mtag->m_tag_id >= IPFW_CALLSTACK_SIZE)) {
2762 printf("ipfw: call stack error, "
2763 "go to next rule\n");
2764 l = 0; /* exit inner loop */
2768 IPFW_INC_RULE_COUNTER(f, pktlen);
2769 stack = (uint16_t *)(mtag + 1);
2772 * The `call' action may use cached f_pos
2773 * (in f->next_rule), whose version is written
2775 * The `return' action, however, doesn't have
2776 * fixed jump address in cmd->arg1 and can't use
2780 stack[mtag->m_tag_id] = f->rulenum;
2782 f_pos = JUMP(chain, f, cmd->arg1,
2784 } else { /* `return' action */
2786 jmpto = stack[mtag->m_tag_id] + 1;
2787 f_pos = ipfw_find_rule(chain, jmpto, 0);
2791 * Skip disabled rules, and re-enter
2792 * the inner loop with the correct
2793 * f_pos, f, l and cmd.
2794 * Also clear cmdlen and skip_or
2796 for (; f_pos < chain->n_rules - 1 &&
2798 (1 << chain->map[f_pos]->set)); f_pos++)
2800 /* Re-enter the inner loop at the dest rule. */
2801 f = chain->map[f_pos];
2807 break; /* NOTREACHED */
2814 * Drop the packet and send a reject notice
2815 * if the packet is not ICMP (or is an ICMP
2816 * query), and it is not multicast/broadcast.
2818 if (hlen > 0 && is_ipv4 && offset == 0 &&
2819 (proto != IPPROTO_ICMP ||
2820 is_icmp_query(ICMP(ulp))) &&
2821 !(m->m_flags & (M_BCAST|M_MCAST)) &&
2822 !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2823 send_reject(args, cmd->arg1, iplen, ip);
2829 if (hlen > 0 && is_ipv6 &&
2830 ((offset & IP6F_OFF_MASK) == 0) &&
2831 (proto != IPPROTO_ICMPV6 ||
2832 (is_icmp6_query(icmp6_type) == 1)) &&
2833 !(m->m_flags & (M_BCAST|M_MCAST)) &&
2834 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
2836 args, cmd->arg1, hlen,
2837 (struct ip6_hdr *)ip);
2843 retval = IP_FW_DENY;
2844 l = 0; /* exit inner loop */
2845 done = 1; /* exit outer loop */
2849 if (args->eh) /* not valid on layer2 pkts */
2851 if (q != f || dyn_dir == MATCH_FORWARD) {
2852 struct sockaddr_in *sa;
2854 sa = &(((ipfw_insn_sa *)cmd)->sa);
2855 if (sa->sin_addr.s_addr == INADDR_ANY) {
2858 * We use O_FORWARD_IP opcode for
2859 * fwd rule with tablearg, but tables
2860 * now support IPv6 addresses. And
2861 * when we are inspecting IPv6 packet,
2862 * we can use nh6 field from
2863 * table_value as next_hop6 address.
2866 struct sockaddr_in6 *sa6;
2868 sa6 = args->next_hop6 =
2870 sa6->sin6_family = AF_INET6;
2871 sa6->sin6_len = sizeof(*sa6);
2872 sa6->sin6_addr = TARG_VAL(
2873 chain, tablearg, nh6);
2874 sa6->sin6_port = sa->sin_port;
2876 * Set sin6_scope_id only for
2877 * link-local unicast addresses.
2879 if (IN6_IS_ADDR_LINKLOCAL(
2881 sa6->sin6_scope_id =
2888 args->hopstore.sin_port =
2890 sa = args->next_hop =
2892 sa->sin_family = AF_INET;
2893 sa->sin_len = sizeof(*sa);
2894 sa->sin_addr.s_addr = htonl(
2895 TARG_VAL(chain, tablearg,
2899 args->next_hop = sa;
2902 retval = IP_FW_PASS;
2903 l = 0; /* exit inner loop */
2904 done = 1; /* exit outer loop */
2909 if (args->eh) /* not valid on layer2 pkts */
2911 if (q != f || dyn_dir == MATCH_FORWARD) {
2912 struct sockaddr_in6 *sin6;
2914 sin6 = &(((ipfw_insn_sa6 *)cmd)->sa);
2915 args->next_hop6 = sin6;
2917 retval = IP_FW_PASS;
2918 l = 0; /* exit inner loop */
2919 done = 1; /* exit outer loop */
2925 set_match(args, f_pos, chain);
2926 args->rule.info = TARG(cmd->arg1, netgraph);
2928 args->rule.info |= IPFW_ONEPASS;
2929 retval = (cmd->opcode == O_NETGRAPH) ?
