2 * Copyright (c) 2002 Luigi Rizzo, Universita` di Pisa
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions
7 * 1. Redistributions of source code must retain the above copyright
8 * notice, this list of conditions and the following disclaimer.
9 * 2. Redistributions in binary form must reproduce the above copyright
10 * notice, this list of conditions and the following disclaimer in the
11 * documentation and/or other materials provided with the distribution.
13 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
14 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
15 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
16 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
17 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
18 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
19 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
20 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
21 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
22 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 #include <sys/cdefs.h>
27 __FBSDID("$FreeBSD$");
33 * Implement IP packet firewall (new version)
36 #if !defined(KLD_MODULE)
38 #include "opt_ipdivert.h"
42 #error IPFIREWALL requires INET.
45 #include "opt_inet6.h"
46 #include "opt_ipsec.h"
49 #include <sys/param.h>
50 #include <sys/systm.h>
51 #include <sys/malloc.h>
53 #include <sys/kernel.h>
56 #include <sys/module.h>
59 #include <sys/socket.h>
60 #include <sys/socketvar.h>
61 #include <sys/sysctl.h>
62 #include <sys/syslog.h>
63 #include <sys/ucred.h>
64 #include <net/ethernet.h> /* for ETHERTYPE_IP */
66 #include <net/radix.h>
67 #include <net/route.h>
68 #include <net/pf_mtag.h>
70 #define IPFW_INTERNAL /* Access to protected data structures in ip_fw.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_divert.h>
80 #include <netinet/ip_dummynet.h>
81 #include <netinet/ip_carp.h>
82 #include <netinet/pim.h>
83 #include <netinet/tcp_var.h>
84 #include <netinet/udp.h>
85 #include <netinet/udp_var.h>
86 #include <netinet/sctp.h>
87 #include <netgraph/ng_ipfw.h>
89 #include <netinet/ip6.h>
90 #include <netinet/icmp6.h>
92 #include <netinet6/scope6_var.h>
95 #include <machine/in_cksum.h> /* XXX for in_cksum */
98 #include <security/mac/mac_framework.h>
102 * set_disable contains one bit per set value (0..31).
103 * If the bit is set, all rules with the corresponding set
104 * are disabled. Set RESVD_SET(31) is reserved for the default rule
105 * and rules that are not deleted by the flush command,
106 * and CANNOT be disabled.
107 * Rules in set RESVD_SET can only be deleted explicitly.
109 static u_int32_t set_disable;
110 static int fw_verbose;
111 static struct callout ipfw_timeout;
112 static int verbose_limit;
114 static uma_zone_t ipfw_dyn_rule_zone;
117 * Data structure to cache our ucred related
118 * information. This structure only gets used if
119 * the user specified UID/GID based constraints in
123 gid_t fw_groups[NGROUPS];
130 * list of rules for layer 3
132 struct ip_fw_chain layer3_chain;
134 MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's");
135 MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables");
136 #define IPFW_NAT_LOADED (ipfw_nat_ptr != NULL)
137 ipfw_nat_t *ipfw_nat_ptr = NULL;
138 ipfw_nat_cfg_t *ipfw_nat_cfg_ptr;
139 ipfw_nat_cfg_t *ipfw_nat_del_ptr;
140 ipfw_nat_cfg_t *ipfw_nat_get_cfg_ptr;
141 ipfw_nat_cfg_t *ipfw_nat_get_log_ptr;
144 struct radix_node rn[2];
145 struct sockaddr_in addr, mask;
149 static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */
151 extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS);
154 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
155 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable,
156 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &fw_enable, 0,
157 ipfw_chg_hook, "I", "Enable ipfw");
158 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW,
159 &autoinc_step, 0, "Rule number autincrement step");
160 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
161 CTLFLAG_RW | CTLFLAG_SECURE3,
163 "Only do a single pass through ipfw when using dummynet(4)");
164 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose,
165 CTLFLAG_RW | CTLFLAG_SECURE3,
166 &fw_verbose, 0, "Log matches to ipfw rules");
167 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW,
168 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged");
169 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, default_rule, CTLFLAG_RD,
170 NULL, IPFW_DEFAULT_RULE, "The default/max possible rule number.");
171 SYSCTL_UINT(_net_inet_ip_fw, OID_AUTO, tables_max, CTLFLAG_RD,
172 NULL, IPFW_TABLES_MAX, "The maximum number of tables.");
173 #endif /* SYSCTL_NODE */
176 * Description of dynamic rules.
178 * Dynamic rules are stored in lists accessed through a hash table
179 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can
180 * be modified through the sysctl variable dyn_buckets which is
181 * updated when the table becomes empty.
183 * XXX currently there is only one list, ipfw_dyn.
185 * When a packet is received, its address fields are first masked
186 * with the mask defined for the rule, then hashed, then matched
187 * against the entries in the corresponding list.
188 * Dynamic rules can be used for different purposes:
190 * + enforcing limits on the number of sessions;
191 * + in-kernel NAT (not implemented yet)
193 * The lifetime of dynamic rules is regulated by dyn_*_lifetime,
194 * measured in seconds and depending on the flags.
196 * The total number of dynamic rules is stored in dyn_count.
197 * The max number of dynamic rules is dyn_max. When we reach
198 * the maximum number of rules we do not create anymore. This is
199 * done to avoid consuming too much memory, but also too much
200 * time when searching on each packet (ideally, we should try instead
201 * to put a limit on the length of the list on each bucket...).
203 * Each dynamic rule holds a pointer to the parent ipfw rule so
204 * we know what action to perform. Dynamic rules are removed when
205 * the parent rule is deleted. XXX we should make them survive.
207 * There are some limitations with dynamic rules -- we do not
208 * obey the 'randomized match', and we do not do multiple
209 * passes through the firewall. XXX check the latter!!!
211 static ipfw_dyn_rule **ipfw_dyn_v = NULL;
212 static u_int32_t dyn_buckets = 256; /* must be power of 2 */
213 static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */
215 static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */
216 #define IPFW_DYN_LOCK_INIT() \
217 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF)
218 #define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx)
219 #define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx)
220 #define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx)
221 #define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED)
224 * Timeouts for various events in handing dynamic rules.
226 static u_int32_t dyn_ack_lifetime = 300;
227 static u_int32_t dyn_syn_lifetime = 20;
228 static u_int32_t dyn_fin_lifetime = 1;
229 static u_int32_t dyn_rst_lifetime = 1;
230 static u_int32_t dyn_udp_lifetime = 10;
231 static u_int32_t dyn_short_lifetime = 5;
234 * Keepalives are sent if dyn_keepalive is set. They are sent every
235 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval
236 * seconds of lifetime of a rule.
237 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower
238 * than dyn_keepalive_period.
241 static u_int32_t dyn_keepalive_interval = 20;
242 static u_int32_t dyn_keepalive_period = 5;
243 static u_int32_t dyn_keepalive = 1; /* do send keepalives */
245 static u_int32_t static_count; /* # of static rules */
246 static u_int32_t static_len; /* size in bytes of static rules */
247 static u_int32_t dyn_count; /* # of dynamic rules */
248 static u_int32_t dyn_max = 4096; /* max # of dynamic rules */
251 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW,
252 &dyn_buckets, 0, "Number of dyn. buckets");
253 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD,
254 &curr_dyn_buckets, 0, "Current Number of dyn. buckets");
255 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD,
256 &dyn_count, 0, "Number of dyn. rules");
257 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW,
258 &dyn_max, 0, "Max number of dyn. rules");
259 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD,
260 &static_count, 0, "Number of static rules");
261 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW,
262 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks");
263 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW,
264 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn");
265 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW,
266 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin");
267 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW,
268 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst");
269 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW,
270 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP");
271 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW,
272 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations");
273 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW,
274 &dyn_keepalive, 0, "Enable keepalives for dyn. rules");
275 #endif /* SYSCTL_NODE */
279 * IPv6 specific variables
282 SYSCTL_DECL(_net_inet6_ip6);
283 #endif /* SYSCTL_NODE */
285 static struct sysctl_ctx_list ip6_fw_sysctl_ctx;
286 static struct sysctl_oid *ip6_fw_sysctl_tree;
289 static int fw_deny_unknown_exthdrs = 1;
293 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
294 * Other macros just cast void * into the appropriate type
296 #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
297 #define TCP(p) ((struct tcphdr *)(p))
298 #define SCTP(p) ((struct sctphdr *)(p))
299 #define UDP(p) ((struct udphdr *)(p))
300 #define ICMP(p) ((struct icmphdr *)(p))
301 #define ICMP6(p) ((struct icmp6_hdr *)(p))
304 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
306 int type = icmp->icmp_type;
308 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
311 #define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
312 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
315 is_icmp_query(struct icmphdr *icmp)
317 int type = icmp->icmp_type;
319 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
324 * The following checks use two arrays of 8 or 16 bits to store the
325 * bits that we want set or clear, respectively. They are in the
326 * low and high half of cmd->arg1 or cmd->d[0].
328 * We scan options and store the bits we find set. We succeed if
330 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
332 * The code is sometimes optimized not to store additional variables.
336 flags_match(ipfw_insn *cmd, u_int8_t bits)
341 if ( ((cmd->arg1 & 0xff) & bits) != 0)
342 return 0; /* some bits we want set were clear */
343 want_clear = (cmd->arg1 >> 8) & 0xff;
344 if ( (want_clear & bits) != want_clear)
345 return 0; /* some bits we want clear were set */
350 ipopts_match(struct ip *ip, ipfw_insn *cmd)
352 int optlen, bits = 0;
353 u_char *cp = (u_char *)(ip + 1);
354 int x = (ip->ip_hl << 2) - sizeof (struct ip);
356 for (; x > 0; x -= optlen, cp += optlen) {
357 int opt = cp[IPOPT_OPTVAL];
359 if (opt == IPOPT_EOL)
361 if (opt == IPOPT_NOP)
364 optlen = cp[IPOPT_OLEN];
365 if (optlen <= 0 || optlen > x)
366 return 0; /* invalid or truncated */
374 bits |= IP_FW_IPOPT_LSRR;
378 bits |= IP_FW_IPOPT_SSRR;
382 bits |= IP_FW_IPOPT_RR;
386 bits |= IP_FW_IPOPT_TS;
390 return (flags_match(cmd, bits));
394 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
396 int optlen, bits = 0;
397 u_char *cp = (u_char *)(tcp + 1);
398 int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
400 for (; x > 0; x -= optlen, cp += optlen) {
402 if (opt == TCPOPT_EOL)
404 if (opt == TCPOPT_NOP)
418 bits |= IP_FW_TCPOPT_MSS;
422 bits |= IP_FW_TCPOPT_WINDOW;
425 case TCPOPT_SACK_PERMITTED:
427 bits |= IP_FW_TCPOPT_SACK;
430 case TCPOPT_TIMESTAMP:
431 bits |= IP_FW_TCPOPT_TS;
436 return (flags_match(cmd, bits));
440 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd)
442 if (ifp == NULL) /* no iface with this packet, match fails */
444 /* Check by name or by IP address */
445 if (cmd->name[0] != '\0') { /* match by name */
448 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
451 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
458 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
459 if (ia->ifa_addr->sa_family != AF_INET)
461 if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
462 (ia->ifa_addr))->sin_addr.s_addr)
463 return(1); /* match */
466 return(0); /* no match, fail ... */
470 * The verify_path function checks if a route to the src exists and
471 * if it is reachable via ifp (when provided).
473 * The 'verrevpath' option checks that the interface that an IP packet
474 * arrives on is the same interface that traffic destined for the
475 * packet's source address would be routed out of. The 'versrcreach'
476 * option just checks that the source address is reachable via any route
477 * (except default) in the routing table. These two are a measure to block
478 * forged packets. This is also commonly known as "anti-spoofing" or Unicast
479 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
480 * is purposely reminiscent of the Cisco IOS command,
482 * ip verify unicast reverse-path
483 * ip verify unicast source reachable-via any
485 * which implements the same functionality. But note that syntax is
486 * misleading. The check may be performed on all IP packets whether unicast,
487 * multicast, or broadcast.
490 verify_path(struct in_addr src, struct ifnet *ifp, u_int fib)
493 struct sockaddr_in *dst;
495 bzero(&ro, sizeof(ro));
497 dst = (struct sockaddr_in *)&(ro.ro_dst);
498 dst->sin_family = AF_INET;
499 dst->sin_len = sizeof(*dst);
501 in_rtalloc_ign(&ro, RTF_CLONING, fib);
503 if (ro.ro_rt == NULL)
507 * If ifp is provided, check for equality with rtentry.
508 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
509 * in order to pass packets injected back by if_simloop():
510 * if useloopback == 1 routing entry (via lo0) for our own address
511 * may exist, so we need to handle routing assymetry.
513 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
518 /* if no ifp provided, check if rtentry is not default route */
520 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) {
525 /* or if this is a blackhole/reject route */
526 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
531 /* found valid route */
538 * ipv6 specific rules here...
