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
32 * Implement IP packet firewall (new version)
35 #if !defined(KLD_MODULE)
37 #include "opt_ipdivert.h"
41 #error IPFIREWALL requires INET.
44 #include "opt_inet6.h"
45 #include "opt_ipsec.h"
48 #include <sys/param.h>
49 #include <sys/systm.h>
50 #include <sys/condvar.h>
51 #include <sys/eventhandler.h>
52 #include <sys/malloc.h>
54 #include <sys/kernel.h>
57 #include <sys/module.h>
60 #include <sys/rwlock.h>
61 #include <sys/socket.h>
62 #include <sys/socketvar.h>
63 #include <sys/sysctl.h>
64 #include <sys/syslog.h>
65 #include <sys/ucred.h>
67 #include <net/radix.h>
68 #include <net/route.h>
69 #include <netinet/in.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/in_var.h>
72 #include <netinet/in_pcb.h>
73 #include <netinet/ip.h>
74 #include <netinet/ip_var.h>
75 #include <netinet/ip_icmp.h>
76 #include <netinet/ip_fw.h>
77 #include <netinet/ip_divert.h>
78 #include <netinet/ip_dummynet.h>
79 #include <netinet/ip_carp.h>
80 #include <netinet/pim.h>
81 #include <netinet/tcp.h>
82 #include <netinet/tcp_timer.h>
83 #include <netinet/tcp_var.h>
84 #include <netinet/tcpip.h>
85 #include <netinet/udp.h>
86 #include <netinet/udp_var.h>
87 #include <netinet/sctp.h>
89 #include <netinet/libalias/alias.h>
90 #include <netinet/libalias/alias_local.h>
92 #include <netgraph/ng_ipfw.h>
94 #include <altq/if_altq.h>
96 #include <netinet/ip6.h>
97 #include <netinet/icmp6.h>
99 #include <netinet6/scope6_var.h>
102 #include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */
104 #include <machine/in_cksum.h> /* XXX for in_cksum */
106 #include <security/mac/mac_framework.h>
109 * set_disable contains one bit per set value (0..31).
110 * If the bit is set, all rules with the corresponding set
111 * are disabled. Set RESVD_SET(31) is reserved for the default rule
112 * and rules that are not deleted by the flush command,
113 * and CANNOT be disabled.
114 * Rules in set RESVD_SET can only be deleted explicitly.
116 static u_int32_t set_disable;
118 static int fw_verbose;
119 static int verbose_limit;
121 static struct callout ipfw_timeout;
122 static uma_zone_t ipfw_dyn_rule_zone;
123 #define IPFW_DEFAULT_RULE 65535
126 * Data structure to cache our ucred related
127 * information. This structure only gets used if
128 * the user specified UID/GID based constraints in
132 gid_t fw_groups[NGROUPS];
138 #define IPFW_TABLES_MAX 128
140 struct ip_fw *rules; /* list of rules */
141 struct ip_fw *reap; /* list of rules to reap */
142 LIST_HEAD(, cfg_nat) nat; /* list of nat entries */
143 struct radix_node_head *tables[IPFW_TABLES_MAX];
146 #define IPFW_LOCK_INIT(_chain) \
147 rw_init(&(_chain)->rwmtx, "IPFW static rules")
148 #define IPFW_LOCK_DESTROY(_chain) rw_destroy(&(_chain)->rwmtx)
149 #define IPFW_WLOCK_ASSERT(_chain) do { \
150 rw_assert(&(_chain)->rwmtx, RA_WLOCKED); \
151 NET_ASSERT_GIANT(); \
154 #define IPFW_RLOCK(p) rw_rlock(&(p)->rwmtx)
155 #define IPFW_RUNLOCK(p) rw_runlock(&(p)->rwmtx)
156 #define IPFW_WLOCK(p) rw_wlock(&(p)->rwmtx)
157 #define IPFW_WUNLOCK(p) rw_wunlock(&(p)->rwmtx)
160 * list of rules for layer 3
162 static struct ip_fw_chain layer3_chain;
164 MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's");
165 MALLOC_DEFINE(M_IPFW_TBL, "ipfw_tbl", "IpFw tables");
168 struct radix_node rn[2];
169 struct sockaddr_in addr, mask;
173 static int fw_debug = 1;
174 static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */
176 extern int ipfw_chg_hook(SYSCTL_HANDLER_ARGS);
179 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
180 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable,
181 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3, &fw_enable, 0,
182 ipfw_chg_hook, "I", "Enable ipfw");
183 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW,
184 &autoinc_step, 0, "Rule number autincrement step");
185 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
186 CTLFLAG_RW | CTLFLAG_SECURE3,
188 "Only do a single pass through ipfw when using dummynet(4)");
189 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW,
190 &fw_debug, 0, "Enable printing of debug ip_fw statements");
191 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose,
192 CTLFLAG_RW | CTLFLAG_SECURE3,
193 &fw_verbose, 0, "Log matches to ipfw rules");
194 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW,
195 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged");
198 * Description of dynamic rules.
200 * Dynamic rules are stored in lists accessed through a hash table
201 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can
202 * be modified through the sysctl variable dyn_buckets which is
203 * updated when the table becomes empty.
205 * XXX currently there is only one list, ipfw_dyn.
207 * When a packet is received, its address fields are first masked
208 * with the mask defined for the rule, then hashed, then matched
209 * against the entries in the corresponding list.
210 * Dynamic rules can be used for different purposes:
212 * + enforcing limits on the number of sessions;
213 * + in-kernel NAT (not implemented yet)
215 * The lifetime of dynamic rules is regulated by dyn_*_lifetime,
216 * measured in seconds and depending on the flags.
218 * The total number of dynamic rules is stored in dyn_count.
219 * The max number of dynamic rules is dyn_max. When we reach
220 * the maximum number of rules we do not create anymore. This is
221 * done to avoid consuming too much memory, but also too much
222 * time when searching on each packet (ideally, we should try instead
223 * to put a limit on the length of the list on each bucket...).
225 * Each dynamic rule holds a pointer to the parent ipfw rule so
226 * we know what action to perform. Dynamic rules are removed when
227 * the parent rule is deleted. XXX we should make them survive.
229 * There are some limitations with dynamic rules -- we do not
230 * obey the 'randomized match', and we do not do multiple
231 * passes through the firewall. XXX check the latter!!!
233 static ipfw_dyn_rule **ipfw_dyn_v = NULL;
234 static u_int32_t dyn_buckets = 256; /* must be power of 2 */
235 static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */
237 static struct mtx ipfw_dyn_mtx; /* mutex guarding dynamic rules */
238 #define IPFW_DYN_LOCK_INIT() \
239 mtx_init(&ipfw_dyn_mtx, "IPFW dynamic rules", NULL, MTX_DEF)
240 #define IPFW_DYN_LOCK_DESTROY() mtx_destroy(&ipfw_dyn_mtx)
241 #define IPFW_DYN_LOCK() mtx_lock(&ipfw_dyn_mtx)
242 #define IPFW_DYN_UNLOCK() mtx_unlock(&ipfw_dyn_mtx)
243 #define IPFW_DYN_LOCK_ASSERT() mtx_assert(&ipfw_dyn_mtx, MA_OWNED)
246 * Timeouts for various events in handing dynamic rules.
248 static u_int32_t dyn_ack_lifetime = 300;
249 static u_int32_t dyn_syn_lifetime = 20;
250 static u_int32_t dyn_fin_lifetime = 1;
251 static u_int32_t dyn_rst_lifetime = 1;
252 static u_int32_t dyn_udp_lifetime = 10;
253 static u_int32_t dyn_short_lifetime = 5;
256 * Keepalives are sent if dyn_keepalive is set. They are sent every
257 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval
258 * seconds of lifetime of a rule.
259 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower
260 * than dyn_keepalive_period.
263 static u_int32_t dyn_keepalive_interval = 20;
264 static u_int32_t dyn_keepalive_period = 5;
265 static u_int32_t dyn_keepalive = 1; /* do send keepalives */
267 static u_int32_t static_count; /* # of static rules */
268 static u_int32_t static_len; /* size in bytes of static rules */
269 static u_int32_t dyn_count; /* # of dynamic rules */
270 static u_int32_t dyn_max = 4096; /* max # of dynamic rules */
272 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW,
273 &dyn_buckets, 0, "Number of dyn. buckets");
274 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD,
275 &curr_dyn_buckets, 0, "Current Number of dyn. buckets");
276 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD,
277 &dyn_count, 0, "Number of dyn. rules");
278 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW,
279 &dyn_max, 0, "Max number of dyn. rules");
280 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD,
281 &static_count, 0, "Number of static rules");
282 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW,
283 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks");
284 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW,
285 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn");
286 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW,
287 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin");
288 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW,
289 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst");
290 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW,
291 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP");
292 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW,
293 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations");
294 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW,
295 &dyn_keepalive, 0, "Enable keepalives for dyn. rules");
299 * IPv6 specific variables
301 SYSCTL_DECL(_net_inet6_ip6);
303 static struct sysctl_ctx_list ip6_fw_sysctl_ctx;
304 static struct sysctl_oid *ip6_fw_sysctl_tree;
306 #endif /* SYSCTL_NODE */
308 #ifdef IPFIREWALL_NAT
309 MODULE_DEPEND(ipfw, libalias, 1, 1, 1);
311 static int fw_deny_unknown_exthdrs = 1;
315 * L3HDR maps an ipv4 pointer into a layer3 header pointer of type T
316 * Other macros just cast void * into the appropriate type
318 #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
319 #define TCP(p) ((struct tcphdr *)(p))
320 #define SCTP(p) ((struct sctphdr *)(p))
321 #define UDP(p) ((struct udphdr *)(p))
322 #define ICMP(p) ((struct icmphdr *)(p))
323 #define ICMP6(p) ((struct icmp6_hdr *)(p))
326 icmptype_match(struct icmphdr *icmp, ipfw_insn_u32 *cmd)
328 int type = icmp->icmp_type;
330 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
333 #define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
334 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
337 is_icmp_query(struct icmphdr *icmp)
339 int type = icmp->icmp_type;
341 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
346 * The following checks use two arrays of 8 or 16 bits to store the
347 * bits that we want set or clear, respectively. They are in the
348 * low and high half of cmd->arg1 or cmd->d[0].
350 * We scan options and store the bits we find set. We succeed if
352 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
354 * The code is sometimes optimized not to store additional variables.
358 flags_match(ipfw_insn *cmd, u_int8_t bits)
363 if ( ((cmd->arg1 & 0xff) & bits) != 0)
364 return 0; /* some bits we want set were clear */
365 want_clear = (cmd->arg1 >> 8) & 0xff;
366 if ( (want_clear & bits) != want_clear)
367 return 0; /* some bits we want clear were set */
372 ipopts_match(struct ip *ip, ipfw_insn *cmd)
374 int optlen, bits = 0;
375 u_char *cp = (u_char *)(ip + 1);
376 int x = (ip->ip_hl << 2) - sizeof (struct ip);
378 for (; x > 0; x -= optlen, cp += optlen) {
379 int opt = cp[IPOPT_OPTVAL];
381 if (opt == IPOPT_EOL)
383 if (opt == IPOPT_NOP)
386 optlen = cp[IPOPT_OLEN];
387 if (optlen <= 0 || optlen > x)
388 return 0; /* invalid or truncated */
396 bits |= IP_FW_IPOPT_LSRR;
400 bits |= IP_FW_IPOPT_SSRR;
404 bits |= IP_FW_IPOPT_RR;
408 bits |= IP_FW_IPOPT_TS;
412 return (flags_match(cmd, bits));
416 tcpopts_match(struct tcphdr *tcp, ipfw_insn *cmd)
418 int optlen, bits = 0;
419 u_char *cp = (u_char *)(tcp + 1);
420 int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
422 for (; x > 0; x -= optlen, cp += optlen) {
424 if (opt == TCPOPT_EOL)
426 if (opt == TCPOPT_NOP)
440 bits |= IP_FW_TCPOPT_MSS;
444 bits |= IP_FW_TCPOPT_WINDOW;
447 case TCPOPT_SACK_PERMITTED:
449 bits |= IP_FW_TCPOPT_SACK;
452 case TCPOPT_TIMESTAMP:
453 bits |= IP_FW_TCPOPT_TS;
458 return (flags_match(cmd, bits));
462 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd)
464 if (ifp == NULL) /* no iface with this packet, match fails */
466 /* Check by name or by IP address */
467 if (cmd->name[0] != '\0') { /* match by name */
470 if (fnmatch(cmd->name, ifp->if_xname, 0) == 0)
473 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
480 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
481 if (ia->ifa_addr->sa_family != AF_INET)
483 if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
484 (ia->ifa_addr))->sin_addr.s_addr)
485 return(1); /* match */
488 return(0); /* no match, fail ... */
492 * The verify_path function checks if a route to the src exists and
493 * if it is reachable via ifp (when provided).
495 * The 'verrevpath' option checks that the interface that an IP packet
496 * arrives on is the same interface that traffic destined for the
497 * packet's source address would be routed out of. The 'versrcreach'
498 * option just checks that the source address is reachable via any route
499 * (except default) in the routing table. These two are a measure to block
500 * forged packets. This is also commonly known as "anti-spoofing" or Unicast
501 * Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The name of the knobs
502 * is purposely reminiscent of the Cisco IOS command,
504 * ip verify unicast reverse-path
505 * ip verify unicast source reachable-via any
507 * which implements the same functionality. But note that syntax is
508 * misleading. The check may be performed on all IP packets whether unicast,
509 * multicast, or broadcast.
512 verify_path(struct in_addr src, struct ifnet *ifp)
515 struct sockaddr_in *dst;
517 bzero(&ro, sizeof(ro));
519 dst = (struct sockaddr_in *)&(ro.ro_dst);
520 dst->sin_family = AF_INET;
521 dst->sin_len = sizeof(*dst);
523 rtalloc_ign(&ro, RTF_CLONING);
525 if (ro.ro_rt == NULL)
529 * If ifp is provided, check for equality with rtentry.
530 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
531 * in order to pass packets injected back by if_simloop():
532 * if useloopback == 1 routing entry (via lo0) for our own address
533 * may exist, so we need to handle routing assymetry.
535 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
540 /* if no ifp provided, check if rtentry is not default route */
542 satosin(rt_key(ro.ro_rt))->sin_addr.s_addr == INADDR_ANY) {
547 /* or if this is a blackhole/reject route */
548 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
553 /* found valid route */
560 * ipv6 specific rules here...
563 icmp6type_match (int type, ipfw_insn_u32 *cmd)
565 return (type <= ICMP6_MAXTYPE && (cmd->d[type/32] & (1<<(type%32)) ) );
569 flow6id_match( int curr_flow, ipfw_insn_u32 *cmd )
572 for (i=0; i <= cmd->o.arg1; ++i )
573 if (curr_flow == cmd->d[i] )
578 /* support for IP6_*_ME opcodes */
580 search_ip6_addr_net (struct in6_addr * ip6_addr)
584 struct in6_ifaddr *fdm;
585 struct in6_addr copia;
587 TAILQ_FOREACH(mdc, &ifnet, if_link)
588 TAILQ_FOREACH(mdc2, &mdc->if_addrlist, ifa_list) {
589 if (mdc2->ifa_addr->sa_family == AF_INET6) {
590 fdm = (struct in6_ifaddr *)mdc2;
591 copia = fdm->ia_addr.sin6_addr;
592 /* need for leaving scope_id in the sock_addr */
593 in6_clearscope(&copia);
594 if (IN6_ARE_ADDR_EQUAL(ip6_addr, &copia))
602 verify_path6(struct in6_addr *src, struct ifnet *ifp)
605 struct sockaddr_in6 *dst;
607 bzero(&ro, sizeof(ro));
609 dst = (struct sockaddr_in6 * )&(ro.ro_dst);
610 dst->sin6_family = AF_INET6;
611 dst->sin6_len = sizeof(*dst);
612 dst->sin6_addr = *src;
613 rtalloc_ign((struct route *)&ro, RTF_CLONING);
615 if (ro.ro_rt == NULL)
619 * if ifp is provided, check for equality with rtentry
620 * We should use rt->rt_ifa->ifa_ifp, instead of rt->rt_ifp,
621 * to support the case of sending packets to an address of our own.