2930 IP_FW_NETGRAPH : IP_FW_NGTEE;
2931 l = 0; /* exit inner loop */
2932 done = 1; /* exit outer loop */
2938 IPFW_INC_RULE_COUNTER(f, pktlen);
2939 fib = TARG(cmd->arg1, fib) & 0x7FFF;
2940 if (fib >= rt_numfibs)
2943 args->f_id.fib = fib;
2944 l = 0; /* exit inner loop */
2951 code = TARG(cmd->arg1, dscp) & 0x3F;
2952 l = 0; /* exit inner loop */
2956 old = *(uint16_t *)ip;
2957 ip->ip_tos = (code << 2) |
2958 (ip->ip_tos & 0x03);
2959 ip->ip_sum = cksum_adjust(ip->ip_sum,
2960 old, *(uint16_t *)ip);
2961 } else if (is_ipv6) {
2964 v = &((struct ip6_hdr *)ip)->ip6_vfc;
2965 *v = (*v & 0xF0) | (code >> 2);
2967 *v = (*v & 0x3F) | ((code & 0x03) << 6);
2971 IPFW_INC_RULE_COUNTER(f, pktlen);
2976 l = 0; /* exit inner loop */
2977 done = 1; /* exit outer loop */
2979 * Ensure that we do not invoke NAT handler for
2980 * non IPv4 packets. Libalias expects only IPv4.
2982 if (!is_ipv4 || !IPFW_NAT_LOADED) {
2983 retval = IP_FW_DENY;
2990 set_match(args, f_pos, chain);
2991 /* Check if this is 'global' nat rule */
2992 if (cmd->arg1 == IP_FW_NAT44_GLOBAL) {
2993 retval = ipfw_nat_ptr(args, NULL, m);
2996 t = ((ipfw_insn_nat *)cmd)->nat;
2998 nat_id = TARG(cmd->arg1, nat);
2999 t = (*lookup_nat_ptr)(&chain->nat, nat_id);
3002 retval = IP_FW_DENY;
3005 if (cmd->arg1 != IP_FW_TARG)
3006 ((ipfw_insn_nat *)cmd)->nat = t;
3008 retval = ipfw_nat_ptr(args, t, m);
3014 IPFW_INC_RULE_COUNTER(f, pktlen);
3015 l = 0; /* in any case exit inner loop */
3016 ip_off = ntohs(ip->ip_off);
3018 /* if not fragmented, go to next rule */
3019 if ((ip_off & (IP_MF | IP_OFFMASK)) == 0)
3022 args->m = m = ip_reass(m);
3025 * do IP header checksum fixup.
3027 if (m == NULL) { /* fragment got swallowed */
3028 retval = IP_FW_DENY;
3029 } else { /* good, packet complete */
3032 ip = mtod(m, struct ip *);
3033 hlen = ip->ip_hl << 2;
3035 if (hlen == sizeof(struct ip))
3036 ip->ip_sum = in_cksum_hdr(ip);
3038 ip->ip_sum = in_cksum(m, hlen);
3039 retval = IP_FW_REASS;
3040 set_match(args, f_pos, chain);
3042 done = 1; /* exit outer loop */
3045 case O_EXTERNAL_ACTION:
3046 l = 0; /* in any case exit inner loop */
3047 retval = ipfw_run_eaction(chain, args,
3050 * If both @retval and @done are zero,
3051 * consider this as rule matching and
3054 if (retval == 0 && done == 0) {
3055 IPFW_INC_RULE_COUNTER(f, pktlen);
3057 * Reset the result of the last
3058 * dynamic state lookup.
3059 * External action can change
3060 * @args content, and it may be
3061 * used for new state lookup later.
3063 dyn_dir = MATCH_UNKNOWN;
3068 panic("-- unknown opcode %d\n", cmd->opcode);
3069 } /* end of switch() on opcodes */
3071 * if we get here with l=0, then match is irrelevant.
3074 if (cmd->len & F_NOT)
3078 if (cmd->len & F_OR)
3081 if (!(cmd->len & F_OR)) /* not an OR block, */
3082 break; /* try next rule */
3085 } /* end of inner loop, scan opcodes */
3091 /* next_rule:; */ /* try next rule */
3093 } /* end of outer for, scan rules */
3096 struct ip_fw *rule = chain->map[f_pos];
3097 /* Update statistics */
3098 IPFW_INC_RULE_COUNTER(rule, pktlen);
3100 retval = IP_FW_DENY;
3101 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
3103 IPFW_PF_RUNLOCK(chain);
3105 if (ucred_cache != NULL)
3106 crfree(ucred_cache);
3112 printf("ipfw: pullup failed\n");
3113 return (IP_FW_DENY);
3117 * Set maximum number of tables that can be used in given VNET ipfw instance.
