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 1.60 2008/08/02 03:03:06 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>
61 #include <net/route.h>
62 #include <net/netmsg2.h>
64 #include <netinet/in.h>
65 #include <netinet/in_systm.h>
66 #include <netinet/in_var.h>
67 #include <netinet/in_pcb.h>
68 #include <netinet/ip.h>
69 #include <netinet/ip_var.h>
70 #include <netinet/ip_icmp.h>
72 #include <net/dummynet/ip_dummynet.h>
73 #include <netinet/tcp.h>
74 #include <netinet/tcp_timer.h>
75 #include <netinet/tcp_var.h>
76 #include <netinet/tcpip.h>
77 #include <netinet/udp.h>
78 #include <netinet/udp_var.h>
80 #include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */
82 #define IPFW_AUTOINC_STEP_MIN 1
83 #define IPFW_AUTOINC_STEP_MAX 1000
84 #define IPFW_AUTOINC_STEP_DEF 100
87 * set_disable contains one bit per set value (0..31).
88 * If the bit is set, all rules with the corresponding set
89 * are disabled. Set 31 is reserved for the default rule
90 * and CANNOT be disabled.
92 static uint32_t set_disable;
94 static int fw_verbose;
95 static int verbose_limit;
98 static int ipfw_refcnt;
101 static struct callout ipfw_timeout_h;
102 #define IPFW_DEFAULT_RULE 65535
105 * list of rules for layer 3
107 static struct ip_fw *layer3_chain;
109 MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's");
111 static int fw_debug = 1;
112 static int autoinc_step = IPFW_AUTOINC_STEP_DEF;
115 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
116 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, enable, CTLFLAG_RW,
117 &fw_enable, 0, "Enable ipfw");
118 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW,
119 &autoinc_step, 0, "Rule number autincrement step");
120 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO,one_pass,CTLFLAG_RW,
122 "Only do a single pass through ipfw when using dummynet(4)");
123 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW,
124 &fw_debug, 0, "Enable printing of debug ip_fw statements");
125 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, CTLFLAG_RW,
126 &fw_verbose, 0, "Log matches to ipfw rules");
127 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW,
128 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged");
131 * Description of dynamic rules.
133 * Dynamic rules are stored in lists accessed through a hash table
134 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can
135 * be modified through the sysctl variable dyn_buckets which is
136 * updated when the table becomes empty.
138 * XXX currently there is only one list, ipfw_dyn.
140 * When a packet is received, its address fields are first masked
141 * with the mask defined for the rule, then hashed, then matched
142 * against the entries in the corresponding list.
143 * Dynamic rules can be used for different purposes:
145 * + enforcing limits on the number of sessions;
146 * + in-kernel NAT (not implemented yet)
148 * The lifetime of dynamic rules is regulated by dyn_*_lifetime,
149 * measured in seconds and depending on the flags.
151 * The total number of dynamic rules is stored in dyn_count.
152 * The max number of dynamic rules is dyn_max. When we reach
153 * the maximum number of rules we do not create anymore. This is
154 * done to avoid consuming too much memory, but also too much
155 * time when searching on each packet (ideally, we should try instead
156 * to put a limit on the length of the list on each bucket...).
158 * Each dynamic rule holds a pointer to the parent ipfw rule so
159 * we know what action to perform. Dynamic rules are removed when
160 * the parent rule is deleted. XXX we should make them survive.
162 * There are some limitations with dynamic rules -- we do not
163 * obey the 'randomized match', and we do not do multiple
164 * passes through the firewall. XXX check the latter!!!
166 static ipfw_dyn_rule **ipfw_dyn_v = NULL;
167 static uint32_t dyn_buckets = 256; /* must be power of 2 */
168 static uint32_t curr_dyn_buckets = 256; /* must be power of 2 */
171 * Timeouts for various events in handing dynamic rules.
173 static uint32_t dyn_ack_lifetime = 300;
174 static uint32_t dyn_syn_lifetime = 20;
175 static uint32_t dyn_fin_lifetime = 1;
176 static uint32_t dyn_rst_lifetime = 1;
177 static uint32_t dyn_udp_lifetime = 10;
178 static uint32_t dyn_short_lifetime = 5;
181 * Keepalives are sent if dyn_keepalive is set. They are sent every
182 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval
183 * seconds of lifetime of a rule.
184 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower
185 * than dyn_keepalive_period.
188 static uint32_t dyn_keepalive_interval = 20;
189 static uint32_t dyn_keepalive_period = 5;
190 static uint32_t dyn_keepalive = 1; /* do send keepalives */
192 static uint32_t static_count; /* # of static rules */
193 static uint32_t static_ioc_len; /* bytes of static rules */
194 static uint32_t dyn_count; /* # of dynamic rules */
195 static uint32_t dyn_max = 4096; /* max # of dynamic rules */
197 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW,
198 &dyn_buckets, 0, "Number of dyn. buckets");
199 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD,
200 &curr_dyn_buckets, 0, "Current Number of dyn. buckets");
201 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD,
202 &dyn_count, 0, "Number of dyn. rules");
203 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW,
204 &dyn_max, 0, "Max number of dyn. rules");
205 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD,
206 &static_count, 0, "Number of static rules");
207 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW,
208 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks");
209 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW,
210 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn");
211 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW,
212 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin");
213 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW,
214 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst");
215 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW,
216 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP");
217 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW,
218 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations");
219 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW,
220 &dyn_keepalive, 0, "Enable keepalives for dyn. rules");
222 #endif /* SYSCTL_NODE */
224 static 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) | \
272 (1 << ICMP_ROUTERSOLICIT) | \
273 (1 << ICMP_TSTAMP) | \
278 is_icmp_query(struct ip *ip)
280 int type = L3HDR(struct icmp, ip)->icmp_type;
282 return (type <= ICMP_MAXTYPE && (TT & (1 << type)));
288 * The following checks use two arrays of 8 or 16 bits to store the
289 * bits that we want set or clear, respectively. They are in the
290 * low and high half of cmd->arg1 or cmd->d[0].
292 * We scan options and store the bits we find set. We succeed if
294 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
296 * The code is sometimes optimized not to store additional variables.
300 flags_match(ipfw_insn *cmd, uint8_t bits)
305 if (((cmd->arg1 & 0xff) & bits) != 0)
306 return 0; /* some bits we want set were clear */
308 want_clear = (cmd->arg1 >> 8) & 0xff;
309 if ((want_clear & bits) != want_clear)
310 return 0; /* some bits we want clear were set */
315 ipopts_match(struct ip *ip, ipfw_insn *cmd)
317 int optlen, bits = 0;
318 u_char *cp = (u_char *)(ip + 1);
319 int x = (ip->ip_hl << 2) - sizeof(struct ip);
321 for (; x > 0; x -= optlen, cp += optlen) {
322 int opt = cp[IPOPT_OPTVAL];
324 if (opt == IPOPT_EOL)
327 if (opt == IPOPT_NOP) {
330 optlen = cp[IPOPT_OLEN];
331 if (optlen <= 0 || optlen > x)
332 return 0; /* invalid or truncated */
337 bits |= IP_FW_IPOPT_LSRR;
341 bits |= IP_FW_IPOPT_SSRR;
345 bits |= IP_FW_IPOPT_RR;
349 bits |= IP_FW_IPOPT_TS;
356 return (flags_match(cmd, bits));
360 tcpopts_match(struct ip *ip, ipfw_insn *cmd)
362 int optlen, bits = 0;
363 struct tcphdr *tcp = L3HDR(struct tcphdr,ip);
364 u_char *cp = (u_char *)(tcp + 1);
365 int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
367 for (; x > 0; x -= optlen, cp += optlen) {
370 if (opt == TCPOPT_EOL)
373 if (opt == TCPOPT_NOP) {
383 bits |= IP_FW_TCPOPT_MSS;
387 bits |= IP_FW_TCPOPT_WINDOW;
390 case TCPOPT_SACK_PERMITTED:
392 bits |= IP_FW_TCPOPT_SACK;
395 case TCPOPT_TIMESTAMP:
396 bits |= IP_FW_TCPOPT_TS;
402 bits |= IP_FW_TCPOPT_CC;
409 return (flags_match(cmd, bits));
413 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd)
415 if (ifp == NULL) /* no iface with this packet, match fails */
418 /* Check by name or by IP address */
419 if (cmd->name[0] != '\0') { /* match by name */
422 if (kfnmatch(cmd->name, ifp->if_xname, 0) == 0)
425 if (strncmp(ifp->if_xname, cmd->name, IFNAMSIZ) == 0)
429 struct ifaddr_container *ifac;
431 TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
432 struct ifaddr *ia = ifac->ifa;
434 if (ia->ifa_addr == NULL)
436 if (ia->ifa_addr->sa_family != AF_INET)
438 if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
439 (ia->ifa_addr))->sin_addr.s_addr)
440 return(1); /* match */
443 return(0); /* no match, fail ... */
446 static uint64_t norule_counter; /* counter for ipfw_log(NULL...) */
448 #define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0
451 * We enter here when we have a rule with O_LOG.
452 * XXX this function alone takes about 2Kbytes of code!
