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
25 * $FreeBSD: src/sys/netinet/ip_fw2.c,v 1.6.2.12 2003/04/08 10:42:32 maxim Exp $
26 * $DragonFly: src/sys/net/ipfw/ip_fw2.c,v 1.27 2007/09/02 13:27:23 sephe Exp $
33 * Implement IP packet firewall (new version)
36 #if !defined(KLD_MODULE)
39 #include "opt_ipdivert.h"
42 #error IPFIREWALL requires INET.
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/malloc.h>
50 #include <sys/kernel.h>
52 #include <sys/socket.h>
53 #include <sys/socketvar.h>
54 #include <sys/sysctl.h>
55 #include <sys/syslog.h>
56 #include <sys/thread2.h>
57 #include <sys/ucred.h>
58 #include <sys/in_cksum.h>
60 #include <net/route.h>
61 #include <netinet/in.h>
62 #include <netinet/in_systm.h>
63 #include <netinet/in_var.h>
64 #include <netinet/in_pcb.h>
65 #include <netinet/ip.h>
66 #include <netinet/ip_var.h>
67 #include <netinet/ip_icmp.h>
69 #include <net/dummynet/ip_dummynet.h>
70 #include <netinet/tcp.h>
71 #include <netinet/tcp_timer.h>
72 #include <netinet/tcp_var.h>
73 #include <netinet/tcpip.h>
74 #include <netinet/udp.h>
75 #include <netinet/udp_var.h>
77 #include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */
80 * set_disable contains one bit per set value (0..31).
81 * If the bit is set, all rules with the corresponding set
82 * are disabled. Set 31 is reserved for the default rule
83 * and CANNOT be disabled.
85 static u_int32_t set_disable
;
87 static int fw_verbose
;
88 static int verbose_limit
;
90 static struct callout ipfw_timeout_h
;
91 #define IPFW_DEFAULT_RULE 65535
94 * list of rules for layer 3
96 static struct ip_fw
*layer3_chain
;
98 MALLOC_DEFINE(M_IPFW
, "IpFw/IpAcct", "IpFw/IpAcct chain's");
100 static int fw_debug
= 1;
101 static int autoinc_step
= 100; /* bounded to 1..1000 in add_rule() */
104 SYSCTL_NODE(_net_inet_ip
, OID_AUTO
, fw
, CTLFLAG_RW
, 0, "Firewall");
105 SYSCTL_INT(_net_inet_ip_fw
, OID_AUTO
, enable
, CTLFLAG_RW
,
106 &fw_enable
, 0, "Enable ipfw");
107 SYSCTL_INT(_net_inet_ip_fw
, OID_AUTO
, autoinc_step
, CTLFLAG_RW
,
108 &autoinc_step
, 0, "Rule number autincrement step");
109 SYSCTL_INT(_net_inet_ip_fw
, OID_AUTO
,one_pass
,CTLFLAG_RW
,
111 "Only do a single pass through ipfw when using dummynet(4)");
112 SYSCTL_INT(_net_inet_ip_fw
, OID_AUTO
, debug
, CTLFLAG_RW
,
113 &fw_debug
, 0, "Enable printing of debug ip_fw statements");
114 SYSCTL_INT(_net_inet_ip_fw
, OID_AUTO
, verbose
, CTLFLAG_RW
,
115 &fw_verbose
, 0, "Log matches to ipfw rules");
116 SYSCTL_INT(_net_inet_ip_fw
, OID_AUTO
, verbose_limit
, CTLFLAG_RW
,
117 &verbose_limit
, 0, "Set upper limit of matches of ipfw rules logged");
120 * Description of dynamic rules.
122 * Dynamic rules are stored in lists accessed through a hash table
123 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can
124 * be modified through the sysctl variable dyn_buckets which is
125 * updated when the table becomes empty.
127 * XXX currently there is only one list, ipfw_dyn.
129 * When a packet is received, its address fields are first masked
130 * with the mask defined for the rule, then hashed, then matched
131 * against the entries in the corresponding list.
132 * Dynamic rules can be used for different purposes:
134 * + enforcing limits on the number of sessions;
135 * + in-kernel NAT (not implemented yet)
137 * The lifetime of dynamic rules is regulated by dyn_*_lifetime,
138 * measured in seconds and depending on the flags.
140 * The total number of dynamic rules is stored in dyn_count.
141 * The max number of dynamic rules is dyn_max. When we reach
142 * the maximum number of rules we do not create anymore. This is
143 * done to avoid consuming too much memory, but also too much
144 * time when searching on each packet (ideally, we should try instead
145 * to put a limit on the length of the list on each bucket...).
147 * Each dynamic rule holds a pointer to the parent ipfw rule so
148 * we know what action to perform. Dynamic rules are removed when
149 * the parent rule is deleted. XXX we should make them survive.
151 * There are some limitations with dynamic rules -- we do not
152 * obey the 'randomized match', and we do not do multiple
153 * passes through the firewall. XXX check the latter!!!
155 static ipfw_dyn_rule
**ipfw_dyn_v
= NULL
;
156 static u_int32_t dyn_buckets
= 256; /* must be power of 2 */
157 static u_int32_t curr_dyn_buckets
= 256; /* must be power of 2 */
160 * Timeouts for various events in handing dynamic rules.
162 static u_int32_t dyn_ack_lifetime
= 300;
163 static u_int32_t dyn_syn_lifetime
= 20;
164 static u_int32_t dyn_fin_lifetime
= 1;
165 static u_int32_t dyn_rst_lifetime
= 1;
166 static u_int32_t dyn_udp_lifetime
= 10;
167 static u_int32_t dyn_short_lifetime
= 5;
170 * Keepalives are sent if dyn_keepalive is set. They are sent every
171 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval
172 * seconds of lifetime of a rule.
173 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower
174 * than dyn_keepalive_period.
177 static u_int32_t dyn_keepalive_interval
= 20;
178 static u_int32_t dyn_keepalive_period
= 5;
179 static u_int32_t dyn_keepalive
= 1; /* do send keepalives */
181 static u_int32_t static_count
; /* # of static rules */
182 static u_int32_t static_len
; /* size in bytes of static rules */
183 static u_int32_t dyn_count
; /* # of dynamic rules */
184 static u_int32_t dyn_max
= 4096; /* max # of dynamic rules */
186 SYSCTL_INT(_net_inet_ip_fw
, OID_AUTO
, dyn_buckets
, CTLFLAG_RW
,
187 &dyn_buckets
, 0, "Number of dyn. buckets");
188 SYSCTL_INT(_net_inet_ip_fw
, OID_AUTO
, curr_dyn_buckets
, CTLFLAG_RD
,
189 &curr_dyn_buckets
, 0, "Current Number of dyn. buckets");
190 SYSCTL_INT(_net_inet_ip_fw
, OID_AUTO
, dyn_count
, CTLFLAG_RD
,
191 &dyn_count
, 0, "Number of dyn. rules");
192 SYSCTL_INT(_net_inet_ip_fw
, OID_AUTO
, dyn_max
, CTLFLAG_RW
,
193 &dyn_max
, 0, "Max number of dyn. rules");
194 SYSCTL_INT(_net_inet_ip_fw
, OID_AUTO
, static_count
, CTLFLAG_RD
,
195 &static_count
, 0, "Number of static rules");
196 SYSCTL_INT(_net_inet_ip_fw
, OID_AUTO
, dyn_ack_lifetime
, CTLFLAG_RW
,
197 &dyn_ack_lifetime
, 0, "Lifetime of dyn. rules for acks");
198 SYSCTL_INT(_net_inet_ip_fw
, OID_AUTO
, dyn_syn_lifetime
, CTLFLAG_RW
,
199 &dyn_syn_lifetime
, 0, "Lifetime of dyn. rules for syn");
200 SYSCTL_INT(_net_inet_ip_fw
, OID_AUTO
, dyn_fin_lifetime
, CTLFLAG_RW
,
201 &dyn_fin_lifetime
, 0, "Lifetime of dyn. rules for fin");
202 SYSCTL_INT(_net_inet_ip_fw
, OID_AUTO
, dyn_rst_lifetime
, CTLFLAG_RW
,
203 &dyn_rst_lifetime
, 0, "Lifetime of dyn. rules for rst");
204 SYSCTL_INT(_net_inet_ip_fw
, OID_AUTO
, dyn_udp_lifetime
, CTLFLAG_RW
,
205 &dyn_udp_lifetime
, 0, "Lifetime of dyn. rules for UDP");
206 SYSCTL_INT(_net_inet_ip_fw
, OID_AUTO
, dyn_short_lifetime
, CTLFLAG_RW
,
207 &dyn_short_lifetime
, 0, "Lifetime of dyn. rules for other situations");
208 SYSCTL_INT(_net_inet_ip_fw
, OID_AUTO
, dyn_keepalive
, CTLFLAG_RW
,
209 &dyn_keepalive
, 0, "Enable keepalives for dyn. rules");
211 #endif /* SYSCTL_NODE */
214 static ip_fw_chk_t ipfw_chk
;
216 ip_dn_ruledel_t
*ip_dn_ruledel_ptr
= NULL
; /* hook into dummynet */
219 * This macro maps an ip pointer into a layer3 header pointer of type T
221 #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
224 icmptype_match(struct ip
*ip
, ipfw_insn_u32
*cmd
)
226 int type
= L3HDR(struct icmp
,ip
)->icmp_type
;
228 return (type
<= ICMP_MAXTYPE
&& (cmd
->d
[0] & (1<<type
)) );
231 #define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
232 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
235 is_icmp_query(struct ip
*ip
)
237 int type
= L3HDR(struct icmp
, ip
)->icmp_type
;
238 return (type
<= ICMP_MAXTYPE
&& (TT
& (1<<type
)) );
243 * The following checks use two arrays of 8 or 16 bits to store the
244 * bits that we want set or clear, respectively. They are in the
245 * low and high half of cmd->arg1 or cmd->d[0].
247 * We scan options and store the bits we find set. We succeed if
249 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
251 * The code is sometimes optimized not to store additional variables.
