2 * Copyright (c) 1993 Daniel Boulet
3 * Copyright (c) 1994 Ugen J.S.Antsilevich
4 * Copyright (c) 1996 Alex Nash
5 * Copyright (c) 2000-2001 Luigi Rizzo
7 * Redistribution and use in source forms, with and without modification,
8 * are permitted provided that this entire comment appears intact.
10 * Redistribution in binary form may occur without any restrictions.
11 * Obviously, it would be nice if you gave credit where credit is due
12 * but requiring it would be too onerous.
14 * This software is provided ``AS IS'' without any warranties of any kind.
16 * $FreeBSD: src/sys/netinet/ip_fw.c,v 1.131.2.39 2003/01/20 02:23:07 iedowse Exp $
17 * $DragonFly: src/sys/net/ipfw/Attic/ip_fw.c,v 1.15 2005/12/11 13:00:16 swildner Exp $
24 * Implement IP packet firewall
27 #if !defined(KLD_MODULE)
30 #include "opt_ipdivert.h"
33 #error IPFIREWALL requires INET.
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/malloc.h>
43 #include <sys/kernel.h>
45 #include <sys/socket.h>
46 #include <sys/socketvar.h>
47 #include <sys/sysctl.h>
48 #include <sys/syslog.h>
49 #include <sys/thread2.h>
50 #include <sys/ucred.h>
52 #include <net/route.h>
53 #include <netinet/in.h>
54 #include <netinet/in_systm.h>
55 #include <netinet/in_var.h>
56 #include <netinet/in_pcb.h>
57 #include <netinet/ip.h>
58 #include <netinet/ip_var.h>
59 #include <netinet/ip_icmp.h>
61 #include <net/dummynet/ip_dummynet.h>
62 #include <netinet/tcp.h>
63 #include <netinet/tcp_timer.h>
64 #include <netinet/tcp_var.h>
65 #include <netinet/tcpip.h>
66 #include <netinet/udp.h>
67 #include <netinet/udp_var.h>
69 #include <netinet/if_ether.h> /* XXX ethertype_ip */
71 static int fw_debug = 1;
72 #ifdef IPFIREWALL_VERBOSE
73 static int fw_verbose = 1;
75 static int fw_verbose = 0;
77 #ifdef IPFIREWALL_VERBOSE_LIMIT
78 static int fw_verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
80 static int fw_verbose_limit = 0;
84 * Right now, two fields in the IP header are changed to host format
85 * by the IP layer before calling the firewall. Ideally, we would like
86 * to have them in network format so that the packet can be
87 * used as it comes from the device driver (and is thus readonly).
90 static u_int64_t counter; /* counter for ipfw_report(NULL...) */
92 #define IPFW_DEFAULT_RULE ((u_int)(u_short)~0)
94 LIST_HEAD (ip_fw_head, ip_fw) ip_fw_chain_head;
96 MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's");
99 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
100 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, enable, CTLFLAG_RW,
101 &fw_enable, 0, "Enable ipfw");
102 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO,one_pass,CTLFLAG_RW,
104 "Only do a single pass through ipfw when using dummynet(4)");
105 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug, CTLFLAG_RW,
106 &fw_debug, 0, "Enable printing of debug ip_fw statements");
107 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose, CTLFLAG_RW,
108 &fw_verbose, 0, "Log matches to ipfw rules");
109 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW,
110 &fw_verbose_limit, 0, "Set upper limit of matches of ipfw rules logged");
113 * Extension for stateful ipfw.
115 * Dynamic rules are stored in lists accessed through a hash table
116 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can
117 * be modified through the sysctl variable dyn_buckets which is
118 * updated when the table becomes empty.
120 * XXX currently there is only one list, ipfw_dyn.
122 * When a packet is received, it is first hashed, then matched
123 * against the entries in the corresponding list.
124 * Matching occurs according to the rule type. The default is to
125 * match the four fields and the protocol, and rules are bidirectional.
127 * For a busy proxy/web server we will have lots of connections to
128 * the server. We could decide for a rule type where we ignore
129 * ports (different hashing) and avoid special SYN/RST/FIN handling.
131 * XXX when we decide to support more than one rule type, we should
132 * repeat the hashing multiple times uing only the useful fields.
133 * Or, we could run the various tests in parallel, because the
134 * 'move to front' technique should shorten the average search.
136 * The lifetime of dynamic rules is regulated by dyn_*_lifetime,
137 * measured in seconds and depending on the flags.
139 * The total number of dynamic rules is stored in dyn_count.
140 * The max number of dynamic rules is dyn_max. When we reach
141 * the maximum number of rules we do not create anymore. This is
142 * done to avoid consuming too much memory, but also too much
143 * time when searching on each packet (ideally, we should try instead
144 * to put a limit on the length of the list on each bucket...).
146 * Each dynamic rules holds a pointer to the parent ipfw rule so
147 * we know what action to perform. Dynamic rules are removed when
148 * the parent rule is deleted.
149 * There are some limitations with dynamic rules -- we do not
150 * obey the 'randomized match', and we do not do multiple
151 * passes through the firewall.
152 * XXX check the latter!!!
154 static struct ipfw_dyn_rule **ipfw_dyn_v = NULL ;
155 static u_int32_t dyn_buckets = 256 ; /* must be power of 2 */
156 static u_int32_t curr_dyn_buckets = 256 ; /* must be power of 2 */
159 * timeouts for various events in handing dynamic rules.
161 static u_int32_t dyn_ack_lifetime = 300 ;
162 static u_int32_t dyn_syn_lifetime = 20 ;
163 static u_int32_t dyn_fin_lifetime = 1 ;
164 static u_int32_t dyn_rst_lifetime = 1 ;
165 static u_int32_t dyn_udp_lifetime = 10 ;
166 static u_int32_t dyn_short_lifetime = 5 ;
169 * after reaching 0, dynamic rules are considered still valid for
170 * an additional grace time, unless there is lack of resources.
172 static u_int32_t dyn_grace_time = 10 ;
174 static u_int32_t static_count = 0 ; /* # of static rules */
175 static u_int32_t dyn_count = 0 ; /* # of dynamic rules */
176 static u_int32_t dyn_max = 1000 ; /* max # of dynamic rules */
178 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW,
179 &dyn_buckets, 0, "Number of dyn. buckets");
180 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD,
181 &curr_dyn_buckets, 0, "Current Number of dyn. buckets");
182 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD,
183 &dyn_count, 0, "Number of dyn. rules");
184 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW,
185 &dyn_max, 0, "Max number of dyn. rules");
186 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD,
187 &static_count, 0, "Number of static rules");
188 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW,
189 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks");
190 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW,
191 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn");
192 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW,
193 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin");
194 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW,
195 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst");
196 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW,
197 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP");
198 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW,
199 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations");
200 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_grace_time, CTLFLAG_RD,
201 &dyn_grace_time, 0, "Grace time for dyn. rules");
203 #endif /* SYSCTL_NODE */
205 #define dprintf(a) do { \
209 #define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0
211 static int add_entry (struct ip_fw_head *chainptr, struct ip_fw *frwl);
212 static int del_entry (struct ip_fw_head *chainptr, u_short number);
213 static int zero_entry (struct ip_fw *, int);
214 static int check_ipfw_struct (struct ip_fw *m);
215 static int iface_match (struct ifnet *ifp, union ip_fw_if *ifu,
217 static int ipopts_match (struct ip *ip, struct ip_fw *f);
219 port_match (u_short *portptr, int nports, u_short port,
220 int range_flag, int mask);
221 static int tcpflg_match (struct tcphdr *tcp, struct ip_fw *f);
222 static int icmptype_match (struct icmp * icmp, struct ip_fw * f);
223 static void ipfw_report (struct ip_fw *f, struct ip *ip, int ip_off,
224 int ip_len, struct ifnet *rif,
227 static void flush_rule_ptrs(void);
229 static ip_fw_chk_t ip_fw_chk;
230 static int ip_fw_ctl (struct sockopt *sopt);
232 ip_dn_ruledel_t *ip_dn_ruledel_ptr = NULL;
234 static char err_prefix[] = "ip_fw_ctl:";
237 * Returns 1 if the port is matched by the vector, 0 otherwise
240 port_match(u_short *portptr, int nports, u_short port, int range_flag, int mask)
245 if ( 0 == ((portptr[0] ^ port) & portptr[1]) )
251 if (portptr[0] <= port && port <= portptr[1])
257 if (*portptr++ == port)
263 tcpflg_match(struct tcphdr *tcp, struct ip_fw *f)
265 u_char flg_set, flg_clr;
268 * If an established connection is required, reject packets that
269 * have only SYN of RST|ACK|SYN set. Otherwise, fall through to
270 * other flag requirements.
