xref: /titanic_44/usr/src/uts/common/inet/ipf/ip_state.c (revision ee0eb9f2d12b8a03b5a193cd667735c1962e0c7b)
1 /*
2  * Copyright (C) 1995-2003 by Darren Reed.
3  *
4  * See the IPFILTER.LICENCE file for details on licencing.
5  *
6  * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
7  * Use is subject to license terms.
8  */
9 
10 #pragma ident	"%Z%%M%	%I%	%E% SMI"
11 
12 #if defined(KERNEL) || defined(_KERNEL)
13 # undef KERNEL
14 # undef _KERNEL
15 # define        KERNEL	1
16 # define        _KERNEL	1
17 #endif
18 #include <sys/errno.h>
19 #include <sys/types.h>
20 #include <sys/param.h>
21 #include <sys/file.h>
22 #if defined(__NetBSD__) && (NetBSD >= 199905) && !defined(IPFILTER_LKM) && \
23     defined(_KERNEL)
24 # include "opt_ipfilter_log.h"
25 #endif
26 #if defined(_KERNEL) && defined(__FreeBSD_version) && \
27     (__FreeBSD_version >= 400000) && !defined(KLD_MODULE)
28 #include "opt_inet6.h"
29 #endif
30 #if !defined(_KERNEL) && !defined(__KERNEL__)
31 # include <stdio.h>
32 # include <stdlib.h>
33 # include <string.h>
34 # define _KERNEL
35 # ifdef __OpenBSD__
36 struct file;
37 # endif
38 # include <sys/uio.h>
39 # undef _KERNEL
40 #endif
41 #if defined(_KERNEL) && (__FreeBSD_version >= 220000)
42 # include <sys/filio.h>
43 # include <sys/fcntl.h>
44 # if (__FreeBSD_version >= 300000) && !defined(IPFILTER_LKM)
45 #  include "opt_ipfilter.h"
46 # endif
47 #else
48 # include <sys/ioctl.h>
49 #endif
50 #include <sys/time.h>
51 #if !defined(linux)
52 # include <sys/protosw.h>
53 #endif
54 #include <sys/socket.h>
55 #if defined(_KERNEL)
56 # include <sys/systm.h>
57 # if !defined(__SVR4) && !defined(__svr4__)
58 #  include <sys/mbuf.h>
59 # endif
60 #endif
61 #if defined(__SVR4) || defined(__svr4__)
62 # include <sys/filio.h>
63 # include <sys/byteorder.h>
64 # ifdef _KERNEL
65 #  include <sys/dditypes.h>
66 # endif
67 # include <sys/stream.h>
68 # include <sys/kmem.h>
69 #endif
70 
71 #include <net/if.h>
72 #ifdef sun
73 # include <net/af.h>
74 #endif
75 #include <net/route.h>
76 #include <netinet/in.h>
77 #include <netinet/in_systm.h>
78 #include <netinet/ip.h>
79 #include <netinet/tcp.h>
80 #if !defined(linux)
81 # include <netinet/ip_var.h>
82 #endif
83 #if !defined(__hpux) && !defined(linux)
84 # include <netinet/tcp_fsm.h>
85 #endif
86 #include <netinet/udp.h>
87 #include <netinet/ip_icmp.h>
88 #include "netinet/ip_compat.h"
89 #include <netinet/tcpip.h>
90 #include "netinet/ip_fil.h"
91 #include "netinet/ip_nat.h"
92 #include "netinet/ip_frag.h"
93 #include "netinet/ip_state.h"
94 #include "netinet/ip_proxy.h"
95 #include "netinet/ipf_stack.h"
96 #ifdef	IPFILTER_SYNC
97 #include "netinet/ip_sync.h"
98 #endif
99 #ifdef	IPFILTER_SCAN
100 #include "netinet/ip_scan.h"
101 #endif
102 #ifdef	USE_INET6
103 #include <netinet/icmp6.h>
104 #endif
105 #if (__FreeBSD_version >= 300000)
106 # include <sys/malloc.h>
107 # if defined(_KERNEL) && !defined(IPFILTER_LKM)
108 #  include <sys/libkern.h>
109 #  include <sys/systm.h>
110 # endif
111 #endif
112 /* END OF INCLUDES */
113 
114 
115 #if !defined(lint)
116 static const char sccsid[] = "@(#)ip_state.c	1.8 6/5/96 (C) 1993-2000 Darren Reed";
117 static const char rcsid[] = "@(#)$Id: ip_state.c,v 2.186.2.36 2005/08/11 19:58:03 darrenr Exp $";
118 #endif
119 
120 #ifdef	USE_INET6
121 static ipstate_t *fr_checkicmp6matchingstate __P((fr_info_t *));
122 #endif
123 static ipstate_t *fr_matchsrcdst __P((fr_info_t *, ipstate_t *, i6addr_t *,
124 				      i6addr_t *, tcphdr_t *, u_32_t));
125 static ipstate_t *fr_checkicmpmatchingstate __P((fr_info_t *));
126 static int fr_state_flush __P((int, int, ipf_stack_t *));
127 static ips_stat_t *fr_statetstats __P((ipf_stack_t *));
128 static void fr_delstate __P((ipstate_t *, int, ipf_stack_t *));
129 static int fr_state_remove __P((caddr_t, ipf_stack_t *));
130 static void fr_ipsmove __P((ipstate_t *, u_int, ipf_stack_t *));
131 static int fr_tcpstate __P((fr_info_t *, tcphdr_t *, ipstate_t *));
132 static int fr_tcpoptions __P((fr_info_t *, tcphdr_t *, tcpdata_t *));
133 static ipstate_t *fr_stclone __P((fr_info_t *, tcphdr_t *, ipstate_t *));
134 static void fr_fixinisn __P((fr_info_t *, ipstate_t *));
135 static void fr_fixoutisn __P((fr_info_t *, ipstate_t *));
136 static void fr_checknewisn __P((fr_info_t *, ipstate_t *));
137 static int fr_stateiter __P((ipftoken_t *, ipfgeniter_t *, ipf_stack_t *));
138 
139 int fr_stputent __P((caddr_t, ipf_stack_t *));
140 int fr_stgetent __P((caddr_t, ipf_stack_t *));
141 
142 #define	ONE_DAY		IPF_TTLVAL(1 * 86400)	/* 1 day */
143 #define	FIVE_DAYS	(5 * ONE_DAY)
144 #define	DOUBLE_HASH(x, ifs)	\
145     (((x) + ifs->ifs_ips_seed[(x) % ifs->ifs_fr_statesize]) % ifs->ifs_fr_statesize)
146 
147 
148 
149 /* ------------------------------------------------------------------------ */
150 /* Function:    fr_stateinit                                                */
151 /* Returns:     int - 0 == success, -1 == failure                           */
152 /* Parameters:  Nil                                                         */
153 /*                                                                          */
154 /* Initialise all the global variables used within the state code.          */
155 /* This action also includes initiailising locks.                           */
156 /* ------------------------------------------------------------------------ */
157 int fr_stateinit(ifs)
158 ipf_stack_t *ifs;
159 {
160 #if defined(NEED_LOCAL_RAND) || !defined(_KERNEL)
161 	struct timeval tv;
162 #endif
163 	int i;
164 
165 	KMALLOCS(ifs->ifs_ips_table, ipstate_t **,
166 		 ifs->ifs_fr_statesize * sizeof(ipstate_t *));
167 	if (ifs->ifs_ips_table == NULL)
168 		return -1;
169 	bzero((char *)ifs->ifs_ips_table,
170 	      ifs->ifs_fr_statesize * sizeof(ipstate_t *));
171 
172 	KMALLOCS(ifs->ifs_ips_seed, u_long *,
173 		 ifs->ifs_fr_statesize * sizeof(*ifs->ifs_ips_seed));
174 	if (ifs->ifs_ips_seed == NULL)
175 		return -2;
176 #if defined(NEED_LOCAL_RAND) || !defined(_KERNEL)
177 	tv.tv_sec = 0;
178 	GETKTIME(&tv);
179 #endif
180 	for (i = 0; i < ifs->ifs_fr_statesize; i++) {
181 		/*
182 		 * XXX - ips_seed[X] should be a random number of sorts.