255 flags_match(ipfw_insn
*cmd
, u_int8_t bits
)
260 if ( ((cmd
->arg1
& 0xff) & bits
) != 0)
261 return 0; /* some bits we want set were clear */
262 want_clear
= (cmd
->arg1
>> 8) & 0xff;
263 if ( (want_clear
& bits
) != want_clear
)
264 return 0; /* some bits we want clear were set */
269 ipopts_match(struct ip
*ip
, ipfw_insn
*cmd
)
271 int optlen
, bits
= 0;
272 u_char
*cp
= (u_char
*)(ip
+ 1);
273 int x
= (ip
->ip_hl
<< 2) - sizeof (struct ip
);
275 for (; x
> 0; x
-= optlen
, cp
+= optlen
) {
276 int opt
= cp
[IPOPT_OPTVAL
];
278 if (opt
== IPOPT_EOL
)
280 if (opt
== IPOPT_NOP
)
283 optlen
= cp
[IPOPT_OLEN
];
284 if (optlen
<= 0 || optlen
> x
)
285 return 0; /* invalid or truncated */
293 bits
|= IP_FW_IPOPT_LSRR
;
297 bits
|= IP_FW_IPOPT_SSRR
;
301 bits
|= IP_FW_IPOPT_RR
;
305 bits
|= IP_FW_IPOPT_TS
;
309 return (flags_match(cmd
, bits
));
313 tcpopts_match(struct ip
*ip
, ipfw_insn
*cmd
)
315 int optlen
, bits
= 0;
316 struct tcphdr
*tcp
= L3HDR(struct tcphdr
,ip
);
317 u_char
*cp
= (u_char
*)(tcp
+ 1);
318 int x
= (tcp
->th_off
<< 2) - sizeof(struct tcphdr
);
320 for (; x
> 0; x
-= optlen
, cp
+= optlen
) {
322 if (opt
== TCPOPT_EOL
)
324 if (opt
== TCPOPT_NOP
)
338 bits
|= IP_FW_TCPOPT_MSS
;
342 bits
|= IP_FW_TCPOPT_WINDOW
;
345 case TCPOPT_SACK_PERMITTED
:
347 bits
|= IP_FW_TCPOPT_SACK
;
350 case TCPOPT_TIMESTAMP
:
351 bits
|= IP_FW_TCPOPT_TS
;
357 bits
|= IP_FW_TCPOPT_CC
;
361 return (flags_match(cmd
, bits
));
365 iface_match(struct ifnet
*ifp
, ipfw_insn_if
*cmd
)
367 if (ifp
== NULL
) /* no iface with this packet, match fails */
369 /* Check by name or by IP address */
370 if (cmd
->name
[0] != '\0') { /* match by name */
373 if (kfnmatch(cmd
->name
, ifp
->if_xname
, 0) == 0)
376 if (strncmp(ifp
->if_xname
, cmd
->name
, IFNAMSIZ
) == 0)
382 TAILQ_FOREACH(ia
, &ifp
->if_addrhead
, ifa_link
) {
383 if (ia
->ifa_addr
== NULL
)
385 if (ia
->ifa_addr
->sa_family
!= AF_INET
)
387 if (cmd
->p
.ip
.s_addr
== ((struct sockaddr_in
*)
388 (ia
->ifa_addr
))->sin_addr
.s_addr
)
389 return(1); /* match */
392 return(0); /* no match, fail ... */
395 static u_int64_t norule_counter
; /* counter for ipfw_log(NULL...) */
397 #define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0
398 #define SNP(buf) buf, sizeof(buf)
401 * We enter here when we have a rule with O_LOG.
402 * XXX this function alone takes about 2Kbytes of code!
405 ipfw_log(struct ip_fw
*f
, u_int hlen
, struct ether_header
*eh
,
406 struct mbuf
*m
, struct ifnet
*oif
)
409 int limit_reached
= 0;
410 char action2
[40], proto
[48], fragment
[28];
415 if (f
== NULL
) { /* bogus pkt */
416 if (verbose_limit
!= 0 && norule_counter
>= verbose_limit
)
419 if (norule_counter
== verbose_limit
)
420 limit_reached
= verbose_limit
;
422 } else { /* O_LOG is the first action, find the real one */
423 ipfw_insn
*cmd
= ACTION_PTR(f
);
424 ipfw_insn_log
*l
= (ipfw_insn_log
*)cmd
;
426 if (l
->max_log
!= 0 && l
->log_left
== 0)
429 if (l
->log_left
== 0)
430 limit_reached
= l
->max_log
;
431 cmd
+= F_LEN(cmd
); /* point to first action */
432 if (cmd
->opcode
== O_PROB
)
436 switch (cmd
->opcode
) {
442 if (cmd
->arg1
==ICMP_REJECT_RST
)
444 else if (cmd
->arg1
==ICMP_UNREACH_HOST
)
447 ksnprintf(SNPARGS(action2
, 0), "Unreach %d",
458 ksnprintf(SNPARGS(action2
, 0), "Divert %d",
462 ksnprintf(SNPARGS(action2
, 0), "Tee %d",
466 ksnprintf(SNPARGS(action2
, 0), "SkipTo %d",
470 ksnprintf(SNPARGS(action2
, 0), "Pipe %d",
474 ksnprintf(SNPARGS(action2
, 0), "Queue %d",
478 ipfw_insn_sa
*sa
= (ipfw_insn_sa
*)cmd
;
481 len
= ksnprintf(SNPARGS(action2
, 0), "Forward to %s",
482 inet_ntoa(sa
->sa
.sin_addr
));
484 ksnprintf(SNPARGS(action2
, len
), ":%d",
494 if (hlen
== 0) { /* non-ip */
495 ksnprintf(SNPARGS(proto
, 0), "MAC");
497 struct ip
*ip
= mtod(m
, struct ip
*);
498 /* these three are all aliases to the same thing */
499 struct icmp
*const icmp
= L3HDR(struct icmp
, ip
);
500 struct tcphdr
*const tcp
= (struct tcphdr
*)icmp
;
501 struct udphdr
*const udp
= (struct udphdr
*)icmp
;
503 int ip_off
, offset
, ip_len
;
507 if (eh
!= NULL
) { /* layer 2 packets are as on the wire */
508 ip_off
= ntohs(ip
->ip_off
);
509 ip_len
= ntohs(ip
->ip_len
);
514 offset
= ip_off
& IP_OFFMASK
;
517 len
= ksnprintf(SNPARGS(proto
, 0), "TCP %s",
518 inet_ntoa(ip
->ip_src
));
520 ksnprintf(SNPARGS(proto
, len
), ":%d %s:%d",
521 ntohs(tcp
->th_sport
),
522 inet_ntoa(ip
->ip_dst
),
523 ntohs(tcp
->th_dport
));
525 ksnprintf(SNPARGS(proto
, len
), " %s",
526 inet_ntoa(ip
->ip_dst
));
530 len
= ksnprintf(SNPARGS(proto
, 0), "UDP %s",
531 inet_ntoa(ip
->ip_src
));
533 ksnprintf(SNPARGS(proto
, len
), ":%d %s:%d",
534 ntohs(udp
->uh_sport
),
535 inet_ntoa(ip
->ip_dst
),
536 ntohs(udp
->uh_dport
));
538 ksnprintf(SNPARGS(proto
, len
), " %s",
539 inet_ntoa(ip
->ip_dst
));
544 len
= ksnprintf(SNPARGS(proto
, 0),
546 icmp
->icmp_type
, icmp
->icmp_code
);
548 len
= ksnprintf(SNPARGS(proto
, 0), "ICMP ");
549 len
+= ksnprintf(SNPARGS(proto
, len
), "%s",
550 inet_ntoa(ip
->ip_src
));
551 ksnprintf(SNPARGS(proto
, len
), " %s",
552 inet_ntoa(ip
->ip_dst
));
556 len
= ksnprintf(SNPARGS(proto
, 0), "P:%d %s", ip
->ip_p
,
557 inet_ntoa(ip
->ip_src
));
558 ksnprintf(SNPARGS(proto
, len
), " %s",
559 inet_ntoa(ip
->ip_dst
));
563 if (ip_off
& (IP_MF
| IP_OFFMASK
))
564 ksnprintf(SNPARGS(fragment
, 0), " (frag %d:%d@%d%s)",
565 ntohs(ip
->ip_id
), ip_len
- (ip
->ip_hl
<< 2),
567 (ip_off
& IP_MF
) ? "+" : "");
569 if (oif
|| m
->m_pkthdr
.rcvif
)
570 log(LOG_SECURITY
| LOG_INFO
,
571 "ipfw: %d %s %s %s via %s%s\n",
573 action
, proto
, oif
? "out" : "in",
574 oif
? oif
->if_xname
: m
->m_pkthdr
.rcvif
->if_xname
,
577 log(LOG_SECURITY
| LOG_INFO
,
578 "ipfw: %d %s %s [no if info]%s\n",
580 action
, proto
, fragment
);
582 log(LOG_SECURITY
| LOG_NOTICE
,
583 "ipfw: limit %d reached on entry %d\n",
584 limit_reached
, f
? f
->rulenum
: -1);
588 * IMPORTANT: the hash function for dynamic rules must be commutative
589 * in source and destination (ip,port), because rules are bidirectional
590 * and we want to find both in the same bucket.
593 hash_packet(struct ipfw_flow_id
*id
)
597 i
= (id
->dst_ip
) ^ (id
->src_ip
) ^ (id
->dst_port
) ^ (id
->src_port
);
598 i
&= (curr_dyn_buckets
- 1);
603 * unlink a dynamic rule from a chain. prev is a pointer to
604 * the previous one, q is a pointer to the rule to delete,
605 * head is a pointer to the head of the queue.
606 * Modifies q and potentially also head.
608 #define UNLINK_DYN_RULE(prev, head, q) { \
609 ipfw_dyn_rule *old_q = q; \
611 /* remove a refcount to the parent */ \
612 if (q->dyn_type == O_LIMIT) \
613 q->parent->count--; \
614 DEB(kprintf("-- unlink entry 0x%08x %d -> 0x%08x %d, %d left\n", \
615 (q->id.src_ip), (q->id.src_port), \
616 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \
618 prev->next = q = q->next; \
620 head = q = q->next; \
622 kfree(old_q, M_IPFW); }
624 #define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0)
627 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL.