455 ipfw_log(struct ip_fw *f, u_int hlen, struct ether_header *eh,
456 struct mbuf *m, struct ifnet *oif)
459 int limit_reached = 0;
460 char action2[40], proto[48], fragment[28];
465 if (f == NULL) { /* bogus pkt */
466 if (verbose_limit != 0 && norule_counter >= verbose_limit)
469 if (norule_counter == verbose_limit)
470 limit_reached = verbose_limit;
472 } else { /* O_LOG is the first action, find the real one */
473 ipfw_insn *cmd = ACTION_PTR(f);
474 ipfw_insn_log *l = (ipfw_insn_log *)cmd;
476 if (l->max_log != 0 && l->log_left == 0)
479 if (l->log_left == 0)
480 limit_reached = l->max_log;
481 cmd += F_LEN(cmd); /* point to first action */
482 if (cmd->opcode == O_PROB)
486 switch (cmd->opcode) {
492 if (cmd->arg1==ICMP_REJECT_RST) {
494 } else if (cmd->arg1==ICMP_UNREACH_HOST) {
497 ksnprintf(SNPARGS(action2, 0), "Unreach %d",
511 ksnprintf(SNPARGS(action2, 0), "Divert %d", cmd->arg1);
515 ksnprintf(SNPARGS(action2, 0), "Tee %d", cmd->arg1);
519 ksnprintf(SNPARGS(action2, 0), "SkipTo %d", cmd->arg1);
523 ksnprintf(SNPARGS(action2, 0), "Pipe %d", cmd->arg1);
527 ksnprintf(SNPARGS(action2, 0), "Queue %d", cmd->arg1);
532 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd;
535 len = ksnprintf(SNPARGS(action2, 0),
537 inet_ntoa(sa->sa.sin_addr));
538 if (sa->sa.sin_port) {
539 ksnprintf(SNPARGS(action2, len), ":%d",
551 if (hlen == 0) { /* non-ip */
552 ksnprintf(SNPARGS(proto, 0), "MAC");
554 struct ip *ip = mtod(m, struct ip *);
555 /* these three are all aliases to the same thing */
556 struct icmp *const icmp = L3HDR(struct icmp, ip);
557 struct tcphdr *const tcp = (struct tcphdr *)icmp;
558 struct udphdr *const udp = (struct udphdr *)icmp;
560 int ip_off, offset, ip_len;
563 if (eh != NULL) { /* layer 2 packets are as on the wire */
564 ip_off = ntohs(ip->ip_off);
565 ip_len = ntohs(ip->ip_len);
570 offset = ip_off & IP_OFFMASK;
573 len = ksnprintf(SNPARGS(proto, 0), "TCP %s",
574 inet_ntoa(ip->ip_src));
576 ksnprintf(SNPARGS(proto, len), ":%d %s:%d",
577 ntohs(tcp->th_sport),
578 inet_ntoa(ip->ip_dst),
579 ntohs(tcp->th_dport));
581 ksnprintf(SNPARGS(proto, len), " %s",
582 inet_ntoa(ip->ip_dst));
587 len = ksnprintf(SNPARGS(proto, 0), "UDP %s",
588 inet_ntoa(ip->ip_src));
590 ksnprintf(SNPARGS(proto, len), ":%d %s:%d",
591 ntohs(udp->uh_sport),
592 inet_ntoa(ip->ip_dst),
593 ntohs(udp->uh_dport));
595 ksnprintf(SNPARGS(proto, len), " %s",
596 inet_ntoa(ip->ip_dst));
602 len = ksnprintf(SNPARGS(proto, 0),
607 len = ksnprintf(SNPARGS(proto, 0), "ICMP ");
609 len += ksnprintf(SNPARGS(proto, len), "%s",
610 inet_ntoa(ip->ip_src));
611 ksnprintf(SNPARGS(proto, len), " %s",
612 inet_ntoa(ip->ip_dst));
616 len = ksnprintf(SNPARGS(proto, 0), "P:%d %s", ip->ip_p,
617 inet_ntoa(ip->ip_src));
618 ksnprintf(SNPARGS(proto, len), " %s",
619 inet_ntoa(ip->ip_dst));
623 if (ip_off & (IP_MF | IP_OFFMASK)) {
624 ksnprintf(SNPARGS(fragment, 0), " (frag %d:%d@%d%s)",
625 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2),
626 offset << 3, (ip_off & IP_MF) ? "+" : "");
630 if (oif || m->m_pkthdr.rcvif) {
631 log(LOG_SECURITY | LOG_INFO,
632 "ipfw: %d %s %s %s via %s%s\n",
634 action, proto, oif ? "out" : "in",
635 oif ? oif->if_xname : m->m_pkthdr.rcvif->if_xname,
638 log(LOG_SECURITY | LOG_INFO,
639 "ipfw: %d %s %s [no if info]%s\n",
641 action, proto, fragment);
645 log(LOG_SECURITY | LOG_NOTICE,
646 "ipfw: limit %d reached on entry %d\n",
647 limit_reached, f ? f->rulenum : -1);
654 * IMPORTANT: the hash function for dynamic rules must be commutative
655 * in source and destination (ip,port), because rules are bidirectional
656 * and we want to find both in the same bucket.
659 hash_packet(struct ipfw_flow_id *id)
663 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port);
664 i &= (curr_dyn_buckets - 1);
669 * unlink a dynamic rule from a chain. prev is a pointer to
670 * the previous one, q is a pointer to the rule to delete,
671 * head is a pointer to the head of the queue.
672 * Modifies q and potentially also head.
674 #define UNLINK_DYN_RULE(prev, head, q) \
676 ipfw_dyn_rule *old_q = q; \
678 /* remove a refcount to the parent */ \
679 if (q->dyn_type == O_LIMIT) \
680 q->parent->count--; \
681 DEB(kprintf("-- unlink entry 0x%08x %d -> 0x%08x %d, %d left\n", \
682 (q->id.src_ip), (q->id.src_port), \
683 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \
685 prev->next = q = q->next; \
687 head = q = q->next; \
688 KASSERT(dyn_count > 0, ("invalid dyn count %u\n", dyn_count)); \
690 kfree(old_q, M_IPFW); \
693 #define TIME_LEQ(a, b) ((int)((a) - (b)) <= 0)
696 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL.
698 * If keep_me == NULL, rules are deleted even if not expired,
699 * otherwise only expired rules are removed.
701 * The value of the second parameter is also used to point to identify
702 * a rule we absolutely do not want to remove (e.g. because we are
703 * holding a reference to it -- this is the case with O_LIMIT_PARENT
704 * rules). The pointer is only used for comparison, so any non-null
708 remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me)
710 static uint32_t last_remove = 0;
712 #define FORCE (keep_me == NULL)
714 ipfw_dyn_rule *prev, *q;
715 int i, pass = 0, max_pass = 0;
717 if (ipfw_dyn_v == NULL || dyn_count == 0)
719 /* do not expire more than once per second, it is useless */
720 if (!FORCE && last_remove == time_second)
722 last_remove = time_second;
725 * because O_LIMIT refer to parent rules, during the first pass only
726 * remove child and mark any pending LIMIT_PARENT, and remove
727 * them in a second pass.
730 for (i = 0; i < curr_dyn_buckets; i++) {
731 for (prev = NULL, q = ipfw_dyn_v[i]; q;) {
733 * Logic can become complex here, so we split tests.
737 if (rule != NULL && rule != q->rule)
738 goto next; /* not the one we are looking for */
739 if (q->dyn_type == O_LIMIT_PARENT) {
741 * handle parent in the second pass,
742 * record we need one.
747 if (FORCE && q->count != 0) {
748 /* XXX should not happen! */
749 kprintf("OUCH! cannot remove rule, "
750 "count %d\n", q->count);
753 if (!FORCE && !TIME_LEQ(q->expire, time_second))
756 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
763 if (pass++ < max_pass)
771 * lookup a dynamic rule.
773 static ipfw_dyn_rule *
774 lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction,
778 * stateful ipfw extensions.
779 * Lookup into dynamic session queue
781 #define MATCH_REVERSE 0
782 #define MATCH_FORWARD 1
784 #define MATCH_UNKNOWN 3
785 int i, dir = MATCH_NONE;
786 ipfw_dyn_rule *prev, *q=NULL;
788 if (ipfw_dyn_v == NULL)
789 goto done; /* not found */
791 i = hash_packet(pkt);
792 for (prev = NULL, q = ipfw_dyn_v[i]; q != NULL;) {
793 if (q->dyn_type == O_LIMIT_PARENT)
796 if (TIME_LEQ( q->expire, time_second)) { /* expire entry */
797 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
800 if (pkt->proto == q->id.proto) {
801 if (pkt->src_ip == q->id.src_ip &&
802 pkt->dst_ip == q->id.dst_ip &&
803 pkt->src_port == q->id.src_port &&
804 pkt->dst_port == q->id.dst_port) {
808 if (pkt->src_ip == q->id.dst_ip &&
809 pkt->dst_ip == q->id.src_ip &&
810 pkt->src_port == q->id.dst_port &&
811 pkt->dst_port == q->id.src_port) {
821 goto done; /* q = NULL, not found */
823 if (prev != NULL) { /* found and not in front */
824 prev->next = q->next;
825 q->next = ipfw_dyn_v[i];
829 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */
830 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST);
832 #define BOTH_SYN (TH_SYN | (TH_SYN << 8))
833 #define BOTH_FIN (TH_FIN | (TH_FIN << 8))
835 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8);
837 case TH_SYN: /* opening */
838 q->expire = time_second + dyn_syn_lifetime;
841 case BOTH_SYN: /* move to established */
842 case BOTH_SYN | TH_FIN : /* one side tries to close */
843 case BOTH_SYN | (TH_FIN << 8) :
845 uint32_t ack = ntohl(tcp->th_ack);
847 #define _SEQ_GE(a, b) ((int)(a) - (int)(b) >= 0)
849 if (dir == MATCH_FORWARD) {
850 if (q->ack_fwd == 0 ||
851 _SEQ_GE(ack, q->ack_fwd))
853 else /* ignore out-of-sequence */
856 if (q->ack_rev == 0 ||
857 _SEQ_GE(ack, q->ack_rev))
859 else /* ignore out-of-sequence */
864 q->expire = time_second + dyn_ack_lifetime;
867 case BOTH_SYN | BOTH_FIN: /* both sides closed */
868 if (dyn_fin_lifetime >= dyn_keepalive_period)
869 dyn_fin_lifetime = dyn_keepalive_period - 1;
870 q->expire = time_second + dyn_fin_lifetime;
876 * reset or some invalid combination, but can also
877 * occur if we use keep-state the wrong way.