183 		 */
184 #if !defined(NEED_LOCAL_RAND) && defined(_KERNEL)
185 		ifs->ifs_ips_seed[i] = ipf_random();
186 #else
187 		ifs->ifs_ips_seed[i] = ((u_long)ifs->ifs_ips_seed + i) *
188 		    ifs->ifs_fr_statesize;
189 		ifs->ifs_ips_seed[i] += tv.tv_sec;
190 		ifs->ifs_ips_seed[i] *= (u_long)ifs->ifs_ips_seed;
191 		ifs->ifs_ips_seed[i] ^= 0x5a5aa5a5;
192 		ifs->ifs_ips_seed[i] *= ifs->ifs_fr_statemax;
193 #endif
194 	}
195 
196 	/* fill icmp reply type table */
197 	for (i = 0; i <= ICMP_MAXTYPE; i++)
198 		icmpreplytype4[i] = -1;
199 	icmpreplytype4[ICMP_ECHO] = ICMP_ECHOREPLY;
200 	icmpreplytype4[ICMP_TSTAMP] = ICMP_TSTAMPREPLY;
201 	icmpreplytype4[ICMP_IREQ] = ICMP_IREQREPLY;
202 	icmpreplytype4[ICMP_MASKREQ] = ICMP_MASKREPLY;
203 #ifdef	USE_INET6
204 	/* fill icmp reply type table */
205 	for (i = 0; i <= ICMP6_MAXTYPE; i++)
206 		icmpreplytype6[i] = -1;
207 	icmpreplytype6[ICMP6_ECHO_REQUEST] = ICMP6_ECHO_REPLY;
208 	icmpreplytype6[ICMP6_MEMBERSHIP_QUERY] = ICMP6_MEMBERSHIP_REPORT;
209 	icmpreplytype6[ICMP6_NI_QUERY] = ICMP6_NI_REPLY;
210 	icmpreplytype6[ND_ROUTER_SOLICIT] = ND_ROUTER_ADVERT;
211 	icmpreplytype6[ND_NEIGHBOR_SOLICIT] = ND_NEIGHBOR_ADVERT;
212 #endif
213 
214 	KMALLOCS(ifs->ifs_ips_stats.iss_bucketlen, u_long *,
215 		 ifs->ifs_fr_statesize * sizeof(u_long));
216 	if (ifs->ifs_ips_stats.iss_bucketlen == NULL)
217 		return -1;
218 	bzero((char *)ifs->ifs_ips_stats.iss_bucketlen,
219 	      ifs->ifs_fr_statesize * sizeof(u_long));
220 
221 	if (ifs->ifs_fr_state_maxbucket == 0) {
222 		for (i = ifs->ifs_fr_statesize; i > 0; i >>= 1)
223 			ifs->ifs_fr_state_maxbucket++;
224 		ifs->ifs_fr_state_maxbucket *= 2;
225 	}
226 
227 	fr_sttab_init(ifs->ifs_ips_tqtqb, ifs);
228 	ifs->ifs_ips_tqtqb[IPF_TCP_NSTATES - 1].ifq_next = &ifs->ifs_ips_udptq;
229 	ifs->ifs_ips_udptq.ifq_ttl = (u_long)ifs->ifs_fr_udptimeout;
230 	ifs->ifs_ips_udptq.ifq_ref = 1;
231 	ifs->ifs_ips_udptq.ifq_head = NULL;
232 	ifs->ifs_ips_udptq.ifq_tail = &ifs->ifs_ips_udptq.ifq_head;
233 	MUTEX_INIT(&ifs->ifs_ips_udptq.ifq_lock, "ipftq udp tab");
234 	ifs->ifs_ips_udptq.ifq_next = &ifs->ifs_ips_udpacktq;
235 	ifs->ifs_ips_udpacktq.ifq_ttl = (u_long)ifs->ifs_fr_udpacktimeout;
236 	ifs->ifs_ips_udpacktq.ifq_ref = 1;
237 	ifs->ifs_ips_udpacktq.ifq_head = NULL;
238 	ifs->ifs_ips_udpacktq.ifq_tail = &ifs->ifs_ips_udpacktq.ifq_head;
239 	MUTEX_INIT(&ifs->ifs_ips_udpacktq.ifq_lock, "ipftq udpack tab");
240 	ifs->ifs_ips_udpacktq.ifq_next = &ifs->ifs_ips_icmptq;
241 	ifs->ifs_ips_icmptq.ifq_ttl = (u_long)ifs->ifs_fr_icmptimeout;
242 	ifs->ifs_ips_icmptq.ifq_ref = 1;
243 	ifs->ifs_ips_icmptq.ifq_head = NULL;
244 	ifs->ifs_ips_icmptq.ifq_tail = &ifs->ifs_ips_icmptq.ifq_head;
245 	MUTEX_INIT(&ifs->ifs_ips_icmptq.ifq_lock, "ipftq icmp tab");
246 	ifs->ifs_ips_icmptq.ifq_next = &ifs->ifs_ips_icmpacktq;
247 	ifs->ifs_ips_icmpacktq.ifq_ttl = (u_long)ifs->ifs_fr_icmpacktimeout;
248 	ifs->ifs_ips_icmpacktq.ifq_ref = 1;
249 	ifs->ifs_ips_icmpacktq.ifq_head = NULL;
250 	ifs->ifs_ips_icmpacktq.ifq_tail = &ifs->ifs_ips_icmpacktq.ifq_head;
251 	MUTEX_INIT(&ifs->ifs_ips_icmpacktq.ifq_lock, "ipftq icmpack tab");
252 	ifs->ifs_ips_icmpacktq.ifq_next = &ifs->ifs_ips_iptq;
253 	ifs->ifs_ips_iptq.ifq_ttl = (u_long)ifs->ifs_fr_iptimeout;
254 	ifs->ifs_ips_iptq.ifq_ref = 1;
255 	ifs->ifs_ips_iptq.ifq_head = NULL;
256 	ifs->ifs_ips_iptq.ifq_tail = &ifs->ifs_ips_iptq.ifq_head;
257 	MUTEX_INIT(&ifs->ifs_ips_iptq.ifq_lock, "ipftq ip tab");
258 	ifs->ifs_ips_iptq.ifq_next = &ifs->ifs_ips_deletetq;
259 	/* entry's ttl in deletetq is just 1 tick */
260 	ifs->ifs_ips_deletetq.ifq_ttl = (u_long) 1;
261 	ifs->ifs_ips_deletetq.ifq_ref = 1;
262 	ifs->ifs_ips_deletetq.ifq_head = NULL;
263 	ifs->ifs_ips_deletetq.ifq_tail = &ifs->ifs_ips_deletetq.ifq_head;
264 	MUTEX_INIT(&ifs->ifs_ips_deletetq.ifq_lock, "state delete queue");
265 	ifs->ifs_ips_deletetq.ifq_next = NULL;
266 
267 	RWLOCK_INIT(&ifs->ifs_ipf_state, "ipf IP state rwlock");
268 	MUTEX_INIT(&ifs->ifs_ipf_stinsert, "ipf state insert mutex");
269 	ifs->ifs_fr_state_init = 1;
270 
271 	ifs->ifs_ips_last_force_flush = ifs->ifs_fr_ticks;
272 	return 0;
273 }
274 
275 
276 /* ------------------------------------------------------------------------ */
277 /* Function:    fr_stateunload                                              */
278 /* Returns:     Nil                                                         */
279 /* Parameters:  Nil                                                         */
280 /*                                                                          */
281 /* Release and destroy any resources acquired or initialised so that        */
282 /* IPFilter can be unloaded or re-initialised.                              */
283 /* ------------------------------------------------------------------------ */
284 void fr_stateunload(ifs)
285 ipf_stack_t *ifs;
286 {
287 	ipftq_t *ifq, *ifqnext;
288 	ipstate_t *is;
289 
290 	while ((is = ifs->ifs_ips_list) != NULL)
291 	    fr_delstate(is, 0, ifs);
292 
293 	/*
294 	 * Proxy timeout queues are not cleaned here because although they
295 	 * exist on the state list, appr_unload is called after fr_stateunload
296 	 * and the proxies actually are responsible for them being created.
