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 1.40 2007/12/19 12:13:17 sephe Exp $
33 * Implement IP packet firewall (new version)
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 uint32_t set_disable;
87 static int fw_verbose;
88 static int verbose_limit;
91 static int ipfw_refcnt;
94 static struct callout ipfw_timeout_h;
95 #define IPFW_DEFAULT_RULE 65535
98 * list of rules for layer 3
100 static struct ip_fw *layer3_chain;
102 MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's");
104 static int fw_debug = 1;
105 static int autoinc_step = 100; /* bounded to 1..1000 in ipfw_add_rule() */
108 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
109 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, enable, CTLFLAG_RW,
110 &fw_enable, 0, "Enable ipfw");
111 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW,
112 &autoinc_step, 0, "Rule number autincrement step");
113 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO,one_pass,CTLFLAG_RW,
115 "Only do a single pass through ipfw when using dummynet(4)");
116 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW,
117 &fw_debug, 0, "Enable printing of debug ip_fw statements");
118 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, CTLFLAG_RW,
119 &fw_verbose, 0, "Log matches to ipfw rules");
120 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW,
121 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged");
124 * Description of dynamic rules.
126 * Dynamic rules are stored in lists accessed through a hash table
127 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can
128 * be modified through the sysctl variable dyn_buckets which is
129 * updated when the table becomes empty.
131 * XXX currently there is only one list, ipfw_dyn.
133 * When a packet is received, its address fields are first masked
134 * with the mask defined for the rule, then hashed, then matched
135 * against the entries in the corresponding list.
136 * Dynamic rules can be used for different purposes:
138 * + enforcing limits on the number of sessions;
139 * + in-kernel NAT (not implemented yet)
141 * The lifetime of dynamic rules is regulated by dyn_*_lifetime,
142 * measured in seconds and depending on the flags.
144 * The total number of dynamic rules is stored in dyn_count.
145 * The max number of dynamic rules is dyn_max. When we reach
146 * the maximum number of rules we do not create anymore. This is
147 * done to avoid consuming too much memory, but also too much
148 * time when searching on each packet (ideally, we should try instead
149 * to put a limit on the length of the list on each bucket...).
151 * Each dynamic rule holds a pointer to the parent ipfw rule so
152 * we know what action to perform. Dynamic rules are removed when
153 * the parent rule is deleted. XXX we should make them survive.
155 * There are some limitations with dynamic rules -- we do not
156 * obey the 'randomized match', and we do not do multiple
157 * passes through the firewall. XXX check the latter!!!
159 static ipfw_dyn_rule **ipfw_dyn_v = NULL;
160 static uint32_t dyn_buckets = 256; /* must be power of 2 */
161 static uint32_t curr_dyn_buckets = 256; /* must be power of 2 */
164 * Timeouts for various events in handing dynamic rules.
166 static uint32_t dyn_ack_lifetime = 300;
167 static uint32_t dyn_syn_lifetime = 20;
168 static uint32_t dyn_fin_lifetime = 1;
169 static uint32_t dyn_rst_lifetime = 1;
170 static uint32_t dyn_udp_lifetime = 10;
171 static uint32_t dyn_short_lifetime = 5;
174 * Keepalives are sent if dyn_keepalive is set. They are sent every
175 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval
176 * seconds of lifetime of a rule.
177 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower
178 * than dyn_keepalive_period.
181 static uint32_t dyn_keepalive_interval = 20;
182 static uint32_t dyn_keepalive_period = 5;
183 static uint32_t dyn_keepalive = 1; /* do send keepalives */
185 static uint32_t static_count; /* # of static rules */
186 static uint32_t static_ioc_len; /* bytes of static rules */
187 static uint32_t dyn_count; /* # of dynamic rules */
188 static uint32_t dyn_max = 4096; /* max # of dynamic rules */
190 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW,
191 &dyn_buckets, 0, "Number of dyn. buckets");
192 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD,
193 &curr_dyn_buckets, 0, "Current Number of dyn. buckets");
194 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD,
195 &dyn_count, 0, "Number of dyn. rules");
196 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW,
197 &dyn_max, 0, "Max number of dyn. rules");
198 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD,
199 &static_count, 0, "Number of static rules");
200 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW,
201 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks");
202 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW,
203 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn");
204 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW,
205 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin");
206 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW,
207 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst");
208 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW,
209 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP");
210 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW,
211 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations");
212 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW,
213 &dyn_keepalive, 0, "Enable keepalives for dyn. rules");
215 #endif /* SYSCTL_NODE */
218 * dummynet needs a reference to the default rule, because rules can be
219 * deleted while packets hold a reference to them. When this happens,
220 * dummynet changes the reference to the default rule (it could well be a
221 * NULL pointer, but this way we do not need to check for the special
222 * case, plus here he have info on the default behaviour).
224 struct ip_fw *ip_fw_default_rule;
226 static ip_fw_chk_t ipfw_chk;
229 ipfw_free_rule(struct ip_fw *rule)
231 KASSERT(rule->refcnt > 0, ("invalid refcnt %u\n", rule->refcnt));
232 atomic_subtract_int(&rule->refcnt, 1);
233 if (atomic_cmpset_int(&rule->refcnt, 0, 1)) {
241 ipfw_unref_rule(void *priv)
243 ipfw_free_rule(priv);
245 atomic_subtract_int(&ipfw_refcnt, 1);
250 ipfw_ref_rule(struct ip_fw *rule)
253 atomic_add_int(&ipfw_refcnt, 1);
255 atomic_add_int(&rule->refcnt, 1);
259 * This macro maps an ip pointer into a layer3 header pointer of type T
261 #define L3HDR(T, ip) ((T *)((uint32_t *)(ip) + (ip)->ip_hl))
264 icmptype_match(struct ip *ip, ipfw_insn_u32 *cmd)
266 int type = L3HDR(struct icmp,ip)->icmp_type;
268 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
271 #define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
272 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
275 is_icmp_query(struct ip *ip)
277 int type = L3HDR(struct icmp, ip)->icmp_type;
278 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
283 * The following checks use two arrays of 8 or 16 bits to store the
284 * bits that we want set or clear, respectively. They are in the
285 * low and high half of cmd->arg1 or cmd->d[0].
287 * We scan options and store the bits we find set. We succeed if
289 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
291 * The code is sometimes optimized not to store additional variables.
295 flags_match(ipfw_insn *cmd, uint8_t bits)
300 if ( ((cmd->arg1 & 0xff) & bits) != 0)
301 return 0; /* some bits we want set were clear */
302 want_clear = (cmd->arg1 >> 8) & 0xff;
303 if ( (want_clear & bits) != want_clear)
304 return 0; /* some bits we want clear were set */
309 ipopts_match(struct ip *ip, ipfw_insn *cmd)
311 int optlen, bits = 0;
312 u_char *cp = (u_char *)(ip + 1);
313 int x = (ip->ip_hl << 2) - sizeof (struct ip);
315 for (; x > 0; x -= optlen, cp += optlen) {
316 int opt = cp[IPOPT_OPTVAL];
318 if (opt == IPOPT_EOL)
320 if (opt == IPOPT_NOP)
323 optlen = cp[IPOPT_OLEN];
324 if (optlen <= 0 || optlen > x)
325 return 0; /* invalid or truncated */
333 bits |= IP_FW_IPOPT_LSRR;
337 bits |= IP_FW_IPOPT_SSRR;
341 bits |= IP_FW_IPOPT_RR;
345 bits |= IP_FW_IPOPT_TS;
349 return (flags_match(cmd, bits));
353 tcpopts_match(struct ip *ip, ipfw_insn *cmd)
355 int optlen, bits = 0;
356 struct tcphdr *tcp = L3HDR(struct tcphdr,ip);
357 u_char *cp = (u_char *)(tcp + 1);
358 int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
360 for (; x > 0; x -= optlen, cp += optlen) {
362 if (opt == TCPOPT_EOL)
364 if (opt == TCPOPT_NOP)
378 bits |= IP_FW_TCPOPT_MSS;
382 bits |= IP_FW_TCPOPT_WINDOW;
385 case TCPOPT_SACK_PERMITTED:
387 bits |= IP_FW_TCPOPT_SACK;
390 case TCPOPT_TIMESTAMP:
391 bits |= IP_FW_TCPOPT_TS;
397 bits |= IP_FW_TCPOPT_CC;
401 return (flags_match(cmd, bits));
405 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd)
407 if (ifp == NULL) /* no iface with this packet, match fails */
409 /* Check by name or by IP address */
410 if (cmd->name[0] != '\0') { /* match by name */
413 if (kfnmatch(cmd->name, ifp->if_xname, 0) == 0)
416 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
422 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
423 if (ia->ifa_addr == NULL)
425 if (ia->ifa_addr->sa_family != AF_INET)
427 if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
428 (ia->ifa_addr))->sin_addr.s_addr)
429 return(1); /* match */
432 return(0); /* no match, fail ... */
435 static uint64_t norule_counter; /* counter for ipfw_log(NULL...) */
437 #define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0
438 #define SNP(buf) buf, sizeof(buf)
441 * We enter here when we have a rule with O_LOG.
442 * XXX this function alone takes about 2Kbytes of code!
