]> git.saurik.com Git - apple/xnu.git/blob - bsd/netinet/ip_fw2.c
xnu-2782.40.9.tar.gz
[apple/xnu.git] / bsd / netinet / ip_fw2.c
1 /*
2 * Copyright (c) 2004-2013 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28
29 /*
30 * Copyright (c) 2002 Luigi Rizzo, Universita` di Pisa
31 *
32 * Redistribution and use in source and binary forms, with or without
33 * modification, are permitted provided that the following conditions
34 * are met:
35 * 1. Redistributions of source code must retain the above copyright
36 * notice, this list of conditions and the following disclaimer.
37 * 2. Redistributions in binary form must reproduce the above copyright
38 * notice, this list of conditions and the following disclaimer in the
39 * documentation and/or other materials provided with the distribution.
40 *
41 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
42 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
43 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
44 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
45 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
46 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
47 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
48 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
49 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
50 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
51 * SUCH DAMAGE.
52 *
53 * $FreeBSD: src/sys/netinet/ip_fw2.c,v 1.6.2.18 2003/10/17 11:01:03 scottl Exp $
54 */
55
56 #define DEB(x)
57 #define DDB(x) x
58
59 /*
60 * Implement IP packet firewall (new version)
61 */
62
63 #ifndef INET
64 #error IPFIREWALL requires INET.
65 #endif /* INET */
66
67 #if IPFW2
68 #include <machine/spl.h>
69
70 #include <sys/param.h>
71 #include <sys/systm.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/mcache.h>
75 #include <sys/kernel.h>
76 #include <sys/proc.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/sysctl.h>
80 #include <sys/syslog.h>
81 #include <sys/ucred.h>
82 #include <sys/kern_event.h>
83 #include <sys/kauth.h>
84
85 #include <net/if.h>
86 #include <net/route.h>
87 #include <netinet/in.h>
88 #include <netinet/in_systm.h>
89 #include <netinet/in_var.h>
90 #include <netinet/in_pcb.h>
91 #include <netinet/ip.h>
92 #include <netinet/ip_var.h>
93 #include <netinet/ip_icmp.h>
94 #include <netinet/ip_fw.h>
95 #include <netinet/ip_divert.h>
96
97 #if DUMMYNET
98 #include <netinet/ip_dummynet.h>
99 #endif /* DUMMYNET */
100
101 #include <netinet/tcp.h>
102 #include <netinet/tcp_timer.h>
103 #include <netinet/tcp_var.h>
104 #include <netinet/tcpip.h>
105 #include <netinet/udp.h>
106 #include <netinet/udp_var.h>
107
108 #ifdef IPSEC
109 #include <netinet6/ipsec.h>
110 #endif
111
112 #include <netinet/if_ether.h> /* XXX for ETHERTYPE_IP */
113
114 #include "ip_fw2_compat.h"
115
116 #include <sys/kern_event.h>
117 #include <stdarg.h>
118
119 /*
120 #include <machine/in_cksum.h>
121 */ /* XXX for in_cksum */
122
123 /*
124 * XXX This one should go in sys/mbuf.h. It is used to avoid that
125 * a firewall-generated packet loops forever through the firewall.
126 */
127 #ifndef M_SKIP_FIREWALL
128 #define M_SKIP_FIREWALL 0x4000
129 #endif
130
131 /*
132 * set_disable contains one bit per set value (0..31).
133 * If the bit is set, all rules with the corresponding set
134 * are disabled. Set RESVD_SET(31) is reserved for the default rule
135 * and rules that are not deleted by the flush command,
136 * and CANNOT be disabled.
137 * Rules in set RESVD_SET can only be deleted explicitly.
138 */
139 static u_int32_t set_disable;
140
141 int fw_verbose;
142 static int verbose_limit;
143 extern int fw_bypass;
144
145 #define IPFW_RULE_INACTIVE 1
146
147 /*
148 * list of rules for layer 3
149 */
150 static struct ip_fw *layer3_chain;
151
152 MALLOC_DEFINE(M_IPFW, "IpFw/IpAcct", "IpFw/IpAcct chain's");
153
154 static int fw_debug = 0;
155 static int autoinc_step = 100; /* bounded to 1..1000 in add_rule() */
156
157 static void ipfw_kev_post_msg(u_int32_t );
158
159 static int Get32static_len(void);
160 static int Get64static_len(void);
161
162 #ifdef SYSCTL_NODE
163
164 static int ipfw_sysctl SYSCTL_HANDLER_ARGS;
165
166 SYSCTL_NODE(_net_inet_ip, OID_AUTO, fw, CTLFLAG_RW|CTLFLAG_LOCKED, 0, "Firewall");
167 SYSCTL_PROC(_net_inet_ip_fw, OID_AUTO, enable,
168 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
169 &fw_enable, 0, ipfw_sysctl, "I", "Enable ipfw");
170 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, autoinc_step, CTLFLAG_RW | CTLFLAG_LOCKED,
171 &autoinc_step, 0, "Rule number autincrement step");
172 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, one_pass,
173 CTLFLAG_RW | CTLFLAG_LOCKED,
174 &fw_one_pass, 0,
175 "Only do a single pass through ipfw when using dummynet(4)");
176 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, debug,
177 CTLFLAG_RW | CTLFLAG_LOCKED,
178 &fw_debug, 0, "Enable printing of debug ip_fw statements");
179 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose,
180 CTLFLAG_RW | CTLFLAG_LOCKED,
181 &fw_verbose, 0, "Log matches to ipfw rules");
182 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, verbose_limit, CTLFLAG_RW | CTLFLAG_LOCKED,
183 &verbose_limit, 0, "Set upper limit of matches of ipfw rules logged");
184
185 /*
186 * IP FW Stealth Logging:
187 */
188 typedef enum ipfw_stealth_stats_type {
189 IPFW_STEALTH_STATS_UDP,
190 IPFW_STEALTH_STATS_TCP,
191 IPFW_STEALTH_STATS_UDPv6,
192 IPFW_STEALTH_STATS_TCPv6,
193 IPFW_STEALTH_STATS_MAX,
194 } ipfw_stealth_stats_type_t;
195
196 #define IPFW_STEALTH_TIMEOUT_SEC 30
197
198 #define DYN_KEEPALIVE_LEEWAY 15
199
200 // Piggybagging Stealth stats with ipfw_tick().
201 #define IPFW_STEALTH_TIMEOUT_FREQUENCY (30 / dyn_keepalive_period)
202
203 static const char* ipfw_stealth_stats_str [IPFW_STEALTH_STATS_MAX] = {
204 "UDP", "TCP", "UDP v6", "TCP v6",
205 };
206
207 static uint32_t ipfw_stealth_stats_needs_flush = FALSE;
208 static uint32_t ipfw_stealth_stats[IPFW_STEALTH_STATS_MAX];
209
210 static void ipfw_stealth_flush_stats(void);
211 void ipfw_stealth_stats_incr_udp(void);
212 void ipfw_stealth_stats_incr_tcp(void);
213 void ipfw_stealth_stats_incr_udpv6(void);
214 void ipfw_stealth_stats_incr_tcpv6(void);
215
216 /*
217 * Description of dynamic rules.
218 *
219 * Dynamic rules are stored in lists accessed through a hash table
220 * (ipfw_dyn_v) whose size is curr_dyn_buckets. This value can
221 * be modified through the sysctl variable dyn_buckets which is
222 * updated when the table becomes empty.
223 *
224 * XXX currently there is only one list, ipfw_dyn.
225 *
226 * When a packet is received, its address fields are first masked
227 * with the mask defined for the rule, then hashed, then matched
228 * against the entries in the corresponding list.
229 * Dynamic rules can be used for different purposes:
230 * + stateful rules;
231 * + enforcing limits on the number of sessions;
232 * + in-kernel NAT (not implemented yet)
233 *
234 * The lifetime of dynamic rules is regulated by dyn_*_lifetime,
235 * measured in seconds and depending on the flags.
236 *
237 * The total number of dynamic rules is stored in dyn_count.
238 * The max number of dynamic rules is dyn_max. When we reach
239 * the maximum number of rules we do not create anymore. This is
240 * done to avoid consuming too much memory, but also too much
241 * time when searching on each packet (ideally, we should try instead
242 * to put a limit on the length of the list on each bucket...).
243 *
244 * Each dynamic rule holds a pointer to the parent ipfw rule so
245 * we know what action to perform. Dynamic rules are removed when
246 * the parent rule is deleted. XXX we should make them survive.
247 *
248 * There are some limitations with dynamic rules -- we do not
249 * obey the 'randomized match', and we do not do multiple
250 * passes through the firewall. XXX check the latter!!!
251 */
252 static ipfw_dyn_rule **ipfw_dyn_v = NULL;
253 static u_int32_t dyn_buckets = 256; /* must be power of 2 */
254 static u_int32_t curr_dyn_buckets = 256; /* must be power of 2 */
255
256 /*
257 * Timeouts for various events in handing dynamic rules.
258 */
259 static u_int32_t dyn_ack_lifetime = 300;
260 static u_int32_t dyn_syn_lifetime = 20;
261 static u_int32_t dyn_fin_lifetime = 1;
262 static u_int32_t dyn_rst_lifetime = 1;
263 static u_int32_t dyn_udp_lifetime = 10;
264 static u_int32_t dyn_short_lifetime = 5;
265
266 /*
267 * Keepalives are sent if dyn_keepalive is set. They are sent every
268 * dyn_keepalive_period seconds, in the last dyn_keepalive_interval
269 * seconds of lifetime of a rule.
270 * dyn_rst_lifetime and dyn_fin_lifetime should be strictly lower
271 * than dyn_keepalive_period.
272 */
273
274 static u_int32_t dyn_keepalive_interval = 25;
275 static u_int32_t dyn_keepalive_period = 5;
276 static u_int32_t dyn_keepalive = 1; /* do send keepalives */
277
278 static u_int32_t static_count; /* # of static rules */
279 static u_int32_t static_len; /* size in bytes of static rules */
280 static u_int32_t static_len_32; /* size in bytes of static rules for 32 bit client */
281 static u_int32_t static_len_64; /* size in bytes of static rules for 64 bit client */
282 static u_int32_t dyn_count; /* # of dynamic rules */
283 static u_int32_t dyn_max = 4096; /* max # of dynamic rules */
284
285 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_buckets, CTLFLAG_RW | CTLFLAG_LOCKED,
286 &dyn_buckets, 0, "Number of dyn. buckets");
287 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, curr_dyn_buckets, CTLFLAG_RD | CTLFLAG_LOCKED,
288 &curr_dyn_buckets, 0, "Current Number of dyn. buckets");
289 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_count, CTLFLAG_RD | CTLFLAG_LOCKED,
290 &dyn_count, 0, "Number of dyn. rules");
291 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_max, CTLFLAG_RW | CTLFLAG_LOCKED,
292 &dyn_max, 0, "Max number of dyn. rules");
293 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, static_count, CTLFLAG_RD | CTLFLAG_LOCKED,
294 &static_count, 0, "Number of static rules");
295 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_ack_lifetime, CTLFLAG_RW | CTLFLAG_LOCKED,
296 &dyn_ack_lifetime, 0, "Lifetime of dyn. rules for acks");
297 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_syn_lifetime, CTLFLAG_RW | CTLFLAG_LOCKED,
298 &dyn_syn_lifetime, 0, "Lifetime of dyn. rules for syn");
299 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_fin_lifetime, CTLFLAG_RW | CTLFLAG_LOCKED,
300 &dyn_fin_lifetime, 0, "Lifetime of dyn. rules for fin");
301 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_rst_lifetime, CTLFLAG_RW | CTLFLAG_LOCKED,
302 &dyn_rst_lifetime, 0, "Lifetime of dyn. rules for rst");
303 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_udp_lifetime, CTLFLAG_RW | CTLFLAG_LOCKED,
304 &dyn_udp_lifetime, 0, "Lifetime of dyn. rules for UDP");
305 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_short_lifetime, CTLFLAG_RW | CTLFLAG_LOCKED,
306 &dyn_short_lifetime, 0, "Lifetime of dyn. rules for other situations");
307 SYSCTL_INT(_net_inet_ip_fw, OID_AUTO, dyn_keepalive, CTLFLAG_RW | CTLFLAG_LOCKED,
308 &dyn_keepalive, 0, "Enable keepalives for dyn. rules");
309
310
311 static int
312 ipfw_sysctl SYSCTL_HANDLER_ARGS
313 {
314 #pragma unused(arg1, arg2)
315 int error;
316
317 error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req);
318 if (error || !req->newptr)
319 return (error);
320
321 ipfw_kev_post_msg(KEV_IPFW_ENABLE);
322
323 return error;
324 }
325
326 #endif /* SYSCTL_NODE */
327
328
329 static ip_fw_chk_t ipfw_chk;
330
331 /* firewall lock */
332 lck_grp_t *ipfw_mutex_grp;
333 lck_grp_attr_t *ipfw_mutex_grp_attr;
334 lck_attr_t *ipfw_mutex_attr;
335 decl_lck_mtx_data(,ipfw_mutex_data);
336 lck_mtx_t *ipfw_mutex = &ipfw_mutex_data;
337
338 extern void ipfwsyslog( int level, const char *format,...);
339
340 #define KEV_LOG_SUBCLASS 10
341 #define IPFWLOGEVENT 0
342
343 #define ipfwstring "ipfw:"
344 static size_t ipfwstringlen;
345
346 #define dolog( a ) { \
347 if ( fw_verbose == 2 ) /* Apple logging, log to ipfw.log */ \
348 ipfwsyslog a ; \
349 else log a ; \
350 }
351
352 #define RULESIZE64(rule) (sizeof(struct ip_fw_64) + \
353 ((struct ip_fw *)(rule))->cmd_len * 4 - 4)
354
355 #define RULESIZE32(rule) (sizeof(struct ip_fw_32) + \
356 ((struct ip_fw *)(rule))->cmd_len * 4 - 4)
357
358 void ipfwsyslog( int level, const char *format,...)
359 {
360 #define msgsize 100
361
362 struct kev_msg ev_msg;
363 va_list ap;
364 char msgBuf[msgsize];
365 char *dptr = msgBuf;
366 unsigned char pri;
367 int loglen;
368
369 bzero(msgBuf, msgsize);
370 bzero(&ev_msg, sizeof(struct kev_msg));
371 va_start( ap, format );
372 loglen = vsnprintf(msgBuf, msgsize, format, ap);
373 va_end( ap );
374
375 ev_msg.vendor_code = KEV_VENDOR_APPLE;
376 ev_msg.kev_class = KEV_NETWORK_CLASS;
377 ev_msg.kev_subclass = KEV_LOG_SUBCLASS;
378 ev_msg.event_code = IPFWLOGEVENT;
379
380 /* get rid of the trailing \n */
381 if (loglen < msgsize)
382 dptr[loglen-1] = 0;
383 else
384 dptr[msgsize-1] = 0;
385
386 pri = LOG_PRI(level);
387
388 /* remove "ipfw:" prefix if logging to ipfw log */
389 if ( !(strncmp( ipfwstring, msgBuf, ipfwstringlen))){
390 dptr = msgBuf+ipfwstringlen;
391 }
392
393 ev_msg.dv[0].data_ptr = &pri;
394 ev_msg.dv[0].data_length = 1;
395 ev_msg.dv[1].data_ptr = dptr;
396 ev_msg.dv[1].data_length = 100; /* bug in kern_post_msg, it can't handle size > 256-msghdr */
397 ev_msg.dv[2].data_length = 0;
398
399 kev_post_msg(&ev_msg);
400 }
401
402 static inline void ipfw_stealth_stats_incr(uint32_t type)
403 {
404 if (type >= IPFW_STEALTH_STATS_MAX)
405 return;
406
407 ipfw_stealth_stats[type]++;
408
409 if (!ipfw_stealth_stats_needs_flush) {
410 ipfw_stealth_stats_needs_flush = TRUE;
411 }
412 }
413
414 void ipfw_stealth_stats_incr_udp(void)
415 {
416 ipfw_stealth_stats_incr(IPFW_STEALTH_STATS_UDP);
417 }
418
419 void ipfw_stealth_stats_incr_tcp(void)
420 {
421 ipfw_stealth_stats_incr(IPFW_STEALTH_STATS_TCP);
422 }
423
424 void ipfw_stealth_stats_incr_udpv6(void)
425 {
426 ipfw_stealth_stats_incr(IPFW_STEALTH_STATS_UDPv6);
427 }
428
429 void ipfw_stealth_stats_incr_tcpv6(void)
430 {
431 ipfw_stealth_stats_incr(IPFW_STEALTH_STATS_TCPv6);
432 }
433
434 static void ipfw_stealth_flush_stats(void)
435 {
436 int i;
437
438 for (i = 0; i < IPFW_STEALTH_STATS_MAX; i++) {
439 if (ipfw_stealth_stats[i]) {
440 ipfwsyslog (LOG_INFO, "Stealth Mode connection attempt to %s %d times",
441 ipfw_stealth_stats_str[i], ipfw_stealth_stats[i]);
442 ipfw_stealth_stats[i] = 0;
443 }
444 }
445 ipfw_stealth_stats_needs_flush = FALSE;
446 }
447
448 /*
449 * This macro maps an ip pointer into a layer3 header pointer of type T
450 */
451 #define L3HDR(T, ip) ((T *)((u_int32_t *)(ip) + (ip)->ip_hl))
452
453 static __inline int
454 icmptype_match(struct ip *ip, ipfw_insn_u32 *cmd)
455 {
456 int type = L3HDR(struct icmp,ip)->icmp_type;
457
458 return (type <= ICMP_MAXTYPE && (cmd->d[0] & (1<<type)) );
459 }
460
461 #define TT ( (1 << ICMP_ECHO) | (1 << ICMP_ROUTERSOLICIT) | \
462 (1 << ICMP_TSTAMP) | (1 << ICMP_IREQ) | (1 << ICMP_MASKREQ) )
463
464 static int
465 is_icmp_query(struct ip *ip)
466 {
467 int type = L3HDR(struct icmp, ip)->icmp_type;
468 return (type <= ICMP_MAXTYPE && (TT & (1<<type)) );
469 }
470 #undef TT
471
472 static int
473 Get32static_len()
474 {
475 int diff;
476 int len = static_len_32;
477 struct ip_fw *rule;
478 char *useraction;
479
480 for (rule = layer3_chain; rule ; rule = rule->next) {