297 	 * Should the proxy timeouts have their own list?  There's no real
298 	 * justification as this is the only complicationA
299 	 */
300 	for (ifq = ifs->ifs_ips_utqe; ifq != NULL; ifq = ifqnext) {
301 		ifqnext = ifq->ifq_next;
302 		if (((ifq->ifq_flags & IFQF_PROXY) == 0) &&
303 		    (fr_deletetimeoutqueue(ifq) == 0))
304 			fr_freetimeoutqueue(ifq, ifs);
305 	}
306 
307 	ifs->ifs_ips_stats.iss_inuse = 0;
308 	ifs->ifs_ips_num = 0;
309 
310 	if (ifs->ifs_fr_state_init == 1) {
311 		fr_sttab_destroy(ifs->ifs_ips_tqtqb);
312 		MUTEX_DESTROY(&ifs->ifs_ips_udptq.ifq_lock);
313 		MUTEX_DESTROY(&ifs->ifs_ips_icmptq.ifq_lock);
314 		MUTEX_DESTROY(&ifs->ifs_ips_udpacktq.ifq_lock);
315 		MUTEX_DESTROY(&ifs->ifs_ips_icmpacktq.ifq_lock);
316 		MUTEX_DESTROY(&ifs->ifs_ips_iptq.ifq_lock);
317 		MUTEX_DESTROY(&ifs->ifs_ips_deletetq.ifq_lock);
318 	}
319 
320 	if (ifs->ifs_ips_table != NULL) {
321 		KFREES(ifs->ifs_ips_table,
322 		       ifs->ifs_fr_statesize * sizeof(*ifs->ifs_ips_table));
323 		ifs->ifs_ips_table = NULL;
324 	}
325 
326 	if (ifs->ifs_ips_seed != NULL) {
327 		KFREES(ifs->ifs_ips_seed,
328 		       ifs->ifs_fr_statesize * sizeof(*ifs->ifs_ips_seed));
329 		ifs->ifs_ips_seed = NULL;
330 	}
331 
332 	if (ifs->ifs_ips_stats.iss_bucketlen != NULL) {
333 		KFREES(ifs->ifs_ips_stats.iss_bucketlen,
334 		       ifs->ifs_fr_statesize * sizeof(u_long));
335 		ifs->ifs_ips_stats.iss_bucketlen = NULL;
336 	}
337 
338 	if (ifs->ifs_fr_state_maxbucket_reset == 1)
339 		ifs->ifs_fr_state_maxbucket = 0;
340 
341 	if (ifs->ifs_fr_state_init == 1) {
342 		ifs->ifs_fr_state_init = 0;
343 		RW_DESTROY(&ifs->ifs_ipf_state);
344 		MUTEX_DESTROY(&ifs->ifs_ipf_stinsert);
345 	}
346 }
347 
348 
349 /* ------------------------------------------------------------------------ */
350 /* Function:    fr_statetstats                                              */
351 /* Returns:     ips_state_t* - pointer to state stats structure             */
352 /* Parameters:  Nil                                                         */
353 /*                                                                          */
354 /* Put all the current numbers and pointers into a single struct and return */
355 /* a pointer to it.                                                         */
356 /* ------------------------------------------------------------------------ */
357 static ips_stat_t *fr_statetstats(ifs)
358 ipf_stack_t *ifs;
359 {
360 	ifs->ifs_ips_stats.iss_active = ifs->ifs_ips_num;
361 	ifs->ifs_ips_stats.iss_statesize = ifs->ifs_fr_statesize;
362 	ifs->ifs_ips_stats.iss_statemax = ifs->ifs_fr_statemax;
363 	ifs->ifs_ips_stats.iss_table = ifs->ifs_ips_table;
364 	ifs->ifs_ips_stats.iss_list = ifs->ifs_ips_list;
365 	ifs->ifs_ips_stats.iss_ticks = ifs->ifs_fr_ticks;
366 	return &ifs->ifs_ips_stats;
367 }
368 
369 /* ------------------------------------------------------------------------ */
370 /* Function:    fr_state_remove                                             */
371 /* Returns:     int - 0 == success, != 0 == failure                         */
372 /* Parameters:  data(I) - pointer to state structure to delete from table   */
373 /*                                                                          */
374 /* Search for a state structure that matches the one passed, according to   */
375 /* the IP addresses and other protocol specific information.                */
376 /* ------------------------------------------------------------------------ */
377 static int fr_state_remove(data, ifs)
378 caddr_t data;
379 ipf_stack_t *ifs;
380 {
381 	ipstate_t *sp, st;
382 	int error;
383 
384 	sp = &st;
385 	error = fr_inobj(data, &st, IPFOBJ_IPSTATE);
386 	if (error)
387 		return EFAULT;
388 
389 	WRITE_ENTER(&ifs->ifs_ipf_state);
390 	for (sp = ifs->ifs_ips_list; sp; sp = sp->is_next)
391 		if ((sp->is_p == st.is_p) && (sp->is_v == st.is_v) &&
392 		    !bcmp((caddr_t)&sp->is_src, (caddr_t)&st.is_src,
393 			  sizeof(st.is_src)) &&
394 		    !bcmp((caddr_t)&sp->is_dst, (caddr_t)&st.is_src,
395 			  sizeof(st.is_dst)) &&
396 		    !bcmp((caddr_t)&sp->is_ps, (caddr_t)&st.is_ps,
397 			  sizeof(st.is_ps))) {
398 			fr_delstate(sp, ISL_REMOVE, ifs);
399 			RWLOCK_EXIT(&ifs->ifs_ipf_state);
400 			return 0;
401 		}
402 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
403 	return ESRCH;
404 }
405 
406 
407 /* ------------------------------------------------------------------------ */
408 /* Function:    fr_state_ioctl                                              */
409 /* Returns:     int - 0 == success, != 0 == failure                         */
410 /* Parameters:  data(I) - pointer to ioctl data                             */
411 /*              cmd(I)  - ioctl command integer                             */
412 /*              mode(I) - file mode bits used with open                     */
413 /*                                                                          */
414 /* Processes an ioctl call made to operate on the IP Filter state device.   */
415 /* ------------------------------------------------------------------------ */
416 int fr_state_ioctl(data, cmd, mode, uid, ctx, ifs)
417 caddr_t data;
418 ioctlcmd_t cmd;
419 int mode, uid;
420 void *ctx;
421 ipf_stack_t *ifs;
422 {
423 	int arg, ret, error = 0;
424 
425 	switch (cmd)
426 	{
427 	/*
428 	 * Delete an entry from the state table.
429 	 */
430 	case SIOCDELST :
431 	    error = fr_state_remove(data, ifs);
432 		break;
433 	/*
434 	 * Flush the state table
435 	 */
436 	case SIOCIPFFL :
437 		BCOPYIN(data, (char *)&arg, sizeof(arg));
438 		if (arg == 0 || arg == 1) {
439 			WRITE_ENTER(&ifs->ifs_ipf_state);
440 			ret = fr_state_flush(arg, 4, ifs);
441 			RWLOCK_EXIT(&ifs->ifs_ipf_state);
442 			BCOPYOUT((char *)&ret, data, sizeof(ret));
443 		} else
444 			error = EINVAL;
445 		break;
446 #ifdef	USE_INET6
447 	case SIOCIPFL6 :
448 		BCOPYIN(data, (char *)&arg, sizeof(arg));
449 		if (arg == 0 || arg == 1) {
450 			WRITE_ENTER(&ifs->ifs_ipf_state);
451 			ret = fr_state_flush(arg, 6, ifs);
452 			RWLOCK_EXIT(&ifs->ifs_ipf_state);
453 			BCOPYOUT((char *)&ret, data, sizeof(ret));
454 		} else
455 			error = EINVAL;
456 		break;
457 #endif
458 #ifdef	IPFILTER_LOG
459 	/*
460 	 * Flush the state log.
461 	 */
462 	case SIOCIPFFB :
463 		if (!(mode & FWRITE))
464 			error = EPERM;
465 		else {
466 			int tmp;
467 
468 			tmp = ipflog_clear(IPL_LOGSTATE, ifs);
469 			BCOPYOUT((char *)&tmp, data, sizeof(tmp));
470 		}
471 		break;
472 	/*
473 	 * Turn logging of state information on/off.
474 	 */
475 	case SIOCSETLG :
476 		if (!(mode & FWRITE))
477 			error = EPERM;
478 		else {
479 			BCOPYIN((char *)data,
480 				       (char *)&ifs->ifs_ipstate_logging,
481 				       sizeof(ifs->ifs_ipstate_logging));
482 		}
483 		break;
484 	/*
485 	 * Return the current state of logging.
486 	 */
487 	case SIOCGETLG :
488 		BCOPYOUT((char *)&ifs->ifs_ipstate_logging, (char *)data,
489 			sizeof(ifs->ifs_ipstate_logging));
490 		break;
491 	/*
492 	 * Return the number of bytes currently waiting to be read.
493 	 */
494 	case FIONREAD :
495 		arg = ifs->ifs_iplused[IPL_LOGSTATE]; /* returned in an int */
496 		BCOPYOUT((char *)&arg, data, sizeof(arg));
497 		break;
498 #endif
499 	/*
500 	 * Get the current state statistics.
501 	 */
502 	case SIOCGETFS :
503 		error = fr_outobj(data, fr_statetstats(ifs), IPFOBJ_STATESTAT);
504 		break;
505 	/*
506 	 * Lock/Unlock the state table.  (Locking prevents any changes, which
507 	 * means no packets match).
508 	 */
509 	case SIOCSTLCK :
510 		if (!(mode & FWRITE)) {
511 			error = EPERM;
512 		} else {
513 			fr_lock(data, &ifs->ifs_fr_state_lock);
514 		}
515 		break;
516 	/*
517 	 * Add an entry to the current state table.
518 	 */
519 	case SIOCSTPUT :
520 		if (!ifs->ifs_fr_state_lock || !(mode &FWRITE)) {
521 			error = EACCES;
522 			break;
523 		}
524 		error = fr_stputent(data, ifs);
525 		break;
526 	/*
527 	 * Get a state table entry.
