FreeBSD/Linux Kernel Cross Reference
sys/netinet/ip_fw2.c
1 /*
2 * Copyright (c) 2002 Luigi Rizzo, Universita` di Pisa
3 *
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions
6 * are met:
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.
12 *
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
23 * SUCH DAMAGE.
24 *
25 * $FreeBSD: releng/5.1/sys/netinet/ip_fw2.c 115796 2003-06-04 02:19:36Z ticso $
26 */
27
28 #define DEB(x)
29 #define DDB(x) x
30
31 /*
32 * Implement IP packet firewall (new version)
33 */
34
35 #if !defined(KLD_MODULE)
36 #include "opt_ipfw.h"
37 #include "opt_ipdn.h"
38 #include "opt_ipdivert.h"
39 #include "opt_inet.h"
40 #ifndef INET
41 #error IPFIREWALL requires INET.
42 #endif /* INET */
43 #endif
44
45 #define IPFW2 1
46 #if IPFW2
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/malloc.h>
50 #include <sys/mbuf.h>
51 #include <sys/kernel.h>
52 #include <sys/proc.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/sysctl.h>
56 #include <sys/syslog.h>
57 #include <sys/ucred.h>
58 #include <net/if.h>
59 #include <net/route.h>
60 #include <netinet/in.h>
61 #include <netinet/in_systm.h>
62 #include <netinet/in_var.h>
63 #include <netinet/in_pcb.h>
64 #include <netinet/ip.h>
65 #include <netinet/ip_var.h>
66 #include <netinet/ip_icmp.h>
67 #include <netinet/ip_fw.h>
68 #include <netinet/ip_dummynet.h>
69 #include <netinet/tcp.h>
70 #include <netinet/tcp_timer.h>
71 #include <netinet/tcp_var.h>
72 #include <netinet/tcpip.h>
73 #include <netinet/udp.h>
74 #include <netinet/udp_var.h>
75
76 #include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */
77
78 #include <machine/in_cksum.h> /* XXX for in_cksum */
79
80 /*
81 * XXX This one should go in sys/mbuf.h. It is used to avoid that
82 * a firewall-generated packet loops forever through the firewall.
83 */
84 #ifndef M_SKIP_FIREWALL
85 #define M_SKIP_FIREWALL 0x4000
86 #endif
87
88 /*
89 * set_disable contains one bit per set value (0..31).
90 * If the bit is set, all rules with the corresponding set
91 * are disabled. Set 31 is reserved for the default rule
92 * and CANNOT be disabled.
93 */
94 static u_int32_t set_disable;
95
96 static int fw_verbose;
97 static int verbose_limit;
98
99 static struct callout_handle ipfw_timeout_h;
100 #define IPFW_DEFAULT_RULE 65535
101
102 /*
103 * list of rules for layer 3
104 */
105 static struct ip_fw *layer3_chain;
106
107 MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's");
108
109 static int fw_debug = 1;
110 static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */
111
112 #ifdef SYSCTL_NODE
113 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW, 0, "Firewall");
114 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, enable,
115 CTLFLAG_RW | CTLFLAG_SECURE3,
116 &fw_enable, 0, "Enable ipfw");
117 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW,
118 &autoinc_step, 0, "Rule number autincrement step");
119 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
120 CTLFLAG_RW | CTLFLAG_SECURE3,
121 &fw_one_pass, 0,
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,
126 CTLFLAG_RW | CTLFLAG_SECURE3,
127 &fw_verbose, 0, "Log matches to ipfw rules");
128 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW,
129 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged");
130
131 /*
132 * Description of dynamic rules.
133 *
134 * Dynamic rules are stored in lists accessed through a hash table
135 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can
136 * be modified through the sysctl variable dyn_buckets which is
137 * updated when the table becomes empty.
138 *
139 * XXX currently there is only one list, ipfw_dyn.
140 *
141 * When a packet is received, its address fields are first masked
142 * with the mask defined for the rule, then hashed, then matched
143 * against the entries in the corresponding list.
144 * Dynamic rules can be used for different purposes:
145 * + stateful rules;
146 * + enforcing limits on the number of sessions;
147 * + in-kernel NAT (not implemented yet)
148 *
149 * The lifetime of dynamic rules is regulated by dyn_*_lifetime,
150 * measured in seconds and depending on the flags.
151 *
152 * The total number of dynamic rules is stored in dyn_count.
153 * The max number of dynamic rules is dyn_max. When we reach
154 * the maximum number of rules we do not create anymore. This is
155 * done to avoid consuming too much memory, but also too much
156 * time when searching on each packet (ideally, we should try instead
157 * to put a limit on the length of the list on each bucket...).
158 *
159 * Each dynamic rule holds a pointer to the parent ipfw rule so
160 * we know what action to perform. Dynamic rules are removed when
161 * the parent rule is deleted. XXX we should make them survive.
162 *
163 * There are some limitations with dynamic rules -- we do not
164 * obey the 'randomized match', and we do not do multiple
165 * passes through the firewall. XXX check the latter!!!
166 */
167 static ipfw_dyn_rule **ipfw_dyn_v = NULL;
168 static u_int32_t dyn_buckets = 256; /* must be power of 2 */
169 static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */
170
171 /*
172 * Timeouts for various events in handing dynamic rules.
173 */
174 static u_int32_t dyn_ack_lifetime = 300;
175 static u_int32_t dyn_syn_lifetime = 20;
176 static u_int32_t dyn_fin_lifetime = 1;
177 static u_int32_t dyn_rst_lifetime = 1;
178 static u_int32_t dyn_udp_lifetime = 10;
179 static u_int32_t dyn_short_lifetime = 5;
180
181 /*
182 * Keepalives are sent if dyn_keepalive is set. They are sent every
183 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval
184 * seconds of lifetime of a rule.
185 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower
186 * than dyn_keepalive_period.
187 */
188
189 static u_int32_t dyn_keepalive_interval = 20;
190 static u_int32_t dyn_keepalive_period = 5;
191 static u_int32_t dyn_keepalive = 1; /* do send keepalives */
192
193 static u_int32_t static_count; /* # of static rules */
194 static u_int32_t static_len; /* size in bytes of static rules */
195 static u_int32_t dyn_count; /* # of dynamic rules */
196 static u_int32_t dyn_max = 4096; /* max # of dynamic rules */
197
198 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW,
199 &dyn_buckets, 0, "Number of dyn. buckets");
200 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD,
201 &curr_dyn_buckets, 0, "Current Number of dyn. buckets");
202 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD,
203 &dyn_count, 0, "Number of dyn. rules");
204 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW,
205 &dyn_max, 0, "Max number of dyn. rules");
206 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD,
207 &static_count, 0, "Number of static rules");
208 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW,
209 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks");
210 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW,
211 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn");
212 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW,
213 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin");
214 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW,
215 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst");
216 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW,
217 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP");
218 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW,
219 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations");
220 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW,
221 &dyn_keepalive, 0, "Enable keepalives for dyn. rules");
222
223 #endif /* SYSCTL_NODE */
224
225
226 static ip_fw_chk_t ipfw_chk;
227
228 ip_dn_ruledel_t *ip_dn_ruledel_ptr = NULL; /* hook into dummynet */
229
230 /*
231 * This macro maps an ip pointer into a layer3 header pointer of type T
232 */
233 #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
234
235 static __inline int
236 icmptype_match(struct ip *ip, ipfw_insn_u32 *cmd)
237 {
238 int type = L3HDR(struct icmp,ip)->icmp_type;
239
240 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
241 }
242
243 #define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
244 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
245
246 static int
247 is_icmp_query(struct ip *ip)
248 {
249 int type = L3HDR(struct icmp, ip)->icmp_type;
250 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
251 }
252 #undef TT
253
254 /*
255 * The following checks use two arrays of 8 or 16 bits to store the
256 * bits that we want set or clear, respectively. They are in the
257 * low and high half of cmd->arg1 or cmd->d[0].
258 *
259 * We scan options and store the bits we find set. We succeed if
260 *
261 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
262 *
263 * The code is sometimes optimized not to store additional variables.
264 */
265
266 static int
267 flags_match(ipfw_insn *cmd, u_int8_t bits)
268 {
269 u_char want_clear;
270 bits = ~bits;
271
272 if ( ((cmd->arg1 & 0xff) & bits) != 0)
273 return 0; /* some bits we want set were clear */
274 want_clear = (cmd->arg1 >> 8) & 0xff;
275 if ( (want_clear & bits) != want_clear)
276 return 0; /* some bits we want clear were set */
277 return 1;
278 }
279
280 static int
281 ipopts_match(struct ip *ip, ipfw_insn *cmd)
282 {
283 int optlen, bits = 0;
284 u_char *cp = (u_char *)(ip + 1);
285 int x = (ip->ip_hl << 2) - sizeof (struct ip);
286
287 for (; x > 0; x -= optlen, cp += optlen) {
288 int opt = cp[IPOPT_OPTVAL];
289
290 if (opt == IPOPT_EOL)
291 break;
292 if (opt == IPOPT_NOP)
293 optlen = 1;
294 else {
295 optlen = cp[IPOPT_OLEN];
296 if (optlen <= 0 || optlen > x)
297 return 0; /* invalid or truncated */
298 }
299 switch (opt) {
300
301 default:
302 break;
303
304 case IPOPT_LSRR:
305 bits |= IP_FW_IPOPT_LSRR;
306 break;
307
308 case IPOPT_SSRR:
309 bits |= IP_FW_IPOPT_SSRR;
310 break;
311
312 case IPOPT_RR:
313 bits |= IP_FW_IPOPT_RR;
314 break;
315
316 case IPOPT_TS:
317 bits |= IP_FW_IPOPT_TS;
318 break;
319 }
320 }
321 return (flags_match(cmd, bits));
322 }
323
324 static int
325 tcpopts_match(struct ip *ip, ipfw_insn *cmd)
326 {
327 int optlen, bits = 0;
328 struct tcphdr *tcp = L3HDR(struct tcphdr,ip);
329 u_char *cp = (u_char *)(tcp + 1);
330 int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
331
332 for (; x > 0; x -= optlen, cp += optlen) {
333 int opt = cp[0];
334 if (opt == TCPOPT_EOL)
335 break;
336 if (opt == TCPOPT_NOP)
337 optlen = 1;
338 else {
339 optlen = cp[1];
340 if (optlen <= 0)
341 break;
342 }
343
344 switch (opt) {
345
346 default:
347 break;
348
349 case TCPOPT_MAXSEG:
350 bits |= IP_FW_TCPOPT_MSS;
351 break;
352
353 case TCPOPT_WINDOW:
354 bits |= IP_FW_TCPOPT_WINDOW;
355 break;
356
357 case TCPOPT_SACK_PERMITTED:
358 case TCPOPT_SACK:
359 bits |= IP_FW_TCPOPT_SACK;
360 break;
361
362 case TCPOPT_TIMESTAMP:
363 bits |= IP_FW_TCPOPT_TS;
364 break;
365
366 case TCPOPT_CC:
367 case TCPOPT_CCNEW:
368 case TCPOPT_CCECHO:
369 bits |= IP_FW_TCPOPT_CC;
370 break;
371 }
372 }
373 return (flags_match(cmd, bits));
374 }
375
376 static int
377 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd)
378 {
379 if (ifp == NULL) /* no iface with this packet, match fails */
380 return 0;
381 /* Check by name or by IP address */
382 if (cmd->name[0] != '\0') { /* match by name */
383 /* Check unit number (-1 is wildcard) */
384 if (cmd->p.unit != -1 && cmd->p.unit != ifp->if_unit)
385 return(0);
386 /* Check name */
387 if (!strncmp(ifp->if_name, cmd->name, IFNAMSIZ))
388 return(1);
389 } else {
390 struct ifaddr *ia;
391
392 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
393 if (ia->ifa_addr == NULL)
394 continue;
395 if (ia->ifa_addr->sa_family != AF_INET)
396 continue;
397 if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
398 (ia->ifa_addr))->sin_addr.s_addr)
399 return(1); /* match */
400 }
401 }
402 return(0); /* no match, fail ... */
403 }
404
405 /*
406 * The 'verrevpath' option checks that the interface that an IP packet
407 * arrives on is the same interface that traffic destined for the
408 * packet's source address would be routed out of. This is a measure
409 * to block forged packets. This is also commonly known as "anti-spoofing"
410 * or Unicast Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The
411 * name of the knob is purposely reminisent of the Cisco IOS command,
412 *
413 * ip verify unicast reverse-path
414 *
415 * which implements the same functionality. But note that syntax is
416 * misleading. The check may be performed on all IP packets whether unicast,
417 * multicast, or broadcast.