481 if (rule->reserved_1 == IPFW_RULE_INACTIVE) {
482 continue;
483 }
484 if ( rule->act_ofs ){
485 useraction = (char*)ACTION_PTR( rule );
486 if ( ((ipfw_insn*)useraction)->opcode == O_QUEUE || ((ipfw_insn*)useraction)->opcode == O_PIPE){
487 diff = sizeof(ipfw_insn_pipe) - sizeof(ipfw_insn_pipe_32);
488 if (diff)
489 len -= diff;
490 }
491 }
492 }
493 return len;
494 }
495
496 static int
497 Get64static_len()
498 {
499 int diff;
500 int len = static_len_64;
501 struct ip_fw *rule;
502 char *useraction;
503
504 for (rule = layer3_chain; rule ; rule = rule->next) {
505 if (rule->reserved_1 == IPFW_RULE_INACTIVE) {
506 continue;
507 }
508 if ( rule->act_ofs ){
509 useraction = (char *)ACTION_PTR( rule );
510 if ( ((ipfw_insn*)useraction)->opcode == O_QUEUE || ((ipfw_insn*)useraction)->opcode == O_PIPE){
511 diff = sizeof(ipfw_insn_pipe_64) - sizeof(ipfw_insn_pipe);
512 if (diff)
513 len += diff;
514 }
515 }
516 }
517 return len;
518 }
519
520 static void
521 copyto32fw_insn( struct ip_fw_32 *fw32 , struct ip_fw *user_ip_fw, int cmdsize)
522 {
523 char *end;
524 char *fw32action;
525 char *useraction;
526 int justcmdsize;
527 int diff=0;
528 int actioncopysize;
529
530 end = ((char*)user_ip_fw->cmd) + cmdsize;
531 useraction = (char*)ACTION_PTR( user_ip_fw );
532 fw32action = (char*)fw32->cmd + (user_ip_fw->act_ofs * sizeof(uint32_t));
533 if ( ( justcmdsize = ( fw32action - (char*)fw32->cmd)))
534 bcopy( user_ip_fw->cmd, fw32->cmd, justcmdsize);
535 while ( useraction < end ){
536 if ( ((ipfw_insn*)useraction)->opcode == O_QUEUE || ((ipfw_insn*)useraction)->opcode == O_PIPE){
537 actioncopysize = sizeof(ipfw_insn_pipe_32);
538 ((ipfw_insn*)fw32action)->opcode = ((ipfw_insn*)useraction)->opcode;
539 ((ipfw_insn*)fw32action)->arg1 = ((ipfw_insn*)useraction)->arg1;
540 ((ipfw_insn*)fw32action)->len = F_INSN_SIZE(ipfw_insn_pipe_32);
541 diff = ((ipfw_insn*)useraction)->len - ((ipfw_insn*)fw32action)->len;
542 if ( diff ){
543 fw32->cmd_len -= diff;
544 }
545 } else{
546 actioncopysize = (F_LEN((ipfw_insn*)useraction) ? (F_LEN((ipfw_insn*)useraction)) : 1 ) * sizeof(uint32_t);
547 bcopy( useraction, fw32action, actioncopysize );
548 }
549 useraction += (F_LEN((ipfw_insn*)useraction) ? (F_LEN((ipfw_insn*)useraction)) : 1 ) * sizeof(uint32_t);
550 fw32action += actioncopysize;
551 }
552 }
553
554 static void
555 copyto64fw_insn( struct ip_fw_64 *fw64 , struct ip_fw *user_ip_fw, int cmdsize)
556 {
557 char *end;
558 char *fw64action;
559 char *useraction;
560 int justcmdsize;
561 int diff;
562 int actioncopysize;
563
564 end = ((char *)user_ip_fw->cmd) + cmdsize;
565 useraction = (char*)ACTION_PTR( user_ip_fw );
566 if ( (justcmdsize = (useraction - (char*)user_ip_fw->cmd)))
567 bcopy( user_ip_fw->cmd, fw64->cmd, justcmdsize);
568 fw64action = (char*)fw64->cmd + justcmdsize;
569 while ( useraction < end ){
570 if ( ((ipfw_insn*)user_ip_fw)->opcode == O_QUEUE || ((ipfw_insn*)user_ip_fw)->opcode == O_PIPE){
571 actioncopysize = sizeof(ipfw_insn_pipe_64);
572 ((ipfw_insn*)fw64action)->opcode = ((ipfw_insn*)useraction)->opcode;
573 ((ipfw_insn*)fw64action)->arg1 = ((ipfw_insn*)useraction)->arg1;
574 ((ipfw_insn*)fw64action)->len = F_INSN_SIZE(ipfw_insn_pipe_64);
575 diff = ((ipfw_insn*)fw64action)->len - ((ipfw_insn*)useraction)->len;
576 if (diff)
577 fw64->cmd_len += diff;
578
579 } else{
580 actioncopysize = (F_LEN((ipfw_insn*)useraction) ? (F_LEN((ipfw_insn*)useraction)) : 1 ) * sizeof(uint32_t);
581 bcopy( useraction, fw64action, actioncopysize );
582 }
583 useraction += (F_LEN((ipfw_insn*)useraction) ? (F_LEN((ipfw_insn*)useraction)) : 1 ) * sizeof(uint32_t);
584 fw64action += actioncopysize;
585 }
586 }
587
588 static void
589 copyto32fw( struct ip_fw *user_ip_fw, struct ip_fw_32 *fw32 , __unused size_t copysize)
590 {
591 size_t rulesize, cmdsize;
592
593 fw32->version = user_ip_fw->version;
594 fw32->context = CAST_DOWN_EXPLICIT( user32_addr_t, user_ip_fw->context);
595 fw32->next = CAST_DOWN_EXPLICIT(user32_addr_t, user_ip_fw->next);
596 fw32->next_rule = CAST_DOWN_EXPLICIT(user32_addr_t, user_ip_fw->next_rule);
597 fw32->act_ofs = user_ip_fw->act_ofs;
598 fw32->cmd_len = user_ip_fw->cmd_len;
599 fw32->rulenum = user_ip_fw->rulenum;
600 fw32->set = user_ip_fw->set;
601 fw32->set_masks[0] = user_ip_fw->set_masks[0];
602 fw32->set_masks[1] = user_ip_fw->set_masks[1];
603 fw32->pcnt = user_ip_fw->pcnt;
604 fw32->bcnt = user_ip_fw->bcnt;
605 fw32->timestamp = user_ip_fw->timestamp;
606 fw32->reserved_1 = user_ip_fw->reserved_1;
607 fw32->reserved_2 = user_ip_fw->reserved_2;
608 rulesize = sizeof(struct ip_fw_32) + (user_ip_fw->cmd_len * sizeof(ipfw_insn) - 4);
609 cmdsize = user_ip_fw->cmd_len * sizeof(u_int32_t);
610 copyto32fw_insn( fw32, user_ip_fw, cmdsize );
611 }
612
613 static void
614 copyto64fw( struct ip_fw *user_ip_fw, struct ip_fw_64 *fw64, size_t copysize)
615 {
616 size_t rulesize, cmdsize;
617
618 fw64->version = user_ip_fw->version;
619 fw64->context = CAST_DOWN_EXPLICIT(__uint64_t, user_ip_fw->context);
620 fw64->next = CAST_DOWN_EXPLICIT(user64_addr_t, user_ip_fw->next);
621 fw64->next_rule = CAST_DOWN_EXPLICIT(user64_addr_t, user_ip_fw->next_rule);
622 fw64->act_ofs = user_ip_fw->act_ofs;
623 fw64->cmd_len = user_ip_fw->cmd_len;
624 fw64->rulenum = user_ip_fw->rulenum;
625 fw64->set = user_ip_fw->set;
626 fw64->set_masks[0] = user_ip_fw->set_masks[0];
627 fw64->set_masks[1] = user_ip_fw->set_masks[1];
628 fw64->pcnt = user_ip_fw->pcnt;
629 fw64->bcnt = user_ip_fw->bcnt;
630 fw64->timestamp = user_ip_fw->timestamp;
631 fw64->reserved_1 = user_ip_fw->reserved_1;
632 fw64->reserved_2 = user_ip_fw->reserved_2;
633 rulesize = sizeof(struct ip_fw_64) + (user_ip_fw->cmd_len * sizeof(ipfw_insn) - 4);
634 if (rulesize > copysize)
635 cmdsize = copysize - sizeof(struct ip_fw_64) + 4;
636 else
637 cmdsize = user_ip_fw->cmd_len * sizeof(u_int32_t);
638 copyto64fw_insn( fw64, user_ip_fw, cmdsize);
639 }
640
641 static int
642 copyfrom32fw_insn( struct ip_fw_32 *fw32 , struct ip_fw *user_ip_fw, int cmdsize)
643 {
644 char *end;
645 char *fw32action;
646 char *useraction;
647 int justcmdsize;
648 int diff;
649 int actioncopysize;
650
651 end = ((char*)fw32->cmd) + cmdsize;
652 fw32action = (char*)ACTION_PTR( fw32 );
653 if ((justcmdsize = (fw32action - (char*)fw32->cmd)))
654 bcopy( fw32->cmd, user_ip_fw->cmd, justcmdsize);
655 useraction = (char*)user_ip_fw->cmd + justcmdsize;
656 while ( fw32action < end ){
657 if ( ((ipfw_insn*)fw32action)->opcode == O_QUEUE || ((ipfw_insn*)fw32action)->opcode == O_PIPE){
658 actioncopysize = sizeof(ipfw_insn_pipe);
659 ((ipfw_insn*)useraction)->opcode = ((ipfw_insn*)fw32action)->opcode;
660 ((ipfw_insn*)useraction)->arg1 = ((ipfw_insn*)fw32action)->arg1;
661 ((ipfw_insn*)useraction)->len = F_INSN_SIZE(ipfw_insn_pipe);
662 diff = ((ipfw_insn*)useraction)->len - ((ipfw_insn*)fw32action)->len;
663 if (diff){
664 /* readjust the cmd_len */
665 user_ip_fw->cmd_len += diff;
666 }
667 } else{
668 actioncopysize = (F_LEN((ipfw_insn*)fw32action) ? (F_LEN((ipfw_insn*)fw32action)) : 1 ) * sizeof(uint32_t);
669 bcopy( fw32action, useraction, actioncopysize );
670 }
671 fw32action += (F_LEN((ipfw_insn*)fw32action) ? (F_LEN((ipfw_insn*)fw32action)) : 1 ) * sizeof(uint32_t);
672 useraction += actioncopysize;
673 }
674
675 return( useraction - (char*)user_ip_fw->cmd );
676 }
677
678 static int
679 copyfrom64fw_insn( struct ip_fw_64 *fw64 , struct ip_fw *user_ip_fw, int cmdsize)
680 {
681 char *end;
682 char *fw64action;
683 char *useraction;
684 int justcmdsize;
685 int diff;
686 int actioncopysize;
687
688 end = ((char *)fw64->cmd) + cmdsize ;
689 fw64action = (char*)ACTION_PTR( fw64 );
690 if ( (justcmdsize = (fw64action - (char*)fw64->cmd)))
691 bcopy( fw64->cmd, user_ip_fw->cmd, justcmdsize);
692 useraction = (char*)user_ip_fw->cmd + justcmdsize;
693 while ( fw64action < end ){
694 if ( ((ipfw_insn*)fw64action)->opcode == O_QUEUE || ((ipfw_insn*)fw64action)->opcode == O_PIPE){
695 actioncopysize = sizeof(ipfw_insn_pipe);
696 ((ipfw_insn*)useraction)->opcode = ((ipfw_insn*)fw64action)->opcode;
697 ((ipfw_insn*)useraction)->arg1 = ((ipfw_insn*)fw64action)->arg1;
698 ((ipfw_insn*)useraction)->len = F_INSN_SIZE(ipfw_insn_pipe);
699 diff = ((ipfw_insn*)fw64action)->len - ((ipfw_insn*)useraction)->len;
700 if (diff) {
701 /* readjust the cmd_len */
702 user_ip_fw->cmd_len -= diff;
703 }
704 } else{
705 actioncopysize = (F_LEN((ipfw_insn*)fw64action) ? (F_LEN((ipfw_insn*)fw64action)) : 1 ) * sizeof(uint32_t);
706 bcopy( fw64action, useraction, actioncopysize );
707 }
708 fw64action += (F_LEN((ipfw_insn*)fw64action) ? (F_LEN((ipfw_insn*)fw64action)) : 1 ) * sizeof(uint32_t);
709 useraction += actioncopysize;
710 }
711 return( useraction - (char*)user_ip_fw->cmd );
712 }
713
714 static size_t
715 copyfrom32fw( struct ip_fw_32 *fw32, struct ip_fw *user_ip_fw, size_t copysize)
716 {
717 size_t rulesize, cmdsize;
718
719 user_ip_fw->version = fw32->version;
720 user_ip_fw->context = CAST_DOWN(void *, fw32->context);
721 user_ip_fw->next = CAST_DOWN(struct ip_fw*, fw32->next);
722 user_ip_fw->next_rule = CAST_DOWN_EXPLICIT(struct ip_fw*, fw32->next_rule);
723 user_ip_fw->act_ofs = fw32->act_ofs;
724 user_ip_fw->cmd_len = fw32->cmd_len;
725 user_ip_fw->rulenum = fw32->rulenum;
726 user_ip_fw->set = fw32->set;
727 user_ip_fw->set_masks[0] = fw32->set_masks[0];
728 user_ip_fw->set_masks[1] = fw32->set_masks[1];
729 user_ip_fw->pcnt = fw32->pcnt;
730 user_ip_fw->bcnt = fw32->bcnt;
731 user_ip_fw->timestamp = fw32->timestamp;
732 user_ip_fw->reserved_1 = fw32->reserved_1;
733 user_ip_fw->reserved_2 = fw32->reserved_2;
734 rulesize = sizeof(struct ip_fw_32) + (fw32->cmd_len * sizeof(ipfw_insn) - 4);
735 if ( rulesize > copysize )
736 cmdsize = copysize - sizeof(struct ip_fw_32)-4;
737 else
738 cmdsize = fw32->cmd_len * sizeof(ipfw_insn);
739 cmdsize = copyfrom32fw_insn( fw32, user_ip_fw, cmdsize);
740 return( sizeof(struct ip_fw) + cmdsize - 4);
741 }
742
743 static size_t
744 copyfrom64fw( struct ip_fw_64 *fw64, struct ip_fw *user_ip_fw, size_t copysize)
745 {
746 size_t rulesize, cmdsize;
747
748 user_ip_fw->version = fw64->version;
749 user_ip_fw->context = CAST_DOWN_EXPLICIT( void *, fw64->context);
750 user_ip_fw->next = CAST_DOWN_EXPLICIT(struct ip_fw*, fw64->next);
751 user_ip_fw->next_rule = CAST_DOWN_EXPLICIT(struct ip_fw*, fw64->next_rule);
752 user_ip_fw->act_ofs = fw64->act_ofs;
753 user_ip_fw->cmd_len = fw64->cmd_len;
754 user_ip_fw->rulenum = fw64->rulenum;
755 user_ip_fw->set = fw64->set;
756 user_ip_fw->set_masks[0] = fw64->set_masks[0];
757 user_ip_fw->set_masks[1] = fw64->set_masks[1];
758 user_ip_fw->pcnt = fw64->pcnt;
759 user_ip_fw->bcnt = fw64->bcnt;
760 user_ip_fw->timestamp = fw64->timestamp;
761 user_ip_fw->reserved_1 = fw64->reserved_1;
762 user_ip_fw->reserved_2 = fw64->reserved_2;
763 //bcopy( fw64->cmd, user_ip_fw->cmd, fw64->cmd_len * sizeof(ipfw_insn));
764 rulesize = sizeof(struct ip_fw_64) + (fw64->cmd_len * sizeof(ipfw_insn) - 4);
765 if ( rulesize > copysize )
766 cmdsize = copysize - sizeof(struct ip_fw_64)-4;
767 else
768 cmdsize = fw64->cmd_len * sizeof(ipfw_insn);
769 cmdsize = copyfrom64fw_insn( fw64, user_ip_fw, cmdsize);
770 return( sizeof(struct ip_fw) + cmdsize - 4);
771 }
772
773 void
774 externalize_flow_id(struct ipfw_flow_id *dst, struct ip_flow_id *src);
775 void
776 externalize_flow_id(struct ipfw_flow_id *dst, struct ip_flow_id *src)
777 {
778 dst->dst_ip = src->dst_ip;
779 dst->src_ip = src->src_ip;
780 dst->dst_port = src->dst_port;
781 dst->src_port = src->src_port;
782 dst->proto = src->proto;
783 dst->flags = src->flags;
784 }
785
786 static
787 void cp_dyn_to_comp_32( struct ipfw_dyn_rule_compat_32 *dyn_rule_vers1, int *len)
788 {
789 struct ipfw_dyn_rule_compat_32 *dyn_last=NULL;
790 ipfw_dyn_rule *p;
791 int i;
792
793 if (ipfw_dyn_v) {
794 for (i = 0; i < curr_dyn_buckets; i++) {
795 for ( p = ipfw_dyn_v[i] ; p != NULL ; p = p->next) {
796 dyn_rule_vers1->chain = (user32_addr_t)(p->rule->rulenum);
797 externalize_flow_id(&dyn_rule_vers1->id, &p->id);
798 externalize_flow_id(&dyn_rule_vers1->mask, &p->id);
799 dyn_rule_vers1->type = p->dyn_type;
800 dyn_rule_vers1->expire = p->expire;
801 dyn_rule_vers1->pcnt = p->pcnt;
802 dyn_rule_vers1->bcnt = p->bcnt;
803 dyn_rule_vers1->bucket = p->bucket;
804 dyn_rule_vers1->state = p->state;
805
806 dyn_rule_vers1->next = CAST_DOWN_EXPLICIT( user32_addr_t, p->next);
807 dyn_last = dyn_rule_vers1;
808
809 *len += sizeof(*dyn_rule_vers1);
810 dyn_rule_vers1++;
811 }
812 }
813
814 if (dyn_last != NULL) {
815 dyn_last->next = ((user32_addr_t)0);
816 }
817 }
818 }
819
820
821 static
822 void cp_dyn_to_comp_64( struct ipfw_dyn_rule_compat_64 *dyn_rule_vers1, int *len)
823 {
824 struct ipfw_dyn_rule_compat_64 *dyn_last=NULL;
825 ipfw_dyn_rule *p;
826 int i;
827
828 if (ipfw_dyn_v) {
829 for (i = 0; i < curr_dyn_buckets; i++) {
830 for ( p = ipfw_dyn_v[i] ; p != NULL ; p = p->next) {
831 dyn_rule_vers1->chain = (user64_addr_t) p->rule->rulenum;
832 externalize_flow_id(&dyn_rule_vers1->id, &p->id);
833 externalize_flow_id(&dyn_rule_vers1->mask, &p->id);
834 dyn_rule_vers1->type = p->dyn_type;
835 dyn_rule_vers1->expire = p->expire;
836 dyn_rule_vers1->pcnt = p->pcnt;
837 dyn_rule_vers1->bcnt = p->bcnt;
838 dyn_rule_vers1->bucket = p->bucket;
839 dyn_rule_vers1->state = p->state;
840
841 dyn_rule_vers1->next = CAST_DOWN(user64_addr_t, p->next);
842 dyn_last = dyn_rule_vers1;
843
844 *len += sizeof(*dyn_rule_vers1);
845 dyn_rule_vers1++;
846 }
847 }
848
849 if (dyn_last != NULL) {
850 dyn_last->next = CAST_DOWN(user64_addr_t, NULL);
851 }
852 }
853 }
854
855 static int
856 sooptcopyin_fw( struct sockopt *sopt, struct ip_fw *user_ip_fw, size_t *size )
857 {
858 size_t valsize, copyinsize = 0;
859 int error = 0;
860
861 valsize = sopt->sopt_valsize;
862 if ( size )
863 copyinsize = *size;
864 if (proc_is64bit(sopt->sopt_p)) {
865 struct ip_fw_64 *fw64=NULL;
866
867 if ( valsize < sizeof(struct ip_fw_64) ) {
868 return(EINVAL);
869 }
870 if ( !copyinsize )
871 copyinsize = sizeof(struct ip_fw_64);
872 if ( valsize > copyinsize )
873 sopt->sopt_valsize = valsize = copyinsize;
874
875 if ( sopt->sopt_p != 0) {
876 fw64 = _MALLOC(copyinsize, M_TEMP, M_WAITOK);
877 if ( fw64 == NULL )
878 return(ENOBUFS);
879 if ((error = copyin(sopt->sopt_val, fw64, valsize)) != 0){
880 _FREE(fw64, M_TEMP);
881 return error;
882 }
883 }
884 else {
885 bcopy(CAST_DOWN(caddr_t, sopt->sopt_val), fw64, valsize);
886 }
887 valsize = copyfrom64fw( fw64, user_ip_fw, valsize );
888 _FREE( fw64, M_TEMP);
889 }else {
890 struct ip_fw_32 *fw32=NULL;
891
892 if ( valsize < sizeof(struct ip_fw_32) ) {
893 return(EINVAL);
894 }
895 if ( !copyinsize)
896 copyinsize = sizeof(struct ip_fw_32);
897 if ( valsize > copyinsize)
898 sopt->sopt_valsize = valsize = copyinsize;
899
900 if ( sopt->sopt_p != 0) {
901 fw32 = _MALLOC(copyinsize, M_TEMP, M_WAITOK);
902 if ( fw32 == NULL )
903 return(ENOBUFS);
904 if ( (error = copyin(sopt->sopt_val, fw32, valsize)) != 0){
905 _FREE( fw32, M_TEMP);
906 return( error );
907 }
908 }
909 else {
910 bcopy(CAST_DOWN(caddr_t, sopt->sopt_val), fw32, valsize);
911 }
912 valsize = copyfrom32fw( fw32, user_ip_fw, valsize);
913 _FREE( fw32, M_TEMP);
914 }
915 if ( size )
916 *size = valsize;
917 return error;
918 }
919
920 /*
921 * The following checks use two arrays of 8 or 16 bits to store the
922 * bits that we want set or clear, respectively. They are in the
923 * low and high half of cmd->arg1 or cmd->d[0].