528 	 */
529 	case SIOCSTGET :
530 		if (!ifs->ifs_fr_state_lock) {
531 			error = EACCES;
532 			break;
533 		}
534 		error = fr_stgetent(data, ifs);
535 		break;
536 
537 	case SIOCGENITER :
538 	    {
539 		ipftoken_t *token;
540 		ipfgeniter_t iter;
541 
542 		error = fr_inobj(data, &iter, IPFOBJ_GENITER);
543 		if (error != 0)
544 			break;
545 
546 		token = ipf_findtoken(IPFGENITER_STATE, uid, ctx, ifs);
547 		if (token != NULL)
548 			error = fr_stateiter(token, &iter, ifs);
549 		else
550 			error = ESRCH;
551 		RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
552 		break;
553 	    }
554 
555 	case SIOCIPFDELTOK :
556 		(void) BCOPYIN(data, (char *)&arg, sizeof(arg));
557 		error = ipf_deltoken(arg, uid, ctx, ifs);
558 		break;
559 
560 	default :
561 		error = EINVAL;
562 		break;
563 	}
564 	return error;
565 }
566 
567 
568 /* ------------------------------------------------------------------------ */
569 /* Function:    fr_stgetent                                                 */
570 /* Returns:     int - 0 == success, != 0 == failure                         */
571 /* Parameters:  data(I) - pointer to state structure to retrieve from table */
572 /*                                                                          */
573 /* Copy out state information from the kernel to a user space process.  If  */
574 /* there is a filter rule associated with the state entry, copy that out    */
575 /* as well.  The entry to copy out is taken from the value of "ips_next" in */
576 /* the struct passed in and if not null and not found in the list of current*/
577 /* state entries, the retrieval fails.                                      */
578 /* ------------------------------------------------------------------------ */
579 int fr_stgetent(data, ifs)
580 caddr_t data;
581 ipf_stack_t *ifs;
582 {
583 	ipstate_t *is, *isn;
584 	ipstate_save_t ips;
585 	int error;
586 
587 	error = fr_inobj(data, &ips, IPFOBJ_STATESAVE);
588 	if (error)
589 		return EFAULT;
590 
591 	isn = ips.ips_next;
592 	if (isn == NULL) {
593 		isn = ifs->ifs_ips_list;
594 		if (isn == NULL) {
595 			if (ips.ips_next == NULL)
596 				return ENOENT;
597 			return 0;
598 		}
599 	} else {
600 		/*
601 		 * Make sure the pointer we're copying from exists in the
602 		 * current list of entries.  Security precaution to prevent
603 		 * copying of random kernel data.
604 		 */
605 		for (is = ifs->ifs_ips_list; is; is = is->is_next)
606 			if (is == isn)
607 				break;
608 		if (!is)
609 			return ESRCH;
610 	}
611 	ips.ips_next = isn->is_next;
612 	bcopy((char *)isn, (char *)&ips.ips_is, sizeof(ips.ips_is));
613 	ips.ips_rule = isn->is_rule;
614 	if (isn->is_rule != NULL)
615 		bcopy((char *)isn->is_rule, (char *)&ips.ips_fr,
616 		      sizeof(ips.ips_fr));
617 	error = fr_outobj(data, &ips, IPFOBJ_STATESAVE);
618 	if (error)
619 		return EFAULT;
620 	return 0;
621 }
622 
623 
624 /* ------------------------------------------------------------------------ */
625 /* Function:    fr_stputent                                                 */
626 /* Returns:     int - 0 == success, != 0 == failure                         */
627 /* Parameters:  data(I) - pointer to state information struct               */
628 /*                                                                          */
629 /* This function implements the SIOCSTPUT ioctl: insert a state entry into  */
630 /* the state table.  If the state info. includes a pointer to a filter rule */
631 /* then also add in an orphaned rule (will not show up in any "ipfstat -io" */
632 /* output.                                                                  */
633 /* ------------------------------------------------------------------------ */
634 int fr_stputent(data, ifs)
635 caddr_t data;
636 ipf_stack_t *ifs;
637 {
638 	ipstate_t *is, *isn;
639 	ipstate_save_t ips;
640 	int error, i;
641 	frentry_t *fr;
642 	char *name;
643 
644 	error = fr_inobj(data, &ips, IPFOBJ_STATESAVE);
645 	if (error)
646 		return EFAULT;
647 
648 	KMALLOC(isn, ipstate_t *);
649 	if (isn == NULL)
650 		return ENOMEM;
651 
652 	bcopy((char *)&ips.ips_is, (char *)isn, sizeof(*isn));
653 	bzero((char *)isn, offsetof(struct ipstate, is_pkts));
654 	isn->is_sti.tqe_pnext = NULL;
655 	isn->is_sti.tqe_next = NULL;
656 	isn->is_sti.tqe_ifq = NULL;
657 	isn->is_sti.tqe_parent = isn;
658 	isn->is_ifp[0] = NULL;
659 	isn->is_ifp[1] = NULL;
660 	isn->is_ifp[2] = NULL;
661 	isn->is_ifp[3] = NULL;
662 	isn->is_sync = NULL;
663 	fr = ips.ips_rule;
664 
665 	if (fr == NULL) {
666 		READ_ENTER(&ifs->ifs_ipf_state);
667 		fr_stinsert(isn, 0, ifs);
668 		MUTEX_EXIT(&isn->is_lock);
669 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
670 		return 0;
671 	}
672 
673 	if (isn->is_flags & SI_NEWFR) {
674 		KMALLOC(fr, frentry_t *);
675 		if (fr == NULL) {
676 			KFREE(isn);
677 			return ENOMEM;
678 		}
679 		bcopy((char *)&ips.ips_fr, (char *)fr, sizeof(*fr));
680 		isn->is_rule = fr;
681 		ips.ips_is.is_rule = fr;
682 		MUTEX_NUKE(&fr->fr_lock);
683 		MUTEX_INIT(&fr->fr_lock, "state filter rule lock");
684 
685 		/*
686 		 * Look up all the interface names in the rule.
687 		 */
688 		for (i = 0; i < 4; i++) {
689 			name = fr->fr_ifnames[i];
690 			fr->fr_ifas[i] = fr_resolvenic(name, fr->fr_v, ifs);
691 			name = isn->is_ifname[i];
692 			isn->is_ifp[i] = fr_resolvenic(name, isn->is_v, ifs);
693 		}
694 
695 		fr->fr_ref = 0;
696 		fr->fr_dsize = 0;
697 		fr->fr_data = NULL;
698 		fr->fr_type = FR_T_NONE;
699 
700 		fr_resolvedest(&fr->fr_tif, fr->fr_v, ifs);
701 		fr_resolvedest(&fr->fr_dif, fr->fr_v, ifs);
702 		fr_resolvedest(&fr->fr_rif, fr->fr_v, ifs);
703 
704 		/*
705 		 * send a copy back to userland of what we ended up
706 		 * to allow for verification.
707 		 */
708 		error = fr_outobj(data, &ips, IPFOBJ_STATESAVE);
709 		if (error) {
710 			KFREE(isn);
711 			MUTEX_DESTROY(&fr->fr_lock);
712 			KFREE(fr);
713 			return EFAULT;
714 		}
715 		READ_ENTER(&ifs->ifs_ipf_state);
716 		fr_stinsert(isn, 0, ifs);
717 		MUTEX_EXIT(&isn->is_lock);
718 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
719 
720 	} else {
721 		READ_ENTER(&ifs->ifs_ipf_state);
722 		for (is = ifs->ifs_ips_list; is; is = is->is_next)
723 			if (is->is_rule == fr) {
724 				fr_stinsert(isn, 0, ifs);
725 				MUTEX_EXIT(&isn->is_lock);
726 				break;
727 			}
728 
729 		if (is == NULL) {
730 			KFREE(isn);
731 			isn = NULL;
732 		}
733 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
734 
735 		return (isn == NULL) ? ESRCH : 0;
736 	}
737 
738 	return 0;
739 }
740 
741 
742 /* ------------------------------------------------------------------------ */
743 /* Function:   fr_stinsert                                                  */
744 /* Returns:    Nil                                                          */
745 /* Parameters: is(I)  - pointer to state structure                          */
746 /*             rev(I) - flag indicating forward/reverse direction of packet */
747 /*                                                                          */
748 /* Inserts a state structure into the hash table (for lookups) and the list */
749 /* of state entries (for enumeration).  Resolves all of the interface names */
750 /* to pointers and adjusts running stats for the hash table as appropriate. */
751 /*                                                                          */
752 /* Locking: it is assumed that some kind of lock on ipf_state is held.      */
753 /*          Exits with is_lock initialised and held.                        */
754 /* ------------------------------------------------------------------------ */
755 void fr_stinsert(is, rev, ifs)
756 ipstate_t *is;
757 int rev;
758 ipf_stack_t *ifs;
759 {
760 	frentry_t *fr;
761 	u_int hv;
762 	int i;
763 
764 	MUTEX_INIT(&is->is_lock, "ipf state entry");
765 
766 	fr = is->is_rule;
767 	if (fr != NULL) {
768 		MUTEX_ENTER(&fr->fr_lock);
769 		fr->fr_ref++;
770 		fr->fr_statecnt++;
771 		MUTEX_EXIT(&fr->fr_lock);
772 	}
773 
774 	/*
775 	 * Look up all the interface names in the state entry.