629 * If keep_me == NULL, rules are deleted even if not expired,
630 * otherwise only expired rules are removed.
632 * The value of the second parameter is also used to point to identify
633 * a rule we absolutely do not want to remove (e.g. because we are
634 * holding a reference to it -- this is the case with O_LIMIT_PARENT
635 * rules). The pointer is only used for comparison, so any non-null
639 remove_dyn_rule(struct ip_fw
*rule
, ipfw_dyn_rule
*keep_me
)
641 static u_int32_t last_remove
= 0;
643 #define FORCE (keep_me == NULL)
645 ipfw_dyn_rule
*prev
, *q
;
646 int i
, pass
= 0, max_pass
= 0;
648 if (ipfw_dyn_v
== NULL
|| dyn_count
== 0)
650 /* do not expire more than once per second, it is useless */
651 if (!FORCE
&& last_remove
== time_second
)
653 last_remove
= time_second
;
656 * because O_LIMIT refer to parent rules, during the first pass only
657 * remove child and mark any pending LIMIT_PARENT, and remove
658 * them in a second pass.
661 for (i
= 0 ; i
< curr_dyn_buckets
; i
++) {
662 for (prev
=NULL
, q
= ipfw_dyn_v
[i
] ; q
; ) {
664 * Logic can become complex here, so we split tests.
668 if (rule
!= NULL
&& rule
!= q
->rule
)
669 goto next
; /* not the one we are looking for */
670 if (q
->dyn_type
== O_LIMIT_PARENT
) {
672 * handle parent in the second pass,
673 * record we need one.
678 if (FORCE
&& q
->count
!= 0 ) {
679 /* XXX should not happen! */
680 kprintf( "OUCH! cannot remove rule,"
681 " count %d\n", q
->count
);
685 !TIME_LEQ( q
->expire
, time_second
))
688 UNLINK_DYN_RULE(prev
, ipfw_dyn_v
[i
], q
);
695 if (pass
++ < max_pass
)
701 * lookup a dynamic rule.
703 static ipfw_dyn_rule
*
704 lookup_dyn_rule(struct ipfw_flow_id
*pkt
, int *match_direction
,
708 * stateful ipfw extensions.
709 * Lookup into dynamic session queue
711 #define MATCH_REVERSE 0
712 #define MATCH_FORWARD 1
714 #define MATCH_UNKNOWN 3
715 int i
, dir
= MATCH_NONE
;
716 ipfw_dyn_rule
*prev
, *q
=NULL
;
718 if (ipfw_dyn_v
== NULL
)
719 goto done
; /* not found */
720 i
= hash_packet( pkt
);
721 for (prev
=NULL
, q
= ipfw_dyn_v
[i
] ; q
!= NULL
; ) {
722 if (q
->dyn_type
== O_LIMIT_PARENT
)
724 if (TIME_LEQ( q
->expire
, time_second
)) { /* expire entry */
725 UNLINK_DYN_RULE(prev
, ipfw_dyn_v
[i
], q
);
728 if ( pkt
->proto
== q
->id
.proto
) {
729 if (pkt
->src_ip
== q
->id
.src_ip
&&
730 pkt
->dst_ip
== q
->id
.dst_ip
&&
731 pkt
->src_port
== q
->id
.src_port
&&
732 pkt
->dst_port
== q
->id
.dst_port
) {
736 if (pkt
->src_ip
== q
->id
.dst_ip
&&
737 pkt
->dst_ip
== q
->id
.src_ip
&&
738 pkt
->src_port
== q
->id
.dst_port
&&
739 pkt
->dst_port
== q
->id
.src_port
) {
749 goto done
; /* q = NULL, not found */
751 if ( prev
!= NULL
) { /* found and not in front */
752 prev
->next
= q
->next
;
753 q
->next
= ipfw_dyn_v
[i
];
756 if (pkt
->proto
== IPPROTO_TCP
) { /* update state according to flags */
757 u_char flags
= pkt
->flags
& (TH_FIN
|TH_SYN
|TH_RST
);
759 #define BOTH_SYN (TH_SYN | (TH_SYN << 8))
760 #define BOTH_FIN (TH_FIN | (TH_FIN << 8))
761 q
->state
|= (dir
== MATCH_FORWARD
) ? flags
: (flags
<< 8);
763 case TH_SYN
: /* opening */
764 q
->expire
= time_second
+ dyn_syn_lifetime
;
767 case BOTH_SYN
: /* move to established */
768 case BOTH_SYN
| TH_FIN
: /* one side tries to close */
769 case BOTH_SYN
| (TH_FIN
<< 8) :
771 #define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0)
772 u_int32_t ack
= ntohl(tcp
->th_ack
);
773 if (dir
== MATCH_FORWARD
) {
774 if (q
->ack_fwd
== 0 || _SEQ_GE(ack
, q
->ack_fwd
))
776 else { /* ignore out-of-sequence */
780 if (q
->ack_rev
== 0 || _SEQ_GE(ack
, q
->ack_rev
))
782 else { /* ignore out-of-sequence */
787 q
->expire
= time_second
+ dyn_ack_lifetime
;
790 case BOTH_SYN
| BOTH_FIN
: /* both sides closed */
791 if (dyn_fin_lifetime
>= dyn_keepalive_period
)
792 dyn_fin_lifetime
= dyn_keepalive_period
- 1;
793 q
->expire
= time_second
+ dyn_fin_lifetime
;
799 * reset or some invalid combination, but can also
800 * occur if we use keep-state the wrong way.
802 if ( (q
->state
& ((TH_RST
<< 8)|TH_RST
)) == 0)
803 kprintf("invalid state: 0x%x\n", q
->state
);
805 if (dyn_rst_lifetime
>= dyn_keepalive_period
)
806 dyn_rst_lifetime
= dyn_keepalive_period
- 1;
807 q
->expire
= time_second
+ dyn_rst_lifetime
;
810 } else if (pkt
->proto
== IPPROTO_UDP
) {
811 q
->expire
= time_second
+ dyn_udp_lifetime
;
813 /* other protocols */
814 q
->expire
= time_second
+ dyn_short_lifetime
;
818 *match_direction
= dir
;
823 realloc_dynamic_table(void)
826 * Try reallocation, make sure we have a power of 2 and do
827 * not allow more than 64k entries. In case of overflow,
831 if (dyn_buckets
> 65536)
833 if ((dyn_buckets
& (dyn_buckets
-1)) != 0) { /* not a power of 2 */
834 dyn_buckets
= curr_dyn_buckets
; /* reset */
837 curr_dyn_buckets
= dyn_buckets
;
838 if (ipfw_dyn_v
!= NULL
)
839 kfree(ipfw_dyn_v
, M_IPFW
);
841 ipfw_dyn_v
= kmalloc(curr_dyn_buckets
* sizeof(ipfw_dyn_rule
*),
842 M_IPFW
, M_WAITOK
| M_ZERO
);
843 if (ipfw_dyn_v
!= NULL
|| curr_dyn_buckets
<= 2)
845 curr_dyn_buckets
/= 2;
850 * Install state of type 'type' for a dynamic session.
851 * The hash table contains two type of rules:
852 * - regular rules (O_KEEP_STATE)
853 * - rules for sessions with limited number of sess per user
854 * (O_LIMIT). When they are created, the parent is
855 * increased by 1, and decreased on delete. In this case,
856 * the third parameter is the parent rule and not the chain.
857 * - "parent" rules for the above (O_LIMIT_PARENT).
859 static ipfw_dyn_rule
*
860 add_dyn_rule(struct ipfw_flow_id
*id
, u_int8_t dyn_type
, struct ip_fw
*rule
)
865 if (ipfw_dyn_v
== NULL
||
866 (dyn_count
== 0 && dyn_buckets
!= curr_dyn_buckets
)) {
867 realloc_dynamic_table();
868 if (ipfw_dyn_v
== NULL
)
869 return NULL
; /* failed ! */
873 r
= kmalloc(sizeof *r
, M_IPFW
, M_WAITOK
| M_ZERO
);
875 kprintf ("sorry cannot allocate state\n");
879 /* increase refcount on parent, and set pointer */
880 if (dyn_type
== O_LIMIT
) {
881 ipfw_dyn_rule
*parent
= (ipfw_dyn_rule
*)rule
;
882 if ( parent
->dyn_type
!= O_LIMIT_PARENT
)
883 panic("invalid parent");
890 r
->expire
= time_second
+ dyn_syn_lifetime
;
892 r
->dyn_type
= dyn_type
;
893 r
->pcnt
= r
->bcnt
= 0;
897 r
->next
= ipfw_dyn_v
[i
];
900 DEB(kprintf("-- add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n",
902 (r
->id
.src_ip
), (r
->id
.src_port
),
903 (r
->id
.dst_ip
), (r
->id
.dst_port
),
909 * lookup dynamic parent rule using pkt and rule as search keys.
910 * If the lookup fails, then install one.
912 static ipfw_dyn_rule
*
913 lookup_dyn_parent(struct ipfw_flow_id
*pkt
, struct ip_fw
*rule
)
919 i
= hash_packet( pkt
);
920 for (q
= ipfw_dyn_v
[i
] ; q
!= NULL
; q
=q
->next
)
921 if (q
->dyn_type
== O_LIMIT_PARENT
&&
923 pkt
->proto
== q
->id
.proto
&&
924 pkt
->src_ip
== q
->id
.src_ip
&&
925 pkt
->dst_ip
== q
->id
.dst_ip
&&
926 pkt
->src_port
== q
->id
.src_port
&&
927 pkt
->dst_port
== q
->id
.dst_port
) {
928 q
->expire
= time_second
+ dyn_short_lifetime
;
929 DEB(kprintf("lookup_dyn_parent found 0x%p\n",q
);)
933 return add_dyn_rule(pkt
, O_LIMIT_PARENT
, rule
);
937 * Install dynamic state for rule type cmd->o.opcode
939 * Returns 1 (failure) if state is not installed because of errors or because
940 * session limitations are enforced.