924 *
925 * We scan options and store the bits we find set. We succeed if
926 *
927 * (want_set & ~bits) == 0 && (want_clear & ~bits) == want_clear
928 *
929 * The code is sometimes optimized not to store additional variables.
930 */
931
932 static int
933 flags_match(ipfw_insn *cmd, u_int8_t bits)
934 {
935 u_char want_clear;
936 bits = ~bits;
937
938 if ( ((cmd->arg1 & 0xff) & bits) != 0)
939 return 0; /* some bits we want set were clear */
940 want_clear = (cmd->arg1 >> 8) & 0xff;
941 if ( (want_clear & bits) != want_clear)
942 return 0; /* some bits we want clear were set */
943 return 1;
944 }
945
946 static int
947 ipopts_match(struct ip *ip, ipfw_insn *cmd)
948 {
949 int optlen, bits = 0;
950 u_char *cp = (u_char *)(ip + 1);
951 int x = (ip->ip_hl << 2) - sizeof (struct ip);
952
953 for (; x > 0; x -= optlen, cp += optlen) {
954 int opt = cp[IPOPT_OPTVAL];
955
956 if (opt == IPOPT_EOL)
957 break;
958 if (opt == IPOPT_NOP)
959 optlen = 1;
960 else {
961 optlen = cp[IPOPT_OLEN];
962 if (optlen <= 0 || optlen > x)
963 return 0; /* invalid or truncated */
964 }
965 switch (opt) {
966
967 default:
968 break;
969
970 case IPOPT_LSRR:
971 bits |= IP_FW_IPOPT_LSRR;
972 break;
973
974 case IPOPT_SSRR:
975 bits |= IP_FW_IPOPT_SSRR;
976 break;
977
978 case IPOPT_RR:
979 bits |= IP_FW_IPOPT_RR;
980 break;
981
982 case IPOPT_TS:
983 bits |= IP_FW_IPOPT_TS;
984 break;
985 }
986 }
987 return (flags_match(cmd, bits));
988 }
989
990 static int
991 tcpopts_match(struct ip *ip, ipfw_insn *cmd)
992 {
993 int optlen, bits = 0;
994 struct tcphdr *tcp = L3HDR(struct tcphdr,ip);
995 u_char *cp = (u_char *)(tcp + 1);
996 int x = (tcp->th_off << 2) - sizeof(struct tcphdr);
997
998 for (; x > 0; x -= optlen, cp += optlen) {
999 int opt = cp[0];
1000 if (opt == TCPOPT_EOL)
1001 break;
1002 if (opt == TCPOPT_NOP)
1003 optlen = 1;
1004 else {
1005 optlen = cp[1];
1006 if (optlen <= 0)
1007 break;
1008 }
1009
1010 switch (opt) {
1011
1012 default:
1013 break;
1014
1015 case TCPOPT_MAXSEG:
1016 bits |= IP_FW_TCPOPT_MSS;
1017 break;
1018
1019 case TCPOPT_WINDOW:
1020 bits |= IP_FW_TCPOPT_WINDOW;
1021 break;
1022
1023 case TCPOPT_SACK_PERMITTED:
1024 case TCPOPT_SACK:
1025 bits |= IP_FW_TCPOPT_SACK;
1026 break;
1027
1028 case TCPOPT_TIMESTAMP:
1029 bits |= IP_FW_TCPOPT_TS;
1030 break;
1031
1032 case TCPOPT_CC:
1033 case TCPOPT_CCNEW:
1034 case TCPOPT_CCECHO:
1035 bits |= IP_FW_TCPOPT_CC;
1036 break;
1037 }
1038 }
1039 return (flags_match(cmd, bits));
1040 }
1041
1042 static int
1043 iface_match(struct ifnet *ifp, ipfw_insn_if *cmd)
1044 {
1045 if (ifp == NULL) /* no iface with this packet, match fails */
1046 return 0;
1047 /* Check by name or by IP address */
1048 if (cmd->name[0] != '\0') { /* match by name */
1049 /* Check unit number (-1 is wildcard) */
1050 if (cmd->p.unit != -1 && cmd->p.unit != ifp->if_unit)
1051 return(0);
1052 /* Check name */
1053 if (!strncmp(ifp->if_name, cmd->name, IFNAMSIZ))
1054 return(1);
1055 } else {
1056 struct ifaddr *ia;
1057
1058 ifnet_lock_shared(ifp);
1059 TAILQ_FOREACH(ia, &ifp->if_addrhead, ifa_link) {
1060 IFA_LOCK(ia);
1061 if (ia->ifa_addr->sa_family != AF_INET) {
1062 IFA_UNLOCK(ia);
1063 continue;
1064 }
1065 if (cmd->p.ip.s_addr == ((struct sockaddr_in *)
1066 (ia->ifa_addr))->sin_addr.s_addr) {
1067 IFA_UNLOCK(ia);
1068 ifnet_lock_done(ifp);
1069 return(1); /* match */
1070 }
1071 IFA_UNLOCK(ia);
1072 }
1073 ifnet_lock_done(ifp);
1074 }
1075 return(0); /* no match, fail ... */
1076 }
1077
1078 /*
1079 * The 'verrevpath' option checks that the interface that an IP packet
1080 * arrives on is the same interface that traffic destined for the
1081 * packet's source address would be routed out of. This is a measure
1082 * to block forged packets. This is also commonly known as "anti-spoofing"
1083 * or Unicast Reverse Path Forwarding (Unicast RFP) in Cisco-ese. The
1084 * name of the knob is purposely reminisent of the Cisco IOS command,
1085 *
1086 * ip verify unicast reverse-path
1087 *
1088 * which implements the same functionality. But note that syntax is
1089 * misleading. The check may be performed on all IP packets whether unicast,
1090 * multicast, or broadcast.
1091 */
1092 static int
1093 verify_rev_path(struct in_addr src, struct ifnet *ifp)
1094 {
1095 static struct route ro;
1096 struct sockaddr_in *dst;
1097
1098 bzero(&ro, sizeof (ro));
1099 dst = (struct sockaddr_in *)&(ro.ro_dst);
1100
1101 /* Check if we've cached the route from the previous call. */
1102 if (src.s_addr != dst->sin_addr.s_addr) {
1103 dst->sin_family = AF_INET;
1104 dst->sin_len = sizeof(*dst);
1105 dst->sin_addr = src;
1106
1107 rtalloc_ign(&ro, RTF_CLONING|RTF_PRCLONING);
1108 }
1109 if (ro.ro_rt != NULL) {
1110 RT_LOCK_SPIN(ro.ro_rt);
1111 } else {
1112 ROUTE_RELEASE(&ro);
1113 return 0; /* No route */
1114 }
1115 if ((ifp == NULL) ||
1116 (ro.ro_rt->rt_ifp->if_index != ifp->if_index)) {
1117 RT_UNLOCK(ro.ro_rt);
1118 ROUTE_RELEASE(&ro);
1119 return 0;
1120 }
1121 RT_UNLOCK(ro.ro_rt);
1122 ROUTE_RELEASE(&ro);
1123 return 1;
1124 }
1125
1126
1127 static u_int64_t norule_counter; /* counter for ipfw_log(NULL...) */
1128
1129 #define SNPARGS(buf, len) buf + len, sizeof(buf) > len ? sizeof(buf) - len : 0
1130 #define SNP(buf) buf, sizeof(buf)
1131
1132 /*
1133 * We enter here when we have a rule with O_LOG.
1134 * XXX this function alone takes about 2Kbytes of code!
1135 */
1136 static void
1137 ipfw_log(struct ip_fw *f, u_int hlen, struct ether_header *eh,
1138 struct mbuf *m, struct ifnet *oif)
1139 {
1140 const char *action;
1141 int limit_reached = 0;
1142 char ipv4str[MAX_IPv4_STR_LEN];
1143 char action2[40], proto[48], fragment[28];
1144
1145 fragment[0] = '\0';
1146 proto[0] = '\0';
1147
1148 if (f == NULL) { /* bogus pkt */
1149 if (verbose_limit != 0 && norule_counter >= verbose_limit)
1150 return;
1151 norule_counter++;
1152 if (norule_counter == verbose_limit)
1153 limit_reached = verbose_limit;
1154 action = "Refuse";
1155 } else { /* O_LOG is the first action, find the real one */
1156 ipfw_insn *cmd = ACTION_PTR(f);
1157 ipfw_insn_log *l = (ipfw_insn_log *)cmd;
1158
1159 if (l->max_log != 0 && l->log_left == 0)
1160 return;
1161 l->log_left--;
1162 if (l->log_left == 0)
1163 limit_reached = l->max_log;
1164 cmd += F_LEN(cmd); /* point to first action */
1165 if (cmd->opcode == O_PROB)
1166 cmd += F_LEN(cmd);
1167
1168 action = action2;
1169 switch (cmd->opcode) {
1170 case O_DENY:
1171 action = "Deny";
1172 break;
1173
1174 case O_REJECT:
1175 if (cmd->arg1==ICMP_REJECT_RST)
1176 action = "Reset";
1177 else if (cmd->arg1==ICMP_UNREACH_HOST)
1178 action = "Reject";
1179 else
1180 snprintf(SNPARGS(action2, 0), "Unreach %d",
1181 cmd->arg1);
1182 break;
1183
1184 case O_ACCEPT:
1185 action = "Accept";
1186 break;
1187 case O_COUNT:
1188 action = "Count";
1189 break;
1190 case O_DIVERT:
1191 snprintf(SNPARGS(action2, 0), "Divert %d",
1192 cmd->arg1);
1193 break;
1194 case O_TEE:
1195 snprintf(SNPARGS(action2, 0), "Tee %d",
1196 cmd->arg1);
1197 break;
1198 case O_SKIPTO:
1199 snprintf(SNPARGS(action2, 0), "SkipTo %d",
1200 cmd->arg1);
1201 break;
1202 case O_PIPE:
1203 snprintf(SNPARGS(action2, 0), "Pipe %d",
1204 cmd->arg1);
1205 break;
1206 case O_QUEUE:
1207 snprintf(SNPARGS(action2, 0), "Queue %d",
1208 cmd->arg1);
1209 break;
1210 case O_FORWARD_IP: {
1211 ipfw_insn_sa *sa = (ipfw_insn_sa *)cmd;
1212 int len;
1213
1214 if (f->reserved_1 == IPFW_RULE_INACTIVE) {
1215 break;
1216 }
1217 len = snprintf(SNPARGS(action2, 0), "Forward to %s",
1218 inet_ntop(AF_INET, &sa->sa.sin_addr, ipv4str, sizeof(ipv4str)));
1219 if (sa->sa.sin_port)
1220 snprintf(SNPARGS(action2, len), ":%d",
1221 sa->sa.sin_port);
1222 }
1223 break;
1224 default:
1225 action = "UNKNOWN";
1226 break;
1227 }
1228 }
1229
1230 if (hlen == 0) { /* non-ip */
1231 snprintf(SNPARGS(proto, 0), "MAC");
1232 } else {
1233 struct ip *ip = mtod(m, struct ip *);
1234 /* these three are all aliases to the same thing */
1235 struct icmp *const icmp = L3HDR(struct icmp, ip);
1236 struct tcphdr *const tcp = (struct tcphdr *)icmp;
1237 struct udphdr *const udp = (struct udphdr *)icmp;
1238
1239 int ip_off, offset, ip_len;
1240
1241 int len;
1242
1243 if (eh != NULL) { /* layer 2 packets are as on the wire */
1244 ip_off = ntohs(ip->ip_off);
1245 ip_len = ntohs(ip->ip_len);
1246 } else {
1247 ip_off = ip->ip_off;
1248 ip_len = ip->ip_len;
1249 }
1250 offset = ip_off & IP_OFFMASK;
1251 switch (ip->ip_p) {
1252 case IPPROTO_TCP:
1253 len = snprintf(SNPARGS(proto, 0), "TCP %s",
1254 inet_ntop(AF_INET, &ip->ip_src, ipv4str, sizeof(ipv4str)));
1255 if (offset == 0)
1256 snprintf(SNPARGS(proto, len), ":%d %s:%d",
1257 ntohs(tcp->th_sport),
1258 inet_ntop(AF_INET, &ip->ip_dst, ipv4str, sizeof(ipv4str)),
1259 ntohs(tcp->th_dport));
1260 else
1261 snprintf(SNPARGS(proto, len), " %s",
1262 inet_ntop(AF_INET, &ip->ip_dst, ipv4str, sizeof(ipv4str)));
1263 break;
1264
1265 case IPPROTO_UDP:
1266 len = snprintf(SNPARGS(proto, 0), "UDP %s",
1267 inet_ntop(AF_INET, &ip->ip_src, ipv4str, sizeof(ipv4str)));
1268 if (offset == 0)
1269 snprintf(SNPARGS(proto, len), ":%d %s:%d",
1270 ntohs(udp->uh_sport),
1271 inet_ntop(AF_INET, &ip->ip_dst, ipv4str, sizeof(ipv4str)),
1272 ntohs(udp->uh_dport));
1273 else
1274 snprintf(SNPARGS(proto, len), " %s",
1275 inet_ntop(AF_INET, &ip->ip_dst, ipv4str, sizeof(ipv4str)));
1276 break;
1277
1278 case IPPROTO_ICMP:
1279 if (offset == 0)
1280 len = snprintf(SNPARGS(proto, 0),
1281 "ICMP:%u.%u ",
1282 icmp->icmp_type, icmp->icmp_code);
1283 else
1284 len = snprintf(SNPARGS(proto, 0), "ICMP ");
1285 len += snprintf(SNPARGS(proto, len), "%s",
1286 inet_ntop(AF_INET, &ip->ip_src, ipv4str, sizeof(ipv4str)));
1287 snprintf(SNPARGS(proto, len), " %s",
1288 inet_ntop(AF_INET, &ip->ip_dst, ipv4str, sizeof(ipv4str)));
1289 break;
1290
1291 default:
1292 len = snprintf(SNPARGS(proto, 0), "P:%d %s", ip->ip_p,
1293 inet_ntop(AF_INET, &ip->ip_src, ipv4str, sizeof(ipv4str)));
1294 snprintf(SNPARGS(proto, len), " %s",
1295 inet_ntop(AF_INET, &ip->ip_dst, ipv4str, sizeof(ipv4str)));
1296 break;
1297 }
1298
1299 if (ip_off & (IP_MF | IP_OFFMASK))
1300 snprintf(SNPARGS(fragment, 0), " (frag %d:%d@%d%s)",
1301 ntohs(ip->ip_id), ip_len - (ip->ip_hl << 2),
1302 offset << 3,
1303 (ip_off & IP_MF) ? "+" : "");
1304 }
1305 if (oif || m->m_pkthdr.rcvif)
1306 {
1307 dolog((LOG_AUTHPRIV | LOG_INFO,
1308 "ipfw: %d %s %s %s via %s%d%s\n",
1309 f ? f->rulenum : -1,
1310 action, proto, oif ? "out" : "in",
1311 oif ? oif->if_name : m->m_pkthdr.rcvif->if_name,
1312 oif ? oif->if_unit : m->m_pkthdr.rcvif->if_unit,
1313 fragment));
1314 }
1315 else{
1316 dolog((LOG_AUTHPRIV | LOG_INFO,
1317 "ipfw: %d %s %s [no if info]%s\n",
1318 f ? f->rulenum : -1,
1319 action, proto, fragment));
1320 }
1321 if (limit_reached){
1322 dolog((LOG_AUTHPRIV | LOG_NOTICE,
1323 "ipfw: limit %d reached on entry %d\n",
1324 limit_reached, f ? f->rulenum : -1));
1325 }
1326 }
1327
1328 /*
1329 * IMPORTANT: the hash function for dynamic rules must be commutative
1330 * in source and destination (ip,port), because rules are bidirectional
1331 * and we want to find both in the same bucket.
1332 */
1333 static __inline int
1334 hash_packet(struct ip_flow_id *id)
1335 {
1336 u_int32_t i;
1337
1338 i = (id->dst_ip) ^ (id->src_ip) ^ (id->dst_port) ^ (id->src_port);
1339 i &= (curr_dyn_buckets - 1);
1340 return i;
1341 }
1342
1343 /**
1344 * unlink a dynamic rule from a chain. prev is a pointer to
1345 * the previous one, q is a pointer to the rule to delete,
1346 * head is a pointer to the head of the queue.
1347 * Modifies q and potentially also head.
1348 */
1349 #define UNLINK_DYN_RULE(prev, head, q) { \
1350 ipfw_dyn_rule *old_q = q; \
1351 \
1352 /* remove a refcount to the parent */ \
1353 if (q->dyn_type == O_LIMIT) \
1354 q->parent->count--; \
1355 DEB(printf("ipfw: unlink entry 0x%08x %d -> 0x%08x %d, %d left\n",\
1356 (q->id.src_ip), (q->id.src_port), \
1357 (q->id.dst_ip), (q->id.dst_port), dyn_count-1 ); ) \
1358 if (prev != NULL) \
1359 prev->next = q = q->next; \
1360 else \
1361 head = q = q->next; \
1362 dyn_count--; \
1363 _FREE(old_q, M_IPFW); }
1364
1365 #define TIME_LEQ(a,b) ((int)((a)-(b)) <= 0)
1366
1367 /**
1368 * Remove dynamic rules pointing to "rule", or all of them if rule == NULL.
1369 *
1370 * If keep_me == NULL, rules are deleted even if not expired,
1371 * otherwise only expired rules are removed.
1372 *
1373 * The value of the second parameter is also used to point to identify
1374 * a rule we absolutely do not want to remove (e.g. because we are
1375 * holding a reference to it -- this is the case with O_LIMIT_PARENT
1376 * rules). The pointer is only used for comparison, so any non-null
1377 * value will do.
1378 */
1379 static void
1380 remove_dyn_rule(struct ip_fw *rule, ipfw_dyn_rule *keep_me)
1381 {
1382 static u_int32_t last_remove = 0;
1383
1384 #define FORCE (keep_me == NULL)
1385
1386 ipfw_dyn_rule *prev, *q;
1387 int i, pass = 0, max_pass = 0;
1388 struct timeval timenow;
1389
1390 getmicrotime(&timenow);
1391
1392 if (ipfw_dyn_v == NULL || dyn_count == 0)
1393 return;
1394 /* do not expire more than once per second, it is useless */
1395 if (!FORCE && last_remove == timenow.tv_sec)
1396 return;
1397 last_remove = timenow.tv_sec;
1398
1399 /*
1400 * because O_LIMIT refer to parent rules, during the first pass only
1401 * remove child and mark any pending LIMIT_PARENT, and remove
1402 * them in a second pass.
1403 */
1404 next_pass:
1405 for (i = 0 ; i < curr_dyn_buckets ; i++) {
1406 for (prev=NULL, q = ipfw_dyn_v[i] ; q ; ) {
1407 /*
1408 * Logic can become complex here, so we split tests.
1409 */
1410 if (q == keep_me)
1411 goto next;
1412 if (rule != NULL && rule != q->rule)
1413 goto next; /* not the one we are looking for */
1414 if (q->dyn_type == O_LIMIT_PARENT) {
1415 /*
1416 * handle parent in the second pass,
1417 * record we need one.
1418 */
1419 max_pass = 1;
1420 if (pass == 0)
1421 goto next;
1422 if (FORCE && q->count != 0 ) {
1423 /* XXX should not happen! */
1424 printf("ipfw: OUCH! cannot remove rule,"
1425 " count %d\n", q->count);
1426 }
1427 } else {
1428 if (!FORCE &&
1429 !TIME_LEQ( q->expire, timenow.tv_sec ))
1430 goto next;
1431 }
1432 if (q->dyn_type != O_LIMIT_PARENT || !q->count) {
1433 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
1434 continue;
1435 }
1436 next:
1437 prev=q;
1438 q=q->next;
1439 }
1440 }
1441 if (pass++ < max_pass)
1442 goto next_pass;
1443 }
1444
1445
1446 /**
1447 * lookup a dynamic rule.