776 	 */
777 	for (i = 0; i < 4; i++) {
778 		if (is->is_ifp[i] != NULL)
779 			continue;
780 		is->is_ifp[i] = fr_resolvenic(is->is_ifname[i], is->is_v, ifs);
781 	}
782 
783 	/*
784 	 * If we could trust is_hv, then the modulous would not be needed, but
785 	 * when running with IPFILTER_SYNC, this stops bad values.
786 	 */
787 	hv = is->is_hv % ifs->ifs_fr_statesize;
788 	is->is_hv = hv;
789 
790 	/*
791 	 * We need to get both of these locks...the first because it is
792 	 * possible that once the insert is complete another packet might
793 	 * come along, match the entry and want to update it.
794 	 */
795 	MUTEX_ENTER(&is->is_lock);
796 	MUTEX_ENTER(&ifs->ifs_ipf_stinsert);
797 
798 	/*
799 	 * add into list table.
800 	 */
801 	if (ifs->ifs_ips_list != NULL)
802 		ifs->ifs_ips_list->is_pnext = &is->is_next;
803 	is->is_pnext = &ifs->ifs_ips_list;
804 	is->is_next = ifs->ifs_ips_list;
805 	ifs->ifs_ips_list = is;
806 
807 	if (ifs->ifs_ips_table[hv] != NULL)
808 		ifs->ifs_ips_table[hv]->is_phnext = &is->is_hnext;
809 	else
810 		ifs->ifs_ips_stats.iss_inuse++;
811 	is->is_phnext = ifs->ifs_ips_table + hv;
812 	is->is_hnext = ifs->ifs_ips_table[hv];
813 	ifs->ifs_ips_table[hv] = is;
814 	ifs->ifs_ips_stats.iss_bucketlen[hv]++;
815 	ifs->ifs_ips_num++;
816 	MUTEX_EXIT(&ifs->ifs_ipf_stinsert);
817 
818 	fr_setstatequeue(is, rev, ifs);
819 }
820 
821 /* ------------------------------------------------------------------------ */
822 /* Function:	fr_match_ipv4addrs					    */
823 /* Returns:	int -	2 strong match (same addresses, same direction)	    */
824 /*			1 weak match (same address, opposite direction)	    */
825 /*			0 no match					    */
826 /*									    */
827 /* Function matches IPv4 addresses.					    */
828 /* ------------------------------------------------------------------------ */
829 static int fr_match_ipv4addrs(is1, is2)
830 ipstate_t *is1;
831 ipstate_t *is2;
832 {
833 	int	rv;
834 
835 	if (is1->is_saddr == is2->is_saddr && is1->is_daddr == is2->is_daddr)
836 		rv = 2;
837 	else if (is1->is_saddr == is2->is_daddr &&
838 	    is1->is_daddr == is2->is_saddr)
839 		rv = 1;
840 	else
841 		rv = 0;
842 
843 	return (rv);
844 }
845 
846 /* ------------------------------------------------------------------------ */
847 /* Function:	fr_match_ipv6addrs					    */
848 /* Returns:	int - 	2 strong match (same addresses, same direction)	    */
849 /*			1 weak match (same addresses, opposite direction)   */
850 /*			0 no match					    */
851 /*									    */
852 /* Function matches IPv6 addresses.					    */
853 /* ------------------------------------------------------------------------ */
854 static int fr_match_ipv6addrs(is1, is2)
855 ipstate_t *is1;
856 ipstate_t *is2;
857 {
858 	int	rv;
859 
860 	if (IP6_EQ(&is1->is_src, &is2->is_src) &&
861 	    IP6_EQ(&is1->is_dst, &is2->is_dst))
862 		rv = 2;
863 	else if (IP6_EQ(&is1->is_src, &is2->is_dst) &&
864 	    IP6_EQ(&is1->is_dst, &is2->is_src)) {
865 		rv = 1;
866 	}
867 	else
868 		rv = 0;
869 
870 	return (rv);
871 }
872 /* ------------------------------------------------------------------------ */
873 /* Function:	fr_match_addresses					    */
874 /* Returns:	int - 	2 strong match (same addresses, same direction)	    */
875 /*			1 weak match (same address, opposite directions)    */
876 /* 			0 no match					    */
877 /* Parameters:	is1, is2 pointers to states we are checking		    */
878 /*									    */
879 /* Matches addresses, function uses fr_match_ipvXaddrs() to deal with IPv4  */
880 /* and IPv6 address format.						    */
881 /* ------------------------------------------------------------------------ */
882 static int fr_match_addresses(is1, is2)
883 ipstate_t *is1;
884 ipstate_t *is2;
885 {
886 	int	rv;
887 
888 	if (is1->is_v == 4) {
889 		rv = fr_match_ipv4addrs(is1, is2);
890 	}
891 	else {
892 		rv = fr_match_ipv6addrs(is1, is2);
893 	}
894 
895 	return (rv);
896 }
897 
898 /* ------------------------------------------------------------------------ */
899 /* Function:	fr_match_ppairs						    */
900 /* Returns:	int - 	2 strong match (same ports, same direction)	    */
901 /*			1 weak match (same ports, different direction)	    */
902 /*			0 no match					    */
903 /* Parameters	ppairs1, ppairs - src, dst ports we want to match.	    */
904 /*									    */
905 /* Matches two port_pair_t types (port pairs). Each port pair contains	    */
906 /* src, dst port, which belong to session (state entry).		    */
907 /* ------------------------------------------------------------------------ */
908 static int fr_match_ppairs(ppairs1, ppairs2)
909 port_pair_t *ppairs1;
910 port_pair_t *ppairs2;
911 {
912 	int	rv;
913 
914 	if (ppairs1->pp_sport == ppairs2->pp_sport &&
915 	    ppairs1->pp_dport == ppairs2->pp_dport)
916 		rv = 2;
917 	else if (ppairs1->pp_sport == ppairs2->pp_dport &&
918 		    ppairs1->pp_dport == ppairs2->pp_sport)
919 		rv = 1;
920 	else
921 		rv = 0;
922 
923 	return (rv);
924 }
925 
926 /* ------------------------------------------------------------------------ */
927 /* Function:	fr_match_l4_hdr						    */
928 /* Returns:	int - 	0 no match, 					    */
929 /*			1 weak match (same ports, different directions)	    */
930 /*			2 strong match (same ports, same direction)	    */
931 /* Parameters	is1, is2 - states we want to match			    */
932 /*									    */
933 /* Function matches L4 header data (source ports for TCP, UDP, CallIds for  */
934 /* GRE protocol).							    */
935 /* ------------------------------------------------------------------------ */
936 static int fr_match_l4_hdr(is1, is2)
937 ipstate_t *is1;
938 ipstate_t *is2;
939 {
940 	int	rv = 0;
941 	port_pair_t	pp1;
942 	port_pair_t	pp2;
943 
944 	if (is1->is_p != is2->is_p)
945 		return (0);
946 
947 	switch (is1->is_p) {
948 		case	IPPROTO_TCP:
949 			pp1.pp_sport = is1->is_ps.is_ts.ts_sport;
950 			pp1.pp_dport = is1->is_ps.is_ts.ts_dport;
951 			pp2.pp_sport = is2->is_ps.is_ts.ts_sport;
952 			pp2.pp_dport = is2->is_ps.is_ts.ts_dport;
953 			rv = fr_match_ppairs(&pp1, &pp2);
954 			break;
955 		case	IPPROTO_UDP:
956 			pp1.pp_sport = is1->is_ps.is_us.us_sport;
957 			pp1.pp_dport = is1->is_ps.is_us.us_dport;
958 			pp2.pp_sport = is2->is_ps.is_us.us_sport;
959 			pp2.pp_dport = is2->is_ps.is_us.us_dport;
960 			rv = fr_match_ppairs(&pp1, &pp2);
961 			break;
962 		case	IPPROTO_GRE:
963 			/* greinfo_t can be also interprted as port pair */
964 			pp1.pp_sport = is1->is_ps.is_ug.gs_call[0];
965 			pp1.pp_dport = is1->is_ps.is_ug.gs_call[1];
966 			pp2.pp_sport = is2->is_ps.is_ug.gs_call[0];
967 			pp2.pp_dport = is2->is_ps.is_ug.gs_call[1];
968 			rv = fr_match_ppairs(&pp1, &pp2);
969 			break;
970 		case	IPPROTO_ICMP:
971 		case	IPPROTO_ICMPV6:
972 			if (bcmp(&is1->is_ps, &is2->is_ps, sizeof(icmpinfo_t)))
973 				rv = 1;
974 			else
975 				rv = 0;
976 			break;
977 		default:
978 			rv = 0;
979 	}
980 
981 	return (rv);
982 }
983 
984 /* ------------------------------------------------------------------------ */
985 /* Function:	fr_matchstates						    */