445 ipfw_log(struct ip_fw *f, u_int hlen, struct ether_header *eh,
446 struct mbuf *m, struct ifnet *oif)
449 int limit_reached = 0;
450 char action2[40], proto[48], fragment[28];
455 if (f == NULL) { /* bogus pkt */
456 if (verbose_limit != 0 && norule_counter >= verbose_limit)
459 if (norule_counter == verbose_limit)
460 limit_reached = verbose_limit;
462 } else { /* O_LOG is the first action, find the real one */
463 ipfw_insn *cmd = ACTION_PTR(f);
464 ipfw_insn_log *l = (ipfw_insn_log *)cmd;
466 if (l->max_log != 0 && l->log_left == 0)
469 if (l->log_left == 0)
470 limit_reached = l->max_log;
471 cmd += F_LEN(cmd); /* point to first action */
472 if (cmd->opcode == O_PROB)
476 switch (cmd->opcode) {
482 if (cmd->arg1==ICMP_REJECT_RST)
484 else if (cmd->arg1==ICMP_UNREACH_HOST)
487 ksnprintf(SNPARGS(action2, 0), "Unreach %d",
498 ksnprintf(SNPARGS(action2, 0), "Divert %d",
502 ksnprintf(SNPARGS(action2, 0), "Tee %d",
506 ksnprintf(SNPARGS(action2, 0), "SkipTo %d",
510 ksnprintf(SNPARGS(action2, 0), "Pipe %d",
514 ksnprintf(SNPARGS(action2, 0), "Queue %d",
518 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd;
521 len = ksnprintf(SNPARGS(action2, 0), "Forward to %s",
522 inet_ntoa(sa->sa.sin_addr));
524 ksnprintf(SNPARGS(action2, len), ":%d",
534 if (hlen == 0) { /* non-ip */
535 ksnprintf(SNPARGS(proto, 0), "MAC");
537 struct ip *ip = mtod(m, struct ip *);
538 /* these three are all aliases to the same thing */
539 struct icmp *const icmp = L3HDR(struct icmp, ip);
540 struct tcphdr *const tcp = (struct tcphdr *)icmp;
541 struct udphdr *const udp = (struct udphdr *)icmp;
543 int ip_off, offset, ip_len;
547 if (eh != NULL) { /* layer 2 packets are as on the wire */
548 ip_off = ntohs(ip->ip_off);
549 ip_len = ntohs(ip->ip_len);
554 offset = ip_off & IP_OFFMASK;
557 len = ksnprintf(SNPARGS(proto, 0), "TCP %s",
558 inet_ntoa(ip->ip_src));
560 ksnprintf(SNPARGS(proto, len), ":%d %s:%d",
561 ntohs(tcp->th_sport),
562 inet_ntoa(ip->ip_dst),
563 ntohs(tcp->th_dport));
565 ksnprintf(SNPARGS(proto, len), " %s",
566 inet_ntoa(ip->ip_dst));
570 len = ksnprintf(SNPARGS(proto, 0), "UDP %s",
571 inet_ntoa(ip->ip_src));
573 ksnprintf(SNPARGS(proto, len), ":%d %s:%d",
574 ntohs(udp->uh_sport),
575 inet_ntoa(ip->ip_dst),
576 ntohs(udp->uh_dport));
578 ksnprintf(SNPARGS(proto, len), " %s",
579 inet_ntoa(ip->ip_dst));
584 len = ksnprintf(SNPARGS(proto, 0),
586 icmp->icmp_type, icmp->icmp_code);
588 len = ksnprintf(SNPARGS(proto, 0), "ICMP ");
589 len += ksnprintf(SNPARGS(proto, len), "%s",
590 inet_ntoa(ip->ip_src));
591 ksnprintf(SNPARGS(proto, len), " %s",
592 inet_ntoa(ip->ip_dst));
596 len = ksnprintf(SNPARGS(proto, 0), "P:%d %s", ip->ip_p,
597 inet_ntoa(ip->ip_src));
598 ksnprintf(SNPARGS(proto, len), " %s",
599 inet_ntoa(ip->ip_dst));
603 if (ip_off & (IP_MF | IP_OFFMASK))
604 ksnprintf(SNPARGS(fragment, 0), " (frag %d:%d@%d%s)",
605 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2),
607 (ip_off & IP_MF) ? "+" : "");
609 if (oif || m->m_pkthdr.rcvif)
610 log(LOG_SECURITY | LOG_INFO,
611 "ipfw: %d %s %s %s via %s%s\n",
613 action, proto, oif ? "out" : "in",
614 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname,
617 log(LOG_SECURITY | LOG_INFO,
618 "ipfw: %d %s %s [no if info]%s\n",
620 action, proto, fragment);
622 log(LOG_SECURITY | LOG_NOTICE,
623 "ipfw: limit %d reached on entry %d\n",
624 limit_reached, f ? f->rulenum : -1);
628 * IMPORTANT: the hash function for dynamic rules must be commutative
629 * in source and destination (ip,port), because rules are bidirectional
630 * and we want to find both in the same bucket.
633 hash_packet(struct ipfw_flow_id *id)
637 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port);
638 i &= (curr_dyn_buckets - 1);
643 * unlink a dynamic rule from a chain. prev is a pointer to
644 * the previous one, q is a pointer to the rule to delete,
645 * head is a pointer to the head of the queue.
646 * Modifies q and potentially also head.
648 #define UNLINK_DYN_RULE(prev, head, q) { \
649 ipfw_dyn_rule *old_q = q; \
651 /* remove a refcount to the parent */ \
652 if (q->dyn_type == O_LIMIT) \
653 q->parent->count--; \
654 DEB(kprintf("-- unlink entry 0x%08x %d -> 0x%08x %d, %d left\n", \
655 (q->id.src_ip), (q->id.src_port), \
656 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \
658 prev->next = q = q->next; \
660 head = q = q->next; \
661 KASSERT(dyn_count > 0, ("invalid dyn count %u\n", dyn_count)); \
663 kfree(old_q, M_IPFW); }
665 #define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0)
668 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL.
670 * If keep_me == NULL, rules are deleted even if not expired,
671 * otherwise only expired rules are removed.
673 * The value of the second parameter is also used to point to identify
674 * a rule we absolutely do not want to remove (e.g. because we are
675 * holding a reference to it -- this is the case with O_LIMIT_PARENT
676 * rules). The pointer is only used for comparison, so any non-null
680 remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me)
682 static uint32_t last_remove = 0;
684 #define FORCE (keep_me == NULL)
686 ipfw_dyn_rule *prev, *q;
687 int i, pass = 0, max_pass = 0;
689 if (ipfw_dyn_v == NULL || dyn_count == 0)
691 /* do not expire more than once per second, it is useless */
692 if (!FORCE && last_remove == time_second)
694 last_remove = time_second;
697 * because O_LIMIT refer to parent rules, during the first pass only
698 * remove child and mark any pending LIMIT_PARENT, and remove
699 * them in a second pass.
702 for (i = 0 ; i < curr_dyn_buckets ; i++) {
703 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) {
705 * Logic can become complex here, so we split tests.
709 if (rule != NULL && rule != q->rule)
710 goto next; /* not the one we are looking for */
711 if (q->dyn_type == O_LIMIT_PARENT) {
713 * handle parent in the second pass,
714 * record we need one.
719 if (FORCE && q->count != 0 ) {
720 /* XXX should not happen! */
721 kprintf( "OUCH! cannot remove rule,"
722 " count %d\n", q->count);
726 !TIME_LEQ( q->expire, time_second ))
729 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
736 if (pass++ < max_pass)
742 * lookup a dynamic rule.
744 static ipfw_dyn_rule *
745 lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction,
749 * stateful ipfw extensions.
750 * Lookup into dynamic session queue
752 #define MATCH_REVERSE 0
753 #define MATCH_FORWARD 1
755 #define MATCH_UNKNOWN 3
756 int i, dir = MATCH_NONE;
757 ipfw_dyn_rule *prev, *q=NULL;
759 if (ipfw_dyn_v == NULL)
760 goto done; /* not found */
761 i = hash_packet( pkt );
762 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) {
763 if (q->dyn_type == O_LIMIT_PARENT)
765 if (TIME_LEQ( q->expire, time_second)) { /* expire entry */
766 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
769 if ( pkt->proto == q->id.proto) {
770 if (pkt->src_ip == q->id.src_ip &&
771 pkt->dst_ip == q->id.dst_ip &&
772 pkt->src_port == q->id.src_port &&
773 pkt->dst_port == q->id.dst_port ) {
777 if (pkt->src_ip == q->id.dst_ip &&
778 pkt->dst_ip == q->id.src_ip &&
779 pkt->src_port == q->id.dst_port &&
780 pkt->dst_port == q->id.src_port ) {
790 goto done; /* q = NULL, not found */
792 if ( prev != NULL) { /* found and not in front */
793 prev->next = q->next;
794 q->next = ipfw_dyn_v[i];
797 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */
798 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST);
800 #define BOTH_SYN (TH_SYN | (TH_SYN << 8))
801 #define BOTH_FIN (TH_FIN | (TH_FIN << 8))
802 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8);
804 case TH_SYN: /* opening */
805 q->expire = time_second + dyn_syn_lifetime;
808 case BOTH_SYN: /* move to established */
809 case BOTH_SYN | TH_FIN : /* one side tries to close */
810 case BOTH_SYN | (TH_FIN << 8) :
812 #define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0)
813 uint32_t ack = ntohl(tcp->th_ack);
814 if (dir == MATCH_FORWARD) {
815 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd))
817 else { /* ignore out-of-sequence */
821 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev))
823 else { /* ignore out-of-sequence */
828 q->expire = time_second + dyn_ack_lifetime;
831 case BOTH_SYN | BOTH_FIN: /* both sides closed */
832 if (dyn_fin_lifetime >= dyn_keepalive_period)
833 dyn_fin_lifetime = dyn_keepalive_period - 1;
834 q->expire = time_second + dyn_fin_lifetime;
840 * reset or some invalid combination, but can also
841 * occur if we use keep-state the wrong way.