1448 */
1449 static ipfw_dyn_rule *
1450 lookup_dyn_rule(struct ip_flow_id *pkt, int *match_direction,
1451 struct tcphdr *tcp)
1452 {
1453 /*
1454 * stateful ipfw extensions.
1455 * Lookup into dynamic session queue
1456 */
1457 #define MATCH_REVERSE 0
1458 #define MATCH_FORWARD 1
1459 #define MATCH_NONE 2
1460 #define MATCH_UNKNOWN 3
1461 #define BOTH_SYN (TH_SYN | (TH_SYN << 8))
1462 #define BOTH_FIN (TH_FIN | (TH_FIN << 8))
1463
1464 int i, dir = MATCH_NONE;
1465 ipfw_dyn_rule *prev, *q=NULL;
1466 struct timeval timenow;
1467
1468 getmicrotime(&timenow);
1469
1470 if (ipfw_dyn_v == NULL)
1471 goto done; /* not found */
1472 i = hash_packet( pkt );
1473 for (prev=NULL, q = ipfw_dyn_v[i] ; q != NULL ; ) {
1474 if (q->dyn_type == O_LIMIT_PARENT && q->count)
1475 goto next;
1476 if (TIME_LEQ( q->expire, timenow.tv_sec)) { /* expire entry */
1477 int dounlink = 1;
1478
1479 /* check if entry is TCP */
1480 if ( q->id.proto == IPPROTO_TCP )
1481 {
1482 /* do not delete an established TCP connection which hasn't been closed by both sides */
1483 if ( (q->state & (BOTH_SYN | BOTH_FIN)) != (BOTH_SYN | BOTH_FIN) )
1484 dounlink = 0;
1485 }
1486 if ( dounlink ){
1487 UNLINK_DYN_RULE(prev, ipfw_dyn_v[i], q);
1488 continue;
1489 }
1490 }
1491 if (pkt->proto == q->id.proto &&
1492 q->dyn_type != O_LIMIT_PARENT) {
1493 if (pkt->src_ip == q->id.src_ip &&
1494 pkt->dst_ip == q->id.dst_ip &&
1495 pkt->src_port == q->id.src_port &&
1496 pkt->dst_port == q->id.dst_port ) {
1497 dir = MATCH_FORWARD;
1498 break;
1499 }
1500 if (pkt->src_ip == q->id.dst_ip &&
1501 pkt->dst_ip == q->id.src_ip &&
1502 pkt->src_port == q->id.dst_port &&
1503 pkt->dst_port == q->id.src_port ) {
1504 dir = MATCH_REVERSE;
1505 break;
1506 }
1507 }
1508 next:
1509 prev = q;
1510 q = q->next;
1511 }
1512 if (q == NULL)
1513 goto done; /* q = NULL, not found */
1514
1515 if ( prev != NULL) { /* found and not in front */
1516 prev->next = q->next;
1517 q->next = ipfw_dyn_v[i];
1518 ipfw_dyn_v[i] = q;
1519 }
1520 if (pkt->proto == IPPROTO_TCP) { /* update state according to flags */
1521 u_char flags = pkt->flags & (TH_FIN|TH_SYN|TH_RST);
1522
1523 q->state |= (dir == MATCH_FORWARD ) ? flags : (flags << 8);
1524 switch (q->state) {
1525 case TH_SYN: /* opening */
1526 q->expire = timenow.tv_sec + dyn_syn_lifetime;
1527 break;
1528
1529 case BOTH_SYN: /* move to established */
1530 case BOTH_SYN | TH_FIN : /* one side tries to close */
1531 case BOTH_SYN | (TH_FIN << 8) :
1532 if (tcp) {
1533 #define _SEQ_GE(a,b) ((int)(a) - (int)(b) >= 0)
1534 u_int32_t ack = ntohl(tcp->th_ack);
1535 if (dir == MATCH_FORWARD) {
1536 if (q->ack_fwd == 0 || _SEQ_GE(ack, q->ack_fwd))
1537 q->ack_fwd = ack;
1538 else { /* ignore out-of-sequence */
1539 break;
1540 }
1541 } else {
1542 if (q->ack_rev == 0 || _SEQ_GE(ack, q->ack_rev))
1543 q->ack_rev = ack;
1544 else { /* ignore out-of-sequence */
1545 break;
1546 }
1547 }
1548 }
1549 q->expire = timenow.tv_sec + dyn_ack_lifetime;
1550 break;
1551
1552 case BOTH_SYN | BOTH_FIN: /* both sides closed */
1553 if (dyn_fin_lifetime >= dyn_keepalive_period)
1554 dyn_fin_lifetime = dyn_keepalive_period - 1;
1555 q->expire = timenow.tv_sec + dyn_fin_lifetime;
1556 break;
1557
1558 default:
1559 #if 0
1560 /*
1561 * reset or some invalid combination, but can also
1562 * occur if we use keep-state the wrong way.
1563 */
1564 if ( (q->state & ((TH_RST << 8)|TH_RST)) == 0)
1565 printf("invalid state: 0x%x\n", q->state);
1566 #endif
1567 if (dyn_rst_lifetime >= dyn_keepalive_period)
1568 dyn_rst_lifetime = dyn_keepalive_period - 1;
1569 q->expire = timenow.tv_sec + dyn_rst_lifetime;
1570 break;
1571 }
1572 } else if (pkt->proto == IPPROTO_UDP) {
1573 q->expire = timenow.tv_sec + dyn_udp_lifetime;
1574 } else {
1575 /* other protocols */
1576 q->expire = timenow.tv_sec + dyn_short_lifetime;
1577 }
1578 done:
1579 if (match_direction)
1580 *match_direction = dir;
1581 return q;
1582 }
1583
1584 static void
1585 realloc_dynamic_table(void)
1586 {
1587 /*
1588 * Try reallocation, make sure we have a power of 2 and do
1589 * not allow more than 64k entries. In case of overflow,
1590 * default to 1024.
1591 */
1592
1593 if (dyn_buckets > 65536)
1594 dyn_buckets = 1024;
1595 if ((dyn_buckets & (dyn_buckets-1)) != 0) { /* not a power of 2 */
1596 dyn_buckets = curr_dyn_buckets; /* reset */
1597 return;
1598 }
1599 curr_dyn_buckets = dyn_buckets;
1600 if (ipfw_dyn_v != NULL)
1601 _FREE(ipfw_dyn_v, M_IPFW);
1602 for (;;) {
1603 ipfw_dyn_v = _MALLOC(curr_dyn_buckets * sizeof(ipfw_dyn_rule *),
1604 M_IPFW, M_NOWAIT | M_ZERO);
1605 if (ipfw_dyn_v != NULL || curr_dyn_buckets <= 2)
1606 break;
1607 curr_dyn_buckets /= 2;
1608 }
1609 }
1610
1611 /**
1612 * Install state of type 'type' for a dynamic session.
1613 * The hash table contains two type of rules:
1614 * - regular rules (O_KEEP_STATE)
1615 * - rules for sessions with limited number of sess per user
1616 * (O_LIMIT). When they are created, the parent is
1617 * increased by 1, and decreased on delete. In this case,
1618 * the third parameter is the parent rule and not the chain.
1619 * - "parent" rules for the above (O_LIMIT_PARENT).
1620 */
1621 static ipfw_dyn_rule *
1622 add_dyn_rule(struct ip_flow_id *id, u_int8_t dyn_type, struct ip_fw *rule)
1623 {
1624 ipfw_dyn_rule *r;
1625 int i;
1626 struct timeval timenow;
1627
1628 getmicrotime(&timenow);
1629
1630 if (ipfw_dyn_v == NULL ||
1631 (dyn_count == 0 && dyn_buckets != curr_dyn_buckets)) {
1632 realloc_dynamic_table();
1633 if (ipfw_dyn_v == NULL)
1634 return NULL; /* failed ! */
1635 }
1636 i = hash_packet(id);
1637
1638 r = _MALLOC(sizeof *r, M_IPFW, M_NOWAIT | M_ZERO);
1639 if (r == NULL) {
1640 #if IPFW_DEBUG
1641 printf ("ipfw: sorry cannot allocate state\n");
1642 #endif
1643 return NULL;
1644 }
1645
1646 /* increase refcount on parent, and set pointer */
1647 if (dyn_type == O_LIMIT) {
1648 ipfw_dyn_rule *parent = (ipfw_dyn_rule *)rule;
1649 if ( parent->dyn_type != O_LIMIT_PARENT)
1650 panic("invalid parent");
1651 parent->count++;
1652 r->parent = parent;
1653 rule = parent->rule;
1654 }
1655
1656 r->id = *id;
1657 r->expire = timenow.tv_sec + dyn_syn_lifetime;
1658 r->rule = rule;
1659 r->dyn_type = dyn_type;
1660 r->pcnt = r->bcnt = 0;
1661 r->count = 0;
1662
1663 r->bucket = i;
1664 r->next = ipfw_dyn_v[i];
1665 ipfw_dyn_v[i] = r;
1666 dyn_count++;
1667 DEB(printf("ipfw: add dyn entry ty %d 0x%08x %d -> 0x%08x %d, total %d\n",
1668 dyn_type,
1669 (r->id.src_ip), (r->id.src_port),
1670 (r->id.dst_ip), (r->id.dst_port),
1671 dyn_count ); )
1672 return r;
1673 }
1674
1675 /**
1676 * lookup dynamic parent rule using pkt and rule as search keys.
1677 * If the lookup fails, then install one.
1678 */
1679 static ipfw_dyn_rule *
1680 lookup_dyn_parent(struct ip_flow_id *pkt, struct ip_fw *rule)
1681 {
1682 ipfw_dyn_rule *q;
1683 int i;
1684 struct timeval timenow;
1685
1686 getmicrotime(&timenow);
1687
1688 if (ipfw_dyn_v) {
1689 i = hash_packet( pkt );
1690 for (q = ipfw_dyn_v[i] ; q != NULL ; q=q->next)
1691 if (q->dyn_type == O_LIMIT_PARENT &&
1692 rule== q->rule &&
1693 pkt->proto == q->id.proto &&
1694 pkt->src_ip == q->id.src_ip &&
1695 pkt->dst_ip == q->id.dst_ip &&
1696 pkt->src_port == q->id.src_port &&
1697 pkt->dst_port == q->id.dst_port) {
1698 q->expire = timenow.tv_sec + dyn_short_lifetime;
1699 DEB(printf("ipfw: lookup_dyn_parent found "
1700 "0x%llx\n", (uint64_t)VM_KERNEL_ADDRPERM(q));)
1701 return q;
1702 }
1703 }
1704 return add_dyn_rule(pkt, O_LIMIT_PARENT, rule);
1705 }
1706
1707 /**
1708 * Install dynamic state for rule type cmd->o.opcode
1709 *
1710 * Returns 1 (failure) if state is not installed because of errors or because
1711 * session limitations are enforced.
1712 */
1713 static int
1714 install_state(struct ip_fw *rule, ipfw_insn_limit *cmd,
1715 struct ip_fw_args *args)
1716 {
1717 static int last_log;
1718 struct timeval timenow;
1719
1720 ipfw_dyn_rule *q;
1721 getmicrotime(&timenow);
1722
1723 DEB(printf("ipfw: install state type %d 0x%08x %u -> 0x%08x %u\n",
1724 cmd->o.opcode,
1725 (args->fwa_id.src_ip), (args->fwa_id.src_port),
1726 (args->fwa_id.dst_ip), (args->fwa_id.dst_port) );)
1727
1728 q = lookup_dyn_rule(&args->fwa_id, NULL, NULL);
1729
1730 if (q != NULL) { /* should never occur */
1731 if (last_log != timenow.tv_sec) {
1732 last_log = timenow.tv_sec;
1733 printf("ipfw: install_state: entry already present, done\n");
1734 }
1735 return 0;
1736 }
1737
1738 if (dyn_count >= dyn_max)
1739 /*
1740 * Run out of slots, try to remove any expired rule.
1741 */
1742 remove_dyn_rule(NULL, (ipfw_dyn_rule *)1);
1743
1744 if (dyn_count >= dyn_max) {
1745 if (last_log != timenow.tv_sec) {
1746 last_log = timenow.tv_sec;
1747 printf("ipfw: install_state: Too many dynamic rules\n");
1748 }
1749 return 1; /* cannot install, notify caller */
1750 }
1751
1752 switch (cmd->o.opcode) {
1753 case O_KEEP_STATE: /* bidir rule */
1754 add_dyn_rule(&args->fwa_id, O_KEEP_STATE, rule);
1755 break;
1756
1757 case O_LIMIT: /* limit number of sessions */
1758 {
1759 u_int16_t limit_mask = cmd->limit_mask;
1760 struct ip_flow_id id;
1761 ipfw_dyn_rule *parent;
1762
1763 DEB(printf("ipfw: installing dyn-limit rule %d\n",
1764 cmd->conn_limit);)
1765
1766 id.dst_ip = id.src_ip = 0;
1767 id.dst_port = id.src_port = 0;
1768 id.proto = args->fwa_id.proto;
1769
1770 if (limit_mask & DYN_SRC_ADDR)
1771 id.src_ip = args->fwa_id.src_ip;
1772 if (limit_mask & DYN_DST_ADDR)
1773 id.dst_ip = args->fwa_id.dst_ip;
1774 if (limit_mask & DYN_SRC_PORT)
1775 id.src_port = args->fwa_id.src_port;
1776 if (limit_mask & DYN_DST_PORT)
1777 id.dst_port = args->fwa_id.dst_port;
1778 parent = lookup_dyn_parent(&id, rule);
1779 if (parent == NULL) {
1780 printf("ipfw: add parent failed\n");
1781 return 1;
1782 }
1783 if (parent->count >= cmd->conn_limit) {
1784 /*
1785 * See if we can remove some expired rule.
1786 */
1787 remove_dyn_rule(rule, parent);
1788 if (parent->count >= cmd->conn_limit) {
1789 if (fw_verbose && last_log != timenow.tv_sec) {
1790 last_log = timenow.tv_sec;
1791 dolog((LOG_AUTHPRIV | LOG_DEBUG,
1792 "drop session, too many entries\n"));
1793 }
1794 return 1;
1795 }
1796 }
1797 add_dyn_rule(&args->fwa_id, O_LIMIT, (struct ip_fw *)parent);
1798 }
1799 break;
1800 default:
1801 printf("ipfw: unknown dynamic rule type %u\n", cmd->o.opcode);
1802 return 1;
1803 }
1804 lookup_dyn_rule(&args->fwa_id, NULL, NULL); /* XXX just set lifetime */
1805 return 0;
1806 }
1807
1808 /*
1809 * Generate a TCP packet, containing either a RST or a keepalive.
1810 * When flags & TH_RST, we are sending a RST packet, because of a
1811 * "reset" action matched the packet.
1812 * Otherwise we are sending a keepalive, and flags & TH_
1813 */
1814 static struct mbuf *
1815 send_pkt(struct ip_flow_id *id, u_int32_t seq, u_int32_t ack, int flags)
1816 {
1817 struct mbuf *m;
1818 struct ip *ip;
1819 struct tcphdr *tcp;
1820
1821 MGETHDR(m, M_DONTWAIT, MT_HEADER); /* MAC-OK */
1822 if (m == 0)
1823 return NULL;
1824 m->m_pkthdr.rcvif = (struct ifnet *)0;
1825 m->m_pkthdr.len = m->m_len = sizeof(struct ip) + sizeof(struct tcphdr);
1826 m->m_data += max_linkhdr;
1827
1828 ip = mtod(m, struct ip *);
1829 bzero(ip, m->m_len);
1830 tcp = (struct tcphdr *)(ip + 1); /* no IP options */
1831 ip->ip_p = IPPROTO_TCP;
1832 tcp->th_off = 5;
1833 /*
1834 * Assume we are sending a RST (or a keepalive in the reverse
1835 * direction), swap src and destination addresses and ports.
1836 */
1837 ip->ip_src.s_addr = htonl(id->dst_ip);
1838 ip->ip_dst.s_addr = htonl(id->src_ip);
1839 tcp->th_sport = htons(id->dst_port);
1840 tcp->th_dport = htons(id->src_port);
1841 if (flags & TH_RST) { /* we are sending a RST */
1842 if (flags & TH_ACK) {
1843 tcp->th_seq = htonl(ack);
1844 tcp->th_ack = htonl(0);
1845 tcp->th_flags = TH_RST;
1846 } else {
1847 if (flags & TH_SYN)
1848 seq++;
1849 tcp->th_seq = htonl(0);
1850 tcp->th_ack = htonl(seq);
1851 tcp->th_flags = TH_RST | TH_ACK;
1852 }
1853 } else {
1854 /*
1855 * We are sending a keepalive. flags & TH_SYN determines
1856 * the direction, forward if set, reverse if clear.
1857 * NOTE: seq and ack are always assumed to be correct
1858 * as set by the caller. This may be confusing...
1859 */
1860 if (flags & TH_SYN) {
1861 /*
1862 * we have to rewrite the correct addresses!
1863 */
1864 ip->ip_dst.s_addr = htonl(id->dst_ip);
1865 ip->ip_src.s_addr = htonl(id->src_ip);
1866 tcp->th_dport = htons(id->dst_port);
1867 tcp->th_sport = htons(id->src_port);
1868 }
1869 tcp->th_seq = htonl(seq);
1870 tcp->th_ack = htonl(ack);
1871 tcp->th_flags = TH_ACK;
1872 }
1873 /*
1874 * set ip_len to the payload size so we can compute
1875 * the tcp checksum on the pseudoheader
1876 * XXX check this, could save a couple of words ?
1877 */
1878 ip->ip_len = htons(sizeof(struct tcphdr));
1879 tcp->th_sum = in_cksum(m, m->m_pkthdr.len);
1880 /*
1881 * now fill fields left out earlier
1882 */
1883 ip->ip_ttl = ip_defttl;
1884 ip->ip_len = m->m_pkthdr.len;
1885 m->m_flags |= M_SKIP_FIREWALL;
1886
1887 return m;
1888 }
1889
1890 /*
1891 * sends a reject message, consuming the mbuf passed as an argument.
1892 */
1893 static void
1894 send_reject(struct ip_fw_args *args, int code, int offset, __unused int ip_len)
1895 {
1896
1897 if (code != ICMP_REJECT_RST) { /* Send an ICMP unreach */
1898 /* We need the IP header in host order for icmp_error(). */
1899 if (args->fwa_eh != NULL) {
1900 struct ip *ip = mtod(args->fwa_m, struct ip *);
1901 ip->ip_len = ntohs(ip->ip_len);
1902 ip->ip_off = ntohs(ip->ip_off);
1903 }
1904 args->fwa_m->m_flags |= M_SKIP_FIREWALL;
1905 icmp_error(args->fwa_m, ICMP_UNREACH, code, 0L, 0);
1906 } else if (offset == 0 && args->fwa_id.proto == IPPROTO_TCP) {
1907 struct tcphdr *const tcp =
1908 L3HDR(struct tcphdr, mtod(args->fwa_m, struct ip *));
1909 if ( (tcp->th_flags & TH_RST) == 0) {
1910 struct mbuf *m;
1911
1912 m = send_pkt(&(args->fwa_id), ntohl(tcp->th_seq),
1913 ntohl(tcp->th_ack),
1914 tcp->th_flags | TH_RST);
1915 if (m != NULL) {
1916 struct route sro; /* fake route */
1917
1918 bzero (&sro, sizeof (sro));
1919 ip_output(m, NULL, &sro, 0, NULL, NULL);
1920 ROUTE_RELEASE(&sro);
1921 }
1922 }
1923 m_freem(args->fwa_m);
1924 } else
1925 m_freem(args->fwa_m);
1926 args->fwa_m = NULL;
1927 }
1928
1929 /**
1930 *
1931 * Given an ip_fw *, lookup_next_rule will return a pointer
1932 * to the next rule, which can be either the jump
1933 * target (for skipto instructions) or the next one in the list (in
1934 * all other cases including a missing jump target).