943 install_state(struct ip_fw
*rule
, ipfw_insn_limit
*cmd
,
944 struct ip_fw_args
*args
)
950 DEB(kprintf("-- install state type %d 0x%08x %u -> 0x%08x %u\n",
952 (args
->f_id
.src_ip
), (args
->f_id
.src_port
),
953 (args
->f_id
.dst_ip
), (args
->f_id
.dst_port
) );)
955 q
= lookup_dyn_rule(&args
->f_id
, NULL
, NULL
);
957 if (q
!= NULL
) { /* should never occur */
958 if (last_log
!= time_second
) {
959 last_log
= time_second
;
960 kprintf(" install_state: entry already present, done\n");
965 if (dyn_count
>= dyn_max
)
967 * Run out of slots, try to remove any expired rule.
969 remove_dyn_rule(NULL
, (ipfw_dyn_rule
*)1);
971 if (dyn_count
>= dyn_max
) {
972 if (last_log
!= time_second
) {
973 last_log
= time_second
;
974 kprintf("install_state: Too many dynamic rules\n");
976 return 1; /* cannot install, notify caller */
979 switch (cmd
->o
.opcode
) {
980 case O_KEEP_STATE
: /* bidir rule */
981 add_dyn_rule(&args
->f_id
, O_KEEP_STATE
, rule
);
984 case O_LIMIT
: /* limit number of sessions */
986 u_int16_t limit_mask
= cmd
->limit_mask
;
987 struct ipfw_flow_id id
;
988 ipfw_dyn_rule
*parent
;
990 DEB(kprintf("installing dyn-limit rule %d\n", cmd
->conn_limit
);)
992 id
.dst_ip
= id
.src_ip
= 0;
993 id
.dst_port
= id
.src_port
= 0;
994 id
.proto
= args
->f_id
.proto
;
996 if (limit_mask
& DYN_SRC_ADDR
)
997 id
.src_ip
= args
->f_id
.src_ip
;
998 if (limit_mask
& DYN_DST_ADDR
)
999 id
.dst_ip
= args
->f_id
.dst_ip
;
1000 if (limit_mask
& DYN_SRC_PORT
)
1001 id
.src_port
= args
->f_id
.src_port
;
1002 if (limit_mask
& DYN_DST_PORT
)
1003 id
.dst_port
= args
->f_id
.dst_port
;
1004 parent
= lookup_dyn_parent(&id
, rule
);
1005 if (parent
== NULL
) {
1006 kprintf("add parent failed\n");
1009 if (parent
->count
>= cmd
->conn_limit
) {
1011 * See if we can remove some expired rule.
1013 remove_dyn_rule(rule
, parent
);
1014 if (parent
->count
>= cmd
->conn_limit
) {
1015 if (fw_verbose
&& last_log
!= time_second
) {
1016 last_log
= time_second
;
1017 log(LOG_SECURITY
| LOG_DEBUG
,
1018 "drop session, too many entries\n");
1023 add_dyn_rule(&args
->f_id
, O_LIMIT
, (struct ip_fw
*)parent
);
1027 kprintf("unknown dynamic rule type %u\n", cmd
->o
.opcode
);
1030 lookup_dyn_rule(&args
->f_id
, NULL
, NULL
); /* XXX just set lifetime */
1035 * Transmit a TCP packet, containing either a RST or a keepalive.
1036 * When flags & TH_RST, we are sending a RST packet, because of a
1037 * "reset" action matched the packet.
1038 * Otherwise we are sending a keepalive, and flags & TH_
1041 send_pkt(struct ipfw_flow_id
*id
, u_int32_t seq
, u_int32_t ack
, int flags
)
1046 struct route sro
; /* fake route */
1048 MGETHDR(m
, MB_DONTWAIT
, MT_HEADER
);
1051 m
->m_pkthdr
.rcvif
= (struct ifnet
*)0;
1052 m
->m_pkthdr
.len
= m
->m_len
= sizeof(struct ip
) + sizeof(struct tcphdr
);
1053 m
->m_data
+= max_linkhdr
;
1055 ip
= mtod(m
, struct ip
*);
1056 bzero(ip
, m
->m_len
);
1057 tcp
= (struct tcphdr
*)(ip
+ 1); /* no IP options */
1058 ip
->ip_p
= IPPROTO_TCP
;
1061 * Assume we are sending a RST (or a keepalive in the reverse
1062 * direction), swap src and destination addresses and ports.
1064 ip
->ip_src
.s_addr
= htonl(id
->dst_ip
);
1065 ip
->ip_dst
.s_addr
= htonl(id
->src_ip
);
1066 tcp
->th_sport
= htons(id
->dst_port
);
1067 tcp
->th_dport
= htons(id
->src_port
);
1068 if (flags
& TH_RST
) { /* we are sending a RST */
1069 if (flags
& TH_ACK
) {
1070 tcp
->th_seq
= htonl(ack
);
1071 tcp
->th_ack
= htonl(0);
1072 tcp
->th_flags
= TH_RST
;
1076 tcp
->th_seq
= htonl(0);
1077 tcp
->th_ack
= htonl(seq
);
1078 tcp
->th_flags
= TH_RST
| TH_ACK
;
1082 * We are sending a keepalive. flags & TH_SYN determines
1083 * the direction, forward if set, reverse if clear.
1084 * NOTE: seq and ack are always assumed to be correct
1085 * as set by the caller. This may be confusing...
1087 if (flags
& TH_SYN
) {
1089 * we have to rewrite the correct addresses!
1091 ip
->ip_dst
.s_addr
= htonl(id
->dst_ip
);
1092 ip
->ip_src
.s_addr
= htonl(id
->src_ip
);
1093 tcp
->th_dport
= htons(id
->dst_port
);
1094 tcp
->th_sport
= htons(id
->src_port
);
1096 tcp
->th_seq
= htonl(seq
);
1097 tcp
->th_ack
= htonl(ack
);
1098 tcp
->th_flags
= TH_ACK
;
1101 * set ip_len to the payload size so we can compute
1102 * the tcp checksum on the pseudoheader
1103 * XXX check this, could save a couple of words ?
1105 ip
->ip_len
= htons(sizeof(struct tcphdr
));
1106 tcp
->th_sum
= in_cksum(m
, m
->m_pkthdr
.len
);
1108 * now fill fields left out earlier
1110 ip
->ip_ttl
= ip_defttl
;
1111 ip
->ip_len
= m
->m_pkthdr
.len
;
1112 bzero (&sro
, sizeof (sro
));
1113 ip_rtaddr(ip
->ip_dst
, &sro
);
1114 m
->m_pkthdr
.fw_flags
|= IPFW_MBUF_SKIP_FIREWALL
;
1115 ip_output(m
, NULL
, &sro
, 0, NULL
, NULL
);
1121 * sends a reject message, consuming the mbuf passed as an argument.
1124 send_reject(struct ip_fw_args
*args
, int code
, int offset
, int ip_len
)
1127 if (code
!= ICMP_REJECT_RST
) { /* Send an ICMP unreach */
1128 /* We need the IP header in host order for icmp_error(). */
1129 if (args
->eh
!= NULL
) {
1130 struct ip
*ip
= mtod(args
->m
, struct ip
*);
1131 ip
->ip_len
= ntohs(ip
->ip_len
);
1132 ip
->ip_off
= ntohs(ip
->ip_off
);
1134 icmp_error(args
->m
, ICMP_UNREACH
, code
, 0L, 0);
1135 } else if (offset
== 0 && args
->f_id
.proto
== IPPROTO_TCP
) {
1136 struct tcphdr
*const tcp
=
1137 L3HDR(struct tcphdr
, mtod(args
->m
, struct ip
*));
1138 if ( (tcp
->th_flags
& TH_RST
) == 0)
1139 send_pkt(&(args
->f_id
), ntohl(tcp
->th_seq
),
1141 tcp
->th_flags
| TH_RST
);
1150 * Given an ip_fw *, lookup_next_rule will return a pointer
1151 * to the next rule, which can be either the jump
1152 * target (for skipto instructions) or the next one in the list (in
1153 * all other cases including a missing jump target).
1154 * The result is also written in the "next_rule" field of the rule.
1155 * Backward jumps are not allowed, so start looking from the next
1158 * This never returns NULL -- in case we do not have an exact match,
1159 * the next rule is returned. When the ruleset is changed,
1160 * pointers are flushed so we are always correct.
1163 static struct ip_fw
*
1164 lookup_next_rule(struct ip_fw
*me
)
1166 struct ip_fw
*rule
= NULL
;
1169 /* look for action, in case it is a skipto */
1170 cmd
= ACTION_PTR(me
);
1171 if (cmd
->opcode
== O_LOG
)
1173 if ( cmd
->opcode
== O_SKIPTO
)
1174 for (rule
= me
->next
; rule
; rule
= rule
->next
)
1175 if (rule
->rulenum
>= cmd
->arg1
)
1177 if (rule
== NULL
) /* failure or not a skipto */
1179 me
->next_rule
= rule
;
1184 * The main check routine for the firewall.
1186 * All arguments are in args so we can modify them and return them
1187 * back to the caller.
1191 * args->m (in/out) The packet; we set to NULL when/if we nuke it.
1192 * Starts with the IP header.
1193 * args->eh (in) Mac header if present, or NULL for layer3 packet.
1194 * args->oif Outgoing interface, or NULL if packet is incoming.
1195 * The incoming interface is in the mbuf. (in)
1197 * args->rule Pointer to the last matching rule (in/out)
1198 * args->next_hop Socket we are forwarding to (out).
1199 * args->f_id Addresses grabbed from the packet (out)
1203 * IP_FW_PORT_DENY_FLAG the packet must be dropped.
1204 * 0 The packet is to be accepted and routed normally OR
1205 * the packet was denied/rejected and has been dropped;
1206 * in the latter case, *m is equal to NULL upon return.
1207 * port Divert the packet to port, with these caveats:
1209 * - If IP_FW_PORT_TEE_FLAG is set, tee the packet instead
1210 * of diverting it (ie, 'ipfw tee').
1212 * - If IP_FW_PORT_DYNT_FLAG is set, interpret the lower
1213 * 16 bits as a dummynet pipe number instead of diverting
1217 ipfw_chk(struct ip_fw_args
*args
)
1220 * Local variables hold state during the processing of a packet.