879 if ((q->state & ((TH_RST << 8) | TH_RST)) == 0)
880 kprintf("invalid state: 0x%x\n", q->state);
882 if (dyn_rst_lifetime >= dyn_keepalive_period)
883 dyn_rst_lifetime = dyn_keepalive_period - 1;
884 q->expire = time_second + dyn_rst_lifetime;
887 } else if (pkt->proto == IPPROTO_UDP) {
888 q->expire = time_second + dyn_udp_lifetime;
890 /* other protocols */
891 q->expire = time_second + dyn_short_lifetime;
895 *match_direction = dir;
900 realloc_dynamic_table(void)
903 * Try reallocation, make sure we have a power of 2 and do
904 * not allow more than 64k entries. In case of overflow,
908 if (dyn_buckets > 65536)
910 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */
911 dyn_buckets = curr_dyn_buckets; /* reset */
914 curr_dyn_buckets = dyn_buckets;
916 if (ipfw_dyn_v != NULL)
917 kfree(ipfw_dyn_v, M_IPFW);
920 ipfw_dyn_v = kmalloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *),
921 M_IPFW, M_INTWAIT | M_NULLOK | M_ZERO);
922 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2)
924 curr_dyn_buckets /= 2;
929 * Install state of type 'type' for a dynamic session.
930 * The hash table contains two type of rules:
931 * - regular rules (O_KEEP_STATE)
932 * - rules for sessions with limited number of sess per user
933 * (O_LIMIT). When they are created, the parent is
934 * increased by 1, and decreased on delete. In this case,
935 * the third parameter is the parent rule and not the chain.
936 * - "parent" rules for the above (O_LIMIT_PARENT).
938 static ipfw_dyn_rule *
939 add_dyn_rule(struct ipfw_flow_id *id, uint8_t dyn_type, struct ip_fw *rule)
944 if (ipfw_dyn_v == NULL ||
945 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) {
946 realloc_dynamic_table();
947 if (ipfw_dyn_v == NULL)
948 return NULL; /* failed ! */
952 r = kmalloc(sizeof(*r), M_IPFW, M_INTWAIT | M_NULLOK | M_ZERO);
954 kprintf ("sorry cannot allocate state\n");
958 /* increase refcount on parent, and set pointer */
959 if (dyn_type == O_LIMIT) {
960 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule;
962 if (parent->dyn_type != O_LIMIT_PARENT)
963 panic("invalid parent");
970 r->expire = time_second + dyn_syn_lifetime;
972 r->dyn_type = dyn_type;
973 r->pcnt = r->bcnt = 0;
977 r->next = ipfw_dyn_v[i];
980 DEB(kprintf("-- add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n",
982 (r->id.src_ip), (r->id.src_port),
983 (r->id.dst_ip), (r->id.dst_port),
989 * lookup dynamic parent rule using pkt and rule as search keys.
990 * If the lookup fails, then install one.
992 static ipfw_dyn_rule *
993 lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule)
999 i = hash_packet(pkt);
1000 for (q = ipfw_dyn_v[i]; q != NULL; q = q->next) {
1001 if (q->dyn_type == O_LIMIT_PARENT &&
1003 pkt->proto == q->id.proto &&
1004 pkt->src_ip == q->id.src_ip &&
1005 pkt->dst_ip == q->id.dst_ip &&
1006 pkt->src_port == q->id.src_port &&
1007 pkt->dst_port == q->id.dst_port) {
1008 q->expire = time_second + dyn_short_lifetime;
1009 DEB(kprintf("lookup_dyn_parent found 0x%p\n",q);)
1014 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule);
1018 * Install dynamic state for rule type cmd->o.opcode
1020 * Returns 1 (failure) if state is not installed because of errors or because
1021 * session limitations are enforced.
1024 install_state(struct ip_fw *rule, ipfw_insn_limit *cmd,
1025 struct ip_fw_args *args)
1027 static int last_log;
1031 DEB(kprintf("-- install state type %d 0x%08x %u -> 0x%08x %u\n",
1033 (args->f_id.src_ip), (args->f_id.src_port),
1034 (args->f_id.dst_ip), (args->f_id.dst_port) );)
1036 q = lookup_dyn_rule(&args->f_id, NULL, NULL);
1037 if (q != NULL) { /* should never occur */
1038 if (last_log != time_second) {
1039 last_log = time_second;
1040 kprintf(" install_state: entry already present, done\n");
1045 if (dyn_count >= dyn_max) {
1047 * Run out of slots, try to remove any expired rule.
1049 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1);
1052 if (dyn_count >= dyn_max) {
1053 if (last_log != time_second) {
1054 last_log = time_second;
1055 kprintf("install_state: Too many dynamic rules\n");
1057 return 1; /* cannot install, notify caller */
1060 switch (cmd->o.opcode) {
1061 case O_KEEP_STATE: /* bidir rule */
1062 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule);
1065 case O_LIMIT: /* limit number of sessions */
1067 uint16_t limit_mask = cmd->limit_mask;
1068 struct ipfw_flow_id id;
1069 ipfw_dyn_rule *parent;
1071 DEB(kprintf("installing dyn-limit rule %d\n",
1074 id.dst_ip = id.src_ip = 0;
1075 id.dst_port = id.src_port = 0;
1076 id.proto = args->f_id.proto;
1078 if (limit_mask & DYN_SRC_ADDR)
1079 id.src_ip = args->f_id.src_ip;
1080 if (limit_mask & DYN_DST_ADDR)
1081 id.dst_ip = args->f_id.dst_ip;
1082 if (limit_mask & DYN_SRC_PORT)
1083 id.src_port = args->f_id.src_port;
1084 if (limit_mask & DYN_DST_PORT)
1085 id.dst_port = args->f_id.dst_port;
1087 parent = lookup_dyn_parent(&id, rule);
1088 if (parent == NULL) {
1089 kprintf("add parent failed\n");
1093 if (parent->count >= cmd->conn_limit) {
1095 * See if we can remove some expired rule.
1097 remove_dyn_rule(rule, parent);
1098 if (parent->count >= cmd->conn_limit) {
1100 last_log != time_second) {
1101 last_log = time_second;
1102 log(LOG_SECURITY | LOG_DEBUG,
1104 "too many entries\n");
1109 add_dyn_rule(&args->f_id, O_LIMIT,
1110 (struct ip_fw *)parent);
1114 kprintf("unknown dynamic rule type %u\n", cmd->o.opcode);
1117 lookup_dyn_rule(&args->f_id, NULL, NULL); /* XXX just set lifetime */
1122 * Transmit a TCP packet, containing either a RST or a keepalive.
1123 * When flags & TH_RST, we are sending a RST packet, because of a
1124 * "reset" action matched the packet.
1125 * Otherwise we are sending a keepalive, and flags & TH_
1128 send_pkt(struct ipfw_flow_id *id, uint32_t seq, uint32_t ack, int flags)
1133 struct route sro; /* fake route */
1135 MGETHDR(m, MB_DONTWAIT, MT_HEADER);
1138 m->m_pkthdr.rcvif = NULL;
1139 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr);
1140 m->m_data += max_linkhdr;
1142 ip = mtod(m, struct ip *);
1143 bzero(ip, m->m_len);
1144 tcp = (struct tcphdr *)(ip + 1); /* no IP options */
1145 ip->ip_p = IPPROTO_TCP;
1149 * Assume we are sending a RST (or a keepalive in the reverse
1150 * direction), swap src and destination addresses and ports.
1152 ip->ip_src.s_addr = htonl(id->dst_ip);
1153 ip->ip_dst.s_addr = htonl(id->src_ip);
1154 tcp->th_sport = htons(id->dst_port);
1155 tcp->th_dport = htons(id->src_port);
1156 if (flags & TH_RST) { /* we are sending a RST */
1157 if (flags & TH_ACK) {
1158 tcp->th_seq = htonl(ack);
1159 tcp->th_ack = htonl(0);
1160 tcp->th_flags = TH_RST;
1164 tcp->th_seq = htonl(0);
1165 tcp->th_ack = htonl(seq);
1166 tcp->th_flags = TH_RST | TH_ACK;
1170 * We are sending a keepalive. flags & TH_SYN determines
1171 * the direction, forward if set, reverse if clear.
1172 * NOTE: seq and ack are always assumed to be correct
1173 * as set by the caller. This may be confusing...
1175 if (flags & TH_SYN) {
1177 * we have to rewrite the correct addresses!