541 icmp6type_match (int type, ipfw_insn_u32 *cmd)
543 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
547 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
550 for (i=0; i <= cmd->o.arg1; ++i )
551 if (curr_flow == cmd->d[i] )
556 /* support for IP6_*_ME opcodes */
558 search_ip6_addr_net (struct in6_addr * ip6_addr)
562 struct in6_ifaddr *fdm;
563 struct in6_addr copia;
565 TAILQ_FOREACH(mdc, &ifnet, if_link)
566 TAILQ_FOREACH(mdc2, &mdc->if_addrlist, ifa_list) {
567 if (mdc2->ifa_addr->sa_family == AF_INET6) {
568 fdm = (struct in6_ifaddr *)mdc2;
569 copia = fdm->ia_addr.sin6_addr;
570 /* need for leaving scope_id in the sock_addr */
571 in6_clearscope(&copia);
572 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia))
580 verify_path6(struct in6_addr *src, struct ifnet *ifp)
583 struct sockaddr_in6 *dst;
585 bzero(&ro, sizeof(ro));
587 dst = (struct sockaddr_in6 * )&(ro.ro_dst);
588 dst->sin6_family = AF_INET6;
589 dst->sin6_len = sizeof(*dst);
590 dst->sin6_addr = *src;
591 /* XXX MRT 0 for ipv6 at this time */
592 rtalloc_ign((struct route *)&ro, RTF_CLONING);
594 if (ro.ro_rt == NULL)
598 * if ifp is provided, check for equality with rtentry
599 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
600 * to support the case of sending packets to an address of our own.
601 * (where the former interface is the first argument of if_simloop()
602 * (=ifp), the latter is lo0)
604 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
609 /* if no ifp provided, check if rtentry is not default route */
611 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) {
616 /* or if this is a blackhole/reject route */
617 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
622 /* found valid route */
628 hash_packet6(struct ipfw_flow_id *id)
631 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^
632 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^
633 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^
634 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^
635 (id->dst_port) ^ (id->src_port);
640 is_icmp6_query(int icmp6_type)
642 if ((icmp6_type <= ICMP6_MAXTYPE) &&
643 (icmp6_type == ICMP6_ECHO_REQUEST ||
644 icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
645 icmp6_type == ICMP6_WRUREQUEST ||
646 icmp6_type == ICMP6_FQDN_QUERY ||
647 icmp6_type == ICMP6_NI_QUERY))
654 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
659 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
667 tcp = (struct tcphdr *)((char *)ip6 + hlen);
669 if ((tcp->th_flags & TH_RST) != 0) {
677 ti.th.th_seq = ntohl(ti.th.th_seq);
678 ti.th.th_ack = ntohl(ti.th.th_ack);
679 ti.ip6.ip6_nxt = IPPROTO_TCP;
681 if (ti.th.th_flags & TH_ACK) {
687 if ((m->m_flags & M_PKTHDR) != 0) {
689 * total new data to ACK is:
690 * total packet length,
691 * minus the header length,
692 * minus the tcp header length.
694 ack += m->m_pkthdr.len - hlen
695 - (ti.th.th_off << 2);
696 } else if (ip6->ip6_plen) {
697 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) -
698 hlen - (ti.th.th_off << 2);
703 if (tcp->th_flags & TH_SYN)
706 flags = TH_RST|TH_ACK;
708 bcopy(&ti, ip6, sizeof(ti));
710 * m is only used to recycle the mbuf
711 * The data in it is never read so we don't need
712 * to correct the offsets or anything
714 tcp_respond(NULL, ip6, tcp, m, ack, seq, flags);
715 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
718 * Unlike above, the mbufs need to line up with the ip6 hdr,
719 * as the contents are read. We need to m_adj() the
721 * The mbuf will however be thrown away so we can adjust it.
722 * Remember we did an m_pullup on it already so we
723 * can make some assumptions about contiguousness.
726 m_adj(m, args->L3offset);
728 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
737 static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */
739 #define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0
740 #define SNP(buf) buf, sizeof(buf)
743 * We enter here when we have a rule with O_LOG.
744 * XXX this function alone takes about 2Kbytes of code!
747 ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args,
748 struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg,
751 struct ether_header *eh = args->eh;
753 int limit_reached = 0;
754 char action2[40], proto[128], fragment[32];
759 if (f == NULL) { /* bogus pkt */
760 if (verbose_limit != 0 && norule_counter >= verbose_limit)
763 if (norule_counter == verbose_limit)
764 limit_reached = verbose_limit;
766 } else { /* O_LOG is the first action, find the real one */
767 ipfw_insn *cmd = ACTION_PTR(f);
768 ipfw_insn_log *l = (ipfw_insn_log *)cmd;
770 if (l->max_log != 0 && l->log_left == 0)
773 if (l->log_left == 0)
774 limit_reached = l->max_log;
775 cmd += F_LEN(cmd); /* point to first action */
776 if (cmd->opcode == O_ALTQ) {
777 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
779 snprintf(SNPARGS(action2, 0), "Altq %d",
783 if (cmd->opcode == O_PROB)
786 if (cmd->opcode == O_TAG)
790 switch (cmd->opcode) {
796 if (cmd->arg1==ICMP_REJECT_RST)
798 else if (cmd->arg1==ICMP_UNREACH_HOST)
801 snprintf(SNPARGS(action2, 0), "Unreach %d",
806 if (cmd->arg1==ICMP6_UNREACH_RST)
809 snprintf(SNPARGS(action2, 0), "Unreach %d",
820 snprintf(SNPARGS(action2, 0), "Divert %d",
824 snprintf(SNPARGS(action2, 0), "Tee %d",
828 snprintf(SNPARGS(action2, 0), "SetFib %d",
832 snprintf(SNPARGS(action2, 0), "SkipTo %d",
836 snprintf(SNPARGS(action2, 0), "Pipe %d",
840 snprintf(SNPARGS(action2, 0), "Queue %d",
844 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd;
846 struct in_addr dummyaddr;
847 if (sa->sa.sin_addr.s_addr == INADDR_ANY)
848 dummyaddr.s_addr = htonl(tablearg);
850 dummyaddr.s_addr = sa->sa.sin_addr.s_addr;
852 len = snprintf(SNPARGS(action2, 0), "Forward to %s",
853 inet_ntoa(dummyaddr));
856 snprintf(SNPARGS(action2, len), ":%d",
861 snprintf(SNPARGS(action2, 0), "Netgraph %d",
865 snprintf(SNPARGS(action2, 0), "Ngtee %d",
877 if (hlen == 0) { /* non-ip */
878 snprintf(SNPARGS(proto, 0), "MAC");
882 char src[48], dst[48];
883 struct icmphdr *icmp;
887 struct ip6_hdr *ip6 = NULL;
888 struct icmp6_hdr *icmp6;
893 if (IS_IP6_FLOW_ID(&(args->f_id))) {
894 char ip6buf[INET6_ADDRSTRLEN];
895 snprintf(src, sizeof(src), "[%s]",
896 ip6_sprintf(ip6buf, &args->f_id.src_ip6));
897 snprintf(dst, sizeof(dst), "[%s]",
898 ip6_sprintf(ip6buf, &args->f_id.dst_ip6));
900 ip6 = (struct ip6_hdr *)ip;
901 tcp = (struct tcphdr *)(((char *)ip) + hlen);
902 udp = (struct udphdr *)(((char *)ip) + hlen);
906 tcp = L3HDR(struct tcphdr, ip);
907 udp = L3HDR(struct udphdr, ip);
909 inet_ntoa_r(ip->ip_src, src);
910 inet_ntoa_r(ip->ip_dst, dst);
913 switch (args->f_id.proto) {
915 len = snprintf(SNPARGS(proto, 0), "TCP %s", src);
917 snprintf(SNPARGS(proto, len), ":%d %s:%d",
918 ntohs(tcp->th_sport),
920 ntohs(tcp->th_dport));
922 snprintf(SNPARGS(proto, len), " %s", dst);
926 len = snprintf(SNPARGS(proto, 0), "UDP %s", src);
928 snprintf(SNPARGS(proto, len), ":%d %s:%d",
929 ntohs(udp->uh_sport),
931 ntohs(udp->uh_dport));
933 snprintf(SNPARGS(proto, len), " %s", dst);
937 icmp = L3HDR(struct icmphdr, ip);
939 len = snprintf(SNPARGS(proto, 0),
941 icmp->icmp_type, icmp->icmp_code);
943 len = snprintf(SNPARGS(proto, 0), "ICMP ");
944 len += snprintf(SNPARGS(proto, len), "%s", src);
945 snprintf(SNPARGS(proto, len), " %s", dst);
949 icmp6 = (struct icmp6_hdr *)(((char *)ip) + hlen);
951 len = snprintf(SNPARGS(proto, 0),
953 icmp6->icmp6_type, icmp6->icmp6_code);
955 len = snprintf(SNPARGS(proto, 0), "ICMPv6 ");
956 len += snprintf(SNPARGS(proto, len), "%s", src);
957 snprintf(SNPARGS(proto, len), " %s", dst);
961 len = snprintf(SNPARGS(proto, 0), "P:%d %s",
962 args->f_id.proto, src);
963 snprintf(SNPARGS(proto, len), " %s", dst);
968 if (IS_IP6_FLOW_ID(&(args->f_id))) {
969 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG))
970 snprintf(SNPARGS(fragment, 0),
971 " (frag %08x:%d@%d%s)",
973 ntohs(ip6->ip6_plen) - hlen,
974 ntohs(offset & IP6F_OFF_MASK) << 3,
975 (offset & IP6F_MORE_FRAG) ? "+" : "");
980 if (eh != NULL) { /* layer 2 packets are as on the wire */
981 ip_off = ntohs(ip->ip_off);
982 ip_len = ntohs(ip->ip_len);
987 if (ip_off & (IP_MF | IP_OFFMASK))
988 snprintf(SNPARGS(fragment, 0),
989 " (frag %d:%d@%d%s)",
990 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2),
992 (ip_off & IP_MF) ? "+" : "");
995 if (oif || m->m_pkthdr.rcvif)
996 log(LOG_SECURITY | LOG_INFO,
997 "ipfw: %d %s %s %s via %s%s\n",
999 action, proto, oif ? "out" : "in",
1000 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname,
1003 log(LOG_SECURITY | LOG_INFO,
1004 "ipfw: %d %s %s [no if info]%s\n",
1005 f ? f->rulenum : -1,
1006 action, proto, fragment);
1008 log(LOG_SECURITY | LOG_NOTICE,
1009 "ipfw: limit %d reached on entry %d\n",
1010 limit_reached, f ? f->rulenum : -1);
1014 * IMPORTANT: the hash function for dynamic rules must be commutative
1015 * in source and destination (ip,port), because rules are bidirectional
1016 * and we want to find both in the same bucket.
1019 hash_packet(struct ipfw_flow_id *id)
1024 if (IS_IP6_FLOW_ID(id))
1025 i = hash_packet6(id);
1028 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port);
1029 i &= (curr_dyn_buckets - 1);
1034 * unlink a dynamic rule from a chain. prev is a pointer to
1035 * the previous one, q is a pointer to the rule to delete,
1036 * head is a pointer to the head of the queue.
1037 * Modifies q and potentially also head.
1039 #define UNLINK_DYN_RULE(prev, head, q) { \
1040 ipfw_dyn_rule *old_q = q; \
1042 /* remove a refcount to the parent */ \
1043 if (q->dyn_type == O_LIMIT) \
1044 q->parent->count--; \
1045 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\
1046 (q->id.src_ip), (q->id.src_port), \
1047 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \
1049 prev->next = q = q->next; \
1051 head = q = q->next; \
1053 uma_zfree(ipfw_dyn_rule_zone, old_q); }
1055 #define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0)
1058 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL.
1060 * If keep_me == NULL, rules are deleted even if not expired,
1061 * otherwise only expired rules are removed.
1063 * The value of the second parameter is also used to point to identify
1064 * a rule we absolutely do not want to remove (e.g. because we are
1065 * holding a reference to it -- this is the case with O_LIMIT_PARENT
1066 * rules). The pointer is only used for comparison, so any non-null
1070 remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me)
1072 static u_int32_t last_remove = 0;
1074 #define FORCE (keep_me == NULL)
1076 ipfw_dyn_rule *prev, *q;
1077 int i, pass = 0, max_pass = 0;
1079 IPFW_DYN_LOCK_ASSERT();
1081 if (ipfw_dyn_v == NULL || dyn_count == 0)
1083 /* do not expire more than once per second, it is useless */
1084 if (!FORCE && last_remove == time_uptime)
1086 last_remove = time_uptime;
1089 * because O_LIMIT refer to parent rules, during the first pass only
1090 * remove child and mark any pending LIMIT_PARENT, and remove
1091 * them in a second pass.
1094 for (i = 0 ; i < curr_dyn_buckets ; i++) {
1095 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) {
1097 * Logic can become complex here, so we split tests.
1101 if (rule != NULL && rule != q->rule)
1102 goto next; /* not the one we are looking for */
1103 if (q->dyn_type == O_LIMIT_PARENT) {
1105 * handle parent in the second pass,
1106 * record we need one.
1111 if (FORCE && q->count != 0 ) {
1112 /* XXX should not happen! */
1113 printf("ipfw: OUCH! cannot remove rule,"
1114 " count %d\n", q->count);
1118 !TIME_LEQ( q->expire, time_uptime ))
1121 if (q->dyn_type != O_LIMIT_PARENT || !q->count) {
1122 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
1130 if (pass++ < max_pass)
1136 * lookup a dynamic rule.
1138 static ipfw_dyn_rule *
1139 lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction,
1143 * stateful ipfw extensions.