622 * (where the former interface is the first argument of if_simloop()
623 * (=ifp), the latter is lo0)
625 if (ifp != NULL && ro.ro_rt->rt_ifa->ifa_ifp != ifp) {
630 /* if no ifp provided, check if rtentry is not default route */
632 IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(ro.ro_rt))->sin6_addr)) {
637 /* or if this is a blackhole/reject route */
638 if (ifp == NULL && ro.ro_rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
643 /* found valid route */
649 hash_packet6(struct ipfw_flow_id *id)
652 i = (id->dst_ip6.__u6_addr.__u6_addr32[2]) ^
653 (id->dst_ip6.__u6_addr.__u6_addr32[3]) ^
654 (id->src_ip6.__u6_addr.__u6_addr32[2]) ^
655 (id->src_ip6.__u6_addr.__u6_addr32[3]) ^
656 (id->dst_port) ^ (id->src_port);
661 is_icmp6_query(int icmp6_type)
663 if ((icmp6_type <= ICMP6_MAXTYPE) &&
664 (icmp6_type == ICMP6_ECHO_REQUEST ||
665 icmp6_type == ICMP6_MEMBERSHIP_QUERY ||
666 icmp6_type == ICMP6_WRUREQUEST ||
667 icmp6_type == ICMP6_FQDN_QUERY ||
668 icmp6_type == ICMP6_NI_QUERY))
675 send_reject6(struct ip_fw_args *args, int code, u_int hlen, struct ip6_hdr *ip6)
680 if (code == ICMP6_UNREACH_RST && args->f_id.proto == IPPROTO_TCP) {
688 tcp = (struct tcphdr *)((char *)ip6 + hlen);
690 if ((tcp->th_flags & TH_RST) != 0) {
698 ti.th.th_seq = ntohl(ti.th.th_seq);
699 ti.th.th_ack = ntohl(ti.th.th_ack);
700 ti.ip6.ip6_nxt = IPPROTO_TCP;
702 if (ti.th.th_flags & TH_ACK) {
708 if ((m->m_flags & M_PKTHDR) != 0) {
710 * total new data to ACK is:
711 * total packet length,
712 * minus the header length,
713 * minus the tcp header length.
715 ack += m->m_pkthdr.len - hlen
716 - (ti.th.th_off << 2);
717 } else if (ip6->ip6_plen) {
718 ack += ntohs(ip6->ip6_plen) + sizeof(*ip6) -
719 hlen - (ti.th.th_off << 2);
724 if (tcp->th_flags & TH_SYN)
727 flags = TH_RST|TH_ACK;
729 bcopy(&ti, ip6, sizeof(ti));
731 * m is only used to recycle the mbuf
732 * The data in it is never read so we don't need
733 * to correct the offsets or anything
735 tcp_respond(NULL, ip6, tcp, m, ack, seq, flags);
736 } else if (code != ICMP6_UNREACH_RST) { /* Send an ICMPv6 unreach. */
739 * Unlike above, the mbufs need to line up with the ip6 hdr,
740 * as the contents are read. We need to m_adj() the
742 * The mbuf will however be thrown away so we can adjust it.
743 * Remember we did an m_pullup on it already so we
744 * can make some assumptions about contiguousness.
747 m_adj(m, args->L3offset);
749 icmp6_error(m, ICMP6_DST_UNREACH, code, 0);
758 static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */
760 #define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0
761 #define SNP(buf) buf, sizeof(buf)
764 * We enter here when we have a rule with O_LOG.
765 * XXX this function alone takes about 2Kbytes of code!
768 ipfw_log(struct ip_fw *f, u_int hlen, struct ip_fw_args *args,
769 struct mbuf *m, struct ifnet *oif, u_short offset, uint32_t tablearg,
772 struct ether_header *eh = args->eh;
774 int limit_reached = 0;
775 char action2[40], proto[128], fragment[32];
780 if (f == NULL) { /* bogus pkt */
781 if (verbose_limit != 0 && norule_counter >= verbose_limit)
784 if (norule_counter == verbose_limit)
785 limit_reached = verbose_limit;
787 } else { /* O_LOG is the first action, find the real one */
788 ipfw_insn *cmd = ACTION_PTR(f);
789 ipfw_insn_log *l = (ipfw_insn_log *)cmd;
791 if (l->max_log != 0 && l->log_left == 0)
794 if (l->log_left == 0)
795 limit_reached = l->max_log;
796 cmd += F_LEN(cmd); /* point to first action */
797 if (cmd->opcode == O_ALTQ) {
798 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
800 snprintf(SNPARGS(action2, 0), "Altq %d",
804 if (cmd->opcode == O_PROB)
807 if (cmd->opcode == O_TAG)
811 switch (cmd->opcode) {
817 if (cmd->arg1==ICMP_REJECT_RST)
819 else if (cmd->arg1==ICMP_UNREACH_HOST)
822 snprintf(SNPARGS(action2, 0), "Unreach %d",
827 if (cmd->arg1==ICMP6_UNREACH_RST)
830 snprintf(SNPARGS(action2, 0), "Unreach %d",
841 snprintf(SNPARGS(action2, 0), "Divert %d",
845 snprintf(SNPARGS(action2, 0), "Tee %d",
849 snprintf(SNPARGS(action2, 0), "SkipTo %d",
853 snprintf(SNPARGS(action2, 0), "Pipe %d",
857 snprintf(SNPARGS(action2, 0), "Queue %d",
861 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd;
863 struct in_addr dummyaddr;
864 if (sa->sa.sin_addr.s_addr == INADDR_ANY)
865 dummyaddr.s_addr = htonl(tablearg);
867 dummyaddr.s_addr = sa->sa.sin_addr.s_addr;
869 len = snprintf(SNPARGS(action2, 0), "Forward to %s",
870 inet_ntoa(dummyaddr));
873 snprintf(SNPARGS(action2, len), ":%d",
878 snprintf(SNPARGS(action2, 0), "Netgraph %d",
882 snprintf(SNPARGS(action2, 0), "Ngtee %d",
894 if (hlen == 0) { /* non-ip */
895 snprintf(SNPARGS(proto, 0), "MAC");
899 char src[48], dst[48];
900 struct icmphdr *icmp;
904 struct ip6_hdr *ip6 = NULL;
905 struct icmp6_hdr *icmp6;
910 if (IS_IP6_FLOW_ID(&(args->f_id))) {
911 char ip6buf[INET6_ADDRSTRLEN];
912 snprintf(src, sizeof(src), "[%s]",
913 ip6_sprintf(ip6buf, &args->f_id.src_ip6));
914 snprintf(dst, sizeof(dst), "[%s]",
915 ip6_sprintf(ip6buf, &args->f_id.dst_ip6));
917 ip6 = (struct ip6_hdr *)ip;
918 tcp = (struct tcphdr *)(((char *)ip) + hlen);
919 udp = (struct udphdr *)(((char *)ip) + hlen);
923 tcp = L3HDR(struct tcphdr, ip);
924 udp = L3HDR(struct udphdr, ip);
926 inet_ntoa_r(ip->ip_src, src);
927 inet_ntoa_r(ip->ip_dst, dst);
930 switch (args->f_id.proto) {
932 len = snprintf(SNPARGS(proto, 0), "TCP %s", src);
934 snprintf(SNPARGS(proto, len), ":%d %s:%d",
935 ntohs(tcp->th_sport),
937 ntohs(tcp->th_dport));
939 snprintf(SNPARGS(proto, len), " %s", dst);
943 len = snprintf(SNPARGS(proto, 0), "UDP %s", src);
945 snprintf(SNPARGS(proto, len), ":%d %s:%d",
946 ntohs(udp->uh_sport),
948 ntohs(udp->uh_dport));
950 snprintf(SNPARGS(proto, len), " %s", dst);
954 icmp = L3HDR(struct icmphdr, ip);
956 len = snprintf(SNPARGS(proto, 0),
958 icmp->icmp_type, icmp->icmp_code);
960 len = snprintf(SNPARGS(proto, 0), "ICMP ");
961 len += snprintf(SNPARGS(proto, len), "%s", src);
962 snprintf(SNPARGS(proto, len), " %s", dst);
966 icmp6 = (struct icmp6_hdr *)(((char *)ip) + hlen);
968 len = snprintf(SNPARGS(proto, 0),
970 icmp6->icmp6_type, icmp6->icmp6_code);
972 len = snprintf(SNPARGS(proto, 0), "ICMPv6 ");
973 len += snprintf(SNPARGS(proto, len), "%s", src);
974 snprintf(SNPARGS(proto, len), " %s", dst);
978 len = snprintf(SNPARGS(proto, 0), "P:%d %s",
979 args->f_id.proto, src);
980 snprintf(SNPARGS(proto, len), " %s", dst);
985 if (IS_IP6_FLOW_ID(&(args->f_id))) {
986 if (offset & (IP6F_OFF_MASK | IP6F_MORE_FRAG))
987 snprintf(SNPARGS(fragment, 0),
988 " (frag %08x:%d@%d%s)",
990 ntohs(ip6->ip6_plen) - hlen,
991 ntohs(offset & IP6F_OFF_MASK) << 3,
992 (offset & IP6F_MORE_FRAG) ? "+" : "");
997 if (eh != NULL) { /* layer 2 packets are as on the wire */
998 ip_off = ntohs(ip->ip_off);
999 ip_len = ntohs(ip->ip_len);
1001 ip_off = ip->ip_off;
1002 ip_len = ip->ip_len;
1004 if (ip_off & (IP_MF | IP_OFFMASK))
1005 snprintf(SNPARGS(fragment, 0),
1006 " (frag %d:%d@%d%s)",
1007 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2),
1009 (ip_off & IP_MF) ? "+" : "");
1012 if (oif || m->m_pkthdr.rcvif)
1013 log(LOG_SECURITY | LOG_INFO,
1014 "ipfw: %d %s %s %s via %s%s\n",
1015 f ? f->rulenum : -1,
1016 action, proto, oif ? "out" : "in",
1017 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname,
1020 log(LOG_SECURITY | LOG_INFO,
1021 "ipfw: %d %s %s [no if info]%s\n",
1022 f ? f->rulenum : -1,
1023 action, proto, fragment);
1025 log(LOG_SECURITY | LOG_NOTICE,
1026 "ipfw: limit %d reached on entry %d\n",
1027 limit_reached, f ? f->rulenum : -1);
1031 * IMPORTANT: the hash function for dynamic rules must be commutative
1032 * in source and destination (ip,port), because rules are bidirectional
1033 * and we want to find both in the same bucket.
1036 hash_packet(struct ipfw_flow_id *id)
1041 if (IS_IP6_FLOW_ID(id))
1042 i = hash_packet6(id);
1045 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port);
1046 i &= (curr_dyn_buckets - 1);
1051 * unlink a dynamic rule from a chain. prev is a pointer to
1052 * the previous one, q is a pointer to the rule to delete,
1053 * head is a pointer to the head of the queue.
1054 * Modifies q and potentially also head.
1056 #define UNLINK_DYN_RULE(prev, head, q) { \
1057 ipfw_dyn_rule *old_q = q; \
1059 /* remove a refcount to the parent */ \
1060 if (q->dyn_type == O_LIMIT) \
1061 q->parent->count--; \
1062 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\
1063 (q->id.src_ip), (q->id.src_port), \
1064 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \
1066 prev->next = q = q->next; \
1068 head = q = q->next; \
1070 uma_zfree(ipfw_dyn_rule_zone, old_q); }
1072 #define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0)
1075 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL.
1077 * If keep_me == NULL, rules are deleted even if not expired,
1078 * otherwise only expired rules are removed.
1080 * The value of the second parameter is also used to point to identify
1081 * a rule we absolutely do not want to remove (e.g. because we are
1082 * holding a reference to it -- this is the case with O_LIMIT_PARENT
1083 * rules). The pointer is only used for comparison, so any non-null
1087 remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me)
1089 static u_int32_t last_remove = 0;
1091 #define FORCE (keep_me == NULL)
1093 ipfw_dyn_rule *prev, *q;
1094 int i, pass = 0, max_pass = 0;
1096 IPFW_DYN_LOCK_ASSERT();
1098 if (ipfw_dyn_v == NULL || dyn_count == 0)
1100 /* do not expire more than once per second, it is useless */
1101 if (!FORCE && last_remove == time_uptime)
1103 last_remove = time_uptime;
1106 * because O_LIMIT refer to parent rules, during the first pass only
1107 * remove child and mark any pending LIMIT_PARENT, and remove
1108 * them in a second pass.
1111 for (i = 0 ; i < curr_dyn_buckets ; i++) {
1112 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) {
1114 * Logic can become complex here, so we split tests.
1118 if (rule != NULL && rule != q->rule)
1119 goto next; /* not the one we are looking for */
1120 if (q->dyn_type == O_LIMIT_PARENT) {
1122 * handle parent in the second pass,
1123 * record we need one.
1128 if (FORCE && q->count != 0 ) {
1129 /* XXX should not happen! */
1130 printf("ipfw: OUCH! cannot remove rule,"
1131 " count %d\n", q->count);
1135 !TIME_LEQ( q->expire, time_uptime ))
1138 if (q->dyn_type != O_LIMIT_PARENT || !q->count) {
1139 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
1147 if (pass++ < max_pass)
1153 * lookup a dynamic rule.
1155 static ipfw_dyn_rule *
1156 lookup_dyn_rule_locked(struct ipfw_flow_id *pkt, int *match_direction,
1160 * stateful ipfw extensions.
1161 * Lookup into dynamic session queue
1163 #define MATCH_REVERSE 0
1164 #define MATCH_FORWARD 1
1165 #define MATCH_NONE 2
1166 #define MATCH_UNKNOWN 3
1167 int i, dir = MATCH_NONE;
1168 ipfw_dyn_rule *prev, *q=NULL;
1170 IPFW_DYN_LOCK_ASSERT();
1172 if (ipfw_dyn_v == NULL)
1173 goto done; /* not found */
1174 i = hash_packet( pkt );
1175 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) {
1176 if (q->dyn_type == O_LIMIT_PARENT && q->count)
1178 if (TIME_LEQ( q->expire, time_uptime)) { /* expire entry */
1179 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
1182 if (pkt->proto == q->id.proto &&
1183 q->dyn_type != O_LIMIT_PARENT) {
1184 if (IS_IP6_FLOW_ID(pkt)) {
1185 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6),
1186 &(q->id.src_ip6)) &&
1187 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6),
1188 &(q->id.dst_ip6)) &&
1189 pkt->src_port == q->id.src_port &&
1190 pkt->dst_port == q->id.dst_port ) {
1191 dir = MATCH_FORWARD;
1194 if (IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6),
1195 &(q->id.dst_ip6)) &&
1196 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6),
1197 &(q->id.src_ip6)) &&
1198 pkt->src_port == q->id.dst_port &&
1199 pkt->dst_port == q->id.src_port ) {
1200 dir = MATCH_REVERSE;
1204 if (pkt->src_ip == q->id.src_ip &&
1205 pkt->dst_ip == q->id.dst_ip &&
1206 pkt->src_port == q->id.src_port &&
1207 pkt->dst_port == q->id.dst_port ) {
1208 dir = MATCH_FORWARD;
1211 if (pkt->src_ip == q->id.dst_ip &&
1212 pkt->dst_ip == q->id.src_ip &&
1213 pkt->src_port == q->id.dst_port &&
1214 pkt->dst_port == q->id.src_port ) {
1215 dir = MATCH_REVERSE;
1225 goto done; /* q = NULL, not found */
1227 if ( prev != NULL) { /* found and not in front */
1228 prev->next = q->next;
1229 q->next = ipfw_dyn_v[i];
1232 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */
1233 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST);
1235 #define BOTH_SYN (TH_SYN | (TH_SYN << 8))
1236 #define BOTH_FIN (TH_FIN | (TH_FIN << 8))
1237 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8);
1239 case TH_SYN: /* opening */
1240 q->expire = time_uptime + dyn_syn_lifetime;
1243 case BOTH_SYN: /* move to established */
1244 case BOTH_SYN | TH_FIN : /* one side tries to close */
1245 case BOTH_SYN | (TH_FIN << 8) :
1247 #define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0)
1248 u_int32_t ack = ntohl(tcp->th_ack);
1249 if (dir == MATCH_FORWARD) {
1250 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd))
1252 else { /* ignore out-of-sequence */
1256 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev))
1258 else { /* ignore out-of-sequence */
1263 q->expire = time_uptime + dyn_ack_lifetime;
1266 case BOTH_SYN | BOTH_FIN: /* both sides closed */
1267 if (dyn_fin_lifetime >= dyn_keepalive_period)
1268 dyn_fin_lifetime = dyn_keepalive_period - 1;
1269 q->expire = time_uptime + dyn_fin_lifetime;
1275 * reset or some invalid combination, but can also
1276 * occur if we use keep-state the wrong way.