272 if ((f->fw_ipflg & IP_FW_IF_TCPEST) &&
273 ((tcp->th_flags & (IP_FW_TCPF_RST | IP_FW_TCPF_ACK |
274 IP_FW_TCPF_SYN)) == IP_FW_TCPF_SYN))
277 flg_set = tcp->th_flags & f->fw_tcpf;
278 flg_clr = tcp->th_flags & f->fw_tcpnf;
280 if (flg_set != f->fw_tcpf)
289 icmptype_match(struct icmp *icmp, struct ip_fw *f)
293 if (!(f->fw_flg & IP_FW_F_ICMPBIT))
296 type = icmp->icmp_type;
298 /* check for matching type in the bitmap */
299 if (type < IP_FW_ICMPTYPES_MAX &&
300 (f->fw_uar.fw_icmptypes[type / (sizeof(unsigned) * NBBY)] &
301 (1U << (type % (sizeof(unsigned) * NBBY)))))
304 return(0); /* no match */
308 is_icmp_query(struct ip *ip)
310 const struct icmp *icmp;
313 icmp = (struct icmp *)((u_int32_t *)ip + ip->ip_hl);
314 icmp_type = icmp->icmp_type;
316 if (icmp_type == ICMP_ECHO || icmp_type == ICMP_ROUTERSOLICIT ||
317 icmp_type == ICMP_TSTAMP || icmp_type == ICMP_IREQ ||
318 icmp_type == ICMP_MASKREQ)
325 ipopts_match(struct ip *ip, struct ip_fw *f)
328 int opt, optlen, cnt;
329 u_char opts, nopts, nopts_sve;
331 cp = (u_char *)(ip + 1);
332 cnt = (ip->ip_hl << 2) - sizeof (struct ip);
334 nopts = nopts_sve = f->fw_ipnopt;
336 for (; cnt > 0; cnt -= optlen, cp += optlen) {
337 opt = cp[IPOPT_OPTVAL];
338 if (opt == IPOPT_EOL)
340 if (opt == IPOPT_NOP)
343 optlen = cp[IPOPT_OLEN];
344 if (optlen <= 0 || optlen > cnt) {
354 opts &= ~IP_FW_IPOPT_LSRR;
355 nopts &= ~IP_FW_IPOPT_LSRR;
359 opts &= ~IP_FW_IPOPT_SSRR;
360 nopts &= ~IP_FW_IPOPT_SSRR;
364 opts &= ~IP_FW_IPOPT_RR;
365 nopts &= ~IP_FW_IPOPT_RR;
368 opts &= ~IP_FW_IPOPT_TS;
369 nopts &= ~IP_FW_IPOPT_TS;
375 if (opts == 0 && nopts == nopts_sve)
382 tcpopts_match(struct tcphdr *tcp, struct ip_fw *f)
385 int opt, optlen, cnt;
386 u_char opts, nopts, nopts_sve;
388 cp = (u_char *)(tcp + 1);
389 cnt = (tcp->th_off << 2) - sizeof (struct tcphdr);
391 nopts = nopts_sve = f->fw_tcpnopt;
393 for (; cnt > 0; cnt -= optlen, cp += optlen) {
395 if (opt == TCPOPT_EOL)
397 if (opt == TCPOPT_NOP)
412 opts &= ~IP_FW_TCPOPT_MSS;
413 nopts &= ~IP_FW_TCPOPT_MSS;
417 opts &= ~IP_FW_TCPOPT_WINDOW;
418 nopts &= ~IP_FW_TCPOPT_WINDOW;
421 case TCPOPT_SACK_PERMITTED:
423 opts &= ~IP_FW_TCPOPT_SACK;
424 nopts &= ~IP_FW_TCPOPT_SACK;
427 case TCPOPT_TIMESTAMP:
428 opts &= ~IP_FW_TCPOPT_TS;
429 nopts &= ~IP_FW_TCPOPT_TS;
435 opts &= ~IP_FW_TCPOPT_CC;
436 nopts &= ~IP_FW_TCPOPT_CC;
442 if (opts == 0 && nopts == nopts_sve)
449 iface_match(struct ifnet *ifp, union ip_fw_if *ifu, int byname)
451 /* Check by name or by IP address */
454 if (ifu->fu_via_if.glob) {
455 if (fnmatch(ifu->fu_via_if.name, ifp->if_xname, 0)
459 if (strncmp(ifp->if_xname, ifu->fu_via_if.name,
464 } else if (ifu->fu_via_ip.s_addr != 0) { /* Zero == wildcard */
467 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
468 if (ia->ifa_addr == NULL)
470 if (ia->ifa_addr->sa_family != AF_INET)
472 if (ifu->fu_via_ip.s_addr != ((struct sockaddr_in *)
473 (ia->ifa_addr))->sin_addr.s_addr)
483 ipfw_report(struct ip_fw *f, struct ip *ip, int ip_off, int ip_len,
484 struct ifnet *rif, struct ifnet *oif)
486 struct tcphdr *const tcp = (struct tcphdr *) ((u_int32_t *) ip+ ip->ip_hl);
487 struct udphdr *const udp = (struct udphdr *) ((u_int32_t *) ip+ ip->ip_hl);
488 struct icmp *const icmp = (struct icmp *) ((u_int32_t *) ip + ip->ip_hl);
491 char action2[32], proto[47], name[18], fragment[27];
493 int offset = ip_off & IP_OFFMASK;
495 count = f ? f->fw_pcnt : ++counter;
496 if ((f == NULL && fw_verbose_limit != 0 && count > fw_verbose_limit) ||
497 (f && f->fw_logamount != 0 && count > f->fw_loghighest))
500 /* Print command name */
501 snprintf(SNPARGS(name, 0), "ipfw: %d", f ? f->fw_number : -1);
507 switch (f->fw_flg & IP_FW_F_COMMAND) {
512 if (f->fw_reject_code == IP_FW_REJECT_RST)
525 snprintf(SNPARGS(action2, 0), "Divert %d",
529 snprintf(SNPARGS(action2, 0), "Tee %d",
534 snprintf(SNPARGS(action2, 0), "SkipTo %d",
538 snprintf(SNPARGS(action2, 0), "Pipe %d",
542 snprintf(SNPARGS(action2, 0), "Queue %d",
547 if (f->fw_fwd_ip.sin_port)
548 snprintf(SNPARGS(action2, 0),
550 inet_ntoa(f->fw_fwd_ip.sin_addr),
551 f->fw_fwd_ip.sin_port);
553 snprintf(SNPARGS(action2, 0), "Forward to %s",
554 inet_ntoa(f->fw_fwd_ip.sin_addr));
565 len = snprintf(SNPARGS(proto, 0), "TCP %s",
566 inet_ntoa(ip->ip_src));
568 len += snprintf(SNPARGS(proto, len), ":%d ",
569 ntohs(tcp->th_sport));
571 len += snprintf(SNPARGS(proto, len), " ");
572 len += snprintf(SNPARGS(proto, len), "%s",
573 inet_ntoa(ip->ip_dst));
575 snprintf(SNPARGS(proto, len), ":%d",
576 ntohs(tcp->th_dport));
579 len = snprintf(SNPARGS(proto, 0), "UDP %s",
580 inet_ntoa(ip->ip_src));
582 len += snprintf(SNPARGS(proto, len), ":%d ",
583 ntohs(udp->uh_sport));
585 len += snprintf(SNPARGS(proto, len), " ");
586 len += snprintf(SNPARGS(proto, len), "%s",
587 inet_ntoa(ip->ip_dst));
589 snprintf(SNPARGS(proto, len), ":%d",
590 ntohs(udp->uh_dport));
594 len = snprintf(SNPARGS(proto, 0), "ICMP:%u.%u ",
595 icmp->icmp_type, icmp->icmp_code);
597 len = snprintf(SNPARGS(proto, 0), "ICMP ");
598 len += snprintf(SNPARGS(proto, len), "%s",
599 inet_ntoa(ip->ip_src));
600 snprintf(SNPARGS(proto, len), " %s", inet_ntoa(ip->ip_dst));
603 len = snprintf(SNPARGS(proto, 0), "P:%d %s", ip->ip_p,
604 inet_ntoa(ip->ip_src));
605 snprintf(SNPARGS(proto, len), " %s", inet_ntoa(ip->ip_dst));
609 if (ip_off & (IP_MF | IP_OFFMASK))
610 snprintf(SNPARGS(fragment, 0), " (frag %d:%d@%d%s)",
611 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2),
613 (ip_off & IP_MF) ? "+" : "");
617 log(LOG_SECURITY | LOG_INFO, "%s %s %s out via %s%s\n",
618 name, action, proto, oif->if_xname, fragment);
620 log(LOG_SECURITY | LOG_INFO, "%s %s %s in via %s%s\n", name,
621 action, proto, rif->if_xname, fragment);
623 log(LOG_SECURITY | LOG_INFO, "%s %s %s%s\n", name, action,
625 if ((f ? f->fw_logamount != 0 : 1) &&
626 count == (f ? f->fw_loghighest : fw_verbose_limit))
627 log(LOG_SECURITY | LOG_NOTICE,
628 "ipfw: limit %d reached on entry %d\n",
629 f ? f->fw_logamount : fw_verbose_limit,
630 f ? f->fw_number : -1);
634 hash_packet(struct ipfw_flow_id *id)
638 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port);
639 i &= (curr_dyn_buckets - 1) ;
644 * unlink a dynamic rule from a chain. prev is a pointer to
645 * the previous one, q is a pointer to the rule to delete,
646 * head is a pointer to the head of the queue.