418 */
419 static int
420 verify_rev_path(struct in_addr src, struct ifnet *ifp)
421 {
422 static struct route ro;
423 struct sockaddr_in *dst;
424
425 dst = (struct sockaddr_in *)&(ro.ro_dst);
426
427 /* Check if we've cached the route from the previous call. */
428 if (src.s_addr != dst->sin_addr.s_addr) {
429 ro.ro_rt = NULL;
430
431 bzero(dst, sizeof(*dst));
432 dst->sin_family = AF_INET;
433 dst->sin_len = sizeof(*dst);
434 dst->sin_addr = src;
435
436 rtalloc_ign(&ro, RTF_CLONING|RTF_PRCLONING);
437 }
438
439 if ((ro.ro_rt == NULL) || (ifp == NULL) ||
440 (ro.ro_rt->rt_ifp->if_index != ifp->if_index))
441 return 0;
442
443 return 1;
444 }
445
446
447 static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */
448
449 #define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0
450 #define SNP(buf) buf, sizeof(buf)
451
452 /*
453 * We enter here when we have a rule with O_LOG.
454 * XXX this function alone takes about 2Kbytes of code!
455 */
456 static void
457 ipfw_log(struct ip_fw *f, u_int hlen, struct ether_header *eh,
458 struct mbuf *m, struct ifnet *oif)
459 {
460 char *action;
461 int limit_reached = 0;
462 char action2[40], proto[48], fragment[28];
463
464 fragment[0] = '\0';
465 proto[0] = '\0';
466
467 if (f == NULL) { /* bogus pkt */
468 if (verbose_limit != 0 && norule_counter >= verbose_limit)
469 return;
470 norule_counter++;
471 if (norule_counter == verbose_limit)
472 limit_reached = verbose_limit;
473 action = "Refuse";
474 } else { /* O_LOG is the first action, find the real one */
475 ipfw_insn *cmd = ACTION_PTR(f);
476 ipfw_insn_log *l = (ipfw_insn_log *)cmd;
477
478 if (l->max_log != 0 && l->log_left == 0)
479 return;
480 l->log_left--;
481 if (l->log_left == 0)
482 limit_reached = l->max_log;
483 cmd += F_LEN(cmd); /* point to first action */
484 if (cmd->opcode == O_PROB)
485 cmd += F_LEN(cmd);
486
487 action = action2;
488 switch (cmd->opcode) {
489 case O_DENY:
490 action = "Deny";
491 break;
492
493 case O_REJECT:
494 if (cmd->arg1==ICMP_REJECT_RST)
495 action = "Reset";
496 else if (cmd->arg1==ICMP_UNREACH_HOST)
497 action = "Reject";
498 else
499 snprintf(SNPARGS(action2, 0), "Unreach %d",
500 cmd->arg1);
501 break;
502
503 case O_ACCEPT:
504 action = "Accept";
505 break;
506 case O_COUNT:
507 action = "Count";
508 break;
509 case O_DIVERT:
510 snprintf(SNPARGS(action2, 0), "Divert %d",
511 cmd->arg1);
512 break;
513 case O_TEE:
514 snprintf(SNPARGS(action2, 0), "Tee %d",
515 cmd->arg1);
516 break;
517 case O_SKIPTO:
518 snprintf(SNPARGS(action2, 0), "SkipTo %d",
519 cmd->arg1);
520 break;
521 case O_PIPE:
522 snprintf(SNPARGS(action2, 0), "Pipe %d",
523 cmd->arg1);
524 break;
525 case O_QUEUE:
526 snprintf(SNPARGS(action2, 0), "Queue %d",
527 cmd->arg1);
528 break;
529 case O_FORWARD_IP: {
530 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd;
531 int len;
532
533 len = snprintf(SNPARGS(action2, 0), "Forward to %s",
534 inet_ntoa(sa->sa.sin_addr));
535 if (sa->sa.sin_port)
536 snprintf(SNPARGS(action2, len), ":%d",
537 sa->sa.sin_port);
538 }
539 break;
540 default:
541 action = "UNKNOWN";
542 break;
543 }
544 }
545
546 if (hlen == 0) { /* non-ip */
547 snprintf(SNPARGS(proto, 0), "MAC");
548 } else {
549 struct ip *ip = mtod(m, struct ip *);
550 /* these three are all aliases to the same thing */
551 struct icmp *const icmp = L3HDR(struct icmp, ip);
552 struct tcphdr *const tcp = (struct tcphdr *)icmp;
553 struct udphdr *const udp = (struct udphdr *)icmp;
554
555 int ip_off, offset, ip_len;
556
557 int len;
558
559 if (eh != NULL) { /* layer 2 packets are as on the wire */
560 ip_off = ntohs(ip->ip_off);
561 ip_len = ntohs(ip->ip_len);
562 } else {
563 ip_off = ip->ip_off;
564 ip_len = ip->ip_len;
565 }
566 offset = ip_off & IP_OFFMASK;
567 switch (ip->ip_p) {
568 case IPPROTO_TCP:
569 len = snprintf(SNPARGS(proto, 0), "TCP %s",
570 inet_ntoa(ip->ip_src));
571 if (offset == 0)
572 snprintf(SNPARGS(proto, len), ":%d %s:%d",
573 ntohs(tcp->th_sport),
574 inet_ntoa(ip->ip_dst),
575 ntohs(tcp->th_dport));
576 else
577 snprintf(SNPARGS(proto, len), " %s",
578 inet_ntoa(ip->ip_dst));
579 break;
580
581 case IPPROTO_UDP:
582 len = snprintf(SNPARGS(proto, 0), "UDP %s",
583 inet_ntoa(ip->ip_src));
584 if (offset == 0)
585 snprintf(SNPARGS(proto, len), ":%d %s:%d",
586 ntohs(udp->uh_sport),
587 inet_ntoa(ip->ip_dst),
588 ntohs(udp->uh_dport));
589 else
590 snprintf(SNPARGS(proto, len), " %s",
591 inet_ntoa(ip->ip_dst));
592 break;
593
594 case IPPROTO_ICMP:
595 if (offset == 0)
596 len = snprintf(SNPARGS(proto, 0),
597 "ICMP:%u.%u ",
598 icmp->icmp_type, icmp->icmp_code);
599 else
600 len = snprintf(SNPARGS(proto, 0), "ICMP ");
601 len += snprintf(SNPARGS(proto, len), "%s",
602 inet_ntoa(ip->ip_src));
603 snprintf(SNPARGS(proto, len), " %s",
604 inet_ntoa(ip->ip_dst));
605 break;
606
607 default:
608 len = snprintf(SNPARGS(proto, 0), "P:%d %s", ip->ip_p,
609 inet_ntoa(ip->ip_src));
610 snprintf(SNPARGS(proto, len), " %s",
611 inet_ntoa(ip->ip_dst));
612 break;
613 }
614
615 if (ip_off & (IP_MF | IP_OFFMASK))
616 snprintf(SNPARGS(fragment, 0), " (frag %d:%d@%d%s)",
617 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2),
618 offset << 3,
619 (ip_off & IP_MF) ? "+" : "");
620 }
621 if (oif || m->m_pkthdr.rcvif)
622 log(LOG_SECURITY | LOG_INFO,
623 "ipfw: %d %s %s %s via %s%d%s\n",
624 f ? f->rulenum : -1,
625 action, proto, oif ? "out" : "in",
626 oif ? oif->if_name : m->m_pkthdr.rcvif->if_name,
627 oif ? oif->if_unit : m->m_pkthdr.rcvif->if_unit,
628 fragment);
629 else
630 log(LOG_SECURITY | LOG_INFO,
631 "ipfw: %d %s %s [no if info]%s\n",
632 f ? f->rulenum : -1,
633 action, proto, fragment);
634 if (limit_reached)
635 log(LOG_SECURITY | LOG_NOTICE,
636 "ipfw: limit %d reached on entry %d\n",
637 limit_reached, f ? f->rulenum : -1);
638 }
639
640 /*
641 * IMPORTANT: the hash function for dynamic rules must be commutative
642 * in source and destination (ip,port), because rules are bidirectional
643 * and we want to find both in the same bucket.
644 */
645 static __inline int
646 hash_packet(struct ipfw_flow_id *id)
647 {
648 u_int32_t i;
649
650 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port);
651 i &= (curr_dyn_buckets - 1);
652 return i;
653 }
654
655 /**
656 * unlink a dynamic rule from a chain. prev is a pointer to
657 * the previous one, q is a pointer to the rule to delete,
658 * head is a pointer to the head of the queue.
659 * Modifies q and potentially also head.
660 */
661 #define UNLINK_DYN_RULE(prev, head, q) { \
662 ipfw_dyn_rule *old_q = q; \
663 \
664 /* remove a refcount to the parent */ \
665 if (q->dyn_type == O_LIMIT) \
666 q->parent->count--; \
667 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\
668 (q->id.src_ip), (q->id.src_port), \
669 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \
670 if (prev != NULL) \
671 prev->next = q = q->next; \
672 else \
673 head = q = q->next; \
674 dyn_count--; \
675 free(old_q, M_IPFW); }
676
677 #define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0)
678
679 /**
680 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL.
681 *
682 * If keep_me == NULL, rules are deleted even if not expired,
683 * otherwise only expired rules are removed.
684 *
685 * The value of the second parameter is also used to point to identify
686 * a rule we absolutely do not want to remove (e.g. because we are
687 * holding a reference to it -- this is the case with O_LIMIT_PARENT
688 * rules). The pointer is only used for comparison, so any non-null
689 * value will do.
690 */
691 static void
692 remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me)
693 {
694 static u_int32_t last_remove = 0;
695
696 #define FORCE (keep_me == NULL)
697
698 ipfw_dyn_rule *prev, *q;
699 int i, pass = 0, max_pass = 0;
700
701 if (ipfw_dyn_v == NULL || dyn_count == 0)
702 return;
703 /* do not expire more than once per second, it is useless */
704 if (!FORCE && last_remove == time_second)
705 return;
706 last_remove = time_second;
707
708 /*
709 * because O_LIMIT refer to parent rules, during the first pass only
710 * remove child and mark any pending LIMIT_PARENT, and remove
711 * them in a second pass.
712 */
713 next_pass:
714 for (i = 0 ; i < curr_dyn_buckets ; i++) {
715 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) {
716 /*
717 * Logic can become complex here, so we split tests.
718 */
719 if (q == keep_me)
720 goto next;
721 if (rule != NULL && rule != q->rule)
722 goto next; /* not the one we are looking for */
723 if (q->dyn_type == O_LIMIT_PARENT) {
724 /*
725 * handle parent in the second pass,
726 * record we need one.
727 */
728 max_pass = 1;
729 if (pass == 0)
730 goto next;
731 if (FORCE && q->count != 0 ) {
732 /* XXX should not happen! */
733 printf("ipfw: OUCH! cannot remove rule,"
734 " count %d\n", q->count);
735 }
736 } else {
737 if (!FORCE &&
738 !TIME_LEQ( q->expire, time_second ))
739 goto next;
740 }
741 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
742 continue;
743 next:
744 prev=q;
745 q=q->next;
746 }
747 }
748 if (pass++ < max_pass)
749 goto next_pass;
750 }
751
752
753 /**
754 * lookup a dynamic rule.
755 */
756 static ipfw_dyn_rule *
757 lookup_dyn_rule(struct ipfw_flow_id *pkt, int *match_direction,
758 struct tcphdr *tcp)
759 {
760 /*
761 * stateful ipfw extensions.