1179 ip->ip_dst.s_addr = htonl(id->dst_ip);
1180 ip->ip_src.s_addr = htonl(id->src_ip);
1181 tcp->th_dport = htons(id->dst_port);
1182 tcp->th_sport = htons(id->src_port);
1184 tcp->th_seq = htonl(seq);
1185 tcp->th_ack = htonl(ack);
1186 tcp->th_flags = TH_ACK;
1190 * set ip_len to the payload size so we can compute
1191 * the tcp checksum on the pseudoheader
1192 * XXX check this, could save a couple of words ?
1194 ip->ip_len = htons(sizeof(struct tcphdr));
1195 tcp->th_sum = in_cksum(m, m->m_pkthdr.len);
1198 * now fill fields left out earlier
1200 ip->ip_ttl = ip_defttl;
1201 ip->ip_len = m->m_pkthdr.len;
1203 bzero(&sro, sizeof(sro));
1204 ip_rtaddr(ip->ip_dst, &sro);
1206 m->m_pkthdr.fw_flags |= IPFW_MBUF_GENERATED;
1207 ip_output(m, NULL, &sro, 0, NULL, NULL);
1213 * sends a reject message, consuming the mbuf passed as an argument.
1216 send_reject(struct ip_fw_args *args, int code, int offset, int ip_len)
1218 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
1219 /* We need the IP header in host order for icmp_error(). */
1220 if (args->eh != NULL) {
1221 struct ip *ip = mtod(args->m, struct ip *);
1223 ip->ip_len = ntohs(ip->ip_len);
1224 ip->ip_off = ntohs(ip->ip_off);
1226 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
1227 } else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) {
1228 struct tcphdr *const tcp =
1229 L3HDR(struct tcphdr, mtod(args->m, struct ip *));
1231 if ((tcp->th_flags & TH_RST) == 0) {
1232 send_pkt(&args->f_id, ntohl(tcp->th_seq),
1233 ntohl(tcp->th_ack), tcp->th_flags | TH_RST);
1244 * Given an ip_fw *, lookup_next_rule will return a pointer
1245 * to the next rule, which can be either the jump
1246 * target (for skipto instructions) or the next one in the list (in
1247 * all other cases including a missing jump target).
1248 * The result is also written in the "next_rule" field of the rule.
1249 * Backward jumps are not allowed, so start looking from the next
1252 * This never returns NULL -- in case we do not have an exact match,
1253 * the next rule is returned. When the ruleset is changed,
1254 * pointers are flushed so we are always correct.
1257 static struct ip_fw *
1258 lookup_next_rule(struct ip_fw *me)
1260 struct ip_fw *rule = NULL;
1263 /* look for action, in case it is a skipto */
1264 cmd = ACTION_PTR(me);
1265 if (cmd->opcode == O_LOG)
1267 if (cmd->opcode == O_SKIPTO) {
1268 for (rule = me->next; rule; rule = rule->next) {
1269 if (rule->rulenum >= cmd->arg1)
1273 if (rule == NULL) /* failure or not a skipto */
1275 me->next_rule = rule;
1280 * The main check routine for the firewall.
1282 * All arguments are in args so we can modify them and return them
1283 * back to the caller.
1287 * args->m (in/out) The packet; we set to NULL when/if we nuke it.
1288 * Starts with the IP header.
1289 * args->eh (in) Mac header if present, or NULL for layer3 packet.
1290 * args->oif Outgoing interface, or NULL if packet is incoming.
1291 * The incoming interface is in the mbuf. (in)
1293 * args->rule Pointer to the last matching rule (in/out)
1294 * args->next_hop Socket we are forwarding to (out).
1295 * args->f_id Addresses grabbed from the packet (out)
1299 * IP_FW_PORT_DENY_FLAG the packet must be dropped.
1300 * 0 The packet is to be accepted and routed normally OR
1301 * the packet was denied/rejected and has been dropped;
1302 * in the latter case, *m is equal to NULL upon return.
1303 * port Divert the packet to port, with these caveats:
1305 * - If IP_FW_PORT_TEE_FLAG is set, tee the packet instead
1306 * of diverting it (ie, 'ipfw tee').
1308 * - If IP_FW_PORT_DYNT_FLAG is set, interpret the lower
1309 * 16 bits as a dummynet pipe number instead of diverting
1313 ipfw_chk(struct ip_fw_args *args)
1316 * Local variables hold state during the processing of a packet.
1318 * IMPORTANT NOTE: to speed up the processing of rules, there
1319 * are some assumption on the values of the variables, which
1320 * are documented here. Should you change them, please check
1321 * the implementation of the various instructions to make sure
1322 * that they still work.
1324 * args->eh The MAC header. It is non-null for a layer2
1325 * packet, it is NULL for a layer-3 packet.
1327 * m | args->m Pointer to the mbuf, as received from the caller.
1328 * It may change if ipfw_chk() does an m_pullup, or if it
1329 * consumes the packet because it calls send_reject().
1330 * XXX This has to change, so that ipfw_chk() never modifies
1331 * or consumes the buffer.
1332 * ip is simply an alias of the value of m, and it is kept
1333 * in sync with it (the packet is supposed to start with
1336 struct mbuf *m = args->m;
1337 struct ip *ip = mtod(m, struct ip *);
1340 * oif | args->oif If NULL, ipfw_chk has been called on the
1341 * inbound path (ether_input, ip_input).
1342 * If non-NULL, ipfw_chk has been called on the outbound path
1343 * (ether_output, ip_output).
1345 struct ifnet *oif = args->oif;
1347 struct ip_fw *f = NULL; /* matching rule */
1352 * hlen The length of the IPv4 header.
1353 * hlen >0 means we have an IPv4 packet.
1355 u_int hlen = 0; /* hlen >0 means we have an IP pkt */
1358 * offset The offset of a fragment. offset != 0 means that
1359 * we have a fragment at this offset of an IPv4 packet.
1360 * offset == 0 means that (if this is an IPv4 packet)
1361 * this is the first or only fragment.
1366 * Local copies of addresses. They are only valid if we have
1369 * proto The protocol. Set to 0 for non-ip packets,
1370 * or to the protocol read from the packet otherwise.
1371 * proto != 0 means that we have an IPv4 packet.
1373 * src_port, dst_port port numbers, in HOST format. Only
1374 * valid for TCP and UDP packets.
1376 * src_ip, dst_ip ip addresses, in NETWORK format.
1377 * Only valid for IPv4 packets.
1380 uint16_t src_port = 0, dst_port = 0; /* NOTE: host format */
1381 struct in_addr src_ip, dst_ip; /* NOTE: network format */
1382 uint16_t ip_len = 0;
1383 int dyn_dir = MATCH_UNKNOWN;
1384 ipfw_dyn_rule *q = NULL;
1386 if (m->m_pkthdr.fw_flags & IPFW_MBUF_GENERATED)
1387 return 0; /* accept */
1389 * dyn_dir = MATCH_UNKNOWN when rules unchecked,
1390 * MATCH_NONE when checked and not matched (q = NULL),
1391 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
1394 if (args->eh == NULL || /* layer 3 packet */
1395 (m->m_pkthdr.len >= sizeof(struct ip) &&
1396 ntohs(args->eh->ether_type) == ETHERTYPE_IP))
1397 hlen = ip->ip_hl << 2;
1400 * Collect parameters into local variables for faster matching.
1402 if (hlen == 0) { /* do not grab addresses for non-ip pkts */
1403 proto = args->f_id.proto = 0; /* mark f_id invalid */
1404 goto after_ip_checks;
1407 proto = args->f_id.proto = ip->ip_p;
1408 src_ip = ip->ip_src;
1409 dst_ip = ip->ip_dst;
1410 if (args->eh != NULL) { /* layer 2 packets are as on the wire */
1411 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1412 ip_len = ntohs(ip->ip_len);
1414 offset = ip->ip_off & IP_OFFMASK;
1415 ip_len = ip->ip_len;
1418 #define PULLUP_TO(len) \
1420 if (m->m_len < (len)) { \
1421 args->m = m = m_pullup(m, (len));\
1423 goto pullup_failed; \
1424 ip = mtod(m, struct ip *); \
1434 PULLUP_TO(hlen + sizeof(struct tcphdr));
1435 tcp = L3HDR(struct tcphdr, ip);
1436 dst_port = tcp->th_dport;
1437 src_port = tcp->th_sport;
1438 args->f_id.flags = tcp->th_flags;
1446 PULLUP_TO(hlen + sizeof(struct udphdr));
1447 udp = L3HDR(struct udphdr, ip);
1448 dst_port = udp->uh_dport;
1449 src_port = udp->uh_sport;
1454 PULLUP_TO(hlen + 4); /* type, code and checksum. */
1455 args->f_id.flags = L3HDR(struct icmp, ip)->icmp_type;
1465 args->f_id.src_ip = ntohl(src_ip.s_addr);
1466 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1467 args->f_id.src_port = src_port = ntohs(src_port);
1468 args->f_id.dst_port = dst_port = ntohs(dst_port);
1473 * Packet has already been tagged. Look for the next rule
1474 * to restart processing.
1476 * If fw_one_pass != 0 then just accept it.
1477 * XXX should not happen here, but optimized out in
1483 /* This rule was deleted */
1484 if (args->rule->rule_flags & IPFW_RULE_F_INVALID)
1485 return IP_FW_PORT_DENY_FLAG;
1487 f = args->rule->next_rule;
1489 f = lookup_next_rule(args->rule);
1492 * Find the starting rule. It can be either the first
1493 * one, or the one after divert_rule if asked so.