1222 * IMPORTANT NOTE: to speed up the processing of rules, there
1223 * are some assumption on the values of the variables, which
1224 * are documented here. Should you change them, please check
1225 * the implementation of the various instructions to make sure
1226 * that they still work.
1228 * args->eh The MAC header. It is non-null for a layer2
1229 * packet, it is NULL for a layer-3 packet.
1231 * m | args->m Pointer to the mbuf, as received from the caller.
1232 * It may change if ipfw_chk() does an m_pullup, or if it
1233 * consumes the packet because it calls send_reject().
1234 * XXX This has to change, so that ipfw_chk() never modifies
1235 * or consumes the buffer.
1236 * ip is simply an alias of the value of m, and it is kept
1237 * in sync with it (the packet is supposed to start with
1240 struct mbuf
*m
= args
->m
;
1241 struct ip
*ip
= mtod(m
, struct ip
*);
1244 * oif | args->oif If NULL, ipfw_chk has been called on the
1245 * inbound path (ether_input, ip_input).
1246 * If non-NULL, ipfw_chk has been called on the outbound path
1247 * (ether_output, ip_output).
1249 struct ifnet
*oif
= args
->oif
;
1251 struct ip_fw
*f
= NULL
; /* matching rule */
1256 * hlen The length of the IPv4 header.
1257 * hlen >0 means we have an IPv4 packet.
1259 u_int hlen
= 0; /* hlen >0 means we have an IP pkt */
1262 * offset The offset of a fragment. offset != 0 means that
1263 * we have a fragment at this offset of an IPv4 packet.
1264 * offset == 0 means that (if this is an IPv4 packet)
1265 * this is the first or only fragment.
1270 * Local copies of addresses. They are only valid if we have
1273 * proto The protocol. Set to 0 for non-ip packets,
1274 * or to the protocol read from the packet otherwise.
1275 * proto != 0 means that we have an IPv4 packet.
1277 * src_port, dst_port port numbers, in HOST format. Only
1278 * valid for TCP and UDP packets.
1280 * src_ip, dst_ip ip addresses, in NETWORK format.
1281 * Only valid for IPv4 packets.
1284 u_int16_t src_port
= 0, dst_port
= 0; /* NOTE: host format */
1285 struct in_addr src_ip
, dst_ip
; /* NOTE: network format */
1287 int dyn_dir
= MATCH_UNKNOWN
;
1288 ipfw_dyn_rule
*q
= NULL
;
1290 if (m
->m_pkthdr
.fw_flags
& IPFW_MBUF_SKIP_FIREWALL
)
1291 return 0; /* accept */
1293 * dyn_dir = MATCH_UNKNOWN when rules unchecked,
1294 * MATCH_NONE when checked and not matched (q = NULL),
1295 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
1298 if (args
->eh
== NULL
|| /* layer 3 packet */
1299 ( m
->m_pkthdr
.len
>= sizeof(struct ip
) &&
1300 ntohs(args
->eh
->ether_type
) == ETHERTYPE_IP
))
1301 hlen
= ip
->ip_hl
<< 2;
1304 * Collect parameters into local variables for faster matching.
1306 if (hlen
== 0) { /* do not grab addresses for non-ip pkts */
1307 proto
= args
->f_id
.proto
= 0; /* mark f_id invalid */
1308 goto after_ip_checks
;
1311 proto
= args
->f_id
.proto
= ip
->ip_p
;
1312 src_ip
= ip
->ip_src
;
1313 dst_ip
= ip
->ip_dst
;
1314 if (args
->eh
!= NULL
) { /* layer 2 packets are as on the wire */
1315 offset
= ntohs(ip
->ip_off
) & IP_OFFMASK
;
1316 ip_len
= ntohs(ip
->ip_len
);
1318 offset
= ip
->ip_off
& IP_OFFMASK
;
1319 ip_len
= ip
->ip_len
;
1322 #define PULLUP_TO(len) \
1324 if ((m)->m_len < (len)) { \
1325 args->m = m = m_pullup(m, (len)); \
1327 goto pullup_failed; \
1328 ip = mtod(m, struct ip *); \
1338 PULLUP_TO(hlen
+ sizeof(struct tcphdr
));
1339 tcp
= L3HDR(struct tcphdr
, ip
);
1340 dst_port
= tcp
->th_dport
;
1341 src_port
= tcp
->th_sport
;
1342 args
->f_id
.flags
= tcp
->th_flags
;
1350 PULLUP_TO(hlen
+ sizeof(struct udphdr
));
1351 udp
= L3HDR(struct udphdr
, ip
);
1352 dst_port
= udp
->uh_dport
;
1353 src_port
= udp
->uh_sport
;
1358 PULLUP_TO(hlen
+ 4); /* type, code and checksum. */
1359 args
->f_id
.flags
= L3HDR(struct icmp
, ip
)->icmp_type
;
1368 args
->f_id
.src_ip
= ntohl(src_ip
.s_addr
);
1369 args
->f_id
.dst_ip
= ntohl(dst_ip
.s_addr
);
1370 args
->f_id
.src_port
= src_port
= ntohs(src_port
);
1371 args
->f_id
.dst_port
= dst_port
= ntohs(dst_port
);
1376 * Packet has already been tagged. Look for the next rule
1377 * to restart processing.
1379 * If fw_one_pass != 0 then just accept it.
1380 * XXX should not happen here, but optimized out in
1386 f
= args
->rule
->next_rule
;
1388 f
= lookup_next_rule(args
->rule
);
1391 * Find the starting rule. It can be either the first
1392 * one, or the one after divert_rule if asked so.
1396 mtag
= m_tag_find(m
, PACKET_TAG_IPFW_DIVERT
, NULL
);
1398 skipto
= *(u_int16_t
*)m_tag_data(mtag
);
1403 if (args
->eh
== NULL
&& skipto
!= 0) {
1404 if (skipto
>= IPFW_DEFAULT_RULE
)
1405 return(IP_FW_PORT_DENY_FLAG
); /* invalid */
1406 while (f
&& f
->rulenum
<= skipto
)
1408 if (f
== NULL
) /* drop packet */
1409 return(IP_FW_PORT_DENY_FLAG
);
1412 if ((mtag
= m_tag_find(m
, PACKET_TAG_IPFW_DIVERT
, NULL
)) != NULL
)
1413 m_tag_delete(m
, mtag
);
1416 * Now scan the rules, and parse microinstructions for each rule.
1418 for (; f
; f
= f
->next
) {
1421 int skip_or
; /* skip rest of OR block */
1424 if (set_disable
& (1 << f
->set
) )
1428 for (l
= f
->cmd_len
, cmd
= f
->cmd
; l
> 0 ;
1429 l
-= cmdlen
, cmd
+= cmdlen
) {
1433 * check_body is a jump target used when we find a
1434 * CHECK_STATE, and need to jump to the body of
1439 cmdlen
= F_LEN(cmd
);
1441 * An OR block (insn_1 || .. || insn_n) has the
1442 * F_OR bit set in all but the last instruction.
1443 * The first match will set "skip_or", and cause
1444 * the following instructions to be skipped until
1445 * past the one with the F_OR bit clear.
1447 if (skip_or
) { /* skip this instruction */
1448 if ((cmd
->len
& F_OR
) == 0)
1449 skip_or
= 0; /* next one is good */
1452 match
= 0; /* set to 1 if we succeed */
1454 switch (cmd
->opcode
) {
1456 * The first set of opcodes compares the packet's
1457 * fields with some pattern, setting 'match' if a
1458 * match is found. At the end of the loop there is
1459 * logic to deal with F_NOT and F_OR flags associated
1467 kprintf("ipfw: opcode %d unimplemented\n",
1474 * We only check offset == 0 && proto != 0,
1475 * as this ensures that we have an IPv4
1476 * packet with the ports info.
1481 struct inpcbinfo
*pi
;
1485 if (proto
== IPPROTO_TCP
) {
1487 pi
= &tcbinfo
[mycpu
->gd_cpuid
];
1488 } else if (proto
== IPPROTO_UDP
) {
1495 in_pcblookup_hash(pi
,
1496 dst_ip
, htons(dst_port
),
1497 src_ip
, htons(src_port
),
1499 in_pcblookup_hash(pi
,
1500 src_ip
, htons(src_port
),
1501 dst_ip
, htons(dst_port
),
1504 if (pcb
== NULL
|| pcb
->inp_socket
== NULL
)
1506 #if defined(__DragonFly__) || (defined(__FreeBSD__) && __FreeBSD_version < 500034)
1507 #define socheckuid(a,b) ((a)->so_cred->cr_uid != (b))
1509 if (cmd
->opcode
== O_UID
) {
1511 !socheckuid(pcb
->inp_socket
,
1512 (uid_t
)((ipfw_insn_u32
*)cmd
)->d
[0]);
1514 match
= groupmember(
1515 (uid_t
)((ipfw_insn_u32
*)cmd
)->d
[0],
1516 pcb
->inp_socket
->so_cred
);
1522 match
= iface_match(m
->m_pkthdr
.rcvif
,
1523 (ipfw_insn_if
*)cmd
);
1527 match
= iface_match(oif
, (ipfw_insn_if
*)cmd
);
1531 match
= iface_match(oif
? oif
:
1532 m
->m_pkthdr
.rcvif
, (ipfw_insn_if
*)cmd
);
1536 if (args
->eh
!= NULL
) { /* have MAC header */
1537 u_int32_t
*want
= (u_int32_t
*)
1538 ((ipfw_insn_mac
*)cmd
)->addr
;
1539 u_int32_t
*mask
= (u_int32_t
*)
1540 ((ipfw_insn_mac
*)cmd
)->mask
;
1541 u_int32_t
*hdr
= (u_int32_t
*)args
->eh
;
1544 ( want
[0] == (hdr
[0] & mask
[0]) &&
1545 want
[1] == (hdr
[1] & mask
[1]) &&
1546 want
[2] == (hdr
[2] & mask
[2]) );
1551 if (args
->eh
!= NULL
) {
1553 ntohs(args
->eh
->ether_type
);
1555 ((ipfw_insn_u16
*)cmd
)->ports
;
1558 for (i
= cmdlen
- 1; !match
&& i
>0;
1560 match
= (t
>=p
[0] && t
<=p
[1]);
1565 match
= (hlen
> 0 && offset
!= 0);
1568 case O_IN
: /* "out" is "not in" */
1569 match
= (oif
== NULL
);
1573 match
= (args
->eh
!= NULL
);
1578 * We do not allow an arg of 0 so the
1579 * check of "proto" only suffices.