762 * Lookup into dynamic session queue
763 */
764 #define MATCH_REVERSE 0
765 #define MATCH_FORWARD 1
766 #define MATCH_NONE 2
767 #define MATCH_UNKNOWN 3
768 int i, dir = MATCH_NONE;
769 ipfw_dyn_rule *prev, *q=NULL;
770
771 if (ipfw_dyn_v == NULL)
772 goto done; /* not found */
773 i = hash_packet( pkt );
774 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) {
775 if (q->dyn_type == O_LIMIT_PARENT)
776 goto next;
777 if (TIME_LEQ( q->expire, time_second)) { /* expire entry */
778 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
779 continue;
780 }
781 if ( pkt->proto == q->id.proto) {
782 if (pkt->src_ip == q->id.src_ip &&
783 pkt->dst_ip == q->id.dst_ip &&
784 pkt->src_port == q->id.src_port &&
785 pkt->dst_port == q->id.dst_port ) {
786 dir = MATCH_FORWARD;
787 break;
788 }
789 if (pkt->src_ip == q->id.dst_ip &&
790 pkt->dst_ip == q->id.src_ip &&
791 pkt->src_port == q->id.dst_port &&
792 pkt->dst_port == q->id.src_port ) {
793 dir = MATCH_REVERSE;
794 break;
795 }
796 }
797 next:
798 prev = q;
799 q = q->next;
800 }
801 if (q == NULL)
802 goto done; /* q = NULL, not found */
803
804 if ( prev != NULL) { /* found and not in front */
805 prev->next = q->next;
806 q->next = ipfw_dyn_v[i];
807 ipfw_dyn_v[i] = q;
808 }
809 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */
810 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST);
811
812 #define BOTH_SYN (TH_SYN | (TH_SYN << 8))
813 #define BOTH_FIN (TH_FIN | (TH_FIN << 8))
814 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8);
815 switch (q->state) {
816 case TH_SYN: /* opening */
817 q->expire = time_second + dyn_syn_lifetime;
818 break;
819
820 case BOTH_SYN: /* move to established */
821 case BOTH_SYN | TH_FIN : /* one side tries to close */
822 case BOTH_SYN | (TH_FIN << 8) :
823 if (tcp) {
824 #define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0)
825 u_int32_t ack = ntohl(tcp->th_ack);
826 if (dir == MATCH_FORWARD) {
827 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd))
828 q->ack_fwd = ack;
829 else { /* ignore out-of-sequence */
830 break;
831 }
832 } else {
833 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev))
834 q->ack_rev = ack;
835 else { /* ignore out-of-sequence */
836 break;
837 }
838 }
839 }
840 q->expire = time_second + dyn_ack_lifetime;
841 break;
842
843 case BOTH_SYN | BOTH_FIN: /* both sides closed */
844 if (dyn_fin_lifetime >= dyn_keepalive_period)
845 dyn_fin_lifetime = dyn_keepalive_period - 1;
846 q->expire = time_second + dyn_fin_lifetime;
847 break;
848
849 default:
850 #if 0
851 /*
852 * reset or some invalid combination, but can also
853 * occur if we use keep-state the wrong way.
854 */
855 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0)
856 printf("invalid state: 0x%x\n", q->state);
857 #endif
858 if (dyn_rst_lifetime >= dyn_keepalive_period)
859 dyn_rst_lifetime = dyn_keepalive_period - 1;
860 q->expire = time_second + dyn_rst_lifetime;
861 break;
862 }
863 } else if (pkt->proto == IPPROTO_UDP) {
864 q->expire = time_second + dyn_udp_lifetime;
865 } else {
866 /* other protocols */
867 q->expire = time_second + dyn_short_lifetime;
868 }
869 done:
870 if (match_direction)
871 *match_direction = dir;
872 return q;
873 }
874
875 static void
876 realloc_dynamic_table(void)
877 {
878 /*
879 * Try reallocation, make sure we have a power of 2 and do
880 * not allow more than 64k entries. In case of overflow,
881 * default to 1024.
882 */
883
884 if (dyn_buckets > 65536)
885 dyn_buckets = 1024;
886 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */
887 dyn_buckets = curr_dyn_buckets; /* reset */
888 return;
889 }
890 curr_dyn_buckets = dyn_buckets;
891 if (ipfw_dyn_v != NULL)
892 free(ipfw_dyn_v, M_IPFW);
893 for (;;) {
894 ipfw_dyn_v = malloc(curr_dyn_buckets * sizeof(ipfw_dyn_rule *),
895 M_IPFW, M_NOWAIT | M_ZERO);
896 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2)
897 break;
898 curr_dyn_buckets /= 2;
899 }
900 }
901
902 /**
903 * Install state of type 'type' for a dynamic session.
904 * The hash table contains two type of rules:
905 * - regular rules (O_KEEP_STATE)
906 * - rules for sessions with limited number of sess per user
907 * (O_LIMIT). When they are created, the parent is
908 * increased by 1, and decreased on delete. In this case,
909 * the third parameter is the parent rule and not the chain.
910 * - "parent" rules for the above (O_LIMIT_PARENT).
911 */
912 static ipfw_dyn_rule *
913 add_dyn_rule(struct ipfw_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule)
914 {
915 ipfw_dyn_rule *r;
916 int i;
917
918 if (ipfw_dyn_v == NULL ||
919 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) {
920 realloc_dynamic_table();
921 if (ipfw_dyn_v == NULL)
922 return NULL; /* failed ! */
923 }
924 i = hash_packet(id);
925
926 r = malloc(sizeof *r, M_IPFW, M_NOWAIT | M_ZERO);
927 if (r == NULL) {
928 printf ("ipfw: sorry cannot allocate state\n");
929 return NULL;
930 }
931
932 /* increase refcount on parent, and set pointer */
933 if (dyn_type == O_LIMIT) {
934 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule;
935 if ( parent->dyn_type != O_LIMIT_PARENT)
936 panic("invalid parent");
937 parent->count++;
938 r->parent = parent;
939 rule = parent->rule;
940 }
941
942 r->id = *id;
943 r->expire = time_second + dyn_syn_lifetime;
944 r->rule = rule;
945 r->dyn_type = dyn_type;
946 r->pcnt = r->bcnt = 0;
947 r->count = 0;
948
949 r->bucket = i;
950 r->next = ipfw_dyn_v[i];
951 ipfw_dyn_v[i] = r;
952 dyn_count++;
953 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n",
954 dyn_type,
955 (r->id.src_ip), (r->id.src_port),
956 (r->id.dst_ip), (r->id.dst_port),
957 dyn_count ); )
958 return r;
959 }
960
961 /**
962 * lookup dynamic parent rule using pkt and rule as search keys.
963 * If the lookup fails, then install one.
964 */
965 static ipfw_dyn_rule *
966 lookup_dyn_parent(struct ipfw_flow_id *pkt, struct ip_fw *rule)
967 {
968 ipfw_dyn_rule *q;
969 int i;
970
971 if (ipfw_dyn_v) {
972 i = hash_packet( pkt );
973 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next)
974 if (q->dyn_type == O_LIMIT_PARENT &&
975 rule== q->rule &&
976 pkt->proto == q->id.proto &&
977 pkt->src_ip == q->id.src_ip &&
978 pkt->dst_ip == q->id.dst_ip &&
979 pkt->src_port == q->id.src_port &&
980 pkt->dst_port == q->id.dst_port) {
981 q->expire = time_second + dyn_short_lifetime;
982 DEB(printf("ipfw: lookup_dyn_parent found 0x%p\n",q);)
983 return q;
984 }
985 }
986 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule);
987 }
988
989 /**
990 * Install dynamic state for rule type cmd->o.opcode
991 *
992 * Returns 1 (failure) if state is not installed because of errors or because
993 * session limitations are enforced.
994 */
995 static int
996 install_state(struct ip_fw *rule, ipfw_insn_limit *cmd,
997 struct ip_fw_args *args)
998 {
999 static int last_log;
1000
1001 ipfw_dyn_rule *q;
1002
1003 DEB(printf("ipfw: install state type %d 0x%08x %u -> 0x%08x %u\n",
1004 cmd->o.opcode,
1005 (args->f_id.src_ip), (args->f_id.src_port),
1006 (args->f_id.dst_ip), (args->f_id.dst_port) );)
1007
1008 q = lookup_dyn_rule(&args->f_id, NULL, NULL);
1009
1010 if (q != NULL) { /* should never occur */
1011 if (last_log != time_second) {
1012 last_log = time_second;
1013 printf("ipfw: install_state: entry already present, done\n");
1014 }
1015 return 0;
1016 }
1017
1018 if (dyn_count >= dyn_max)
1019 /*
1020 * Run out of slots, try to remove any expired rule.
1021 */
1022 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1);
1023
1024 if (dyn_count >= dyn_max) {
1025 if (last_log != time_second) {
1026 last_log = time_second;
1027 printf("ipfw: install_state: Too many dynamic rules\n");
1028 }
1029 return 1; /* cannot install, notify caller */
1030 }
1031
1032 switch (cmd->o.opcode) {
1033 case O_KEEP_STATE: /* bidir rule */
1034 add_dyn_rule(&args->f_id, O_KEEP_STATE, rule);
1035 break;
1036
1037 case O_LIMIT: /* limit number of sessions */
1038 {
1039 u_int16_t limit_mask = cmd->limit_mask;
1040 struct ipfw_flow_id id;
1041 ipfw_dyn_rule *parent;
1042
1043 DEB(printf("ipfw: installing dyn-limit rule %d\n",
1044 cmd->conn_limit);)
1045
1046 id.dst_ip = id.src_ip = 0;
1047 id.dst_port = id.src_port = 0;
1048 id.proto = args->f_id.proto;
1049
1050 if (limit_mask & DYN_SRC_ADDR)
1051 id.src_ip = args->f_id.src_ip;
1052 if (limit_mask & DYN_DST_ADDR)
1053 id.dst_ip = args->f_id.dst_ip;
1054 if (limit_mask & DYN_SRC_PORT)
1055 id.src_port = args->f_id.src_port;
1056 if (limit_mask & DYN_DST_PORT)
1057 id.dst_port = args->f_id.dst_port;
1058 parent = lookup_dyn_parent(&id, rule);
1059 if (parent == NULL) {
1060 printf("ipfw: add parent failed\n");
1061 return 1;
1062 }
1063 if (parent->count >= cmd->conn_limit) {
1064 /*
1065 * See if we can remove some expired rule.
1066 */
1067 remove_dyn_rule(rule, parent);
1068 if (parent->count >= cmd->conn_limit) {
1069 if (fw_verbose && last_log != time_second) {
1070 last_log = time_second;
1071 log(LOG_SECURITY | LOG_DEBUG,
1072 "drop session, too many entries\n");
1073 }
1074 return 1;
1075 }
1076 }
1077 add_dyn_rule(&args->f_id, O_LIMIT, (struct ip_fw *)parent);
1078 }
1079 break;
1080 default:
1081 printf("ipfw: unknown dynamic rule type %u\n", cmd->o.opcode);
1082 return 1;
1083 }
1084 lookup_dyn_rule(&args->f_id, NULL, NULL); /* XXX just set lifetime */
1085 return 0;
1086 }
1087
1088 /*
1089 * Transmit a TCP packet, containing either a RST or a keepalive.
1090 * When flags & TH_RST, we are sending a RST packet, because of a
1091 * "reset" action matched the packet.
1092 * Otherwise we are sending a keepalive, and flags & TH_
1093 */
1094 static void
1095 send_pkt(struct ipfw_flow_id *id, u_int32_t seq, u_int32_t ack, int flags)
1096 {
1097 struct mbuf *m;
1098 struct ip *ip;
1099 struct tcphdr *tcp;
1100 struct route sro; /* fake route */
1101
1102 MGETHDR(m, M_DONTWAIT, MT_HEADER);
1103 if (m == 0)
1104 return;
1105 m->m_pkthdr.rcvif = (struct ifnet *)0;
1106 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr);
1107 m->m_data += max_linkhdr;
1108
1109 ip = mtod(m, struct ip *);
1110 bzero(ip, m->m_len);
1111 tcp = (struct tcphdr *)(ip + 1); /* no IP options */
1112 ip->ip_p = IPPROTO_TCP;
1113 tcp->th_off = 5;
1114 /*
1115 * Assume we are sending a RST (or a keepalive in the reverse
1116 * direction), swap src and destination addresses and ports.