647 * Modifies q and potentially also head.
649 #define UNLINK_DYN_RULE(prev, head, q) { \
650 struct ipfw_dyn_rule *old_q = q; \
652 /* remove a refcount to the parent */ \
653 if (q->dyn_type == DYN_LIMIT) \
654 q->parent->count--; \
655 DEB(printf("-- unlink entry 0x%08x %d -> 0x%08x %d, %d left\n", \
656 (q->id.src_ip), (q->id.src_port), \
657 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \
659 prev->next = q = q->next ; \
661 ipfw_dyn_v[i] = q = q->next ; \
663 free(old_q, M_IPFW); }
665 #define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0)
667 * Remove all dynamic rules pointing to a given rule, or all
668 * rules if rule == NULL. Second parameter is 1 if we want to
669 * delete unconditionally, otherwise only expired rules are removed.
672 remove_dyn_rule(struct ip_fw *rule, int force)
674 struct ipfw_dyn_rule *prev, *q;
675 int i, pass, max_pass ;
676 static u_int32_t last_remove = 0 ;
678 if (ipfw_dyn_v == NULL || dyn_count == 0)
680 /* do not expire more than once per second, it is useless */
681 if (force == 0 && last_remove == time_second)
683 last_remove = time_second ;
686 * because DYN_LIMIT refer to parent rules, during the first pass only
687 * remove child and mark any pending LIMIT_PARENT, and remove
688 * them in a second pass.
690 for (pass = max_pass = 0; pass <= max_pass ; pass++ ) {
691 for (i = 0 ; i < curr_dyn_buckets ; i++) {
692 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) {
694 * logic can become complex here, so we split tests.
695 * First, test if we match any rule,
696 * then make sure the rule is expired or we want to kill it,
697 * and possibly more in the future.
699 int zap = ( rule == NULL || rule == q->rule);
701 zap = force || TIME_LEQ( q->expire , time_second );
702 /* do not zap parent in first pass, record we need a second pass */
703 if (zap && q->dyn_type == DYN_LIMIT_PARENT) {
704 max_pass = 1; /* we need a second pass */
705 if (pass == 0 || q->count != 0) {
707 if (pass == 1 && force) /* should not happen */
708 printf("OUCH! cannot remove rule, count %d\n",
713 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
723 #define EXPIRE_DYN_CHAIN(rule) remove_dyn_rule(rule, 0 /* expired ones */)
724 #define EXPIRE_DYN_CHAINS() remove_dyn_rule(NULL, 0 /* expired ones */)
725 #define DELETE_DYN_CHAIN(rule) remove_dyn_rule(rule, 1 /* force removal */)
726 #define DELETE_DYN_CHAINS() remove_dyn_rule(NULL, 1 /* force removal */)
729 * lookup a dynamic rule.
731 static struct ipfw_dyn_rule *
732 lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction)
735 * stateful ipfw extensions.
736 * Lookup into dynamic session queue
738 struct ipfw_dyn_rule *prev, *q ;
740 #define MATCH_FORWARD 1
742 if (ipfw_dyn_v == NULL)
744 i = hash_packet( pkt );
745 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) {
746 if (q->dyn_type == DYN_LIMIT_PARENT)
748 if (TIME_LEQ( q->expire , time_second ) ) { /* expire entry */
749 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
752 if ( pkt->proto == q->id.proto) {
753 if (pkt->src_ip == q->id.src_ip &&
754 pkt->dst_ip == q->id.dst_ip &&
755 pkt->src_port == q->id.src_port &&
756 pkt->dst_port == q->id.dst_port ) {
757 dir = MATCH_FORWARD ;
760 if (pkt->src_ip == q->id.dst_ip &&
761 pkt->dst_ip == q->id.src_ip &&
762 pkt->src_port == q->id.dst_port &&
763 pkt->dst_port == q->id.src_port ) {
764 dir = 0 ; /* reverse match */
772 return NULL ; /* clearly not found */
774 if ( prev != NULL) { /* found and not in front */
775 prev->next = q->next ;
776 q->next = ipfw_dyn_v[i] ;
779 if (pkt->proto == IPPROTO_TCP) {
780 /* update state according to flags */
781 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST);
782 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8);
786 q->expire = time_second + dyn_syn_lifetime ;
788 case TH_SYN | (TH_SYN << 8) :
789 /* move to established */
790 q->expire = time_second + dyn_ack_lifetime ;
792 case TH_SYN | (TH_SYN << 8) | TH_FIN :
793 case TH_SYN | (TH_SYN << 8) | (TH_FIN << 8) :
794 /* one side tries to close */
795 q->expire = time_second + dyn_ack_lifetime ;
797 case TH_SYN | (TH_SYN << 8) | TH_FIN | (TH_FIN << 8) :
798 /* both sides closed */
799 q->expire = time_second + dyn_fin_lifetime ;
804 * reset or some invalid combination, but can also
805 * occur if we use keep-state the wrong way.
807 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0)
808 printf("invalid state: 0x%x\n", q->state);
810 q->expire = time_second + dyn_rst_lifetime ;
813 } else if (pkt->proto == IPPROTO_UDP) {
814 q->expire = time_second + dyn_udp_lifetime ;
816 /* other protocols */
817 q->expire = time_second + dyn_short_lifetime ;
820 *match_direction = dir ;
825 * Install state of type 'type' for a dynamic session.
826 * The hash table contains two type of rules:
827 * - regular rules (DYN_KEEP_STATE)
828 * - rules for sessions with limited number of sess per user
829 * (DYN_LIMIT). When they are created, the parent is
830 * increased by 1, and decreased on delete. In this case,
831 * the third parameter is the parent rule and not the chain.
832 * - "parent" rules for the above (DYN_LIMIT_PARENT).