1497 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
1499 skipto = *(uint16_t *)m_tag_data(mtag);
1504 if (args->eh == NULL && skipto != 0) {
1505 if (skipto >= IPFW_DEFAULT_RULE)
1506 return(IP_FW_PORT_DENY_FLAG); /* invalid */
1507 while (f && f->rulenum <= skipto)
1509 if (f == NULL) /* drop packet */
1510 return(IP_FW_PORT_DENY_FLAG);
1513 if ((mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
1514 m_tag_delete(m, mtag);
1517 * Now scan the rules, and parse microinstructions for each rule.
1519 for (; f; f = f->next) {
1522 int skip_or; /* skip rest of OR block */
1525 if (set_disable & (1 << f->set))
1529 for (l = f->cmd_len, cmd = f->cmd; l > 0;
1530 l -= cmdlen, cmd += cmdlen) {
1534 * check_body is a jump target used when we find a
1535 * CHECK_STATE, and need to jump to the body of
1540 cmdlen = F_LEN(cmd);
1542 * An OR block (insn_1 || .. || insn_n) has the
1543 * F_OR bit set in all but the last instruction.
1544 * The first match will set "skip_or", and cause
1545 * the following instructions to be skipped until
1546 * past the one with the F_OR bit clear.
1548 if (skip_or) { /* skip this instruction */
1549 if ((cmd->len & F_OR) == 0)
1550 skip_or = 0; /* next one is good */
1553 match = 0; /* set to 1 if we succeed */
1555 switch (cmd->opcode) {
1557 * The first set of opcodes compares the packet's
1558 * fields with some pattern, setting 'match' if a
1559 * match is found. At the end of the loop there is
1560 * logic to deal with F_NOT and F_OR flags associated
1568 kprintf("ipfw: opcode %d unimplemented\n",
1575 * We only check offset == 0 && proto != 0,
1576 * as this ensures that we have an IPv4
1577 * packet with the ports info.
1582 struct inpcbinfo *pi;
1586 if (proto == IPPROTO_TCP) {
1588 pi = &tcbinfo[mycpu->gd_cpuid];
1589 } else if (proto == IPPROTO_UDP) {
1596 in_pcblookup_hash(pi,
1597 dst_ip, htons(dst_port),
1598 src_ip, htons(src_port),
1600 in_pcblookup_hash(pi,
1601 src_ip, htons(src_port),
1602 dst_ip, htons(dst_port),
1605 if (pcb == NULL || pcb->inp_socket == NULL)
1608 if (cmd->opcode == O_UID) {
1609 #define socheckuid(a,b) ((a)->so_cred->cr_uid != (b))
1611 !socheckuid(pcb->inp_socket,
1612 (uid_t)((ipfw_insn_u32 *)cmd)->d[0]);
1615 match = groupmember(
1616 (uid_t)((ipfw_insn_u32 *)cmd)->d[0],
1617 pcb->inp_socket->so_cred);
1623 match = iface_match(m->m_pkthdr.rcvif,
1624 (ipfw_insn_if *)cmd);
1628 match = iface_match(oif, (ipfw_insn_if *)cmd);
1632 match = iface_match(oif ? oif :
1633 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd);
1637 if (args->eh != NULL) { /* have MAC header */
1638 uint32_t *want = (uint32_t *)
1639 ((ipfw_insn_mac *)cmd)->addr;
1640 uint32_t *mask = (uint32_t *)
1641 ((ipfw_insn_mac *)cmd)->mask;
1642 uint32_t *hdr = (uint32_t *)args->eh;
1645 (want[0] == (hdr[0] & mask[0]) &&
1646 want[1] == (hdr[1] & mask[1]) &&
1647 want[2] == (hdr[2] & mask[2]));
1652 if (args->eh != NULL) {
1654 ntohs(args->eh->ether_type);
1656 ((ipfw_insn_u16 *)cmd)->ports;
1659 /* Special vlan handling */
1660 if (m->m_flags & M_VLANTAG)
1663 for (i = cmdlen - 1; !match && i > 0;
1666 (t >= p[0] && t <= p[1]);
1672 match = (hlen > 0 && offset != 0);
1675 case O_IN: /* "out" is "not in" */
1676 match = (oif == NULL);
1680 match = (args->eh != NULL);
1685 * We do not allow an arg of 0 so the
1686 * check of "proto" only suffices.
1688 match = (proto == cmd->arg1);
1692 match = (hlen > 0 &&
1693 ((ipfw_insn_ip *)cmd)->addr.s_addr ==
1698 match = (hlen > 0 &&
1699 ((ipfw_insn_ip *)cmd)->addr.s_addr ==
1701 ((ipfw_insn_ip *)cmd)->mask.s_addr));
1708 tif = INADDR_TO_IFP(&src_ip);
1709 match = (tif != NULL);
1716 uint32_t *d = (uint32_t *)(cmd+1);
1718 cmd->opcode == O_IP_DST_SET ?
1724 addr -= d[0]; /* subtract base */
1726 (addr < cmd->arg1) &&
1727 (d[1 + (addr >> 5)] &
1728 (1 << (addr & 0x1f)));
1733 match = (hlen > 0 &&
1734 ((ipfw_insn_ip *)cmd)->addr.s_addr ==
1739 match = (hlen > 0) &&
1740 (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1742 ((ipfw_insn_ip *)cmd)->mask.s_addr));
1749 tif = INADDR_TO_IFP(&dst_ip);
1750 match = (tif != NULL);
1757 * offset == 0 && proto != 0 is enough
1758 * to guarantee that we have an IPv4
1759 * packet with port info.
1761 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
1764 (cmd->opcode == O_IP_SRCPORT) ?
1765 src_port : dst_port ;
1767 ((ipfw_insn_u16 *)cmd)->ports;
1770 for (i = cmdlen - 1; !match && i > 0;
1773 (x >= p[0] && x <= p[1]);
1779 match = (offset == 0 && proto==IPPROTO_ICMP &&
1780 icmptype_match(ip, (ipfw_insn_u32 *)cmd));
1784 match = (hlen > 0 && ipopts_match(ip, cmd));
1788 match = (hlen > 0 && cmd->arg1 == ip->ip_v);
1792 match = (hlen > 0 && cmd->arg1 == ip->ip_ttl);
1796 match = (hlen > 0 &&
1797 cmd->arg1 == ntohs(ip->ip_id));
1801 match = (hlen > 0 && cmd->arg1 == ip_len);
1804 case O_IPPRECEDENCE:
1805 match = (hlen > 0 &&
1806 (cmd->arg1 == (ip->ip_tos & 0xe0)));
1810 match = (hlen > 0 &&
1811 flags_match(cmd, ip->ip_tos));
1815 match = (proto == IPPROTO_TCP && offset == 0 &&
1817 L3HDR(struct tcphdr,ip)->th_flags));
1821 match = (proto == IPPROTO_TCP && offset == 0 &&
1822 tcpopts_match(ip, cmd));
1826 match = (proto == IPPROTO_TCP && offset == 0 &&
1827 ((ipfw_insn_u32 *)cmd)->d[0] ==
1828 L3HDR(struct tcphdr,ip)->th_seq);
1832 match = (proto == IPPROTO_TCP && offset == 0 &&
1833 ((ipfw_insn_u32 *)cmd)->d[0] ==
1834 L3HDR(struct tcphdr,ip)->th_ack);
1838 match = (proto == IPPROTO_TCP && offset == 0 &&
1840 L3HDR(struct tcphdr,ip)->th_win);
1844 /* reject packets which have SYN only */
1845 /* XXX should i also check for TH_ACK ? */
1846 match = (proto == IPPROTO_TCP && offset == 0 &&
1847 (L3HDR(struct tcphdr,ip)->th_flags &
1848 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
1853 ipfw_log(f, hlen, args->eh, m, oif);
1858 match = (krandom() <
1859 ((ipfw_insn_u32 *)cmd)->d[0]);
1863 * The second set of opcodes represents 'actions',
1864 * i.e. the terminal part of a rule once the packet
1865 * matches all previous patterns.
1866 * Typically there is only one action for each rule,
1867 * and the opcode is stored at the end of the rule
1868 * (but there are exceptions -- see below).
1870 * In general, here we set retval and terminate the
1871 * outer loop (would be a 'break 3' in some language,
1872 * but we need to do a 'goto done').
1875 * O_COUNT and O_SKIPTO actions:
1876 * instead of terminating, we jump to the next rule
1877 * ('goto next_rule', equivalent to a 'break 2'),
1878 * or to the SKIPTO target ('goto again' after
1879 * having set f, cmd and l), respectively.
1881 * O_LIMIT and O_KEEP_STATE: these opcodes are
1882 * not real 'actions', and are stored right
1883 * before the 'action' part of the rule.
1884 * These opcodes try to install an entry in the
1885 * state tables; if successful, we continue with
1886 * the next opcode (match=1; break;), otherwise
1887 * the packet * must be dropped
1888 * ('goto done' after setting retval);
1890 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
1891 * cause a lookup of the state table, and a jump
1892 * to the 'action' part of the parent rule
1893 * ('goto check_body') if an entry is found, or
1894 * (CHECK_STATE only) a jump to the next rule if
1895 * the entry is not found ('goto next_rule').
1896 * The result of the lookup is cached to make
1897 * further instances of these opcodes are
1902 if (install_state(f,
1903 (ipfw_insn_limit *)cmd, args)) {
1904 retval = IP_FW_PORT_DENY_FLAG;
1905 goto done; /* error/limit violation */
1913 * dynamic rules are checked at the first
1914 * keep-state or check-state occurrence,
1915 * with the result being stored in dyn_dir.