1144 * Lookup into dynamic session queue
1146 #define MATCH_REVERSE 0
1147 #define MATCH_FORWARD 1
1148 #define MATCH_NONE 2
1149 #define MATCH_UNKNOWN 3
1150 int i, dir = MATCH_NONE;
1151 ipfw_dyn_rule *prev, *q=NULL;
1153 IPFW_DYN_LOCK_ASSERT();
1155 if (ipfw_dyn_v == NULL)
1156 goto done; /* not found */
1157 i = hash_packet( pkt );
1158 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) {
1159 if (q->dyn_type == O_LIMIT_PARENT && q->count)
1161 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */
1162 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
1165 if (pkt->proto == q->id.proto &&
1166 q->dyn_type != O_LIMIT_PARENT) {
1167 if (IS_IP6_FLOW_ID(pkt)) {
1168 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6),
1169 &(q->id.src_ip6)) &&
1170 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6),
1171 &(q->id.dst_ip6)) &&
1172 pkt->src_port == q->id.src_port &&
1173 pkt->dst_port == q->id.dst_port ) {
1174 dir = MATCH_FORWARD;
1177 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6),
1178 &(q->id.dst_ip6)) &&
1179 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6),
1180 &(q->id.src_ip6)) &&
1181 pkt->src_port == q->id.dst_port &&
1182 pkt->dst_port == q->id.src_port ) {
1183 dir = MATCH_REVERSE;
1187 if (pkt->src_ip == q->id.src_ip &&
1188 pkt->dst_ip == q->id.dst_ip &&
1189 pkt->src_port == q->id.src_port &&
1190 pkt->dst_port == q->id.dst_port ) {
1191 dir = MATCH_FORWARD;
1194 if (pkt->src_ip == q->id.dst_ip &&
1195 pkt->dst_ip == q->id.src_ip &&
1196 pkt->src_port == q->id.dst_port &&
1197 pkt->dst_port == q->id.src_port ) {
1198 dir = MATCH_REVERSE;
1208 goto done; /* q = NULL, not found */
1210 if ( prev != NULL) { /* found and not in front */
1211 prev->next = q->next;
1212 q->next = ipfw_dyn_v[i];
1215 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */
1216 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST);
1218 #define BOTH_SYN (TH_SYN | (TH_SYN << 8))
1219 #define BOTH_FIN (TH_FIN | (TH_FIN << 8))
1220 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8);
1222 case TH_SYN: /* opening */
1223 q->expire = time_uptime + dyn_syn_lifetime;
1226 case BOTH_SYN: /* move to established */
1227 case BOTH_SYN | TH_FIN : /* one side tries to close */
1228 case BOTH_SYN | (TH_FIN << 8) :
1230 #define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0)
1231 u_int32_t ack = ntohl(tcp->th_ack);
1232 if (dir == MATCH_FORWARD) {
1233 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd))
1235 else { /* ignore out-of-sequence */
1239 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev))
1241 else { /* ignore out-of-sequence */
1246 q->expire = time_uptime + dyn_ack_lifetime;
1249 case BOTH_SYN | BOTH_FIN: /* both sides closed */
1250 if (dyn_fin_lifetime >= dyn_keepalive_period)
1251 dyn_fin_lifetime = dyn_keepalive_period - 1;
1252 q->expire = time_uptime + dyn_fin_lifetime;
1258 * reset or some invalid combination, but can also
1259 * occur if we use keep-state the wrong way.
1261 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0)
1262 printf("invalid state: 0x%x\n", q->state);
1264 if (dyn_rst_lifetime >= dyn_keepalive_period)
1265 dyn_rst_lifetime = dyn_keepalive_period - 1;
1266 q->expire = time_uptime + dyn_rst_lifetime;
1269 } else if (pkt->proto == IPPROTO_UDP) {
1270 q->expire = time_uptime + dyn_udp_lifetime;
1272 /* other protocols */
1273 q->expire = time_uptime + dyn_short_lifetime;
1276 if (match_direction)
1277 *match_direction = dir;
1281 static ipfw_dyn_rule *
1282 lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction,
1288 q = lookup_dyn_rule_locked(pkt, match_direction, tcp);
1291 /* NB: return table locked when q is not NULL */
1296 realloc_dynamic_table(void)
1298 IPFW_DYN_LOCK_ASSERT();
1301 * Try reallocation, make sure we have a power of 2 and do
1302 * not allow more than 64k entries. In case of overflow,
1306 if (dyn_buckets > 65536)
1308 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */
1309 dyn_buckets = curr_dyn_buckets; /* reset */
1312 curr_dyn_buckets = dyn_buckets;
1313 if (ipfw_dyn_v != NULL)
1314 free(ipfw_dyn_v, M_IPFW);
1316 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *),
1317 M_IPFW, M_NOWAIT | M_ZERO);
1318 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2)
1320 curr_dyn_buckets /= 2;
1325 * Install state of type 'type' for a dynamic session.
1326 * The hash table contains two type of rules:
1327 * - regular rules (O_KEEP_STATE)
1328 * - rules for sessions with limited number of sess per user
1329 * (O_LIMIT). When they are created, the parent is
1330 * increased by 1, and decreased on delete. In this case,
1331 * the third parameter is the parent rule and not the chain.
1332 * - "parent" rules for the above (O_LIMIT_PARENT).
1334 static ipfw_dyn_rule *
1335 add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule)
1340 IPFW_DYN_LOCK_ASSERT();
1342 if (ipfw_dyn_v == NULL ||
1343 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) {
1344 realloc_dynamic_table();
1345 if (ipfw_dyn_v == NULL)
1346 return NULL; /* failed ! */
1348 i = hash_packet(id);
1350 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO);
1352 printf ("ipfw: sorry cannot allocate state\n");
1356 /* increase refcount on parent, and set pointer */
1357 if (dyn_type == O_LIMIT) {
1358 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule;
1359 if ( parent->dyn_type != O_LIMIT_PARENT)
1360 panic("invalid parent");
1363 rule = parent->rule;
1367 r->expire = time_uptime + dyn_syn_lifetime;
1369 r->dyn_type = dyn_type;
1370 r->pcnt = r->bcnt = 0;
1374 r->next = ipfw_dyn_v[i];
1377 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n",
1379 (r->id.src_ip), (r->id.src_port),
1380 (r->id.dst_ip), (r->id.dst_port),
1386 * lookup dynamic parent rule using pkt and rule as search keys.
1387 * If the lookup fails, then install one.
1389 static ipfw_dyn_rule *
1390 lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule)
1395 IPFW_DYN_LOCK_ASSERT();
1398 int is_v6 = IS_IP6_FLOW_ID(pkt);
1399 i = hash_packet( pkt );
1400 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next)
1401 if (q->dyn_type == O_LIMIT_PARENT &&
1403 pkt->proto == q->id.proto &&
1404 pkt->src_port == q->id.src_port &&
1405 pkt->dst_port == q->id.dst_port &&
1408 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6),
1409 &(q->id.src_ip6)) &&
1410 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6),
1411 &(q->id.dst_ip6))) ||
1413 pkt->src_ip == q->id.src_ip &&
1414 pkt->dst_ip == q->id.dst_ip)
1417 q->expire = time_uptime + dyn_short_lifetime;
1418 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);)
1422 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule);
1426 * Install dynamic state for rule type cmd->o.opcode
1428 * Returns 1 (failure) if state is not installed because of errors or because
1429 * session limitations are enforced.
1432 install_state(struct ip_fw *rule, ipfw_insn_limit *cmd,
1433 struct ip_fw_args *args, uint32_t tablearg)
1435 static int last_log;
1438 char src[48], dst[48];
1444 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n",
1445 __func__, cmd->o.opcode,
1446 (args->f_id.src_ip), (args->f_id.src_port),
1447 (args->f_id.dst_ip), (args->f_id.dst_port));
1452 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL);
1454 if (q != NULL) { /* should never occur */
1455 if (last_log != time_uptime) {
1456 last_log = time_uptime;
1457 printf("ipfw: %s: entry already present, done\n",
1464 if (dyn_count >= dyn_max)
1465 /* Run out of slots, try to remove any expired rule. */
1466 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1);
1468 if (dyn_count >= dyn_max) {
1469 if (last_log != time_uptime) {
1470 last_log = time_uptime;
1471 printf("ipfw: %s: Too many dynamic rules\n", __func__);
1474 return (1); /* cannot install, notify caller */
1477 switch (cmd->o.opcode) {
1478 case O_KEEP_STATE: /* bidir rule */
1479 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule);
1482 case O_LIMIT: { /* limit number of sessions */
1483 struct ipfw_flow_id id;
1484 ipfw_dyn_rule *parent;
1485 uint32_t conn_limit;
1486 uint16_t limit_mask = cmd->limit_mask;
1488 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ?
1489 tablearg : cmd->conn_limit;
1492 if (cmd->conn_limit == IP_FW_TABLEARG)
1493 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u "
1494 "(tablearg)\n", __func__, conn_limit);
1496 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n",
1497 __func__, conn_limit);
1500 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0;
1501 id.proto = args->f_id.proto;
1502 id.addr_type = args->f_id.addr_type;
1503 id.fib = M_GETFIB(args->m);
1505 if (IS_IP6_FLOW_ID (&(args->f_id))) {
1506 if (limit_mask & DYN_SRC_ADDR)
1507 id.src_ip6 = args->f_id.src_ip6;
1508 if (limit_mask & DYN_DST_ADDR)
1509 id.dst_ip6 = args->f_id.dst_ip6;
1511 if (limit_mask & DYN_SRC_ADDR)
1512 id.src_ip = args->f_id.src_ip;
1513 if (limit_mask & DYN_DST_ADDR)
1514 id.dst_ip = args->f_id.dst_ip;
1516 if (limit_mask & DYN_SRC_PORT)
1517 id.src_port = args->f_id.src_port;
1518 if (limit_mask & DYN_DST_PORT)
1519 id.dst_port = args->f_id.dst_port;
1520 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) {
1521 printf("ipfw: %s: add parent failed\n", __func__);
1526 if (parent->count >= conn_limit) {
1527 /* See if we can remove some expired rule. */
1528 remove_dyn_rule(rule, parent);
1529 if (parent->count >= conn_limit) {
1530 if (fw_verbose && last_log != time_uptime) {
1531 last_log = time_uptime;
1534 * XXX IPv6 flows are not
1537 if (IS_IP6_FLOW_ID(&(args->f_id))) {
1538 char ip6buf[INET6_ADDRSTRLEN];
1539 snprintf(src, sizeof(src),
1540 "[%s]", ip6_sprintf(ip6buf,
1541 &args->f_id.src_ip6));
1542 snprintf(dst, sizeof(dst),
1543 "[%s]", ip6_sprintf(ip6buf,
1544 &args->f_id.dst_ip6));
1549 htonl(args->f_id.src_ip);
1550 inet_ntoa_r(da, src);
1552 htonl(args->f_id.dst_ip);
1553 inet_ntoa_r(da, dst);
1555 log(LOG_SECURITY | LOG_DEBUG,
1556 "ipfw: %d %s %s:%u -> %s:%u, %s\n",
1557 parent->rule->rulenum,
1559 src, (args->f_id.src_port),
1560 dst, (args->f_id.dst_port),
1561 "too many entries");
1567 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent);
1571 printf("ipfw: %s: unknown dynamic rule type %u\n",
1572 __func__, cmd->o.opcode);
1577 /* XXX just set lifetime */
1578 lookup_dyn_rule_locked(&args->f_id, NULL, NULL);
1585 * Generate a TCP packet, containing either a RST or a keepalive.
1586 * When flags & TH_RST, we are sending a RST packet, because of a
1587 * "reset" action matched the packet.
1588 * Otherwise we are sending a keepalive, and flags & TH_
1589 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required
1590 * so that MAC can label the reply appropriately.
1592 static struct mbuf *
1593 send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq,
1594 u_int32_t ack, int flags)
1600 MGETHDR(m, M_DONTWAIT, MT_DATA);
1603 m->m_pkthdr.rcvif = (struct ifnet *)0;
1605 M_SETFIB(m, id->fib);
1607 if (replyto != NULL)
1608 mac_create_mbuf_netlayer(replyto, m);
1610 mac_create_mbuf_from_firewall(m);
1612 (void)replyto; /* don't warn about unused arg */
1615 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr);
1616 m->m_data += max_linkhdr;
1618 ip = mtod(m, struct ip *);
1619 bzero(ip, m->m_len);
1620 tcp = (struct tcphdr *)(ip + 1); /* no IP options */
1621 ip->ip_p = IPPROTO_TCP;
1624 * Assume we are sending a RST (or a keepalive in the reverse
1625 * direction), swap src and destination addresses and ports.
1627 ip->ip_src.s_addr = htonl(id->dst_ip);
1628 ip->ip_dst.s_addr = htonl(id->src_ip);
1629 tcp->th_sport = htons(id->dst_port);
1630 tcp->th_dport = htons(id->src_port);
1631 if (flags & TH_RST) { /* we are sending a RST */
1632 if (flags & TH_ACK) {
1633 tcp->th_seq = htonl(ack);
1634 tcp->th_ack = htonl(0);
1635 tcp->th_flags = TH_RST;
1639 tcp->th_seq = htonl(0);
1640 tcp->th_ack = htonl(seq);
1641 tcp->th_flags = TH_RST | TH_ACK;
1645 * We are sending a keepalive. flags & TH_SYN determines
1646 * the direction, forward if set, reverse if clear.
1647 * NOTE: seq and ack are always assumed to be correct
1648 * as set by the caller. This may be confusing...
1650 if (flags & TH_SYN) {
1652 * we have to rewrite the correct addresses!
1654 ip->ip_dst.s_addr = htonl(id->dst_ip);
1655 ip->ip_src.s_addr = htonl(id->src_ip);
1656 tcp->th_dport = htons(id->dst_port);
1657 tcp->th_sport = htons(id->src_port);
1659 tcp->th_seq = htonl(seq);
1660 tcp->th_ack = htonl(ack);
1661 tcp->th_flags = TH_ACK;
1664 * set ip_len to the payload size so we can compute
1665 * the tcp checksum on the pseudoheader
1666 * XXX check this, could save a couple of words ?