843 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0)
844 kprintf("invalid state: 0x%x\n", q->state);
846 if (dyn_rst_lifetime >= dyn_keepalive_period)
847 dyn_rst_lifetime = dyn_keepalive_period - 1;
848 q->expire = time_second + dyn_rst_lifetime;
851 } else if (pkt->proto == IPPROTO_UDP) {
852 q->expire = time_second + dyn_udp_lifetime;
854 /* other protocols */
855 q->expire = time_second + dyn_short_lifetime;
859 *match_direction = dir;
864 realloc_dynamic_table(void)
867 * Try reallocation, make sure we have a power of 2 and do
868 * not allow more than 64k entries. In case of overflow,
872 if (dyn_buckets > 65536)
874 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */
875 dyn_buckets = curr_dyn_buckets; /* reset */
878 curr_dyn_buckets = dyn_buckets;
879 if (ipfw_dyn_v != NULL)
880 kfree(ipfw_dyn_v, M_IPFW);
882 ipfw_dyn_v = kmalloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *),
883 M_IPFW, M_INTWAIT | M_NULLOK | M_ZERO);
884 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2)
886 curr_dyn_buckets /= 2;
891 * Install state of type 'type' for a dynamic session.
892 * The hash table contains two type of rules:
893 * - regular rules (O_KEEP_STATE)
894 * - rules for sessions with limited number of sess per user
895 * (O_LIMIT). When they are created, the parent is
896 * increased by 1, and decreased on delete. In this case,
897 * the third parameter is the parent rule and not the chain.
898 * - "parent" rules for the above (O_LIMIT_PARENT).
900 static ipfw_dyn_rule *
901 add_dyn_rule(struct ipfw_flow_id *id, uint8_t dyn_type, struct ip_fw *rule)
906 if (ipfw_dyn_v == NULL ||
907 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) {
908 realloc_dynamic_table();
909 if (ipfw_dyn_v == NULL)
910 return NULL; /* failed ! */
914 r = kmalloc(sizeof *r, M_IPFW, M_INTWAIT | M_NULLOK | M_ZERO);
916 kprintf ("sorry cannot allocate state\n");
920 /* increase refcount on parent, and set pointer */
921 if (dyn_type == O_LIMIT) {
922 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule;
923 if ( parent->dyn_type != O_LIMIT_PARENT)
924 panic("invalid parent");
931 r->expire = time_second + dyn_syn_lifetime;
933 r->dyn_type = dyn_type;
934 r->pcnt = r->bcnt = 0;
938 r->next = ipfw_dyn_v[i];
941 DEB(kprintf("-- add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n",
943 (r->id.src_ip), (r->id.src_port),
944 (r->id.dst_ip), (r->id.dst_port),
950 * lookup dynamic parent rule using pkt and rule as search keys.
951 * If the lookup fails, then install one.
953 static ipfw_dyn_rule *
954 lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule)
960 i = hash_packet( pkt );
961 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next)
962 if (q->dyn_type == O_LIMIT_PARENT &&
964 pkt->proto == q->id.proto &&
965 pkt->src_ip == q->id.src_ip &&
966 pkt->dst_ip == q->id.dst_ip &&
967 pkt->src_port == q->id.src_port &&
968 pkt->dst_port == q->id.dst_port) {
969 q->expire = time_second + dyn_short_lifetime;
970 DEB(kprintf("lookup_dyn_parent found 0x%p\n",q);)
974 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule);
978 * Install dynamic state for rule type cmd->o.opcode
980 * Returns 1 (failure) if state is not installed because of errors or because
981 * session limitations are enforced.
984 install_state(struct ip_fw *rule, ipfw_insn_limit *cmd,
985 struct ip_fw_args *args)
991 DEB(kprintf("-- install state type %d 0x%08x %u -> 0x%08x %u\n",
993 (args->f_id.src_ip), (args->f_id.src_port),
994 (args->f_id.dst_ip), (args->f_id.dst_port) );)
996 q = lookup_dyn_rule(&args->f_id, NULL, NULL);
998 if (q != NULL) { /* should never occur */
999 if (last_log != time_second) {
1000 last_log = time_second;
1001 kprintf(" install_state: entry already present, done\n");
1006 if (dyn_count >= dyn_max)
1008 * Run out of slots, try to remove any expired rule.
1010 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1);
1012 if (dyn_count >= dyn_max) {
1013 if (last_log != time_second) {
1014 last_log = time_second;
1015 kprintf("install_state: Too many dynamic rules\n");
1017 return 1; /* cannot install, notify caller */
1020 switch (cmd->o.opcode) {
1021 case O_KEEP_STATE: /* bidir rule */
1022 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule);
1025 case O_LIMIT: /* limit number of sessions */
1027 uint16_t limit_mask = cmd->limit_mask;
1028 struct ipfw_flow_id id;
1029 ipfw_dyn_rule *parent;
1031 DEB(kprintf("installing dyn-limit rule %d\n", cmd->conn_limit);)
1033 id.dst_ip = id.src_ip = 0;
1034 id.dst_port = id.src_port = 0;
1035 id.proto = args->f_id.proto;
1037 if (limit_mask & DYN_SRC_ADDR)
1038 id.src_ip = args->f_id.src_ip;
1039 if (limit_mask & DYN_DST_ADDR)
1040 id.dst_ip = args->f_id.dst_ip;
1041 if (limit_mask & DYN_SRC_PORT)
1042 id.src_port = args->f_id.src_port;
1043 if (limit_mask & DYN_DST_PORT)
1044 id.dst_port = args->f_id.dst_port;
1045 parent = lookup_dyn_parent(&id, rule);
1046 if (parent == NULL) {
1047 kprintf("add parent failed\n");
1050 if (parent->count >= cmd->conn_limit) {
1052 * See if we can remove some expired rule.
1054 remove_dyn_rule(rule, parent);
1055 if (parent->count >= cmd->conn_limit) {
1056 if (fw_verbose && last_log != time_second) {
1057 last_log = time_second;
1058 log(LOG_SECURITY | LOG_DEBUG,
1059 "drop session, too many entries\n");
1064 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent);
1068 kprintf("unknown dynamic rule type %u\n", cmd->o.opcode);
1071 lookup_dyn_rule(&args->f_id, NULL, NULL); /* XXX just set lifetime */
1076 * Transmit a TCP packet, containing either a RST or a keepalive.
1077 * When flags & TH_RST, we are sending a RST packet, because of a
1078 * "reset" action matched the packet.
1079 * Otherwise we are sending a keepalive, and flags & TH_
1082 send_pkt(struct ipfw_flow_id *id, uint32_t seq, uint32_t ack, int flags)
1087 struct route sro; /* fake route */
1089 MGETHDR(m, MB_DONTWAIT, MT_HEADER);
1092 m->m_pkthdr.rcvif = (struct ifnet *)0;
1093 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr);
1094 m->m_data += max_linkhdr;
1096 ip = mtod(m, struct ip *);
1097 bzero(ip, m->m_len);
1098 tcp = (struct tcphdr *)(ip + 1); /* no IP options */
1099 ip->ip_p = IPPROTO_TCP;
1102 * Assume we are sending a RST (or a keepalive in the reverse
1103 * direction), swap src and destination addresses and ports.
1105 ip->ip_src.s_addr = htonl(id->dst_ip);
1106 ip->ip_dst.s_addr = htonl(id->src_ip);
1107 tcp->th_sport = htons(id->dst_port);
1108 tcp->th_dport = htons(id->src_port);
1109 if (flags & TH_RST) { /* we are sending a RST */
1110 if (flags & TH_ACK) {
1111 tcp->th_seq = htonl(ack);
1112 tcp->th_ack = htonl(0);
1113 tcp->th_flags = TH_RST;
1117 tcp->th_seq = htonl(0);
1118 tcp->th_ack = htonl(seq);
1119 tcp->th_flags = TH_RST | TH_ACK;
1123 * We are sending a keepalive. flags & TH_SYN determines
1124 * the direction, forward if set, reverse if clear.
1125 * NOTE: seq and ack are always assumed to be correct
1126 * as set by the caller. This may be confusing...
1128 if (flags & TH_SYN) {
1130 * we have to rewrite the correct addresses!