1581 match
= (proto
== cmd
->arg1
);
1585 match
= (hlen
> 0 &&
1586 ((ipfw_insn_ip
*)cmd
)->addr
.s_addr
==
1591 match
= (hlen
> 0 &&
1592 ((ipfw_insn_ip
*)cmd
)->addr
.s_addr
==
1594 ((ipfw_insn_ip
*)cmd
)->mask
.s_addr
));
1601 INADDR_TO_IFP(src_ip
, tif
);
1602 match
= (tif
!= NULL
);
1609 u_int32_t
*d
= (u_int32_t
*)(cmd
+1);
1611 cmd
->opcode
== O_IP_DST_SET
?
1617 addr
-= d
[0]; /* subtract base */
1618 match
= (addr
< cmd
->arg1
) &&
1619 ( d
[ 1 + (addr
>>5)] &
1620 (1<<(addr
& 0x1f)) );
1625 match
= (hlen
> 0 &&
1626 ((ipfw_insn_ip
*)cmd
)->addr
.s_addr
==
1631 match
= (hlen
> 0) &&
1632 (((ipfw_insn_ip
*)cmd
)->addr
.s_addr
==
1634 ((ipfw_insn_ip
*)cmd
)->mask
.s_addr
));
1641 INADDR_TO_IFP(dst_ip
, tif
);
1642 match
= (tif
!= NULL
);
1649 * offset == 0 && proto != 0 is enough
1650 * to guarantee that we have an IPv4
1651 * packet with port info.
1653 if ((proto
==IPPROTO_UDP
|| proto
==IPPROTO_TCP
)
1656 (cmd
->opcode
== O_IP_SRCPORT
) ?
1657 src_port
: dst_port
;
1659 ((ipfw_insn_u16
*)cmd
)->ports
;
1662 for (i
= cmdlen
- 1; !match
&& i
>0;
1664 match
= (x
>=p
[0] && x
<=p
[1]);
1669 match
= (offset
== 0 && proto
==IPPROTO_ICMP
&&
1670 icmptype_match(ip
, (ipfw_insn_u32
*)cmd
) );
1674 match
= (hlen
> 0 && ipopts_match(ip
, cmd
) );
1678 match
= (hlen
> 0 && cmd
->arg1
== ip
->ip_v
);
1682 match
= (hlen
> 0 && cmd
->arg1
== ip
->ip_ttl
);
1686 match
= (hlen
> 0 &&
1687 cmd
->arg1
== ntohs(ip
->ip_id
));
1691 match
= (hlen
> 0 && cmd
->arg1
== ip_len
);
1694 case O_IPPRECEDENCE
:
1695 match
= (hlen
> 0 &&
1696 (cmd
->arg1
== (ip
->ip_tos
& 0xe0)) );
1700 match
= (hlen
> 0 &&
1701 flags_match(cmd
, ip
->ip_tos
));
1705 match
= (proto
== IPPROTO_TCP
&& offset
== 0 &&
1707 L3HDR(struct tcphdr
,ip
)->th_flags
));
1711 match
= (proto
== IPPROTO_TCP
&& offset
== 0 &&
1712 tcpopts_match(ip
, cmd
));
1716 match
= (proto
== IPPROTO_TCP
&& offset
== 0 &&
1717 ((ipfw_insn_u32
*)cmd
)->d
[0] ==
1718 L3HDR(struct tcphdr
,ip
)->th_seq
);
1722 match
= (proto
== IPPROTO_TCP
&& offset
== 0 &&
1723 ((ipfw_insn_u32
*)cmd
)->d
[0] ==
1724 L3HDR(struct tcphdr
,ip
)->th_ack
);
1728 match
= (proto
== IPPROTO_TCP
&& offset
== 0 &&
1730 L3HDR(struct tcphdr
,ip
)->th_win
);
1734 /* reject packets which have SYN only */
1735 /* XXX should i also check for TH_ACK ? */
1736 match
= (proto
== IPPROTO_TCP
&& offset
== 0 &&
1737 (L3HDR(struct tcphdr
,ip
)->th_flags
&
1738 (TH_RST
| TH_ACK
| TH_SYN
)) != TH_SYN
);
1743 ipfw_log(f
, hlen
, args
->eh
, m
, oif
);
1748 match
= (krandom() <
1749 ((ipfw_insn_u32
*)cmd
)->d
[0]);
1753 * The second set of opcodes represents 'actions',
1754 * i.e. the terminal part of a rule once the packet
1755 * matches all previous patterns.
1756 * Typically there is only one action for each rule,
1757 * and the opcode is stored at the end of the rule
1758 * (but there are exceptions -- see below).
1760 * In general, here we set retval and terminate the
1761 * outer loop (would be a 'break 3' in some language,
1762 * but we need to do a 'goto done').
1765 * O_COUNT and O_SKIPTO actions:
1766 * instead of terminating, we jump to the next rule
1767 * ('goto next_rule', equivalent to a 'break 2'),
1768 * or to the SKIPTO target ('goto again' after
1769 * having set f, cmd and l), respectively.
1771 * O_LIMIT and O_KEEP_STATE: these opcodes are
1772 * not real 'actions', and are stored right
1773 * before the 'action' part of the rule.
1774 * These opcodes try to install an entry in the
1775 * state tables; if successful, we continue with
1776 * the next opcode (match=1; break;), otherwise
1777 * the packet * must be dropped
1778 * ('goto done' after setting retval);
1780 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
1781 * cause a lookup of the state table, and a jump
1782 * to the 'action' part of the parent rule
1783 * ('goto check_body') if an entry is found, or
1784 * (CHECK_STATE only) a jump to the next rule if
1785 * the entry is not found ('goto next_rule').
1786 * The result of the lookup is cached to make
1787 * further instances of these opcodes are
1792 if (install_state(f
,
1793 (ipfw_insn_limit
*)cmd
, args
)) {
1794 retval
= IP_FW_PORT_DENY_FLAG
;
1795 goto done
; /* error/limit violation */
1803 * dynamic rules are checked at the first
1804 * keep-state or check-state occurrence,
1805 * with the result being stored in dyn_dir.
1806 * The compiler introduces a PROBE_STATE
1807 * instruction for us when we have a
1808 * KEEP_STATE (because PROBE_STATE needs
1811 if (dyn_dir
== MATCH_UNKNOWN
&&
1812 (q
= lookup_dyn_rule(&args
->f_id
,
1813 &dyn_dir
, proto
== IPPROTO_TCP
?
1814 L3HDR(struct tcphdr
, ip
) : NULL
))
1817 * Found dynamic entry, update stats
1818 * and jump to the 'action' part of
1824 cmd
= ACTION_PTR(f
);
1825 l
= f
->cmd_len
- f
->act_ofs
;
1829 * Dynamic entry not found. If CHECK_STATE,
1830 * skip to next rule, if PROBE_STATE just
1831 * ignore and continue with next opcode.
1833 if (cmd
->opcode
== O_CHECK_STATE
)
1839 retval
= 0; /* accept */
1844 args
->rule
= f
; /* report matching rule */
1845 retval
= cmd
->arg1
| IP_FW_PORT_DYNT_FLAG
;
1850 if (args
->eh
) /* not on layer 2 */
1853 mtag
= m_tag_get(PACKET_TAG_IPFW_DIVERT
,
1854 sizeof(u_int16_t
), M_NOWAIT
);
1856 retval
= IP_FW_PORT_DENY_FLAG
;
1859 *(u_int16_t
*)m_tag_data(mtag
) = f
->rulenum
;
1860 m_tag_prepend(m
, mtag
);
1861 retval
= (cmd
->opcode
== O_DIVERT
) ?
1863 cmd
->arg1
| IP_FW_PORT_TEE_FLAG
;
1868 f
->pcnt
++; /* update stats */
1870 f
->timestamp
= time_second
;
1871 if (cmd
->opcode
== O_COUNT
)
1874 if (f
->next_rule
== NULL
)
1875 lookup_next_rule(f
);
1881 * Drop the packet and send a reject notice
1882 * if the packet is not ICMP (or is an ICMP
1883 * query), and it is not multicast/broadcast.
1886 (proto
!= IPPROTO_ICMP
||
1887 is_icmp_query(ip
)) &&
1888 !(m
->m_flags
& (M_BCAST
|M_MCAST
)) &&
1889 !IN_MULTICAST(ntohl(dst_ip
.s_addr
))) {
1890 send_reject(args
, cmd
->arg1
,
1896 retval
= IP_FW_PORT_DENY_FLAG
;
1900 if (args
->eh
) /* not valid on layer2 pkts */
1902 if (!q
|| dyn_dir
== MATCH_FORWARD
)
1904 &((ipfw_insn_sa
*)cmd
)->sa
;
1909 panic("-- unknown opcode %d\n", cmd
->opcode
);
1910 } /* end of switch() on opcodes */
1912 if (cmd
->len
& F_NOT
)
1916 if (cmd
->len
& F_OR
)
1919 if (!(cmd
->len
& F_OR
)) /* not an OR block, */
1920 break; /* try next rule */
1923 } /* end of inner for, scan opcodes */
1925 next_rule
:; /* try next rule */
1927 } /* end of outer for, scan rules */
1928 kprintf("+++ ipfw: ouch!, skip past end of rules, denying packet\n");
1929 return(IP_FW_PORT_DENY_FLAG
);
1932 /* Update statistics */
1935 f
->timestamp
= time_second
;
1940 kprintf("pullup failed\n");
1941 return(IP_FW_PORT_DENY_FLAG
);
1945 * When a rule is added/deleted, clear the next_rule pointers in all rules.
1946 * These will be reconstructed on the fly as packets are matched.
1947 * Must be called at splimp().
1950 flush_rule_ptrs(void)
1954 for (rule
= layer3_chain
; rule
; rule
= rule
->next
)
1955 rule
->next_rule
= NULL
;
1959 * When pipes/queues are deleted, clear the "pipe_ptr" pointer to a given
1960 * pipe/queue, or to all of them (match == NULL).