1117 */
1118 ip->ip_src.s_addr = htonl(id->dst_ip);
1119 ip->ip_dst.s_addr = htonl(id->src_ip);
1120 tcp->th_sport = htons(id->dst_port);
1121 tcp->th_dport = htons(id->src_port);
1122 if (flags & TH_RST) { /* we are sending a RST */
1123 if (flags & TH_ACK) {
1124 tcp->th_seq = htonl(ack);
1125 tcp->th_ack = htonl(0);
1126 tcp->th_flags = TH_RST;
1127 } else {
1128 if (flags & TH_SYN)
1129 seq++;
1130 tcp->th_seq = htonl(0);
1131 tcp->th_ack = htonl(seq);
1132 tcp->th_flags = TH_RST | TH_ACK;
1133 }
1134 } else {
1135 /*
1136 * We are sending a keepalive. flags & TH_SYN determines
1137 * the direction, forward if set, reverse if clear.
1138 * NOTE: seq and ack are always assumed to be correct
1139 * as set by the caller. This may be confusing...
1140 */
1141 if (flags & TH_SYN) {
1142 /*
1143 * we have to rewrite the correct addresses!
1144 */
1145 ip->ip_dst.s_addr = htonl(id->dst_ip);
1146 ip->ip_src.s_addr = htonl(id->src_ip);
1147 tcp->th_dport = htons(id->dst_port);
1148 tcp->th_sport = htons(id->src_port);
1149 }
1150 tcp->th_seq = htonl(seq);
1151 tcp->th_ack = htonl(ack);
1152 tcp->th_flags = TH_ACK;
1153 }
1154 /*
1155 * set ip_len to the payload size so we can compute
1156 * the tcp checksum on the pseudoheader
1157 * XXX check this, could save a couple of words ?
1158 */
1159 ip->ip_len = htons(sizeof(struct tcphdr));
1160 tcp->th_sum = in_cksum(m, m->m_pkthdr.len);
1161 /*
1162 * now fill fields left out earlier
1163 */
1164 ip->ip_ttl = ip_defttl;
1165 ip->ip_len = m->m_pkthdr.len;
1166 bzero (&sro, sizeof (sro));
1167 ip_rtaddr(ip->ip_dst, &sro);
1168 m->m_flags |= M_SKIP_FIREWALL;
1169 ip_output(m, NULL, &sro, 0, NULL, NULL);
1170 if (sro.ro_rt)
1171 RTFREE(sro.ro_rt);
1172 }
1173
1174 /*
1175 * sends a reject message, consuming the mbuf passed as an argument.
1176 */
1177 static void
1178 send_reject(struct ip_fw_args *args, int code, int offset, int ip_len)
1179 {
1180
1181 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
1182 /* We need the IP header in host order for icmp_error(). */
1183 if (args->eh != NULL) {
1184 struct ip *ip = mtod(args->m, struct ip *);
1185 ip->ip_len = ntohs(ip->ip_len);
1186 ip->ip_off = ntohs(ip->ip_off);
1187 }
1188 icmp_error(args->m, ICMP_UNREACH, code, 0L, 0);
1189 } else if (offset == 0 && args->f_id.proto == IPPROTO_TCP) {
1190 struct tcphdr *const tcp =
1191 L3HDR(struct tcphdr, mtod(args->m, struct ip *));
1192 if ( (tcp->th_flags & TH_RST) == 0)
1193 send_pkt(&(args->f_id), ntohl(tcp->th_seq),
1194 ntohl(tcp->th_ack),
1195 tcp->th_flags | TH_RST);
1196 m_freem(args->m);
1197 } else
1198 m_freem(args->m);
1199 args->m = NULL;
1200 }
1201
1202 /**
1203 *
1204 * Given an ip_fw *, lookup_next_rule will return a pointer
1205 * to the next rule, which can be either the jump
1206 * target (for skipto instructions) or the next one in the list (in
1207 * all other cases including a missing jump target).
1208 * The result is also written in the "next_rule" field of the rule.
1209 * Backward jumps are not allowed, so start looking from the next
1210 * rule...
1211 *
1212 * This never returns NULL -- in case we do not have an exact match,
1213 * the next rule is returned. When the ruleset is changed,
1214 * pointers are flushed so we are always correct.
1215 */
1216
1217 static struct ip_fw *
1218 lookup_next_rule(struct ip_fw *me)
1219 {
1220 struct ip_fw *rule = NULL;
1221 ipfw_insn *cmd;
1222
1223 /* look for action, in case it is a skipto */
1224 cmd = ACTION_PTR(me);
1225 if (cmd->opcode == O_LOG)
1226 cmd += F_LEN(cmd);
1227 if ( cmd->opcode == O_SKIPTO )
1228 for (rule = me->next; rule ; rule = rule->next)
1229 if (rule->rulenum >= cmd->arg1)
1230 break;
1231 if (rule == NULL) /* failure or not a skipto */
1232 rule = me->next;
1233 me->next_rule = rule;
1234 return rule;
1235 }
1236
1237 /*
1238 * The main check routine for the firewall.
1239 *
1240 * All arguments are in args so we can modify them and return them
1241 * back to the caller.
1242 *
1243 * Parameters:
1244 *
1245 * args->m (in/out) The packet; we set to NULL when/if we nuke it.
1246 * Starts with the IP header.
1247 * args->eh (in) Mac header if present, or NULL for layer3 packet.
1248 * args->oif Outgoing interface, or NULL if packet is incoming.
1249 * The incoming interface is in the mbuf. (in)
1250 * args->divert_rule (in/out)
1251 * Skip up to the first rule past this rule number;
1252 * upon return, non-zero port number for divert or tee.
1253 *
1254 * args->rule Pointer to the last matching rule (in/out)
1255 * args->next_hop Socket we are forwarding to (out).
1256 * args->f_id Addresses grabbed from the packet (out)
1257 *
1258 * Return value:
1259 *
1260 * IP_FW_PORT_DENY_FLAG the packet must be dropped.
1261 * 0 The packet is to be accepted and routed normally OR
1262 * the packet was denied/rejected and has been dropped;
1263 * in the latter case, *m is equal to NULL upon return.
1264 * port Divert the packet to port, with these caveats:
1265 *
1266 * - If IP_FW_PORT_TEE_FLAG is set, tee the packet instead
1267 * of diverting it (ie, 'ipfw tee').
1268 *
1269 * - If IP_FW_PORT_DYNT_FLAG is set, interpret the lower
1270 * 16 bits as a dummynet pipe number instead of diverting
1271 */
1272
1273 static int
1274 ipfw_chk(struct ip_fw_args *args)
1275 {
1276 /*
1277 * Local variables hold state during the processing of a packet.
1278 *
1279 * IMPORTANT NOTE: to speed up the processing of rules, there
1280 * are some assumption on the values of the variables, which
1281 * are documented here. Should you change them, please check
1282 * the implementation of the various instructions to make sure
1283 * that they still work.
1284 *
1285 * args->eh The MAC header. It is non-null for a layer2
1286 * packet, it is NULL for a layer-3 packet.
1287 *
1288 * m | args->m Pointer to the mbuf, as received from the caller.
1289 * It may change if ipfw_chk() does an m_pullup, or if it
1290 * consumes the packet because it calls send_reject().
1291 * XXX This has to change, so that ipfw_chk() never modifies
1292 * or consumes the buffer.
1293 * ip is simply an alias of the value of m, and it is kept
1294 * in sync with it (the packet is supposed to start with
1295 * the ip header).
1296 */
1297 struct mbuf *m = args->m;
1298 struct ip *ip = mtod(m, struct ip *);
1299
1300 /*
1301 * oif | args->oif If NULL, ipfw_chk has been called on the
1302 * inbound path (ether_input, bdg_forward, ip_input).
1303 * If non-NULL, ipfw_chk has been called on the outbound path
1304 * (ether_output, ip_output).
1305 */
1306 struct ifnet *oif = args->oif;
1307
1308 struct ip_fw *f = NULL; /* matching rule */
1309 int retval = 0;
1310
1311 /*
1312 * hlen The length of the IPv4 header.
1313 * hlen >0 means we have an IPv4 packet.
1314 */
1315 u_int hlen = 0; /* hlen >0 means we have an IP pkt */
1316
1317 /*
1318 * offset The offset of a fragment. offset != 0 means that
1319 * we have a fragment at this offset of an IPv4 packet.
1320 * offset == 0 means that (if this is an IPv4 packet)
1321 * this is the first or only fragment.
1322 */
1323 u_short offset = 0;
1324
1325 /*
1326 * Local copies of addresses. They are only valid if we have
1327 * an IP packet.
1328 *
1329 * proto The protocol. Set to 0 for non-ip packets,
1330 * or to the protocol read from the packet otherwise.
1331 * proto != 0 means that we have an IPv4 packet.
1332 *
1333 * src_port, dst_port port numbers, in HOST format. Only
1334 * valid for TCP and UDP packets.
1335 *
1336 * src_ip, dst_ip ip addresses, in NETWORK format.
1337 * Only valid for IPv4 packets.
1338 */
1339 u_int8_t proto;
1340 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */
1341 struct in_addr src_ip, dst_ip; /* NOTE: network format */
1342 u_int16_t ip_len=0;
1343 int dyn_dir = MATCH_UNKNOWN;
1344 ipfw_dyn_rule *q = NULL;
1345
1346 if (m->m_flags & M_SKIP_FIREWALL)
1347 return 0; /* accept */
1348 /*
1349 * dyn_dir = MATCH_UNKNOWN when rules unchecked,
1350 * MATCH_NONE when checked and not matched (q = NULL),
1351 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
1352 */
1353
1354 if (args->eh == NULL || /* layer 3 packet */
1355 ( m->m_pkthdr.len >= sizeof(struct ip) &&
1356 ntohs(args->eh->ether_type) == ETHERTYPE_IP))
1357 hlen = ip->ip_hl << 2;
1358
1359 /*
1360 * Collect parameters into local variables for faster matching.
1361 */
1362 if (hlen == 0) { /* do not grab addresses for non-ip pkts */
1363 proto = args->f_id.proto = 0; /* mark f_id invalid */
1364 goto after_ip_checks;
1365 }
1366
1367 proto = args->f_id.proto = ip->ip_p;
1368 src_ip = ip->ip_src;
1369 dst_ip = ip->ip_dst;
1370 if (args->eh != NULL) { /* layer 2 packets are as on the wire */
1371 offset = ntohs(ip->ip_off) & IP_OFFMASK;
1372 ip_len = ntohs(ip->ip_len);
1373 } else {
1374 offset = ip->ip_off & IP_OFFMASK;
1375 ip_len = ip->ip_len;
1376 }
1377
1378 #define PULLUP_TO(len) \
1379 do { \
1380 if ((m)->m_len < (len)) { \
1381 args->m = m = m_pullup(m, (len)); \
1382 if (m == 0) \
1383 goto pullup_failed; \
1384 ip = mtod(m, struct ip *); \
1385 } \
1386 } while (0)
1387
1388 if (offset == 0) {
1389 switch (proto) {
1390 case IPPROTO_TCP:
1391 {
1392 struct tcphdr *tcp;
1393
1394 PULLUP_TO(hlen + sizeof(struct tcphdr));
1395 tcp = L3HDR(struct tcphdr, ip);
1396 dst_port = tcp->th_dport;
1397 src_port = tcp->th_sport;
1398 args->f_id.flags = tcp->th_flags;
1399 }
1400 break;
1401
1402 case IPPROTO_UDP:
1403 {
1404 struct udphdr *udp;
1405
1406 PULLUP_TO(hlen + sizeof(struct udphdr));
1407 udp = L3HDR(struct udphdr, ip);
1408 dst_port = udp->uh_dport;
1409 src_port = udp->uh_sport;
1410 }
1411 break;
1412
1413 case IPPROTO_ICMP:
1414 PULLUP_TO(hlen + 4); /* type, code and checksum. */
1415 args->f_id.flags = L3HDR(struct icmp, ip)->icmp_type;
1416 break;
1417
1418 default:
1419 break;
1420 }
1421 #undef PULLUP_TO
1422 }
1423
1424 args->f_id.src_ip = ntohl(src_ip.s_addr);
1425 args->f_id.dst_ip = ntohl(dst_ip.s_addr);
1426 args->f_id.src_port = src_port = ntohs(src_port);
1427 args->f_id.dst_port = dst_port = ntohs(dst_port);
1428
1429 after_ip_checks:
1430 if (args->rule) {
1431 /*
1432 * Packet has already been tagged. Look for the next rule
1433 * to restart processing.