1278 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0)
1279 printf("invalid state: 0x%x\n", q->state);
1281 if (dyn_rst_lifetime >= dyn_keepalive_period)
1282 dyn_rst_lifetime = dyn_keepalive_period - 1;
1283 q->expire = time_uptime + dyn_rst_lifetime;
1286 } else if (pkt->proto == IPPROTO_UDP) {
1287 q->expire = time_uptime + dyn_udp_lifetime;
1289 /* other protocols */
1290 q->expire = time_uptime + dyn_short_lifetime;
1293 if (match_direction)
1294 *match_direction = dir;
1298 static ipfw_dyn_rule *
1299 lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction,
1305 q = lookup_dyn_rule_locked(pkt, match_direction, tcp);
1308 /* NB: return table locked when q is not NULL */
1313 realloc_dynamic_table(void)
1315 IPFW_DYN_LOCK_ASSERT();
1318 * Try reallocation, make sure we have a power of 2 and do
1319 * not allow more than 64k entries. In case of overflow,
1323 if (dyn_buckets > 65536)
1325 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */
1326 dyn_buckets = curr_dyn_buckets; /* reset */
1329 curr_dyn_buckets = dyn_buckets;
1330 if (ipfw_dyn_v != NULL)
1331 free(ipfw_dyn_v, M_IPFW);
1333 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *),
1334 M_IPFW, M_NOWAIT | M_ZERO);
1335 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2)
1337 curr_dyn_buckets /= 2;
1342 * Install state of type 'type' for a dynamic session.
1343 * The hash table contains two type of rules:
1344 * - regular rules (O_KEEP_STATE)
1345 * - rules for sessions with limited number of sess per user
1346 * (O_LIMIT). When they are created, the parent is
1347 * increased by 1, and decreased on delete. In this case,
1348 * the third parameter is the parent rule and not the chain.
1349 * - "parent" rules for the above (O_LIMIT_PARENT).
1351 static ipfw_dyn_rule *
1352 add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule)
1357 IPFW_DYN_LOCK_ASSERT();
1359 if (ipfw_dyn_v == NULL ||
1360 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) {
1361 realloc_dynamic_table();
1362 if (ipfw_dyn_v == NULL)
1363 return NULL; /* failed ! */
1365 i = hash_packet(id);
1367 r = uma_zalloc(ipfw_dyn_rule_zone, M_NOWAIT | M_ZERO);
1369 printf ("ipfw: sorry cannot allocate state\n");
1373 /* increase refcount on parent, and set pointer */
1374 if (dyn_type == O_LIMIT) {
1375 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule;
1376 if ( parent->dyn_type != O_LIMIT_PARENT)
1377 panic("invalid parent");
1380 rule = parent->rule;
1384 r->expire = time_uptime + dyn_syn_lifetime;
1386 r->dyn_type = dyn_type;
1387 r->pcnt = r->bcnt = 0;
1391 r->next = ipfw_dyn_v[i];
1394 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n",
1396 (r->id.src_ip), (r->id.src_port),
1397 (r->id.dst_ip), (r->id.dst_port),
1403 * lookup dynamic parent rule using pkt and rule as search keys.
1404 * If the lookup fails, then install one.
1406 static ipfw_dyn_rule *
1407 lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule)
1412 IPFW_DYN_LOCK_ASSERT();
1415 int is_v6 = IS_IP6_FLOW_ID(pkt);
1416 i = hash_packet( pkt );
1417 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next)
1418 if (q->dyn_type == O_LIMIT_PARENT &&
1420 pkt->proto == q->id.proto &&
1421 pkt->src_port == q->id.src_port &&
1422 pkt->dst_port == q->id.dst_port &&
1425 IN6_ARE_ADDR_EQUAL(&(pkt->src_ip6),
1426 &(q->id.src_ip6)) &&
1427 IN6_ARE_ADDR_EQUAL(&(pkt->dst_ip6),
1428 &(q->id.dst_ip6))) ||
1430 pkt->src_ip == q->id.src_ip &&
1431 pkt->dst_ip == q->id.dst_ip)
1434 q->expire = time_uptime + dyn_short_lifetime;
1435 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);)
1439 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule);
1443 * Install dynamic state for rule type cmd->o.opcode
1445 * Returns 1 (failure) if state is not installed because of errors or because
1446 * session limitations are enforced.
1449 install_state(struct ip_fw *rule, ipfw_insn_limit *cmd,
1450 struct ip_fw_args *args, uint32_t tablearg)
1452 static int last_log;
1455 char src[48], dst[48];
1461 printf("ipfw: %s: type %d 0x%08x %u -> 0x%08x %u\n",
1462 __func__, cmd->o.opcode,
1463 (args->f_id.src_ip), (args->f_id.src_port),
1464 (args->f_id.dst_ip), (args->f_id.dst_port));
1469 q = lookup_dyn_rule_locked(&args->f_id, NULL, NULL);
1471 if (q != NULL) { /* should never occur */
1472 if (last_log != time_uptime) {
1473 last_log = time_uptime;
1474 printf("ipfw: %s: entry already present, done\n",
1481 if (dyn_count >= dyn_max)
1482 /* Run out of slots, try to remove any expired rule. */
1483 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1);
1485 if (dyn_count >= dyn_max) {
1486 if (last_log != time_uptime) {
1487 last_log = time_uptime;
1488 printf("ipfw: %s: Too many dynamic rules\n", __func__);
1491 return (1); /* cannot install, notify caller */
1494 switch (cmd->o.opcode) {
1495 case O_KEEP_STATE: /* bidir rule */
1496 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule);
1499 case O_LIMIT: { /* limit number of sessions */
1500 struct ipfw_flow_id id;
1501 ipfw_dyn_rule *parent;
1502 uint32_t conn_limit;
1503 uint16_t limit_mask = cmd->limit_mask;
1505 conn_limit = (cmd->conn_limit == IP_FW_TABLEARG) ?
1506 tablearg : cmd->conn_limit;
1509 if (cmd->conn_limit == IP_FW_TABLEARG)
1510 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u "
1511 "(tablearg)\n", __func__, conn_limit);
1513 printf("ipfw: %s: O_LIMIT rule, conn_limit: %u\n",
1514 __func__, conn_limit);
1517 id.dst_ip = id.src_ip = id.dst_port = id.src_port = 0;
1518 id.proto = args->f_id.proto;
1519 id.addr_type = args->f_id.addr_type;
1521 if (IS_IP6_FLOW_ID (&(args->f_id))) {
1522 if (limit_mask & DYN_SRC_ADDR)
1523 id.src_ip6 = args->f_id.src_ip6;
1524 if (limit_mask & DYN_DST_ADDR)
1525 id.dst_ip6 = args->f_id.dst_ip6;
1527 if (limit_mask & DYN_SRC_ADDR)
1528 id.src_ip = args->f_id.src_ip;
1529 if (limit_mask & DYN_DST_ADDR)
1530 id.dst_ip = args->f_id.dst_ip;
1532 if (limit_mask & DYN_SRC_PORT)
1533 id.src_port = args->f_id.src_port;
1534 if (limit_mask & DYN_DST_PORT)
1535 id.dst_port = args->f_id.dst_port;
1536 if ((parent = lookup_dyn_parent(&id, rule)) == NULL) {
1537 printf("ipfw: %s: add parent failed\n", __func__);
1542 if (parent->count >= conn_limit) {
1543 /* See if we can remove some expired rule. */
1544 remove_dyn_rule(rule, parent);
1545 if (parent->count >= conn_limit) {
1546 if (fw_verbose && last_log != time_uptime) {
1547 last_log = time_uptime;
1550 * XXX IPv6 flows are not
1553 if (IS_IP6_FLOW_ID(&(args->f_id))) {
1554 char ip6buf[INET6_ADDRSTRLEN];
1555 snprintf(src, sizeof(src),
1556 "[%s]", ip6_sprintf(ip6buf,
1557 &args->f_id.src_ip6));
1558 snprintf(dst, sizeof(dst),
1559 "[%s]", ip6_sprintf(ip6buf,
1560 &args->f_id.dst_ip6));
1565 htonl(args->f_id.src_ip);
1566 inet_ntoa_r(da, src);
1568 htonl(args->f_id.dst_ip);
1569 inet_ntoa_r(da, dst);
1571 log(LOG_SECURITY | LOG_DEBUG,
1572 "%s %s:%u -> %s:%u, %s\n",
1574 src, (args->f_id.src_port),
1575 dst, (args->f_id.dst_port),
1576 "too many entries");
1582 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent);
1586 printf("ipfw: %s: unknown dynamic rule type %u\n",
1587 __func__, cmd->o.opcode);
1592 /* XXX just set lifetime */
1593 lookup_dyn_rule_locked(&args->f_id, NULL, NULL);
1600 * Generate a TCP packet, containing either a RST or a keepalive.
1601 * When flags & TH_RST, we are sending a RST packet, because of a
1602 * "reset" action matched the packet.
1603 * Otherwise we are sending a keepalive, and flags & TH_
1604 * The 'replyto' mbuf is the mbuf being replied to, if any, and is required
1605 * so that MAC can label the reply appropriately.
1607 static struct mbuf *
1608 send_pkt(struct mbuf *replyto, struct ipfw_flow_id *id, u_int32_t seq,
1609 u_int32_t ack, int flags)
1615 MGETHDR(m, M_DONTWAIT, MT_DATA);
1618 m->m_pkthdr.rcvif = (struct ifnet *)0;
1621 if (replyto != NULL)
1622 mac_create_mbuf_netlayer(replyto, m);
1624 mac_create_mbuf_from_firewall(m);
1626 (void)replyto; /* don't warn about unused arg */
1629 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr);
1630 m->m_data += max_linkhdr;
1632 ip = mtod(m, struct ip *);
1633 bzero(ip, m->m_len);
1634 tcp = (struct tcphdr *)(ip + 1); /* no IP options */
1635 ip->ip_p = IPPROTO_TCP;
1638 * Assume we are sending a RST (or a keepalive in the reverse
1639 * direction), swap src and destination addresses and ports.
1641 ip->ip_src.s_addr = htonl(id->dst_ip);
1642 ip->ip_dst.s_addr = htonl(id->src_ip);
1643 tcp->th_sport = htons(id->dst_port);
1644 tcp->th_dport = htons(id->src_port);
1645 if (flags & TH_RST) { /* we are sending a RST */
1646 if (flags & TH_ACK) {
1647 tcp->th_seq = htonl(ack);
1648 tcp->th_ack = htonl(0);
1649 tcp->th_flags = TH_RST;
1653 tcp->th_seq = htonl(0);
1654 tcp->th_ack = htonl(seq);
1655 tcp->th_flags = TH_RST | TH_ACK;
1659 * We are sending a keepalive. flags & TH_SYN determines
1660 * the direction, forward if set, reverse if clear.
1661 * NOTE: seq and ack are always assumed to be correct
1662 * as set by the caller. This may be confusing...
1664 if (flags & TH_SYN) {
1666 * we have to rewrite the correct addresses!
1668 ip->ip_dst.s_addr = htonl(id->dst_ip);
1669 ip->ip_src.s_addr = htonl(id->src_ip);
1670 tcp->th_dport = htons(id->dst_port);
1671 tcp->th_sport = htons(id->src_port);
1673 tcp->th_seq = htonl(seq);
1674 tcp->th_ack = htonl(ack);
1675 tcp->th_flags = TH_ACK;
1678 * set ip_len to the payload size so we can compute
1679 * the tcp checksum on the pseudoheader
1680 * XXX check this, could save a couple of words ?
1682 ip->ip_len = htons(sizeof(struct tcphdr));
1683 tcp->th_sum = in_cksum(m, m->m_pkthdr.len);
1685 * now fill fields left out earlier
1687 ip->ip_ttl = ip_defttl;
1688 ip->ip_len = m->m_pkthdr.len;
1689 m->m_flags |= M_SKIP_FIREWALL;
1694 * sends a reject message, consuming the mbuf passed as an argument.
1697 send_reject(struct ip_fw_args *args, int code, int ip_len, struct ip *ip)
1701 /* XXX When ip is not guaranteed to be at mtod() we will
1702 * need to account for this */
1703 * The mbuf will however be thrown away so we can adjust it.
1704 * Remember we did an m_pullup on it already so we
1705 * can make some assumptions about contiguousness.
1708 m_adj(m, args->L3offset);
1710 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
1711 /* We need the IP header in host order for icmp_error(). */
1712 if (args->eh != NULL) {
1713 ip->ip_len = ntohs(ip->ip_len);
1714 ip->ip_off = ntohs(ip->ip_off);
1716 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
1717 } else if (args->f_id.proto == IPPROTO_TCP) {
1718 struct tcphdr *const tcp =
1719 L3HDR(struct tcphdr, mtod(args->m, struct ip *));
1720 if ( (tcp->th_flags & TH_RST) == 0) {
1722 m = send_pkt(args->m, &(args->f_id),
1723 ntohl(tcp->th_seq), ntohl(tcp->th_ack),
1724 tcp->th_flags | TH_RST);
1726 ip_output(m, NULL, NULL, 0, NULL, NULL);
1736 * Given an ip_fw *, lookup_next_rule will return a pointer
1737 * to the next rule, which can be either the jump
1738 * target (for skipto instructions) or the next one in the list (in
1739 * all other cases including a missing jump target).
1740 * The result is also written in the "next_rule" field of the rule.
1741 * Backward jumps are not allowed, so start looking from the next
1744 * This never returns NULL -- in case we do not have an exact match,
1745 * the next rule is returned. When the ruleset is changed,
1746 * pointers are flushed so we are always correct.