1132 ip->ip_dst.s_addr = htonl(id->dst_ip);
1133 ip->ip_src.s_addr = htonl(id->src_ip);
1134 tcp->th_dport = htons(id->dst_port);
1135 tcp->th_sport = htons(id->src_port);
1137 tcp->th_seq = htonl(seq);
1138 tcp->th_ack = htonl(ack);
1139 tcp->th_flags = TH_ACK;
1142 * set ip_len to the payload size so we can compute
1143 * the tcp checksum on the pseudoheader
1144 * XXX check this, could save a couple of words ?
1146 ip->ip_len = htons(sizeof(struct tcphdr));
1147 tcp->th_sum = in_cksum(m, m->m_pkthdr.len);
1149 * now fill fields left out earlier
1151 ip->ip_ttl = ip_defttl;
1152 ip->ip_len = m->m_pkthdr.len;
1153 bzero (&sro, sizeof (sro));
1154 ip_rtaddr(ip->ip_dst, &sro);
1155 m->m_pkthdr.fw_flags |= IPFW_MBUF_GENERATED;
1156 ip_output(m, NULL, &sro, 0, NULL, NULL);
1162 * sends a reject message, consuming the mbuf passed as an argument.
1165 send_reject(struct ip_fw_args *args, int code, int offset, int ip_len)
1168 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
1169 /* We need the IP header in host order for icmp_error(). */
1170 if (args->eh != NULL) {
1171 struct ip *ip = mtod(args->m, struct ip *);
1172 ip->ip_len = ntohs(ip->ip_len);
1173 ip->ip_off = ntohs(ip->ip_off);
1175 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
1176 } else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) {
1177 struct tcphdr *const tcp =
1178 L3HDR(struct tcphdr, mtod(args->m, struct ip *));
1179 if ( (tcp->th_flags & TH_RST) == 0)
1180 send_pkt(&(args->f_id), ntohl(tcp->th_seq),
1182 tcp->th_flags | TH_RST);
1191 * Given an ip_fw *, lookup_next_rule will return a pointer
1192 * to the next rule, which can be either the jump
1193 * target (for skipto instructions) or the next one in the list (in
1194 * all other cases including a missing jump target).
1195 * The result is also written in the "next_rule" field of the rule.
1196 * Backward jumps are not allowed, so start looking from the next
1199 * This never returns NULL -- in case we do not have an exact match,
1200 * the next rule is returned. When the ruleset is changed,
1201 * pointers are flushed so we are always correct.
1204 static struct ip_fw *
1205 lookup_next_rule(struct ip_fw *me)
1207 struct ip_fw *rule = NULL;
1210 /* look for action, in case it is a skipto */
1211 cmd = ACTION_PTR(me);
1212 if (cmd->opcode == O_LOG)
1214 if ( cmd->opcode == O_SKIPTO )
1215 for (rule = me->next; rule ; rule = rule->next)
1216 if (rule->rulenum >= cmd->arg1)
1218 if (rule == NULL) /* failure or not a skipto */
1220 me->next_rule = rule;
1225 * The main check routine for the firewall.
1227 * All arguments are in args so we can modify them and return them
1228 * back to the caller.
1232 * args->m (in/out) The packet; we set to NULL when/if we nuke it.
1233 * Starts with the IP header.
1234 * args->eh (in) Mac header if present, or NULL for layer3 packet.
1235 * args->oif Outgoing interface, or NULL if packet is incoming.
1236 * The incoming interface is in the mbuf. (in)
1238 * args->rule Pointer to the last matching rule (in/out)
1239 * args->next_hop Socket we are forwarding to (out).
1240 * args->f_id Addresses grabbed from the packet (out)
1244 * IP_FW_PORT_DENY_FLAG the packet must be dropped.
1245 * 0 The packet is to be accepted and routed normally OR
1246 * the packet was denied/rejected and has been dropped;
1247 * in the latter case, *m is equal to NULL upon return.
1248 * port Divert the packet to port, with these caveats:
1250 * - If IP_FW_PORT_TEE_FLAG is set, tee the packet instead
1251 * of diverting it (ie, 'ipfw tee').
1253 * - If IP_FW_PORT_DYNT_FLAG is set, interpret the lower
1254 * 16 bits as a dummynet pipe number instead of diverting
1258 ipfw_chk(struct ip_fw_args *args)
1261 * Local variables hold state during the processing of a packet.
1263 * IMPORTANT NOTE: to speed up the processing of rules, there
1264 * are some assumption on the values of the variables, which
1265 * are documented here. Should you change them, please check
1266 * the implementation of the various instructions to make sure
1267 * that they still work.
1269 * args->eh The MAC header. It is non-null for a layer2
1270 * packet, it is NULL for a layer-3 packet.
1272 * m | args->m Pointer to the mbuf, as received from the caller.
1273 * It may change if ipfw_chk() does an m_pullup, or if it
1274 * consumes the packet because it calls send_reject().
1275 * XXX This has to change, so that ipfw_chk() never modifies
1276 * or consumes the buffer.
1277 * ip is simply an alias of the value of m, and it is kept
1278 * in sync with it (the packet is supposed to start with
1281 struct mbuf *m = args->m;
1282 struct ip *ip = mtod(m, struct ip *);
1285 * oif | args->oif If NULL, ipfw_chk has been called on the
1286 * inbound path (ether_input, ip_input).
1287 * If non-NULL, ipfw_chk has been called on the outbound path
1288 * (ether_output, ip_output).
1290 struct ifnet *oif = args->oif;
1292 struct ip_fw *f = NULL; /* matching rule */
1297 * hlen The length of the IPv4 header.
1298 * hlen >0 means we have an IPv4 packet.
1300 u_int hlen = 0; /* hlen >0 means we have an IP pkt */
1303 * offset The offset of a fragment. offset != 0 means that
1304 * we have a fragment at this offset of an IPv4 packet.
1305 * offset == 0 means that (if this is an IPv4 packet)
1306 * this is the first or only fragment.
1311 * Local copies of addresses. They are only valid if we have
1314 * proto The protocol. Set to 0 for non-ip packets,
1315 * or to the protocol read from the packet otherwise.
1316 * proto != 0 means that we have an IPv4 packet.
1318 * src_port, dst_port port numbers, in HOST format. Only
1319 * valid for TCP and UDP packets.
1321 * src_ip, dst_ip ip addresses, in NETWORK format.
1322 * Only valid for IPv4 packets.
1325 uint16_t src_port = 0, dst_port = 0; /* NOTE: host format */
1326 struct in_addr src_ip, dst_ip; /* NOTE: network format */
1328 int dyn_dir = MATCH_UNKNOWN;
1329 ipfw_dyn_rule *q = NULL;
1331 if (m->m_pkthdr.fw_flags & IPFW_MBUF_GENERATED)
1332 return 0; /* accept */
1334 * dyn_dir = MATCH_UNKNOWN when rules unchecked,
1335 * MATCH_NONE when checked and not matched (q = NULL),
1336 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
1339 if (args->eh == NULL || /* layer 3 packet */
1340 ( m->m_pkthdr.len >= sizeof(struct ip) &&
1341 ntohs(args->eh->ether_type) == ETHERTYPE_IP))
1342 hlen = ip->ip_hl << 2;
1345 * Collect parameters into local variables for faster matching.
1347 if (hlen == 0) { /* do not grab addresses for non-ip pkts */
1348 proto = args->f_id.proto = 0; /* mark f_id invalid */
1349 goto after_ip_checks;
1352 proto = args->f_id.proto = ip->ip_p;
1353 src_ip = ip->ip_src;
1354 dst_ip = ip->ip_dst;
1355 if (args->eh != NULL) { /* layer 2 packets are as on the wire */
1356 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1357 ip_len = ntohs(ip->ip_len);
1359 offset = ip->ip_off & IP_OFFMASK;
1360 ip_len = ip->ip_len;
1363 #define PULLUP_TO(len) \
1365 if ((m)->m_len < (len)) { \
1366 args->m = m = m_pullup(m, (len)); \
1368 goto pullup_failed; \
1369 ip = mtod(m, struct ip *); \
1379 PULLUP_TO(hlen + sizeof(struct tcphdr));
1380 tcp = L3HDR(struct tcphdr, ip);
1381 dst_port = tcp->th_dport;
1382 src_port = tcp->th_sport;
1383 args->f_id.flags = tcp->th_flags;
1391 PULLUP_TO(hlen + sizeof(struct udphdr));
1392 udp = L3HDR(struct udphdr, ip);
1393 dst_port = udp->uh_dport;
1394 src_port = udp->uh_sport;
1399 PULLUP_TO(hlen + 4); /* type, code and checksum. */
1400 args->f_id.flags = L3HDR(struct icmp, ip)->icmp_type;
1409 args->f_id.src_ip = ntohl(src_ip.s_addr);
1410 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1411 args->f_id.src_port = src_port = ntohs(src_port);
1412 args->f_id.dst_port = dst_port = ntohs(dst_port);
1417 * Packet has already been tagged. Look for the next rule
1418 * to restart processing.
1420 * If fw_one_pass != 0 then just accept it.