3121 sysctl_ipfw_table_num(SYSCTL_HANDLER_ARGS)
3124 unsigned int ntables;
3126 ntables = V_fw_tables_max;
3128 error = sysctl_handle_int(oidp, &ntables, 0, req);
3129 /* Read operation or some error */
3130 if ((error != 0) || (req->newptr == NULL))
3133 return (ipfw_resize_tables(&V_layer3_chain, ntables));
3137 * Switches table namespace between global and per-set.
3140 sysctl_ipfw_tables_sets(SYSCTL_HANDLER_ARGS)
3145 sets = V_fw_tables_sets;
3147 error = sysctl_handle_int(oidp, &sets, 0, req);
3148 /* Read operation or some error */
3149 if ((error != 0) || (req->newptr == NULL))
3152 return (ipfw_switch_tables_namespace(&V_layer3_chain, sets));
3157 * Module and VNET glue
3161 * Stuff that must be initialised only on boot or module load
3169 * Only print out this stuff the first time around,
3170 * when called from the sysinit code.
3176 "initialized, divert %s, nat %s, "
3177 "default to %s, logging ",
3183 #ifdef IPFIREWALL_NAT
3188 default_to_accept ? "accept" : "deny");
3191 * Note: V_xxx variables can be accessed here but the vnet specific
3192 * initializer may not have been called yet for the VIMAGE case.
3193 * Tuneables will have been processed. We will print out values for
3195 * XXX This should all be rationalized AFTER 8.0
3197 if (V_fw_verbose == 0)
3198 printf("disabled\n");
3199 else if (V_verbose_limit == 0)
3200 printf("unlimited\n");
3202 printf("limited to %d packets/entry by default\n",
3205 /* Check user-supplied table count for validness */
3206 if (default_fw_tables > IPFW_TABLES_MAX)
3207 default_fw_tables = IPFW_TABLES_MAX;
3209 ipfw_init_sopt_handler();
3210 ipfw_init_obj_rewriter();
3216 * Called for the removal of the last instance only on module unload.
3222 ipfw_iface_destroy();
3223 ipfw_destroy_sopt_handler();
3224 ipfw_destroy_obj_rewriter();
3225 printf("IP firewall unloaded\n");
3229 * Stuff that must be initialized for every instance
3230 * (including the first of course).
3233 vnet_ipfw_init(const void *unused)
3236 struct ip_fw *rule = NULL;
3237 struct ip_fw_chain *chain;
3239 chain = &V_layer3_chain;
3241 first = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3243 /* First set up some values that are compile time options */
3244 V_autoinc_step = 100; /* bounded to 1..1000 in add_rule() */
3245 V_fw_deny_unknown_exthdrs = 1;
3246 #ifdef IPFIREWALL_VERBOSE
3249 #ifdef IPFIREWALL_VERBOSE_LIMIT
3250 V_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
3252 #ifdef IPFIREWALL_NAT
3253 LIST_INIT(&chain->nat);
3256 /* Init shared services hash table */
3257 ipfw_init_srv(chain);
3259 ipfw_init_counters();
3260 /* Set initial number of tables */
3261 V_fw_tables_max = default_fw_tables;
3262 error = ipfw_init_tables(chain, first);
3264 printf("ipfw2: setting up tables failed\n");
3265 free(chain->map, M_IPFW);
3270 IPFW_LOCK_INIT(chain);
3272 /* fill and insert the default rule */
3273 rule = ipfw_alloc_rule(chain, sizeof(struct ip_fw));
3275 rule->cmd[0].len = 1;
3276 rule->cmd[0].opcode = default_to_accept ? O_ACCEPT : O_DENY;
3277 chain->default_rule = rule;
3278 ipfw_add_protected_rule(chain, rule, 0);
3280 ipfw_dyn_init(chain);
3281 ipfw_eaction_init(chain, first);
3282 #ifdef LINEAR_SKIPTO
3283 ipfw_init_skipto_cache(chain);
3285 ipfw_bpf_init(first);
3287 /* First set up some values that are compile time options */
3288 V_ipfw_vnet_ready = 1; /* Open for business */
3291 * Hook the sockopt handler and pfil hooks for ipv4 and ipv6.
3292 * Even if the latter two fail we still keep the module alive
3293 * because the sockopt and layer2 paths are still useful.
3294 * ipfw[6]_hook return 0 on success, ENOENT on failure,
3295 * so we can ignore the exact return value and just set a flag.