1961 * Must be called at splimp().
1964 flush_pipe_ptrs(struct dn_flow_set
*match
)
1968 for (rule
= layer3_chain
; rule
; rule
= rule
->next
) {
1969 ipfw_insn_pipe
*cmd
= (ipfw_insn_pipe
*)ACTION_PTR(rule
);
1971 if (cmd
->o
.opcode
!= O_PIPE
&& cmd
->o
.opcode
!= O_QUEUE
)
1973 if (match
== NULL
|| cmd
->pipe_ptr
== match
)
1974 cmd
->pipe_ptr
= NULL
;
1979 * Add a new rule to the list. Copy the rule into a malloc'ed area, then
1980 * possibly create a rule number and add the rule to the list.
1981 * Update the rule_number in the input struct so the caller knows it as well.
1984 add_rule(struct ip_fw
**head
, struct ip_fw
*input_rule
)
1986 struct ip_fw
*rule
, *f
, *prev
;
1987 int l
= RULESIZE(input_rule
);
1989 if (*head
== NULL
&& input_rule
->rulenum
!= IPFW_DEFAULT_RULE
)
1992 rule
= kmalloc(l
, M_IPFW
, M_WAITOK
| M_ZERO
);
1996 bcopy(input_rule
, rule
, l
);
1999 rule
->next_rule
= NULL
;
2003 rule
->timestamp
= 0;
2007 if (*head
== NULL
) { /* default rule */
2013 * If rulenum is 0, find highest numbered rule before the
2014 * default rule, and add autoinc_step
2016 if (autoinc_step
< 1)
2018 else if (autoinc_step
> 1000)
2019 autoinc_step
= 1000;
2020 if (rule
->rulenum
== 0) {
2022 * locate the highest numbered rule before default
2024 for (f
= *head
; f
; f
= f
->next
) {
2025 if (f
->rulenum
== IPFW_DEFAULT_RULE
)
2027 rule
->rulenum
= f
->rulenum
;
2029 if (rule
->rulenum
< IPFW_DEFAULT_RULE
- autoinc_step
)
2030 rule
->rulenum
+= autoinc_step
;
2031 input_rule
->rulenum
= rule
->rulenum
;
2035 * Now insert the new rule in the right place in the sorted list.
2037 for (prev
= NULL
, f
= *head
; f
; prev
= f
, f
= f
->next
) {
2038 if (f
->rulenum
> rule
->rulenum
) { /* found the location */
2042 } else { /* head insert */
2054 DEB(kprintf("++ installed rule %d, static count now %d\n",
2055 rule
->rulenum
, static_count
);)
2060 * Free storage associated with a static rule (including derived
2062 * The caller is in charge of clearing rule pointers to avoid
2063 * dangling pointers.
2064 * @return a pointer to the next entry.
2065 * Arguments are not checked, so they better be correct.
2066 * Must be called at splimp().
2068 static struct ip_fw
*
2069 delete_rule(struct ip_fw
**head
, struct ip_fw
*prev
, struct ip_fw
*rule
)
2072 int l
= RULESIZE(rule
);
2075 remove_dyn_rule(rule
, NULL
/* force removal */);
2083 if (DUMMYNET_LOADED
)
2084 ip_dn_ruledel_ptr(rule
);
2085 kfree(rule
, M_IPFW
);
2090 * Deletes all rules from a chain (including the default rule
2091 * if the second argument is set).
2092 * Must be called at splimp().
2095 free_chain(struct ip_fw
**chain
, int kill_default
)
2099 flush_rule_ptrs(); /* more efficient to do outside the loop */
2101 while ( (rule
= *chain
) != NULL
&&
2102 (kill_default
|| rule
->rulenum
!= IPFW_DEFAULT_RULE
) )
2103 delete_rule(chain
, NULL
, rule
);
2107 * Remove all rules with given number, and also do set manipulation.
2109 * The argument is an u_int32_t. The low 16 bit are the rule or set number,
2110 * the next 8 bits are the new set, the top 8 bits are the command:
2112 * 0 delete rules with given number
2113 * 1 delete rules with given set number
2114 * 2 move rules with given number to new set
2115 * 3 move rules with given set number to new set
2116 * 4 swap sets with given numbers
2119 del_entry(struct ip_fw
**chain
, u_int32_t arg
)
2121 struct ip_fw
*prev
, *rule
;
2123 u_int8_t cmd
, new_set
;
2125 rulenum
= arg
& 0xffff;
2126 cmd
= (arg
>> 24) & 0xff;
2127 new_set
= (arg
>> 16) & 0xff;
2133 if (cmd
== 0 || cmd
== 2) {
2134 if (rulenum
== IPFW_DEFAULT_RULE
)
2142 case 0: /* delete rules with given number */
2144 * locate first rule to delete
2146 for (prev
= NULL
, rule
= *chain
;
2147 rule
&& rule
->rulenum
< rulenum
;
2148 prev
= rule
, rule
= rule
->next
)
2150 if (rule
->rulenum
!= rulenum
)
2153 crit_enter(); /* no access to rules while removing */
2155 * flush pointers outside the loop, then delete all matching
2156 * rules. prev remains the same throughout the cycle.
2159 while (rule
&& rule
->rulenum
== rulenum
)
2160 rule
= delete_rule(chain
, prev
, rule
);
2164 case 1: /* delete all rules with given set number */
2167 for (prev
= NULL
, rule
= *chain
; rule
; )
2168 if (rule
->set
== rulenum
)
2169 rule
= delete_rule(chain
, prev
, rule
);
2177 case 2: /* move rules with given number to new set */
2179 for (rule
= *chain
; rule
; rule
= rule
->next
)
2180 if (rule
->rulenum
== rulenum
)
2181 rule
->set
= new_set
;
2185 case 3: /* move rules with given set number to new set */
2187 for (rule
= *chain
; rule
; rule
= rule
->next
)
2188 if (rule
->set
== rulenum
)
2189 rule
->set
= new_set
;
2193 case 4: /* swap two sets */
2195 for (rule
= *chain
; rule
; rule
= rule
->next
)
2196 if (rule
->set
== rulenum
)
2197 rule
->set
= new_set
;
2198 else if (rule
->set
== new_set
)
2199 rule
->set
= rulenum
;
2207 * Clear counters for a specific rule.
2210 clear_counters(struct ip_fw
*rule
, int log_only
)
2212 ipfw_insn_log
*l
= (ipfw_insn_log
*)ACTION_PTR(rule
);
2214 if (log_only
== 0) {
2215 rule
->bcnt
= rule
->pcnt
= 0;
2216 rule
->timestamp
= 0;
2218 if (l
->o
.opcode
== O_LOG
)
2219 l
->log_left
= l
->max_log
;
2223 * Reset some or all counters on firewall rules.
2224 * @arg frwl is null to clear all entries, or contains a specific
2226 * @arg log_only is 1 if we only want to reset logs, zero otherwise.
2229 zero_entry(int rulenum
, int log_only
)
2237 for (rule
= layer3_chain
; rule
; rule
= rule
->next
)
2238 clear_counters(rule
, log_only
);
2240 msg
= log_only
? "ipfw: All logging counts reset.\n" :
2241 "ipfw: Accounting cleared.\n";
2245 * We can have multiple rules with the same number, so we
2246 * need to clear them all.
2248 for (rule
= layer3_chain
; rule
; rule
= rule
->next
)
2249 if (rule
->rulenum
== rulenum
) {
2251 while (rule
&& rule
->rulenum
== rulenum
) {
2252 clear_counters(rule
, log_only
);
2259 if (!cleared
) /* we did not find any matching rules */
2261 msg
= log_only
? "ipfw: Entry %d logging count reset.\n" :
2262 "ipfw: Entry %d cleared.\n";
2265 log(LOG_SECURITY
| LOG_NOTICE
, msg
, rulenum
);
2270 * Check validity of the structure before insert.
2271 * Fortunately rules are simple, so this mostly need to check rule sizes.
2274 check_ipfw_struct(struct ip_fw
*rule
, int size
)
2280 if (size
< sizeof(*rule
)) {
2281 kprintf("ipfw: rule too short\n");
2284 /* first, check for valid size */
2287 kprintf("ipfw: size mismatch (have %d want %d)\n", size
, l
);
2291 * Now go for the individual checks. Very simple ones, basically only
2292 * instruction sizes.
2294 for (l
= rule
->cmd_len
, cmd
= rule
->cmd
;
2295 l
> 0 ; l
-= cmdlen
, cmd
+= cmdlen
) {
2296 cmdlen
= F_LEN(cmd
);
2298 kprintf("ipfw: opcode %d size truncated\n",
2302 DEB(kprintf("ipfw: opcode %d\n", cmd
->opcode
);)
2303 switch (cmd
->opcode
) {
2317 case O_IPPRECEDENCE
:
2324 if (cmdlen
!= F_INSN_SIZE(ipfw_insn
))
2336 if (cmdlen
!= F_INSN_SIZE(ipfw_insn_u32
))
2341 if (cmdlen
!= F_INSN_SIZE(ipfw_insn_limit
))
2346 if (cmdlen
!= F_INSN_SIZE(ipfw_insn_log
))
2349 ((ipfw_insn_log
*)cmd
)->log_left
=
2350 ((ipfw_insn_log
*)cmd
)->max_log
;
2356 if (cmdlen
!= F_INSN_SIZE(ipfw_insn_ip
))
2358 if (((ipfw_insn_ip
*)cmd
)->mask
.s_addr
== 0) {
2359 kprintf("ipfw: opcode %d, useless rule\n",
2367 if (cmd
->arg1
== 0 || cmd
->arg1
> 256) {
2368 kprintf("ipfw: invalid set size %d\n",
2372 if (cmdlen
!= F_INSN_SIZE(ipfw_insn_u32
) +
2378 if (cmdlen
!= F_INSN_SIZE(ipfw_insn_mac
))
2384 case O_IP_DSTPORT
: /* XXX artificial limit, 30 port pairs */
2385 if (cmdlen
< 2 || cmdlen
> 31)
2392 if (cmdlen
!= F_INSN_SIZE(ipfw_insn_if
))
2398 if (cmdlen
!= F_INSN_SIZE(ipfw_insn_pipe
))
2403 if (cmdlen
!= F_INSN_SIZE(ipfw_insn_sa
))
2407 case O_FORWARD_MAC
: /* XXX not implemented yet */
2416 if (cmdlen
!= F_INSN_SIZE(ipfw_insn
))
2420 kprintf("ipfw: opcode %d, multiple actions"
2427 kprintf("ipfw: opcode %d, action must be"
2434 kprintf("ipfw: opcode %d, unknown opcode\n",
2439 if (have_action
== 0) {
2440 kprintf("ipfw: missing action\n");
2446 kprintf("ipfw: opcode %d size %d wrong\n",
2447 cmd
->opcode
, cmdlen
);
2453 * {set|get}sockopt parser.