1434 *
1435 * If fw_one_pass != 0 then just accept it.
1436 * XXX should not happen here, but optimized out in
1437 * the caller.
1438 */
1439 if (fw_one_pass)
1440 return 0;
1441
1442 f = args->rule->next_rule;
1443 if (f == NULL)
1444 f = lookup_next_rule(args->rule);
1445 } else {
1446 /*
1447 * Find the starting rule. It can be either the first
1448 * one, or the one after divert_rule if asked so.
1449 */
1450 int skipto = args->divert_rule;
1451
1452 f = layer3_chain;
1453 if (args->eh == NULL && skipto != 0) {
1454 if (skipto >= IPFW_DEFAULT_RULE)
1455 return(IP_FW_PORT_DENY_FLAG); /* invalid */
1456 while (f && f->rulenum <= skipto)
1457 f = f->next;
1458 if (f == NULL) /* drop packet */
1459 return(IP_FW_PORT_DENY_FLAG);
1460 }
1461 }
1462 args->divert_rule = 0; /* reset to avoid confusion later */
1463
1464 /*
1465 * Now scan the rules, and parse microinstructions for each rule.
1466 */
1467 for (; f; f = f->next) {
1468 int l, cmdlen;
1469 ipfw_insn *cmd;
1470 int skip_or; /* skip rest of OR block */
1471
1472 again:
1473 if (set_disable & (1 << f->set) )
1474 continue;
1475
1476 skip_or = 0;
1477 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
1478 l -= cmdlen, cmd += cmdlen) {
1479 int match;
1480
1481 /*
1482 * check_body is a jump target used when we find a
1483 * CHECK_STATE, and need to jump to the body of
1484 * the target rule.
1485 */
1486
1487 check_body:
1488 cmdlen = F_LEN(cmd);
1489 /*
1490 * An OR block (insn_1 || .. || insn_n) has the
1491 * F_OR bit set in all but the last instruction.
1492 * The first match will set "skip_or", and cause
1493 * the following instructions to be skipped until
1494 * past the one with the F_OR bit clear.
1495 */
1496 if (skip_or) { /* skip this instruction */
1497 if ((cmd->len & F_OR) == 0)
1498 skip_or = 0; /* next one is good */
1499 continue;
1500 }
1501 match = 0; /* set to 1 if we succeed */
1502
1503 switch (cmd->opcode) {
1504 /*
1505 * The first set of opcodes compares the packet's
1506 * fields with some pattern, setting 'match' if a
1507 * match is found. At the end of the loop there is
1508 * logic to deal with F_NOT and F_OR flags associated
1509 * with the opcode.
1510 */
1511 case O_NOP:
1512 match = 1;
1513 break;
1514
1515 case O_FORWARD_MAC:
1516 printf("ipfw: opcode %d unimplemented\n",
1517 cmd->opcode);
1518 break;
1519
1520 case O_GID:
1521 case O_UID:
1522 /*
1523 * We only check offset == 0 && proto != 0,
1524 * as this ensures that we have an IPv4
1525 * packet with the ports info.
1526 */
1527 if (offset!=0)
1528 break;
1529 {
1530 struct inpcbinfo *pi;
1531 int wildcard;
1532 struct inpcb *pcb;
1533
1534 if (proto == IPPROTO_TCP) {
1535 wildcard = 0;
1536 pi = &tcbinfo;
1537 } else if (proto == IPPROTO_UDP) {
1538 wildcard = 1;
1539 pi = &udbinfo;
1540 } else
1541 break;
1542
1543 pcb = (oif) ?
1544 in_pcblookup_hash(pi,
1545 dst_ip, htons(dst_port),
1546 src_ip, htons(src_port),
1547 wildcard, oif) :
1548 in_pcblookup_hash(pi,
1549 src_ip, htons(src_port),
1550 dst_ip, htons(dst_port),
1551 wildcard, NULL);
1552
1553 if (pcb == NULL || pcb->inp_socket == NULL)
1554 break;
1555 #if __FreeBSD_version < 500034
1556 #define socheckuid(a,b) ((a)->so_cred->cr_uid != (b))
1557 #endif
1558 if (cmd->opcode == O_UID) {
1559 match =
1560 !socheckuid(pcb->inp_socket,
1561 (uid_t)((ipfw_insn_u32 *)cmd)->d[0]);
1562 } else {
1563 match = groupmember(
1564 (uid_t)((ipfw_insn_u32 *)cmd)->d[0],
1565 pcb->inp_socket->so_cred);
1566 }
1567 }
1568 break;
1569
1570 case O_RECV:
1571 match = iface_match(m->m_pkthdr.rcvif,
1572 (ipfw_insn_if *)cmd);
1573 break;
1574
1575 case O_XMIT:
1576 match = iface_match(oif, (ipfw_insn_if *)cmd);
1577 break;
1578
1579 case O_VIA:
1580 match = iface_match(oif ? oif :
1581 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd);
1582 break;
1583
1584 case O_MACADDR2:
1585 if (args->eh != NULL) { /* have MAC header */
1586 u_int32_t *want = (u_int32_t *)
1587 ((ipfw_insn_mac *)cmd)->addr;
1588 u_int32_t *mask = (u_int32_t *)
1589 ((ipfw_insn_mac *)cmd)->mask;
1590 u_int32_t *hdr = (u_int32_t *)args->eh;
1591
1592 match =
1593 ( want[0] == (hdr[0] & mask[0]) &&
1594 want[1] == (hdr[1] & mask[1]) &&
1595 want[2] == (hdr[2] & mask[2]) );
1596 }
1597 break;
1598
1599 case O_MAC_TYPE:
1600 if (args->eh != NULL) {
1601 u_int16_t t =
1602 ntohs(args->eh->ether_type);
1603 u_int16_t *p =
1604 ((ipfw_insn_u16 *)cmd)->ports;
1605 int i;
1606
1607 for (i = cmdlen - 1; !match && i>0;
1608 i--, p += 2)
1609 match = (t>=p[0] && t<=p[1]);
1610 }
1611 break;
1612
1613 case O_FRAG:
1614 match = (hlen > 0 && offset != 0);
1615 break;
1616
1617 case O_IN: /* "out" is "not in" */
1618 match = (oif == NULL);
1619 break;
1620
1621 case O_LAYER2:
1622 match = (args->eh != NULL);
1623 break;
1624
1625 case O_PROTO:
1626 /*
1627 * We do not allow an arg of 0 so the
1628 * check of "proto" only suffices.
1629 */
1630 match = (proto == cmd->arg1);
1631 break;
1632
1633 case O_IP_SRC:
1634 match = (hlen > 0 &&
1635 ((ipfw_insn_ip *)cmd)->addr.s_addr ==
1636 src_ip.s_addr);
1637 break;
1638
1639 case O_IP_SRC_MASK:
1640 match = (hlen > 0 &&
1641 ((ipfw_insn_ip *)cmd)->addr.s_addr ==
1642 (src_ip.s_addr &
1643 ((ipfw_insn_ip *)cmd)->mask.s_addr));
1644 break;
1645
1646 case O_IP_SRC_ME:
1647 if (hlen > 0) {
1648 struct ifnet *tif;
1649
1650 INADDR_TO_IFP(src_ip, tif);
1651 match = (tif != NULL);
1652 }
1653 break;
1654
1655 case O_IP_DST_SET:
1656 case O_IP_SRC_SET:
1657 if (hlen > 0) {
1658 u_int32_t *d = (u_int32_t *)(cmd+1);
1659 u_int32_t addr =
1660 cmd->opcode == O_IP_DST_SET ?
1661 args->f_id.dst_ip :
1662 args->f_id.src_ip;
1663
1664 if (addr < d[0])
1665 break;
1666 addr -= d[0]; /* subtract base */
1667 match = (addr < cmd->arg1) &&
1668 ( d[ 1 + (addr>>5)] &
1669 (1<<(addr & 0x1f)) );
1670 }
1671 break;
1672
1673 case O_IP_DST:
1674 match = (hlen > 0 &&
1675 ((ipfw_insn_ip *)cmd)->addr.s_addr ==
1676 dst_ip.s_addr);
1677 break;
1678
1679 case O_IP_DST_MASK:
1680 match = (hlen > 0) &&
1681 (((ipfw_insn_ip *)cmd)->addr.s_addr ==
1682 (dst_ip.s_addr &
1683 ((ipfw_insn_ip *)cmd)->mask.s_addr));
1684 break;
1685
1686 case O_IP_DST_ME:
1687 if (hlen > 0) {
1688 struct ifnet *tif;
1689
1690 INADDR_TO_IFP(dst_ip, tif);
1691 match = (tif != NULL);
1692 }
1693 break;
1694
1695 case O_IP_SRCPORT:
1696 case O_IP_DSTPORT:
1697 /*
1698 * offset == 0 && proto != 0 is enough
1699 * to guarantee that we have an IPv4
1700 * packet with port info.
1701 */
1702 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
1703 && offset == 0) {
1704 u_int16_t x =
1705 (cmd->opcode == O_IP_SRCPORT) ?
1706 src_port : dst_port ;
1707 u_int16_t *p =
1708 ((ipfw_insn_u16 *)cmd)->ports;
1709 int i;
1710
1711 for (i = cmdlen - 1; !match && i>0;
1712 i--, p += 2)
1713 match = (x>=p[0] && x<=p[1]);
1714 }
1715 break;
1716
1717 case O_ICMPTYPE:
1718 match = (offset == 0 && proto==IPPROTO_ICMP &&
1719 icmptype_match(ip, (ipfw_insn_u32 *)cmd) );
1720 break;
1721
1722 case O_IPOPT:
1723 match = (hlen > 0 && ipopts_match(ip, cmd) );
1724 break;
1725
1726 case O_IPVER:
1727 match = (hlen > 0 && cmd->arg1 == ip->ip_v);
1728 break;
1729
1730 case O_IPTTL:
1731 match = (hlen > 0 && cmd->arg1 == ip->ip_ttl);
1732 break;
1733
1734 case O_IPID:
1735 match = (hlen > 0 &&
1736 cmd->arg1 == ntohs(ip->ip_id));
1737 break;
1738
1739 case O_IPLEN:
1740 match = (hlen > 0 && cmd->arg1 == ip_len);
1741 break;
1742
1743 case O_IPPRECEDENCE:
1744 match = (hlen > 0 &&
1745 (cmd->arg1 == (ip->ip_tos & 0xe0)) );
1746 break;
1747
1748 case O_IPTOS:
1749 match = (hlen > 0 &&
1750 flags_match(cmd, ip->ip_tos));
1751 break;
1752
1753 case O_TCPFLAGS:
1754 match = (proto == IPPROTO_TCP && offset == 0 &&
1755 flags_match(cmd,
1756 L3HDR(struct tcphdr,ip)->th_flags));
1757 break;
1758
1759 case O_TCPOPTS:
1760 match = (proto == IPPROTO_TCP && offset == 0 &&
1761 tcpopts_match(ip, cmd));
1762 break;
1763
1764 case O_TCPSEQ:
1765 match = (proto == IPPROTO_TCP && offset == 0 &&
1766 ((ipfw_insn_u32 *)cmd)->d[0] ==
1767 L3HDR(struct tcphdr,ip)->th_seq);
1768 break;
1769
1770 case O_TCPACK:
1771 match = (proto == IPPROTO_TCP && offset == 0 &&
1772 ((ipfw_insn_u32 *)cmd)->d[0] ==
1773 L3HDR(struct tcphdr,ip)->th_ack);
1774 break;
1775
1776 case O_TCPWIN:
1777 match = (proto == IPPROTO_TCP && offset == 0 &&
1778 cmd->arg1 ==
1779 L3HDR(struct tcphdr,ip)->th_win);
1780 break;
1781
1782 case O_ESTAB:
1783 /* reject packets which have SYN only */
1784 /* XXX should i also check for TH_ACK ? */
1785 match = (proto == IPPROTO_TCP && offset == 0 &&
1786 (L3HDR(struct tcphdr,ip)->th_flags &
1787 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
1788 break;
1789
1790 case O_LOG:
1791 if (fw_verbose)
1792 ipfw_log(f, hlen, args->eh, m, oif);
1793 match = 1;
1794 break;
1795
1796 case O_PROB:
1797 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
1798 break;
1799
1800 case O_VERREVPATH:
1801 /* Outgoing packets automatically pass/match */
1802 match = ((oif != NULL) ||
1803 (m->m_pkthdr.rcvif == NULL) ||
1804 verify_rev_path(src_ip, m->m_pkthdr.rcvif));
1805 break;
1806
1807 /*
1808 * The second set of opcodes represents 'actions',
1809 * i.e. the terminal part of a rule once the packet
1810 * matches all previous patterns.