1935 * The result is also written in the "next_rule" field of the rule.
1936 * Backward jumps are not allowed, so start looking from the next
1937 * rule...
1938 *
1939 * This never returns NULL -- in case we do not have an exact match,
1940 * the next rule is returned. When the ruleset is changed,
1941 * pointers are flushed so we are always correct.
1942 */
1943
1944 static struct ip_fw *
1945 lookup_next_rule(struct ip_fw *me)
1946 {
1947 struct ip_fw *rule = NULL;
1948 ipfw_insn *cmd;
1949
1950 /* look for action, in case it is a skipto */
1951 cmd = ACTION_PTR(me);
1952 if (cmd->opcode == O_LOG)
1953 cmd += F_LEN(cmd);
1954 if ( cmd->opcode == O_SKIPTO )
1955 for (rule = me->next; rule ; rule = rule->next)
1956 if (rule->rulenum >= cmd->arg1)
1957 break;
1958 if (rule == NULL) /* failure or not a skipto */
1959 rule = me->next;
1960 me->next_rule = rule;
1961 return rule;
1962 }
1963
1964 /*
1965 * The main check routine for the firewall.
1966 *
1967 * All arguments are in args so we can modify them and return them
1968 * back to the caller.
1969 *
1970 * Parameters:
1971 *
1972 * args->fwa_m (in/out) The packet; we set to NULL when/if we nuke it.
1973 * Starts with the IP header.
1974 * args->fwa_eh (in) Mac header if present, or NULL for layer3 packet.
1975 * args->fwa_oif Outgoing interface, or NULL if packet is incoming.
1976 * The incoming interface is in the mbuf. (in)
1977 * args->fwa_divert_rule (in/out)
1978 * Skip up to the first rule past this rule number;
1979 * upon return, non-zero port number for divert or tee.
1980 *
1981 * args->fwa_ipfw_rule Pointer to the last matching rule (in/out)
1982 * args->fwa_next_hop Socket we are forwarding to (out).
1983 * args->fwa_id Addresses grabbed from the packet (out)
1984 *
1985 * Return value:
1986 *
1987 * IP_FW_PORT_DENY_FLAG the packet must be dropped.
1988 * 0 The packet is to be accepted and routed normally OR
1989 * the packet was denied/rejected and has been dropped;
1990 * in the latter case, *m is equal to NULL upon return.
1991 * port Divert the packet to port, with these caveats:
1992 *
1993 * - If IP_FW_PORT_TEE_FLAG is set, tee the packet instead
1994 * of diverting it (ie, 'ipfw tee').
1995 *
1996 * - If IP_FW_PORT_DYNT_FLAG is set, interpret the lower
1997 * 16 bits as a dummynet pipe number instead of diverting
1998 */
1999
2000 static int
2001 ipfw_chk(struct ip_fw_args *args)
2002 {
2003 /*
2004 * Local variables hold state during the processing of a packet.
2005 *
2006 * IMPORTANT NOTE: to speed up the processing of rules, there
2007 * are some assumption on the values of the variables, which
2008 * are documented here. Should you change them, please check
2009 * the implementation of the various instructions to make sure
2010 * that they still work.
2011 *
2012 * args->fwa_eh The MAC header. It is non-null for a layer2
2013 * packet, it is NULL for a layer-3 packet.
2014 *
2015 * m | args->fwa_m Pointer to the mbuf, as received from the caller.
2016 * It may change if ipfw_chk() does an m_pullup, or if it
2017 * consumes the packet because it calls send_reject().
2018 * XXX This has to change, so that ipfw_chk() never modifies
2019 * or consumes the buffer.
2020 * ip is simply an alias of the value of m, and it is kept
2021 * in sync with it (the packet is supposed to start with
2022 * the ip header).
2023 */
2024 struct mbuf *m = args->fwa_m;
2025 struct ip *ip = mtod(m, struct ip *);
2026
2027 /*
2028 * oif | args->fwa_oif If NULL, ipfw_chk has been called on the
2029 * inbound path (ether_input, bdg_forward, ip_input).
2030 * If non-NULL, ipfw_chk has been called on the outbound path
2031 * (ether_output, ip_output).
2032 */
2033 struct ifnet *oif = args->fwa_oif;
2034
2035 struct ip_fw *f = NULL; /* matching rule */
2036 int retval = 0;
2037
2038 /*
2039 * hlen The length of the IPv4 header.
2040 * hlen >0 means we have an IPv4 packet.
2041 */
2042 u_int hlen = 0; /* hlen >0 means we have an IP pkt */
2043
2044 /*
2045 * offset The offset of a fragment. offset != 0 means that
2046 * we have a fragment at this offset of an IPv4 packet.
2047 * offset == 0 means that (if this is an IPv4 packet)
2048 * this is the first or only fragment.
2049 */
2050 u_short offset = 0;
2051
2052 /*
2053 * Local copies of addresses. They are only valid if we have
2054 * an IP packet.
2055 *
2056 * proto The protocol. Set to 0 for non-ip packets,
2057 * or to the protocol read from the packet otherwise.
2058 * proto != 0 means that we have an IPv4 packet.
2059 *
2060 * src_port, dst_port port numbers, in HOST format. Only
2061 * valid for TCP and UDP packets.
2062 *
2063 * src_ip, dst_ip ip addresses, in NETWORK format.
2064 * Only valid for IPv4 packets.
2065 */
2066 u_int8_t proto;
2067 u_int16_t src_port = 0, dst_port = 0; /* NOTE: host format */
2068 struct in_addr src_ip = { 0 } , dst_ip = { 0 }; /* NOTE: network format */
2069 u_int16_t ip_len=0;
2070 int pktlen;
2071 int dyn_dir = MATCH_UNKNOWN;
2072 ipfw_dyn_rule *q = NULL;
2073 struct timeval timenow;
2074
2075 if (m->m_flags & M_SKIP_FIREWALL || fw_bypass) {
2076 return 0; /* accept */
2077 }
2078
2079 /*
2080 * Clear packet chain if we find one here.
2081 */
2082
2083 if (m->m_nextpkt != NULL) {
2084 m_freem_list(m->m_nextpkt);
2085 m->m_nextpkt = NULL;
2086 }
2087
2088 lck_mtx_lock(ipfw_mutex);
2089
2090 getmicrotime(&timenow);
2091 /*
2092 * dyn_dir = MATCH_UNKNOWN when rules unchecked,
2093 * MATCH_NONE when checked and not matched (q = NULL),
2094 * MATCH_FORWARD or MATCH_REVERSE otherwise (q != NULL)
2095 */
2096
2097 pktlen = m->m_pkthdr.len;
2098 if (args->fwa_eh == NULL || /* layer 3 packet */
2099 ( m->m_pkthdr.len >= sizeof(struct ip) &&
2100 ntohs(args->fwa_eh->ether_type) == ETHERTYPE_IP))
2101 hlen = ip->ip_hl << 2;
2102
2103 /*
2104 * Collect parameters into local variables for faster matching.
2105 */
2106 if (hlen == 0) { /* do not grab addresses for non-ip pkts */
2107 proto = args->fwa_id.proto = 0; /* mark f_id invalid */
2108 goto after_ip_checks;
2109 }
2110
2111 proto = args->fwa_id.proto = ip->ip_p;
2112 src_ip = ip->ip_src;
2113 dst_ip = ip->ip_dst;
2114 if (args->fwa_eh != NULL) { /* layer 2 packets are as on the wire */
2115 offset = ntohs(ip->ip_off) & IP_OFFMASK;
2116 ip_len = ntohs(ip->ip_len);
2117 } else {
2118 offset = ip->ip_off & IP_OFFMASK;
2119 ip_len = ip->ip_len;
2120 }
2121 pktlen = ip_len < pktlen ? ip_len : pktlen;
2122
2123 #define PULLUP_TO(len) \
2124 do { \
2125 if ((m)->m_len < (len)) { \
2126 args->fwa_m = m = m_pullup(m, (len)); \
2127 if (m == 0) \
2128 goto pullup_failed; \
2129 ip = mtod(m, struct ip *); \
2130 } \
2131 } while (0)
2132
2133 if (offset == 0) {
2134 switch (proto) {
2135 case IPPROTO_TCP:
2136 {
2137 struct tcphdr *tcp;
2138
2139 PULLUP_TO(hlen + sizeof(struct tcphdr));
2140 tcp = L3HDR(struct tcphdr, ip);
2141 dst_port = tcp->th_dport;
2142 src_port = tcp->th_sport;
2143 args->fwa_id.flags = tcp->th_flags;
2144 }
2145 break;
2146
2147 case IPPROTO_UDP:
2148 {
2149 struct udphdr *udp;
2150
2151 PULLUP_TO(hlen + sizeof(struct udphdr));
2152 udp = L3HDR(struct udphdr, ip);
2153 dst_port = udp->uh_dport;
2154 src_port = udp->uh_sport;
2155 }
2156 break;
2157
2158 case IPPROTO_ICMP:
2159 PULLUP_TO(hlen + 4); /* type, code and checksum. */
2160 args->fwa_id.flags = L3HDR(struct icmp, ip)->icmp_type;
2161 break;
2162
2163 default:
2164 break;
2165 }
2166 #undef PULLUP_TO
2167 }
2168
2169 args->fwa_id.src_ip = ntohl(src_ip.s_addr);
2170 args->fwa_id.dst_ip = ntohl(dst_ip.s_addr);
2171 args->fwa_id.src_port = src_port = ntohs(src_port);
2172 args->fwa_id.dst_port = dst_port = ntohs(dst_port);
2173
2174 after_ip_checks:
2175 if (args->fwa_ipfw_rule) {
2176 /*
2177 * Packet has already been tagged. Look for the next rule
2178 * to restart processing.
2179 *
2180 * If fw_one_pass != 0 then just accept it.
2181 * XXX should not happen here, but optimized out in
2182 * the caller.
2183 */
2184 if (fw_one_pass) {
2185 lck_mtx_unlock(ipfw_mutex);
2186 return 0;
2187 }
2188
2189 f = args->fwa_ipfw_rule->next_rule;
2190 if (f == NULL)
2191 f = lookup_next_rule(args->fwa_ipfw_rule);
2192 } else {
2193 /*
2194 * Find the starting rule. It can be either the first
2195 * one, or the one after divert_rule if asked so.
2196 */
2197 int skipto = args->fwa_divert_rule;
2198
2199 f = layer3_chain;
2200 if (args->fwa_eh == NULL && skipto != 0) {
2201 if (skipto >= IPFW_DEFAULT_RULE) {
2202 lck_mtx_unlock(ipfw_mutex);
2203 return(IP_FW_PORT_DENY_FLAG); /* invalid */
2204 }
2205 while (f && f->rulenum <= skipto)
2206 f = f->next;
2207 if (f == NULL) { /* drop packet */
2208 lck_mtx_unlock(ipfw_mutex);
2209 return(IP_FW_PORT_DENY_FLAG);
2210 }
2211 }
2212 }
2213 args->fwa_divert_rule = 0; /* reset to avoid confusion later */
2214
2215 /*
2216 * Now scan the rules, and parse microinstructions for each rule.
2217 */
2218 for (; f; f = f->next) {
2219 int l, cmdlen;
2220 ipfw_insn *cmd;
2221 int skip_or; /* skip rest of OR block */
2222
2223 again:
2224 if (f->reserved_1 == IPFW_RULE_INACTIVE) {
2225 continue;
2226 }
2227
2228 if (set_disable & (1 << f->set) )
2229 continue;
2230
2231 skip_or = 0;
2232 for (l = f->cmd_len, cmd = f->cmd ; l > 0 ;
2233 l -= cmdlen, cmd += cmdlen) {
2234 int match;
2235
2236 /*
2237 * check_body is a jump target used when we find a
2238 * CHECK_STATE, and need to jump to the body of
2239 * the target rule.
2240 */
2241
2242 check_body:
2243 cmdlen = F_LEN(cmd);
2244 /*
2245 * An OR block (insn_1 || .. || insn_n) has the
2246 * F_OR bit set in all but the last instruction.
2247 * The first match will set "skip_or", and cause
2248 * the following instructions to be skipped until
2249 * past the one with the F_OR bit clear.
2250 */
2251 if (skip_or) { /* skip this instruction */
2252 if ((cmd->len & F_OR) == 0)
2253 skip_or = 0; /* next one is good */
2254 continue;
2255 }
2256 match = 0; /* set to 1 if we succeed */
2257
2258 switch (cmd->opcode) {
2259 /*
2260 * The first set of opcodes compares the packet's
2261 * fields with some pattern, setting 'match' if a
2262 * match is found. At the end of the loop there is
2263 * logic to deal with F_NOT and F_OR flags associated
2264 * with the opcode.
2265 */
2266 case O_NOP:
2267 match = 1;
2268 break;
2269
2270 case O_FORWARD_MAC:
2271 printf("ipfw: opcode %d unimplemented\n",
2272 cmd->opcode);
2273 break;
2274
2275 #ifndef __APPLE__
2276 case O_GID:
2277 #endif
2278 case O_UID:
2279 /*
2280 * We only check offset == 0 && proto != 0,
2281 * as this ensures that we have an IPv4
2282 * packet with the ports info.
2283 */
2284 if (offset!=0)
2285 break;
2286
2287 {
2288 struct inpcbinfo *pi;
2289 int wildcard;
2290 struct inpcb *pcb;
2291
2292 if (proto == IPPROTO_TCP) {
2293 wildcard = 0;
2294 pi = &tcbinfo;
2295 } else if (proto == IPPROTO_UDP) {
2296 wildcard = 1;
2297 pi = &udbinfo;
2298 } else
2299 break;
2300
2301 pcb = (oif) ?
2302 in_pcblookup_hash(pi,
2303 dst_ip, htons(dst_port),
2304 src_ip, htons(src_port),
2305 wildcard, oif) :
2306 in_pcblookup_hash(pi,
2307 src_ip, htons(src_port),
2308 dst_ip, htons(dst_port),
2309 wildcard, NULL);
2310
2311 if (pcb == NULL || pcb->inp_socket == NULL)
2312 break;
2313 #if __FreeBSD_version < 500034
2314 #define socheckuid(a,b) (kauth_cred_getuid((a)->so_cred) != (b))
2315 #endif
2316 if (cmd->opcode == O_UID) {
2317 match =
2318 #ifdef __APPLE__
2319 (kauth_cred_getuid(pcb->inp_socket->so_cred) == (uid_t)((ipfw_insn_u32 *)cmd)->d[0]);
2320 #else
2321 !socheckuid(pcb->inp_socket,
2322 (uid_t)((ipfw_insn_u32 *)cmd)->d[0]);
2323 #endif
2324 }
2325 #ifndef __APPLE__
2326 else {
2327 match = 0;
2328 kauth_cred_ismember_gid(pcb->inp_socket->so_cred,
2329 (gid_t)((ipfw_insn_u32 *)cmd)->d[0], &match);
2330 }
2331 #endif
2332 /* release reference on pcb */
2333 in_pcb_checkstate(pcb, WNT_RELEASE, 0);
2334 }
2335
2336 break;
2337
2338 case O_RECV:
2339 match = iface_match(m->m_pkthdr.rcvif,
2340 (ipfw_insn_if *)cmd);
2341 break;
2342
2343 case O_XMIT:
2344 match = iface_match(oif, (ipfw_insn_if *)cmd);
2345 break;
2346
2347 case O_VIA:
2348 match = iface_match(oif ? oif :
2349 m->m_pkthdr.rcvif, (ipfw_insn_if *)cmd);
2350 break;
2351
2352 case O_MACADDR2:
2353 if (args->fwa_eh != NULL) { /* have MAC header */
2354 u_int32_t *want = (u_int32_t *)
2355 ((ipfw_insn_mac *)cmd)->addr;
2356 u_int32_t *mask = (u_int32_t *)
2357 ((ipfw_insn_mac *)cmd)->mask;
2358 u_int32_t *hdr = (u_int32_t *)args->fwa_eh;
2359
2360 match =
2361 ( want[0] == (hdr[0] & mask[0]) &&
2362 want[1] == (hdr[1] & mask[1]) &&
2363 want[2] == (hdr[2] & mask[2]) );
2364 }
2365 break;
2366
2367 case O_MAC_TYPE:
2368 if (args->fwa_eh != NULL) {
2369 u_int16_t t =
2370 ntohs(args->fwa_eh->ether_type);
2371 u_int16_t *p =
2372 ((ipfw_insn_u16 *)cmd)->ports;
2373 int i;
2374
2375 for (i = cmdlen - 1; !match && i>0;
2376 i--, p += 2)
2377 match = (t>=p[0] && t<=p[1]);
2378 }
2379 break;
2380
2381 case O_FRAG:
2382 match = (hlen > 0 && offset != 0);
2383 break;
2384
2385 case O_IN: /* "out" is "not in" */
2386 match = (oif == NULL);
2387 break;
2388
2389 case O_LAYER2:
2390 match = (args->fwa_eh != NULL);
2391 break;
2392
2393 case O_PROTO:
2394 /*
2395 * We do not allow an arg of 0 so the
2396 * check of "proto" only suffices.
2397 */
2398 match = (proto == cmd->arg1);
2399 break;
2400
2401 case O_IP_SRC:
2402 match = (hlen > 0 &&
2403 ((ipfw_insn_ip *)cmd)->addr.s_addr ==
2404 src_ip.s_addr);
2405 break;
2406
2407 case O_IP_SRC_MASK:
2408 case O_IP_DST_MASK:
2409 if (hlen > 0) {
2410 uint32_t a =
2411 (cmd->opcode == O_IP_DST_MASK) ?
2412 dst_ip.s_addr : src_ip.s_addr;
2413 uint32_t *p = ((ipfw_insn_u32 *)cmd)->d;
2414 int i = cmdlen-1;
2415
2416 for (; !match && i>0; i-= 2, p+= 2)
2417 match = (p[0] == (a & p[1]));
2418 }
2419 break;
2420
2421 case O_IP_SRC_ME:
2422 if (hlen > 0) {
2423 struct ifnet *tif;
2424
2425 INADDR_TO_IFP(src_ip, tif);
2426 match = (tif != NULL);
2427 }
2428 break;
2429
2430 case O_IP_DST_SET:
2431 case O_IP_SRC_SET:
2432 if (hlen > 0) {
2433 u_int32_t *d = (u_int32_t *)(cmd+1);
2434 u_int32_t addr =
2435 cmd->opcode == O_IP_DST_SET ?
2436 args->fwa_id.dst_ip :
2437 args->fwa_id.src_ip;
2438
2439 if (addr < d[0])
2440 break;
2441 addr -= d[0]; /* subtract base */
2442 match = (addr < cmd->arg1) &&
2443 ( d[ 1 + (addr>>5)] &
2444 (1<<(addr & 0x1f)) );
2445 }
2446 break;
2447
2448 case O_IP_DST:
2449 match = (hlen > 0 &&
2450 ((ipfw_insn_ip *)cmd)->addr.s_addr ==
2451 dst_ip.s_addr);
2452 break;
2453
2454 case O_IP_DST_ME:
2455 if (hlen > 0) {
2456 struct ifnet *tif;
2457
2458 INADDR_TO_IFP(dst_ip, tif);
2459 match = (tif != NULL);
2460 }
2461 break;
2462
2463 case O_IP_SRCPORT:
2464 case O_IP_DSTPORT:
2465 /*
2466 * offset == 0 && proto != 0 is enough
2467 * to guarantee that we have an IPv4
2468 * packet with port info.
2469 */
2470 if ((proto==IPPROTO_UDP || proto==IPPROTO_TCP)
2471 && offset == 0) {
2472 u_int16_t x =
2473 (cmd->opcode == O_IP_SRCPORT) ?