3297 * Note that V_fw[6]_enable are manipulated by a SYSCTL_PROC so
3298 * changes in the underlying (per-vnet) variables trigger
3299 * immediate hook()/unhook() calls.
3300 * In layer2 we have the same behaviour, except that V_ether_ipfw
3301 * is checked on each packet because there are no pfil hooks.
3303 V_ip_fw_ctl_ptr = ipfw_ctl3;
3304 error = ipfw_attach_hooks(1);
3309 * Called for the removal of each instance.
3312 vnet_ipfw_uninit(const void *unused)
3315 struct ip_fw_chain *chain = &V_layer3_chain;
3318 V_ipfw_vnet_ready = 0; /* tell new callers to go away */
3320 * disconnect from ipv4, ipv6, layer2 and sockopt.
3321 * Then grab, release and grab again the WLOCK so we make
3322 * sure the update is propagated and nobody will be in.
3324 (void)ipfw_attach_hooks(0 /* detach */);
3325 V_ip_fw_ctl_ptr = NULL;
3327 last = IS_DEFAULT_VNET(curvnet) ? 1 : 0;
3329 IPFW_UH_WLOCK(chain);
3330 IPFW_UH_WUNLOCK(chain);
3332 ipfw_dyn_uninit(0); /* run the callout_drain */
3334 IPFW_UH_WLOCK(chain);
3338 for (i = 0; i < chain->n_rules; i++)
3339 ipfw_reap_add(chain, &reap, chain->map[i]);
3340 free(chain->map, M_IPFW);
3341 #ifdef LINEAR_SKIPTO
3342 ipfw_destroy_skipto_cache(chain);
3344 IPFW_WUNLOCK(chain);
3345 IPFW_UH_WUNLOCK(chain);
3346 ipfw_destroy_tables(chain, last);
3347 ipfw_eaction_uninit(chain, last);
3349 ipfw_reap_rules(reap);
3350 vnet_ipfw_iface_destroy(chain);
3351 ipfw_destroy_srv(chain);
3352 IPFW_LOCK_DESTROY(chain);
3353 ipfw_dyn_uninit(1); /* free the remaining parts */
3354 ipfw_destroy_counters();
3355 ipfw_bpf_uninit(last);
3360 * Module event handler.
3361 * In general we have the choice of handling most of these events by the
3362 * event handler or by the (VNET_)SYS(UN)INIT handlers. I have chosen to
3363 * use the SYSINIT handlers as they are more capable of expressing the
3364 * flow of control during module and vnet operations, so this is just
3365 * a skeleton. Note there is no SYSINIT equivalent of the module
3366 * SHUTDOWN handler, but we don't have anything to do in that case anyhow.
3369 ipfw_modevent(module_t mod, int type, void *unused)
3375 /* Called once at module load or
3376 * system boot if compiled in. */
3379 /* Called before unload. May veto unloading. */
3382 /* Called during unload. */
3385 /* Called during system shutdown. */
3394 static moduledata_t ipfwmod = {
3400 /* Define startup order. */
3401 #define IPFW_SI_SUB_FIREWALL SI_SUB_PROTO_FIREWALL
3402 #define IPFW_MODEVENT_ORDER (SI_ORDER_ANY - 255) /* On boot slot in here. */
3403 #define IPFW_MODULE_ORDER (IPFW_MODEVENT_ORDER + 1) /* A little later. */
3404 #define IPFW_VNET_ORDER (IPFW_MODEVENT_ORDER + 2) /* Later still. */
3406 DECLARE_MODULE(ipfw, ipfwmod, IPFW_SI_SUB_FIREWALL, IPFW_MODEVENT_ORDER);
3407 FEATURE(ipfw_ctl3, "ipfw new sockopt calls");
3408 MODULE_VERSION(ipfw, 3);
3409 /* should declare some dependencies here */
3412 * Starting up. Done in order after ipfwmod() has been called.
3413 * VNET_SYSINIT is also called for each existing vnet and each new vnet.
3415 SYSINIT(ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3417 VNET_SYSINIT(vnet_ipfw_init, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3418 vnet_ipfw_init, NULL);
3421 * Closing up shop. These are done in REVERSE ORDER, but still
3422 * after ipfwmod() has been called. Not called on reboot.
3423 * VNET_SYSUNINIT is also called for each exiting vnet as it exits.
3424 * or when the module is unloaded.
3426 SYSUNINIT(ipfw_destroy, IPFW_SI_SUB_FIREWALL, IPFW_MODULE_ORDER,
3427 ipfw_destroy, NULL);
3428 VNET_SYSUNINIT(vnet_ipfw_uninit, IPFW_SI_SUB_FIREWALL, IPFW_VNET_ORDER,
3429 vnet_ipfw_uninit, NULL);