986 /* Returns:	int - nonzero match, zero no match			    */
987 /* Parameters	is1, is2 - states we want to match			    */
988 /*									    */
989 /* The state entries are equal (identical match) if they belong to the same */
990 /* session. Any time new state entry is being added the fr_addstate() 	    */
991 /* function creates temporal state entry from the data it gets from IP and  */
992 /* L4 header. The fr_matchstats() must be also aware of packet direction,   */
993 /* which is also stored within the state entry. We should keep in mind the  */
994 /* information about packet direction is spread accross L3 (addresses) and  */
995 /* L4 (ports). There are three possible relationships betwee is1, is2:	    */
996 /* 		- no match (match(is1, is2) == 0))			    */
997 /*		- weak match same addresses (ports), but different 	    */
998 /*			directions (1)	(fr_match_xxxx(is1, is2) == 1)	    */
999 /*		- strong match same addresses (ports) and same directions   */
1000 /*			 (2) (fr_match_xxxx(is1, is2) == 2)		    */
1001 /*									    */
1002 /* There are functions, which match match addresses (L3 header) in is1, is2 */
1003 /* and functions, which are used to compare ports (L4 header) data. We say  */
1004 /* the is1 and is2 are same (identical) if there is a match		    */
1005 /* (fr_match_l4_hdr(is1, is2) != 0) and matchlevels are same for entries    */
1006 /* (fr_match_l3_hdr(is1, is2) == fr_match_l4_hdr(is1, is2)) for is1, is2.   */
1007 /* Such requirement deals with case as follows:				    */
1008 /*	suppose there are two connections between hosts A, B. Connection 1: */
1009 /*			a.a.a.a:12345 <=> b.b.b.b:54321			    */
1010 /*		Connection 2:						    */
1011 /*			a.a.a.a:54321 <=> b.b.b.b:12345			    */
1012 /* since we've introduced match levels into our fr_matchstates(), we are    */
1013 /* able to identify, which packets belong to connection A and which belong  */
1014 /* to connection B.	Assume there are two entries is1, is2. is1 has been */
1015 /* from con. 1 packet, which travelled from A to B:			    */
1016 /*			a.a.a.a:12345 -> b.b.b.b:54321			    */
1017 /* while s2, has been created from packet which belongs to con. 2 and is    */
1018 /* also coming from A to B:						    */
1019 /*			a.a.a.a:54321 -> b.b.b.b:12345			    */
1020 /* fr_match_l3_hdr(is1, is2) == 2 -> strong match, while		    */
1021 /* fr_match_l4_hdr(is1, is2) == 1 -> weak match. Since match levels are	    */
1022 /* different the state entries are not identical -> no match as a final	    */
1023 /* result.								    */
1024 /* ------------------------------------------------------------------------ */
1025 static int fr_matchstates(is1, is2)
1026 ipstate_t *is1;
1027 ipstate_t *is2;
1028 {
1029 	int	rv;
1030 	int	amatch;
1031 	int	pmatch;
1032 
1033 	if (bcmp(&is1->is_pass, &is2->is_pass,
1034 		 offsetof(struct ipstate, is_ps) -
1035 		 offsetof(struct ipstate, is_pass)) == 0) {
1036 
1037 		pmatch = fr_match_l4_hdr(is1, is2);
1038 		amatch = fr_match_addresses(is1, is2);
1039 		/*
1040 		 * If addresses match (amatch != 0), then 'match levels'
1041 		 * must be same for matching entries. If amatch and pmatch
1042 		 * have different values (different match levels), then
1043 		 * is1 and is2 belong to different sessions.
1044 		 */
1045 		rv = (amatch != 0) && (amatch == pmatch);
1046 	}
1047 	else
1048 		rv = 0;
1049 
1050 	return (rv);
1051 }
1052 
1053 /* ------------------------------------------------------------------------ */
1054 /* Function:    fr_addstate                                                 */
1055 /* Returns:     ipstate_t* - NULL == failure, else pointer to new state     */
1056 /* Parameters:  fin(I)    - pointer to packet information                   */
1057 /*              stsave(O) - pointer to place to save pointer to created     */
1058 /*                          state structure.                                */
1059 /*              flags(I)  - flags to use when creating the structure        */
1060 /*                                                                          */
1061 /* Creates a new IP state structure from the packet information collected.  */
1062 /* Inserts it into the state table and appends to the bottom of the active  */
1063 /* list.  If the capacity of the table has reached the maximum allowed then */
1064 /* the call will fail and a flush is scheduled for the next timeout call.   */
1065 /* ------------------------------------------------------------------------ */
1066 ipstate_t *fr_addstate(fin, stsave, flags)
1067 fr_info_t *fin;
1068 ipstate_t **stsave;
1069 u_int flags;
1070 {
1071 	ipstate_t *is, ips;
1072 	struct icmp *ic;
1073 	u_int pass, hv;
1074 	frentry_t *fr;
1075 	tcphdr_t *tcp;
1076 	grehdr_t *gre;
1077 	void *ifp;
1078 	int out;
1079 	ipf_stack_t *ifs = fin->fin_ifs;
1080 
1081 	if (ifs->ifs_fr_state_lock ||
1082 	    (fin->fin_flx & (FI_SHORT|FI_STATE|FI_FRAGBODY|FI_BAD)))
1083 		return NULL;
1084 
1085 	if ((fin->fin_flx & FI_OOW) && !(fin->fin_tcpf & TH_SYN))
1086 		return NULL;
1087 
1088 	/*
1089 	 * If a "keep state" rule has reached the maximum number of references
1090 	 * to it, then schedule an automatic flush in case we can clear out
1091 	 * some "dead old wood".  Note that because the lock isn't held on
1092 	 * fr it is possible that we could overflow.  The cost of overflowing
1093 	 * is being ignored here as the number by which it can overflow is
1094 	 * a product of the number of simultaneous threads that could be
1095 	 * executing in here, so a limit of 100 won't result in 200, but could
1096 	 * result in 101 or 102.
1097 	 */
1098 	fr = fin->fin_fr;
1099 	if (fr != NULL) {
1100 		if ((ifs->ifs_ips_num == ifs->ifs_fr_statemax) && (fr->fr_statemax == 0)) {
1101 			ATOMIC_INCL(ifs->ifs_ips_stats.iss_max);
1102 			ifs->ifs_fr_state_doflush = 1;
1103 			return NULL;
1104 		}
1105 		if ((fr->fr_statemax != 0) &&
1106 		    (fr->fr_statecnt >= fr->fr_statemax)) {
1107 			ATOMIC_INCL(ifs->ifs_ips_stats.iss_maxref);
1108 			ifs->ifs_fr_state_doflush = 1;
1109 			return NULL;
1110 		}
1111 	}
1112 
1113 	ic = NULL;
1114 	tcp = NULL;
1115 	out = fin->fin_out;
1116 	is = &ips;
1117 	bzero((char *)is, sizeof(*is));
1118 
1119 	if (fr == NULL) {
1120 		pass = ifs->ifs_fr_flags;
1121 		is->is_tag = FR_NOLOGTAG;
1122 	}
1123 	else {
1124 		pass = fr->fr_flags;
1125 	}
1126 
1127 	is->is_die = 1 + ifs->ifs_fr_ticks;
1128 	/*
1129 	 * We want to check everything that is a property of this packet,
1130 	 * but we don't (automatically) care about it's fragment status as
1131 	 * this may change.
1132 	 */
1133 	is->is_pass = pass;
1134 	is->is_v = fin->fin_v;
1135 	is->is_opt[0] = fin->fin_optmsk;
1136 	is->is_optmsk[0] = 0xffffffff;
1137 	is->is_optmsk[1] = 0xffffffff;
1138 	if (is->is_v == 6) {
1139 		is->is_opt[0] &= ~0x8;
1140 		is->is_optmsk[0] &= ~0x8;
1141 		is->is_optmsk[1] &= ~0x8;
1142 	}
1143 	is->is_sec = fin->fin_secmsk;
1144 	is->is_secmsk = 0xffff;
1145 	is->is_auth = fin->fin_auth;
1146 	is->is_authmsk = 0xffff;
1147 
1148 	/*
1149 	 * Copy and calculate...