1668 ip->ip_len = htons(sizeof(struct tcphdr));
1669 tcp->th_sum = in_cksum(m, m->m_pkthdr.len);
1671 * now fill fields left out earlier
1673 ip->ip_ttl = ip_defttl;
1674 ip->ip_len = m->m_pkthdr.len;
1675 m->m_flags |= M_SKIP_FIREWALL;
1680 * sends a reject message, consuming the mbuf passed as an argument.
1683 send_reject(struct ip_fw_args *args, int code, int ip_len, struct ip *ip)
1687 /* XXX When ip is not guaranteed to be at mtod() we will
1688 * need to account for this */
1689 * The mbuf will however be thrown away so we can adjust it.
1690 * Remember we did an m_pullup on it already so we
1691 * can make some assumptions about contiguousness.
1694 m_adj(m, args->L3offset);
1696 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
1697 /* We need the IP header in host order for icmp_error(). */
1698 if (args->eh != NULL) {
1699 ip->ip_len = ntohs(ip->ip_len);
1700 ip->ip_off = ntohs(ip->ip_off);
1702 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
1703 } else if (args->f_id.proto == IPPROTO_TCP) {
1704 struct tcphdr *const tcp =
1705 L3HDR(struct tcphdr, mtod(args->m, struct ip *));
1706 if ( (tcp->th_flags & TH_RST) == 0) {
1708 m = send_pkt(args->m, &(args->f_id),
1709 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
1710 tcp->th_flags | TH_RST);
1712 ip_output(m, NULL, NULL, 0, NULL, NULL);
1722 * Given an ip_fw *, lookup_next_rule will return a pointer
1723 * to the next rule, which can be either the jump
1724 * target (for skipto instructions) or the next one in the list (in
1725 * all other cases including a missing jump target).
1726 * The result is also written in the "next_rule" field of the rule.
1727 * Backward jumps are not allowed, so start looking from the next
1730 * This never returns NULL -- in case we do not have an exact match,
1731 * the next rule is returned. When the ruleset is changed,
1732 * pointers are flushed so we are always correct.
1735 static struct ip_fw *
1736 lookup_next_rule(struct ip_fw *me, u_int32_t tablearg)
1738 struct ip_fw *rule = NULL;
1742 /* look for action, in case it is a skipto */
1743 cmd = ACTION_PTR(me);
1744 if (cmd->opcode == O_LOG)
1746 if (cmd->opcode == O_ALTQ)
1748 if (cmd->opcode == O_TAG)
1750 if (cmd->opcode == O_SKIPTO ) {
1751 if (tablearg != 0) {
1752 rulenum = (u_int16_t)tablearg;
1754 rulenum = cmd->arg1;
1756 for (rule = me->next; rule ; rule = rule->next) {
1757 if (rule->rulenum >= rulenum) {
1762 if (rule == NULL) /* failure or not a skipto */
1764 me->next_rule = rule;
1769 add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr,
1770 uint8_t mlen, uint32_t value)
1772 struct radix_node_head *rnh;
1773 struct table_entry *ent;
1774 struct radix_node *rn;
1776 if (tbl >= IPFW_TABLES_MAX)
1778 rnh = ch->tables[tbl];
1779 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO);
1783 ent->addr.sin_len = ent->mask.sin_len = 8;
1784 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0);
1785 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr;
1787 rn = rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent);
1790 free(ent, M_IPFW_TBL);
1798 del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr,
1801 struct radix_node_head *rnh;
1802 struct table_entry *ent;
1803 struct sockaddr_in sa, mask;
1805 if (tbl >= IPFW_TABLES_MAX)
1807 rnh = ch->tables[tbl];
1808 sa.sin_len = mask.sin_len = 8;
1809 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0);
1810 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr;
1812 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh);
1818 free(ent, M_IPFW_TBL);
1823 flush_table_entry(struct radix_node *rn, void *arg)
1825 struct radix_node_head * const rnh = arg;
1826 struct table_entry *ent;
1828 ent = (struct table_entry *)
1829 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh);
1831 free(ent, M_IPFW_TBL);
1836 flush_table(struct ip_fw_chain *ch, uint16_t tbl)
1838 struct radix_node_head *rnh;
1840 IPFW_WLOCK_ASSERT(ch);
1842 if (tbl >= IPFW_TABLES_MAX)
1844 rnh = ch->tables[tbl];
1845 KASSERT(rnh != NULL, ("NULL IPFW table"));
1846 rnh->rnh_walktree(rnh, flush_table_entry, rnh);
1851 flush_tables(struct ip_fw_chain *ch)
1855 IPFW_WLOCK_ASSERT(ch);
1857 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++)
1858 flush_table(ch, tbl);
1862 init_tables(struct ip_fw_chain *ch)
1867 for (i = 0; i < IPFW_TABLES_MAX; i++) {
1868 if (!rn_inithead((void **)&ch->tables[i], 32)) {
1869 for (j = 0; j < i; j++) {
1870 (void) flush_table(ch, j);
1879 lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr,
1882 struct radix_node_head *rnh;
1883 struct table_entry *ent;
1884 struct sockaddr_in sa;
1886 if (tbl >= IPFW_TABLES_MAX)
1888 rnh = ch->tables[tbl];
1890 sa.sin_addr.s_addr = addr;
1891 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh));
1900 count_table_entry(struct radix_node *rn, void *arg)
1902 u_int32_t * const cnt = arg;
1909 count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt)
1911 struct radix_node_head *rnh;
1913 if (tbl >= IPFW_TABLES_MAX)
1915 rnh = ch->tables[tbl];
1917 rnh->rnh_walktree(rnh, count_table_entry, cnt);
1922 dump_table_entry(struct radix_node *rn, void *arg)
1924 struct table_entry * const n = (struct table_entry *)rn;
1925 ipfw_table * const tbl = arg;
1926 ipfw_table_entry *ent;
1928 if (tbl->cnt == tbl->size)
1930 ent = &tbl->ent[tbl->cnt];
1931 ent->tbl = tbl->tbl;
1932 if (in_nullhost(n->mask.sin_addr))
1935 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr));
1936 ent->addr = n->addr.sin_addr.s_addr;
1937 ent->value = n->value;
1943 dump_table(struct ip_fw_chain *ch, ipfw_table *tbl)
1945 struct radix_node_head *rnh;
1947 if (tbl->tbl >= IPFW_TABLES_MAX)
1949 rnh = ch->tables[tbl->tbl];
1951 rnh->rnh_walktree(rnh, dump_table_entry, tbl);
1956 fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp)
1961 ugp->fw_prid = jailed(cr) ? cr->cr_prison->pr_id : -1;
1962 ugp->fw_uid = cr->cr_uid;
1963 ugp->fw_ngroups = cr->cr_ngroups;
1964 bcopy(cr->cr_groups, ugp->fw_groups, sizeof(ugp->fw_groups));
1968 check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif,
1969 struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip,
1970 u_int16_t src_port, struct ip_fw_ugid *ugp, int *ugid_lookupp,
1973 struct inpcbinfo *pi;
1980 * Check to see if the UDP or TCP stack supplied us with
1981 * the PCB. If so, rather then holding a lock and looking
1982 * up the PCB, we can use the one that was supplied.
1984 if (inp && *ugid_lookupp == 0) {
1985 INP_LOCK_ASSERT(inp);
1986 if (inp->inp_socket != NULL) {
1987 fill_ugid_cache(inp, ugp);
1993 * If we have already been here and the packet has no
1994 * PCB entry associated with it, then we can safely
1995 * assume that this is a no match.
1997 if (*ugid_lookupp == -1)
1999 if (proto == IPPROTO_TCP) {
2002 } else if (proto == IPPROTO_UDP) {
2003 wildcard = INPLOOKUP_WILDCARD;
2008 if (*ugid_lookupp == 0) {
2011 in_pcblookup_hash(pi,
2012 dst_ip, htons(dst_port),
2013 src_ip, htons(src_port),
2015 in_pcblookup_hash(pi,
2016 src_ip, htons(src_port),
2017 dst_ip, htons(dst_port),
2020 fill_ugid_cache(pcb, ugp);
2023 INP_INFO_RUNLOCK(pi);
2024 if (*ugid_lookupp == 0) {
2026 * If the lookup did not yield any results, there
2027 * is no sense in coming back and trying again. So
2028 * we can set lookup to -1 and ensure that we wont
2029 * bother the pcb system again.
2035 if (insn->o.opcode == O_UID)
2036 match = (ugp->fw_uid == (uid_t)insn->d[0]);
2037 else if (insn->o.opcode == O_GID) {
2038 for (gp = ugp->fw_groups;
2039 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++)
2040 if (*gp == (gid_t)insn->d[0]) {
2044 } else if (insn->o.opcode == O_JAIL)
2045 match = (ugp->fw_prid == (int)insn->d[0]);
2050 * The main check routine for the firewall.
2052 * All arguments are in args so we can modify them and return them
2053 * back to the caller.
2057 * args->m (in/out) The packet; we set to NULL when/if we nuke it.
2058 * Starts with the IP header.
2059 * args->eh (in) Mac header if present, or NULL for layer3 packet.
2060 * args->L3offset Number of bytes bypassed if we came from L2.
2061 * e.g. often sizeof(eh) ** NOTYET **
2062 * args->oif Outgoing interface, or NULL if packet is incoming.
2063 * The incoming interface is in the mbuf. (in)
2064 * args->divert_rule (in/out)
2065 * Skip up to the first rule past this rule number;
2066 * upon return, non-zero port number for divert or tee.
2068 * args->rule Pointer to the last matching rule (in/out)
2069 * args->next_hop Socket we are forwarding to (out).
2070 * args->f_id Addresses grabbed from the packet (out)
2071 * args->cookie a cookie depending on rule action
2075 * IP_FW_PASS the packet must be accepted
2076 * IP_FW_DENY the packet must be dropped
2077 * IP_FW_DIVERT divert packet, port in m_tag
2078 * IP_FW_TEE tee packet, port in m_tag
2079 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie
2080 * IP_FW_NETGRAPH into netgraph, cookie args->cookie
2084 ipfw_chk(struct ip_fw_args *args)
2087 * Local variables holding state during the processing of a packet:
2089 * IMPORTANT NOTE: to speed up the processing of rules, there
2090 * are some assumption on the values of the variables, which
2091 * are documented here. Should you change them, please check
2092 * the implementation of the various instructions to make sure
2093 * that they still work.
2095 * args->eh The MAC header. It is non-null for a layer2
2096 * packet, it is NULL for a layer-3 packet.
2098 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
2100 * m | args->m Pointer to the mbuf, as received from the caller.
2101 * It may change if ipfw_chk() does an m_pullup, or if it
2102 * consumes the packet because it calls send_reject().
2103 * XXX This has to change, so that ipfw_chk() never modifies
2104 * or consumes the buffer.
2105 * ip is the beginning of the ip(4 or 6) header.
2106 * Calculated by adding the L3offset to the start of data.
2107 * (Until we start using L3offset, the packet is
2108 * supposed to start with the ip header).
2110 struct mbuf *m = args->m;
2111 struct ip *ip = mtod(m, struct ip *);
2114 * For rules which contain uid/gid or jail constraints, cache
2115 * a copy of the users credentials after the pcb lookup has been
2116 * executed. This will speed up the processing of rules with
2117 * these types of constraints, as well as decrease contention
2118 * on pcb related locks.
2120 struct ip_fw_ugid fw_ugid_cache;
2121 int ugid_lookup = 0;
2124 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG
2125 * associated with a packet input on a divert socket. This
2126 * will allow to distinguish traffic and its direction when
2127 * it originates from a divert socket.
2129 u_int divinput_flags = 0;
2132 * oif | args->oif If NULL, ipfw_chk has been called on the
2133 * inbound path (ether_input, ip_input).
2134 * If non-NULL, ipfw_chk has been called on the outbound path
2135 * (ether_output, ip_output).
2137 struct ifnet *oif = args->oif;
2139 struct ip_fw *f = NULL; /* matching rule */
2143 * hlen The length of the IP header.
2145 u_int hlen = 0; /* hlen >0 means we have an IP pkt */
2148 * offset The offset of a fragment. offset != 0 means that
2149 * we have a fragment at this offset of an IPv4 packet.
2150 * offset == 0 means that (if this is an IPv4 packet)
2151 * this is the first or only fragment.
2152 * For IPv6 offset == 0 means there is no Fragment Header.
2153 * If offset != 0 for IPv6 always use correct mask to
2154 * get the correct offset because we add IP6F_MORE_FRAG
2155 * to be able to dectect the first fragment which would
2156 * otherwise have offset = 0.
2161 * Local copies of addresses. They are only valid if we have
2164 * proto The protocol. Set to 0 for non-ip packets,
2165 * or to the protocol read from the packet otherwise.
2166 * proto != 0 means that we have an IPv4 packet.
2168 * src_port, dst_port port numbers, in HOST format. Only
2169 * valid for TCP and UDP packets.
2171 * src_ip, dst_ip ip addresses, in NETWORK format.
2172 * Only valid for IPv4 packets.
2175 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */
2176 struct in_addr src_ip, dst_ip; /* NOTE: network format */
2179 u_int16_t etype = 0; /* Host order stored ether type */
2182 * dyn_dir = MATCH_UNKNOWN when rules unchecked,
2183 * MATCH_NONE when checked and not matched (q = NULL),
2184 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
2186 int dyn_dir = MATCH_UNKNOWN;
2187 ipfw_dyn_rule *q = NULL;
2188 struct ip_fw_chain *chain = &layer3_chain;
2192 * We store in ulp a pointer to the upper layer protocol header.