1421 * XXX should not happen here, but optimized out in
1427 /* This rule was deleted */
1428 if (args->rule->rule_flags & IPFW_RULE_F_INVALID)
1429 return IP_FW_PORT_DENY_FLAG;
1431 f = args->rule->next_rule;
1433 f = lookup_next_rule(args->rule);
1436 * Find the starting rule. It can be either the first
1437 * one, or the one after divert_rule if asked so.
1441 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
1443 skipto = *(uint16_t *)m_tag_data(mtag);
1448 if (args->eh == NULL && skipto != 0) {
1449 if (skipto >= IPFW_DEFAULT_RULE)
1450 return(IP_FW_PORT_DENY_FLAG); /* invalid */
1451 while (f && f->rulenum <= skipto)
1453 if (f == NULL) /* drop packet */
1454 return(IP_FW_PORT_DENY_FLAG);
1457 if ((mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
1458 m_tag_delete(m, mtag);
1461 * Now scan the rules, and parse microinstructions for each rule.
1463 for (; f; f = f->next) {
1466 int skip_or; /* skip rest of OR block */
1469 if (set_disable & (1 << f->set) )
1473 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1474 l -= cmdlen, cmd += cmdlen) {
1478 * check_body is a jump target used when we find a
1479 * CHECK_STATE, and need to jump to the body of
1484 cmdlen = F_LEN(cmd);
1486 * An OR block (insn_1 || .. || insn_n) has the
1487 * F_OR bit set in all but the last instruction.
1488 * The first match will set "skip_or", and cause
1489 * the following instructions to be skipped until
1490 * past the one with the F_OR bit clear.
1492 if (skip_or) { /* skip this instruction */
1493 if ((cmd->len & F_OR) == 0)
1494 skip_or = 0; /* next one is good */
1497 match = 0; /* set to 1 if we succeed */
1499 switch (cmd->opcode) {
1501 * The first set of opcodes compares the packet's
1502 * fields with some pattern, setting 'match' if a
1503 * match is found. At the end of the loop there is
1504 * logic to deal with F_NOT and F_OR flags associated
1512 kprintf("ipfw: opcode %d unimplemented\n",
1519 * We only check offset == 0 && proto != 0,
1520 * as this ensures that we have an IPv4
1521 * packet with the ports info.
1526 struct inpcbinfo *pi;
1530 if (proto == IPPROTO_TCP) {
1532 pi = &tcbinfo[mycpu->gd_cpuid];
1533 } else if (proto == IPPROTO_UDP) {
1540 in_pcblookup_hash(pi,
1541 dst_ip, htons(dst_port),
1542 src_ip, htons(src_port),
1544 in_pcblookup_hash(pi,
1545 src_ip, htons(src_port),
1546 dst_ip, htons(dst_port),
1549 if (pcb == NULL || pcb->inp_socket == NULL)
1551 #if defined(__DragonFly__) || (defined(__FreeBSD__) && __FreeBSD_version < 500034)
1552 #define socheckuid(a,b) ((a)->so_cred->cr_uid != (b))
1554 if (cmd->opcode == O_UID) {
1556 !socheckuid(pcb->inp_socket,
1557 (uid_t)((ipfw_insn_u32 *)cmd)->d[0]);
1559 match = groupmember(
1560 (uid_t)((ipfw_insn_u32 *)cmd)->d[0],
1561 pcb->inp_socket->so_cred);
1567 match = iface_match(m->m_pkthdr.rcvif,
1568 (ipfw_insn_if *)cmd);
1572 match = iface_match(oif, (ipfw_insn_if *)cmd);
1576 match = iface_match(oif ? oif :
1577 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd);
1581 if (args->eh != NULL) { /* have MAC header */
1582 uint32_t *want = (uint32_t *)
1583 ((ipfw_insn_mac *)cmd)->addr;
1584 uint32_t *mask = (uint32_t *)
1585 ((ipfw_insn_mac *)cmd)->mask;
1586 uint32_t *hdr = (uint32_t *)args->eh;
1589 ( want[0] == (hdr[0] & mask[0]) &&
1590 want[1] == (hdr[1] & mask[1]) &&
1591 want[2] == (hdr[2] & mask[2]) );
1596 if (args->eh != NULL) {
1598 ntohs(args->eh->ether_type);
1600 ((ipfw_insn_u16 *)cmd)->ports;
1603 for (i = cmdlen - 1; !match && i>0;
1605 match = (t>=p[0] && t<=p[1]);
1610 match = (hlen > 0 && offset != 0);
1613 case O_IN: /* "out" is "not in" */
1614 match = (oif == NULL);
1618 match = (args->eh != NULL);
1623 * We do not allow an arg of 0 so the
1624 * check of "proto" only suffices.
1626 match = (proto == cmd->arg1);
1630 match = (hlen > 0 &&
1631 ((ipfw_insn_ip *)cmd)->addr.s_addr ==
1636 match = (hlen > 0 &&
1637 ((ipfw_insn_ip *)cmd)->addr.s_addr ==
1639 ((ipfw_insn_ip *)cmd)->mask.s_addr));
1646 INADDR_TO_IFP(src_ip, tif);
1647 match = (tif != NULL);
1654 uint32_t *d = (uint32_t *)(cmd+1);
1656 cmd->opcode == O_IP_DST_SET ?
1662 addr -= d[0]; /* subtract base */
1663 match = (addr < cmd->arg1) &&
1664 ( d[ 1 + (addr>>5)] &
1665 (1<<(addr & 0x1f)) );
1670 match = (hlen > 0 &&
1671 ((ipfw_insn_ip *)cmd)->addr.s_addr ==
1676 match = (hlen > 0) &&
1677 (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1679 ((ipfw_insn_ip *)cmd)->mask.s_addr));
1686 INADDR_TO_IFP(dst_ip, tif);
1687 match = (tif != NULL);
1694 * offset == 0 && proto != 0 is enough
1695 * to guarantee that we have an IPv4
1696 * packet with port info.
1698 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
1701 (cmd->opcode == O_IP_SRCPORT) ?
1702 src_port : dst_port ;
1704 ((ipfw_insn_u16 *)cmd)->ports;
1707 for (i = cmdlen - 1; !match && i>0;
1709 match = (x>=p[0] && x<=p[1]);
1714 match = (offset == 0 && proto==IPPROTO_ICMP &&
1715 icmptype_match(ip, (ipfw_insn_u32 *)cmd) );
1719 match = (hlen > 0 && ipopts_match(ip, cmd) );
1723 match = (hlen > 0 && cmd->arg1 == ip->ip_v);
1727 match = (hlen > 0 && cmd->arg1 == ip->ip_ttl);
1731 match = (hlen > 0 &&
1732 cmd->arg1 == ntohs(ip->ip_id));
1736 match = (hlen > 0 && cmd->arg1 == ip_len);
1739 case O_IPPRECEDENCE:
1740 match = (hlen > 0 &&
1741 (cmd->arg1 == (ip->ip_tos & 0xe0)) );
1745 match = (hlen > 0 &&
1746 flags_match(cmd, ip->ip_tos));
1750 match = (proto == IPPROTO_TCP && offset == 0 &&
1752 L3HDR(struct tcphdr,ip)->th_flags));
1756 match = (proto == IPPROTO_TCP && offset == 0 &&
1757 tcpopts_match(ip, cmd));
1761 match = (proto == IPPROTO_TCP && offset == 0 &&
1762 ((ipfw_insn_u32 *)cmd)->d[0] ==
1763 L3HDR(struct tcphdr,ip)->th_seq);
1767 match = (proto == IPPROTO_TCP && offset == 0 &&
1768 ((ipfw_insn_u32 *)cmd)->d[0] ==
1769 L3HDR(struct tcphdr,ip)->th_ack);
1773 match = (proto == IPPROTO_TCP && offset == 0 &&
1775 L3HDR(struct tcphdr,ip)->th_win);
1779 /* reject packets which have SYN only */
1780 /* XXX should i also check for TH_ACK ? */
1781 match = (proto == IPPROTO_TCP && offset == 0 &&
1782 (L3HDR(struct tcphdr,ip)->th_flags &
1783 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
1788 ipfw_log(f, hlen, args->eh, m, oif);
1793 match = (krandom() <
1794 ((ipfw_insn_u32 *)cmd)->d[0]);
1798 * The second set of opcodes represents 'actions',
1799 * i.e. the terminal part of a rule once the packet
1800 * matches all previous patterns.
1801 * Typically there is only one action for each rule,
1802 * and the opcode is stored at the end of the rule
1803 * (but there are exceptions -- see below).
1805 * In general, here we set retval and terminate the
1806 * outer loop (would be a 'break 3' in some language,
1807 * but we need to do a 'goto done').
1810 * O_COUNT and O_SKIPTO actions:
1811 * instead of terminating, we jump to the next rule
1812 * ('goto next_rule', equivalent to a 'break 2'),
1813 * or to the SKIPTO target ('goto again' after
1814 * having set f, cmd and l), respectively.
1816 * O_LIMIT and O_KEEP_STATE: these opcodes are
1817 * not real 'actions', and are stored right
1818 * before the 'action' part of the rule.