1916 * The compiler introduces a PROBE_STATE
1917 * instruction for us when we have a
1918 * KEEP_STATE (because PROBE_STATE needs
1921 if (dyn_dir == MATCH_UNKNOWN &&
1922 (q = lookup_dyn_rule(&args->f_id,
1923 &dyn_dir, proto == IPPROTO_TCP ?
1924 L3HDR(struct tcphdr, ip) : NULL))
1927 * Found dynamic entry, update stats
1928 * and jump to the 'action' part of
1934 cmd = ACTION_PTR(f);
1935 l = f->cmd_len - f->act_ofs;
1939 * Dynamic entry not found. If CHECK_STATE,
1940 * skip to next rule, if PROBE_STATE just
1941 * ignore and continue with next opcode.
1943 if (cmd->opcode == O_CHECK_STATE)
1949 retval = 0; /* accept */
1954 args->rule = f; /* report matching rule */
1955 retval = cmd->arg1 | IP_FW_PORT_DYNT_FLAG;
1960 if (args->eh) /* not on layer 2 */
1963 mtag = m_tag_get(PACKET_TAG_IPFW_DIVERT,
1964 sizeof(uint16_t), MB_DONTWAIT);
1966 retval = IP_FW_PORT_DENY_FLAG;
1969 *(uint16_t *)m_tag_data(mtag) = f->rulenum;
1970 m_tag_prepend(m, mtag);
1971 retval = (cmd->opcode == O_DIVERT) ?
1973 cmd->arg1 | IP_FW_PORT_TEE_FLAG;
1978 f->pcnt++; /* update stats */
1980 f->timestamp = time_second;
1981 if (cmd->opcode == O_COUNT)
1984 if (f->next_rule == NULL)
1985 lookup_next_rule(f);
1991 * Drop the packet and send a reject notice
1992 * if the packet is not ICMP (or is an ICMP
1993 * query), and it is not multicast/broadcast.
1996 (proto != IPPROTO_ICMP ||
1997 is_icmp_query(ip)) &&
1998 !(m->m_flags & (M_BCAST|M_MCAST)) &&
1999 !IN_MULTICAST(ntohl(dst_ip.s_addr))) {
2000 send_reject(args, cmd->arg1,
2006 retval = IP_FW_PORT_DENY_FLAG;
2010 if (args->eh) /* not valid on layer2 pkts */
2012 if (!q || dyn_dir == MATCH_FORWARD) {
2014 &((ipfw_insn_sa *)cmd)->sa;
2020 panic("-- unknown opcode %d\n", cmd->opcode);
2021 } /* end of switch() on opcodes */
2023 if (cmd->len & F_NOT)
2027 if (cmd->len & F_OR)
2030 if (!(cmd->len & F_OR)) /* not an OR block, */
2031 break; /* try next rule */
2034 } /* end of inner for, scan opcodes */
2036 next_rule:; /* try next rule */
2038 } /* end of outer for, scan rules */
2039 kprintf("+++ ipfw: ouch!, skip past end of rules, denying packet\n");
2040 return(IP_FW_PORT_DENY_FLAG);
2043 /* Update statistics */
2046 f->timestamp = time_second;
2051 kprintf("pullup failed\n");
2052 return(IP_FW_PORT_DENY_FLAG);
2056 ipfw_dummynet_io(struct mbuf *m, int pipe_nr, int dir, struct ip_fw_args *fwa)
2061 const struct ipfw_flow_id *id;
2062 struct dn_flow_id *fid;
2066 mtag = m_tag_get(PACKET_TAG_DUMMYNET, sizeof(*pkt), MB_DONTWAIT);
2071 m_tag_prepend(m, mtag);
2073 pkt = m_tag_data(mtag);
2074 bzero(pkt, sizeof(*pkt));
2076 cmd = fwa->rule->cmd + fwa->rule->act_ofs;
2077 if (cmd->opcode == O_LOG)
2079 KASSERT(cmd->opcode == O_PIPE || cmd->opcode == O_QUEUE,
2080 ("Rule is not PIPE or QUEUE, opcode %d\n", cmd->opcode));
2083 pkt->dn_flags = (dir & DN_FLAGS_DIR_MASK);
2084 pkt->ifp = fwa->oif;
2085 pkt->cpuid = mycpu->gd_cpuid;
2086 pkt->pipe_nr = pipe_nr;
2090 fid->fid_dst_ip = id->dst_ip;
2091 fid->fid_src_ip = id->src_ip;
2092 fid->fid_dst_port = id->dst_port;
2093 fid->fid_src_port = id->src_port;
2094 fid->fid_proto = id->proto;
2095 fid->fid_flags = id->flags;
2097 ipfw_ref_rule(fwa->rule);
2098 pkt->dn_priv = fwa->rule;
2099 pkt->dn_unref_priv = ipfw_unref_rule;
2101 if (cmd->opcode == O_PIPE)
2102 pkt->dn_flags |= DN_FLAGS_IS_PIPE;
2104 if (dir == DN_TO_IP_OUT) {
2106 * We need to copy *ro because for ICMP pkts (and maybe
2107 * others) the caller passed a pointer into the stack;
2108 * dst might also be a pointer into *ro so it needs to
2111 pkt->ro = *(fwa->ro);
2113 fwa->ro->ro_rt->rt_refcnt++;
2114 if (fwa->dst == (struct sockaddr_in *)&fwa->ro->ro_dst) {
2115 /* 'dst' points into 'ro' */
2116 fwa->dst = (struct sockaddr_in *)&(pkt->ro.ro_dst);
2118 pkt->dn_dst = fwa->dst;
2119 pkt->flags = fwa->flags;
2122 m->m_pkthdr.fw_flags |= DUMMYNET_MBUF_TAGGED;
2127 * When a rule is added/deleted, clear the next_rule pointers in all rules.
2128 * These will be reconstructed on the fly as packets are matched.
2129 * Must be called at splimp().
2132 flush_rule_ptrs(void)
2136 for (rule = layer3_chain; rule; rule = rule->next)
2137 rule->next_rule = NULL;
2140 static __inline void
2141 ipfw_inc_static_count(struct ip_fw *rule)
2143 KASSERT(mycpuid == 0,
2144 ("adding static rule not on cpu0 (%d)", mycpuid));
2147 static_ioc_len += IOC_RULESIZE(rule);
2150 static __inline void
2151 ipfw_dec_static_count(struct ip_fw *rule)
2153 int l = IOC_RULESIZE(rule);
2155 KASSERT(mycpuid == 0,
2156 ("deleting static rule not on cpu0 (%d)", mycpuid));
2158 KASSERT(static_count > 0, ("invalid static count %u\n", static_count));
2161 KASSERT(static_ioc_len >= l,
2162 ("invalid static len %u\n", static_ioc_len));
2163 static_ioc_len -= l;
2166 static struct ip_fw *
2167 ipfw_create_rule(const struct ipfw_ioc_rule *ioc_rule)
2171 rule = kmalloc(RULESIZE(ioc_rule), M_IPFW, M_WAITOK | M_ZERO);
2173 rule->act_ofs = ioc_rule->act_ofs;
2174 rule->cmd_len = ioc_rule->cmd_len;
2175 rule->rulenum = ioc_rule->rulenum;
2176 rule->set = ioc_rule->set;
2177 rule->usr_flags = ioc_rule->usr_flags;
2179 bcopy(ioc_rule->cmd, rule->cmd, rule->cmd_len * 4 /* XXX */);
2187 * Add a new rule to the list. Copy the rule into a malloc'ed area,
2188 * then possibly create a rule number and add the rule to the list.
2189 * Update the rule_number in the input struct so the caller knows
2193 ipfw_add_rule(struct ip_fw **head, struct ipfw_ioc_rule *ioc_rule)
2195 struct ip_fw *rule, *f, *prev;
2197 KKASSERT(*head != NULL);
2198 KASSERT(mycpuid == 0,
2199 ("adding static rule not on cpu0 (%d)", mycpuid));
2201 rule = ipfw_create_rule(ioc_rule);
2206 * If rulenum is 0, find highest numbered rule before the
2207 * default rule, and add rule number incremental step
2209 if (rule->rulenum == 0) {
2210 int step = autoinc_step;
2213 * Make sure that rule number incremental step
2216 if (step < IPFW_AUTOINC_STEP_MIN)
2217 step = IPFW_AUTOINC_STEP_MIN;
2218 else if (step > IPFW_AUTOINC_STEP_MAX)
2219 step = IPFW_AUTOINC_STEP_MAX;
2222 * Locate the highest numbered rule before default
2224 for (f = *head; f; f = f->next) {
2225 if (f->rulenum == IPFW_DEFAULT_RULE)
2227 rule->rulenum = f->rulenum;
2229 if (rule->rulenum < IPFW_DEFAULT_RULE - step)
2230 rule->rulenum += step;
2232 /* Update the input structure */
2233 ioc_rule->rulenum = rule->rulenum;
2237 * Now insert the new rule in the right place in the sorted list.
2239 for (prev = NULL, f = *head; f; prev = f, f = f->next) {
2240 if (f->rulenum > rule->rulenum) {
2241 /* Found the location */
2254 ipfw_inc_static_count(rule);
2258 DEB(kprintf("++ installed rule %d, static count now %d\n",
2259 rule->rulenum, static_count);)
2263 * Free storage associated with a static rule (including derived
2265 * The caller is in charge of clearing rule pointers to avoid
2266 * dangling pointers.