2193 * In the ipv4 case this is easy to determine from the header,
2194 * but for ipv6 we might have some additional headers in the middle.
2195 * ulp is NULL if not found.
2197 void *ulp = NULL; /* upper layer protocol pointer. */
2198 /* XXX ipv6 variables */
2200 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */
2201 /* end of ipv6 variables */
2204 if (m->m_flags & M_SKIP_FIREWALL)
2205 return (IP_FW_PASS); /* accept */
2207 dst_ip.s_addr = 0; /* make sure it is initialized */
2208 src_ip.s_addr = 0; /* make sure it is initialized */
2209 pktlen = m->m_pkthdr.len;
2210 args->f_id.fib = M_GETFIB(m); /* note mbuf not altered) */
2211 proto = args->f_id.proto = 0; /* mark f_id invalid */
2212 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
2215 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
2216 * then it sets p to point at the offset "len" in the mbuf. WARNING: the
2217 * pointer might become stale after other pullups (but we never use it
2220 #define PULLUP_TO(_len, p, T) \
2222 int x = (_len) + sizeof(T); \
2223 if ((m)->m_len < x) { \
2224 args->m = m = m_pullup(m, x); \
2226 goto pullup_failed; \
2228 p = (mtod(m, char *) + (_len)); \
2232 * if we have an ether header,
2235 etype = ntohs(args->eh->ether_type);
2237 /* Identify IP packets and fill up variables. */
2238 if (pktlen >= sizeof(struct ip6_hdr) &&
2239 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
2240 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
2242 args->f_id.addr_type = 6;
2243 hlen = sizeof(struct ip6_hdr);
2244 proto = ip6->ip6_nxt;
2246 /* Search extension headers to find upper layer protocols */
2247 while (ulp == NULL) {
2249 case IPPROTO_ICMPV6:
2250 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
2251 args->f_id.flags = ICMP6(ulp)->icmp6_type;
2255 PULLUP_TO(hlen, ulp, struct tcphdr);
2256 dst_port = TCP(ulp)->th_dport;
2257 src_port = TCP(ulp)->th_sport;
2258 args->f_id.flags = TCP(ulp)->th_flags;
2262 PULLUP_TO(hlen, ulp, struct sctphdr);
2263 src_port = SCTP(ulp)->src_port;
2264 dst_port = SCTP(ulp)->dest_port;
2268 PULLUP_TO(hlen, ulp, struct udphdr);
2269 dst_port = UDP(ulp)->uh_dport;
2270 src_port = UDP(ulp)->uh_sport;
2273 case IPPROTO_HOPOPTS: /* RFC 2460 */
2274 PULLUP_TO(hlen, ulp, struct ip6_hbh);
2275 ext_hd |= EXT_HOPOPTS;
2276 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
2277 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
2281 case IPPROTO_ROUTING: /* RFC 2460 */
2282 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
2283 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
2285 ext_hd |= EXT_RTHDR0;
2288 ext_hd |= EXT_RTHDR2;
2291 printf("IPFW2: IPV6 - Unknown Routing "
2292 "Header type(%d)\n",
2293 ((struct ip6_rthdr *)ulp)->ip6r_type);
2294 if (fw_deny_unknown_exthdrs)
2295 return (IP_FW_DENY);
2298 ext_hd |= EXT_ROUTING;
2299 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
2300 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
2304 case IPPROTO_FRAGMENT: /* RFC 2460 */
2305 PULLUP_TO(hlen, ulp, struct ip6_frag);
2306 ext_hd |= EXT_FRAGMENT;
2307 hlen += sizeof (struct ip6_frag);
2308 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
2309 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
2311 /* Add IP6F_MORE_FRAG for offset of first
2312 * fragment to be != 0. */
2313 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg &
2316 printf("IPFW2: IPV6 - Invalid Fragment "
2318 if (fw_deny_unknown_exthdrs)
2319 return (IP_FW_DENY);
2322 args->f_id.frag_id6 =
2323 ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
2327 case IPPROTO_DSTOPTS: /* RFC 2460 */
2328 PULLUP_TO(hlen, ulp, struct ip6_hbh);
2329 ext_hd |= EXT_DSTOPTS;
2330 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
2331 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
2335 case IPPROTO_AH: /* RFC 2402 */
2336 PULLUP_TO(hlen, ulp, struct ip6_ext);
2338 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
2339 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
2343 case IPPROTO_ESP: /* RFC 2406 */
2344 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
2345 /* Anything past Seq# is variable length and
2346 * data past this ext. header is encrypted. */
2350 case IPPROTO_NONE: /* RFC 2460 */
2352 * Packet ends here, and IPv6 header has
2353 * already been pulled up. If ip6e_len!=0
2354 * then octets must be ignored.
2356 ulp = ip; /* non-NULL to get out of loop. */
2359 case IPPROTO_OSPFIGP:
2360 /* XXX OSPF header check? */
2361 PULLUP_TO(hlen, ulp, struct ip6_ext);
2365 /* XXX PIM header check? */
2366 PULLUP_TO(hlen, ulp, struct pim);
2370 PULLUP_TO(hlen, ulp, struct carp_header);
2371 if (((struct carp_header *)ulp)->carp_version !=
2373 return (IP_FW_DENY);
2374 if (((struct carp_header *)ulp)->carp_type !=
2376 return (IP_FW_DENY);
2379 case IPPROTO_IPV6: /* RFC 2893 */
2380 PULLUP_TO(hlen, ulp, struct ip6_hdr);
2383 case IPPROTO_IPV4: /* RFC 2893 */
2384 PULLUP_TO(hlen, ulp, struct ip);
2388 printf("IPFW2: IPV6 - Unknown Extension "
2389 "Header(%d), ext_hd=%x\n", proto, ext_hd);
2390 if (fw_deny_unknown_exthdrs)
2391 return (IP_FW_DENY);
2392 PULLUP_TO(hlen, ulp, struct ip6_ext);
2396 ip = mtod(m, struct ip *);
2397 ip6 = (struct ip6_hdr *)ip;
2398 args->f_id.src_ip6 = ip6->ip6_src;
2399 args->f_id.dst_ip6 = ip6->ip6_dst;
2400 args->f_id.src_ip = 0;
2401 args->f_id.dst_ip = 0;
2402 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
2403 } else if (pktlen >= sizeof(struct ip) &&
2404 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
2406 hlen = ip->ip_hl << 2;
2407 args->f_id.addr_type = 4;
2410 * Collect parameters into local variables for faster matching.
2413 src_ip = ip->ip_src;
2414 dst_ip = ip->ip_dst;
2415 if (args->eh != NULL) { /* layer 2 packets are as on the wire */
2416 offset = ntohs(ip->ip_off) & IP_OFFMASK;
2417 ip_len = ntohs(ip->ip_len);
2419 offset = ip->ip_off & IP_OFFMASK;
2420 ip_len = ip->ip_len;
2422 pktlen = ip_len < pktlen ? ip_len : pktlen;
2427 PULLUP_TO(hlen, ulp, struct tcphdr);
2428 dst_port = TCP(ulp)->th_dport;
2429 src_port = TCP(ulp)->th_sport;
2430 args->f_id.flags = TCP(ulp)->th_flags;
2434 PULLUP_TO(hlen, ulp, struct udphdr);
2435 dst_port = UDP(ulp)->uh_dport;
2436 src_port = UDP(ulp)->uh_sport;
2440 PULLUP_TO(hlen, ulp, struct icmphdr);
2441 args->f_id.flags = ICMP(ulp)->icmp_type;
2449 ip = mtod(m, struct ip *);
2450 args->f_id.src_ip = ntohl(src_ip.s_addr);
2451 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
2454 if (proto) { /* we may have port numbers, store them */
2455 args->f_id.proto = proto;
2456 args->f_id.src_port = src_port = ntohs(src_port);
2457 args->f_id.dst_port = dst_port = ntohs(dst_port);
2461 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL);
2464 * Packet has already been tagged. Look for the next rule
2465 * to restart processing.
2467 * If fw_one_pass != 0 then just accept it.
2468 * XXX should not happen here, but optimized out in
2472 IPFW_RUNLOCK(chain);
2473 return (IP_FW_PASS);
2476 f = args->rule->next_rule;
2478 f = lookup_next_rule(args->rule, 0);
2481 * Find the starting rule. It can be either the first
2482 * one, or the one after divert_rule if asked so.
2484 int skipto = mtag ? divert_cookie(mtag) : 0;
2487 if (args->eh == NULL && skipto != 0) {
2488 if (skipto >= IPFW_DEFAULT_RULE) {
2489 IPFW_RUNLOCK(chain);
2490 return (IP_FW_DENY); /* invalid */
2492 while (f && f->rulenum <= skipto)
2494 if (f == NULL) { /* drop packet */
2495 IPFW_RUNLOCK(chain);
2496 return (IP_FW_DENY);
2500 /* reset divert rule to avoid confusion later */
2502 divinput_flags = divert_info(mtag) &
2503 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG);
2504 m_tag_delete(m, mtag);
2508 * Now scan the rules, and parse microinstructions for each rule.
2510 for (; f; f = f->next) {
2512 uint32_t tablearg = 0;
2513 int l, cmdlen, skip_or; /* skip rest of OR block */
2516 if (set_disable & (1 << f->set) )
2520 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
2521 l -= cmdlen, cmd += cmdlen) {
2525 * check_body is a jump target used when we find a
2526 * CHECK_STATE, and need to jump to the body of
2531 cmdlen = F_LEN(cmd);
2533 * An OR block (insn_1 || .. || insn_n) has the
2534 * F_OR bit set in all but the last instruction.
2535 * The first match will set "skip_or", and cause
2536 * the following instructions to be skipped until
2537 * past the one with the F_OR bit clear.
2539 if (skip_or) { /* skip this instruction */
2540 if ((cmd->len & F_OR) == 0)
2541 skip_or = 0; /* next one is good */
2544 match = 0; /* set to 1 if we succeed */
2546 switch (cmd->opcode) {
2548 * The first set of opcodes compares the packet's
2549 * fields with some pattern, setting 'match' if a
2550 * match is found. At the end of the loop there is
2551 * logic to deal with F_NOT and F_OR flags associated
2559 printf("ipfw: opcode %d unimplemented\n",
2567 * We only check offset == 0 && proto != 0,
2568 * as this ensures that we have a
2569 * packet with the ports info.
2573 if (is_ipv6) /* XXX to be fixed later */
2575 if (proto == IPPROTO_TCP ||
2576 proto == IPPROTO_UDP)
2577 match = check_uidgid(
2578 (ipfw_insn_u32 *)cmd,
2581 src_ip, src_port, &fw_ugid_cache,
2582 &ugid_lookup, args->inp);
2586 match = iface_match(m->m_pkthdr.rcvif,
2587 (ipfw_insn_if *)cmd);
2591 match = iface_match(oif, (ipfw_insn_if *)cmd);
2595 match = iface_match(oif ? oif :
2596 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd);
2600 if (args->eh != NULL) { /* have MAC header */
2601 u_int32_t *want = (u_int32_t *)
2602 ((ipfw_insn_mac *)cmd)->addr;
2603 u_int32_t *mask = (u_int32_t *)
2604 ((ipfw_insn_mac *)cmd)->mask;
2605 u_int32_t *hdr = (u_int32_t *)args->eh;
2608 ( want[0] == (hdr[0] & mask[0]) &&
2609 want[1] == (hdr[1] & mask[1]) &&
2610 want[2] == (hdr[2] & mask[2]) );
2615 if (args->eh != NULL) {
2617 ((ipfw_insn_u16 *)cmd)->ports;
2620 for (i = cmdlen - 1; !match && i>0;
2622 match = (etype >= p[0] &&
2628 match = (offset != 0);
2631 case O_IN: /* "out" is "not in" */
2632 match = (oif == NULL);
2636 match = (args->eh != NULL);
2640 match = (cmd->arg1 & 1 && divinput_flags &
2641 IP_FW_DIVERT_LOOPBACK_FLAG) ||
2642 (cmd->arg1 & 2 && divinput_flags &
2643 IP_FW_DIVERT_OUTPUT_FLAG);
2648 * We do not allow an arg of 0 so the
2649 * check of "proto" only suffices.
2651 match = (proto == cmd->arg1);
2656 (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2660 case O_IP_SRC_LOOKUP:
2661 case O_IP_DST_LOOKUP:
2664 (cmd->opcode == O_IP_DST_LOOKUP) ?
2665 dst_ip.s_addr : src_ip.s_addr;
2668 match = lookup_table(chain, cmd->arg1, a,
2672 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
2674 ((ipfw_insn_u32 *)cmd)->d[0] == v;
2684 (cmd->opcode == O_IP_DST_MASK) ?
2685 dst_ip.s_addr : src_ip.s_addr;
2686 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2689 for (; !match && i>0; i-= 2, p+= 2)
2690 match = (p[0] == (a & p[1]));
2698 INADDR_TO_IFP(src_ip, tif);
2699 match = (tif != NULL);
2706 u_int32_t *d = (u_int32_t *)(cmd+1);
2708 cmd->opcode == O_IP_DST_SET ?
2714 addr -= d[0]; /* subtract base */
2715 match = (addr < cmd->arg1) &&
2716 ( d[ 1 + (addr>>5)] &
2717 (1<<(addr & 0x1f)) );
2723 (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2731 INADDR_TO_IFP(dst_ip, tif);
2732 match = (tif != NULL);
2739 * offset == 0 && proto != 0 is enough
2740 * to guarantee that we have a
2741 * packet with port info.
2743 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
2746 (cmd->opcode == O_IP_SRCPORT) ?