1819 * These opcodes try to install an entry in the
1820 * state tables; if successful, we continue with
1821 * the next opcode (match=1; break;), otherwise
1822 * the packet * must be dropped
1823 * ('goto done' after setting retval);
1825 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
1826 * cause a lookup of the state table, and a jump
1827 * to the 'action' part of the parent rule
1828 * ('goto check_body') if an entry is found, or
1829 * (CHECK_STATE only) a jump to the next rule if
1830 * the entry is not found ('goto next_rule').
1831 * The result of the lookup is cached to make
1832 * further instances of these opcodes are
1837 if (install_state(f,
1838 (ipfw_insn_limit *)cmd, args)) {
1839 retval = IP_FW_PORT_DENY_FLAG;
1840 goto done; /* error/limit violation */
1848 * dynamic rules are checked at the first
1849 * keep-state or check-state occurrence,
1850 * with the result being stored in dyn_dir.
1851 * The compiler introduces a PROBE_STATE
1852 * instruction for us when we have a
1853 * KEEP_STATE (because PROBE_STATE needs
1856 if (dyn_dir == MATCH_UNKNOWN &&
1857 (q = lookup_dyn_rule(&args->f_id,
1858 &dyn_dir, proto == IPPROTO_TCP ?
1859 L3HDR(struct tcphdr, ip) : NULL))
1862 * Found dynamic entry, update stats
1863 * and jump to the 'action' part of
1869 cmd = ACTION_PTR(f);
1870 l = f->cmd_len - f->act_ofs;
1874 * Dynamic entry not found. If CHECK_STATE,
1875 * skip to next rule, if PROBE_STATE just
1876 * ignore and continue with next opcode.
1878 if (cmd->opcode == O_CHECK_STATE)
1884 retval = 0; /* accept */
1889 args->rule = f; /* report matching rule */
1890 retval = cmd->arg1 | IP_FW_PORT_DYNT_FLAG;
1895 if (args->eh) /* not on layer 2 */
1898 mtag = m_tag_get(PACKET_TAG_IPFW_DIVERT,
1899 sizeof(uint16_t), MB_DONTWAIT);
1901 retval = IP_FW_PORT_DENY_FLAG;
1904 *(uint16_t *)m_tag_data(mtag) = f->rulenum;
1905 m_tag_prepend(m, mtag);
1906 retval = (cmd->opcode == O_DIVERT) ?
1908 cmd->arg1 | IP_FW_PORT_TEE_FLAG;
1913 f->pcnt++; /* update stats */
1915 f->timestamp = time_second;
1916 if (cmd->opcode == O_COUNT)
1919 if (f->next_rule == NULL)
1920 lookup_next_rule(f);
1926 * Drop the packet and send a reject notice
1927 * if the packet is not ICMP (or is an ICMP
1928 * query), and it is not multicast/broadcast.
1931 (proto != IPPROTO_ICMP ||
1932 is_icmp_query(ip)) &&
1933 !(m->m_flags & (M_BCAST|M_MCAST)) &&
1934 !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
1935 send_reject(args, cmd->arg1,
1941 retval = IP_FW_PORT_DENY_FLAG;
1945 if (args->eh) /* not valid on layer2 pkts */
1947 if (!q || dyn_dir == MATCH_FORWARD)
1949 &((ipfw_insn_sa *)cmd)->sa;
1954 panic("-- unknown opcode %d\n", cmd->opcode);
1955 } /* end of switch() on opcodes */
1957 if (cmd->len & F_NOT)
1961 if (cmd->len & F_OR)
1964 if (!(cmd->len & F_OR)) /* not an OR block, */
1965 break; /* try next rule */
1968 } /* end of inner for, scan opcodes */
1970 next_rule:; /* try next rule */
1972 } /* end of outer for, scan rules */
1973 kprintf("+++ ipfw: ouch!, skip past end of rules, denying packet\n");
1974 return(IP_FW_PORT_DENY_FLAG);
1977 /* Update statistics */
1980 f->timestamp = time_second;
1985 kprintf("pullup failed\n");
1986 return(IP_FW_PORT_DENY_FLAG);
1990 ipfw_dummynet_io(struct mbuf *m, int pipe_nr, int dir, struct ip_fw_args *fwa)
1995 const struct ipfw_flow_id *id;
1996 struct dn_flow_id *fid;
2000 mtag = m_tag_get(PACKET_TAG_DUMMYNET, sizeof(*pkt), MB_DONTWAIT);
2005 m_tag_prepend(m, mtag);
2007 pkt = m_tag_data(mtag);
2008 bzero(pkt, sizeof(*pkt));
2010 cmd = fwa->rule->cmd + fwa->rule->act_ofs;
2011 if (cmd->opcode == O_LOG)
2013 KASSERT(cmd->opcode == O_PIPE || cmd->opcode == O_QUEUE,
2014 ("Rule is not PIPE or QUEUE, opcode %d\n", cmd->opcode));
2017 pkt->dn_flags = (dir & DN_FLAGS_DIR_MASK);
2018 pkt->ifp = fwa->oif;
2019 pkt->cpuid = mycpu->gd_cpuid;
2020 pkt->pipe_nr = pipe_nr;
2024 fid->fid_dst_ip = id->dst_ip;
2025 fid->fid_src_ip = id->src_ip;
2026 fid->fid_dst_port = id->dst_port;
2027 fid->fid_src_port = id->src_port;
2028 fid->fid_proto = id->proto;
2029 fid->fid_flags = id->flags;
2031 ipfw_ref_rule(fwa->rule);
2032 pkt->dn_priv = fwa->rule;
2033 pkt->dn_unref_priv = ipfw_unref_rule;
2035 if (cmd->opcode == O_PIPE)
2036 pkt->dn_flags |= DN_FLAGS_IS_PIPE;
2038 if (dir == DN_TO_IP_OUT) {
2040 * We need to copy *ro because for ICMP pkts (and maybe
2041 * others) the caller passed a pointer into the stack;
2042 * dst might also be a pointer into *ro so it needs to
2045 pkt->ro = *(fwa->ro);
2047 fwa->ro->ro_rt->rt_refcnt++;
2048 if (fwa->dst == (struct sockaddr_in *)&fwa->ro->ro_dst) {
2049 /* 'dst' points into 'ro' */
2050 fwa->dst = (struct sockaddr_in *)&(pkt->ro.ro_dst);
2052 pkt->dn_dst = fwa->dst;
2053 pkt->flags = fwa->flags;
2056 m->m_pkthdr.fw_flags |= DUMMYNET_MBUF_TAGGED;
2061 * When a rule is added/deleted, clear the next_rule pointers in all rules.
2062 * These will be reconstructed on the fly as packets are matched.
2063 * Must be called at splimp().
2066 flush_rule_ptrs(void)
2070 for (rule = layer3_chain; rule; rule = rule->next)
2071 rule->next_rule = NULL;
2074 static __inline void
2075 ipfw_inc_static_count(struct ip_fw *rule)
2078 static_ioc_len += IOC_RULESIZE(rule);
2081 static __inline void
2082 ipfw_dec_static_count(struct ip_fw *rule)
2084 int l = IOC_RULESIZE(rule);
2086 KASSERT(static_count > 0, ("invalid static count %u\n", static_count));
2089 KASSERT(static_ioc_len >= l,
2090 ("invalid static len %u\n", static_ioc_len));
2091 static_ioc_len -= l;
2094 static struct ip_fw *
2095 ipfw_create_rule(const struct ipfw_ioc_rule *ioc_rule)
2099 rule = kmalloc(RULESIZE(ioc_rule), M_IPFW, M_WAITOK | M_ZERO);
2101 rule->act_ofs = ioc_rule->act_ofs;
2102 rule->cmd_len = ioc_rule->cmd_len;
2103 rule->rulenum = ioc_rule->rulenum;
2104 rule->set = ioc_rule->set;
2105 rule->usr_flags = ioc_rule->usr_flags;
2107 bcopy(ioc_rule->cmd, rule->cmd, rule->cmd_len * 4 /* XXX */);
2115 * Add a new rule to the list. Copy the rule into a malloc'ed area, then
2116 * possibly create a rule number and add the rule to the list.
2117 * Update the rule_number in the input struct so the caller knows it as well.
2120 ipfw_add_rule(struct ip_fw **head, struct ipfw_ioc_rule *ioc_rule)
2122 struct ip_fw *rule, *f, *prev;
2124 KKASSERT(*head != NULL);
2126 rule = ipfw_create_rule(ioc_rule);
2131 * If rulenum is 0, find highest numbered rule before the
2132 * default rule, and add autoinc_step
2134 if (autoinc_step < 1)
2136 else if (autoinc_step > 1000)
2137 autoinc_step = 1000;
2138 if (rule->rulenum == 0) {
2140 * locate the highest numbered rule before default
2142 for (f = *head; f; f = f->next) {
2143 if (f->rulenum == IPFW_DEFAULT_RULE)
2145 rule->rulenum = f->rulenum;
2147 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step)
2148 rule->rulenum += autoinc_step;
2149 ioc_rule->rulenum = rule->rulenum;
2153 * Now insert the new rule in the right place in the sorted list.