1749 static struct ip_fw *
1750 lookup_next_rule(struct ip_fw *me)
1752 struct ip_fw *rule = NULL;
1755 /* look for action, in case it is a skipto */
1756 cmd = ACTION_PTR(me);
1757 if (cmd->opcode == O_LOG)
1759 if (cmd->opcode == O_ALTQ)
1761 if (cmd->opcode == O_TAG)
1763 if ( cmd->opcode == O_SKIPTO )
1764 for (rule = me->next; rule ; rule = rule->next)
1765 if (rule->rulenum >= cmd->arg1)
1767 if (rule == NULL) /* failure or not a skipto */
1769 me->next_rule = rule;
1774 add_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr,
1775 uint8_t mlen, uint32_t value)
1777 struct radix_node_head *rnh;
1778 struct table_entry *ent;
1780 if (tbl >= IPFW_TABLES_MAX)
1782 rnh = ch->tables[tbl];
1783 ent = malloc(sizeof(*ent), M_IPFW_TBL, M_NOWAIT | M_ZERO);
1787 ent->addr.sin_len = ent->mask.sin_len = 8;
1788 ent->mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0);
1789 ent->addr.sin_addr.s_addr = addr & ent->mask.sin_addr.s_addr;
1790 IPFW_WLOCK(&layer3_chain);
1791 if (rnh->rnh_addaddr(&ent->addr, &ent->mask, rnh, (void *)ent) ==
1793 IPFW_WUNLOCK(&layer3_chain);
1794 free(ent, M_IPFW_TBL);
1797 IPFW_WUNLOCK(&layer3_chain);
1802 del_table_entry(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr,
1805 struct radix_node_head *rnh;
1806 struct table_entry *ent;
1807 struct sockaddr_in sa, mask;
1809 if (tbl >= IPFW_TABLES_MAX)
1811 rnh = ch->tables[tbl];
1812 sa.sin_len = mask.sin_len = 8;
1813 mask.sin_addr.s_addr = htonl(mlen ? ~((1 << (32 - mlen)) - 1) : 0);
1814 sa.sin_addr.s_addr = addr & mask.sin_addr.s_addr;
1816 ent = (struct table_entry *)rnh->rnh_deladdr(&sa, &mask, rnh);
1822 free(ent, M_IPFW_TBL);
1827 flush_table_entry(struct radix_node *rn, void *arg)
1829 struct radix_node_head * const rnh = arg;
1830 struct table_entry *ent;
1832 ent = (struct table_entry *)
1833 rnh->rnh_deladdr(rn->rn_key, rn->rn_mask, rnh);
1835 free(ent, M_IPFW_TBL);
1840 flush_table(struct ip_fw_chain *ch, uint16_t tbl)
1842 struct radix_node_head *rnh;
1844 IPFW_WLOCK_ASSERT(ch);
1846 if (tbl >= IPFW_TABLES_MAX)
1848 rnh = ch->tables[tbl];
1849 KASSERT(rnh != NULL, ("NULL IPFW table"));
1850 rnh->rnh_walktree(rnh, flush_table_entry, rnh);
1855 flush_tables(struct ip_fw_chain *ch)
1859 IPFW_WLOCK_ASSERT(ch);
1861 for (tbl = 0; tbl < IPFW_TABLES_MAX; tbl++)
1862 flush_table(ch, tbl);
1866 init_tables(struct ip_fw_chain *ch)
1871 for (i = 0; i < IPFW_TABLES_MAX; i++) {
1872 if (!rn_inithead((void **)&ch->tables[i], 32)) {
1873 for (j = 0; j < i; j++) {
1874 (void) flush_table(ch, j);
1883 lookup_table(struct ip_fw_chain *ch, uint16_t tbl, in_addr_t addr,
1886 struct radix_node_head *rnh;
1887 struct table_entry *ent;
1888 struct sockaddr_in sa;
1890 if (tbl >= IPFW_TABLES_MAX)
1892 rnh = ch->tables[tbl];
1894 sa.sin_addr.s_addr = addr;
1895 ent = (struct table_entry *)(rnh->rnh_lookup(&sa, NULL, rnh));
1904 count_table_entry(struct radix_node *rn, void *arg)
1906 u_int32_t * const cnt = arg;
1913 count_table(struct ip_fw_chain *ch, uint32_t tbl, uint32_t *cnt)
1915 struct radix_node_head *rnh;
1917 if (tbl >= IPFW_TABLES_MAX)
1919 rnh = ch->tables[tbl];
1921 rnh->rnh_walktree(rnh, count_table_entry, cnt);
1926 dump_table_entry(struct radix_node *rn, void *arg)
1928 struct table_entry * const n = (struct table_entry *)rn;
1929 ipfw_table * const tbl = arg;
1930 ipfw_table_entry *ent;
1932 if (tbl->cnt == tbl->size)
1934 ent = &tbl->ent[tbl->cnt];
1935 ent->tbl = tbl->tbl;
1936 if (in_nullhost(n->mask.sin_addr))
1939 ent->masklen = 33 - ffs(ntohl(n->mask.sin_addr.s_addr));
1940 ent->addr = n->addr.sin_addr.s_addr;
1941 ent->value = n->value;
1947 dump_table(struct ip_fw_chain *ch, ipfw_table *tbl)
1949 struct radix_node_head *rnh;
1951 if (tbl->tbl >= IPFW_TABLES_MAX)
1953 rnh = ch->tables[tbl->tbl];
1955 rnh->rnh_walktree(rnh, dump_table_entry, tbl);
1960 fill_ugid_cache(struct inpcb *inp, struct ip_fw_ugid *ugp)
1964 if (inp->inp_socket != NULL) {
1965 cr = inp->inp_socket->so_cred;
1966 ugp->fw_prid = jailed(cr) ?
1967 cr->cr_prison->pr_id : -1;
1968 ugp->fw_uid = cr->cr_uid;
1969 ugp->fw_ngroups = cr->cr_ngroups;
1970 bcopy(cr->cr_groups, ugp->fw_groups,
1971 sizeof(ugp->fw_groups));
1976 check_uidgid(ipfw_insn_u32 *insn, int proto, struct ifnet *oif,
1977 struct in_addr dst_ip, u_int16_t dst_port, struct in_addr src_ip,
1978 u_int16_t src_port, struct ip_fw_ugid *ugp, int *lookup,
1981 struct inpcbinfo *pi;
1988 * Check to see if the UDP or TCP stack supplied us with
1989 * the PCB. If so, rather then holding a lock and looking
1990 * up the PCB, we can use the one that was supplied.
1992 if (inp && *lookup == 0) {
1993 INP_LOCK_ASSERT(inp);
1994 if (inp->inp_socket != NULL) {
1995 fill_ugid_cache(inp, ugp);
2000 * If we have already been here and the packet has no
2001 * PCB entry associated with it, then we can safely
2002 * assume that this is a no match.
2006 if (proto == IPPROTO_TCP) {
2009 } else if (proto == IPPROTO_UDP) {
2010 wildcard = INPLOOKUP_WILDCARD;
2018 in_pcblookup_hash(pi,
2019 dst_ip, htons(dst_port),
2020 src_ip, htons(src_port),
2022 in_pcblookup_hash(pi,
2023 src_ip, htons(src_port),
2024 dst_ip, htons(dst_port),
2028 if (pcb->inp_socket != NULL) {
2029 fill_ugid_cache(pcb, ugp);
2034 INP_INFO_RUNLOCK(pi);
2037 * If the lookup did not yield any results, there
2038 * is no sense in coming back and trying again. So
2039 * we can set lookup to -1 and ensure that we wont
2040 * bother the pcb system again.
2046 if (insn->o.opcode == O_UID)
2047 match = (ugp->fw_uid == (uid_t)insn->d[0]);
2048 else if (insn->o.opcode == O_GID) {
2049 for (gp = ugp->fw_groups;
2050 gp < &ugp->fw_groups[ugp->fw_ngroups]; gp++)
2051 if (*gp == (gid_t)insn->d[0]) {
2055 } else if (insn->o.opcode == O_JAIL)
2056 match = (ugp->fw_prid == (int)insn->d[0]);
2060 #ifdef IPFIREWALL_NAT
2061 static eventhandler_tag ifaddr_event_tag;
2064 ifaddr_change(void *arg __unused, struct ifnet *ifp)
2066 struct cfg_nat *ptr;
2069 IPFW_WLOCK(&layer3_chain);
2070 /* Check every nat entry... */
2071 LIST_FOREACH(ptr, &layer3_chain.nat, _next) {
2072 /* ...using nic 'ifp->if_xname' as dynamic alias address. */
2073 if (strncmp(ptr->if_name, ifp->if_xname, IF_NAMESIZE) == 0) {
2074 mtx_lock(&ifp->if_addr_mtx);
2075 TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
2076 if (ifa->ifa_addr == NULL)
2078 if (ifa->ifa_addr->sa_family != AF_INET)
2080 ptr->ip = ((struct sockaddr_in *)
2081 (ifa->ifa_addr))->sin_addr;
2082 LibAliasSetAddress(ptr->lib, ptr->ip);
2084 mtx_unlock(&ifp->if_addr_mtx);
2087 IPFW_WUNLOCK(&layer3_chain);
2091 flush_nat_ptrs(const int i)
2095 IPFW_WLOCK_ASSERT(&layer3_chain);
2096 for (rule = layer3_chain.rules; rule; rule = rule->next) {
2097 ipfw_insn_nat *cmd = (ipfw_insn_nat *)ACTION_PTR(rule);
2098 if (cmd->o.opcode != O_NAT)
2100 if (cmd->nat != NULL && cmd->nat->id == i)
2105 static struct cfg_nat *
2106 lookup_nat(const int i)
2108 struct cfg_nat *ptr;
2110 LIST_FOREACH(ptr, &layer3_chain.nat, _next)
2116 #define HOOK_NAT(b, p) do { \
2117 IPFW_WLOCK_ASSERT(&layer3_chain); \
2118 LIST_INSERT_HEAD(b, p, _next); \
2121 #define UNHOOK_NAT(p) do { \
2122 IPFW_WLOCK_ASSERT(&layer3_chain); \
2123 LIST_REMOVE(p, _next); \
2126 #define HOOK_REDIR(b, p) do { \
2127 LIST_INSERT_HEAD(b, p, _next); \
2130 #define HOOK_SPOOL(b, p) do { \
2131 LIST_INSERT_HEAD(b, p, _next); \
2135 del_redir_spool_cfg(struct cfg_nat *n, struct redir_chain *head)
2137 struct cfg_redir *r, *tmp_r;
2138 struct cfg_spool *s, *tmp_s;
2141 LIST_FOREACH_SAFE(r, head, _next, tmp_r) {
2142 num = 1; /* Number of alias_link to delete. */
2149 /* Delete all libalias redirect entry. */
2150 for (i = 0; i < num; i++)
2151 LibAliasRedirectDelete(n->lib, r->alink[i]);
2152 /* Del spool cfg if any. */
2153 LIST_FOREACH_SAFE(s, &r->spool_chain, _next, tmp_s) {
2154 LIST_REMOVE(s, _next);
2157 free(r->alink, M_IPFW);
2158 LIST_REMOVE(r, _next);
2162 printf("unknown redirect mode: %u\n", r->mode);
2163 /* XXX - panic?!?!? */
2170 add_redir_spool_cfg(char *buf, struct cfg_nat *ptr)
2172 struct cfg_redir *r, *ser_r;
2173 struct cfg_spool *s, *ser_s;
2177 for (cnt = 0, off = 0; cnt < ptr->redir_cnt; cnt++) {
2178 ser_r = (struct cfg_redir *)&buf[off];
2179 r = malloc(SOF_REDIR, M_IPFW, M_WAITOK | M_ZERO);
2180 memcpy(r, ser_r, SOF_REDIR);
2181 LIST_INIT(&r->spool_chain);
2183 r->alink = malloc(sizeof(struct alias_link *) * r->pport_cnt,
2184 M_IPFW, M_WAITOK | M_ZERO);
2187 r->alink[0] = LibAliasRedirectAddr(ptr->lib, r->laddr,
2191 for (i = 0 ; i < r->pport_cnt; i++) {
2192 /* If remotePort is all ports, set it to 0. */
2193 u_short remotePortCopy = r->rport + i;
2194 if (r->rport_cnt == 1 && r->rport == 0)
2196 r->alink[i] = LibAliasRedirectPort(ptr->lib,
2197 r->laddr, htons(r->lport + i), r->raddr,
2198 htons(remotePortCopy), r->paddr,
2199 htons(r->pport + i), r->proto);
2200 if (r->alink[i] == NULL) {
2207 r->alink[0] = LibAliasRedirectProto(ptr->lib ,r->laddr,
2208 r->raddr, r->paddr, r->proto);
2211 printf("unknown redirect mode: %u\n", r->mode);
2214 if (r->alink[0] == NULL) {
2215 panic_err = "LibAliasRedirect* returned NULL";
2217 } else /* LSNAT handling. */
2218 for (i = 0; i < r->spool_cnt; i++) {
2219 ser_s = (struct cfg_spool *)&buf[off];
2220 s = malloc(SOF_REDIR, M_IPFW,
2222 memcpy(s, ser_s, SOF_SPOOL);
2223 LibAliasAddServer(ptr->lib, r->alink[0],
2224 s->addr, htons(s->port));
2226 /* Hook spool entry. */
2227 HOOK_SPOOL(&r->spool_chain, s);
2229 /* And finally hook this redir entry. */
2230 HOOK_REDIR(&ptr->redir_chain, r);
2234 /* something really bad happened: panic! */
2235 panic("%s\n", panic_err);
2240 * The main check routine for the firewall.
2242 * All arguments are in args so we can modify them and return them
2243 * back to the caller.
2247 * args->m (in/out) The packet; we set to NULL when/if we nuke it.
2248 * Starts with the IP header.
2249 * args->eh (in) Mac header if present, or NULL for layer3 packet.
2250 * args->L3offset Number of bytes bypassed if we came from L2.
2251 * e.g. often sizeof(eh) ** NOTYET **
2252 * args->oif Outgoing interface, or NULL if packet is incoming.
2253 * The incoming interface is in the mbuf. (in)
2254 * args->divert_rule (in/out)
2255 * Skip up to the first rule past this rule number;
2256 * upon return, non-zero port number for divert or tee.
2258 * args->rule Pointer to the last matching rule (in/out)
2259 * args->next_hop Socket we are forwarding to (out).
2260 * args->f_id Addresses grabbed from the packet (out)
2261 * args->cookie a cookie depending on rule action
2265 * IP_FW_PASS the packet must be accepted
2266 * IP_FW_DENY the packet must be dropped
2267 * IP_FW_DIVERT divert packet, port in m_tag
2268 * IP_FW_TEE tee packet, port in m_tag
2269 * IP_FW_DUMMYNET to dummynet, pipe in args->cookie
2270 * IP_FW_NETGRAPH into netgraph, cookie args->cookie
2274 ipfw_chk(struct ip_fw_args *args)
2277 * Local variables holding state during the processing of a packet:
2279 * IMPORTANT NOTE: to speed up the processing of rules, there
2280 * are some assumption on the values of the variables, which
2281 * are documented here. Should you change them, please check
2282 * the implementation of the various instructions to make sure
2283 * that they still work.
2285 * args->eh The MAC header. It is non-null for a layer2
2286 * packet, it is NULL for a layer-3 packet.
2288 * args->L3offset Offset in the packet to the L3 (IP or equiv.) header.
2290 * m | args->m Pointer to the mbuf, as received from the caller.
2291 * It may change if ipfw_chk() does an m_pullup, or if it
2292 * consumes the packet because it calls send_reject().
2293 * XXX This has to change, so that ipfw_chk() never modifies
2294 * or consumes the buffer.
2295 * ip is the beginning of the ip(4 or 6) header.
2296 * Calculated by adding the L3offset to the start of data.
2297 * (Until we start using L3offset, the packet is
2298 * supposed to start with the ip header).
2300 struct mbuf *m = args->m;
2301 struct ip *ip = mtod(m, struct ip *);
2304 * For rules which contain uid/gid or jail constraints, cache
2305 * a copy of the users credentials after the pcb lookup has been
2306 * executed. This will speed up the processing of rules with
2307 * these types of constraints, as well as decrease contention
2308 * on pcb related locks.
2310 struct ip_fw_ugid fw_ugid_cache;
2311 int ugid_lookup = 0;
2314 * divinput_flags If non-zero, set to the IP_FW_DIVERT_*_FLAG
2315 * associated with a packet input on a divert socket. This
2316 * will allow to distinguish traffic and its direction when
2317 * it originates from a divert socket.
2319 u_int divinput_flags = 0;
2322 * oif | args->oif If NULL, ipfw_chk has been called on the
2323 * inbound path (ether_input, ip_input).
2324 * If non-NULL, ipfw_chk has been called on the outbound path
2325 * (ether_output, ip_output).
2327 struct ifnet *oif = args->oif;
2329 struct ip_fw *f = NULL; /* matching rule */
2333 * hlen The length of the IP header.
2335 u_int hlen = 0; /* hlen >0 means we have an IP pkt */
2338 * offset The offset of a fragment. offset != 0 means that
2339 * we have a fragment at this offset of an IPv4 packet.
2340 * offset == 0 means that (if this is an IPv4 packet)
2341 * this is the first or only fragment.
2342 * For IPv6 offset == 0 means there is no Fragment Header.
2343 * If offset != 0 for IPv6 always use correct mask to
2344 * get the correct offset because we add IP6F_MORE_FRAG
2345 * to be able to dectect the first fragment which would
2346 * otherwise have offset = 0.
2351 * Local copies of addresses. They are only valid if we have
2354 * proto The protocol. Set to 0 for non-ip packets,
2355 * or to the protocol read from the packet otherwise.
2356 * proto != 0 means that we have an IPv4 packet.
2358 * src_port, dst_port port numbers, in HOST format. Only
2359 * valid for TCP and UDP packets.
2361 * src_ip, dst_ip ip addresses, in NETWORK format.
2362 * Only valid for IPv4 packets.
2365 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */
2366 struct in_addr src_ip, dst_ip; /* NOTE: network format */
2369 u_int16_t etype = 0; /* Host order stored ether type */
2372 * dyn_dir = MATCH_UNKNOWN when rules unchecked,
2373 * MATCH_NONE when checked and not matched (q = NULL),
2374 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
2376 int dyn_dir = MATCH_UNKNOWN;
2377 ipfw_dyn_rule *q = NULL;
2378 struct ip_fw_chain *chain = &layer3_chain;
2382 * We store in ulp a pointer to the upper layer protocol header.
2383 * In the ipv4 case this is easy to determine from the header,
2384 * but for ipv6 we might have some additional headers in the middle.
2385 * ulp is NULL if not found.