1811 * Typically there is only one action for each rule,
1812 * and the opcode is stored at the end of the rule
1813 * (but there are exceptions -- see below).
1814 *
1815 * In general, here we set retval and terminate the
1816 * outer loop (would be a 'break 3' in some language,
1817 * but we need to do a 'goto done').
1818 *
1819 * Exceptions:
1820 * O_COUNT and O_SKIPTO actions:
1821 * instead of terminating, we jump to the next rule
1822 * ('goto next_rule', equivalent to a 'break 2'),
1823 * or to the SKIPTO target ('goto again' after
1824 * having set f, cmd and l), respectively.
1825 *
1826 * O_LIMIT and O_KEEP_STATE: these opcodes are
1827 * not real 'actions', and are stored right
1828 * before the 'action' part of the rule.
1829 * These opcodes try to install an entry in the
1830 * state tables; if successful, we continue with
1831 * the next opcode (match=1; break;), otherwise
1832 * the packet * must be dropped
1833 * ('goto done' after setting retval);
1834 *
1835 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
1836 * cause a lookup of the state table, and a jump
1837 * to the 'action' part of the parent rule
1838 * ('goto check_body') if an entry is found, or
1839 * (CHECK_STATE only) a jump to the next rule if
1840 * the entry is not found ('goto next_rule').
1841 * The result of the lookup is cached to make
1842 * further instances of these opcodes are
1843 * effectively NOPs.
1844 */
1845 case O_LIMIT:
1846 case O_KEEP_STATE:
1847 if (install_state(f,
1848 (ipfw_insn_limit *)cmd, args)) {
1849 retval = IP_FW_PORT_DENY_FLAG;
1850 goto done; /* error/limit violation */
1851 }
1852 match = 1;
1853 break;
1854
1855 case O_PROBE_STATE:
1856 case O_CHECK_STATE:
1857 /*
1858 * dynamic rules are checked at the first
1859 * keep-state or check-state occurrence,
1860 * with the result being stored in dyn_dir.
1861 * The compiler introduces a PROBE_STATE
1862 * instruction for us when we have a
1863 * KEEP_STATE (because PROBE_STATE needs
1864 * to be run first).
1865 */
1866 if (dyn_dir == MATCH_UNKNOWN &&
1867 (q = lookup_dyn_rule(&args->f_id,
1868 &dyn_dir, proto == IPPROTO_TCP ?
1869 L3HDR(struct tcphdr, ip) : NULL))
1870 != NULL) {
1871 /*
1872 * Found dynamic entry, update stats
1873 * and jump to the 'action' part of
1874 * the parent rule.
1875 */
1876 q->pcnt++;
1877 q->bcnt += ip_len;
1878 f = q->rule;
1879 cmd = ACTION_PTR(f);
1880 l = f->cmd_len - f->act_ofs;
1881 goto check_body;
1882 }
1883 /*
1884 * Dynamic entry not found. If CHECK_STATE,
1885 * skip to next rule, if PROBE_STATE just
1886 * ignore and continue with next opcode.
1887 */
1888 if (cmd->opcode == O_CHECK_STATE)
1889 goto next_rule;
1890 match = 1;
1891 break;
1892
1893 case O_ACCEPT:
1894 retval = 0; /* accept */
1895 goto done;
1896
1897 case O_PIPE:
1898 case O_QUEUE:
1899 args->rule = f; /* report matching rule */
1900 retval = cmd->arg1 | IP_FW_PORT_DYNT_FLAG;
1901 goto done;
1902
1903 case O_DIVERT:
1904 case O_TEE:
1905 if (args->eh) /* not on layer 2 */
1906 break;
1907 args->divert_rule = f->rulenum;
1908 retval = (cmd->opcode == O_DIVERT) ?
1909 cmd->arg1 :
1910 cmd->arg1 | IP_FW_PORT_TEE_FLAG;
1911 goto done;
1912
1913 case O_COUNT:
1914 case O_SKIPTO:
1915 f->pcnt++; /* update stats */
1916 f->bcnt += ip_len;
1917 f->timestamp = time_second;
1918 if (cmd->opcode == O_COUNT)
1919 goto next_rule;
1920 /* handle skipto */
1921 if (f->next_rule == NULL)
1922 lookup_next_rule(f);
1923 f = f->next_rule;
1924 goto again;
1925
1926 case O_REJECT:
1927 /*
1928 * Drop the packet and send a reject notice
1929 * if the packet is not ICMP (or is an ICMP
1930 * query), and it is not multicast/broadcast.
1931 */
1932 if (hlen > 0 &&
1933 (proto != IPPROTO_ICMP ||
1934 is_icmp_query(ip)) &&
1935 !(m->m_flags & (M_BCAST|M_MCAST)) &&
1936 !IN_MULTICAST(dst_ip.s_addr)) {
1937 send_reject(args, cmd->arg1,
1938 offset,ip_len);
1939 m = args->m;
1940 }
1941 /* FALLTHROUGH */
1942 case O_DENY:
1943 retval = IP_FW_PORT_DENY_FLAG;
1944 goto done;
1945
1946 case O_FORWARD_IP:
1947 if (args->eh) /* not valid on layer2 pkts */
1948 break;
1949 if (!q || dyn_dir == MATCH_FORWARD)
1950 args->next_hop =
1951 &((ipfw_insn_sa *)cmd)->sa;
1952 retval = 0;
1953 goto done;
1954
1955 default:
1956 panic("-- unknown opcode %d\n", cmd->opcode);
1957 } /* end of switch() on opcodes */
1958
1959 if (cmd->len & F_NOT)
1960 match = !match;
1961
1962 if (match) {
1963 if (cmd->len & F_OR)
1964 skip_or = 1;
1965 } else {
1966 if (!(cmd->len & F_OR)) /* not an OR block, */
1967 break; /* try next rule */
1968 }
1969
1970 } /* end of inner for, scan opcodes */
1971
1972 next_rule:; /* try next rule */
1973
1974 } /* end of outer for, scan rules */
1975 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
1976 return(IP_FW_PORT_DENY_FLAG);
1977
1978 done:
1979 /* Update statistics */
1980 f->pcnt++;
1981 f->bcnt += ip_len;
1982 f->timestamp = time_second;
1983 return retval;
1984
1985 pullup_failed:
1986 if (fw_verbose)
1987 printf("ipfw: pullup failed\n");
1988 return(IP_FW_PORT_DENY_FLAG);
1989 }
1990
1991 /*
1992 * When a rule is added/deleted, clear the next_rule pointers in all rules.
1993 * These will be reconstructed on the fly as packets are matched.
1994 * Must be called at splimp().
1995 */
1996 static void
1997 flush_rule_ptrs(void)
1998 {
1999 struct ip_fw *rule;
2000
2001 for (rule = layer3_chain; rule; rule = rule->next)
2002 rule->next_rule = NULL;
2003 }
2004
2005 /*
2006 * When pipes/queues are deleted, clear the "pipe_ptr" pointer to a given
2007 * pipe/queue, or to all of them (match == NULL).
2008 * Must be called at splimp().
2009 */
2010 void
2011 flush_pipe_ptrs(struct dn_flow_set *match)
2012 {
2013 struct ip_fw *rule;
2014
2015 for (rule = layer3_chain; rule; rule = rule->next) {
2016 ipfw_insn_pipe *cmd = (ipfw_insn_pipe *)ACTION_PTR(rule);
2017
2018 if (cmd->o.opcode != O_PIPE && cmd->o.opcode != O_QUEUE)
2019 continue;
2020 /*
2021 * XXX Use bcmp/bzero to handle pipe_ptr to overcome
2022 * possible alignment problems on 64-bit architectures.
2023 * This code is seldom used so we do not worry too
2024 * much about efficiency.
2025 */
2026 if (match == NULL ||
2027 !bcmp(&cmd->pipe_ptr, &match, sizeof(match)) )
2028 bzero(&cmd->pipe_ptr, sizeof(cmd->pipe_ptr));
2029 }
2030 }
2031
2032 /*
2033 * Add a new rule to the list. Copy the rule into a malloc'ed area, then
2034 * possibly create a rule number and add the rule to the list.
2035 * Update the rule_number in the input struct so the caller knows it as well.
2036 */
2037 static int
2038 add_rule(struct ip_fw **head, struct ip_fw *input_rule)
2039 {
2040 struct ip_fw *rule, *f, *prev;
2041 int s;
2042 int l = RULESIZE(input_rule);
2043
2044 if (*head == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE)
2045 return (EINVAL);
2046
2047 rule = malloc(l, M_IPFW, M_NOWAIT | M_ZERO);
2048 if (rule == NULL)
2049 return (ENOSPC);
2050
2051 bcopy(input_rule, rule, l);
2052
2053 rule->next = NULL;
2054 rule->next_rule = NULL;
2055
2056 rule->pcnt = 0;
2057 rule->bcnt = 0;
2058 rule->timestamp = 0;
2059
2060 s = splimp();
2061
2062 if (*head == NULL) { /* default rule */
2063 *head = rule;
2064 goto done;
2065 }
2066
2067 /*
2068 * If rulenum is 0, find highest numbered rule before the
2069 * default rule, and add autoinc_step
2070 */
2071 if (autoinc_step < 1)
2072 autoinc_step = 1;
2073 else if (autoinc_step > 1000)
2074 autoinc_step = 1000;
2075 if (rule->rulenum == 0) {
2076 /*
2077 * locate the highest numbered rule before default
2078 */
2079 for (f = *head; f; f = f->next) {
2080 if (f->rulenum == IPFW_DEFAULT_RULE)
2081 break;
2082 rule->rulenum = f->rulenum;
2083 }
2084 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step)
2085 rule->rulenum += autoinc_step;
2086 input_rule->rulenum = rule->rulenum;
2087 }
2088
2089 /*
2090 * Now insert the new rule in the right place in the sorted list.
2091 */
2092 for (prev = NULL, f = *head; f; prev = f, f = f->next) {
2093 if (f->rulenum > rule->rulenum) { /* found the location */
2094 if (prev) {
2095 rule->next = f;
2096 prev->next = rule;
2097 } else { /* head insert */
2098 rule->next = *head;
2099 *head = rule;
2100 }
2101 break;
2102 }
2103 }
2104 flush_rule_ptrs();
2105 done:
2106 static_count++;
2107 static_len += l;
2108 splx(s);
2109 DEB(printf("ipfw: installed rule %d, static count now %d\n",
2110 rule->rulenum, static_count);)
2111 return (0);
2112 }
2113
2114 /**
2115 * Free storage associated with a static rule (including derived
2116 * dynamic rules).
2117 * The caller is in charge of clearing rule pointers to avoid
2118 * dangling pointers.
2119 * @return a pointer to the next entry.