2155 for (prev = NULL, f = *head; f; prev = f, f = f->next) {
2156 if (f->rulenum > rule->rulenum) { /* found the location */
2160 } else { /* head insert */
2169 ipfw_inc_static_count(rule);
2173 DEB(kprintf("++ installed rule %d, static count now %d\n",
2174 rule->rulenum, static_count);)
2179 * Free storage associated with a static rule (including derived
2181 * The caller is in charge of clearing rule pointers to avoid
2182 * dangling pointers.
2183 * @return a pointer to the next entry.
2184 * Arguments are not checked, so they better be correct.
2185 * Must be called at splimp().
2187 static struct ip_fw *
2188 delete_rule(struct ip_fw **head, struct ip_fw *prev, struct ip_fw *rule)
2193 remove_dyn_rule(rule, NULL /* force removal */);
2198 ipfw_dec_static_count(rule);
2200 /* Mark the rule as invalid */
2201 rule->rule_flags |= IPFW_RULE_F_INVALID;
2202 rule->next_rule = NULL;
2204 /* Try to free this rule */
2205 ipfw_free_rule(rule);
2211 * Deletes all rules from a chain (including the default rule
2212 * if the second argument is set).
2213 * Must be called at splimp().
2216 free_chain(struct ip_fw **chain, int kill_default)
2220 flush_rule_ptrs(); /* more efficient to do outside the loop */
2222 while ( (rule = *chain) != NULL &&
2223 (kill_default || rule->rulenum != IPFW_DEFAULT_RULE) )
2224 delete_rule(chain, NULL, rule);
2226 KASSERT(dyn_count == 0, ("%u dyn rule remains\n", dyn_count));
2229 ip_fw_default_rule = NULL; /* Reset default rule */
2231 if (ipfw_dyn_v != NULL) {
2233 * Free dynamic rules(state) hash table
2235 kfree(ipfw_dyn_v, M_IPFW);
2239 KASSERT(static_count == 0,
2240 ("%u static rules remains\n", static_count));
2241 KASSERT(static_ioc_len == 0,
2242 ("%u bytes of static rules remains\n", static_ioc_len));
2244 KASSERT(static_count == 1,
2245 ("%u static rules remains\n", static_count));
2246 KASSERT(static_ioc_len == IOC_RULESIZE(ip_fw_default_rule),
2247 ("%u bytes of static rules remains, should be %u\n",
2248 static_ioc_len, IOC_RULESIZE(ip_fw_default_rule)));
2253 * Remove all rules with given number, and also do set manipulation.
2255 * The argument is an uint32_t. The low 16 bit are the rule or set number,
2256 * the next 8 bits are the new set, the top 8 bits are the command:
2258 * 0 delete rules with given number
2259 * 1 delete rules with given set number
2260 * 2 move rules with given number to new set
2261 * 3 move rules with given set number to new set
2262 * 4 swap sets with given numbers
2265 del_entry(struct ip_fw **chain, uint32_t arg)
2267 struct ip_fw *prev, *rule;
2269 uint8_t cmd, new_set;
2271 rulenum = arg & 0xffff;
2272 cmd = (arg >> 24) & 0xff;
2273 new_set = (arg >> 16) & 0xff;
2279 if (cmd == 0 || cmd == 2) {
2280 if (rulenum == IPFW_DEFAULT_RULE)
2288 case 0: /* delete rules with given number */
2290 * locate first rule to delete
2292 for (prev = NULL, rule = *chain;
2293 rule && rule->rulenum < rulenum;
2294 prev = rule, rule = rule->next)
2296 if (rule->rulenum != rulenum)
2299 crit_enter(); /* no access to rules while removing */
2301 * flush pointers outside the loop, then delete all matching
2302 * rules. prev remains the same throughout the cycle.
2305 while (rule && rule->rulenum == rulenum)
2306 rule = delete_rule(chain, prev, rule);
2310 case 1: /* delete all rules with given set number */
2313 for (prev = NULL, rule = *chain; rule ; )
2314 if (rule->set == rulenum)
2315 rule = delete_rule(chain, prev, rule);
2323 case 2: /* move rules with given number to new set */
2325 for (rule = *chain; rule ; rule = rule->next)
2326 if (rule->rulenum == rulenum)
2327 rule->set = new_set;
2331 case 3: /* move rules with given set number to new set */
2333 for (rule = *chain; rule ; rule = rule->next)
2334 if (rule->set == rulenum)
2335 rule->set = new_set;
2339 case 4: /* swap two sets */
2341 for (rule = *chain; rule ; rule = rule->next)
2342 if (rule->set == rulenum)
2343 rule->set = new_set;
2344 else if (rule->set == new_set)
2345 rule->set = rulenum;
2353 * Clear counters for a specific rule.
2356 clear_counters(struct ip_fw *rule, int log_only)
2358 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule);
2360 if (log_only == 0) {
2361 rule->bcnt = rule->pcnt = 0;
2362 rule->timestamp = 0;
2364 if (l->o.opcode == O_LOG)
2365 l->log_left = l->max_log;
2369 * Reset some or all counters on firewall rules.
2370 * @arg frwl is null to clear all entries, or contains a specific
2372 * @arg log_only is 1 if we only want to reset logs, zero otherwise.
2375 zero_entry(int rulenum, int log_only)
2383 for (rule = layer3_chain; rule; rule = rule->next)
2384 clear_counters(rule, log_only);
2386 msg = log_only ? "ipfw: All logging counts reset.\n" :
2387 "ipfw: Accounting cleared.\n";
2391 * We can have multiple rules with the same number, so we
2392 * need to clear them all.
2394 for (rule = layer3_chain; rule; rule = rule->next)
2395 if (rule->rulenum == rulenum) {
2397 while (rule && rule->rulenum == rulenum) {
2398 clear_counters(rule, log_only);
2405 if (!cleared) /* we did not find any matching rules */
2407 msg = log_only ? "ipfw: Entry %d logging count reset.\n" :
2408 "ipfw: Entry %d cleared.\n";
2411 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum);
2416 * Check validity of the structure before insert.
2417 * Fortunately rules are simple, so this mostly need to check rule sizes.
2420 ipfw_ctl_check_rule(struct ipfw_ioc_rule *rule, int size)
2423 int have_action = 0;
2426 /* Check for valid size */
2427 if (size < sizeof(*rule)) {
2428 kprintf("ipfw: rule too short\n");
2431 l = IOC_RULESIZE(rule);
2433 kprintf("ipfw: size mismatch (have %d want %d)\n", size, l);
2438 * Now go for the individual checks. Very simple ones, basically only
2439 * instruction sizes.
2441 for (l = rule->cmd_len, cmd = rule->cmd; l > 0;
2442 l -= cmdlen, cmd += cmdlen) {
2443 cmdlen = F_LEN(cmd);
2445 kprintf("ipfw: opcode %d size truncated\n",
2449 DEB(kprintf("ipfw: opcode %d\n", cmd->opcode);)
2450 switch (cmd->opcode) {
2464 case O_IPPRECEDENCE:
2471 if (cmdlen != F_INSN_SIZE(ipfw_insn))
2483 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32))
2488 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit))
2493 if (cmdlen != F_INSN_SIZE(ipfw_insn_log))
2496 ((ipfw_insn_log *)cmd)->log_left =
2497 ((ipfw_insn_log *)cmd)->max_log;
2503 if (cmdlen != F_INSN_SIZE(ipfw_insn_ip))
2505 if (((ipfw_insn_ip *)cmd)->mask.s_addr == 0) {
2506 kprintf("ipfw: opcode %d, useless rule\n",
2514 if (cmd->arg1 == 0 || cmd->arg1 > 256) {
2515 kprintf("ipfw: invalid set size %d\n",
2519 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
2525 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac))
2531 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */
2532 if (cmdlen < 2 || cmdlen > 31)
2539 if (cmdlen != F_INSN_SIZE(ipfw_insn_if))
2545 if (cmdlen != F_INSN_SIZE(ipfw_insn_pipe))
2550 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa))
2554 case O_FORWARD_MAC: /* XXX not implemented yet */
2563 if (cmdlen != F_INSN_SIZE(ipfw_insn))
2567 kprintf("ipfw: opcode %d, multiple actions"
2574 kprintf("ipfw: opcode %d, action must be"
2581 kprintf("ipfw: opcode %d, unknown opcode\n",
2586 if (have_action == 0) {
2587 kprintf("ipfw: missing action\n");
2593 kprintf("ipfw: opcode %d size %d wrong\n",
2594 cmd->opcode, cmdlen);
2599 ipfw_ctl_add_rule(struct sockopt *sopt)
2601 struct ipfw_ioc_rule *ioc_rule;
2602 uint32_t rule_buf[IPFW_RULE_SIZE_MAX];
2606 ioc_rule = (struct ipfw_ioc_rule *)rule_buf;
2607 error = sooptcopyin(sopt, ioc_rule, sizeof(rule_buf),
2612 size = sopt->sopt_valsize;
2613 error = ipfw_ctl_check_rule(ioc_rule, size);
2617 error = ipfw_add_rule(&layer3_chain, ioc_rule);
2621 if (sopt->sopt_dir == SOPT_GET)
2622 error = sooptcopyout(sopt, ioc_rule, IOC_RULESIZE(ioc_rule));
2627 ipfw_copy_rule(const struct ip_fw *rule, struct ipfw_ioc_rule *ioc_rule)
2629 ioc_rule->act_ofs = rule->act_ofs;
2630 ioc_rule->cmd_len = rule->cmd_len;
2631 ioc_rule->rulenum = rule->rulenum;
2632 ioc_rule->set = rule->set;
2633 ioc_rule->usr_flags = rule->usr_flags;
2635 ioc_rule->set_disable = set_disable;
2636 ioc_rule->static_count = static_count;
2637 ioc_rule->static_len = static_ioc_len;
2639 ioc_rule->pcnt = rule->pcnt;
2640 ioc_rule->bcnt = rule->bcnt;
2641 ioc_rule->timestamp = rule->timestamp;
2643 bcopy(rule->cmd, ioc_rule->cmd, ioc_rule->cmd_len * 4 /* XXX */);
2645 return ((uint8_t *)ioc_rule + IOC_RULESIZE(ioc_rule));
2649 ipfw_copy_state(const ipfw_dyn_rule *dyn_rule,
2650 struct ipfw_ioc_state *ioc_state)
2652 const struct ipfw_flow_id *id;
2653 struct ipfw_ioc_flowid *ioc_id;
2655 ioc_state->expire = TIME_LEQ(dyn_rule->expire, time_second) ?