2474 src_port : dst_port ;
2475 u_int16_t *p =
2476 ((ipfw_insn_u16 *)cmd)->ports;
2477 int i;
2478
2479 for (i = cmdlen - 1; !match && i>0;
2480 i--, p += 2)
2481 match = (x>=p[0] && x<=p[1]);
2482 }
2483 break;
2484
2485 case O_ICMPTYPE:
2486 match = (offset == 0 && proto==IPPROTO_ICMP &&
2487 icmptype_match(ip, (ipfw_insn_u32 *)cmd) );
2488 break;
2489
2490 case O_IPOPT:
2491 match = (hlen > 0 && ipopts_match(ip, cmd) );
2492 break;
2493
2494 case O_IPVER:
2495 match = (hlen > 0 && cmd->arg1 == ip->ip_v);
2496 break;
2497
2498 case O_IPID:
2499 case O_IPLEN:
2500 case O_IPTTL:
2501 if (hlen > 0) { /* only for IP packets */
2502 uint16_t x;
2503 uint16_t *p;
2504 int i;
2505
2506 if (cmd->opcode == O_IPLEN)
2507 x = ip_len;
2508 else if (cmd->opcode == O_IPTTL)
2509 x = ip->ip_ttl;
2510 else /* must be IPID */
2511 x = ntohs(ip->ip_id);
2512 if (cmdlen == 1) {
2513 match = (cmd->arg1 == x);
2514 break;
2515 }
2516 /* otherwise we have ranges */
2517 p = ((ipfw_insn_u16 *)cmd)->ports;
2518 i = cmdlen - 1;
2519 for (; !match && i>0; i--, p += 2)
2520 match = (x >= p[0] && x <= p[1]);
2521 }
2522 break;
2523
2524 case O_IPPRECEDENCE:
2525 match = (hlen > 0 &&
2526 (cmd->arg1 == (ip->ip_tos & 0xe0)) );
2527 break;
2528
2529 case O_IPTOS:
2530 match = (hlen > 0 &&
2531 flags_match(cmd, ip->ip_tos));
2532 break;
2533
2534 case O_TCPFLAGS:
2535 match = (proto == IPPROTO_TCP && offset == 0 &&
2536 flags_match(cmd,
2537 L3HDR(struct tcphdr,ip)->th_flags));
2538 break;
2539
2540 case O_TCPOPTS:
2541 match = (proto == IPPROTO_TCP && offset == 0 &&
2542 tcpopts_match(ip, cmd));
2543 break;
2544
2545 case O_TCPSEQ:
2546 match = (proto == IPPROTO_TCP && offset == 0 &&
2547 ((ipfw_insn_u32 *)cmd)->d[0] ==
2548 L3HDR(struct tcphdr,ip)->th_seq);
2549 break;
2550
2551 case O_TCPACK:
2552 match = (proto == IPPROTO_TCP && offset == 0 &&
2553 ((ipfw_insn_u32 *)cmd)->d[0] ==
2554 L3HDR(struct tcphdr,ip)->th_ack);
2555 break;
2556
2557 case O_TCPWIN:
2558 match = (proto == IPPROTO_TCP && offset == 0 &&
2559 cmd->arg1 ==
2560 L3HDR(struct tcphdr,ip)->th_win);
2561 break;
2562
2563 case O_ESTAB:
2564 /* reject packets which have SYN only */
2565 /* XXX should i also check for TH_ACK ? */
2566 match = (proto == IPPROTO_TCP && offset == 0 &&
2567 (L3HDR(struct tcphdr,ip)->th_flags &
2568 (TH_RST | TH_ACK | TH_SYN)) != TH_SYN);
2569 break;
2570
2571 case O_LOG:
2572 if (fw_verbose)
2573 ipfw_log(f, hlen, args->fwa_eh, m, oif);
2574 match = 1;
2575 break;
2576
2577 case O_PROB:
2578 match = (random()<((ipfw_insn_u32 *)cmd)->d[0]);
2579 break;
2580
2581 case O_VERREVPATH:
2582 /* Outgoing packets automatically pass/match */
2583 match = ((oif != NULL) ||
2584 (m->m_pkthdr.rcvif == NULL) ||
2585 verify_rev_path(src_ip, m->m_pkthdr.rcvif));
2586 break;
2587
2588 case O_IPSEC:
2589 #ifdef FAST_IPSEC
2590 match = (m_tag_find(m,
2591 PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL);
2592 #endif
2593 #ifdef IPSEC
2594 match = (ipsec_gethist(m, NULL) != NULL);
2595 #endif
2596 /* otherwise no match */
2597 break;
2598
2599 /*
2600 * The second set of opcodes represents 'actions',
2601 * i.e. the terminal part of a rule once the packet
2602 * matches all previous patterns.
2603 * Typically there is only one action for each rule,
2604 * and the opcode is stored at the end of the rule
2605 * (but there are exceptions -- see below).
2606 *
2607 * In general, here we set retval and terminate the
2608 * outer loop (would be a 'break 3' in some language,
2609 * but we need to do a 'goto done').
2610 *
2611 * Exceptions:
2612 * O_COUNT and O_SKIPTO actions:
2613 * instead of terminating, we jump to the next rule
2614 * ('goto next_rule', equivalent to a 'break 2'),
2615 * or to the SKIPTO target ('goto again' after
2616 * having set f, cmd and l), respectively.
2617 *
2618 * O_LIMIT and O_KEEP_STATE: these opcodes are
2619 * not real 'actions', and are stored right
2620 * before the 'action' part of the rule.
2621 * These opcodes try to install an entry in the
2622 * state tables; if successful, we continue with
2623 * the next opcode (match=1; break;), otherwise
2624 * the packet * must be dropped
2625 * ('goto done' after setting retval);
2626 *
2627 * O_PROBE_STATE and O_CHECK_STATE: these opcodes
2628 * cause a lookup of the state table, and a jump
2629 * to the 'action' part of the parent rule
2630 * ('goto check_body') if an entry is found, or
2631 * (CHECK_STATE only) a jump to the next rule if
2632 * the entry is not found ('goto next_rule').
2633 * The result of the lookup is cached to make
2634 * further instances of these opcodes are
2635 * effectively NOPs.
2636 */
2637 case O_LIMIT:
2638 case O_KEEP_STATE:
2639 if (install_state(f,
2640 (ipfw_insn_limit *)cmd, args)) {
2641 retval = IP_FW_PORT_DENY_FLAG;
2642 goto done; /* error/limit violation */
2643 }
2644 match = 1;
2645 break;
2646
2647 case O_PROBE_STATE:
2648 case O_CHECK_STATE:
2649 /*
2650 * dynamic rules are checked at the first
2651 * keep-state or check-state occurrence,
2652 * with the result being stored in dyn_dir.
2653 * The compiler introduces a PROBE_STATE
2654 * instruction for us when we have a
2655 * KEEP_STATE (because PROBE_STATE needs
2656 * to be run first).
2657 */
2658 if (dyn_dir == MATCH_UNKNOWN &&
2659 (q = lookup_dyn_rule(&args->fwa_id,
2660 &dyn_dir, proto == IPPROTO_TCP ?
2661 L3HDR(struct tcphdr, ip) : NULL))
2662 != NULL) {
2663 /*
2664 * Found dynamic entry, update stats
2665 * and jump to the 'action' part of
2666 * the parent rule.
2667 */
2668 q->pcnt++;
2669 q->bcnt += pktlen;
2670 f = q->rule;
2671 cmd = ACTION_PTR(f);
2672 l = f->cmd_len - f->act_ofs;
2673 goto check_body;
2674 }
2675 /*
2676 * Dynamic entry not found. If CHECK_STATE,
2677 * skip to next rule, if PROBE_STATE just
2678 * ignore and continue with next opcode.
2679 */
2680 if (cmd->opcode == O_CHECK_STATE)
2681 goto next_rule;
2682 match = 1;
2683 break;
2684
2685 case O_ACCEPT:
2686 retval = 0; /* accept */
2687 goto done;
2688
2689 case O_PIPE:
2690 case O_QUEUE:
2691 args->fwa_ipfw_rule = f; /* report matching rule */
2692 retval = cmd->arg1 | IP_FW_PORT_DYNT_FLAG;
2693 goto done;
2694
2695 case O_DIVERT:
2696 case O_TEE:
2697 if (args->fwa_eh) /* not on layer 2 */
2698 break;
2699 args->fwa_divert_rule = f->rulenum;
2700 retval = (cmd->opcode == O_DIVERT) ?
2701 cmd->arg1 :
2702 cmd->arg1 | IP_FW_PORT_TEE_FLAG;
2703 goto done;
2704
2705 case O_COUNT:
2706 case O_SKIPTO:
2707 f->pcnt++; /* update stats */
2708 f->bcnt += pktlen;
2709 f->timestamp = timenow.tv_sec;
2710 if (cmd->opcode == O_COUNT)
2711 goto next_rule;
2712 /* handle skipto */
2713 if (f->next_rule == NULL)
2714 lookup_next_rule(f);
2715 f = f->next_rule;
2716 goto again;
2717
2718 case O_REJECT:
2719 /*
2720 * Drop the packet and send a reject notice
2721 * if the packet is not ICMP (or is an ICMP
2722 * query), and it is not multicast/broadcast.
2723 */
2724 if (hlen > 0 && offset == 0 &&
2725 (proto != IPPROTO_ICMP ||
2726 is_icmp_query(ip)) &&
2727 !(m->m_flags & (M_BCAST|M_MCAST)) &&
2728 !IN_MULTICAST(dst_ip.s_addr)) {
2729 send_reject(args, cmd->arg1,
2730 offset,ip_len);
2731 m = args->fwa_m;
2732 }
2733 /* FALLTHROUGH */
2734 case O_DENY:
2735 retval = IP_FW_PORT_DENY_FLAG;
2736 goto done;
2737
2738 case O_FORWARD_IP:
2739 if (args->fwa_eh) /* not valid on layer2 pkts */
2740 break;
2741 if (!q || dyn_dir == MATCH_FORWARD)
2742 args->fwa_next_hop =
2743 &((ipfw_insn_sa *)cmd)->sa;
2744 retval = 0;
2745 goto done;
2746
2747 default:
2748 panic("-- unknown opcode %d\n", cmd->opcode);
2749 } /* end of switch() on opcodes */
2750
2751 if (cmd->len & F_NOT)
2752 match = !match;
2753
2754 if (match) {
2755 if (cmd->len & F_OR)
2756 skip_or = 1;
2757 } else {
2758 if (!(cmd->len & F_OR)) /* not an OR block, */
2759 break; /* try next rule */
2760 }
2761
2762 } /* end of inner for, scan opcodes */
2763
2764 next_rule:; /* try next rule */
2765
2766 } /* end of outer for, scan rules */
2767 printf("ipfw: ouch!, skip past end of rules, denying packet\n");
2768 lck_mtx_unlock(ipfw_mutex);
2769 return(IP_FW_PORT_DENY_FLAG);
2770
2771 done:
2772 /* Update statistics */
2773 f->pcnt++;
2774 f->bcnt += pktlen;
2775 f->timestamp = timenow.tv_sec;
2776 lck_mtx_unlock(ipfw_mutex);
2777 return retval;
2778
2779 pullup_failed:
2780 if (fw_verbose)
2781 printf("ipfw: pullup failed\n");
2782 lck_mtx_unlock(ipfw_mutex);
2783 return(IP_FW_PORT_DENY_FLAG);
2784 }
2785
2786 /*
2787 * When a rule is added/deleted, clear the next_rule pointers in all rules.
2788 * These will be reconstructed on the fly as packets are matched.
2789 * Must be called at splimp().
2790 */
2791 static void
2792 flush_rule_ptrs(void)
2793 {
2794 struct ip_fw *rule;
2795
2796 for (rule = layer3_chain; rule; rule = rule->next)
2797 rule->next_rule = NULL;
2798 }
2799
2800 /*
2801 * When pipes/queues are deleted, clear the "pipe_ptr" pointer to a given
2802 * pipe/queue, or to all of them (match == NULL).
2803 * Must be called at splimp().
2804 */
2805 void
2806 flush_pipe_ptrs(struct dn_flow_set *match)
2807 {
2808 struct ip_fw *rule;
2809
2810 for (rule = layer3_chain; rule; rule = rule->next) {
2811 ipfw_insn_pipe *cmd = (ipfw_insn_pipe *)ACTION_PTR(rule);
2812
2813 if (cmd->o.opcode != O_PIPE && cmd->o.opcode != O_QUEUE)
2814 continue;
2815 /*
2816 * XXX Use bcmp/bzero to handle pipe_ptr to overcome
2817 * possible alignment problems on 64-bit architectures.
2818 * This code is seldom used so we do not worry too
2819 * much about efficiency.
2820 */
2821 if (match == NULL ||
2822 !bcmp(&cmd->pipe_ptr, &match, sizeof(match)) )
2823 bzero(&cmd->pipe_ptr, sizeof(cmd->pipe_ptr));
2824 }
2825 }
2826
2827 /*
2828 * Add a new rule to the list. Copy the rule into a malloc'ed area, then
2829 * possibly create a rule number and add the rule to the list.
2830 * Update the rule_number in the input struct so the caller knows it as well.
2831 */
2832 static int
2833 add_rule(struct ip_fw **head, struct ip_fw *input_rule)
2834 {
2835 struct ip_fw *rule, *f, *prev;
2836 int l = RULESIZE(input_rule);
2837
2838 if (*head == NULL && input_rule->rulenum != IPFW_DEFAULT_RULE)
2839 return (EINVAL);
2840
2841 rule = _MALLOC(l, M_IPFW, M_WAIT);
2842 if (rule == NULL) {
2843 printf("ipfw2: add_rule MALLOC failed\n");
2844 return (ENOSPC);
2845 }
2846
2847 bzero(rule, l);
2848 bcopy(input_rule, rule, l);
2849
2850 rule->next = NULL;
2851 rule->next_rule = NULL;
2852
2853 rule->pcnt = 0;
2854 rule->bcnt = 0;
2855 rule->timestamp = 0;
2856
2857 if (*head == NULL) { /* default rule */
2858 *head = rule;
2859 goto done;
2860 }
2861
2862 /*
2863 * If rulenum is 0, find highest numbered rule before the
2864 * default rule, and add autoinc_step
2865 */
2866 if (autoinc_step < 1)
2867 autoinc_step = 1;
2868 else if (autoinc_step > 1000)
2869 autoinc_step = 1000;
2870 if (rule->rulenum == 0) {
2871 /*
2872 * locate the highest numbered rule before default
2873 */
2874 for (f = *head; f; f = f->next) {
2875 if (f->rulenum == IPFW_DEFAULT_RULE)
2876 break;
2877 rule->rulenum = f->rulenum;
2878 }
2879 if (rule->rulenum < IPFW_DEFAULT_RULE - autoinc_step)
2880 rule->rulenum += autoinc_step;
2881 input_rule->rulenum = rule->rulenum;
2882 }
2883
2884 /*
2885 * Now insert the new rule in the right place in the sorted list.
2886 */
2887 for (prev = NULL, f = *head; f; prev = f, f = f->next) {
2888 if (f->rulenum > rule->rulenum) { /* found the location */
2889 if (prev) {
2890 rule->next = f;
2891 prev->next = rule;
2892 } else { /* head insert */
2893 rule->next = *head;
2894 *head = rule;
2895 }
2896 break;
2897 }
2898 }
2899 flush_rule_ptrs();
2900 done:
2901 static_count++;
2902 static_len += l;
2903 static_len_32 += RULESIZE32(input_rule);
2904 static_len_64 += RULESIZE64(input_rule);
2905 DEB(printf("ipfw: installed rule %d, static count now %d\n",
2906 rule->rulenum, static_count);)
2907 return (0);
2908 }
2909
2910 /**
2911 * Free storage associated with a static rule (including derived
2912 * dynamic rules).
2913 * The caller is in charge of clearing rule pointers to avoid
2914 * dangling pointers.
2915 * @return a pointer to the next entry.
2916 * Arguments are not checked, so they better be correct.
2917 * Must be called at splimp().
2918 */
2919 static struct ip_fw *
2920 delete_rule(struct ip_fw **head, struct ip_fw *prev, struct ip_fw *rule)
2921 {
2922 struct ip_fw *n;
2923 int l = RULESIZE(rule);
2924
2925 n = rule->next;
2926 remove_dyn_rule(rule, NULL /* force removal */);
2927 if (prev == NULL)
2928 *head = n;
2929 else
2930 prev->next = n;
2931 static_count--;
2932 static_len -= l;
2933 static_len_32 -= RULESIZE32(rule);
2934 static_len_64 -= RULESIZE64(rule);
2935
2936 #if DUMMYNET
2937 if (DUMMYNET_LOADED)
2938 dn_ipfw_rule_delete(rule);
2939 #endif /* DUMMYNET */
2940 _FREE(rule, M_IPFW);
2941 return n;
2942 }
2943
2944 #if DEBUG_INACTIVE_RULES
2945 static void
2946 print_chain(struct ip_fw **chain)
2947 {
2948 struct ip_fw *rule = *chain;
2949
2950 for (; rule; rule = rule->next) {
2951 ipfw_insn *cmd = ACTION_PTR(rule);
2952
2953 printf("ipfw: rule->rulenum = %d\n", rule->rulenum);
2954
2955 if (rule->reserved_1 == IPFW_RULE_INACTIVE) {
2956 printf("ipfw: rule->reserved = IPFW_RULE_INACTIVE\n");
2957 }
2958
2959 switch (cmd->opcode) {
2960 case O_DENY:
2961 printf("ipfw: ACTION: Deny\n");
2962 break;
2963
2964 case O_REJECT:
2965 if (cmd->arg1==ICMP_REJECT_RST)
2966 printf("ipfw: ACTION: Reset\n");
2967 else if (cmd->arg1==ICMP_UNREACH_HOST)
2968 printf("ipfw: ACTION: Reject\n");
2969 break;
2970
2971 case O_ACCEPT:
2972 printf("ipfw: ACTION: Accept\n");
2973 break;
2974 case O_COUNT:
2975 printf("ipfw: ACTION: Count\n");
2976 break;
2977 case O_DIVERT:
2978 printf("ipfw: ACTION: Divert\n");
2979 break;
2980 case O_TEE:
2981 printf("ipfw: ACTION: Tee\n");
2982 break;
2983 case O_SKIPTO:
2984 printf("ipfw: ACTION: SkipTo\n");
2985 break;
2986 case O_PIPE:
2987 printf("ipfw: ACTION: Pipe\n");
2988 break;
2989 case O_QUEUE:
2990 printf("ipfw: ACTION: Queue\n");
2991 break;
2992 case O_FORWARD_IP:
2993 printf("ipfw: ACTION: Forward\n");
2994 break;
2995 default:
2996 printf("ipfw: invalid action! %d\n", cmd->opcode);
2997 }
2998 }
2999 }
3000 #endif /* DEBUG_INACTIVE_RULES */
3001
3002 static void
3003 flush_inactive(void *param)
3004 {
3005 struct ip_fw *inactive_rule = (struct ip_fw *)param;
3006 struct ip_fw *rule, *prev;
3007
3008 lck_mtx_lock(ipfw_mutex);
3009
3010 for (rule = layer3_chain, prev = NULL; rule; ) {
3011 if (rule == inactive_rule && rule->reserved_1 == IPFW_RULE_INACTIVE) {
3012 struct ip_fw *n = rule;
3013
3014 if (prev == NULL) {
3015 layer3_chain = rule->next;
3016 }
3017 else {
3018 prev->next = rule->next;
3019 }
3020 rule = rule->next;
3021 _FREE(n, M_IPFW);
3022 }
3023 else {
3024 prev = rule;
3025 rule = rule->next;
3026 }
3027 }
3028
3029 #if DEBUG_INACTIVE_RULES
3030 print_chain(&layer3_chain);
3031 #endif
3032 lck_mtx_unlock(ipfw_mutex);
3033 }
3034
3035 static void
3036 mark_inactive(struct ip_fw **prev, struct ip_fw **rule)
3037 {
3038 int l = RULESIZE(*rule);
3039
3040 if ((*rule)->reserved_1 != IPFW_RULE_INACTIVE) {
3041 (*rule)->reserved_1 = IPFW_RULE_INACTIVE;
3042 static_count--;
3043 static_len -= l;
3044 static_len_32 -= RULESIZE32(*rule);
3045 static_len_64 -= RULESIZE64(*rule);
3046
3047 timeout(flush_inactive, *rule, 30*hz); /* 30 sec. */
3048 }
3049
3050 *prev = *rule;
3051 *rule = (*rule)->next;
3052 }
3053
3054 /*
3055 * Deletes all rules from a chain (except rules in set RESVD_SET
3056 * unless kill_default = 1).