2747 src_port : dst_port ;
2749 ((ipfw_insn_u16 *)cmd)->ports;
2752 for (i = cmdlen - 1; !match && i>0;
2754 match = (x>=p[0] && x<=p[1]);
2759 match = (offset == 0 && proto==IPPROTO_ICMP &&
2760 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
2765 match = is_ipv6 && offset == 0 &&
2766 proto==IPPROTO_ICMPV6 &&
2768 ICMP6(ulp)->icmp6_type,
2769 (ipfw_insn_u32 *)cmd);
2775 ipopts_match(ip, cmd) );
2780 cmd->arg1 == ip->ip_v);
2786 if (is_ipv4) { /* only for IP packets */
2791 if (cmd->opcode == O_IPLEN)
2793 else if (cmd->opcode == O_IPTTL)
2795 else /* must be IPID */
2796 x = ntohs(ip->ip_id);
2798 match = (cmd->arg1 == x);
2801 /* otherwise we have ranges */
2802 p = ((ipfw_insn_u16 *)cmd)->ports;
2804 for (; !match && i>0; i--, p += 2)
2805 match = (x >= p[0] && x <= p[1]);
2809 case O_IPPRECEDENCE:
2811 (cmd->arg1 == (ip->ip_tos & 0xe0)) );
2816 flags_match(cmd, ip->ip_tos));
2820 if (proto == IPPROTO_TCP && offset == 0) {
2828 ((ip->ip_hl + tcp->th_off) << 2);
2830 match = (cmd->arg1 == x);
2833 /* otherwise we have ranges */
2834 p = ((ipfw_insn_u16 *)cmd)->ports;
2836 for (; !match && i>0; i--, p += 2)
2837 match = (x >= p[0] && x <= p[1]);
2842 match = (proto == IPPROTO_TCP && offset == 0 &&
2843 flags_match(cmd, TCP(ulp)->th_flags));
2847 match = (proto == IPPROTO_TCP && offset == 0 &&
2848 tcpopts_match(TCP(ulp), cmd));
2852 match = (proto == IPPROTO_TCP && offset == 0 &&
2853 ((ipfw_insn_u32 *)cmd)->d[0] ==
2858 match = (proto == IPPROTO_TCP && offset == 0 &&
2859 ((ipfw_insn_u32 *)cmd)->d[0] ==
2864 match = (proto == IPPROTO_TCP && offset == 0 &&
2865 cmd->arg1 == TCP(ulp)->th_win);
2869 /* reject packets which have SYN only */
2870 /* XXX should i also check for TH_ACK ? */
2871 match = (proto == IPPROTO_TCP && offset == 0 &&
2872 (TCP(ulp)->th_flags &
2873 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
2878 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
2881 at = pf_find_mtag(m);
2882 if (at != NULL && at->qid != 0)
2884 at = pf_get_mtag(m);
2887 * Let the packet fall back to the
2892 at->qid = altq->qid;
2903 ipfw_log(f, hlen, args, m,
2904 oif, offset, tablearg, ip);
2909 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
2913 /* Outgoing packets automatically pass/match */
2914 match = ((oif != NULL) ||
2915 (m->m_pkthdr.rcvif == NULL) ||
2919 verify_path6(&(args->f_id.src_ip6),
2920 m->m_pkthdr.rcvif) :
2922 verify_path(src_ip, m->m_pkthdr.rcvif,
2927 /* Outgoing packets automatically pass/match */
2928 match = (hlen > 0 && ((oif != NULL) ||
2931 verify_path6(&(args->f_id.src_ip6),
2934 verify_path(src_ip, NULL, args->f_id.fib)));
2938 /* Outgoing packets automatically pass/match */
2939 if (oif == NULL && hlen > 0 &&
2940 ( (is_ipv4 && in_localaddr(src_ip))
2943 in6_localaddr(&(args->f_id.src_ip6)))
2948 is_ipv6 ? verify_path6(
2949 &(args->f_id.src_ip6),
2950 m->m_pkthdr.rcvif) :
2961 match = (m_tag_find(m,
2962 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
2964 /* otherwise no match */
2970 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
2971 &((ipfw_insn_ip6 *)cmd)->addr6);
2976 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
2977 &((ipfw_insn_ip6 *)cmd)->addr6);
2979 case O_IP6_SRC_MASK:
2980 case O_IP6_DST_MASK:
2984 struct in6_addr *d =
2985 &((ipfw_insn_ip6 *)cmd)->addr6;
2987 for (; !match && i > 0; d += 2,
2988 i -= F_INSN_SIZE(struct in6_addr)
2994 APPLY_MASK(&p, &d[1]);
2996 IN6_ARE_ADDR_EQUAL(&d[0],
3003 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6);
3007 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6);
3012 flow6id_match(args->f_id.flow_id6,
3013 (ipfw_insn_u32 *) cmd);
3018 (ext_hd & ((ipfw_insn *) cmd)->arg1);
3031 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
3032 tablearg : cmd->arg1;
3034 /* Packet is already tagged with this tag? */
3035 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
3037 /* We have `untag' action when F_NOT flag is
3038 * present. And we must remove this mtag from
3039 * mbuf and reset `match' to zero (`match' will
3040 * be inversed later).
3041 * Otherwise we should allocate new mtag and
3042 * push it into mbuf.
3044 if (cmd->len & F_NOT) { /* `untag' action */
3046 m_tag_delete(m, mtag);
3047 } else if (mtag == NULL) {
3048 if ((mtag = m_tag_alloc(MTAG_IPFW,
3049 tag, 0, M_NOWAIT)) != NULL)
3050 m_tag_prepend(m, mtag);
3052 match = (cmd->len & F_NOT) ? 0: 1;
3056 case O_FIB: /* try match the specified fib */
3057 if (args->f_id.fib == cmd->arg1)
3062 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
3063 tablearg : cmd->arg1;
3066 match = m_tag_locate(m, MTAG_IPFW,
3071 /* we have ranges */
3072 for (mtag = m_tag_first(m);
3073 mtag != NULL && !match;
3074 mtag = m_tag_next(m, mtag)) {
3078 if (mtag->m_tag_cookie != MTAG_IPFW)
3081 p = ((ipfw_insn_u16 *)cmd)->ports;
3083 for(; !match && i > 0; i--, p += 2)
3085 mtag->m_tag_id >= p[0] &&
3086 mtag->m_tag_id <= p[1];
3092 * The second set of opcodes represents 'actions',
3093 * i.e. the terminal part of a rule once the packet
3094 * matches all previous patterns.
3095 * Typically there is only one action for each rule,
3096 * and the opcode is stored at the end of the rule
3097 * (but there are exceptions -- see below).
3099 * In general, here we set retval and terminate the
3100 * outer loop (would be a 'break 3' in some language,
3101 * but we need to do a 'goto done').
3104 * O_COUNT and O_SKIPTO actions:
3105 * instead of terminating, we jump to the next rule
3106 * ('goto next_rule', equivalent to a 'break 2'),
3107 * or to the SKIPTO target ('goto again' after
3108 * having set f, cmd and l), respectively.
3110 * O_TAG, O_LOG and O_ALTQ action parameters:
3111 * perform some action and set match = 1;
3113 * O_LIMIT and O_KEEP_STATE: these opcodes are
3114 * not real 'actions', and are stored right
3115 * before the 'action' part of the rule.
3116 * These opcodes try to install an entry in the
3117 * state tables; if successful, we continue with
3118 * the next opcode (match=1; break;), otherwise
3119 * the packet * must be dropped
3120 * ('goto done' after setting retval);
3122 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
3123 * cause a lookup of the state table, and a jump
3124 * to the 'action' part of the parent rule
3125 * ('goto check_body') if an entry is found, or
3126 * (CHECK_STATE only) a jump to the next rule if
3127 * the entry is not found ('goto next_rule').
3128 * The result of the lookup is cached to make
3129 * further instances of these opcodes are
3134 if (install_state(f,
3135 (ipfw_insn_limit *)cmd, args, tablearg)) {
3136 retval = IP_FW_DENY;
3137 goto done; /* error/limit violation */
3145 * dynamic rules are checked at the first
3146 * keep-state or check-state occurrence,
3147 * with the result being stored in dyn_dir.
3148 * The compiler introduces a PROBE_STATE
3149 * instruction for us when we have a
3150 * KEEP_STATE (because PROBE_STATE needs
3153 if (dyn_dir == MATCH_UNKNOWN &&
3154 (q = lookup_dyn_rule(&args->f_id,
3155 &dyn_dir, proto == IPPROTO_TCP ?
3159 * Found dynamic entry, update stats
3160 * and jump to the 'action' part of
3166 cmd = ACTION_PTR(f);
3167 l = f->cmd_len - f->act_ofs;
3172 * Dynamic entry not found. If CHECK_STATE,
3173 * skip to next rule, if PROBE_STATE just
3174 * ignore and continue with next opcode.
3176 if (cmd->opcode == O_CHECK_STATE)
3182 retval = 0; /* accept */
3187 args->rule = f; /* report matching rule */
3188 if (cmd->arg1 == IP_FW_TABLEARG)
3189 args->cookie = tablearg;
3191 args->cookie = cmd->arg1;
3192 retval = IP_FW_DUMMYNET;
3197 struct divert_tag *dt;
3199 if (args->eh) /* not on layer 2 */
3201 mtag = m_tag_get(PACKET_TAG_DIVERT,
3202 sizeof(struct divert_tag),
3207 IPFW_RUNLOCK(chain);
3208 return (IP_FW_DENY);
3210 dt = (struct divert_tag *)(mtag+1);
3211 dt->cookie = f->rulenum;
3212 if (cmd->arg1 == IP_FW_TABLEARG)
3213 dt->info = tablearg;
3215 dt->info = cmd->arg1;
3216 m_tag_prepend(m, mtag);
3217 retval = (cmd->opcode == O_DIVERT) ?
3218 IP_FW_DIVERT : IP_FW_TEE;
3224 f->pcnt++; /* update stats */
3226 f->timestamp = time_uptime;
3227 if (cmd->opcode == O_COUNT)
3230 if (cmd->arg1 == IP_FW_TABLEARG) {
3231 f = lookup_next_rule(f, tablearg);
3233 if (f->next_rule == NULL)
3234 lookup_next_rule(f, 0);
3241 * Drop the packet and send a reject notice
3242 * if the packet is not ICMP (or is an ICMP
3243 * query), and it is not multicast/broadcast.
3245 if (hlen > 0 && is_ipv4 && offset == 0 &&
3246 (proto != IPPROTO_ICMP ||
3247 is_icmp_query(ICMP(ulp))) &&
3248 !(m->m_flags & (M_BCAST|M_MCAST)) &&
3249 !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
3250 send_reject(args, cmd->arg1, ip_len, ip);
3256 if (hlen > 0 && is_ipv6 &&
3257 ((offset & IP6F_OFF_MASK) == 0) &&
3258 (proto != IPPROTO_ICMPV6 ||
3259 (is_icmp6_query(args->f_id.flags) == 1)) &&
3260 !(m->m_flags & (M_BCAST|M_MCAST)) &&
3261 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
3263 args, cmd->arg1, hlen,
3264 (struct ip6_hdr *)ip);
3270 retval = IP_FW_DENY;
3273 case O_FORWARD_IP: {
3274 struct sockaddr_in *sa;
3275 sa = &(((ipfw_insn_sa *)cmd)->sa);
3276 if (args->eh) /* not valid on layer2 pkts */
3278 if (!q || dyn_dir == MATCH_FORWARD) {
3279 if (sa->sin_addr.s_addr == INADDR_ANY) {
3280 bcopy(sa, &args->hopstore,
3282 args->hopstore.sin_addr.s_addr =
3287 args->next_hop = sa;
3290 retval = IP_FW_PASS;
3296 args->rule = f; /* report matching rule */
3297 if (cmd->arg1 == IP_FW_TABLEARG)
3298 args->cookie = tablearg;
3300 args->cookie = cmd->arg1;
3301 retval = (cmd->opcode == O_NETGRAPH) ?
3302 IP_FW_NETGRAPH : IP_FW_NGTEE;
3306 f->pcnt++; /* update stats */
3308 f->timestamp = time_uptime;
3309 M_SETFIB(m, cmd->arg1);
3310 args->f_id.fib = cmd->arg1;
3317 if (IPFW_NAT_LOADED) {
3318 args->rule = f; /* Report matching rule. */
3319 t = ((ipfw_insn_nat *)cmd)->nat;
3321 nat_id = (cmd->arg1 == IP_FW_TABLEARG) ?
3322 tablearg : cmd->arg1;
3323 LOOKUP_NAT(layer3_chain, nat_id, t);
3325 retval = IP_FW_DENY;
3328 if (cmd->arg1 != IP_FW_TABLEARG)
3329 ((ipfw_insn_nat *)cmd)->nat = t;
3331 retval = ipfw_nat_ptr(args, t, m);
3333 retval = IP_FW_DENY;
3338 panic("-- unknown opcode %d\n", cmd->opcode);
3339 } /* end of switch() on opcodes */
3341 if (cmd->len & F_NOT)
3345 if (cmd->len & F_OR)
3348 if (!(cmd->len & F_OR)) /* not an OR block, */
3349 break; /* try next rule */
3352 } /* end of inner for, scan opcodes */
3354 next_rule:; /* try next rule */
3356 } /* end of outer for, scan rules */
3357 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
3358 IPFW_RUNLOCK(chain);
3359 return (IP_FW_DENY);
3362 /* Update statistics */
3365 f->timestamp = time_uptime;
3366 IPFW_RUNLOCK(chain);
3371 printf("ipfw: pullup failed\n");
3372 return (IP_FW_DENY);
3376 * When a rule is added/deleted, clear the next_rule pointers in all rules.