1150 	 */
1151 	hv = (is->is_p = fin->fin_fi.fi_p);
1152 	is->is_src = fin->fin_fi.fi_src;
1153 	hv += is->is_saddr;
1154 	is->is_dst = fin->fin_fi.fi_dst;
1155 	hv += is->is_daddr;
1156 #ifdef	USE_INET6
1157 	if (fin->fin_v == 6) {
1158 		/*
1159 		 * For ICMPv6, we check to see if the destination address is
1160 		 * a multicast address.  If it is, do not include it in the
1161 		 * calculation of the hash because the correct reply will come
1162 		 * back from a real address, not a multicast address.
1163 		 */
1164 		if ((is->is_p == IPPROTO_ICMPV6) &&
1165 		    IN6_IS_ADDR_MULTICAST(&is->is_dst.in6)) {
1166 			/*
1167 			 * So you can do keep state with neighbour discovery.
1168 			 *
1169 			 * Here we could use the address from the neighbour
1170 			 * solicit message to put in the state structure and
1171 			 * we could use that without a wildcard flag too...
1172 			 */
1173 			is->is_flags |= SI_W_DADDR;
1174 			hv -= is->is_daddr;
1175 		} else {
1176 			hv += is->is_dst.i6[1];
1177 			hv += is->is_dst.i6[2];
1178 			hv += is->is_dst.i6[3];
1179 		}
1180 		hv += is->is_src.i6[1];
1181 		hv += is->is_src.i6[2];
1182 		hv += is->is_src.i6[3];
1183 	}
1184 #endif
1185 
1186 	switch (is->is_p)
1187 	{
1188 #ifdef	USE_INET6
1189 	case IPPROTO_ICMPV6 :
1190 		ic = fin->fin_dp;
1191 
1192 		switch (ic->icmp_type)
1193 		{
1194 		case ICMP6_ECHO_REQUEST :
1195 			is->is_icmp.ici_type = ic->icmp_type;
1196 			hv += (is->is_icmp.ici_id = ic->icmp_id);
1197 			break;
1198 		case ICMP6_MEMBERSHIP_QUERY :
1199 		case ND_ROUTER_SOLICIT :
1200 		case ND_NEIGHBOR_SOLICIT :
1201 		case ICMP6_NI_QUERY :
1202 			is->is_icmp.ici_type = ic->icmp_type;
1203 			break;
1204 		default :
1205 			return NULL;
1206 		}
1207 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_icmp);
1208 		break;
1209 #endif
1210 	case IPPROTO_ICMP :
1211 		ic = fin->fin_dp;
1212 
1213 		switch (ic->icmp_type)
1214 		{
1215 		case ICMP_ECHO :
1216 		case ICMP_TSTAMP :
1217 		case ICMP_IREQ :
1218 		case ICMP_MASKREQ :
1219 			is->is_icmp.ici_type = ic->icmp_type;
1220 			hv += (is->is_icmp.ici_id = ic->icmp_id);
1221 			break;
1222 		default :
1223 			return NULL;
1224 		}
1225 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_icmp);
1226 		break;
1227 
1228 	case IPPROTO_GRE :
1229 		gre = fin->fin_dp;
1230 
1231 		is->is_gre.gs_flags = gre->gr_flags;
1232 		is->is_gre.gs_ptype = gre->gr_ptype;
1233 		if (GRE_REV(is->is_gre.gs_flags) == 1) {
1234 			is->is_call[0] = fin->fin_data[0];
1235 			is->is_call[1] = fin->fin_data[1];
1236 		}
1237 		break;
1238 
1239 	case IPPROTO_TCP :
1240 		tcp = fin->fin_dp;
1241 
1242 		if (tcp->th_flags & TH_RST)
1243 			return NULL;
1244 		/*
1245 		 * The endian of the ports doesn't matter, but the ack and
1246 		 * sequence numbers do as we do mathematics on them later.
1247 		 */
1248 		is->is_sport = htons(fin->fin_data[0]);
1249 		is->is_dport = htons(fin->fin_data[1]);
1250 		if ((flags & (SI_W_DPORT|SI_W_SPORT)) == 0) {
1251 			hv += is->is_sport;
1252 			hv += is->is_dport;
1253 		}
1254 
1255 		/*
1256 		 * If this is a real packet then initialise fields in the
1257 		 * state information structure from the TCP header information.
1258 		 */
1259 
1260 		is->is_maxdwin = 1;
1261 		is->is_maxswin = ntohs(tcp->th_win);
1262 		if (is->is_maxswin == 0)
1263 			is->is_maxswin = 1;
1264 
1265 		if ((fin->fin_flx & FI_IGNORE) == 0) {
1266 			is->is_send = ntohl(tcp->th_seq) + fin->fin_dlen -
1267 				      (TCP_OFF(tcp) << 2) +
1268 				      ((tcp->th_flags & TH_SYN) ? 1 : 0) +
1269 				      ((tcp->th_flags & TH_FIN) ? 1 : 0);
1270 			is->is_maxsend = is->is_send;
1271 
1272 			/*
1273 			 * Window scale option is only present in
1274 			 * SYN/SYN-ACK packet.
1275 			 */
1276 			if ((tcp->th_flags & ~(TH_FIN|TH_ACK|TH_ECNALL)) ==
1277 			    TH_SYN &&
1278 			    (TCP_OFF(tcp) > (sizeof(tcphdr_t) >> 2))) {
1279 				if (fr_tcpoptions(fin, tcp,
1280 					      &is->is_tcp.ts_data[0]) == -1) {
1281 					fin->fin_flx |= FI_BAD;
1282 				}
1283 			}
1284 
1285 			if ((fin->fin_out != 0) && (pass & FR_NEWISN) != 0) {
1286 				fr_checknewisn(fin, is);
1287 				fr_fixoutisn(fin, is);
1288 			}
1289 
1290 			if ((tcp->th_flags & TH_OPENING) == TH_SYN)
1291 				flags |= IS_TCPFSM;
1292 			else {
1293 				is->is_maxdwin = is->is_maxswin * 2;
1294 				is->is_dend = ntohl(tcp->th_ack);
1295 				is->is_maxdend = ntohl(tcp->th_ack);
1296 				is->is_maxdwin *= 2;
1297 			}
1298 		}
1299 
1300 		/*
1301 		 * If we're creating state for a starting connection, start the
1302 		 * timer on it as we'll never see an error if it fails to
1303 		 * connect.
1304 		 */
1305 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_tcp);
1306 		break;
1307 
1308 	case IPPROTO_UDP :
1309 		tcp = fin->fin_dp;
1310 
1311 		is->is_sport = htons(fin->fin_data[0]);
1312 		is->is_dport = htons(fin->fin_data[1]);
1313 		if ((flags & (SI_W_DPORT|SI_W_SPORT)) == 0) {
1314 			hv += tcp->th_dport;
1315 			hv += tcp->th_sport;
1316 		}
1317 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_udp);
1318 		break;
1319 
1320 	default :
1321 		break;
1322 	}
1323 	hv = DOUBLE_HASH(hv, ifs);
1324 	is->is_hv = hv;
1325 	is->is_rule = fr;
1326 	is->is_flags = flags & IS_INHERITED;
1327 
1328 	/*
1329 	 * Look for identical state.
1330 	 */
1331 	for (is = ifs->ifs_ips_table[is->is_hv % ifs->ifs_fr_statesize];
1332 	     is != NULL;
1333 	     is = is->is_hnext) {
1334 		if (fr_matchstates(&ips, is) == 1)
1335 			break;
1336 	}
1337 
1338 	/*
1339 	 * we've found a matching state -> state already exists,
1340 	 * we are not going to add a duplicate record.
1341 	 */
1342 	if (is != NULL)
1343 		return NULL;
1344 
1345 	if (ifs->ifs_ips_stats.iss_bucketlen[hv] >= ifs->ifs_fr_state_maxbucket) {
1346 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_bucketfull);
1347 		return NULL;
1348 	}
1349 	KMALLOC(is, ipstate_t *);
1350 	if (is == NULL) {
1351 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_nomem);
1352 		return NULL;
1353 	}
1354 	bcopy((char *)&ips, (char *)is, sizeof(*is));
1355 	/*
1356 	 * Do not do the modulous here, it is done in fr_stinsert().