2387 void *ulp = NULL; /* upper layer protocol pointer. */
2388 /* XXX ipv6 variables */
2390 u_int16_t ext_hd = 0; /* bits vector for extension header filtering */
2391 /* end of ipv6 variables */
2394 if (m->m_flags & M_SKIP_FIREWALL)
2395 return (IP_FW_PASS); /* accept */
2397 pktlen = m->m_pkthdr.len;
2398 proto = args->f_id.proto = 0; /* mark f_id invalid */
2399 /* XXX 0 is a valid proto: IP/IPv6 Hop-by-Hop Option */
2402 * PULLUP_TO(len, p, T) makes sure that len + sizeof(T) is contiguous,
2403 * then it sets p to point at the offset "len" in the mbuf. WARNING: the
2404 * pointer might become stale after other pullups (but we never use it
2407 #define PULLUP_TO(len, p, T) \
2409 int x = (len) + sizeof(T); \
2410 if ((m)->m_len < x) { \
2411 args->m = m = m_pullup(m, x); \
2413 goto pullup_failed; \
2415 p = (mtod(m, char *) + (len)); \
2419 * if we have an ether header,
2422 etype = ntohs(args->eh->ether_type);
2424 /* Identify IP packets and fill up variables. */
2425 if (pktlen >= sizeof(struct ip6_hdr) &&
2426 (args->eh == NULL || etype == ETHERTYPE_IPV6) && ip->ip_v == 6) {
2427 struct ip6_hdr *ip6 = (struct ip6_hdr *)ip;
2429 args->f_id.addr_type = 6;
2430 hlen = sizeof(struct ip6_hdr);
2431 proto = ip6->ip6_nxt;
2433 /* Search extension headers to find upper layer protocols */
2434 while (ulp == NULL) {
2436 case IPPROTO_ICMPV6:
2437 PULLUP_TO(hlen, ulp, struct icmp6_hdr);
2438 args->f_id.flags = ICMP6(ulp)->icmp6_type;
2442 PULLUP_TO(hlen, ulp, struct tcphdr);
2443 dst_port = TCP(ulp)->th_dport;
2444 src_port = TCP(ulp)->th_sport;
2445 args->f_id.flags = TCP(ulp)->th_flags;
2449 PULLUP_TO(hlen, ulp, struct sctphdr);
2450 src_port = SCTP(ulp)->src_port;
2451 dst_port = SCTP(ulp)->dest_port;
2455 PULLUP_TO(hlen, ulp, struct udphdr);
2456 dst_port = UDP(ulp)->uh_dport;
2457 src_port = UDP(ulp)->uh_sport;
2460 case IPPROTO_HOPOPTS: /* RFC 2460 */
2461 PULLUP_TO(hlen, ulp, struct ip6_hbh);
2462 ext_hd |= EXT_HOPOPTS;
2463 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
2464 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
2468 case IPPROTO_ROUTING: /* RFC 2460 */
2469 PULLUP_TO(hlen, ulp, struct ip6_rthdr);
2470 switch (((struct ip6_rthdr *)ulp)->ip6r_type) {
2472 ext_hd |= EXT_RTHDR0;
2475 ext_hd |= EXT_RTHDR2;
2478 printf("IPFW2: IPV6 - Unknown Routing "
2479 "Header type(%d)\n",
2480 ((struct ip6_rthdr *)ulp)->ip6r_type);
2481 if (fw_deny_unknown_exthdrs)
2482 return (IP_FW_DENY);
2485 ext_hd |= EXT_ROUTING;
2486 hlen += (((struct ip6_rthdr *)ulp)->ip6r_len + 1) << 3;
2487 proto = ((struct ip6_rthdr *)ulp)->ip6r_nxt;
2491 case IPPROTO_FRAGMENT: /* RFC 2460 */
2492 PULLUP_TO(hlen, ulp, struct ip6_frag);
2493 ext_hd |= EXT_FRAGMENT;
2494 hlen += sizeof (struct ip6_frag);
2495 proto = ((struct ip6_frag *)ulp)->ip6f_nxt;
2496 offset = ((struct ip6_frag *)ulp)->ip6f_offlg &
2498 /* Add IP6F_MORE_FRAG for offset of first
2499 * fragment to be != 0. */
2500 offset |= ((struct ip6_frag *)ulp)->ip6f_offlg &
2503 printf("IPFW2: IPV6 - Invalid Fragment "
2505 if (fw_deny_unknown_exthdrs)
2506 return (IP_FW_DENY);
2509 args->f_id.frag_id6 =
2510 ntohl(((struct ip6_frag *)ulp)->ip6f_ident);
2514 case IPPROTO_DSTOPTS: /* RFC 2460 */
2515 PULLUP_TO(hlen, ulp, struct ip6_hbh);
2516 ext_hd |= EXT_DSTOPTS;
2517 hlen += (((struct ip6_hbh *)ulp)->ip6h_len + 1) << 3;
2518 proto = ((struct ip6_hbh *)ulp)->ip6h_nxt;
2522 case IPPROTO_AH: /* RFC 2402 */
2523 PULLUP_TO(hlen, ulp, struct ip6_ext);
2525 hlen += (((struct ip6_ext *)ulp)->ip6e_len + 2) << 2;
2526 proto = ((struct ip6_ext *)ulp)->ip6e_nxt;
2530 case IPPROTO_ESP: /* RFC 2406 */
2531 PULLUP_TO(hlen, ulp, uint32_t); /* SPI, Seq# */
2532 /* Anything past Seq# is variable length and
2533 * data past this ext. header is encrypted. */
2537 case IPPROTO_NONE: /* RFC 2460 */
2538 PULLUP_TO(hlen, ulp, struct ip6_ext);
2539 /* Packet ends here. if ip6e_len!=0 octets
2540 * must be ignored. */
2543 case IPPROTO_OSPFIGP:
2544 /* XXX OSPF header check? */
2545 PULLUP_TO(hlen, ulp, struct ip6_ext);
2549 /* XXX PIM header check? */
2550 PULLUP_TO(hlen, ulp, struct pim);
2554 PULLUP_TO(hlen, ulp, struct carp_header);
2555 if (((struct carp_header *)ulp)->carp_version !=
2557 return (IP_FW_DENY);
2558 if (((struct carp_header *)ulp)->carp_type !=
2560 return (IP_FW_DENY);
2563 case IPPROTO_IPV6: /* RFC 2893 */
2564 PULLUP_TO(hlen, ulp, struct ip6_hdr);
2567 case IPPROTO_IPV4: /* RFC 2893 */
2568 PULLUP_TO(hlen, ulp, struct ip);
2572 printf("IPFW2: IPV6 - Unknown Extension "
2573 "Header(%d), ext_hd=%x\n", proto, ext_hd);
2574 if (fw_deny_unknown_exthdrs)
2575 return (IP_FW_DENY);
2576 PULLUP_TO(hlen, ulp, struct ip6_ext);
2580 ip = mtod(m, struct ip *);
2581 ip6 = (struct ip6_hdr *)ip;
2582 args->f_id.src_ip6 = ip6->ip6_src;
2583 args->f_id.dst_ip6 = ip6->ip6_dst;
2584 args->f_id.src_ip = 0;
2585 args->f_id.dst_ip = 0;
2586 args->f_id.flow_id6 = ntohl(ip6->ip6_flow);
2587 } else if (pktlen >= sizeof(struct ip) &&
2588 (args->eh == NULL || etype == ETHERTYPE_IP) && ip->ip_v == 4) {
2590 hlen = ip->ip_hl << 2;
2591 args->f_id.addr_type = 4;
2594 * Collect parameters into local variables for faster matching.
2597 src_ip = ip->ip_src;
2598 dst_ip = ip->ip_dst;
2599 if (args->eh != NULL) { /* layer 2 packets are as on the wire */
2600 offset = ntohs(ip->ip_off) & IP_OFFMASK;
2601 ip_len = ntohs(ip->ip_len);
2603 offset = ip->ip_off & IP_OFFMASK;
2604 ip_len = ip->ip_len;
2606 pktlen = ip_len < pktlen ? ip_len : pktlen;
2611 PULLUP_TO(hlen, ulp, struct tcphdr);
2612 dst_port = TCP(ulp)->th_dport;
2613 src_port = TCP(ulp)->th_sport;
2614 args->f_id.flags = TCP(ulp)->th_flags;
2618 PULLUP_TO(hlen, ulp, struct udphdr);
2619 dst_port = UDP(ulp)->uh_dport;
2620 src_port = UDP(ulp)->uh_sport;
2624 PULLUP_TO(hlen, ulp, struct icmphdr);
2625 args->f_id.flags = ICMP(ulp)->icmp_type;
2633 ip = mtod(m, struct ip *);
2634 args->f_id.src_ip = ntohl(src_ip.s_addr);
2635 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
2638 if (proto) { /* we may have port numbers, store them */
2639 args->f_id.proto = proto;
2640 args->f_id.src_port = src_port = ntohs(src_port);
2641 args->f_id.dst_port = dst_port = ntohs(dst_port);
2645 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL);
2648 * Packet has already been tagged. Look for the next rule
2649 * to restart processing.
2651 * If fw_one_pass != 0 then just accept it.
2652 * XXX should not happen here, but optimized out in
2656 IPFW_RUNLOCK(chain);
2657 return (IP_FW_PASS);
2660 f = args->rule->next_rule;
2662 f = lookup_next_rule(args->rule);
2665 * Find the starting rule. It can be either the first
2666 * one, or the one after divert_rule if asked so.
2668 int skipto = mtag ? divert_cookie(mtag) : 0;
2671 if (args->eh == NULL && skipto != 0) {
2672 if (skipto >= IPFW_DEFAULT_RULE) {
2673 IPFW_RUNLOCK(chain);
2674 return (IP_FW_DENY); /* invalid */
2676 while (f && f->rulenum <= skipto)
2678 if (f == NULL) { /* drop packet */
2679 IPFW_RUNLOCK(chain);
2680 return (IP_FW_DENY);
2684 /* reset divert rule to avoid confusion later */
2686 divinput_flags = divert_info(mtag) &
2687 (IP_FW_DIVERT_OUTPUT_FLAG | IP_FW_DIVERT_LOOPBACK_FLAG);
2688 m_tag_delete(m, mtag);
2692 * Now scan the rules, and parse microinstructions for each rule.
2694 for (; f; f = f->next) {
2696 uint32_t tablearg = 0;
2697 int l, cmdlen, skip_or; /* skip rest of OR block */
2700 if (set_disable & (1 << f->set) )
2704 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
2705 l -= cmdlen, cmd += cmdlen) {
2709 * check_body is a jump target used when we find a
2710 * CHECK_STATE, and need to jump to the body of
2715 cmdlen = F_LEN(cmd);
2717 * An OR block (insn_1 || .. || insn_n) has the
2718 * F_OR bit set in all but the last instruction.
2719 * The first match will set "skip_or", and cause
2720 * the following instructions to be skipped until
2721 * past the one with the F_OR bit clear.
2723 if (skip_or) { /* skip this instruction */
2724 if ((cmd->len & F_OR) == 0)
2725 skip_or = 0; /* next one is good */
2728 match = 0; /* set to 1 if we succeed */
2730 switch (cmd->opcode) {
2732 * The first set of opcodes compares the packet's
2733 * fields with some pattern, setting 'match' if a
2734 * match is found. At the end of the loop there is
2735 * logic to deal with F_NOT and F_OR flags associated
2743 printf("ipfw: opcode %d unimplemented\n",
2751 * We only check offset == 0 && proto != 0,
2752 * as this ensures that we have a
2753 * packet with the ports info.
2757 if (is_ipv6) /* XXX to be fixed later */
2759 if (proto == IPPROTO_TCP ||
2760 proto == IPPROTO_UDP)
2761 match = check_uidgid(
2762 (ipfw_insn_u32 *)cmd,
2765 src_ip, src_port, &fw_ugid_cache,
2766 &ugid_lookup, args->inp);
2770 match = iface_match(m->m_pkthdr.rcvif,
2771 (ipfw_insn_if *)cmd);
2775 match = iface_match(oif, (ipfw_insn_if *)cmd);
2779 match = iface_match(oif ? oif :
2780 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd);
2784 if (args->eh != NULL) { /* have MAC header */
2785 u_int32_t *want = (u_int32_t *)
2786 ((ipfw_insn_mac *)cmd)->addr;
2787 u_int32_t *mask = (u_int32_t *)
2788 ((ipfw_insn_mac *)cmd)->mask;
2789 u_int32_t *hdr = (u_int32_t *)args->eh;
2792 ( want[0] == (hdr[0] & mask[0]) &&
2793 want[1] == (hdr[1] & mask[1]) &&
2794 want[2] == (hdr[2] & mask[2]) );
2799 if (args->eh != NULL) {
2801 ((ipfw_insn_u16 *)cmd)->ports;
2804 for (i = cmdlen - 1; !match && i>0;
2806 match = (etype >= p[0] &&
2812 match = (offset != 0);
2815 case O_IN: /* "out" is "not in" */
2816 match = (oif == NULL);
2820 match = (args->eh != NULL);
2824 match = (cmd->arg1 & 1 && divinput_flags &
2825 IP_FW_DIVERT_LOOPBACK_FLAG) ||
2826 (cmd->arg1 & 2 && divinput_flags &
2827 IP_FW_DIVERT_OUTPUT_FLAG);
2832 * We do not allow an arg of 0 so the
2833 * check of "proto" only suffices.
2835 match = (proto == cmd->arg1);
2840 (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2844 case O_IP_SRC_LOOKUP:
2845 case O_IP_DST_LOOKUP:
2848 (cmd->opcode == O_IP_DST_LOOKUP) ?
2849 dst_ip.s_addr : src_ip.s_addr;
2852 match = lookup_table(chain, cmd->arg1, a,
2856 if (cmdlen == F_INSN_SIZE(ipfw_insn_u32))
2858 ((ipfw_insn_u32 *)cmd)->d[0] == v;
2868 (cmd->opcode == O_IP_DST_MASK) ?
2869 dst_ip.s_addr : src_ip.s_addr;
2870 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2873 for (; !match && i>0; i-= 2, p+= 2)
2874 match = (p[0] == (a & p[1]));
2882 INADDR_TO_IFP(src_ip, tif);
2883 match = (tif != NULL);
2890 u_int32_t *d = (u_int32_t *)(cmd+1);
2892 cmd->opcode == O_IP_DST_SET ?
2898 addr -= d[0]; /* subtract base */
2899 match = (addr < cmd->arg1) &&
2900 ( d[ 1 + (addr>>5)] &
2901 (1<<(addr & 0x1f)) );
2907 (((ipfw_insn_ip *)cmd)->addr.s_addr ==
2915 INADDR_TO_IFP(dst_ip, tif);
2916 match = (tif != NULL);
2923 * offset == 0 && proto != 0 is enough
2924 * to guarantee that we have a
2925 * packet with port info.
2927 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
2930 (cmd->opcode == O_IP_SRCPORT) ?