835 static struct ipfw_dyn_rule *
836 add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule)
838 struct ipfw_dyn_rule *r ;
841 if (ipfw_dyn_v == NULL ||
842 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) {
843 /* try reallocation, make sure we have a power of 2 */
844 u_int32_t i = dyn_buckets ;
845 while ( i > 0 && (i & 1) == 0 )
847 if (i != 1) /* not a power of 2 */
848 dyn_buckets = curr_dyn_buckets ; /* reset */
850 curr_dyn_buckets = dyn_buckets ;
851 if (ipfw_dyn_v != NULL)
852 free(ipfw_dyn_v, M_IPFW);
853 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof r,
854 M_IPFW, M_WAITOK | M_ZERO);
855 if (ipfw_dyn_v == NULL)
856 return NULL; /* failed ! */
861 r = malloc(sizeof *r, M_IPFW, M_WAITOK | M_ZERO);
863 printf ("sorry cannot allocate state\n");
867 /* increase refcount on parent, and set pointer */
868 if (dyn_type == DYN_LIMIT) {
869 struct ipfw_dyn_rule *parent = (struct ipfw_dyn_rule *)rule;
870 if ( parent->dyn_type != DYN_LIMIT_PARENT)
871 panic("invalid parent");
878 r->expire = time_second + dyn_syn_lifetime ;
880 r->dyn_type = dyn_type ;
881 r->pcnt = r->bcnt = 0 ;
885 r->next = ipfw_dyn_v[i] ;
888 DEB(printf("-- add entry 0x%08x %d -> 0x%08x %d, total %d\n",
889 (r->id.src_ip), (r->id.src_port),
890 (r->id.dst_ip), (r->id.dst_port),
896 * lookup dynamic parent rule using pkt and rule as search keys.
897 * If the lookup fails, then install one.
899 static struct ipfw_dyn_rule *
900 lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule)
902 struct ipfw_dyn_rule *q;
906 i = hash_packet( pkt );
907 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next)
908 if (q->dyn_type == DYN_LIMIT_PARENT && rule == q->rule &&
909 pkt->proto == q->id.proto &&
910 pkt->src_ip == q->id.src_ip &&
911 pkt->dst_ip == q->id.dst_ip &&
912 pkt->src_port == q->id.src_port &&
913 pkt->dst_port == q->id.dst_port) {
914 q->expire = time_second + dyn_short_lifetime ;
915 DEB(printf("lookup_dyn_parent found 0x%p\n", q);)
919 return add_dyn_rule(pkt, DYN_LIMIT_PARENT, rule);
923 * Install dynamic state.
924 * There are different types of dynamic rules which can be installed.
925 * The type is in rule->dyn_type.
926 * Type 0 (default) is a bidirectional rule
928 * Returns 1 (failure) if state is not installed because of errors or because
929 * session limitations are enforced.
932 install_state(struct ip_fw *rule, struct ip_fw_args *args)
934 struct ipfw_dyn_rule *q ;
935 static int last_log ;
937 u_int8_t type = rule->dyn_type ;
939 DEB(printf("-- install state type %d 0x%08x %u -> 0x%08x %u\n",
941 (args->f_id.src_ip), (args->f_id.src_port),
942 (args->f_id.dst_ip), (args->f_id.dst_port) );)
944 q = lookup_dyn_rule(&args->f_id, NULL) ;
945 if (q != NULL) { /* should never occur */
946 if (last_log != time_second) {
947 last_log = time_second ;
948 printf(" entry already present, done\n");
952 if (dyn_count >= dyn_max) /* try remove old ones... */
954 if (dyn_count >= dyn_max) {
955 if (last_log != time_second) {
956 last_log = time_second ;
957 printf(" Too many dynamic rules, sorry\n");
959 return 1; /* cannot install, notify caller */
963 case DYN_KEEP_STATE: /* bidir rule */
964 add_dyn_rule(&args->f_id, DYN_KEEP_STATE, rule);
966 case DYN_LIMIT: /* limit number of sessions */
968 u_int16_t limit_mask = rule->limit_mask ;
969 u_int16_t conn_limit = rule->conn_limit ;
970 struct ipfw_flow_id id;
971 struct ipfw_dyn_rule *parent;
973 DEB(printf("installing dyn-limit rule %d\n", conn_limit);)
975 id.dst_ip = id.src_ip = 0;
976 id.dst_port = id.src_port = 0 ;
977 id.proto = args->f_id.proto ;
979 if (limit_mask & DYN_SRC_ADDR)
980 id.src_ip = args->f_id.src_ip;
981 if (limit_mask & DYN_DST_ADDR)
982 id.dst_ip = args->f_id.dst_ip;
983 if (limit_mask & DYN_SRC_PORT)
984 id.src_port = args->f_id.src_port;
985 if (limit_mask & DYN_DST_PORT)
986 id.dst_port = args->f_id.dst_port;
987 parent = lookup_dyn_parent(&id, rule);
988 if (parent == NULL) {
989 printf("add parent failed\n");
992 if (parent->count >= conn_limit) {
993 EXPIRE_DYN_CHAIN(rule); /* try to expire some */
995 * The expiry might have removed the parent too.
996 * We lookup again, which will re-create if necessary.
998 parent = lookup_dyn_parent(&id, rule);
999 if (parent == NULL) {
1000 printf("add parent failed\n");
1003 if (parent->count >= conn_limit) {
1004 if (fw_verbose && last_log != time_second) {
1005 last_log = time_second;
1006 log(LOG_SECURITY | LOG_DEBUG,
1007 "drop session, too many entries\n");
1012 add_dyn_rule(&args->f_id, DYN_LIMIT, (struct ip_fw *)parent);
1016 printf("unknown dynamic rule type %u\n", type);
1019 lookup_dyn_rule(&args->f_id, NULL) ; /* XXX just set the lifetime */
1024 * given an ip_fw *, lookup_next_rule will return a pointer
1025 * of the same type to the next one. This can be either the jump
1026 * target (for skipto instructions) or the next one in the list (in
1027 * all other cases including a missing jump target).
1028 * Backward jumps are not allowed, so start looking from the next
1031 static struct ip_fw * lookup_next_rule(struct ip_fw *me);
1033 static struct ip_fw *
1034 lookup_next_rule(struct ip_fw *me)
1036 struct ip_fw *rule ;
1037 int rulenum = me->fw_skipto_rule ; /* guess... */
1039 if ( (me->fw_flg & IP_FW_F_COMMAND) == IP_FW_F_SKIPTO )
1040 for (rule = LIST_NEXT(me,next); rule ; rule = LIST_NEXT(rule,next))
1041 if (rule->fw_number >= rulenum)
1043 return LIST_NEXT(me,next) ; /* failure or not a skipto */
1049 * *m The packet; we set to NULL when/if we nuke it.
1050 * oif Outgoing interface, or NULL if packet is incoming
1051 * *cookie Skip up to the first rule past this rule number;
1052 * upon return, non-zero port number for divert or tee.
1053 * Special case: cookie == NULL on input for bridging.
1054 * *flow_id pointer to the last matching rule (in/out)
1055 * *next_hop socket we are forwarding to (in/out).
1059 * IP_FW_PORT_DENY_FLAG the packet must be dropped.
1060 * 0 The packet is to be accepted and routed normally OR
1061 * the packet was denied/rejected and has been dropped;
1062 * in the latter case, *m is equal to NULL upon return.
1063 * port Divert the packet to port, with these caveats:
1065 * - If IP_FW_PORT_TEE_FLAG is set, tee the packet instead
1066 * of diverting it (ie, 'ipfw tee').
1068 * - If IP_FW_PORT_DYNT_FLAG is set, interpret the lower
1069 * 16 bits as a dummynet pipe number instead of diverting
1073 ip_fw_chk(struct ip_fw_args *args)
1076 * grab things into variables to minimize diffs.
1077 * XXX this has to be cleaned up later.