2120 * Arguments are not checked, so they better be correct.
2121 * Must be called at splimp().
2122 */
2123 static struct ip_fw *
2124 delete_rule(struct ip_fw **head, struct ip_fw *prev, struct ip_fw *rule)
2125 {
2126 struct ip_fw *n;
2127 int l = RULESIZE(rule);
2128
2129 n = rule->next;
2130 remove_dyn_rule(rule, NULL /* force removal */);
2131 if (prev == NULL)
2132 *head = n;
2133 else
2134 prev->next = n;
2135 static_count--;
2136 static_len -= l;
2137
2138 if (DUMMYNET_LOADED)
2139 ip_dn_ruledel_ptr(rule);
2140 free(rule, M_IPFW);
2141 return n;
2142 }
2143
2144 /*
2145 * Deletes all rules from a chain (including the default rule
2146 * if the second argument is set).
2147 * Must be called at splimp().
2148 */
2149 static void
2150 free_chain(struct ip_fw **chain, int kill_default)
2151 {
2152 struct ip_fw *rule;
2153
2154 flush_rule_ptrs(); /* more efficient to do outside the loop */
2155
2156 while ( (rule = *chain) != NULL &&
2157 (kill_default || rule->rulenum != IPFW_DEFAULT_RULE) )
2158 delete_rule(chain, NULL, rule);
2159 }
2160
2161 /**
2162 * Remove all rules with given number, and also do set manipulation.
2163 *
2164 * The argument is an u_int32_t. The low 16 bit are the rule or set number,
2165 * the next 8 bits are the new set, the top 8 bits are the command:
2166 *
2167 * 0 delete rules with given number
2168 * 1 delete rules with given set number
2169 * 2 move rules with given number to new set
2170 * 3 move rules with given set number to new set
2171 * 4 swap sets with given numbers
2172 */
2173 static int
2174 del_entry(struct ip_fw **chain, u_int32_t arg)
2175 {
2176 struct ip_fw *prev, *rule;
2177 int s;
2178 u_int16_t rulenum;
2179 u_int8_t cmd, new_set;
2180
2181 rulenum = arg & 0xffff;
2182 cmd = (arg >> 24) & 0xff;
2183 new_set = (arg >> 16) & 0xff;
2184
2185 if (cmd > 4)
2186 return EINVAL;
2187 if (new_set > 30)
2188 return EINVAL;
2189 if (cmd == 0 || cmd == 2) {
2190 if (rulenum == IPFW_DEFAULT_RULE)
2191 return EINVAL;
2192 } else {
2193 if (rulenum > 30)
2194 return EINVAL;
2195 }
2196
2197 switch (cmd) {
2198 case 0: /* delete rules with given number */
2199 /*
2200 * locate first rule to delete
2201 */
2202 for (prev = NULL, rule = *chain;
2203 rule && rule->rulenum < rulenum;
2204 prev = rule, rule = rule->next)
2205 ;
2206 if (rule->rulenum != rulenum)
2207 return EINVAL;
2208
2209 s = splimp(); /* no access to rules while removing */
2210 /*
2211 * flush pointers outside the loop, then delete all matching
2212 * rules. prev remains the same throughout the cycle.
2213 */
2214 flush_rule_ptrs();
2215 while (rule && rule->rulenum == rulenum)
2216 rule = delete_rule(chain, prev, rule);
2217 splx(s);
2218 break;
2219
2220 case 1: /* delete all rules with given set number */
2221 s = splimp();
2222 flush_rule_ptrs();
2223 for (prev = NULL, rule = *chain; rule ; )
2224 if (rule->set == rulenum)
2225 rule = delete_rule(chain, prev, rule);
2226 else {
2227 prev = rule;
2228 rule = rule->next;
2229 }
2230 splx(s);
2231 break;
2232
2233 case 2: /* move rules with given number to new set */
2234 s = splimp();
2235 for (rule = *chain; rule ; rule = rule->next)
2236 if (rule->rulenum == rulenum)
2237 rule->set = new_set;
2238 splx(s);
2239 break;
2240
2241 case 3: /* move rules with given set number to new set */
2242 s = splimp();
2243 for (rule = *chain; rule ; rule = rule->next)
2244 if (rule->set == rulenum)
2245 rule->set = new_set;
2246 splx(s);
2247 break;
2248
2249 case 4: /* swap two sets */
2250 s = splimp();
2251 for (rule = *chain; rule ; rule = rule->next)
2252 if (rule->set == rulenum)
2253 rule->set = new_set;
2254 else if (rule->set == new_set)
2255 rule->set = rulenum;
2256 splx(s);
2257 break;
2258 }
2259 return 0;
2260 }
2261
2262 /*
2263 * Clear counters for a specific rule.
2264 */
2265 static void
2266 clear_counters(struct ip_fw *rule, int log_only)
2267 {
2268 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule);
2269
2270 if (log_only == 0) {
2271 rule->bcnt = rule->pcnt = 0;
2272 rule->timestamp = 0;
2273 }
2274 if (l->o.opcode == O_LOG)
2275 l->log_left = l->max_log;
2276 }
2277
2278 /**
2279 * Reset some or all counters on firewall rules.
2280 * @arg frwl is null to clear all entries, or contains a specific
2281 * rule number.
2282 * @arg log_only is 1 if we only want to reset logs, zero otherwise.
2283 */
2284 static int
2285 zero_entry(int rulenum, int log_only)
2286 {
2287 struct ip_fw *rule;
2288 int s;
2289 char *msg;
2290
2291 if (rulenum == 0) {
2292 s = splimp();
2293 norule_counter = 0;
2294 for (rule = layer3_chain; rule; rule = rule->next)
2295 clear_counters(rule, log_only);
2296 splx(s);
2297 msg = log_only ? "ipfw: All logging counts reset.\n" :
2298 "ipfw: Accounting cleared.\n";
2299 } else {
2300 int cleared = 0;
2301 /*
2302 * We can have multiple rules with the same number, so we
2303 * need to clear them all.
2304 */
2305 for (rule = layer3_chain; rule; rule = rule->next)
2306 if (rule->rulenum == rulenum) {
2307 s = splimp();
2308 while (rule && rule->rulenum == rulenum) {
2309 clear_counters(rule, log_only);
2310 rule = rule->next;
2311 }
2312 splx(s);
2313 cleared = 1;
2314 break;
2315 }
2316 if (!cleared) /* we did not find any matching rules */
2317 return (EINVAL);
2318 msg = log_only ? "ipfw: Entry %d logging count reset.\n" :
2319 "ipfw: Entry %d cleared.\n";
2320 }
2321 if (fw_verbose)
2322 log(LOG_SECURITY | LOG_NOTICE, msg, rulenum);
2323 return (0);
2324 }
2325
2326 /*
2327 * Check validity of the structure before insert.
2328 * Fortunately rules are simple, so this mostly need to check rule sizes.
2329 */
2330 static int
2331 check_ipfw_struct(struct ip_fw *rule, int size)
2332 {
2333 int l, cmdlen = 0;
2334 int have_action=0;
2335 ipfw_insn *cmd;
2336
2337 if (size < sizeof(*rule)) {
2338 printf("ipfw: rule too short\n");
2339 return (EINVAL);
2340 }
2341 /* first, check for valid size */
2342 l = RULESIZE(rule);
2343 if (l != size) {
2344 printf("ipfw: size mismatch (have %d want %d)\n", size, l);
2345 return (EINVAL);
2346 }
2347 /*
2348 * Now go for the individual checks. Very simple ones, basically only
2349 * instruction sizes.
2350 */
2351 for (l = rule->cmd_len, cmd = rule->cmd ;
2352 l > 0 ; l -= cmdlen, cmd += cmdlen) {
2353 cmdlen = F_LEN(cmd);
2354 if (cmdlen > l) {
2355 printf("ipfw: opcode %d size truncated\n",
2356 cmd->opcode);
2357 return EINVAL;
2358 }
2359 DEB(printf("ipfw: opcode %d\n", cmd->opcode);)
2360 switch (cmd->opcode) {
2361 case O_NOP:
2362 case O_PROBE_STATE:
2363 case O_KEEP_STATE:
2364 case O_PROTO:
2365 case O_IP_SRC_ME:
2366 case O_IP_DST_ME:
2367 case O_LAYER2:
2368 case O_IN:
2369 case O_FRAG:
2370 case O_IPOPT:
2371 case O_IPLEN:
2372 case O_IPID:
2373 case O_IPTOS:
2374 case O_IPPRECEDENCE:
2375 case O_IPTTL:
2376 case O_IPVER:
2377 case O_TCPWIN:
2378 case O_TCPFLAGS:
2379 case O_TCPOPTS:
2380 case O_ESTAB:
2381 case O_VERREVPATH:
2382 if (cmdlen != F_INSN_SIZE(ipfw_insn))
2383 goto bad_size;
2384 break;
2385
2386 case O_UID:
2387 case O_GID:
2388 case O_IP_SRC:
2389 case O_IP_DST:
2390 case O_TCPSEQ:
2391 case O_TCPACK:
2392 case O_PROB:
2393 case O_ICMPTYPE:
2394 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32))
2395 goto bad_size;
2396 break;
2397
2398 case O_LIMIT:
2399 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit))
2400 goto bad_size;
2401 break;
2402
2403 case O_LOG:
2404 if (cmdlen != F_INSN_SIZE(ipfw_insn_log))
2405 goto bad_size;
2406
2407 ((ipfw_insn_log *)cmd)->log_left =
2408 ((ipfw_insn_log *)cmd)->max_log;
2409
2410 break;
2411
2412 case O_IP_SRC_MASK:
2413 case O_IP_DST_MASK:
2414 if (cmdlen != F_INSN_SIZE(ipfw_insn_ip))
2415 goto bad_size;
2416 if (((ipfw_insn_ip *)cmd)->mask.s_addr == 0) {
2417 printf("ipfw: opcode %d, useless rule\n",
2418 cmd->opcode);
2419 return EINVAL;
2420 }
2421 break;
2422
2423 case O_IP_SRC_SET:
2424 case O_IP_DST_SET:
2425 if (cmd->arg1 == 0 || cmd->arg1 > 256) {
2426 printf("ipfw: invalid set size %d\n",
2427 cmd->arg1);
2428 return EINVAL;
2429 }
2430 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
2431 (cmd->arg1+31)/32 )
2432 goto bad_size;
2433 break;
2434
2435 case O_MACADDR2:
2436 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac))
2437 goto bad_size;
2438 break;
2439
2440 case O_MAC_TYPE:
2441 case O_IP_SRCPORT:
2442 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */
2443 if (cmdlen < 2 || cmdlen > 31)
2444 goto bad_size;
2445 break;
2446
2447 case O_RECV:
2448 case O_XMIT:
2449 case O_VIA:
2450 if (cmdlen != F_INSN_SIZE(ipfw_insn_if))
2451 goto bad_size;
2452 break;
2453
2454 case O_PIPE:
2455 case O_QUEUE:
2456 if (cmdlen != F_INSN_SIZE(ipfw_insn_pipe))
2457 goto bad_size;
2458 goto check_action;
2459
2460 case O_FORWARD_IP:
2461 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa))
2462 goto bad_size;
2463 goto check_action;
2464
2465 case O_FORWARD_MAC: /* XXX not implemented yet */
2466 case O_CHECK_STATE:
2467 case O_COUNT:
2468 case O_ACCEPT:
2469 case O_DENY:
2470 case O_REJECT:
2471 case O_SKIPTO:
2472 case O_DIVERT:
2473 case O_TEE:
2474 if (cmdlen != F_INSN_SIZE(ipfw_insn))
2475 goto bad_size;
2476 check_action:
2477 if (have_action) {
2478 printf("ipfw: opcode %d, multiple actions"
2479 " not allowed\n",
2480 cmd->opcode);
2481 return EINVAL;
2482 }
2483 have_action = 1;
2484 if (l != cmdlen) {
2485 printf("ipfw: opcode %d, action must be"
2486 " last opcode\n",
2487 cmd->opcode);
2488 return EINVAL;
2489 }
2490 break;
2491 default:
2492 printf("ipfw: opcode %d, unknown opcode\n",
2493 cmd->opcode);
2494 return EINVAL;
2495 }
2496 }
2497 if (have_action == 0) {
2498 printf("ipfw: missing action\n");
2499 return EINVAL;
2500 }
2501 return 0;
2502
2503 bad_size:
2504 printf("ipfw: opcode %d size %d wrong\n",
2505 cmd->opcode, cmdlen);
2506 return EINVAL;
2507 }
2508
2509
2510 /**
2511 * {set|get}sockopt parser.