2931 src_port : dst_port ;
2933 ((ipfw_insn_u16 *)cmd)->ports;
2936 for (i = cmdlen - 1; !match && i>0;
2938 match = (x>=p[0] && x<=p[1]);
2943 match = (offset == 0 && proto==IPPROTO_ICMP &&
2944 icmptype_match(ICMP(ulp), (ipfw_insn_u32 *)cmd) );
2949 match = is_ipv6 && offset == 0 &&
2950 proto==IPPROTO_ICMPV6 &&
2952 ICMP6(ulp)->icmp6_type,
2953 (ipfw_insn_u32 *)cmd);
2959 ipopts_match(ip, cmd) );
2964 cmd->arg1 == ip->ip_v);
2970 if (is_ipv4) { /* only for IP packets */
2975 if (cmd->opcode == O_IPLEN)
2977 else if (cmd->opcode == O_IPTTL)
2979 else /* must be IPID */
2980 x = ntohs(ip->ip_id);
2982 match = (cmd->arg1 == x);
2985 /* otherwise we have ranges */
2986 p = ((ipfw_insn_u16 *)cmd)->ports;
2988 for (; !match && i>0; i--, p += 2)
2989 match = (x >= p[0] && x <= p[1]);
2993 case O_IPPRECEDENCE:
2995 (cmd->arg1 == (ip->ip_tos & 0xe0)) );
3000 flags_match(cmd, ip->ip_tos));
3004 if (proto == IPPROTO_TCP && offset == 0) {
3012 ((ip->ip_hl + tcp->th_off) << 2);
3014 match = (cmd->arg1 == x);
3017 /* otherwise we have ranges */
3018 p = ((ipfw_insn_u16 *)cmd)->ports;
3020 for (; !match && i>0; i--, p += 2)
3021 match = (x >= p[0] && x <= p[1]);
3026 match = (proto == IPPROTO_TCP && offset == 0 &&
3027 flags_match(cmd, TCP(ulp)->th_flags));
3031 match = (proto == IPPROTO_TCP && offset == 0 &&
3032 tcpopts_match(TCP(ulp), cmd));
3036 match = (proto == IPPROTO_TCP && offset == 0 &&
3037 ((ipfw_insn_u32 *)cmd)->d[0] ==
3042 match = (proto == IPPROTO_TCP && offset == 0 &&
3043 ((ipfw_insn_u32 *)cmd)->d[0] ==
3048 match = (proto == IPPROTO_TCP && offset == 0 &&
3049 cmd->arg1 == TCP(ulp)->th_win);
3053 /* reject packets which have SYN only */
3054 /* XXX should i also check for TH_ACK ? */
3055 match = (proto == IPPROTO_TCP && offset == 0 &&
3056 (TCP(ulp)->th_flags &
3057 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
3061 struct altq_tag *at;
3062 ipfw_insn_altq *altq = (ipfw_insn_altq *)cmd;
3065 mtag = m_tag_find(m, PACKET_TAG_PF_QID, NULL);
3068 mtag = m_tag_get(PACKET_TAG_PF_QID,
3069 sizeof(struct altq_tag),
3073 * Let the packet fall back to the
3078 at = (struct altq_tag *)(mtag+1);
3079 at->qid = altq->qid;
3085 m_tag_prepend(m, mtag);
3091 ipfw_log(f, hlen, args, m,
3092 oif, offset, tablearg, ip);
3097 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
3101 /* Outgoing packets automatically pass/match */
3102 match = ((oif != NULL) ||
3103 (m->m_pkthdr.rcvif == NULL) ||
3107 verify_path6(&(args->f_id.src_ip6),
3108 m->m_pkthdr.rcvif) :
3110 verify_path(src_ip, m->m_pkthdr.rcvif)));
3114 /* Outgoing packets automatically pass/match */
3115 match = (hlen > 0 && ((oif != NULL) ||
3118 verify_path6(&(args->f_id.src_ip6),
3121 verify_path(src_ip, NULL)));
3125 /* Outgoing packets automatically pass/match */
3126 if (oif == NULL && hlen > 0 &&
3127 ( (is_ipv4 && in_localaddr(src_ip))
3130 in6_localaddr(&(args->f_id.src_ip6)))
3135 is_ipv6 ? verify_path6(
3136 &(args->f_id.src_ip6),
3137 m->m_pkthdr.rcvif) :
3147 match = (m_tag_find(m,
3148 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
3150 /* otherwise no match */
3156 IN6_ARE_ADDR_EQUAL(&args->f_id.src_ip6,
3157 &((ipfw_insn_ip6 *)cmd)->addr6);
3162 IN6_ARE_ADDR_EQUAL(&args->f_id.dst_ip6,
3163 &((ipfw_insn_ip6 *)cmd)->addr6);
3165 case O_IP6_SRC_MASK:
3166 case O_IP6_DST_MASK:
3170 struct in6_addr *d =
3171 &((ipfw_insn_ip6 *)cmd)->addr6;
3173 for (; !match && i > 0; d += 2,
3174 i -= F_INSN_SIZE(struct in6_addr)
3180 APPLY_MASK(&p, &d[1]);
3182 IN6_ARE_ADDR_EQUAL(&d[0],
3189 match= is_ipv6 && search_ip6_addr_net(&args->f_id.src_ip6);
3193 match= is_ipv6 && search_ip6_addr_net(&args->f_id.dst_ip6);
3198 flow6id_match(args->f_id.flow_id6,
3199 (ipfw_insn_u32 *) cmd);
3204 (ext_hd & ((ipfw_insn *) cmd)->arg1);
3217 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
3218 tablearg : cmd->arg1;
3220 /* Packet is already tagged with this tag? */
3221 mtag = m_tag_locate(m, MTAG_IPFW, tag, NULL);
3223 /* We have `untag' action when F_NOT flag is
3224 * present. And we must remove this mtag from
3225 * mbuf and reset `match' to zero (`match' will
3226 * be inversed later).
3227 * Otherwise we should allocate new mtag and
3228 * push it into mbuf.
3230 if (cmd->len & F_NOT) { /* `untag' action */
3232 m_tag_delete(m, mtag);
3233 } else if (mtag == NULL) {
3234 if ((mtag = m_tag_alloc(MTAG_IPFW,
3235 tag, 0, M_NOWAIT)) != NULL)
3236 m_tag_prepend(m, mtag);
3238 match = (cmd->len & F_NOT) ? 0: 1;
3243 uint32_t tag = (cmd->arg1 == IP_FW_TABLEARG) ?
3244 tablearg : cmd->arg1;
3247 match = m_tag_locate(m, MTAG_IPFW,
3252 /* we have ranges */
3253 for (mtag = m_tag_first(m);
3254 mtag != NULL && !match;
3255 mtag = m_tag_next(m, mtag)) {
3259 if (mtag->m_tag_cookie != MTAG_IPFW)
3262 p = ((ipfw_insn_u16 *)cmd)->ports;
3264 for(; !match && i > 0; i--, p += 2)
3266 mtag->m_tag_id >= p[0] &&
3267 mtag->m_tag_id <= p[1];
3273 * The second set of opcodes represents 'actions',
3274 * i.e. the terminal part of a rule once the packet
3275 * matches all previous patterns.
3276 * Typically there is only one action for each rule,
3277 * and the opcode is stored at the end of the rule
3278 * (but there are exceptions -- see below).
3280 * In general, here we set retval and terminate the
3281 * outer loop (would be a 'break 3' in some language,
3282 * but we need to do a 'goto done').
3285 * O_COUNT and O_SKIPTO actions:
3286 * instead of terminating, we jump to the next rule
3287 * ('goto next_rule', equivalent to a 'break 2'),
3288 * or to the SKIPTO target ('goto again' after
3289 * having set f, cmd and l), respectively.
3291 * O_TAG, O_LOG and O_ALTQ action parameters:
3292 * perform some action and set match = 1;
3294 * O_LIMIT and O_KEEP_STATE: these opcodes are
3295 * not real 'actions', and are stored right
3296 * before the 'action' part of the rule.
3297 * These opcodes try to install an entry in the
3298 * state tables; if successful, we continue with
3299 * the next opcode (match=1; break;), otherwise
3300 * the packet * must be dropped
3301 * ('goto done' after setting retval);
3303 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
3304 * cause a lookup of the state table, and a jump
3305 * to the 'action' part of the parent rule
3306 * ('goto check_body') if an entry is found, or
3307 * (CHECK_STATE only) a jump to the next rule if
3308 * the entry is not found ('goto next_rule').
3309 * The result of the lookup is cached to make
3310 * further instances of these opcodes are
3315 if (install_state(f,
3316 (ipfw_insn_limit *)cmd, args, tablearg)) {
3317 retval = IP_FW_DENY;
3318 goto done; /* error/limit violation */
3326 * dynamic rules are checked at the first
3327 * keep-state or check-state occurrence,
3328 * with the result being stored in dyn_dir.
3329 * The compiler introduces a PROBE_STATE
3330 * instruction for us when we have a
3331 * KEEP_STATE (because PROBE_STATE needs
3334 if (dyn_dir == MATCH_UNKNOWN &&
3335 (q = lookup_dyn_rule(&args->f_id,
3336 &dyn_dir, proto == IPPROTO_TCP ?
3340 * Found dynamic entry, update stats
3341 * and jump to the 'action' part of
3347 cmd = ACTION_PTR(f);
3348 l = f->cmd_len - f->act_ofs;
3353 * Dynamic entry not found. If CHECK_STATE,
3354 * skip to next rule, if PROBE_STATE just
3355 * ignore and continue with next opcode.
3357 if (cmd->opcode == O_CHECK_STATE)
3363 retval = 0; /* accept */
3368 args->rule = f; /* report matching rule */
3369 if (cmd->arg1 == IP_FW_TABLEARG)
3370 args->cookie = tablearg;
3372 args->cookie = cmd->arg1;
3373 retval = IP_FW_DUMMYNET;
3378 struct divert_tag *dt;
3380 if (args->eh) /* not on layer 2 */
3382 mtag = m_tag_get(PACKET_TAG_DIVERT,
3383 sizeof(struct divert_tag),
3388 IPFW_RUNLOCK(chain);
3389 return (IP_FW_DENY);
3391 dt = (struct divert_tag *)(mtag+1);
3392 dt->cookie = f->rulenum;
3393 if (cmd->arg1 == IP_FW_TABLEARG)
3394 dt->info = tablearg;
3396 dt->info = cmd->arg1;
3397 m_tag_prepend(m, mtag);
3398 retval = (cmd->opcode == O_DIVERT) ?
3399 IP_FW_DIVERT : IP_FW_TEE;
3405 f->pcnt++; /* update stats */
3407 f->timestamp = time_uptime;
3408 if (cmd->opcode == O_COUNT)
3411 if (f->next_rule == NULL)
3412 lookup_next_rule(f);
3418 * Drop the packet and send a reject notice
3419 * if the packet is not ICMP (or is an ICMP
3420 * query), and it is not multicast/broadcast.
3422 if (hlen > 0 && is_ipv4 && offset == 0 &&
3423 (proto != IPPROTO_ICMP ||
3424 is_icmp_query(ICMP(ulp))) &&
3425 !(m->m_flags & (M_BCAST|M_MCAST)) &&
3426 !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
3427 send_reject(args, cmd->arg1, ip_len, ip);
3433 if (hlen > 0 && is_ipv6 &&
3434 ((offset & IP6F_OFF_MASK) == 0) &&
3435 (proto != IPPROTO_ICMPV6 ||
3436 (is_icmp6_query(args->f_id.flags) == 1)) &&
3437 !(m->m_flags & (M_BCAST|M_MCAST)) &&
3438 !IN6_IS_ADDR_MULTICAST(&args->f_id.dst_ip6)) {
3440 args, cmd->arg1, hlen,
3441 (struct ip6_hdr *)ip);
3447 retval = IP_FW_DENY;
3450 case O_FORWARD_IP: {
3451 struct sockaddr_in *sa;
3452 sa = &(((ipfw_insn_sa *)cmd)->sa);
3453 if (args->eh) /* not valid on layer2 pkts */
3455 if (!q || dyn_dir == MATCH_FORWARD) {
3456 if (sa->sin_addr.s_addr == INADDR_ANY) {
3457 bcopy(sa, &args->hopstore,
3459 args->hopstore.sin_addr.s_addr =
3464 args->next_hop = sa;
3467 retval = IP_FW_PASS;
3473 args->rule = f; /* report matching rule */
3474 if (cmd->arg1 == IP_FW_TABLEARG)
3475 args->cookie = tablearg;
3477 args->cookie = cmd->arg1;
3478 retval = (cmd->opcode == O_NETGRAPH) ?
3479 IP_FW_NETGRAPH : IP_FW_NGTEE;
3482 #ifdef IPFIREWALL_NAT
3486 /* XXX - libalias duct tape */
3491 args->rule = f; /* Report matching rule. */
3493 t = ((ipfw_insn_nat *)cmd)->nat;
3495 t = lookup_nat(cmd->arg1);
3497 retval = IP_FW_DENY;
3500 ((ipfw_insn_nat *)cmd)->nat =
3503 if ((mcl = m_megapullup(m, m->m_pkthdr.len)) ==
3506 ip = mtod(mcl, struct ip *);
3507 if (args->eh == NULL) {
3508 ip->ip_len = htons(ip->ip_len);
3509 ip->ip_off = htons(ip->ip_off);
3513 * XXX - Libalias checksum offload 'duct tape':
3515 * locally generated packets have only
3516 * pseudo-header checksum calculated
3517 * and libalias will screw it[1], so
3518 * mark them for later fix. Moreover
3519 * there are cases when libalias
3520 * modify tcp packet data[2], mark it
3521 * for later fix too.
3523 * [1] libalias was never meant to run
3524 * in kernel, so it doesn't have any
3525 * knowledge about checksum
3526 * offloading, and it expects a packet
3527 * with a full internet
3528 * checksum. Unfortunately, packets
3529 * generated locally will have just the
3530 * pseudo header calculated, and when
3531 * libalias tries to adjust the
3532 * checksum it will actually screw it.
3534 * [2] when libalias modify tcp's data
3535 * content, full TCP checksum has to
3536 * be recomputed: the problem is that
3537 * libalias doesn't have any idea
3538 * about checksum offloading To
3539 * workaround this, we do not do
3540 * checksumming in LibAlias, but only
3541 * mark the packets in th_x2 field. If
3542 * we receive a marked packet, we
3543 * calculate correct checksum for it
3544 * aware of offloading. Why such a
3545 * terrible hack instead of
3546 * recalculating checksum for each
3547 * packet? Because the previous
3548 * checksum was not checked!
3549 * Recalculating checksums for EVERY
3550 * packet will hide ALL transmission
3551 * errors. Yes, marked packets still
3552 * suffer from this problem. But,
3553 * sigh, natd(8) has this problem,
3556 * TODO: -make libalias mbuf aware (so
3557 * it can handle delayed checksum and tso)
3560 if (mcl->m_pkthdr.rcvif == NULL &&
3561 mcl->m_pkthdr.csum_flags &
3565 c = mtod(mcl, char *);
3567 retval = LibAliasIn(t->lib, c,
3570 retval = LibAliasOut(t->lib, c,
3572 if (retval != PKT_ALIAS_OK) {
3573 /* XXX - should i add some logging? */
3577 retval = IP_FW_DENY;
3580 mcl->m_pkthdr.len = mcl->m_len =
3584 * XXX - libalias checksum offload
3585 * 'duct tape' (see above)
3588 if ((ip->ip_off & htons(IP_OFFMASK)) == 0 &&
3589 ip->ip_p == IPPROTO_TCP) {
3592 th = (struct tcphdr *)(ip + 1);
3602 ip->ip_len = ntohs(ip->ip_len);
3606 htons(ip->ip_p + ip->ip_len -
3612 th = (struct tcphdr *)(ip + 1);
3614 * Maybe it was set in
3619 mcl->m_pkthdr.csum_data =
3620 offsetof(struct tcphdr,
3624 uh = (struct udphdr *)(ip + 1);
3626 mcl->m_pkthdr.csum_data =
3627 offsetof(struct udphdr,
3632 * No hw checksum offloading: do it
3635 if ((mcl->m_pkthdr.csum_flags &
3636 CSUM_DELAY_DATA) == 0) {
3637 in_delayed_cksum(mcl);
3638 mcl->m_pkthdr.csum_flags &=
3641 ip->ip_len = htons(ip->ip_len);
3644 if (args->eh == NULL) {
3645 ip->ip_len = ntohs(ip->ip_len);
3646 ip->ip_off = ntohs(ip->ip_off);
3656 panic("-- unknown opcode %d\n", cmd->opcode);
3657 } /* end of switch() on opcodes */
3659 if (cmd->len & F_NOT)
3663 if (cmd->len & F_OR)
3666 if (!(cmd->len & F_OR)) /* not an OR block, */
3667 break; /* try next rule */
3670 } /* end of inner for, scan opcodes */
3672 next_rule:; /* try next rule */
3674 } /* end of outer for, scan rules */
3675 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
3676 IPFW_RUNLOCK(chain);
3677 return (IP_FW_DENY);
3680 /* Update statistics */
3683 f->timestamp = time_uptime;
3684 IPFW_RUNLOCK(chain);
3689 printf("ipfw: pullup failed\n");
3690 return (IP_FW_DENY);
3694 * When a rule is added/deleted, clear the next_rule pointers in all rules.
3695 * These will be reconstructed on the fly as packets are matched.
3698 flush_rule_ptrs(struct ip_fw_chain *chain)
3702 IPFW_WLOCK_ASSERT(chain);
3704 for (rule = chain->rules; rule; rule = rule->next)
3705 rule->next_rule = NULL;
3709 * Add a new rule to the list. Copy the rule into a malloc'ed area, then
3710 * possibly create a rule number and add the rule to the list.