2267 * @return a pointer to the next entry.
2268 * Arguments are not checked, so they better be correct.
2269 * Must be called at splimp().
2271 static struct ip_fw *
2272 delete_rule(struct ip_fw **head, struct ip_fw *prev, struct ip_fw *rule)
2277 remove_dyn_rule(rule, NULL /* force removal */);
2282 ipfw_dec_static_count(rule);
2284 /* Mark the rule as invalid */
2285 rule->rule_flags |= IPFW_RULE_F_INVALID;
2286 rule->next_rule = NULL;
2288 /* Try to free this rule */
2289 ipfw_free_rule(rule);
2295 * Deletes all rules from a chain (including the default rule
2296 * if the second argument is set).
2297 * Must be called at splimp().
2300 free_chain(struct ip_fw **chain, int kill_default)
2304 flush_rule_ptrs(); /* more efficient to do outside the loop */
2306 while ((rule = *chain) != NULL &&
2307 (kill_default || rule->rulenum != IPFW_DEFAULT_RULE))
2308 delete_rule(chain, NULL, rule);
2310 KASSERT(dyn_count == 0, ("%u dyn rule remains\n", dyn_count));
2313 ip_fw_default_rule = NULL; /* Reset default rule */
2315 if (ipfw_dyn_v != NULL) {
2317 * Free dynamic rules(state) hash table
2319 kfree(ipfw_dyn_v, M_IPFW);
2323 KASSERT(static_count == 0,
2324 ("%u static rules remains\n", static_count));
2325 KASSERT(static_ioc_len == 0,
2326 ("%u bytes of static rules remains\n", static_ioc_len));
2328 KASSERT(static_count == 1,
2329 ("%u static rules remains\n", static_count));
2330 KASSERT(static_ioc_len == IOC_RULESIZE(ip_fw_default_rule),
2331 ("%u bytes of static rules remains, should be %u\n",
2332 static_ioc_len, IOC_RULESIZE(ip_fw_default_rule)));
2337 * Remove all rules with given number, and also do set manipulation.
2339 * The argument is an uint32_t. The low 16 bit are the rule or set number,
2340 * the next 8 bits are the new set, the top 8 bits are the command:
2342 * 0 delete rules with given number
2343 * 1 delete rules with given set number
2344 * 2 move rules with given number to new set
2345 * 3 move rules with given set number to new set
2346 * 4 swap sets with given numbers
2349 del_entry(struct ip_fw **chain, uint32_t arg)
2351 struct ip_fw *prev, *rule;
2353 uint8_t cmd, new_set;
2355 rulenum = arg & 0xffff;
2356 cmd = (arg >> 24) & 0xff;
2357 new_set = (arg >> 16) & 0xff;
2363 if (cmd == 0 || cmd == 2) {
2364 if (rulenum == IPFW_DEFAULT_RULE)
2372 case 0: /* delete rules with given number */
2374 * locate first rule to delete
2376 for (prev = NULL, rule = *chain;
2377 rule && rule->rulenum < rulenum;
2378 prev = rule, rule = rule->next)
2380 if (rule->rulenum != rulenum)
2383 crit_enter(); /* no access to rules while removing */
2385 * flush pointers outside the loop, then delete all matching
2386 * rules. prev remains the same throughout the cycle.
2389 while (rule && rule->rulenum == rulenum)
2390 rule = delete_rule(chain, prev, rule);
2394 case 1: /* delete all rules with given set number */
2397 for (prev = NULL, rule = *chain; rule;) {
2398 if (rule->set == rulenum) {
2399 rule = delete_rule(chain, prev, rule);
2408 case 2: /* move rules with given number to new set */
2410 for (rule = *chain; rule; rule = rule->next) {
2411 if (rule->rulenum == rulenum)
2412 rule->set = new_set;
2417 case 3: /* move rules with given set number to new set */
2419 for (rule = *chain; rule; rule = rule->next) {
2420 if (rule->set == rulenum)
2421 rule->set = new_set;
2426 case 4: /* swap two sets */
2428 for (rule = *chain; rule; rule = rule->next) {
2429 if (rule->set == rulenum)
2430 rule->set = new_set;
2431 else if (rule->set == new_set)
2432 rule->set = rulenum;
2441 * Clear counters for a specific rule.
2444 clear_counters(struct ip_fw *rule, int log_only)
2446 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule);
2448 if (log_only == 0) {
2449 rule->bcnt = rule->pcnt = 0;
2450 rule->timestamp = 0;
2452 if (l->o.opcode == O_LOG)
2453 l->log_left = l->max_log;
2457 * Reset some or all counters on firewall rules.
2458 * @arg frwl is null to clear all entries, or contains a specific
2460 * @arg log_only is 1 if we only want to reset logs, zero otherwise.
2463 zero_entry(int rulenum, int log_only)
2471 for (rule = layer3_chain; rule; rule = rule->next)
2472 clear_counters(rule, log_only);
2474 msg = log_only ? "ipfw: All logging counts reset.\n"
2475 : "ipfw: Accounting cleared.\n";
2480 * We can have multiple rules with the same number, so we
2481 * need to clear them all.
2483 for (rule = layer3_chain; rule; rule = rule->next) {
2484 if (rule->rulenum == rulenum) {
2486 while (rule && rule->rulenum == rulenum) {
2487 clear_counters(rule, log_only);
2495 if (!cleared) /* we did not find any matching rules */
2497 msg = log_only ? "ipfw: Entry %d logging count reset.\n"
2498 : "ipfw: Entry %d cleared.\n";
2501 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum);
2506 * Check validity of the structure before insert.
2507 * Fortunately rules are simple, so this mostly need to check rule sizes.
2510 ipfw_ctl_check_rule(struct ipfw_ioc_rule *rule, int size)
2513 int have_action = 0;
2516 /* Check for valid size */
2517 if (size < sizeof(*rule)) {
2518 kprintf("ipfw: rule too short\n");
2521 l = IOC_RULESIZE(rule);
2523 kprintf("ipfw: size mismatch (have %d want %d)\n", size, l);
2528 * Now go for the individual checks. Very simple ones, basically only
2529 * instruction sizes.
2531 for (l = rule->cmd_len, cmd = rule->cmd; l > 0;
2532 l -= cmdlen, cmd += cmdlen) {
2533 cmdlen = F_LEN(cmd);
2535 kprintf("ipfw: opcode %d size truncated\n",
2539 DEB(kprintf("ipfw: opcode %d\n", cmd->opcode);)
2540 switch (cmd->opcode) {
2554 case O_IPPRECEDENCE:
2561 if (cmdlen != F_INSN_SIZE(ipfw_insn))
2573 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32))
2578 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit))
2583 if (cmdlen != F_INSN_SIZE(ipfw_insn_log))
2586 ((ipfw_insn_log *)cmd)->log_left =
2587 ((ipfw_insn_log *)cmd)->max_log;
2593 if (cmdlen != F_INSN_SIZE(ipfw_insn_ip))
2595 if (((ipfw_insn_ip *)cmd)->mask.s_addr == 0) {
2596 kprintf("ipfw: opcode %d, useless rule\n",
2604 if (cmd->arg1 == 0 || cmd->arg1 > 256) {
2605 kprintf("ipfw: invalid set size %d\n",
2609 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
2615 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac))
2621 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */
2622 if (cmdlen < 2 || cmdlen > 31)
2629 if (cmdlen != F_INSN_SIZE(ipfw_insn_if))
2635 if (cmdlen != F_INSN_SIZE(ipfw_insn_pipe))
2640 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa))
2644 case O_FORWARD_MAC: /* XXX not implemented yet */
2653 if (cmdlen != F_INSN_SIZE(ipfw_insn))
2657 kprintf("ipfw: opcode %d, multiple actions"
2664 kprintf("ipfw: opcode %d, action must be"
2671 kprintf("ipfw: opcode %d, unknown opcode\n",
2676 if (have_action == 0) {
2677 kprintf("ipfw: missing action\n");
2683 kprintf("ipfw: opcode %d size %d wrong\n",
2684 cmd->opcode, cmdlen);
2689 ipfw_ctl_add_rule(struct sockopt *sopt)
2691 struct ipfw_ioc_rule *ioc_rule;
2695 size = sopt->sopt_valsize;
2696 if (size > (sizeof(uint32_t) * IPFW_RULE_SIZE_MAX) ||
2697 size < sizeof(*ioc_rule)) {
2700 if (size != (sizeof(uint32_t) * IPFW_RULE_SIZE_MAX)) {
2701 sopt->sopt_val = krealloc(sopt->sopt_val, sizeof(uint32_t) *
2702 IPFW_RULE_SIZE_MAX, M_TEMP, M_WAITOK);
2704 ioc_rule = sopt->sopt_val;
2706 error = ipfw_ctl_check_rule(ioc_rule, size);
2710 ipfw_add_rule(&layer3_chain, ioc_rule);
2712 if (sopt->sopt_dir == SOPT_GET)
2713 sopt->sopt_valsize = IOC_RULESIZE(ioc_rule);
2718 ipfw_copy_rule(const struct ip_fw *rule, struct ipfw_ioc_rule *ioc_rule)
2720 ioc_rule->act_ofs = rule->act_ofs;
2721 ioc_rule->cmd_len = rule->cmd_len;
2722 ioc_rule->rulenum = rule->rulenum;
2723 ioc_rule->set = rule->set;
2724 ioc_rule->usr_flags = rule->usr_flags;
2726 ioc_rule->set_disable = set_disable;
2727 ioc_rule->static_count = static_count;
2728 ioc_rule->static_len = static_ioc_len;
2730 ioc_rule->pcnt = rule->pcnt;
2731 ioc_rule->bcnt = rule->bcnt;
2732 ioc_rule->timestamp = rule->timestamp;
2734 bcopy(rule->cmd, ioc_rule->cmd, ioc_rule->cmd_len * 4 /* XXX */);
2736 return ((uint8_t *)ioc_rule + IOC_RULESIZE(ioc_rule));
2740 ipfw_copy_state(const ipfw_dyn_rule *dyn_rule,
2741 struct ipfw_ioc_state *ioc_state)
2743 const struct ipfw_flow_id *id;
2744 struct ipfw_ioc_flowid *ioc_id;
2746 ioc_state->expire = TIME_LEQ(dyn_rule->expire, time_second) ?