3377 * These will be reconstructed on the fly as packets are matched.
3380 flush_rule_ptrs(struct ip_fw_chain *chain)
3384 IPFW_WLOCK_ASSERT(chain);
3386 for (rule = chain->rules; rule; rule = rule->next)
3387 rule->next_rule = NULL;
3391 * Add a new rule to the list. Copy the rule into a malloc'ed area, then
3392 * possibly create a rule number and add the rule to the list.
3393 * Update the rule_number in the input struct so the caller knows it as well.
3396 add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule)
3398 struct ip_fw *rule, *f, *prev;
3399 int l = RULESIZE(input_rule);
3401 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE)
3404 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO);
3408 bcopy(input_rule, rule, l);
3411 rule->next_rule = NULL;
3415 rule->timestamp = 0;
3419 if (chain->rules == NULL) { /* default rule */
3420 chain->rules = rule;
3425 * If rulenum is 0, find highest numbered rule before the
3426 * default rule, and add autoinc_step
3428 if (autoinc_step < 1)
3430 else if (autoinc_step > 1000)
3431 autoinc_step = 1000;
3432 if (rule->rulenum == 0) {
3434 * locate the highest numbered rule before default
3436 for (f = chain->rules; f; f = f->next) {
3437 if (f->rulenum == IPFW_DEFAULT_RULE)
3439 rule->rulenum = f->rulenum;
3441 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step)
3442 rule->rulenum += autoinc_step;
3443 input_rule->rulenum = rule->rulenum;
3447 * Now insert the new rule in the right place in the sorted list.
3449 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) {
3450 if (f->rulenum > rule->rulenum) { /* found the location */
3454 } else { /* head insert */
3455 rule->next = chain->rules;
3456 chain->rules = rule;
3461 flush_rule_ptrs(chain);
3465 IPFW_WUNLOCK(chain);
3466 DEB(printf("ipfw: installed rule %d, static count now %d\n",
3467 rule->rulenum, static_count);)
3472 * Remove a static rule (including derived * dynamic rules)
3473 * and place it on the ``reap list'' for later reclamation.
3474 * The caller is in charge of clearing rule pointers to avoid
3475 * dangling pointers.
3476 * @return a pointer to the next entry.
3477 * Arguments are not checked, so they better be correct.
3479 static struct ip_fw *
3480 remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule,
3484 int l = RULESIZE(rule);
3486 IPFW_WLOCK_ASSERT(chain);
3490 remove_dyn_rule(rule, NULL /* force removal */);
3499 rule->next = chain->reap;
3506 * Reclaim storage associated with a list of rules. This is
3507 * typically the list created using remove_rule.
3510 reap_rules(struct ip_fw *head)
3514 while ((rule = head) != NULL) {
3516 if (DUMMYNET_LOADED)
3517 ip_dn_ruledel_ptr(rule);
3523 * Remove all rules from a chain (except rules in set RESVD_SET
3524 * unless kill_default = 1). The caller is responsible for
3525 * reclaiming storage for the rules left in chain->reap.
3528 free_chain(struct ip_fw_chain *chain, int kill_default)
3530 struct ip_fw *prev, *rule;
3532 IPFW_WLOCK_ASSERT(chain);
3534 flush_rule_ptrs(chain); /* more efficient to do outside the loop */
3535 for (prev = NULL, rule = chain->rules; rule ; )
3536 if (kill_default || rule->set != RESVD_SET)
3537 rule = remove_rule(chain, rule, prev);
3545 * Remove all rules with given number, and also do set manipulation.
3546 * Assumes chain != NULL && *chain != NULL.
3548 * The argument is an u_int32_t. The low 16 bit are the rule or set number,
3549 * the next 8 bits are the new set, the top 8 bits are the command:
3551 * 0 delete rules with given number
3552 * 1 delete rules with given set number
3553 * 2 move rules with given number to new set
3554 * 3 move rules with given set number to new set
3555 * 4 swap sets with given numbers
3556 * 5 delete rules with given number and with given set number
3559 del_entry(struct ip_fw_chain *chain, u_int32_t arg)
3561 struct ip_fw *prev = NULL, *rule;
3562 u_int16_t rulenum; /* rule or old_set */
3563 u_int8_t cmd, new_set;
3565 rulenum = arg & 0xffff;
3566 cmd = (arg >> 24) & 0xff;
3567 new_set = (arg >> 16) & 0xff;
3569 if (cmd > 5 || new_set > RESVD_SET)
3571 if (cmd == 0 || cmd == 2 || cmd == 5) {
3572 if (rulenum >= IPFW_DEFAULT_RULE)
3575 if (rulenum > RESVD_SET) /* old_set */
3580 rule = chain->rules;
3583 case 0: /* delete rules with given number */
3585 * locate first rule to delete
3587 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next)
3589 if (rule->rulenum != rulenum) {
3590 IPFW_WUNLOCK(chain);
3595 * flush pointers outside the loop, then delete all matching
3596 * rules. prev remains the same throughout the cycle.
3598 flush_rule_ptrs(chain);
3599 while (rule->rulenum == rulenum)
3600 rule = remove_rule(chain, rule, prev);
3603 case 1: /* delete all rules with given set number */
3604 flush_rule_ptrs(chain);
3605 rule = chain->rules;
3606 while (rule->rulenum < IPFW_DEFAULT_RULE)
3607 if (rule->set == rulenum)
3608 rule = remove_rule(chain, rule, prev);
3615 case 2: /* move rules with given number to new set */
3616 rule = chain->rules;
3617 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
3618 if (rule->rulenum == rulenum)
3619 rule->set = new_set;
3622 case 3: /* move rules with given set number to new set */
3623 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
3624 if (rule->set == rulenum)
3625 rule->set = new_set;
3628 case 4: /* swap two sets */
3629 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
3630 if (rule->set == rulenum)
3631 rule->set = new_set;
3632 else if (rule->set == new_set)
3633 rule->set = rulenum;
3635 case 5: /* delete rules with given number and with given set number.
3636 * rulenum - given rule number;
3637 * new_set - given set number.
3639 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next)
3641 if (rule->rulenum != rulenum) {
3642 IPFW_WUNLOCK(chain);
3645 flush_rule_ptrs(chain);
3646 while (rule->rulenum == rulenum) {
3647 if (rule->set == new_set)
3648 rule = remove_rule(chain, rule, prev);
3656 * Look for rules to reclaim. We grab the list before
3657 * releasing the lock then reclaim them w/o the lock to
3658 * avoid a LOR with dummynet.
3662 IPFW_WUNLOCK(chain);
3669 * Clear counters for a specific rule.
3670 * The enclosing "table" is assumed locked.
3673 clear_counters(struct ip_fw *rule, int log_only)
3675 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule);
3677 if (log_only == 0) {
3678 rule->bcnt = rule->pcnt = 0;
3679 rule->timestamp = 0;
3681 if (l->o.opcode == O_LOG)
3682 l->log_left = l->max_log;
3686 * Reset some or all counters on firewall rules.
3687 * The argument `arg' is an u_int32_t. The low 16 bit are the rule number,
3688 * the next 8 bits are the set number, the top 8 bits are the command:
3689 * 0 work with rules from all set's;
3690 * 1 work with rules only from specified set.
3691 * Specified rule number is zero if we want to clear all entries.
3692 * log_only is 1 if we only want to reset logs, zero otherwise.
3695 zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only)
3700 uint16_t rulenum = arg & 0xffff;
3701 uint8_t set = (arg >> 16) & 0xff;
3702 uint8_t cmd = (arg >> 24) & 0xff;
3706 if (cmd == 1 && set > RESVD_SET)
3712 for (rule = chain->rules; rule; rule = rule->next) {
3713 /* Skip rules from another set. */
3714 if (cmd == 1 && rule->set != set)
3716 clear_counters(rule, log_only);
3718 msg = log_only ? "All logging counts reset" :
3719 "Accounting cleared";
3723 * We can have multiple rules with the same number, so we
3724 * need to clear them all.
3726 for (rule = chain->rules; rule; rule = rule->next)
3727 if (rule->rulenum == rulenum) {
3728 while (rule && rule->rulenum == rulenum) {
3729 if (cmd == 0 || rule->set == set)
3730 clear_counters(rule, log_only);
3736 if (!cleared) { /* we did not find any matching rules */
3737 IPFW_WUNLOCK(chain);
3740 msg = log_only ? "logging count reset" : "cleared";
3742 IPFW_WUNLOCK(chain);
3745 int lev = LOG_SECURITY | LOG_NOTICE;
3748 log(lev, "ipfw: Entry %d %s.\n", rulenum, msg);
3750 log(lev, "ipfw: %s.\n", msg);
3756 * Check validity of the structure before insert.
3757 * Fortunately rules are simple, so this mostly need to check rule sizes.
3760 check_ipfw_struct(struct ip_fw *rule, int size)
3766 if (size < sizeof(*rule)) {
3767 printf("ipfw: rule too short\n");
3770 /* first, check for valid size */
3773 printf("ipfw: size mismatch (have %d want %d)\n", size, l);
3776 if (rule->act_ofs >= rule->cmd_len) {
3777 printf("ipfw: bogus action offset (%u > %u)\n",
3778 rule->act_ofs, rule->cmd_len - 1);
3782 * Now go for the individual checks. Very simple ones, basically only
3783 * instruction sizes.
3785 for (l = rule->cmd_len, cmd = rule->cmd ;
3786 l > 0 ; l -= cmdlen, cmd += cmdlen) {
3787 cmdlen = F_LEN(cmd);
3789 printf("ipfw: opcode %d size truncated\n",
3793 DEB(printf("ipfw: opcode %d\n", cmd->opcode);)
3794 switch (cmd->opcode) {
3806 case O_IPPRECEDENCE:
3824 if (cmdlen != F_INSN_SIZE(ipfw_insn))
3829 if (cmdlen != F_INSN_SIZE(ipfw_insn))
3831 if (cmd->arg1 >= rt_numfibs) {
3832 printf("ipfw: invalid fib number %d\n",
3839 if (cmdlen != F_INSN_SIZE(ipfw_insn))
3841 if (cmd->arg1 >= rt_numfibs) {
3842 printf("ipfw: invalid fib number %d\n",
3857 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32))
3862 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit))
3867 if (cmdlen != F_INSN_SIZE(ipfw_insn_log))
3870 ((ipfw_insn_log *)cmd)->log_left =
3871 ((ipfw_insn_log *)cmd)->max_log;
3877 /* only odd command lengths */
3878 if ( !(cmdlen & 1) || cmdlen > 31)
3884 if (cmd->arg1 == 0 || cmd->arg1 > 256) {
3885 printf("ipfw: invalid set size %d\n",
3889 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
3894 case O_IP_SRC_LOOKUP:
3895 case O_IP_DST_LOOKUP:
3896 if (cmd->arg1 >= IPFW_TABLES_MAX) {
3897 printf("ipfw: invalid table number %d\n",
3901 if (cmdlen != F_INSN_SIZE(ipfw_insn) &&
3902 cmdlen != F_INSN_SIZE(ipfw_insn_u32))
3907 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac))
3917 if (cmdlen < 1 || cmdlen > 31)
3923 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */
3924 if (cmdlen < 2 || cmdlen > 31)
3931 if (cmdlen != F_INSN_SIZE(ipfw_insn_if))
3936 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq))
3942 if (cmdlen != F_INSN_SIZE(ipfw_insn))
3947 #ifdef IPFIREWALL_FORWARD
3948 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa))
3957 if (ip_divert_ptr == NULL)
3963 if (!NG_IPFW_LOADED)
3968 if (!IPFW_NAT_LOADED)
3970 if (cmdlen != F_INSN_SIZE(ipfw_insn_nat))
3973 case O_FORWARD_MAC: /* XXX not implemented yet */
3984 if (cmdlen != F_INSN_SIZE(ipfw_insn))
3988 printf("ipfw: opcode %d, multiple actions"
3995 printf("ipfw: opcode %d, action must be"
4004 if (cmdlen != F_INSN_SIZE(struct in6_addr) +
4005 F_INSN_SIZE(ipfw_insn))
4010 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
4011 ((ipfw_insn_u32 *)cmd)->o.arg1)
4015 case O_IP6_SRC_MASK:
4016 case O_IP6_DST_MASK:
4017 if ( !(cmdlen & 1) || cmdlen > 127)
4021 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) )
4027 switch (cmd->opcode) {
4037 case O_IP6_SRC_MASK:
4038 case O_IP6_DST_MASK:
4040 printf("ipfw: no IPv6 support in kernel\n");
4041 return EPROTONOSUPPORT;
4044 printf("ipfw: opcode %d, unknown opcode\n",
4050 if (have_action == 0) {
4051 printf("ipfw: missing action\n");
4057 printf("ipfw: opcode %d size %d wrong\n",
4058 cmd->opcode, cmdlen);
4063 * Copy the static and dynamic rules to the supplied buffer
4064 * and return the amount of space actually used.
4067 ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space)
4070 char *ep = bp + space;
4073 time_t boot_seconds;
4075 boot_seconds = boottime.tv_sec;
4076 /* XXX this can take a long time and locking will block packet flow */
4078 for (rule = chain->rules; rule ; rule = rule->next) {
4080 * Verify the entry fits in the buffer in case the
4081 * rules changed between calculating buffer space and
4082 * now. This would be better done using a generation
4083 * number but should suffice for now.
4089 * XXX HACK. Store the disable mask in the "next" pointer
4090 * in a wild attempt to keep the ABI the same.
4091 * Why do we do this on EVERY rule?