3711 * Update the rule_number in the input struct so the caller knows it as well.
3714 add_rule(struct ip_fw_chain *chain, struct ip_fw *input_rule)
3716 struct ip_fw *rule, *f, *prev;
3717 int l = RULESIZE(input_rule);
3719 if (chain->rules == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE)
3722 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO);
3726 bcopy(input_rule, rule, l);
3729 rule->next_rule = NULL;
3733 rule->timestamp = 0;
3737 if (chain->rules == NULL) { /* default rule */
3738 chain->rules = rule;
3743 * If rulenum is 0, find highest numbered rule before the
3744 * default rule, and add autoinc_step
3746 if (autoinc_step < 1)
3748 else if (autoinc_step > 1000)
3749 autoinc_step = 1000;
3750 if (rule->rulenum == 0) {
3752 * locate the highest numbered rule before default
3754 for (f = chain->rules; f; f = f->next) {
3755 if (f->rulenum == IPFW_DEFAULT_RULE)
3757 rule->rulenum = f->rulenum;
3759 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step)
3760 rule->rulenum += autoinc_step;
3761 input_rule->rulenum = rule->rulenum;
3765 * Now insert the new rule in the right place in the sorted list.
3767 for (prev = NULL, f = chain->rules; f; prev = f, f = f->next) {
3768 if (f->rulenum > rule->rulenum) { /* found the location */
3772 } else { /* head insert */
3773 rule->next = chain->rules;
3774 chain->rules = rule;
3779 flush_rule_ptrs(chain);
3783 IPFW_WUNLOCK(chain);
3784 DEB(printf("ipfw: installed rule %d, static count now %d\n",
3785 rule->rulenum, static_count);)
3790 * Remove a static rule (including derived * dynamic rules)
3791 * and place it on the ``reap list'' for later reclamation.
3792 * The caller is in charge of clearing rule pointers to avoid
3793 * dangling pointers.
3794 * @return a pointer to the next entry.
3795 * Arguments are not checked, so they better be correct.
3797 static struct ip_fw *
3798 remove_rule(struct ip_fw_chain *chain, struct ip_fw *rule,
3802 int l = RULESIZE(rule);
3804 IPFW_WLOCK_ASSERT(chain);
3808 remove_dyn_rule(rule, NULL /* force removal */);
3817 rule->next = chain->reap;
3824 * Reclaim storage associated with a list of rules. This is
3825 * typically the list created using remove_rule.
3828 reap_rules(struct ip_fw *head)
3832 while ((rule = head) != NULL) {
3834 if (DUMMYNET_LOADED)
3835 ip_dn_ruledel_ptr(rule);
3841 * Remove all rules from a chain (except rules in set RESVD_SET
3842 * unless kill_default = 1). The caller is responsible for
3843 * reclaiming storage for the rules left in chain->reap.
3846 free_chain(struct ip_fw_chain *chain, int kill_default)
3848 struct ip_fw *prev, *rule;
3850 IPFW_WLOCK_ASSERT(chain);
3852 flush_rule_ptrs(chain); /* more efficient to do outside the loop */
3853 for (prev = NULL, rule = chain->rules; rule ; )
3854 if (kill_default || rule->set != RESVD_SET)
3855 rule = remove_rule(chain, rule, prev);
3863 * Remove all rules with given number, and also do set manipulation.
3864 * Assumes chain != NULL && *chain != NULL.
3866 * The argument is an u_int32_t. The low 16 bit are the rule or set number,
3867 * the next 8 bits are the new set, the top 8 bits are the command:
3869 * 0 delete rules with given number
3870 * 1 delete rules with given set number
3871 * 2 move rules with given number to new set
3872 * 3 move rules with given set number to new set
3873 * 4 swap sets with given numbers
3874 * 5 delete rules with given number and with given set number
3877 del_entry(struct ip_fw_chain *chain, u_int32_t arg)
3879 struct ip_fw *prev = NULL, *rule;
3880 u_int16_t rulenum; /* rule or old_set */
3881 u_int8_t cmd, new_set;
3883 rulenum = arg & 0xffff;
3884 cmd = (arg >> 24) & 0xff;
3885 new_set = (arg >> 16) & 0xff;
3887 if (cmd > 5 || new_set > RESVD_SET)
3889 if (cmd == 0 || cmd == 2 || cmd == 5) {
3890 if (rulenum >= IPFW_DEFAULT_RULE)
3893 if (rulenum > RESVD_SET) /* old_set */
3898 rule = chain->rules;
3901 case 0: /* delete rules with given number */
3903 * locate first rule to delete
3905 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next)
3907 if (rule->rulenum != rulenum) {
3908 IPFW_WUNLOCK(chain);
3913 * flush pointers outside the loop, then delete all matching
3914 * rules. prev remains the same throughout the cycle.
3916 flush_rule_ptrs(chain);
3917 while (rule->rulenum == rulenum)
3918 rule = remove_rule(chain, rule, prev);
3921 case 1: /* delete all rules with given set number */
3922 flush_rule_ptrs(chain);
3923 rule = chain->rules;
3924 while (rule->rulenum < IPFW_DEFAULT_RULE)
3925 if (rule->set == rulenum)
3926 rule = remove_rule(chain, rule, prev);
3933 case 2: /* move rules with given number to new set */
3934 rule = chain->rules;
3935 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
3936 if (rule->rulenum == rulenum)
3937 rule->set = new_set;
3940 case 3: /* move rules with given set number to new set */
3941 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
3942 if (rule->set == rulenum)
3943 rule->set = new_set;
3946 case 4: /* swap two sets */
3947 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
3948 if (rule->set == rulenum)
3949 rule->set = new_set;
3950 else if (rule->set == new_set)
3951 rule->set = rulenum;
3953 case 5: /* delete rules with given number and with given set number.
3954 * rulenum - given rule number;
3955 * new_set - given set number.
3957 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next)
3959 if (rule->rulenum != rulenum) {
3960 IPFW_WUNLOCK(chain);
3963 flush_rule_ptrs(chain);
3964 while (rule->rulenum == rulenum) {
3965 if (rule->set == new_set)
3966 rule = remove_rule(chain, rule, prev);
3974 * Look for rules to reclaim. We grab the list before
3975 * releasing the lock then reclaim them w/o the lock to
3976 * avoid a LOR with dummynet.
3980 IPFW_WUNLOCK(chain);
3987 * Clear counters for a specific rule.
3988 * The enclosing "table" is assumed locked.
3991 clear_counters(struct ip_fw *rule, int log_only)
3993 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule);
3995 if (log_only == 0) {
3996 rule->bcnt = rule->pcnt = 0;
3997 rule->timestamp = 0;
3999 if (l->o.opcode == O_LOG)
4000 l->log_left = l->max_log;
4004 * Reset some or all counters on firewall rules.
4005 * The argument `arg' is an u_int32_t. The low 16 bit are the rule number,
4006 * the next 8 bits are the set number, the top 8 bits are the command:
4007 * 0 work with rules from all set's;
4008 * 1 work with rules only from specified set.
4009 * Specified rule number is zero if we want to clear all entries.
4010 * log_only is 1 if we only want to reset logs, zero otherwise.
4013 zero_entry(struct ip_fw_chain *chain, u_int32_t arg, int log_only)
4018 uint16_t rulenum = arg & 0xffff;
4019 uint8_t set = (arg >> 16) & 0xff;
4020 uint8_t cmd = (arg >> 24) & 0xff;
4024 if (cmd == 1 && set > RESVD_SET)
4030 for (rule = chain->rules; rule; rule = rule->next) {
4031 /* Skip rules from another set. */
4032 if (cmd == 1 && rule->set != set)
4034 clear_counters(rule, log_only);
4036 msg = log_only ? "ipfw: All logging counts reset.\n" :
4037 "ipfw: Accounting cleared.\n";
4041 * We can have multiple rules with the same number, so we
4042 * need to clear them all.
4044 for (rule = chain->rules; rule; rule = rule->next)
4045 if (rule->rulenum == rulenum) {
4046 while (rule && rule->rulenum == rulenum) {
4047 if (cmd == 0 || rule->set == set)
4048 clear_counters(rule, log_only);
4054 if (!cleared) { /* we did not find any matching rules */
4055 IPFW_WUNLOCK(chain);
4058 msg = log_only ? "ipfw: Entry %d logging count reset.\n" :
4059 "ipfw: Entry %d cleared.\n";
4061 IPFW_WUNLOCK(chain);
4064 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum);
4069 * Check validity of the structure before insert.
4070 * Fortunately rules are simple, so this mostly need to check rule sizes.
4073 check_ipfw_struct(struct ip_fw *rule, int size)
4079 if (size < sizeof(*rule)) {
4080 printf("ipfw: rule too short\n");
4083 /* first, check for valid size */
4086 printf("ipfw: size mismatch (have %d want %d)\n", size, l);
4089 if (rule->act_ofs >= rule->cmd_len) {
4090 printf("ipfw: bogus action offset (%u > %u)\n",
4091 rule->act_ofs, rule->cmd_len - 1);
4095 * Now go for the individual checks. Very simple ones, basically only
4096 * instruction sizes.
4098 for (l = rule->cmd_len, cmd = rule->cmd ;
4099 l > 0 ; l -= cmdlen, cmd += cmdlen) {
4100 cmdlen = F_LEN(cmd);
4102 printf("ipfw: opcode %d size truncated\n",
4106 DEB(printf("ipfw: opcode %d\n", cmd->opcode);)
4107 switch (cmd->opcode) {
4119 case O_IPPRECEDENCE:
4137 if (cmdlen != F_INSN_SIZE(ipfw_insn))
4150 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32))
4155 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit))
4160 if (cmdlen != F_INSN_SIZE(ipfw_insn_log))
4163 ((ipfw_insn_log *)cmd)->log_left =
4164 ((ipfw_insn_log *)cmd)->max_log;
4170 /* only odd command lengths */
4171 if ( !(cmdlen & 1) || cmdlen > 31)
4177 if (cmd->arg1 == 0 || cmd->arg1 > 256) {
4178 printf("ipfw: invalid set size %d\n",
4182 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
4187 case O_IP_SRC_LOOKUP:
4188 case O_IP_DST_LOOKUP:
4189 if (cmd->arg1 >= IPFW_TABLES_MAX) {
4190 printf("ipfw: invalid table number %d\n",
4194 if (cmdlen != F_INSN_SIZE(ipfw_insn) &&
4195 cmdlen != F_INSN_SIZE(ipfw_insn_u32))
4200 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac))
4210 if (cmdlen < 1 || cmdlen > 31)
4216 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */
4217 if (cmdlen < 2 || cmdlen > 31)
4224 if (cmdlen != F_INSN_SIZE(ipfw_insn_if))
4229 if (cmdlen != F_INSN_SIZE(ipfw_insn_altq))
4235 if (cmdlen != F_INSN_SIZE(ipfw_insn))
4240 #ifdef IPFIREWALL_FORWARD
4241 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa))
4250 if (ip_divert_ptr == NULL)
4256 if (!NG_IPFW_LOADED)
4261 #ifdef IPFIREWALL_NAT
4262 if (cmdlen != F_INSN_SIZE(ipfw_insn_nat))
4268 case O_FORWARD_MAC: /* XXX not implemented yet */
4279 if (cmdlen != F_INSN_SIZE(ipfw_insn))
4283 printf("ipfw: opcode %d, multiple actions"
4290 printf("ipfw: opcode %d, action must be"
4299 if (cmdlen != F_INSN_SIZE(struct in6_addr) +
4300 F_INSN_SIZE(ipfw_insn))
4305 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
4306 ((ipfw_insn_u32 *)cmd)->o.arg1)
4310 case O_IP6_SRC_MASK:
4311 case O_IP6_DST_MASK:
4312 if ( !(cmdlen & 1) || cmdlen > 127)
4316 if( cmdlen != F_INSN_SIZE( ipfw_insn_icmp6 ) )
4322 switch (cmd->opcode) {
4332 case O_IP6_SRC_MASK:
4333 case O_IP6_DST_MASK:
4335 printf("ipfw: no IPv6 support in kernel\n");
4336 return EPROTONOSUPPORT;
4339 printf("ipfw: opcode %d, unknown opcode\n",
4345 if (have_action == 0) {
4346 printf("ipfw: missing action\n");
4352 printf("ipfw: opcode %d size %d wrong\n",
4353 cmd->opcode, cmdlen);
4358 * Copy the static and dynamic rules to the supplied buffer
4359 * and return the amount of space actually used.
4362 ipfw_getrules(struct ip_fw_chain *chain, void *buf, size_t space)
4365 char *ep = bp + space;
4368 time_t boot_seconds;
4370 boot_seconds = boottime.tv_sec;
4371 /* XXX this can take a long time and locking will block packet flow */
4373 for (rule = chain->rules; rule ; rule = rule->next) {
4375 * Verify the entry fits in the buffer in case the
4376 * rules changed between calculating buffer space and
4377 * now. This would be better done using a generation
4378 * number but should suffice for now.
4384 * XXX HACK. Store the disable mask in the "next" pointer
4385 * in a wild attempt to keep the ABI the same.
4386 * Why do we do this on EVERY rule?
4388 bcopy(&set_disable, &(((struct ip_fw *)bp)->next_rule),
4389 sizeof(set_disable));
4390 if (((struct ip_fw *)bp)->timestamp)
4391 ((struct ip_fw *)bp)->timestamp += boot_seconds;
4395 IPFW_RUNLOCK(chain);
4397 ipfw_dyn_rule *p, *last = NULL;
4400 for (i = 0 ; i < curr_dyn_buckets; i++)
4401 for (p = ipfw_dyn_v[i] ; p != NULL; p = p->next) {
4402 if (bp + sizeof *p <= ep) {
4403 ipfw_dyn_rule *dst =
4404 (ipfw_dyn_rule *)bp;
4405 bcopy(p, dst, sizeof *p);
4406 bcopy(&(p->rule->rulenum), &(dst->rule),
4407 sizeof(p->rule->rulenum));
4409 * store set number into high word of
4410 * dst->rule pointer.
4412 bcopy(&(p->rule->set), &dst->rule +
4413 sizeof(p->rule->rulenum),
4414 sizeof(p->rule->set));
4416 * store a non-null value in "next".
4417 * The userland code will interpret a
4418 * NULL here as a marker
4419 * for the last dynamic rule.
4421 bcopy(&dst, &dst->next, sizeof(dst));
4424 TIME_LEQ(dst->expire, time_uptime) ?
4425 0 : dst->expire - time_uptime ;
4426 bp += sizeof(ipfw_dyn_rule);
4430 if (last != NULL) /* mark last dynamic rule */
4431 bzero(&last->next, sizeof(last));
4433 return (bp - (char *)buf);
4438 * {set|get}sockopt parser.
4441 ipfw_ctl(struct sockopt *sopt)
4443 #define RULE_MAXSIZE (256*sizeof(u_int32_t))
4446 struct ip_fw *buf, *rule;
4447 u_int32_t rulenum[2];
4449 error = priv_check(sopt->sopt_td, PRIV_NETINET_IPFW);
4454 * Disallow modifications in really-really secure mode, but still allow
4455 * the logging counters to be reset.
4457 if (sopt->sopt_name == IP_FW_ADD ||
4458 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) {
4459 error = securelevel_ge(sopt->sopt_td->td_ucred, 3);
4466 switch (sopt->sopt_name) {
4469 * pass up a copy of the current rules. Static rules
4470 * come first (the last of which has number IPFW_DEFAULT_RULE),
4471 * followed by a possibly empty list of dynamic rule.
4472 * The last dynamic rule has NULL in the "next" field.
4474 * Note that the calculated size is used to bound the
4475 * amount of data returned to the user. The rule set may
4476 * change between calculating the size and returning the
4477 * data in which case we'll just return what fits.
4479 size = static_len; /* size of static rules */
4480 if (ipfw_dyn_v) /* add size of dyn.rules */
4481 size += (dyn_count * sizeof(ipfw_dyn_rule));
4484 * XXX todo: if the user passes a short length just to know
4485 * how much room is needed, do not bother filling up the
4486 * buffer, just jump to the sooptcopyout.
4488 buf = malloc(size, M_TEMP, M_WAITOK);
4489 error = sooptcopyout(sopt, buf,
4490 ipfw_getrules(&layer3_chain, buf, size));
4496 * Normally we cannot release the lock on each iteration.