3057 * Must be called at splimp().
3058 */
3059 static void
3060 free_chain(struct ip_fw **chain, int kill_default)
3061 {
3062 struct ip_fw *prev, *rule;
3063
3064 flush_rule_ptrs(); /* more efficient to do outside the loop */
3065 for (prev = NULL, rule = *chain; rule ; )
3066 if (kill_default || rule->set != RESVD_SET) {
3067 ipfw_insn *cmd = ACTION_PTR(rule);
3068
3069 /* skip over forwarding rules so struct isn't
3070 * deleted while pointer is still in use elsewhere
3071 */
3072 if (cmd->opcode == O_FORWARD_IP) {
3073 mark_inactive(&prev, &rule);
3074 }
3075 else {
3076 rule = delete_rule(chain, prev, rule);
3077 }
3078 }
3079 else {
3080 prev = rule;
3081 rule = rule->next;
3082 }
3083 }
3084
3085 /**
3086 * Remove all rules with given number, and also do set manipulation.
3087 * Assumes chain != NULL && *chain != NULL.
3088 *
3089 * The argument is an u_int32_t. The low 16 bit are the rule or set number,
3090 * the next 8 bits are the new set, the top 8 bits are the command:
3091 *
3092 * 0 delete rules with given number
3093 * 1 delete rules with given set number
3094 * 2 move rules with given number to new set
3095 * 3 move rules with given set number to new set
3096 * 4 swap sets with given numbers
3097 */
3098 static int
3099 del_entry(struct ip_fw **chain, u_int32_t arg)
3100 {
3101 struct ip_fw *prev = NULL, *rule = *chain;
3102 u_int16_t rulenum; /* rule or old_set */
3103 u_int8_t cmd, new_set;
3104
3105 rulenum = arg & 0xffff;
3106 cmd = (arg >> 24) & 0xff;
3107 new_set = (arg >> 16) & 0xff;
3108
3109 if (cmd > 4)
3110 return EINVAL;
3111 if (new_set > RESVD_SET)
3112 return EINVAL;
3113 if (cmd == 0 || cmd == 2) {
3114 if (rulenum >= IPFW_DEFAULT_RULE)
3115 return EINVAL;
3116 } else {
3117 if (rulenum > RESVD_SET) /* old_set */
3118 return EINVAL;
3119 }
3120
3121 switch (cmd) {
3122 case 0: /* delete rules with given number */
3123 /*
3124 * locate first rule to delete
3125 */
3126 for (; rule->rulenum < rulenum; prev = rule, rule = rule->next)
3127 ;
3128 if (rule->rulenum != rulenum)
3129 return EINVAL;
3130
3131 /*
3132 * flush pointers outside the loop, then delete all matching
3133 * rules. prev remains the same throughout the cycle.
3134 */
3135 flush_rule_ptrs();
3136 while (rule->rulenum == rulenum) {
3137 ipfw_insn *insn = ACTION_PTR(rule);
3138
3139 /* keep forwarding rules around so struct isn't
3140 * deleted while pointer is still in use elsewhere
3141 */
3142 if (insn->opcode == O_FORWARD_IP) {
3143 mark_inactive(&prev, &rule);
3144 }
3145 else {
3146 rule = delete_rule(chain, prev, rule);
3147 }
3148 }
3149 break;
3150
3151 case 1: /* delete all rules with given set number */
3152 flush_rule_ptrs();
3153 while (rule->rulenum < IPFW_DEFAULT_RULE) {
3154 if (rule->set == rulenum) {
3155 ipfw_insn *insn = ACTION_PTR(rule);
3156
3157 /* keep forwarding rules around so struct isn't
3158 * deleted while pointer is still in use elsewhere
3159 */
3160 if (insn->opcode == O_FORWARD_IP) {
3161 mark_inactive(&prev, &rule);
3162 }
3163 else {
3164 rule = delete_rule(chain, prev, rule);
3165 }
3166 }
3167 else {
3168 prev = rule;
3169 rule = rule->next;
3170 }
3171 }
3172 break;
3173
3174 case 2: /* move rules with given number to new set */
3175 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
3176 if (rule->rulenum == rulenum)
3177 rule->set = new_set;
3178 break;
3179
3180 case 3: /* move rules with given set number to new set */
3181 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
3182 if (rule->set == rulenum)
3183 rule->set = new_set;
3184 break;
3185
3186 case 4: /* swap two sets */
3187 for (; rule->rulenum < IPFW_DEFAULT_RULE; rule = rule->next)
3188 if (rule->set == rulenum)
3189 rule->set = new_set;
3190 else if (rule->set == new_set)
3191 rule->set = rulenum;
3192 break;
3193 }
3194 return 0;
3195 }
3196
3197 /*
3198 * Clear counters for a specific rule.
3199 */
3200 static void
3201 clear_counters(struct ip_fw *rule, int log_only)
3202 {
3203 ipfw_insn_log *l = (ipfw_insn_log *)ACTION_PTR(rule);
3204
3205 if (log_only == 0) {
3206 rule->bcnt = rule->pcnt = 0;
3207 rule->timestamp = 0;
3208 }
3209 if (l->o.opcode == O_LOG)
3210 l->log_left = l->max_log;
3211 }
3212
3213 /**
3214 * Reset some or all counters on firewall rules.
3215 * @arg frwl is null to clear all entries, or contains a specific
3216 * rule number.
3217 * @arg log_only is 1 if we only want to reset logs, zero otherwise.
3218 */
3219 static int
3220 zero_entry(int rulenum, int log_only)
3221 {
3222 struct ip_fw *rule;
3223 const char *msg;
3224
3225 if (rulenum == 0) {
3226 norule_counter = 0;
3227 for (rule = layer3_chain; rule; rule = rule->next)
3228 clear_counters(rule, log_only);
3229 msg = log_only ? "ipfw: All logging counts reset.\n" :
3230 "ipfw: Accounting cleared.\n";
3231 } else {
3232 int cleared = 0;
3233 /*
3234 * We can have multiple rules with the same number, so we
3235 * need to clear them all.
3236 */
3237 for (rule = layer3_chain; rule; rule = rule->next)
3238 if (rule->rulenum == rulenum) {
3239 while (rule && rule->rulenum == rulenum) {
3240 clear_counters(rule, log_only);
3241 rule = rule->next;
3242 }
3243 cleared = 1;
3244 break;
3245 }
3246 if (!cleared) /* we did not find any matching rules */
3247 return (EINVAL);
3248 msg = log_only ? "ipfw: Entry %d logging count reset.\n" :
3249 "ipfw: Entry %d cleared.\n";
3250 }
3251 if (fw_verbose)
3252 {
3253 dolog((LOG_AUTHPRIV | LOG_NOTICE, msg, rulenum));
3254 }
3255 return (0);
3256 }
3257
3258 /*
3259 * Check validity of the structure before insert.
3260 * Fortunately rules are simple, so this mostly need to check rule sizes.
3261 */
3262 static int
3263 check_ipfw_struct(struct ip_fw *rule, int size)
3264 {
3265 int l, cmdlen = 0;
3266 int have_action=0;
3267 ipfw_insn *cmd;
3268
3269 if (size < sizeof(*rule)) {
3270 printf("ipfw: rule too short\n");
3271 return (EINVAL);
3272 }
3273 /* first, check for valid size */
3274 l = RULESIZE(rule);
3275 if (l != size) {
3276 printf("ipfw: size mismatch (have %d want %d)\n", size, l);
3277 return (EINVAL);
3278 }
3279 /*
3280 * Now go for the individual checks. Very simple ones, basically only
3281 * instruction sizes.
3282 */
3283 for (l = rule->cmd_len, cmd = rule->cmd ;
3284 l > 0 ; l -= cmdlen, cmd += cmdlen) {
3285 cmdlen = F_LEN(cmd);
3286 if (cmdlen > l) {
3287 printf("ipfw: opcode %d size truncated\n",
3288 cmd->opcode);
3289 return EINVAL;
3290 }
3291 DEB(printf("ipfw: opcode %d\n", cmd->opcode);)
3292 switch (cmd->opcode) {
3293 case O_PROBE_STATE:
3294 case O_KEEP_STATE:
3295 case O_PROTO:
3296 case O_IP_SRC_ME:
3297 case O_IP_DST_ME:
3298 case O_LAYER2:
3299 case O_IN:
3300 case O_FRAG:
3301 case O_IPOPT:
3302 case O_IPTOS:
3303 case O_IPPRECEDENCE:
3304 case O_IPVER:
3305 case O_TCPWIN:
3306 case O_TCPFLAGS:
3307 case O_TCPOPTS:
3308 case O_ESTAB:
3309 case O_VERREVPATH:
3310 case O_IPSEC:
3311 if (cmdlen != F_INSN_SIZE(ipfw_insn))
3312 goto bad_size;
3313 break;
3314 case O_UID:
3315 #ifndef __APPLE__
3316 case O_GID:
3317 #endif /* __APPLE__ */
3318 case O_IP_SRC:
3319 case O_IP_DST:
3320 case O_TCPSEQ:
3321 case O_TCPACK:
3322 case O_PROB:
3323 case O_ICMPTYPE:
3324 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32))
3325 goto bad_size;
3326 break;
3327
3328 case O_LIMIT:
3329 if (cmdlen != F_INSN_SIZE(ipfw_insn_limit))
3330 goto bad_size;
3331 break;
3332
3333 case O_LOG:
3334 if (cmdlen != F_INSN_SIZE(ipfw_insn_log))
3335 goto bad_size;
3336
3337 /* enforce logging limit */
3338 if (fw_verbose &&
3339 ((ipfw_insn_log *)cmd)->max_log == 0 && verbose_limit != 0) {
3340 ((ipfw_insn_log *)cmd)->max_log = verbose_limit;
3341 }
3342
3343 ((ipfw_insn_log *)cmd)->log_left =
3344 ((ipfw_insn_log *)cmd)->max_log;
3345
3346 break;
3347
3348 case O_IP_SRC_MASK:
3349 case O_IP_DST_MASK:
3350 /* only odd command lengths */
3351 if ( !(cmdlen & 1) || cmdlen > 31)
3352 goto bad_size;
3353 break;
3354
3355 case O_IP_SRC_SET:
3356 case O_IP_DST_SET:
3357 if (cmd->arg1 == 0 || cmd->arg1 > 256) {
3358 printf("ipfw: invalid set size %d\n",
3359 cmd->arg1);
3360 return EINVAL;
3361 }
3362 if (cmdlen != F_INSN_SIZE(ipfw_insn_u32) +
3363 (cmd->arg1+31)/32 )
3364 goto bad_size;
3365 break;
3366
3367 case O_MACADDR2:
3368 if (cmdlen != F_INSN_SIZE(ipfw_insn_mac))
3369 goto bad_size;
3370 break;
3371
3372 case O_NOP:
3373 case O_IPID:
3374 case O_IPTTL:
3375 case O_IPLEN:
3376 if (cmdlen < 1 || cmdlen > 31)
3377 goto bad_size;
3378 break;
3379
3380 case O_MAC_TYPE:
3381 case O_IP_SRCPORT:
3382 case O_IP_DSTPORT: /* XXX artificial limit, 30 port pairs */
3383 if (cmdlen < 2 || cmdlen > 31)
3384 goto bad_size;
3385 break;
3386
3387 case O_RECV:
3388 case O_XMIT:
3389 case O_VIA:
3390 if (cmdlen != F_INSN_SIZE(ipfw_insn_if))
3391 goto bad_size;
3392 break;
3393
3394 case O_PIPE:
3395 case O_QUEUE:
3396 if (cmdlen != F_INSN_SIZE(ipfw_insn_pipe))
3397 goto bad_size;
3398 goto check_action;
3399
3400 case O_FORWARD_IP:
3401 if (cmdlen != F_INSN_SIZE(ipfw_insn_sa))
3402 goto bad_size;
3403 goto check_action;
3404
3405 case O_FORWARD_MAC: /* XXX not implemented yet */
3406 case O_CHECK_STATE:
3407 case O_COUNT:
3408 case O_ACCEPT:
3409 case O_DENY:
3410 case O_REJECT:
3411 case O_SKIPTO:
3412 case O_DIVERT:
3413 case O_TEE:
3414 if (cmdlen != F_INSN_SIZE(ipfw_insn))
3415 goto bad_size;
3416 check_action:
3417 if (have_action) {
3418 printf("ipfw: opcode %d, multiple actions"
3419 " not allowed\n",
3420 cmd->opcode);
3421 return EINVAL;
3422 }
3423 have_action = 1;
3424 if (l != cmdlen) {
3425 printf("ipfw: opcode %d, action must be"
3426 " last opcode\n",
3427 cmd->opcode);
3428 return EINVAL;
3429 }
3430 break;
3431 default:
3432 printf("ipfw: opcode %d, unknown opcode\n",
3433 cmd->opcode);
3434 return EINVAL;
3435 }
3436 }
3437 if (have_action == 0) {
3438 printf("ipfw: missing action\n");
3439 return EINVAL;
3440 }
3441 return 0;
3442
3443 bad_size:
3444 printf("ipfw: opcode %d size %d wrong\n",
3445 cmd->opcode, cmdlen);
3446 return EINVAL;
3447 }
3448
3449
3450 static void
3451 ipfw_kev_post_msg(u_int32_t event_code)
3452 {
3453 struct kev_msg ev_msg;
3454
3455 bzero(&ev_msg, sizeof(struct kev_msg));
3456
3457 ev_msg.vendor_code = KEV_VENDOR_APPLE;
3458 ev_msg.kev_class = KEV_FIREWALL_CLASS;
3459 ev_msg.kev_subclass = KEV_IPFW_SUBCLASS;
3460 ev_msg.event_code = event_code;
3461
3462 kev_post_msg(&ev_msg);
3463
3464 }
3465
3466 /**
3467 * {set|get}sockopt parser.
3468 */
3469 static int
3470 ipfw_ctl(struct sockopt *sopt)
3471 {
3472 #define RULE_MAXSIZE (256*sizeof(u_int32_t))
3473 u_int32_t api_version;
3474 int command;
3475 int error;
3476 size_t size;
3477 size_t rulesize = RULE_MAXSIZE;
3478 struct ip_fw *bp , *buf, *rule;
3479 int is64user = 0;
3480
3481 /* copy of orig sopt to send to ipfw_get_command_and_version() */
3482 struct sockopt tmp_sopt = *sopt;
3483 struct timeval timenow;
3484
3485 getmicrotime(&timenow);
3486
3487 /*
3488 * Disallow modifications in really-really secure mode, but still allow
3489 * the logging counters to be reset.
3490 */
3491 if (sopt->sopt_name == IP_FW_ADD ||
3492 (sopt->sopt_dir == SOPT_SET && sopt->sopt_name != IP_FW_RESETLOG)) {
3493 #if __FreeBSD_version >= 500034
3494 error = securelevel_ge(sopt->sopt_td->td_ucred, 3);
3495 if (error)
3496 return (error);
3497 #else /* FreeBSD 4.x */
3498 if (securelevel >= 3)
3499 return (EPERM);
3500 #endif
3501 }
3502
3503 /* first get the command and version, then do conversion as necessary */
3504 error = ipfw_get_command_and_version(&tmp_sopt, &command, &api_version);
3505 if (error) {
3506 /* error getting the version */
3507 return error;
3508 }
3509
3510 if (proc_is64bit(sopt->sopt_p))
3511 is64user = 1;
3512
3513 switch (command) {
3514 case IP_FW_GET:
3515 {
3516 size_t dynrulesize;
3517 /*
3518 * pass up a copy of the current rules. Static rules
3519 * come first (the last of which has number IPFW_DEFAULT_RULE),
3520 * followed by a possibly empty list of dynamic rule.
3521 * The last dynamic rule has NULL in the "next" field.
3522 */
3523 lck_mtx_lock(ipfw_mutex);
3524
3525 if (is64user){
3526 size = Get64static_len();
3527 dynrulesize = sizeof(ipfw_dyn_rule_64);
3528 if (ipfw_dyn_v)
3529 size += (dyn_count * dynrulesize);
3530 }else {
3531 size = Get32static_len();
3532 dynrulesize = sizeof(ipfw_dyn_rule_32);
3533 if (ipfw_dyn_v)
3534 size += (dyn_count * dynrulesize);
3535 }
3536
3537 /*
3538 * XXX todo: if the user passes a short length just to know
3539 * how much room is needed, do not bother filling up the
3540 * buffer, just jump to the sooptcopyout.
3541 */
3542 buf = _MALLOC(size, M_TEMP, M_WAITOK);
3543 if (buf == 0) {
3544 lck_mtx_unlock(ipfw_mutex);
3545 error = ENOBUFS;
3546 break;
3547 }
3548
3549 bzero(buf, size);
3550
3551 bp = buf;
3552 for (rule = layer3_chain; rule ; rule = rule->next) {
3553
3554 if (rule->reserved_1 == IPFW_RULE_INACTIVE) {
3555 continue;
3556 }
3557
3558 if (is64user){
3559 int rulesize_64;
3560
3561 copyto64fw( rule, (struct ip_fw_64 *)bp, size);
3562 bcopy(&set_disable, &(( (struct ip_fw_64*)bp)->next_rule), sizeof(set_disable));
3563 /* do not use macro RULESIZE64 since we want RULESIZE for ip_fw_64 */
3564 rulesize_64 = sizeof(struct ip_fw_64) + ((struct ip_fw_64 *)(bp))->cmd_len * 4 - 4;
3565 bp = (struct ip_fw *)((char *)bp + rulesize_64);
3566 }else{
3567 int rulesize_32;
3568
3569 copyto32fw( rule, (struct ip_fw_32*)bp, size);
3570 bcopy(&set_disable, &(( (struct ip_fw_32*)bp)->next_rule), sizeof(set_disable));
3571 /* do not use macro RULESIZE32 since we want RULESIZE for ip_fw_32 */
3572 rulesize_32 = sizeof(struct ip_fw_32) + ((struct ip_fw_32 *)(bp))->cmd_len * 4 - 4;
3573 bp = (struct ip_fw *)((char *)bp + rulesize_32);
3574 }
3575 }
3576 if (ipfw_dyn_v) {
3577 int i;
3578 ipfw_dyn_rule *p;
3579 char *dst, *last = NULL;
3580
3581 dst = (char *)bp;
3582 for (i = 0 ; i < curr_dyn_buckets ; i++ )
3583 for ( p = ipfw_dyn_v[i] ; p != NULL ;
3584 p = p->next, dst += dynrulesize ) {
3585 if ( is64user ){
3586 ipfw_dyn_rule_64 *ipfw_dyn_dst;
3587
3588 ipfw_dyn_dst = (ipfw_dyn_rule_64 *)dst;
3589 /*
3590 * store a non-null value in "next".
3591 * The userland code will interpret a
3592 * NULL here as a marker
3593 * for the last dynamic rule.
3594 */
3595 ipfw_dyn_dst->next = CAST_DOWN_EXPLICIT(user64_addr_t, dst);
3596 ipfw_dyn_dst->rule = p->rule->rulenum;
3597 ipfw_dyn_dst->parent = CAST_DOWN(user64_addr_t, p->parent);
3598 ipfw_dyn_dst->pcnt = p->pcnt;
3599 ipfw_dyn_dst->bcnt = p->bcnt;
3600 externalize_flow_id(&ipfw_dyn_dst->id, &p->id);
3601 ipfw_dyn_dst->expire =
3602 TIME_LEQ(p->expire, timenow.tv_sec) ?
3603 0 : p->expire - timenow.tv_sec;
3604 ipfw_dyn_dst->bucket = p->bucket;
3605 ipfw_dyn_dst->state = p->state;
3606 ipfw_dyn_dst->ack_fwd = p->ack_fwd;
3607 ipfw_dyn_dst->ack_rev = p->ack_rev;
3608 ipfw_dyn_dst->dyn_type = p->dyn_type;
3609 ipfw_dyn_dst->count = p->count;
3610 last = (char*)&ipfw_dyn_dst->next;
3611 } else {
3612 ipfw_dyn_rule_32 *ipfw_dyn_dst;
3613
3614 ipfw_dyn_dst = (ipfw_dyn_rule_32 *)dst;
3615 /*
3616 * store a non-null value in "next".