2747 0 : dyn_rule->expire - time_second;
2748 ioc_state->pcnt = dyn_rule->pcnt;
2749 ioc_state->bcnt = dyn_rule->bcnt;
2751 ioc_state->dyn_type = dyn_rule->dyn_type;
2752 ioc_state->count = dyn_rule->count;
2754 ioc_state->rulenum = dyn_rule->rule->rulenum;
2757 ioc_id = &ioc_state->id;
2759 ioc_id->type = ETHERTYPE_IP;
2760 ioc_id->u.ip.dst_ip = id->dst_ip;
2761 ioc_id->u.ip.src_ip = id->src_ip;
2762 ioc_id->u.ip.dst_port = id->dst_port;
2763 ioc_id->u.ip.src_port = id->src_port;
2764 ioc_id->u.ip.proto = id->proto;
2768 ipfw_ctl_get_rules(struct sockopt *sopt)
2775 * pass up a copy of the current rules. Static rules
2776 * come first (the last of which has number IPFW_DEFAULT_RULE),
2777 * followed by a possibly empty list of dynamic rule.
2781 size = static_ioc_len; /* size of static rules */
2782 if (ipfw_dyn_v) /* add size of dyn.rules */
2783 size += (dyn_count * sizeof(struct ipfw_ioc_state));
2785 if (sopt->sopt_valsize < size) {
2786 /* short length, no need to return incomplete rules */
2787 /* XXX: if superuser, no need to zero buffer */
2788 bzero(sopt->sopt_val, sopt->sopt_valsize);
2791 bp = sopt->sopt_val;
2793 for (rule = layer3_chain; rule; rule = rule->next)
2794 bp = ipfw_copy_rule(rule, bp);
2797 struct ipfw_ioc_state *ioc_state;
2801 for (i = 0; i < curr_dyn_buckets; i++) {
2804 for (p = ipfw_dyn_v[i]; p != NULL;
2805 p = p->next, ioc_state++)
2806 ipfw_copy_state(p, ioc_state);
2812 sopt->sopt_valsize = size;
2817 * {set|get}sockopt parser.
2820 ipfw_ctl(struct sockopt *sopt)
2828 switch (sopt->sopt_name) {
2830 error = ipfw_ctl_get_rules(sopt);
2835 * Normally we cannot release the lock on each iteration.
2836 * We could do it here only because we start from the head all
2837 * the times so there is no risk of missing some entries.
2838 * On the other hand, the risk is that we end up with
2839 * a very inconsistent ruleset, so better keep the lock
2840 * around the whole cycle.
2842 * XXX this code can be improved by resetting the head of
2843 * the list to point to the default rule, and then freeing
2844 * the old list without the need for a lock.
2848 free_chain(&layer3_chain, 0 /* keep default rule */);
2853 error = ipfw_ctl_add_rule(sopt);
2858 * IP_FW_DEL is used for deleting single rules or sets,
2859 * and (ab)used to atomically manipulate sets. Argument size
2860 * is used to distinguish between the two:
2862 * delete single rule or set of rules,
2863 * or reassign rules (or sets) to a different set.
2864 * 2*sizeof(uint32_t)
2865 * atomic disable/enable sets.
2866 * first uint32_t contains sets to be disabled,
2867 * second uint32_t contains sets to be enabled.
2869 masks = sopt->sopt_val;
2870 size = sopt->sopt_valsize;
2871 if (size == sizeof(*masks)) {
2873 * Delete or reassign static rule
2875 error = del_entry(&layer3_chain, masks[0]);
2876 } else if (size == (2 * sizeof(*masks))) {
2878 * Set enable/disable
2883 (set_disable | masks[0]) & ~masks[1] &
2884 ~(1 << 31); /* set 31 always enabled */
2893 case IP_FW_RESETLOG: /* argument is an int, the rule number */
2896 if (sopt->sopt_val != 0) {
2897 error = soopt_to_kbuf(sopt, &rulenum,
2898 sizeof(int), sizeof(int));
2902 error = zero_entry(rulenum, sopt->sopt_name == IP_FW_RESETLOG);
2906 kprintf("ipfw_ctl invalid option %d\n", sopt->sopt_name);
2913 * This procedure is only used to handle keepalives. It is invoked
2914 * every dyn_keepalive_period
2917 ipfw_tick(void *unused __unused)
2922 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0)
2926 for (i = 0; i < curr_dyn_buckets; i++) {
2927 for (q = ipfw_dyn_v[i]; q; q = q->next) {
2928 if (q->dyn_type == O_LIMIT_PARENT)
2930 if (q->id.proto != IPPROTO_TCP)
2932 if ((q->state & BOTH_SYN) != BOTH_SYN)
2934 if (TIME_LEQ(time_second + dyn_keepalive_interval,
2936 continue; /* too early */
2937 if (TIME_LEQ(q->expire, time_second))
2938 continue; /* too late, rule expired */
2940 send_pkt(&q->id, q->ack_rev - 1, q->ack_fwd, TH_SYN);
2941 send_pkt(&q->id, q->ack_fwd - 1, q->ack_rev, 0);
2946 callout_reset(&ipfw_timeout_h, dyn_keepalive_period * hz,
2951 ipfw_init_default_rule(struct ip_fw **head)
2953 struct ip_fw *def_rule;
2955 KKASSERT(*head == NULL);
2957 def_rule = kmalloc(sizeof(*def_rule), M_IPFW, M_WAITOK | M_ZERO);
2959 def_rule->act_ofs = 0;
2960 def_rule->rulenum = IPFW_DEFAULT_RULE;
2961 def_rule->cmd_len = 1;
2964 def_rule->cmd[0].len = 1;
2965 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
2966 def_rule->cmd[0].opcode = O_ACCEPT;
2968 def_rule->cmd[0].opcode = O_DENY;
2971 def_rule->refcnt = 1;
2974 ipfw_inc_static_count(def_rule);
2976 /* Install the default rule */
2977 ip_fw_default_rule = def_rule;
2981 ipfw_init_dispatch(struct netmsg *nmsg)
2988 kprintf("IP firewall already loaded\n");
2993 ip_fw_chk_ptr = ipfw_chk;
2994 ip_fw_ctl_ptr = ipfw_ctl;
2995 ip_fw_dn_io_ptr = ipfw_dummynet_io;
2997 layer3_chain = NULL;
2998 ipfw_init_default_rule(&layer3_chain);
3000 kprintf("ipfw2 initialized, divert %s, "
3001 "rule-based forwarding enabled, default to %s, logging ",
3007 ip_fw_default_rule->cmd[0].opcode == O_ACCEPT ?
3010 #ifdef IPFIREWALL_VERBOSE
3013 #ifdef IPFIREWALL_VERBOSE_LIMIT
3014 verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
3016 if (fw_verbose == 0) {
3017 kprintf("disabled\n");
3018 } else if (verbose_limit == 0) {
3019 kprintf("unlimited\n");
3021 kprintf("limited to %d packets/entry by default\n",
3024 callout_init(&ipfw_timeout_h);
3027 callout_reset(&ipfw_timeout_h, hz, ipfw_tick, NULL);
3030 lwkt_replymsg(&nmsg->nm_lmsg, error);
3038 netmsg_init(&smsg, &curthread->td_msgport, 0, ipfw_init_dispatch);
3039 return lwkt_domsg(cpu_portfn(0), &smsg.nm_lmsg, 0);
3045 ipfw_fini_dispatch(struct netmsg *nmsg)
3051 if (ipfw_refcnt != 0) {
3056 callout_stop(&ipfw_timeout_h);
3059 netmsg_service_sync();
3061 ip_fw_chk_ptr = NULL;
3062 ip_fw_ctl_ptr = NULL;
3063 ip_fw_dn_io_ptr = NULL;
3064 free_chain(&layer3_chain, 1 /* kill default rule */);
3066 kprintf("IP firewall unloaded\n");
3069 lwkt_replymsg(&nmsg->nm_lmsg, error);
3077 netmsg_init(&smsg, &curthread->td_msgport, 0, ipfw_fini_dispatch);
3078 return lwkt_domsg(cpu_portfn(0), &smsg.nm_lmsg, 0);
3081 #endif /* KLD_MODULE */
3084 ipfw_modevent(module_t mod, int type, void *unused)
3095 kprintf("ipfw statically compiled, cannot unload\n");
3107 static moduledata_t ipfwmod = {
3112 DECLARE_MODULE(ipfw, ipfwmod, SI_SUB_PROTO_END, SI_ORDER_ANY);
3113 MODULE_VERSION(ipfw, 1);