4497 * We could do it here only because we start from the head all
4498 * the times so there is no risk of missing some entries.
4499 * On the other hand, the risk is that we end up with
4500 * a very inconsistent ruleset, so better keep the lock
4501 * around the whole cycle.
4503 * XXX this code can be improved by resetting the head of
4504 * the list to point to the default rule, and then freeing
4505 * the old list without the need for a lock.
4508 IPFW_WLOCK(&layer3_chain);
4509 layer3_chain.reap = NULL;
4510 free_chain(&layer3_chain, 0 /* keep default rule */);
4511 rule = layer3_chain.reap;
4512 layer3_chain.reap = NULL;
4513 IPFW_WUNLOCK(&layer3_chain);
4519 rule = malloc(RULE_MAXSIZE, M_TEMP, M_WAITOK);
4520 error = sooptcopyin(sopt, rule, RULE_MAXSIZE,
4521 sizeof(struct ip_fw) );
4523 error = check_ipfw_struct(rule, sopt->sopt_valsize);
4525 error = add_rule(&layer3_chain, rule);
4526 size = RULESIZE(rule);
4527 if (!error && sopt->sopt_dir == SOPT_GET)
4528 error = sooptcopyout(sopt, rule, size);
4535 * IP_FW_DEL is used for deleting single rules or sets,
4536 * and (ab)used to atomically manipulate sets. Argument size
4537 * is used to distinguish between the two:
4539 * delete single rule or set of rules,
4540 * or reassign rules (or sets) to a different set.
4541 * 2*sizeof(u_int32_t)
4542 * atomic disable/enable sets.
4543 * first u_int32_t contains sets to be disabled,
4544 * second u_int32_t contains sets to be enabled.
4546 error = sooptcopyin(sopt, rulenum,
4547 2*sizeof(u_int32_t), sizeof(u_int32_t));
4550 size = sopt->sopt_valsize;
4551 if (size == sizeof(u_int32_t)) /* delete or reassign */
4552 error = del_entry(&layer3_chain, rulenum[0]);
4553 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */
4555 (set_disable | rulenum[0]) & ~rulenum[1] &
4556 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */
4562 case IP_FW_RESETLOG: /* argument is an u_int_32, the rule number */
4564 if (sopt->sopt_val != 0) {
4565 error = sooptcopyin(sopt, rulenum,
4566 sizeof(u_int32_t), sizeof(u_int32_t));
4570 error = zero_entry(&layer3_chain, rulenum[0],
4571 sopt->sopt_name == IP_FW_RESETLOG);
4574 case IP_FW_TABLE_ADD:
4576 ipfw_table_entry ent;
4578 error = sooptcopyin(sopt, &ent,
4579 sizeof(ent), sizeof(ent));
4582 error = add_table_entry(&layer3_chain, ent.tbl,
4583 ent.addr, ent.masklen, ent.value);
4587 case IP_FW_TABLE_DEL:
4589 ipfw_table_entry ent;
4591 error = sooptcopyin(sopt, &ent,
4592 sizeof(ent), sizeof(ent));
4595 error = del_table_entry(&layer3_chain, ent.tbl,
4596 ent.addr, ent.masklen);
4600 case IP_FW_TABLE_FLUSH:
4604 error = sooptcopyin(sopt, &tbl,
4605 sizeof(tbl), sizeof(tbl));
4608 IPFW_WLOCK(&layer3_chain);
4609 error = flush_table(&layer3_chain, tbl);
4610 IPFW_WUNLOCK(&layer3_chain);
4614 case IP_FW_TABLE_GETSIZE:
4618 if ((error = sooptcopyin(sopt, &tbl, sizeof(tbl),
4621 IPFW_RLOCK(&layer3_chain);
4622 error = count_table(&layer3_chain, tbl, &cnt);
4623 IPFW_RUNLOCK(&layer3_chain);
4626 error = sooptcopyout(sopt, &cnt, sizeof(cnt));
4630 case IP_FW_TABLE_LIST:
4634 if (sopt->sopt_valsize < sizeof(*tbl)) {
4638 size = sopt->sopt_valsize;
4639 tbl = malloc(size, M_TEMP, M_WAITOK);
4640 error = sooptcopyin(sopt, tbl, size, sizeof(*tbl));
4645 tbl->size = (size - sizeof(*tbl)) /
4646 sizeof(ipfw_table_entry);
4647 IPFW_RLOCK(&layer3_chain);
4648 error = dump_table(&layer3_chain, tbl);
4649 IPFW_RUNLOCK(&layer3_chain);
4654 error = sooptcopyout(sopt, tbl, size);
4659 #ifdef IPFIREWALL_NAT
4662 struct cfg_nat *ptr, *ser_n;
4665 buf = malloc(NAT_BUF_LEN, M_IPFW, M_WAITOK | M_ZERO);
4666 error = sooptcopyin(sopt, buf, NAT_BUF_LEN,
4667 sizeof(struct cfg_nat));
4668 ser_n = (struct cfg_nat *)buf;
4671 * Find/create nat rule.
4673 IPFW_WLOCK(&layer3_chain);
4674 ptr = lookup_nat(ser_n->id);
4676 /* New rule: allocate and init new instance. */
4677 ptr = malloc(sizeof(struct cfg_nat),
4678 M_IPFW, M_NOWAIT | M_ZERO);
4680 IPFW_WUNLOCK(&layer3_chain);
4684 ptr->lib = LibAliasInit(NULL);
4685 if (ptr->lib == NULL) {
4686 IPFW_WUNLOCK(&layer3_chain);
4691 LIST_INIT(&ptr->redir_chain);
4693 /* Entry already present: temporarly unhook it. */
4695 flush_nat_ptrs(ser_n->id);
4697 IPFW_WUNLOCK(&layer3_chain);
4700 * Basic nat configuration.
4702 ptr->id = ser_n->id;
4704 * XXX - what if this rule doesn't nat any ip and just
4706 * do we set aliasaddress to 0.0.0.0?
4708 ptr->ip = ser_n->ip;
4709 ptr->redir_cnt = ser_n->redir_cnt;
4710 ptr->mode = ser_n->mode;
4711 LibAliasSetMode(ptr->lib, ser_n->mode, ser_n->mode);
4712 LibAliasSetAddress(ptr->lib, ptr->ip);
4713 memcpy(ptr->if_name, ser_n->if_name, IF_NAMESIZE);
4716 * Redir and LSNAT configuration.
4718 /* Delete old cfgs. */
4719 del_redir_spool_cfg(ptr, &ptr->redir_chain);
4720 /* Add new entries. */
4721 add_redir_spool_cfg(&buf[(sizeof(struct cfg_nat))], ptr);
4723 IPFW_WLOCK(&layer3_chain);
4724 HOOK_NAT(&layer3_chain.nat, ptr);
4725 IPFW_WUNLOCK(&layer3_chain);
4731 struct cfg_nat *ptr;
4734 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
4735 IPFW_WLOCK(&layer3_chain);
4736 ptr = lookup_nat(i);
4739 IPFW_WUNLOCK(&layer3_chain);
4744 IPFW_WUNLOCK(&layer3_chain);
4745 del_redir_spool_cfg(ptr, &ptr->redir_chain);
4746 LibAliasUninit(ptr->lib);
4751 case IP_FW_NAT_GET_CONFIG:
4755 struct cfg_redir *r;
4756 struct cfg_spool *s;
4760 off = sizeof(nat_cnt);
4762 data = malloc(NAT_BUF_LEN, M_IPFW, M_WAITOK | M_ZERO);
4763 IPFW_RLOCK(&layer3_chain);
4764 /* Serialize all the data. */
4765 LIST_FOREACH(n, &layer3_chain.nat, _next) {
4767 if (off + SOF_NAT < NAT_BUF_LEN) {
4768 bcopy(n, &data[off], SOF_NAT);
4770 LIST_FOREACH(r, &n->redir_chain, _next) {
4771 if (off + SOF_REDIR < NAT_BUF_LEN) {
4772 bcopy(r, &data[off],
4775 LIST_FOREACH(s, &r->spool_chain,
4777 if (off + SOF_SPOOL <
4793 bcopy(&nat_cnt, data, sizeof(nat_cnt));
4794 IPFW_RUNLOCK(&layer3_chain);
4795 error = sooptcopyout(sopt, data, NAT_BUF_LEN);
4799 IPFW_RUNLOCK(&layer3_chain);
4800 printf("serialized data buffer not big enough:"
4801 "please increase NAT_BUF_LEN\n");
4806 case IP_FW_NAT_GET_LOG:
4809 struct cfg_nat *ptr;
4810 int i, size, cnt, sof;
4813 sof = LIBALIAS_BUF_SIZE;
4816 IPFW_RLOCK(&layer3_chain);
4818 LIST_FOREACH(ptr, &layer3_chain.nat, _next) {
4819 if (ptr->lib->logDesc == NULL)
4822 size = cnt * (sof + sizeof(int));
4823 data = realloc(data, size, M_IPFW, M_NOWAIT | M_ZERO);
4825 IPFW_RUNLOCK(&layer3_chain);
4828 bcopy(&ptr->id, &data[i], sizeof(int));
4830 bcopy(ptr->lib->logDesc, &data[i], sof);
4833 IPFW_RUNLOCK(&layer3_chain);
4834 error = sooptcopyout(sopt, data, size);
4841 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name);
4850 * dummynet needs a reference to the default rule, because rules can be
4851 * deleted while packets hold a reference to them. When this happens,
4852 * dummynet changes the reference to the default rule (it could well be a
4853 * NULL pointer, but this way we do not need to check for the special
4854 * case, plus here he have info on the default behaviour).
4856 struct ip_fw *ip_fw_default_rule;
4859 * This procedure is only used to handle keepalives. It is invoked
4860 * every dyn_keepalive_period
4863 ipfw_tick(void * __unused unused)
4865 struct mbuf *m0, *m, *mnext, **mtailp;
4869 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0)
4873 * We make a chain of packets to go out here -- not deferring
4874 * until after we drop the IPFW dynamic rule lock would result
4875 * in a lock order reversal with the normal packet input -> ipfw
4881 for (i = 0 ; i < curr_dyn_buckets ; i++) {
4882 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) {
4883 if (q->dyn_type == O_LIMIT_PARENT)
4885 if (q->id.proto != IPPROTO_TCP)
4887 if ( (q->state & BOTH_SYN) != BOTH_SYN)
4889 if (TIME_LEQ( time_uptime+dyn_keepalive_interval,
4891 continue; /* too early */
4892 if (TIME_LEQ(q->expire, time_uptime))
4893 continue; /* too late, rule expired */
4895 *mtailp = send_pkt(NULL, &(q->id), q->ack_rev - 1,
4896 q->ack_fwd, TH_SYN);
4897 if (*mtailp != NULL)
4898 mtailp = &(*mtailp)->m_nextpkt;
4899 *mtailp = send_pkt(NULL, &(q->id), q->ack_fwd - 1,
4901 if (*mtailp != NULL)
4902 mtailp = &(*mtailp)->m_nextpkt;
4906 for (m = mnext = m0; m != NULL; m = mnext) {
4907 mnext = m->m_nextpkt;
4908 m->m_nextpkt = NULL;
4909 ip_output(m, NULL, NULL, 0, NULL, NULL);
4912 callout_reset(&ipfw_timeout, dyn_keepalive_period*hz, ipfw_tick, NULL);
4918 struct ip_fw default_rule;
4922 /* Setup IPv6 fw sysctl tree. */
4923 sysctl_ctx_init(&ip6_fw_sysctl_ctx);
4924 ip6_fw_sysctl_tree = SYSCTL_ADD_NODE(&ip6_fw_sysctl_ctx,
4925 SYSCTL_STATIC_CHILDREN(_net_inet6_ip6), OID_AUTO, "fw",
4926 CTLFLAG_RW | CTLFLAG_SECURE, 0, "Firewall");
4927 SYSCTL_ADD_PROC(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree),
4928 OID_AUTO, "enable", CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_SECURE3,
4929 &fw6_enable, 0, ipfw_chg_hook, "I", "Enable ipfw+6");
4930 SYSCTL_ADD_INT(&ip6_fw_sysctl_ctx, SYSCTL_CHILDREN(ip6_fw_sysctl_tree),
4931 OID_AUTO, "deny_unknown_exthdrs", CTLFLAG_RW | CTLFLAG_SECURE,
4932 &fw_deny_unknown_exthdrs, 0,
4933 "Deny packets with unknown IPv6 Extension Headers");
4936 layer3_chain.rules = NULL;
4937 IPFW_LOCK_INIT(&layer3_chain);
4938 ipfw_dyn_rule_zone = uma_zcreate("IPFW dynamic rule",
4939 sizeof(ipfw_dyn_rule), NULL, NULL, NULL, NULL,
4941 IPFW_DYN_LOCK_INIT();
4942 callout_init(&ipfw_timeout, NET_CALLOUT_MPSAFE);
4944 bzero(&default_rule, sizeof default_rule);
4946 default_rule.act_ofs = 0;
4947 default_rule.rulenum = IPFW_DEFAULT_RULE;
4948 default_rule.cmd_len = 1;
4949 default_rule.set = RESVD_SET;
4951 default_rule.cmd[0].len = 1;
4952 default_rule.cmd[0].opcode =
4953 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
4958 error = add_rule(&layer3_chain, &default_rule);
4960 printf("ipfw2: error %u initializing default rule "
4961 "(support disabled)\n", error);
4962 IPFW_DYN_LOCK_DESTROY();
4963 IPFW_LOCK_DESTROY(&layer3_chain);
4964 uma_zdestroy(ipfw_dyn_rule_zone);
4968 ip_fw_default_rule = layer3_chain.rules;
4973 "initialized, divert %s, "
4974 "rule-based forwarding "
4975 #ifdef IPFIREWALL_FORWARD
4980 "default to %s, logging ",
4986 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny");
4988 #ifdef IPFIREWALL_VERBOSE
4991 #ifdef IPFIREWALL_VERBOSE_LIMIT
4992 verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
4994 if (fw_verbose == 0)
4995 printf("disabled\n");
4996 else if (verbose_limit == 0)
4997 printf("unlimited\n");
4999 printf("limited to %d packets/entry by default\n",
5002 error = init_tables(&layer3_chain);
5004 IPFW_DYN_LOCK_DESTROY();
5005 IPFW_LOCK_DESTROY(&layer3_chain);
5006 uma_zdestroy(ipfw_dyn_rule_zone);
5009 ip_fw_ctl_ptr = ipfw_ctl;
5010 ip_fw_chk_ptr = ipfw_chk;
5011 callout_reset(&ipfw_timeout, hz, ipfw_tick, NULL);
5012 #ifdef IPFIREWALL_NAT
5013 LIST_INIT(&layer3_chain.nat);
5014 ifaddr_event_tag = EVENTHANDLER_REGISTER(ifaddr_event, ifaddr_change,
5015 NULL, EVENTHANDLER_PRI_ANY);
5024 #ifdef IPFIREWALL_NAT
5025 struct cfg_nat *ptr, *ptr_temp;
5028 ip_fw_chk_ptr = NULL;
5029 ip_fw_ctl_ptr = NULL;
5030 callout_drain(&ipfw_timeout);
5031 IPFW_WLOCK(&layer3_chain);
5032 flush_tables(&layer3_chain);
5033 #ifdef IPFIREWALL_NAT
5034 LIST_FOREACH_SAFE(ptr, &layer3_chain.nat, _next, ptr_temp) {
5035 LIST_REMOVE(ptr, _next);
5036 del_redir_spool_cfg(ptr, &ptr->redir_chain);
5037 LibAliasUninit(ptr->lib);
5040 EVENTHANDLER_DEREGISTER(ifaddr_event, ifaddr_event_tag);
5042 layer3_chain.reap = NULL;
5043 free_chain(&layer3_chain, 1 /* kill default rule */);
5044 reap = layer3_chain.reap, layer3_chain.reap = NULL;
5045 IPFW_WUNLOCK(&layer3_chain);
5048 IPFW_DYN_LOCK_DESTROY();
5049 uma_zdestroy(ipfw_dyn_rule_zone);
5050 IPFW_LOCK_DESTROY(&layer3_chain);
5053 /* Free IPv6 fw sysctl tree. */
5054 sysctl_ctx_free(&ip6_fw_sysctl_ctx);
5057 printf("IP firewall unloaded\n");