3617 * The userland code will interpret a
3618 * NULL here as a marker
3619 * for the last dynamic rule.
3620 */
3621 ipfw_dyn_dst->next = CAST_DOWN_EXPLICIT(user32_addr_t, dst);
3622 ipfw_dyn_dst->rule = p->rule->rulenum;
3623 ipfw_dyn_dst->parent = CAST_DOWN_EXPLICIT(user32_addr_t, p->parent);
3624 ipfw_dyn_dst->pcnt = p->pcnt;
3625 ipfw_dyn_dst->bcnt = p->bcnt;
3626 externalize_flow_id(&ipfw_dyn_dst->id, &p->id);
3627 ipfw_dyn_dst->expire =
3628 TIME_LEQ(p->expire, timenow.tv_sec) ?
3629 0 : p->expire - timenow.tv_sec;
3630 ipfw_dyn_dst->bucket = p->bucket;
3631 ipfw_dyn_dst->state = p->state;
3632 ipfw_dyn_dst->ack_fwd = p->ack_fwd;
3633 ipfw_dyn_dst->ack_rev = p->ack_rev;
3634 ipfw_dyn_dst->dyn_type = p->dyn_type;
3635 ipfw_dyn_dst->count = p->count;
3636 last = (char*)&ipfw_dyn_dst->next;
3637 }
3638 }
3639 if (last != NULL) /* mark last dynamic rule */
3640 bzero(last, sizeof(last));
3641 }
3642 lck_mtx_unlock(ipfw_mutex);
3643
3644 /* convert back if necessary and copyout */
3645 if (api_version == IP_FW_VERSION_0) {
3646 int i, len = 0;
3647 struct ip_old_fw *buf2, *rule_vers0;
3648
3649 lck_mtx_lock(ipfw_mutex);
3650 buf2 = _MALLOC(static_count * sizeof(struct ip_old_fw), M_TEMP, M_WAITOK);
3651 if (buf2 == 0) {
3652 lck_mtx_unlock(ipfw_mutex);
3653 error = ENOBUFS;
3654 }
3655
3656 if (!error) {
3657 bp = buf;
3658 rule_vers0 = buf2;
3659
3660 for (i = 0; i < static_count; i++) {
3661 /* static rules have different sizes */
3662 int j = RULESIZE(bp);
3663 ipfw_convert_from_latest(bp, rule_vers0, api_version, is64user);
3664 bp = (struct ip_fw *)((char *)bp + j);
3665 len += sizeof(*rule_vers0);
3666 rule_vers0++;
3667 }
3668 lck_mtx_unlock(ipfw_mutex);
3669 error = sooptcopyout(sopt, buf2, len);
3670 _FREE(buf2, M_TEMP);
3671 }
3672 } else if (api_version == IP_FW_VERSION_1) {
3673 int i, len = 0, buf_size;
3674 struct ip_fw_compat *buf2;
3675 size_t ipfwcompsize;
3676 size_t ipfwdyncompsize;
3677 char *rule_vers1;
3678
3679 lck_mtx_lock(ipfw_mutex);
3680 if ( is64user ){
3681 ipfwcompsize = sizeof(struct ip_fw_compat_64);
3682 ipfwdyncompsize = sizeof(struct ipfw_dyn_rule_compat_64);
3683 } else {
3684 ipfwcompsize = sizeof(struct ip_fw_compat_32);
3685 ipfwdyncompsize = sizeof(struct ipfw_dyn_rule_compat_32);
3686 }
3687
3688 buf_size = static_count * ipfwcompsize +
3689 dyn_count * ipfwdyncompsize;
3690
3691 buf2 = _MALLOC(buf_size, M_TEMP, M_WAITOK);
3692 if (buf2 == 0) {
3693 lck_mtx_unlock(ipfw_mutex);
3694 error = ENOBUFS;
3695 }
3696 if (!error) {
3697 bp = buf;
3698 rule_vers1 = (char*)buf2;
3699
3700 /* first do static rules */
3701 for (i = 0; i < static_count; i++) {
3702 /* static rules have different sizes */
3703 if ( is64user ){
3704 int rulesize_64;
3705 ipfw_convert_from_latest(bp, (void *)rule_vers1, api_version, is64user);
3706 rulesize_64 = sizeof(struct ip_fw_64) + ((struct ip_fw_64 *)(bp))->cmd_len * 4 - 4;
3707 bp = (struct ip_fw *)((char *)bp + rulesize_64);
3708 }else {
3709 int rulesize_32;
3710 ipfw_convert_from_latest(bp, (void *)rule_vers1, api_version, is64user);
3711 rulesize_32 = sizeof(struct ip_fw_32) + ((struct ip_fw_32 *)(bp))->cmd_len * 4 - 4;
3712 bp = (struct ip_fw *)((char *)bp + rulesize_32);
3713 }
3714 len += ipfwcompsize;
3715 rule_vers1 += ipfwcompsize;
3716 }
3717 /* now do dynamic rules */
3718 if ( is64user )
3719 cp_dyn_to_comp_64( (struct ipfw_dyn_rule_compat_64 *)rule_vers1, &len);
3720 else
3721 cp_dyn_to_comp_32( (struct ipfw_dyn_rule_compat_32 *)rule_vers1, &len);
3722
3723 lck_mtx_unlock(ipfw_mutex);
3724 error = sooptcopyout(sopt, buf2, len);
3725 _FREE(buf2, M_TEMP);
3726 }
3727 } else {
3728 error = sooptcopyout(sopt, buf, size);
3729 }
3730
3731 _FREE(buf, M_TEMP);
3732 break;
3733 }
3734
3735 case IP_FW_FLUSH:
3736 /*
3737 * Normally we cannot release the lock on each iteration.
3738 * We could do it here only because we start from the head all
3739 * the times so there is no risk of missing some entries.
3740 * On the other hand, the risk is that we end up with
3741 * a very inconsistent ruleset, so better keep the lock
3742 * around the whole cycle.
3743 *
3744 * XXX this code can be improved by resetting the head of
3745 * the list to point to the default rule, and then freeing
3746 * the old list without the need for a lock.
3747 */
3748
3749 lck_mtx_lock(ipfw_mutex);
3750 free_chain(&layer3_chain, 0 /* keep default rule */);
3751 fw_bypass = 1;
3752 #if DEBUG_INACTIVE_RULES
3753 print_chain(&layer3_chain);
3754 #endif
3755 lck_mtx_unlock(ipfw_mutex);
3756 break;
3757
3758 case IP_FW_ADD:
3759 {
3760 size_t savedsopt_valsize=0;
3761 rule = _MALLOC(RULE_MAXSIZE, M_TEMP, M_WAITOK);
3762 if (rule == 0) {
3763 error = ENOBUFS;
3764 break;
3765 }
3766
3767 bzero(rule, RULE_MAXSIZE);
3768
3769 if (api_version != IP_FW_CURRENT_API_VERSION) {
3770 error = ipfw_convert_to_latest(sopt, rule, api_version, is64user);
3771 }
3772 else {
3773 savedsopt_valsize = sopt->sopt_valsize; /* it might get modified in sooptcopyin_fw */
3774 error = sooptcopyin_fw( sopt, rule, &rulesize);
3775
3776 }
3777
3778 if (!error) {
3779 if ((api_version == IP_FW_VERSION_0) || (api_version == IP_FW_VERSION_1)) {
3780 /* the rule has already been checked so just
3781 * adjust sopt_valsize to match what would be expected.
3782 */
3783 sopt->sopt_valsize = RULESIZE(rule);
3784 rulesize = RULESIZE(rule);
3785 }
3786 error = check_ipfw_struct(rule, rulesize);
3787 if (!error) {
3788 lck_mtx_lock(ipfw_mutex);
3789 error = add_rule(&layer3_chain, rule);
3790 if (!error && fw_bypass)
3791 fw_bypass = 0;
3792 lck_mtx_unlock(ipfw_mutex);
3793
3794 size = RULESIZE(rule);
3795 if (!error && sopt->sopt_dir == SOPT_GET) {
3796 /* convert back if necessary and copyout */
3797 if (api_version == IP_FW_VERSION_0) {
3798 struct ip_old_fw rule_vers0;
3799
3800 ipfw_convert_from_latest(rule, &rule_vers0, api_version, is64user);
3801 sopt->sopt_valsize = sizeof(struct ip_old_fw);
3802
3803 error = sooptcopyout(sopt, &rule_vers0, sizeof(struct ip_old_fw));
3804 } else if (api_version == IP_FW_VERSION_1) {
3805 struct ip_fw_compat rule_vers1;
3806 ipfw_convert_from_latest(rule, &rule_vers1, api_version, is64user);
3807 sopt->sopt_valsize = sizeof(struct ip_fw_compat);
3808
3809 error = sooptcopyout(sopt, &rule_vers1, sizeof(struct ip_fw_compat));
3810 } else {
3811 char *userrule;
3812 userrule = _MALLOC(savedsopt_valsize, M_TEMP, M_WAITOK);
3813 if ( userrule == NULL )
3814 userrule = (char*)rule;
3815 if (proc_is64bit(sopt->sopt_p)){
3816 copyto64fw( rule, (struct ip_fw_64*)userrule, savedsopt_valsize);
3817 }
3818 else {
3819 copyto32fw( rule, (struct ip_fw_32*)userrule, savedsopt_valsize);
3820 }
3821 error = sooptcopyout(sopt, userrule, savedsopt_valsize);
3822 if ( userrule )
3823 _FREE(userrule, M_TEMP);
3824 }
3825 }
3826 }
3827 }
3828
3829 _FREE(rule, M_TEMP);
3830 break;
3831 }
3832 case IP_FW_DEL:
3833 {
3834 /*
3835 * IP_FW_DEL is used for deleting single rules or sets,
3836 * and (ab)used to atomically manipulate sets.
3837 * rule->rulenum != 0 indicates single rule delete
3838 * rule->set_masks used to manipulate sets
3839 * rule->set_masks[0] contains info on sets to be
3840 * disabled, swapped, or moved
3841 * rule->set_masks[1] contains sets to be enabled.
3842 */
3843
3844 /* there is only a simple rule passed in
3845 * (no cmds), so use a temp struct to copy
3846 */
3847 struct ip_fw temp_rule;
3848 u_int32_t arg;
3849 u_int8_t cmd;
3850
3851 bzero(&temp_rule, sizeof(struct ip_fw));
3852 if (api_version != IP_FW_CURRENT_API_VERSION) {
3853 error = ipfw_convert_to_latest(sopt, &temp_rule, api_version, is64user);
3854 }
3855 else {
3856 error = sooptcopyin_fw(sopt, &temp_rule, 0 );
3857 }
3858
3859 if (!error) {
3860 /* set_masks is used to distinguish between deleting
3861 * single rules or atomically manipulating sets
3862 */
3863 lck_mtx_lock(ipfw_mutex);
3864
3865 arg = temp_rule.set_masks[0];
3866 cmd = (arg >> 24) & 0xff;
3867
3868 if (temp_rule.rulenum) {
3869 /* single rule */
3870 error = del_entry(&layer3_chain, temp_rule.rulenum);
3871 #if DEBUG_INACTIVE_RULES
3872 print_chain(&layer3_chain);
3873 #endif
3874 }
3875 else if (cmd) {
3876 /* set reassignment - see comment above del_entry() for details */
3877 error = del_entry(&layer3_chain, temp_rule.set_masks[0]);
3878 #if DEBUG_INACTIVE_RULES
3879 print_chain(&layer3_chain);
3880 #endif
3881 }
3882 else if (temp_rule.set_masks[0] != 0 ||
3883 temp_rule.set_masks[1] != 0) {
3884 /* set enable/disable */
3885 set_disable =
3886 (set_disable | temp_rule.set_masks[0]) & ~temp_rule.set_masks[1] &
3887 ~(1<<RESVD_SET); /* set RESVD_SET always enabled */
3888 }
3889
3890 if (!layer3_chain->next)
3891 fw_bypass = 1;
3892 lck_mtx_unlock(ipfw_mutex);
3893 }
3894 break;
3895 }
3896 case IP_FW_ZERO:
3897 case IP_FW_RESETLOG: /* using rule->rulenum */
3898 {
3899 /* there is only a simple rule passed in
3900 * (no cmds), so use a temp struct to copy
3901 */
3902 struct ip_fw temp_rule;
3903
3904 bzero(&temp_rule, sizeof(struct ip_fw));
3905
3906 if (api_version != IP_FW_CURRENT_API_VERSION) {
3907 error = ipfw_convert_to_latest(sopt, &temp_rule, api_version, is64user);
3908 }
3909 else {
3910 if (sopt->sopt_val != 0) {
3911 error = sooptcopyin_fw( sopt, &temp_rule, 0);
3912 }
3913 }
3914
3915 if (!error) {
3916 lck_mtx_lock(ipfw_mutex);
3917 error = zero_entry(temp_rule.rulenum, sopt->sopt_name == IP_FW_RESETLOG);
3918 lck_mtx_unlock(ipfw_mutex);
3919 }
3920 break;
3921 }
3922 default:
3923 printf("ipfw: ipfw_ctl invalid option %d\n", sopt->sopt_name);
3924 error = EINVAL;
3925 }
3926
3927 if (error != EINVAL) {
3928 switch (command) {
3929 case IP_FW_ADD:
3930 case IP_OLD_FW_ADD:
3931 ipfw_kev_post_msg(KEV_IPFW_ADD);
3932 break;
3933 case IP_OLD_FW_DEL:
3934 case IP_FW_DEL:
3935 ipfw_kev_post_msg(KEV_IPFW_DEL);
3936 break;
3937 case IP_FW_FLUSH:
3938 case IP_OLD_FW_FLUSH:
3939 ipfw_kev_post_msg(KEV_IPFW_FLUSH);
3940 break;
3941
3942 default:
3943 break;
3944 }
3945 }
3946
3947 return (error);
3948 }
3949
3950 /**
3951 * dummynet needs a reference to the default rule, because rules can be
3952 * deleted while packets hold a reference to them. When this happens,
3953 * dummynet changes the reference to the default rule (it could well be a
3954 * NULL pointer, but this way we do not need to check for the special
3955 * case, plus here he have info on the default behaviour).
3956 */
3957 struct ip_fw *ip_fw_default_rule;
3958
3959 /*
3960 * This procedure is only used to handle keepalives. It is invoked
3961 * every dyn_keepalive_period
3962 */
3963 static void
3964 ipfw_tick(__unused void * unused)
3965 {
3966 struct mbuf *m0, *m, *mnext, **mtailp;
3967 int i;
3968 ipfw_dyn_rule *q;
3969 struct timeval timenow;
3970 static int stealth_cnt = 0;
3971
3972 if (ipfw_stealth_stats_needs_flush) {
3973 stealth_cnt++;
3974 if (!(stealth_cnt % IPFW_STEALTH_TIMEOUT_FREQUENCY)) {
3975 ipfw_stealth_flush_stats();
3976 }
3977 }
3978
3979 if (dyn_keepalive == 0 || ipfw_dyn_v == NULL || dyn_count == 0)
3980 goto done;
3981
3982 getmicrotime(&timenow);
3983
3984 /*
3985 * We make a chain of packets to go out here -- not deferring
3986 * until after we drop the ipfw lock would result
3987 * in a lock order reversal with the normal packet input -> ipfw
3988 * call stack.
3989 */
3990 m0 = NULL;
3991 mtailp = &m0;
3992
3993 lck_mtx_lock(ipfw_mutex);
3994 for (i = 0 ; i < curr_dyn_buckets ; i++) {
3995 for (q = ipfw_dyn_v[i] ; q ; q = q->next ) {
3996 if (q->dyn_type == O_LIMIT_PARENT)
3997 continue;
3998 if (q->id.proto != IPPROTO_TCP)
3999 continue;
4000 if ( (q->state & BOTH_SYN) != BOTH_SYN)
4001 continue;
4002 if (TIME_LEQ( timenow.tv_sec+dyn_keepalive_interval,
4003 q->expire))
4004 continue; /* too early */
4005 if (TIME_LEQ(q->expire, timenow.tv_sec))
4006 continue; /* too late, rule expired */
4007
4008 *mtailp = send_pkt(&(q->id), q->ack_rev - 1, q->ack_fwd, TH_SYN);
4009 if (*mtailp != NULL)
4010 mtailp = &(*mtailp)->m_nextpkt;
4011
4012 *mtailp = send_pkt(&(q->id), q->ack_fwd - 1, q->ack_rev, 0);
4013 if (*mtailp != NULL)
4014 mtailp = &(*mtailp)->m_nextpkt;
4015 }
4016 }
4017 lck_mtx_unlock(ipfw_mutex);
4018
4019 for (m = mnext = m0; m != NULL; m = mnext) {
4020 struct route sro; /* fake route */
4021
4022 mnext = m->m_nextpkt;
4023 m->m_nextpkt = NULL;
4024 bzero (&sro, sizeof (sro));
4025 ip_output(m, NULL, &sro, 0, NULL, NULL);
4026 ROUTE_RELEASE(&sro);
4027 }
4028 done:
4029 timeout_with_leeway(ipfw_tick, NULL, dyn_keepalive_period*hz,
4030 DYN_KEEPALIVE_LEEWAY*hz);
4031 }
4032
4033 void
4034 ipfw_init(void)
4035 {
4036 struct ip_fw default_rule;
4037
4038 /* setup locks */
4039 ipfw_mutex_grp_attr = lck_grp_attr_alloc_init();
4040 ipfw_mutex_grp = lck_grp_alloc_init("ipfw", ipfw_mutex_grp_attr);
4041 ipfw_mutex_attr = lck_attr_alloc_init();
4042 lck_mtx_init(ipfw_mutex, ipfw_mutex_grp, ipfw_mutex_attr);
4043
4044 layer3_chain = NULL;
4045
4046 bzero(&default_rule, sizeof default_rule);
4047
4048 default_rule.act_ofs = 0;
4049 default_rule.rulenum = IPFW_DEFAULT_RULE;
4050 default_rule.cmd_len = 1;
4051 default_rule.set = RESVD_SET;
4052
4053 default_rule.cmd[0].len = 1;
4054 default_rule.cmd[0].opcode =
4055 #ifdef IPFIREWALL_DEFAULT_TO_ACCEPT
4056 (1) ? O_ACCEPT :
4057 #endif
4058 O_DENY;
4059
4060 if (add_rule(&layer3_chain, &default_rule)) {
4061 printf("ipfw2: add_rule failed adding default rule\n");
4062 printf("ipfw2 failed initialization!!\n");
4063 fw_enable = 0;
4064 }
4065 else {
4066 ip_fw_default_rule = layer3_chain;
4067
4068 #ifdef IPFIREWALL_VERBOSE
4069 fw_verbose = 1;
4070 #endif
4071 #ifdef IPFIREWALL_VERBOSE_LIMIT
4072 verbose_limit = IPFIREWALL_VERBOSE_LIMIT;
4073 #endif
4074 if (fw_verbose) {
4075 if (!verbose_limit)
4076 printf("ipfw2 verbose logging enabled: unlimited logging by default\n");
4077 else
4078 printf("ipfw2 verbose logging enabled: limited to %d packets/entry by default\n",
4079 verbose_limit);
4080 }
4081 }
4082
4083 ip_fw_chk_ptr = ipfw_chk;
4084 ip_fw_ctl_ptr = ipfw_ctl;
4085
4086 ipfwstringlen = strlen( ipfwstring );
4087
4088 timeout(ipfw_tick, NULL, hz);
4089 }
4090
4091 #endif /* IPFW2 */
4092