2512 */
2513 static int
2514 ipfw_ctl(struct sockopt *sopt)
2515 {
2516 int error, s, rulenum;
2517 size_t size;
2518 struct ip_fw *bp , *buf, *rule;
2519
2520 static u_int32_t rule_buf[255]; /* we copy the data here */
2521
2522 /*
2523 * Disallow modifications in really-really secure mode, but still allow
2524 * the logging counters to be reset.
2525 */
2526 if (sopt->sopt_name == IP_FW_ADD ||
2527 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) {
2528 #if __FreeBSD_version >= 500034
2529 error = securelevel_ge(sopt->sopt_td->td_ucred, 3);
2530 if (error)
2531 return (error);
2532 #else /* FreeBSD 4.x */
2533 if (securelevel >= 3)
2534 return (EPERM);
2535 #endif
2536 }
2537
2538 error = 0;
2539
2540 switch (sopt->sopt_name) {
2541 case IP_FW_GET:
2542 /*
2543 * pass up a copy of the current rules. Static rules
2544 * come first (the last of which has number IPFW_DEFAULT_RULE),
2545 * followed by a possibly empty list of dynamic rule.
2546 * The last dynamic rule has NULL in the "next" field.
2547 */
2548 s = splimp();
2549 size = static_len; /* size of static rules */
2550 if (ipfw_dyn_v) /* add size of dyn.rules */
2551 size += (dyn_count * sizeof(ipfw_dyn_rule));
2552
2553 /*
2554 * XXX todo: if the user passes a short length just to know
2555 * how much room is needed, do not bother filling up the
2556 * buffer, just jump to the sooptcopyout.
2557 */
2558 buf = malloc(size, M_TEMP, M_WAITOK);
2559 if (buf == 0) {
2560 splx(s);
2561 error = ENOBUFS;
2562 break;
2563 }
2564
2565 bp = buf;
2566 for (rule = layer3_chain; rule ; rule = rule->next) {
2567 int i = RULESIZE(rule);
2568 bcopy(rule, bp, i);
2569 bcopy(&set_disable, &(bp->next_rule),
2570 sizeof(set_disable));
2571 bp = (struct ip_fw *)((char *)bp + i);
2572 }
2573 if (ipfw_dyn_v) {
2574 int i;
2575 ipfw_dyn_rule *p, *dst, *last = NULL;
2576
2577 dst = (ipfw_dyn_rule *)bp;
2578 for (i = 0 ; i < curr_dyn_buckets ; i++ )
2579 for ( p = ipfw_dyn_v[i] ; p != NULL ;
2580 p = p->next, dst++ ) {
2581 bcopy(p, dst, sizeof *p);
2582 bcopy(&(p->rule->rulenum), &(dst->rule),
2583 sizeof(p->rule->rulenum));
2584 /*
2585 * store a non-null value in "next".
2586 * The userland code will interpret a
2587 * NULL here as a marker
2588 * for the last dynamic rule.
2589 */
2590 bcopy(&dst, &dst->next, sizeof(dst));
2591 last = dst ;
2592 dst->expire =
2593 TIME_LEQ(dst->expire, time_second) ?
2594 0 : dst->expire - time_second ;
2595 }
2596 if (last != NULL) /* mark last dynamic rule */
2597 bzero(&last->next, sizeof(last));
2598 }
2599 splx(s);
2600
2601 error = sooptcopyout(sopt, buf, size);
2602 free(buf, M_TEMP);
2603 break;
2604
2605 case IP_FW_FLUSH:
2606 /*
2607 * Normally we cannot release the lock on each iteration.
2608 * We could do it here only because we start from the head all
2609 * the times so there is no risk of missing some entries.
2610 * On the other hand, the risk is that we end up with
2611 * a very inconsistent ruleset, so better keep the lock
2612 * around the whole cycle.
2613 *
2614 * XXX this code can be improved by resetting the head of
2615 * the list to point to the default rule, and then freeing
2616 * the old list without the need for a lock.
2617 */
2618
2619 s = splimp();
2620 free_chain(&layer3_chain, 0 /* keep default rule */);
2621 splx(s);
2622 break;
2623
2624 case IP_FW_ADD:
2625 rule = (struct ip_fw *)rule_buf; /* XXX do a malloc */
2626 error = sooptcopyin(sopt, rule, sizeof(rule_buf),
2627 sizeof(struct ip_fw) );
2628 size = sopt->sopt_valsize;
2629 if (error || (error = check_ipfw_struct(rule, size)))
2630 break;
2631
2632 error = add_rule(&layer3_chain, rule);
2633 size = RULESIZE(rule);
2634 if (!error && sopt->sopt_dir == SOPT_GET)
2635 error = sooptcopyout(sopt, rule, size);
2636 break;
2637
2638 case IP_FW_DEL:
2639 /*
2640 * IP_FW_DEL is used for deleting single rules or sets,
2641 * and (ab)used to atomically manipulate sets. Argument size
2642 * is used to distinguish between the two:
2643 * sizeof(u_int32_t)
2644 * delete single rule or set of rules,
2645 * or reassign rules (or sets) to a different set.
2646 * 2*sizeof(u_int32_t)
2647 * atomic disable/enable sets.
2648 * first u_int32_t contains sets to be disabled,
2649 * second u_int32_t contains sets to be enabled.
2650 */
2651 error = sooptcopyin(sopt, rule_buf,
2652 2*sizeof(u_int32_t), sizeof(u_int32_t));
2653 if (error)
2654 break;
2655 size = sopt->sopt_valsize;
2656 if (size == sizeof(u_int32_t)) /* delete or reassign */
2657 error = del_entry(&layer3_chain, rule_buf[0]);
2658 else if (size == 2*sizeof(u_int32_t)) /* set enable/disable */
2659 set_disable =
2660 (set_disable | rule_buf[0]) & ~rule_buf[1] &
2661 ~(1<<31); /* set 31 always enabled */
2662 else
2663 error = EINVAL;
2664 break;
2665
2666 case IP_FW_ZERO:
2667 case IP_FW_RESETLOG: /* argument is an int, the rule number */
2668 rulenum=0;
2669
2670 if (sopt->sopt_val != 0) {
2671 error = sooptcopyin(sopt, &rulenum,
2672 sizeof(int), sizeof(int));
2673 if (error)
2674 break;
2675 }
2676 error = zero_entry(rulenum, sopt->sopt_name == IP_FW_RESETLOG);
2677 break;
2678
2679 default:
2680 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name);
2681 error = EINVAL;
2682 }
2683
2684 return (error);
2685 }
2686
2687 /**
2688 * dummynet needs a reference to the default rule, because rules can be
2689 * deleted while packets hold a reference to them. When this happens,
2690 * dummynet changes the reference to the default rule (it could well be a
2691 * NULL pointer, but this way we do not need to check for the special
2692 * case, plus here he have info on the default behaviour).
2693 */
2694 struct ip_fw *ip_fw_default_rule;
2695
2696 /*
2697 * This procedure is only used to handle keepalives. It is invoked
2698 * every dyn_keepalive_period
2699 */
2700 static void
2701 ipfw_tick(void * __unused unused)
2702 {
2703 int i;
2704 int s;
2705 ipfw_dyn_rule *q;
2706
2707 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0)
2708 goto done;
2709
2710 s = splimp();
2711 for (i = 0 ; i < curr_dyn_buckets ; i++) {
2712 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) {
2713 if (q->dyn_type == O_LIMIT_PARENT)
2714 continue;
2715 if (q->id.proto != IPPROTO_TCP)
2716 continue;
2717 if ( (q->state & BOTH_SYN) != BOTH_SYN)
2718 continue;
2719 if (TIME_LEQ( time_second+dyn_keepalive_interval,
2720 q->expire))
2721 continue; /* too early */
2722 if (TIME_LEQ(q->expire, time_second))
2723 continue; /* too late, rule expired */
2724
2725 send_pkt(&(q->id), q->ack_rev - 1, q->ack_fwd, TH_SYN);
2726 send_pkt(&(q->id), q->ack_fwd - 1, q->ack_rev, 0);
2727 }
2728 }
2729 splx(s);
2730 done:
2731 ipfw_timeout_h = timeout(ipfw_tick, NULL, dyn_keepalive_period*hz);
2732 }
2733
2734 static void
2735 ipfw_init(void)
2736 {
2737 struct ip_fw default_rule;
2738
2739 ip_fw_chk_ptr = ipfw_chk;
2740 ip_fw_ctl_ptr = ipfw_ctl;
2741 layer3_chain = NULL;
2742
2743 bzero(&default_rule, sizeof default_rule);
2744
2745 default_rule.act_ofs = 0;
2746 default_rule.rulenum = IPFW_DEFAULT_RULE;
2747 default_rule.cmd_len = 1;
2748 default_rule.set = 31;
2749
2750 default_rule.cmd[0].len = 1;
2751 default_rule.cmd[0].opcode =
2752 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
2753 1 ? O_ACCEPT :
2754 #endif
2755 O_DENY;
2756
2757 add_rule(&layer3_chain, &default_rule);
2758
2759 ip_fw_default_rule = layer3_chain;
2760 printf("ipfw2 initialized, divert %s, "
2761 "rule-based forwarding enabled, default to %s, logging ",
2762 #ifdef IPDIVERT
2763 "enabled",
2764 #else
2765 "disabled",
2766 #endif
2767 default_rule.cmd[0].opcode == O_ACCEPT ? "accept" : "deny");
2768
2769 #ifdef IPFIREWALL_VERBOSE
2770 fw_verbose = 1;
2771 #endif
2772 #ifdef IPFIREWALL_VERBOSE_LIMIT
2773 verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
2774 #endif
2775 if (fw_verbose == 0)
2776 printf("disabled\n");
2777 else if (verbose_limit == 0)
2778 printf("unlimited\n");
2779 else
2780 printf("limited to %d packets/entry by default\n",
2781 verbose_limit);
2782 bzero(&ipfw_timeout_h, sizeof(struct callout_handle));
2783 ipfw_timeout_h = timeout(ipfw_tick, NULL, hz);
2784 }
2785
2786 static int
2787 ipfw_modevent(module_t mod, int type, void *unused)
2788 {
2789 int s;
2790 int err = 0;
2791
2792 switch (type) {
2793 case MOD_LOAD:
2794 s = splimp();
2795 if (IPFW_LOADED) {
2796 splx(s);
2797 printf("IP firewall already loaded\n");
2798 err = EEXIST;
2799 } else {
2800 ipfw_init();
2801 splx(s);
2802 }
2803 break;
2804
2805 case MOD_UNLOAD:
2806 #if !defined(KLD_MODULE)
2807 printf("ipfw statically compiled, cannot unload\n");
2808 err = EBUSY;
2809 #else
2810 s = splimp();
2811 untimeout(ipfw_tick, NULL, ipfw_timeout_h);
2812 ip_fw_chk_ptr = NULL;
2813 ip_fw_ctl_ptr = NULL;
2814 free_chain(&layer3_chain, 1 /* kill default rule */);
2815 splx(s);
2816 printf("IP firewall unloaded\n");
2817 #endif
2818 break;
2819 default:
2820 break;
2821 }
2822 return err;
2823 }
2824
2825 static moduledata_t ipfwmod = {
2826 "ipfw",
2827 ipfw_modevent,
2828 0
2829 };
2830 DECLARE_MODULE(ipfw, ipfwmod, SI_SUB_PSEUDO, SI_ORDER_ANY);
2831 MODULE_VERSION(ipfw, 1);
2832 #endif /* IPFW2 */
Cache object: f4477a791e3f26a8b1994dc66701b6f7
|