1079 struct mbuf **m = &(args->m);
1080 struct ifnet *oif = args->oif;
1082 struct ip_fw **flow_id = &(args->rule);
1083 struct sockaddr_in **next_hop = &(args->next_hop);
1085 struct ip_fw *f = NULL; /* matching rule */
1086 struct ip *ip = mtod(*m, struct ip *);
1087 struct ifnet *const rif = (*m)->m_pkthdr.rcvif;
1091 u_short ip_off=0, offset = 0;
1092 /* local copy of addresses for faster matching */
1093 u_short src_port = 0, dst_port = 0;
1094 struct in_addr src_ip, dst_ip;
1095 u_int8_t proto= 0, flags = 0;
1099 int dyn_checked = 0 ; /* set after dyn.rules have been checked. */
1100 int direction = MATCH_FORWARD ; /* dirty trick... */
1101 struct ipfw_dyn_rule *q = NULL ;
1103 /* Special hack for bridging (as usual) */
1104 #define BRIDGED (args->eh != NULL)
1105 if (BRIDGED) { /* this is a bridged packet */
1106 if ( (*m)->m_pkthdr.len >= sizeof(struct ip) &&
1107 ntohs(args->eh->ether_type) == ETHERTYPE_IP)
1108 hlen = ip->ip_hl << 2;
1110 return 0; /* XXX ipfw1 always accepts non-ip pkts */
1112 hlen = ip->ip_hl << 2;
1114 /* Grab and reset cookie */
1115 if ((mtag = m_tag_find(*m, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL) {
1116 skipto = *(u_int16_t *)m_tag_data(mtag);
1117 m_tag_delete(*m, mtag);
1123 #define PULLUP_TO(len) do { \
1124 if ((*m)->m_len < (len)) { \
1125 if ((*m = m_pullup(*m, (len))) == 0) \
1127 ip = mtod(*m, struct ip *); \
1131 if (hlen > 0) { /* this is an IP packet */
1133 * Collect parameters into local variables for faster matching.
1136 src_ip = ip->ip_src;
1137 dst_ip = ip->ip_dst;
1138 if (BRIDGED) { /* bridged packets are as on the wire */
1139 ip_off = ntohs(ip->ip_off);
1140 ip_len = ntohs(ip->ip_len);
1142 ip_off = ip->ip_off;
1143 ip_len = ip->ip_len;
1145 offset = ip_off & IP_OFFMASK;
1148 case IPPROTO_TCP : {
1151 PULLUP_TO(hlen + sizeof(struct tcphdr));
1152 tcp =(struct tcphdr *)((u_int32_t *)ip + ip->ip_hl);
1153 dst_port = tcp->th_dport ;
1154 src_port = tcp->th_sport ;
1155 flags = tcp->th_flags ;
1159 case IPPROTO_UDP : {
1162 PULLUP_TO(hlen + sizeof(struct udphdr));
1163 udp =(struct udphdr *)((u_int32_t *)ip + ip->ip_hl);
1164 dst_port = udp->uh_dport ;
1165 src_port = udp->uh_sport ;
1170 PULLUP_TO(hlen + 4); /* type, code and checksum. */
1171 flags = ((struct icmp *)
1172 ((u_int32_t *)ip + ip->ip_hl))->icmp_type ;
1181 args->f_id.src_ip = ntohl(src_ip.s_addr);
1182 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1183 args->f_id.proto = proto;
1184 args->f_id.src_port = ntohs(src_port);
1185 args->f_id.dst_port = ntohs(dst_port);
1186 args->f_id.flags = flags;
1190 * Packet has already been tagged. Look for the next rule
1191 * to restart processing.
1193 if (fw_one_pass) /* just accept if fw_one_pass is set */
1196 f = (*flow_id)->next_rule_ptr ;
1198 f = (*flow_id)->next_rule_ptr = lookup_next_rule(*flow_id);
1203 * Go down the list, looking for enlightment.
1204 * If we've been asked to start at a given rule, do so.
1206 f = LIST_FIRST(&ip_fw_chain_head);
1208 if (skipto >= IPFW_DEFAULT_RULE)
1210 while (f && f->fw_number <= skipto)
1211 f = LIST_NEXT(f, next);
1217 for (; f; f = LIST_NEXT(f, next)) {
1219 if (f->fw_number == IPFW_DEFAULT_RULE)
1223 * dynamic rules are checked at the first keep-state or
1224 * check-state occurrence.
1226 if (f->fw_flg & (IP_FW_F_KEEP_S|IP_FW_F_CHECK_S) &&
1227 dyn_checked == 0 ) {
1229 q = lookup_dyn_rule(&args->f_id, &direction);
1231 DEB(printf("-- dynamic match 0x%08x %d %s 0x%08x %d\n",
1232 (q->id.src_ip), (q->id.src_port),
1233 (direction == MATCH_FORWARD ? "-->" : "<--"),
1234 (q->id.dst_ip), (q->id.dst_port) ); )
1238 goto got_match ; /* random not allowed here */
1240 /* if this was a check-only rule, continue with next */
1241 if (f->fw_flg & IP_FW_F_CHECK_S)
1245 /* Check if rule only valid for bridged packets */
1246 if ((f->fw_flg & IP_FW_BRIDGED) != 0 && !(BRIDGED))
1250 /* Check direction outbound */
1251 if (!(f->fw_flg & IP_FW_F_OUT))
1254 /* Check direction inbound */
1255 if (!(f->fw_flg & IP_FW_F_IN))
1260 if ((f->fw_flg & IP_FW_F_FRAG) && offset == 0 )
1263 if (f->fw_flg & IP_FW_F_SME) {
1264 INADDR_TO_IFP(src_ip, tif);
1268 if (f->fw_flg & IP_FW_F_DME) {
1269 INADDR_TO_IFP(dst_ip, tif);
1273 /* If src-addr doesn't match, not this rule. */
1274 if (((f->fw_flg & IP_FW_F_INVSRC) != 0) ^ ((src_ip.s_addr
1275 & f->fw_smsk.s_addr) != f->fw_src.s_addr))
1278 /* If dest-addr doesn't match, not this rule. */
1279 if (((f->fw_flg & IP_FW_F_INVDST) != 0) ^ ((dst_ip.s_addr
1280 & f->fw_dmsk.s_addr) != f->fw_dst.s_addr))
1283 /* Interface check */
1284 if ((f->fw_flg & IF_FW_F_VIAHACK) == IF_FW_F_VIAHACK) {
1285 struct ifnet *const iface = oif ? oif : rif;
1287 /* Backwards compatibility hack for "via" */
1288 if (!iface || !iface_match(iface,
1289 &f->fw_in_if, f->fw_flg & IP_FW_F_OIFNAME))
1292 /* Check receive interface */
1293 if ((f->fw_flg & IP_FW_F_IIFACE)
1294 && (!rif || !iface_match(rif,
1295 &f->fw_in_if, f->fw_flg & IP_FW_F_IIFNAME)))
1297 /* Check outgoing interface */
1298 if ((f->fw_flg & IP_FW_F_OIFACE)
1299 && (!oif || !iface_match(oif,
1300 &f->fw_out_if, f->fw_flg & IP_FW_F_OIFNAME)))
1304 /* Check IP options */
1305 if (f->fw_ipopt != f->fw_ipnopt && !ipopts_match(ip, f))
1308 /* Check protocol; if wildcard, and no [ug]id, match */
1309 if (f->fw_prot == IPPROTO_IP) {
1310 if (!(f->fw_flg & (IP_FW_F_UID|IP_FW_F_GID)))
1311 goto rnd_then_got_match;
1313 /* If different, don't match */
1314 if (proto != f->fw_prot)
1317 /* Protocol specific checks for uid only */
1318 if (f->fw_flg & (IP_FW_F_UID|IP_FW_F_GID)) {
1324 if (offset == 1) /* cf. RFC 1858 */
1330 P = in_pcblookup_hash(&tcbinfo[mycpu->gd_cpuid],
1331 dst_ip, dst_port, src_ip, src_port, 0, oif);
1333 P = in_pcblookup_hash(&tcbinfo[mycpu->gd_cpuid],
1334 src_ip, src_port, dst_ip, dst_port, 0, NULL);
1336 if (P && P->inp_socket) {
1337 if (f->fw_flg & IP_FW_F_UID) {
1338 if (P->inp_socket->so_cred->cr_uid !=
1341 } else if (!groupmember(f->fw_gid,
1342 P->inp_socket->so_cred))
1357 P = in_pcblookup_hash(&udbinfo, dst_ip,
1358 dst_port, src_ip, src_port, 1,
1361 P = in_pcblookup_hash(&udbinfo, src_ip,
1362 src_port, dst_ip, dst_port, 1,
1365 if (P && P->inp_socket) {
1366 if (f->fw_flg & IP_FW_F_UID) {
1367 if (P->inp_socket->so_cred->cr_uid !=
1370 } else if (!groupmember(f->fw_gid,
1371 P->inp_socket->so_cred))
1383 /* Protocol specific checks */
1389 if (offset == 1) /* cf. RFC 1858 */
1393 * TCP flags and ports aren't available in this
1394 * packet -- if this rule specified either one,
1395 * we consider the rule a non-match.