4093 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule),
4094 sizeof(set_disable));
4095 if (((struct ip_fw *)bp)->timestamp)
4096 ((struct ip_fw *)bp)->timestamp += boot_seconds;
4100 IPFW_RUNLOCK(chain);
4102 ipfw_dyn_rule *p, *last = NULL;
4105 for (i = 0 ; i < curr_dyn_buckets; i++)
4106 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) {
4107 if (bp + sizeof *p <= ep) {
4108 ipfw_dyn_rule *dst =
4109 (ipfw_dyn_rule *)bp;
4110 bcopy(p, dst, sizeof *p);
4111 bcopy(&(p->rule->rulenum), &(dst->rule),
4112 sizeof(p->rule->rulenum));
4114 * store set number into high word of
4115 * dst->rule pointer.
4117 bcopy(&(p->rule->set),
4118 (char *)&dst->rule +
4119 sizeof(p->rule->rulenum),
4120 sizeof(p->rule->set));
4122 * store a non-null value in "next".
4123 * The userland code will interpret a
4124 * NULL here as a marker
4125 * for the last dynamic rule.
4127 bcopy(&dst, &dst->next, sizeof(dst));
4130 TIME_LEQ(dst->expire, time_uptime) ?
4131 0 : dst->expire - time_uptime ;
4132 bp += sizeof(ipfw_dyn_rule);
4136 if (last != NULL) /* mark last dynamic rule */
4137 bzero(&last->next, sizeof(last));
4139 return (bp - (char *)buf);
4144 * {set|get}sockopt parser.
4147 ipfw_ctl(struct sockopt *sopt)
4149 #define RULE_MAXSIZE (256*sizeof(u_int32_t))
4152 struct ip_fw *buf, *rule;
4153 u_int32_t rulenum[2];
4155 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW);
4160 * Disallow modifications in really-really secure mode, but still allow
4161 * the logging counters to be reset.
4163 if (sopt->sopt_name == IP_FW_ADD ||
4164 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) {
4165 error = securelevel_ge(sopt->sopt_td->td_ucred, 3);
4172 switch (sopt->sopt_name) {
4175 * pass up a copy of the current rules. Static rules
4176 * come first (the last of which has number IPFW_DEFAULT_RULE),
4177 * followed by a possibly empty list of dynamic rule.
4178 * The last dynamic rule has NULL in the "next" field.
4180 * Note that the calculated size is used to bound the
4181 * amount of data returned to the user. The rule set may
4182 * change between calculating the size and returning the
4183 * data in which case we'll just return what fits.
4185 size = static_len; /* size of static rules */
4186 if (ipfw_dyn_v) /* add size of dyn.rules */
4187 size += (dyn_count * sizeof(ipfw_dyn_rule));
4190 * XXX todo: if the user passes a short length just to know
4191 * how much room is needed, do not bother filling up the
4192 * buffer, just jump to the sooptcopyout.
4194 buf = malloc(size, M_TEMP, M_WAITOK);
4195 error = sooptcopyout(sopt, buf,
4196 ipfw_getrules(&layer3_chain, buf, size));
4202 * Normally we cannot release the lock on each iteration.
4203 * We could do it here only because we start from the head all
4204 * the times so there is no risk of missing some entries.
4205 * On the other hand, the risk is that we end up with
4206 * a very inconsistent ruleset, so better keep the lock
4207 * around the whole cycle.
4209 * XXX this code can be improved by resetting the head of
4210 * the list to point to the default rule, and then freeing
4211 * the old list without the need for a lock.
4214 IPFW_WLOCK(&layer3_chain);
4215 layer3_chain.reap = NULL;
4216 free_chain(&layer3_chain, 0 /* keep default rule */);
4217 rule = layer3_chain.reap;
4218 layer3_chain.reap = NULL;
4219 IPFW_WUNLOCK(&layer3_chain);
4225 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK);
4226 error = sooptcopyin(sopt, rule, RULE_MAXSIZE,
4227 sizeof(struct ip_fw) );
4229 error = check_ipfw_struct(rule, sopt->sopt_valsize);
4231 error = add_rule(&layer3_chain, rule);
4232 size = RULESIZE(rule);
4233 if (!error && sopt->sopt_dir == SOPT_GET)
4234 error = sooptcopyout(sopt, rule, size);
4241 * IP_FW_DEL is used for deleting single rules or sets,
4242 * and (ab)used to atomically manipulate sets. Argument size
4243 * is used to distinguish between the two:
4245 * delete single rule or set of rules,
4246 * or reassign rules (or sets) to a different set.
4247 * 2*sizeof(u_int32_t)
4248 * atomic disable/enable sets.
4249 * first u_int32_t contains sets to be disabled,
4250 * second u_int32_t contains sets to be enabled.
4252 error = sooptcopyin(sopt, rulenum,
4253 2*sizeof(u_int32_t), sizeof(u_int32_t));
4256 size = sopt->sopt_valsize;
4257 if (size == sizeof(u_int32_t)) /* delete or reassign */
4258 error = del_entry(&layer3_chain, rulenum[0]);
4259 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */
4261 (set_disable | rulenum[0]) & ~rulenum[1] &
4262 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */
4268 case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */
4270 if (sopt->sopt_val != 0) {
4271 error = sooptcopyin(sopt, rulenum,
4272 sizeof(u_int32_t), sizeof(u_int32_t));
4276 error = zero_entry(&layer3_chain, rulenum[0],
4277 sopt->sopt_name == IP_FW_RESETLOG);
4280 case IP_FW_TABLE_ADD:
4282 ipfw_table_entry ent;
4284 error = sooptcopyin(sopt, &ent,
4285 sizeof(ent), sizeof(ent));
4288 error = add_table_entry(&layer3_chain, ent.tbl,
4289 ent.addr, ent.masklen, ent.value);
4293 case IP_FW_TABLE_DEL:
4295 ipfw_table_entry ent;
4297 error = sooptcopyin(sopt, &ent,
4298 sizeof(ent), sizeof(ent));
4301 error = del_table_entry(&layer3_chain, ent.tbl,
4302 ent.addr, ent.masklen);
4306 case IP_FW_TABLE_FLUSH:
4310 error = sooptcopyin(sopt, &tbl,
4311 sizeof(tbl), sizeof(tbl));
4314 IPFW_WLOCK(&layer3_chain);
4315 error = flush_table(&layer3_chain, tbl);
4316 IPFW_WUNLOCK(&layer3_chain);
4320 case IP_FW_TABLE_GETSIZE:
4324 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl),
4327 IPFW_RLOCK(&layer3_chain);
4328 error = count_table(&layer3_chain, tbl, &cnt);
4329 IPFW_RUNLOCK(&layer3_chain);
4332 error = sooptcopyout(sopt, &cnt, sizeof(cnt));
4336 case IP_FW_TABLE_LIST:
4340 if (sopt->sopt_valsize < sizeof(*tbl)) {
4344 size = sopt->sopt_valsize;
4345 tbl = malloc(size, M_TEMP, M_WAITOK);
4346 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl));
4351 tbl->size = (size - sizeof(*tbl)) /
4352 sizeof(ipfw_table_entry);
4353 IPFW_RLOCK(&layer3_chain);
4354 error = dump_table(&layer3_chain, tbl);
4355 IPFW_RUNLOCK(&layer3_chain);
4360 error = sooptcopyout(sopt, tbl, size);
4366 if (IPFW_NAT_LOADED)
4367 error = ipfw_nat_cfg_ptr(sopt);
4369 printf("IP_FW_NAT_CFG: %s\n",
4370 "ipfw_nat not present, please load it");
4376 if (IPFW_NAT_LOADED)
4377 error = ipfw_nat_del_ptr(sopt);
4379 printf("IP_FW_NAT_DEL: %s\n",
4380 "ipfw_nat not present, please load it");
4385 case IP_FW_NAT_GET_CONFIG:
4386 if (IPFW_NAT_LOADED)
4387 error = ipfw_nat_get_cfg_ptr(sopt);
4389 printf("IP_FW_NAT_GET_CFG: %s\n",
4390 "ipfw_nat not present, please load it");
4395 case IP_FW_NAT_GET_LOG:
4396 if (IPFW_NAT_LOADED)
4397 error = ipfw_nat_get_log_ptr(sopt);
4399 printf("IP_FW_NAT_GET_LOG: %s\n",
4400 "ipfw_nat not present, please load it");
4406 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name);
4415 * dummynet needs a reference to the default rule, because rules can be
4416 * deleted while packets hold a reference to them. When this happens,
4417 * dummynet changes the reference to the default rule (it could well be a
4418 * NULL pointer, but this way we do not need to check for the special
4419 * case, plus here he have info on the default behaviour).
4421 struct ip_fw *ip_fw_default_rule;
4424 * This procedure is only used to handle keepalives. It is invoked
4425 * every dyn_keepalive_period
4428 ipfw_tick(void * __unused unused)
4430 struct mbuf *m0, *m, *mnext, **mtailp;
4434 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0)
4438 * We make a chain of packets to go out here -- not deferring
4439 * until after we drop the IPFW dynamic rule lock would result
4440 * in a lock order reversal with the normal packet input -> ipfw
4446 for (i = 0 ; i < curr_dyn_buckets ; i++) {
4447 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) {
4448 if (q->dyn_type == O_LIMIT_PARENT)
4450 if (q->id.proto != IPPROTO_TCP)
4452 if ( (q->state & BOTH_SYN) != BOTH_SYN)
4454 if (TIME_LEQ( time_uptime+dyn_keepalive_interval,
4456 continue; /* too early */
4457 if (TIME_LEQ(q->expire, time_uptime))
4458 continue; /* too late, rule expired */
4460 *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1,
4461 q->ack_fwd, TH_SYN);
4462 if (*mtailp != NULL)
4463 mtailp = &(*mtailp)->m_nextpkt;
4464 *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1,
4466 if (*mtailp != NULL)
4467 mtailp = &(*mtailp)->m_nextpkt;
4471 for (m = mnext = m0; m != NULL; m = mnext) {
4472 mnext = m->m_nextpkt;
4473 m->m_nextpkt = NULL;
4474 ip_output(m, NULL, NULL, 0, NULL, NULL);
4477 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL);
4483 struct ip_fw default_rule;
4487 /* Setup IPv6 fw sysctl tree. */
4488 sysctl_ctx_init(&ip6_fw_sysctl_ctx);
4489 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx,
4490 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw",
4491 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall");
4492 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree),
4493 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3,
4494 &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6");
4495 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree),
4496 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE,
4497 &fw_deny_unknown_exthdrs, 0,
4498 "Deny packets with unknown IPv6 Extension Headers");
4501 layer3_chain.rules = NULL;
4502 IPFW_LOCK_INIT(&layer3_chain);
4503 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule",
4504 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL,
4506 IPFW_DYN_LOCK_INIT();
4507 callout_init(&ipfw_timeout, CALLOUT_MPSAFE);
4509 bzero(&default_rule, sizeof default_rule);
4511 default_rule.act_ofs = 0;
4512 default_rule.rulenum = IPFW_DEFAULT_RULE;
4513 default_rule.cmd_len = 1;
4514 default_rule.set = RESVD_SET;
4516 default_rule.cmd[0].len = 1;
4517 default_rule.cmd[0].opcode =
4518 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
4523 error = add_rule(&layer3_chain, &default_rule);
4525 printf("ipfw2: error %u initializing default rule "
4526 "(support disabled)\n", error);
4527 IPFW_DYN_LOCK_DESTROY();
4528 IPFW_LOCK_DESTROY(&layer3_chain);
4529 uma_zdestroy(ipfw_dyn_rule_zone);
4533 ip_fw_default_rule = layer3_chain.rules;
4538 "initialized, divert %s, nat %s, "
4539 "rule-based forwarding "
4540 #ifdef IPFIREWALL_FORWARD
4545 "default to %s, logging ",
4551 #ifdef IPFIREWALL_NAT
4557 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny");
4559 #ifdef IPFIREWALL_VERBOSE
4562 #ifdef IPFIREWALL_VERBOSE_LIMIT
4563 verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
4565 if (fw_verbose == 0)
4566 printf("disabled\n");
4567 else if (verbose_limit == 0)
4568 printf("unlimited\n");
4570 printf("limited to %d packets/entry by default\n",
4573 error = init_tables(&layer3_chain);
4575 IPFW_DYN_LOCK_DESTROY();
4576 IPFW_LOCK_DESTROY(&layer3_chain);
4577 uma_zdestroy(ipfw_dyn_rule_zone);
4580 ip_fw_ctl_ptr = ipfw_ctl;
4581 ip_fw_chk_ptr = ipfw_chk;
4582 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL);
4583 LIST_INIT(&layer3_chain.nat);
4592 ip_fw_chk_ptr = NULL;
4593 ip_fw_ctl_ptr = NULL;
4594 callout_drain(&ipfw_timeout);
4595 IPFW_WLOCK(&layer3_chain);
4596 flush_tables(&layer3_chain);
4597 layer3_chain.reap = NULL;
4598 free_chain(&layer3_chain, 1 /* kill default rule */);
4599 reap = layer3_chain.reap, layer3_chain.reap = NULL;
4600 IPFW_WUNLOCK(&layer3_chain);
4603 IPFW_DYN_LOCK_DESTROY();
4604 uma_zdestroy(ipfw_dyn_rule_zone);
4605 if (ipfw_dyn_v != NULL)
4606 free(ipfw_dyn_v, M_IPFW);
4607 IPFW_LOCK_DESTROY(&layer3_chain);
4610 /* Free IPv6 fw sysctl tree. */
4611 sysctl_ctx_free(&ip6_fw_sysctl_ctx);
4614 printf("IP firewall unloaded\n");