2656 0 : dyn_rule->expire - time_second;
2657 ioc_state->pcnt = dyn_rule->pcnt;
2658 ioc_state->bcnt = dyn_rule->bcnt;
2660 ioc_state->dyn_type = dyn_rule->dyn_type;
2661 ioc_state->count = dyn_rule->count;
2663 ioc_state->rulenum = dyn_rule->rule->rulenum;
2666 ioc_id = &ioc_state->id;
2668 ioc_id->type = ETHERTYPE_IP;
2669 ioc_id->u.ip.dst_ip = id->dst_ip;
2670 ioc_id->u.ip.src_ip = id->src_ip;
2671 ioc_id->u.ip.dst_port = id->dst_port;
2672 ioc_id->u.ip.src_port = id->src_port;
2673 ioc_id->u.ip.proto = id->proto;
2677 ipfw_ctl_get_rules(struct sockopt *sopt)
2685 * pass up a copy of the current rules. Static rules
2686 * come first (the last of which has number IPFW_DEFAULT_RULE),
2687 * followed by a possibly empty list of dynamic rule.
2691 size = static_ioc_len; /* size of static rules */
2692 if (ipfw_dyn_v) /* add size of dyn.rules */
2693 size += (dyn_count * sizeof(struct ipfw_ioc_state));
2696 * XXX todo: if the user passes a short length just to know
2697 * how much room is needed, do not bother filling up the
2698 * buffer, just jump to the sooptcopyout.
2700 bp = buf = kmalloc(size, M_TEMP, M_WAITOK | M_ZERO);
2702 for (rule = layer3_chain; rule; rule = rule->next)
2703 bp = ipfw_copy_rule(rule, bp);
2706 struct ipfw_ioc_state *ioc_state;
2710 for (i = 0; i < curr_dyn_buckets; i++) {
2713 for (p = ipfw_dyn_v[i]; p != NULL;
2714 p = p->next, ioc_state++)
2715 ipfw_copy_state(p, ioc_state);
2721 error = sooptcopyout(sopt, buf, size);
2727 * {set|get}sockopt parser.
2730 ipfw_ctl(struct sockopt *sopt)
2737 * Disallow modifications in really-really secure mode, but still allow
2738 * the logging counters to be reset.
2740 if (sopt->sopt_name == IP_FW_ADD ||
2741 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) {
2742 if (securelevel >= 3)
2748 switch (sopt->sopt_name) {
2750 error = ipfw_ctl_get_rules(sopt);
2755 * Normally we cannot release the lock on each iteration.
2756 * We could do it here only because we start from the head all
2757 * the times so there is no risk of missing some entries.
2758 * On the other hand, the risk is that we end up with
2759 * a very inconsistent ruleset, so better keep the lock
2760 * around the whole cycle.
2762 * XXX this code can be improved by resetting the head of
2763 * the list to point to the default rule, and then freeing
2764 * the old list without the need for a lock.
2768 free_chain(&layer3_chain, 0 /* keep default rule */);
2773 error = ipfw_ctl_add_rule(sopt);
2778 * IP_FW_DEL is used for deleting single rules or sets,
2779 * and (ab)used to atomically manipulate sets. Argument size
2780 * is used to distinguish between the two:
2782 * delete single rule or set of rules,
2783 * or reassign rules (or sets) to a different set.
2784 * 2*sizeof(uint32_t)
2785 * atomic disable/enable sets.
2786 * first uint32_t contains sets to be disabled,
2787 * second uint32_t contains sets to be enabled.
2789 error = sooptcopyin(sopt, masks,
2790 sizeof(masks), sizeof(masks[0]));
2794 size = sopt->sopt_valsize;
2795 if (size == sizeof(masks[0])) {
2797 * Delete or reassign static rule
2799 error = del_entry(&layer3_chain, masks[0]);
2800 } else if (size == sizeof(masks)) {
2802 * Set enable/disable
2807 (set_disable | masks[0]) & ~masks[1] &
2808 ~(1 << 31); /* set 31 always enabled */
2817 case IP_FW_RESETLOG: /* argument is an int, the rule number */
2820 if (sopt->sopt_val != 0) {
2821 error = sooptcopyin(sopt, &rulenum,
2822 sizeof(int), sizeof(int));
2826 error = zero_entry(rulenum, sopt->sopt_name == IP_FW_RESETLOG);
2830 kprintf("ipfw_ctl invalid option %d\n", sopt->sopt_name);
2838 * This procedure is only used to handle keepalives. It is invoked
2839 * every dyn_keepalive_period
2842 ipfw_tick(void * __unused unused)
2847 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0)
2851 for (i = 0 ; i < curr_dyn_buckets ; i++) {
2852 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) {
2853 if (q->dyn_type == O_LIMIT_PARENT)
2855 if (q->id.proto != IPPROTO_TCP)
2857 if ( (q->state & BOTH_SYN) != BOTH_SYN)
2859 if (TIME_LEQ( time_second+dyn_keepalive_interval,
2861 continue; /* too early */
2862 if (TIME_LEQ(q->expire, time_second))
2863 continue; /* too late, rule expired */
2865 send_pkt(&(q->id), q->ack_rev - 1, q->ack_fwd, TH_SYN);
2866 send_pkt(&(q->id), q->ack_fwd - 1, q->ack_rev, 0);
2871 callout_reset(&ipfw_timeout_h, dyn_keepalive_period * hz,
2876 ipfw_init_default_rule(struct ip_fw **head)
2878 struct ip_fw *def_rule;
2880 KKASSERT(*head == NULL);
2882 def_rule = kmalloc(sizeof(*def_rule), M_IPFW, M_WAITOK | M_ZERO);
2884 def_rule->act_ofs = 0;
2885 def_rule->rulenum = IPFW_DEFAULT_RULE;
2886 def_rule->cmd_len = 1;
2889 def_rule->cmd[0].len = 1;
2890 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
2891 def_rule->cmd[0].opcode = O_ACCEPT;
2893 def_rule->cmd[0].opcode = O_DENY;
2896 def_rule->refcnt = 1;
2899 ipfw_inc_static_count(def_rule);
2901 /* Install the default rule */
2902 ip_fw_default_rule = def_rule;
2908 ip_fw_chk_ptr = ipfw_chk;
2909 ip_fw_ctl_ptr = ipfw_ctl;
2910 ip_fw_dn_io_ptr = ipfw_dummynet_io;
2912 layer3_chain = NULL;
2913 ipfw_init_default_rule(&layer3_chain);
2915 kprintf("ipfw2 initialized, divert %s, "
2916 "rule-based forwarding enabled, default to %s, logging ",
2922 ip_fw_default_rule->cmd[0].opcode == O_ACCEPT ?
2925 #ifdef IPFIREWALL_VERBOSE
2928 #ifdef IPFIREWALL_VERBOSE_LIMIT
2929 verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
2931 if (fw_verbose == 0)
2932 kprintf("disabled\n");
2933 else if (verbose_limit == 0)
2934 kprintf("unlimited\n");
2936 kprintf("limited to %d packets/entry by default\n",
2938 callout_init(&ipfw_timeout_h);
2939 callout_reset(&ipfw_timeout_h, hz, ipfw_tick, NULL);
2943 ipfw_modevent(module_t mod, int type, void *unused)
2952 kprintf("IP firewall already loaded\n");
2962 kprintf("ipfw statically compiled, cannot unload\n");
2965 if (ipfw_refcnt != 0) {
2971 callout_stop(&ipfw_timeout_h);
2972 ip_fw_chk_ptr = NULL;
2973 ip_fw_ctl_ptr = NULL;
2974 ip_fw_dn_io_ptr = NULL;
2975 free_chain(&layer3_chain, 1 /* kill default rule */);
2977 kprintf("IP firewall unloaded\n");
2986 static moduledata_t ipfwmod = {
2991 DECLARE_MODULE(ipfw, ipfwmod, SI_SUB_PROTO_END, SI_ORDER_ANY);
2992 MODULE_VERSION(ipfw, 1);