1397 if (IP_FW_HAVEPORTS(f) != 0 ||
1398 f->fw_tcpopt != f->fw_tcpnopt ||
1399 f->fw_tcpf != f->fw_tcpnf)
1404 tcp = (struct tcphdr *) ((u_int32_t *)ip + ip->ip_hl);
1406 if (f->fw_tcpopt != f->fw_tcpnopt && !tcpopts_match(tcp, f))
1408 if (((f->fw_tcpf != f->fw_tcpnf) ||
1409 (f->fw_ipflg & IP_FW_IF_TCPEST)) &&
1410 !tcpflg_match(tcp, f))
1418 * Port specification is unavailable -- if this
1419 * rule specifies a port, we consider the rule
1422 if (IP_FW_HAVEPORTS(f) )
1428 if (!port_match(&f->fw_uar.fw_pts[0],
1429 IP_FW_GETNSRCP(f), ntohs(src_port),
1430 f->fw_flg & IP_FW_F_SRNG,
1431 f->fw_flg & IP_FW_F_SMSK))
1433 if (!port_match(&f->fw_uar.fw_pts[IP_FW_GETNSRCP(f)],
1434 IP_FW_GETNDSTP(f), ntohs(dst_port),
1435 f->fw_flg & IP_FW_F_DRNG,
1436 f->fw_flg & IP_FW_F_DMSK))
1444 if (offset != 0) /* Type isn't valid */
1446 icmp = (struct icmp *) ((u_int32_t *)ip + ip->ip_hl);
1447 if (!icmptype_match(icmp, f))
1458 ipfw_report(NULL, ip, ip_off, ip_len, rif, oif);
1460 printf("pullup failed\n");
1467 if ( f->dont_match_prob && random() < f->dont_match_prob )
1471 * If not a dynamic match (q == NULL) and keep-state, install
1472 * a new dynamic entry.
1474 if (q == NULL && f->fw_flg & IP_FW_F_KEEP_S) {
1475 if (install_state(f, args)) /* error or limit violation */
1478 /* Update statistics */
1480 f->fw_bcnt += ip_len;
1481 f->timestamp = time_second;
1483 /* Log to console if desired */
1484 if ((f->fw_flg & IP_FW_F_PRN) && fw_verbose && hlen > 0)
1485 ipfw_report(f, ip, offset, ip_len, rif, oif);
1487 /* Take appropriate action */
1488 switch (f->fw_flg & IP_FW_F_COMMAND) {
1489 case IP_FW_F_ACCEPT:
1494 case IP_FW_F_DIVERT:
1495 mtag = m_tag_get(PACKET_TAG_IPFW_DIVERT,
1496 sizeof(u_int16_t), M_NOWAIT);
1499 *(u_int16_t *)m_tag_data(mtag) = f->fw_number;
1500 m_tag_prepend(*m, mtag);
1501 return(f->fw_divert_port);
1503 mtag = m_tag_get(PACKET_TAG_IPFW_DIVERT,
1504 sizeof(u_int16_t), M_NOWAIT);
1507 *(u_int16_t *)m_tag_data(mtag) = f->fw_number;
1508 m_tag_prepend(*m, mtag);
1509 return(f->fw_divert_port | IP_FW_PORT_TEE_FLAG);
1511 case IP_FW_F_SKIPTO: /* XXX check */
1512 if (f->next_rule_ptr == NULL)
1513 f->next_rule_ptr = lookup_next_rule(f) ;
1514 f = f->next_rule_ptr;
1521 *flow_id = f ; /* XXX set flow id */
1522 return(f->fw_pipe_nr | IP_FW_PORT_DYNT_FLAG);
1525 /* Change the next-hop address for this packet.
1526 * Initially we'll only worry about directly
1527 * reachable next-hop's, but ultimately
1528 * we will work out for next-hops that aren't
1529 * direct the route we would take for it. We
1530 * [cs]ould leave this latter problem to
1531 * ip_output.c. We hope to high [name the abode of
1532 * your favourite deity] that ip_output doesn't modify
1533 * the new value of next_hop (which is dst there)
1534 * XXX warning-- there is a dangerous reference here
1535 * from next_hop to a field within the rule. If the
1536 * rule is deleted, weird things might occur.
1538 if (next_hop != NULL /* Make sure, first... */
1539 && (q == NULL || direction == MATCH_FORWARD) )
1540 *next_hop = &(f->fw_fwd_ip);
1541 return(0); /* Allow the packet */
1545 /* Deny/reject this packet using this rule */
1549 /* Rule IPFW_DEFAULT_RULE should always be there and match */
1550 KASSERT(f != NULL, ("ip_fw: no chain"));
1553 * At this point, we're going to drop the packet.
1554 * Send a reject notice if all of the following are true:
1556 * - The packet matched a reject rule
1557 * - The packet is not an ICMP packet, or is an ICMP query packet
1558 * - The packet is not a multicast or broadcast packet
1560 if ((f->fw_flg & IP_FW_F_COMMAND) == IP_FW_F_REJECT
1561 && (proto != IPPROTO_ICMP || is_icmp_query(ip))
1562 && !((*m)->m_flags & (M_BCAST|M_MCAST))
1563 && !IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
1564 /* Must convert to host order for icmp_error() etc. */
1566 ip->ip_len = ntohs(ip->ip_len);
1567 ip->ip_off = ntohs(ip->ip_off);
1569 switch (f->fw_reject_code) {
1570 case IP_FW_REJECT_RST:
1572 /* XXX warning, this code writes into the mbuf */
1573 struct tcphdr *const tcp =
1574 (struct tcphdr *) ((u_int32_t *)ip + ip->ip_hl);
1575 struct tcpiphdr ti, *const tip = (struct tcpiphdr *) ip;
1577 if (offset != 0 || (tcp->th_flags & TH_RST))
1579 ti.ti_i = *((struct ipovly *) ip);
1581 bcopy(&ti, ip, sizeof(ti));
1582 tip->ti_seq = ntohl(tip->ti_seq);
1583 tip->ti_ack = ntohl(tip->ti_ack);
1584 tip->ti_len = ip_len - hlen - (tip->ti_off << 2);
1585 if (tcp->th_flags & TH_ACK) {
1586 tcp_respond(NULL, (void *)ip, tcp, *m,
1587 (tcp_seq)0, tcp->th_ack, TH_RST);
1589 if (tcp->th_flags & TH_SYN)
1591 tcp_respond(NULL, (void *)ip, tcp, *m,
1592 tip->ti_seq + tip->ti_len,
1593 (tcp_seq)0, TH_RST|TH_ACK);
1598 default: /* Send an ICMP unreachable using code */
1599 icmp_error(*m, ICMP_UNREACH,
1600 f->fw_reject_code, 0L, 0);
1608 * Finally, drop the packet.
1610 return(IP_FW_PORT_DENY_FLAG);
1615 * when a rule is added/deleted, zero the direct pointers within
1616 * all firewall rules. These will be reconstructed on the fly
1617 * as packets are matched.