2456 ipfw_ctl(struct sockopt
*sopt
)
2460 struct ip_fw
*bp
, *buf
, *rule
;
2462 static u_int32_t rule_buf
[255]; /* we copy the data here */
2465 * Disallow modifications in really-really secure mode, but still allow
2466 * the logging counters to be reset.
2468 if (sopt
->sopt_name
== IP_FW_ADD
||
2469 (sopt
->sopt_dir
== SOPT_SET
&& sopt
->sopt_name
!= IP_FW_RESETLOG
)) {
2470 #if defined(__FreeBSD__) && __FreeBSD_version >= 500034
2471 error
= securelevel_ge(sopt
->sopt_td
->td_ucred
, 3);
2474 #else /* FreeBSD 4.x */
2475 if (securelevel
>= 3)
2482 switch (sopt
->sopt_name
) {
2485 * pass up a copy of the current rules. Static rules
2486 * come first (the last of which has number IPFW_DEFAULT_RULE),
2487 * followed by a possibly empty list of dynamic rule.
2488 * The last dynamic rule has NULL in the "next" field.
2491 size
= static_len
; /* size of static rules */
2492 if (ipfw_dyn_v
) /* add size of dyn.rules */
2493 size
+= (dyn_count
* sizeof(ipfw_dyn_rule
));
2496 * XXX todo: if the user passes a short length just to know
2497 * how much room is needed, do not bother filling up the
2498 * buffer, just jump to the sooptcopyout.
2500 buf
= kmalloc(size
, M_TEMP
, M_WAITOK
);
2503 for (rule
= layer3_chain
; rule
; rule
= rule
->next
) {
2504 int i
= RULESIZE(rule
);
2507 * abuse 'next_rule' to store the set_disable word
2509 bcopy(&set_disable
, &(((struct ip_fw
*)bp
)->next_rule
),
2510 sizeof(set_disable
));
2511 bp
= (struct ip_fw
*)((char *)bp
+ i
);
2515 ipfw_dyn_rule
*p
, *dst
, *last
= NULL
;
2517 dst
= (ipfw_dyn_rule
*)bp
;
2518 for (i
= 0 ; i
< curr_dyn_buckets
; i
++ )
2519 for ( p
= ipfw_dyn_v
[i
] ; p
!= NULL
;
2520 p
= p
->next
, dst
++ ) {
2521 bcopy(p
, dst
, sizeof *p
);
2522 bcopy(&(p
->rule
->rulenum
), &(dst
->rule
),
2523 sizeof(p
->rule
->rulenum
));
2525 * store a non-null value in "next".
2526 * The userland code will interpret a
2527 * NULL here as a marker
2528 * for the last dynamic rule.
2533 TIME_LEQ(dst
->expire
, time_second
) ?
2534 0 : dst
->expire
- time_second
;
2536 if (last
!= NULL
) /* mark last dynamic rule */
2541 error
= sooptcopyout(sopt
, buf
, size
);
2547 * Normally we cannot release the lock on each iteration.
2548 * We could do it here only because we start from the head all
2549 * the times so there is no risk of missing some entries.
2550 * On the other hand, the risk is that we end up with
2551 * a very inconsistent ruleset, so better keep the lock
2552 * around the whole cycle.
2554 * XXX this code can be improved by resetting the head of
2555 * the list to point to the default rule, and then freeing
2556 * the old list without the need for a lock.
2560 free_chain(&layer3_chain
, 0 /* keep default rule */);
2565 rule
= (struct ip_fw
*)rule_buf
; /* XXX do a malloc */
2566 error
= sooptcopyin(sopt
, rule
, sizeof(rule_buf
),
2567 sizeof(struct ip_fw
) );
2568 size
= sopt
->sopt_valsize
;
2569 if (error
|| (error
= check_ipfw_struct(rule
, size
)))
2572 error
= add_rule(&layer3_chain
, rule
);
2573 size
= RULESIZE(rule
);
2574 if (!error
&& sopt
->sopt_dir
== SOPT_GET
)
2575 error
= sooptcopyout(sopt
, rule
, size
);
2580 * IP_FW_DEL is used for deleting single rules or sets,
2581 * and (ab)used to atomically manipulate sets. Argument size
2582 * is used to distinguish between the two:
2584 * delete single rule or set of rules,
2585 * or reassign rules (or sets) to a different set.
2586 * 2*sizeof(u_int32_t)
2587 * atomic disable/enable sets.
2588 * first u_int32_t contains sets to be disabled,
2589 * second u_int32_t contains sets to be enabled.
2591 error
= sooptcopyin(sopt
, rule_buf
,
2592 2*sizeof(u_int32_t
), sizeof(u_int32_t
));
2595 size
= sopt
->sopt_valsize
;
2596 if (size
== sizeof(u_int32_t
)) /* delete or reassign */
2597 error
= del_entry(&layer3_chain
, rule_buf
[0]);
2598 else if (size
== 2*sizeof(u_int32_t
)) /* set enable/disable */
2600 (set_disable
| rule_buf
[0]) & ~rule_buf
[1] &
2601 ~(1<<31); /* set 31 always enabled */
2607 case IP_FW_RESETLOG
: /* argument is an int, the rule number */
2610 if (sopt
->sopt_val
!= 0) {
2611 error
= sooptcopyin(sopt
, &rulenum
,
2612 sizeof(int), sizeof(int));
2616 error
= zero_entry(rulenum
, sopt
->sopt_name
== IP_FW_RESETLOG
);
2620 kprintf("ipfw_ctl invalid option %d\n", sopt
->sopt_name
);
2628 * dummynet needs a reference to the default rule, because rules can be
2629 * deleted while packets hold a reference to them. When this happens,
2630 * dummynet changes the reference to the default rule (it could well be a
2631 * NULL pointer, but this way we do not need to check for the special
2632 * case, plus here he have info on the default behaviour).
2634 struct ip_fw
*ip_fw_default_rule
;
2637 * This procedure is only used to handle keepalives. It is invoked
2638 * every dyn_keepalive_period
2641 ipfw_tick(void * __unused unused
)
2646 if (dyn_keepalive
== 0 || ipfw_dyn_v
== NULL
|| dyn_count
== 0)
2650 for (i
= 0 ; i
< curr_dyn_buckets
; i
++) {
2651 for (q
= ipfw_dyn_v
[i
] ; q
; q
= q
->next
) {
2652 if (q
->dyn_type
== O_LIMIT_PARENT
)
2654 if (q
->id
.proto
!= IPPROTO_TCP
)
2656 if ( (q
->state
& BOTH_SYN
) != BOTH_SYN
)
2658 if (TIME_LEQ( time_second
+dyn_keepalive_interval
,
2660 continue; /* too early */
2661 if (TIME_LEQ(q
->expire
, time_second
))
2662 continue; /* too late, rule expired */
2664 send_pkt(&(q
->id
), q
->ack_rev
- 1, q
->ack_fwd
, TH_SYN
);
2665 send_pkt(&(q
->id
), q
->ack_fwd
- 1, q
->ack_rev
, 0);
2670 callout_reset(&ipfw_timeout_h
, dyn_keepalive_period
* hz
,
2677 struct ip_fw default_rule
;
2679 ip_fw_chk_ptr
= ipfw_chk
;
2680 ip_fw_ctl_ptr
= ipfw_ctl
;
2681 layer3_chain
= NULL
;
2683 bzero(&default_rule
, sizeof default_rule
);
2685 default_rule
.act_ofs
= 0;
2686 default_rule
.rulenum
= IPFW_DEFAULT_RULE
;
2687 default_rule
.cmd_len
= 1;
2688 default_rule
.set
= 31;
2690 default_rule
.cmd
[0].len
= 1;
2691 default_rule
.cmd
[0].opcode
=
2692 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
2697 add_rule(&layer3_chain
, &default_rule
);
2699 ip_fw_default_rule
= layer3_chain
;
2700 kprintf("ipfw2 initialized, divert %s, "
2701 "rule-based forwarding enabled, default to %s, logging ",
2707 default_rule
.cmd
[0].opcode
== O_ACCEPT
? "accept" : "deny");
2709 #ifdef IPFIREWALL_VERBOSE
2712 #ifdef IPFIREWALL_VERBOSE_LIMIT
2713 verbose_limit
= IPFIREWALL_VERBOSE_LIMIT
;
2715 if (fw_verbose
== 0)
2716 kprintf("disabled\n");
2717 else if (verbose_limit
== 0)
2718 kprintf("unlimited\n");
2720 kprintf("limited to %d packets/entry by default\n",
2722 callout_init(&ipfw_timeout_h
);
2723 callout_reset(&ipfw_timeout_h
, hz
, ipfw_tick
, NULL
);
2727 ipfw_modevent(module_t mod
, int type
, void *unused
)
2736 kprintf("IP firewall already loaded\n");
2745 #if !defined(KLD_MODULE)
2746 kprintf("ipfw statically compiled, cannot unload\n");
2750 callout_stop(&ipfw_timeout_h
);
2751 ip_fw_chk_ptr
= NULL
;
2752 ip_fw_ctl_ptr
= NULL
;
2753 free_chain(&layer3_chain
, 1 /* kill default rule */);
2755 kprintf("IP firewall unloaded\n");
2764 static moduledata_t ipfwmod
= {
2769 DECLARE_MODULE(ipfw
, ipfwmod
, SI_SUB_PSEUDO
, SI_ORDER_ANY
);
2770 MODULE_VERSION(ipfw
, 1);