1618 * Must be called at splimp().
1621 flush_rule_ptrs(void)
1625 LIST_FOREACH(fcp, &ip_fw_chain_head, next) {
1626 fcp->next_rule_ptr = NULL ;
1631 flush_pipe_ptrs(struct dn_flow_set *match)
1635 LIST_FOREACH(fcp, &ip_fw_chain_head, next) {
1636 if (match == NULL || fcp->pipe_ptr == match)
1637 fcp->pipe_ptr = NULL;
1642 add_entry(struct ip_fw_head *head, struct ip_fw *rule)
1644 struct ip_fw *ftmp, *fcp, *fcpl;
1647 ftmp = malloc(sizeof *ftmp, M_IPFW, M_WAITOK | M_ZERO);
1650 bcopy(rule, ftmp, sizeof(*ftmp));
1652 ftmp->fw_in_if.fu_via_if.name[FW_IFNLEN - 1] = '\0';
1655 ftmp->next_rule_ptr = NULL ;
1656 ftmp->pipe_ptr = NULL ;
1660 if (LIST_FIRST(head) == 0) {
1661 LIST_INSERT_HEAD(head, ftmp, next);
1665 /* If entry number is 0, find highest numbered rule and add 100 */
1666 if (ftmp->fw_number == 0) {
1667 LIST_FOREACH(fcp, head, next) {
1668 if (fcp->fw_number != IPFW_DEFAULT_RULE)
1669 nbr = fcp->fw_number;
1673 if (nbr < IPFW_DEFAULT_RULE - 100)
1675 ftmp->fw_number = rule->fw_number = nbr;
1678 /* Got a valid number; now insert it, keeping the list ordered */
1680 LIST_FOREACH(fcp, head, next) {
1681 if (fcp->fw_number > ftmp->fw_number) {
1683 LIST_INSERT_AFTER(fcpl, ftmp, next);
1685 LIST_INSERT_HEAD(head, ftmp, next);
1696 DEB(printf("++ installed rule %d, static count now %d\n",
1697 ftmp->fw_number, static_count);)
1702 * free storage associated with a static rule entry (including
1703 * dependent dynamic rules), and zeroes rule pointers to avoid
1704 * dangling pointer dereferences.
1705 * @return a pointer to the next entry.
1706 * Must be called at splimp() and with a non-null argument.
1708 static struct ip_fw *
1709 free_chain(struct ip_fw *fcp)
1713 n = LIST_NEXT(fcp, next);
1714 DELETE_DYN_CHAIN(fcp);
1715 LIST_REMOVE(fcp, next);
1717 if (DUMMYNET_LOADED)
1718 ip_dn_ruledel_ptr(fcp) ;
1719 flush_rule_ptrs(); /* more efficient to do outside the loop */
1725 * remove all rules with given number.
1728 del_entry(struct ip_fw_head *chainptr, u_short number)
1732 if (number != IPFW_DEFAULT_RULE) {
1733 LIST_FOREACH(rule, chainptr, next) {
1734 if (rule->fw_number == number) {
1735 crit_enter(); /* prevent access to rules while removing */
1736 while (rule && rule->fw_number == number)
1737 rule = free_chain(rule);
1738 /* XXX could move flush_rule_ptrs() here */
1748 * Reset some or all counters on firewall rules.
1749 * @arg frwl is null to clear all entries, or contains a specific
1751 * @arg log_only is 1 if we only want to reset logs, zero otherwise.
1755 zero_entry(struct ip_fw *frwl, int log_only)
1758 u_short number = 0 ;
1763 LIST_FOREACH(rule, &ip_fw_chain_head, next) {
1764 if (log_only == 0) {
1765 rule->fw_bcnt = rule->fw_pcnt = 0;
1766 rule->timestamp = 0;
1768 rule->fw_loghighest = rule->fw_pcnt+rule->fw_logamount;
1771 msg = log_only ? "ipfw: All logging counts cleared.\n" :
1772 "ipfw: Accounting cleared.\n";
1775 number = frwl->fw_number ;
1777 * It is possible to insert multiple chain entries with the
1778 * same number, so we don't stop after finding the first
1779 * match if zeroing a specific entry.
1781 LIST_FOREACH(rule, &ip_fw_chain_head, next)
1782 if (number == rule->fw_number) {
1784 while (rule && number == rule->fw_number) {
1785 if (log_only == 0) {
1786 rule->fw_bcnt = rule->fw_pcnt = 0;
1787 rule->timestamp = 0;
1789 rule->fw_loghighest = rule->fw_pcnt+ rule->fw_logamount;
1790 rule = LIST_NEXT(rule, next);
1796 if (!cleared) /* we did not find any matching rules */
1798 msg = log_only ? "ipfw: Entry %d logging count reset.\n" :
1799 "ipfw: Entry %d cleared.\n";
1802 log(LOG_SECURITY | LOG_NOTICE, msg, number);
1807 check_ipfw_struct(struct ip_fw *frwl)
1809 /* Check for invalid flag bits */
1810 if ((frwl->fw_flg & ~IP_FW_F_MASK) != 0) {
1811 dprintf(("%s undefined flag bits set (flags=%x)\n",
1812 err_prefix, frwl->fw_flg));
1815 if (frwl->fw_flg == IP_FW_F_CHECK_S) {
1819 /* Must apply to incoming or outgoing (or both) */
1820 if (!(frwl->fw_flg & (IP_FW_F_IN | IP_FW_F_OUT))) {
1821 dprintf(("%s neither in nor out\n", err_prefix));
1824 /* Empty interface name is no good */
1825 if (((frwl->fw_flg & IP_FW_F_IIFNAME)
1826 && !*frwl->fw_in_if.fu_via_if.name)
1827 || ((frwl->fw_flg & IP_FW_F_OIFNAME)
1828 && !*frwl->fw_out_if.fu_via_if.name)) {
1829 dprintf(("%s empty interface name\n", err_prefix));
1832 /* Sanity check interface matching */
1833 if ((frwl->fw_flg & IF_FW_F_VIAHACK) == IF_FW_F_VIAHACK) {
1834 ; /* allow "via" backwards compatibility */
1835 } else if ((frwl->fw_flg & IP_FW_F_IN)
1836 && (frwl->fw_flg & IP_FW_F_OIFACE)) {
1837 dprintf(("%s outgoing interface check on incoming\n",
1841 /* Sanity check port ranges */
1842 if ((frwl->fw_flg & IP_FW_F_SRNG) && IP_FW_GETNSRCP(frwl) < 2) {
1843 dprintf(("%s src range set but n_src_p=%d\n",
1844 err_prefix, IP_FW_GETNSRCP(frwl)));
1847 if ((frwl->fw_flg & IP_FW_F_DRNG) && IP_FW_GETNDSTP(frwl) < 2) {
1848 dprintf(("%s dst range set but n_dst_p=%d\n",
1849 err_prefix, IP_FW_GETNDSTP(frwl)));
1852 if (IP_FW_GETNSRCP(frwl) + IP_FW_GETNDSTP(frwl) > IP_FW_MAX_PORTS) {
1853 dprintf(("%s too many ports (%d+%d)\n",
1854 err_prefix, IP_FW_GETNSRCP(frwl), IP_FW_GETNDSTP(frwl)));
1858 * Protocols other than TCP/UDP don't use port range
1860 if ((frwl->fw_prot != IPPROTO_TCP) &&
1861 (frwl->fw_prot != IPPROTO_UDP) &&
1862 (IP_FW_GETNSRCP(frwl) || IP_FW_GETNDSTP(frwl))) {
1863 dprintf(("%s port(s) specified for non TCP/UDP rule\n",
1869 * Rather than modify the entry to make such entries work,
1870 * we reject this rule and require user level utilities
1871 * to enforce whatever policy they deem appropriate.
1873 if ((frwl->fw_src.s_addr & (~frwl->fw_smsk.s_addr)) ||
1874 (frwl->fw_dst.s_addr & (~frwl->fw_dmsk.s_addr))) {
1875 dprintf(("%s rule never matches\n", err_prefix));
1879 if ((frwl->fw_flg & IP_FW_F_FRAG) &&
1880 (frwl->fw_prot == IPPROTO_UDP || frwl->fw_prot == IPPROTO_TCP)) {
1881 if (IP_FW_HAVEPORTS(frwl)) {
1882 dprintf(("%s cannot mix 'frag' and ports\n", err_prefix));
1885 if (frwl->fw_prot == IPPROTO_TCP &&
1886 frwl->fw_tcpf != frwl->fw_tcpnf) {
1887 dprintf(("%s cannot mix 'frag' and TCP flags\n", err_prefix));
1892 /* Check command specific stuff */
1893 switch (frwl->fw_flg & IP_FW_F_COMMAND) {
1894 case IP_FW_F_REJECT:
1895 if (frwl->fw_reject_code >= 0x100
1896 && !(frwl->fw_prot == IPPROTO_TCP
1897 && frwl->fw_reject_code == IP_FW_REJECT_RST)) {
1898 dprintf(("%s unknown reject code\n", err_prefix));
1903 case IP_FW_F_DIVERT: /* Diverting to port zero is invalid */
1906 case IP_FW_F_PIPE: /* pipe 0 is invalid */
1907 case IP_FW_F_QUEUE: /* queue 0 is invalid */
1908 if (frwl->fw_divert_port == 0) {
1909 dprintf(("%s 0 is an invalid argument\n", err_prefix));
1914 case IP_FW_F_ACCEPT:
1916 case IP_FW_F_SKIPTO:
1922 dprintf(("%s invalid command\n", err_prefix));
1930 ip_fw_ctl(struct sockopt *sopt)
1935 struct ip_fw frwl, *bp , *buf;
1938 * Disallow modifications in really-really secure mode, but still allow
1939 * the logging counters to be reset.
1941 if (securelevel >= 3 && (sopt->sopt_name == IP_FW_ADD ||
1942 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)))
1946 switch (sopt->sopt_name) {
1949 * pass up a copy of the current rules. Static rules
1950 * come first (the last of which has number 65535),
1951 * followed by a possibly empty list of dynamic rule.
1952 * The last dynamic rule has NULL in the "next" field.
1955 /* size of static rules */
1956 size = static_count * sizeof(struct ip_fw) ;
1957 if (ipfw_dyn_v) /* add size of dyn.rules */
1958 size += (dyn_count * sizeof(struct ipfw_dyn_rule));
1961 * XXX todo: if the user passes a short length to know how
1962 * much room is needed, do not
1963 * bother filling up the buffer, just jump to the
1966 buf = malloc(size, M_TEMP, M_WAITOK);
1969 LIST_FOREACH(fcp, &ip_fw_chain_head, next) {
1970 bcopy(fcp, bp, sizeof *fcp);
1975 struct ipfw_dyn_rule *p, *dst, *last = NULL ;
1977 dst = (struct ipfw_dyn_rule *)bp ;
1978 for (i = 0 ; i < curr_dyn_buckets ; i++ )
1979 for ( p = ipfw_dyn_v[i] ; p != NULL ; p = p->next, dst++ ) {
1980 bcopy(p, dst, sizeof *p);
1981 dst->rule = (void *)(int)p->rule->fw_number;
1983 * store a non-null value in "next". The userland
1984 * code will interpret a NULL here as a marker
1985 * for the last dynamic rule.
1989 if (TIME_LEQ(dst->expire, time_second) )
1992 dst->expire -= time_second ;
1995 last->next = NULL ; /* mark last dynamic rule */
1999 error = sooptcopyout(sopt, buf, size);
2005 * Normally we cannot release the lock on each iteration.
2006 * We could do it here only because we start from the head all
2007 * the times so there is no risk of missing some entries.
2008 * On the other hand, the risk is that we end up with
2009 * a very inconsistent ruleset, so better keep the lock
2010 * around the whole cycle.
2012 * XXX this code can be improved by resetting the head of
2013 * the list to point to the default rule, and then freeing
2014 * the old list without the need for a lock.
2018 while ( (fcp = LIST_FIRST(&ip_fw_chain_head)) &&
2019 fcp->fw_number != IPFW_DEFAULT_RULE )
2025 error = sooptcopyin(sopt, &frwl, sizeof frwl, sizeof frwl);
2026 if (error || (error = check_ipfw_struct(&frwl)))
2029 if (frwl.fw_number == IPFW_DEFAULT_RULE) {
2030 dprintf(("%s can't add rule %u\n", err_prefix,
2031 (unsigned)IPFW_DEFAULT_RULE));
2034 error = add_entry(&ip_fw_chain_head, &frwl);
2035 if (!error && sopt->sopt_dir == SOPT_GET)
2036 error = sooptcopyout(sopt, &frwl, sizeof frwl);
2041 error = sooptcopyin(sopt, &frwl, sizeof frwl, sizeof frwl);
2045 if (frwl.fw_number == IPFW_DEFAULT_RULE) {
2046 dprintf(("%s can't delete rule %u\n", err_prefix,
2047 (unsigned)IPFW_DEFAULT_RULE));
2050 error = del_entry(&ip_fw_chain_head, frwl.fw_number);
2055 case IP_FW_RESETLOG:
2057 int cmd = (sopt->sopt_name == IP_FW_RESETLOG );
2060 if (sopt->sopt_val != 0) {
2061 error = sooptcopyin(sopt, &frwl, sizeof frwl, sizeof frwl);
2066 error = zero_entry(arg, cmd);
2071 printf("ip_fw_ctl invalid option %d\n", sopt->sopt_name);
2079 * dummynet needs a reference to the default rule, because rules can
2080 * be deleted while packets hold a reference to them (e.g. to resume
2081 * processing at the next rule). When this happens, dummynet changes
2082 * the reference to the default rule (probably it could well be a
2083 * NULL pointer, but this way we do not need to check for the special
2084 * case, plus here he have info on the default behaviour.
2086 struct ip_fw *ip_fw_default_rule ;
2091 struct ip_fw default_rule;
2093 ip_fw_chk_ptr = ip_fw_chk;
2094 ip_fw_ctl_ptr = ip_fw_ctl;
2095 LIST_INIT(&ip_fw_chain_head);
2097 bzero(&default_rule, sizeof default_rule);
2098 default_rule.fw_prot = IPPROTO_IP;
2099 default_rule.fw_number = IPFW_DEFAULT_RULE;
2100 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
2101 default_rule.fw_flg |= IP_FW_F_ACCEPT;
2103 default_rule.fw_flg |= IP_FW_F_DENY;
2105 default_rule.fw_flg |= IP_FW_F_IN | IP_FW_F_OUT;
2106 if (check_ipfw_struct(&default_rule) != 0 ||
2107 add_entry(&ip_fw_chain_head, &default_rule))
2108 panic("ip_fw_init");
2110 ip_fw_default_rule = LIST_FIRST(&ip_fw_chain_head) ;
2111 printf("IP packet filtering initialized, "
2117 "rule-based forwarding enabled, "
2118 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
2119 "default to accept, ");
2121 "default to deny, " );
2123 #ifndef IPFIREWALL_VERBOSE
2124 printf("logging disabled\n");
2126 if (fw_verbose_limit == 0)
2127 printf("unlimited logging\n");
2129 printf("logging limited to %d packets/entry by default\n",
2135 ipfw_modevent(module_t mod, int type, void *unused)
2138 #if defined(KLD_MODULE)
2147 printf("IP firewall already loaded\n");
2155 #if !defined(KLD_MODULE)
2156 printf("ipfw statically compiled, cannot unload\n");
2160 ip_fw_chk_ptr = NULL ;
2161 ip_fw_ctl_ptr = NULL ;
2162 while ( (fcp = LIST_FIRST(&ip_fw_chain_head)) != NULL)
2165 printf("IP firewall unloaded\n");
2175 static moduledata_t ipfwmod = {
2180 DECLARE_MODULE(ipfw, ipfwmod, SI_SUB_PSEUDO, SI_ORDER_ANY);