1357 	 */
1358 	if (fr != NULL) {
1359 		(void) strncpy(is->is_group, fr->fr_group, FR_GROUPLEN);
1360 		if (fr->fr_age[0] != 0) {
1361 			is->is_tqehead[0] =
1362 			    fr_addtimeoutqueue(&ifs->ifs_ips_utqe,
1363 					       fr->fr_age[0], ifs);
1364 			is->is_sti.tqe_flags |= TQE_RULEBASED;
1365 		}
1366 		if (fr->fr_age[1] != 0) {
1367 			is->is_tqehead[1] =
1368 			    fr_addtimeoutqueue(&ifs->ifs_ips_utqe,
1369 					       fr->fr_age[1], ifs);
1370 			is->is_sti.tqe_flags |= TQE_RULEBASED;
1371 		}
1372 		is->is_tag = fr->fr_logtag;
1373 
1374 		is->is_ifp[(out << 1) + 1] = fr->fr_ifas[1];
1375 		is->is_ifp[(1 - out) << 1] = fr->fr_ifas[2];
1376 		is->is_ifp[((1 - out) << 1) + 1] = fr->fr_ifas[3];
1377 
1378 		if (((ifp = fr->fr_ifas[1]) != NULL) &&
1379 		    (ifp != (void *)-1)) {
1380 			COPYIFNAME(ifp, is->is_ifname[(out << 1) + 1], fr->fr_v);
1381 		}
1382 		if (((ifp = fr->fr_ifas[2]) != NULL) &&
1383 		    (ifp != (void *)-1)) {
1384 			COPYIFNAME(ifp, is->is_ifname[(1 - out) << 1], fr->fr_v);
1385 		}
1386 		if (((ifp = fr->fr_ifas[3]) != NULL) &&
1387 		    (ifp != (void *)-1)) {
1388 			COPYIFNAME(ifp, is->is_ifname[((1 - out) << 1) + 1], fr->fr_v);
1389 		}
1390 	}
1391 
1392 	is->is_ifp[out << 1] = fin->fin_ifp;
1393 	if (fin->fin_ifp != NULL) {
1394 		COPYIFNAME(fin->fin_ifp, is->is_ifname[out << 1], fr->fr_v);
1395 	}
1396 
1397 	/*
1398 	 * It may seem strange to set is_ref to 2, but fr_check() will call
1399 	 * fr_statederef() after calling fr_addstate() and the idea is to
1400 	 * have it exist at the end of fr_check() with is_ref == 1.
1401 	 */
1402 	is->is_ref = 2;
1403 	is->is_pkts[0] = 0, is->is_bytes[0] = 0;
1404 	is->is_pkts[1] = 0, is->is_bytes[1] = 0;
1405 	is->is_pkts[2] = 0, is->is_bytes[2] = 0;
1406 	is->is_pkts[3] = 0, is->is_bytes[3] = 0;
1407 	if ((fin->fin_flx & FI_IGNORE) == 0) {
1408 		is->is_pkts[out] = 1;
1409 		is->is_bytes[out] = fin->fin_plen;
1410 		is->is_flx[out][0] = fin->fin_flx & FI_CMP;
1411 		is->is_flx[out][0] &= ~FI_OOW;
1412 	}
1413 
1414 	if (pass & FR_STSTRICT)
1415 		is->is_flags |= IS_STRICT;
1416 
1417 	if (pass & FR_STATESYNC)
1418 		is->is_flags |= IS_STATESYNC;
1419 
1420 	if (flags & (SI_WILDP|SI_WILDA)) {
1421 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_wild);
1422 	}
1423 	is->is_rulen = fin->fin_rule;
1424 
1425 
1426 	if (pass & FR_LOGFIRST)
1427 		is->is_pass &= ~(FR_LOGFIRST|FR_LOG);
1428 
1429 	READ_ENTER(&ifs->ifs_ipf_state);
1430 	is->is_me = stsave;
1431 
1432 	fr_stinsert(is, fin->fin_rev, ifs);
1433 
1434 	if (fin->fin_p == IPPROTO_TCP) {
1435 		/*
1436 		* If we're creating state for a starting connection, start the
1437 		* timer on it as we'll never see an error if it fails to
1438 		* connect.
1439 		*/
1440 		(void) fr_tcp_age(&is->is_sti, fin, ifs->ifs_ips_tqtqb,
1441 				  is->is_flags);
1442 		MUTEX_EXIT(&is->is_lock);
1443 #ifdef	IPFILTER_SCAN
1444 		if ((is->is_flags & SI_CLONE) == 0)
1445 			(void) ipsc_attachis(is);
1446 #endif
1447 	} else {
1448 		MUTEX_EXIT(&is->is_lock);
1449 	}
1450 #ifdef	IPFILTER_SYNC
1451 	if ((is->is_flags & IS_STATESYNC) && ((is->is_flags & SI_CLONE) == 0))
1452 		is->is_sync = ipfsync_new(SMC_STATE, fin, is);
1453 #endif
1454 	if (ifs->ifs_ipstate_logging)
1455 		ipstate_log(is, ISL_NEW, ifs);
1456 
1457 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
1458 	fin->fin_state = is;
1459 	fin->fin_rev = IP6_NEQ(&is->is_dst, &fin->fin_daddr);
1460 	fin->fin_flx |= FI_STATE;
1461 	if (fin->fin_flx & FI_FRAG)
1462 		(void) fr_newfrag(fin, pass ^ FR_KEEPSTATE);
1463 
1464 	return is;
1465 }
1466 
1467 
1468 /* ------------------------------------------------------------------------ */
1469 /* Function:    fr_tcpoptions                                               */
1470 /* Returns:     int - 1 == packet matches state entry, 0 == it does not     */
1471 /* Parameters:  fin(I) - pointer to packet information                      */
1472 /*              tcp(I) - pointer to TCP packet header                       */
1473 /*              td(I)  - pointer to TCP data held as part of the state      */
1474 /*                                                                          */
1475 /* Look after the TCP header for any options and deal with those that are   */
1476 /* present.  Record details about those that we recogise.                   */
1477 /* ------------------------------------------------------------------------ */
1478 static int fr_tcpoptions(fin, tcp, td)
1479 fr_info_t *fin;
1480 tcphdr_t *tcp;
1481 tcpdata_t *td;
1482 {
1483 	int off, mlen, ol, i, len, retval;
1484 	char buf[64], *s, opt;
1485 	mb_t *m = NULL;
1486 
1487 	len = (TCP_OFF(tcp) << 2);
1488 	if (fin->fin_dlen < len)
1489 		return 0;
1490 	len -= sizeof(*tcp);
1491 
1492 	off = fin->fin_plen - fin->fin_dlen + sizeof(*tcp) + fin->fin_ipoff;
1493 
1494 	m = fin->fin_m;
1495 	mlen = MSGDSIZE(m) - off;
1496 	if (len > mlen) {
1497 		len = mlen;
1498 		retval = 0;
1499 	} else {
1500 		retval = 1;
1501 	}
1502 
1503 	COPYDATA(m, off, len, buf);
1504 
1505 	for (s = buf; len > 0; ) {
1506 		opt = *s;
1507 		if (opt == TCPOPT_EOL)
1508 			break;
1509 		else if (opt == TCPOPT_NOP)
1510 			ol = 1;
1511 		else {
1512 			if (len < 2)
1513 				break;
1514 			ol = (int)*(s + 1);
1515 			if (ol < 2 || ol > len)
1516 				break;
1517 
1518 			/*
1519 			 * Extract the TCP options we are interested in out of
1520 			 * the header and store them in the the tcpdata struct.
1521 			 */
1522 			switch (opt)
1523 			{
1524 			case TCPOPT_WINDOW :
1525 				if (ol == TCPOLEN_WINDOW) {
1526 					i = (int)*(s + 2);
1527 					if (i > TCP_WSCALE_MAX)
1528 						i = TCP_WSCALE_MAX;
1529 					else if (i < 0)
1530 						i = 0;
1531 					td->td_winscale = i;
1532 					td->td_winflags |= TCP_WSCALE_SEEN|
1533 							   TCP_WSCALE_FIRST;
1534 				} else
1535 					retval = -1;
1536 				break;
1537 			case TCPOPT_MAXSEG :
1538 				/*
1539 				 * So, if we wanted to set the TCP MAXSEG,
1540 				 * it should be done here...
1541 				 */
1542 				if (ol == TCPOLEN_MAXSEG) {
1543 					i = (int)*(s + 2);
1544 					i <<= 8;
1545 					i += (int)*(s + 3);
1546 					td->td_maxseg = i;
1547 				} else
1548 					retval = -1;
1549 				break;
1550 			case TCPOPT_SACK_PERMITTED :
1551 				if (ol == TCPOLEN_SACK_PERMITTED)
1552 					td->td_winflags |= TCP_SACK_PERMIT;
1553 				else
1554 					retval = -1;
1555 				break;
1556 			}
1557 		}
1558 		len -= ol;
1559 		s += ol;
1560 	}
1561 	return retval;
1562 }
1563 
1564 
1565 /* ------------------------------------------------------------------------ */
1566 /* Function:    fr_tcpstate                                                 */
1567 /* Returns:     int - 1 == packet matches state entry, 0 == it does not     */
1568 /* Parameters:  fin(I)   - pointer to packet information                    */
1569 /*              tcp(I)   - pointer to TCP packet header                     */
1570 /*              is(I)  - pointer to master state structure                  */
1571 /*                                                                          */
1572 /* Check to see if a packet with TCP headers fits within the TCP window.    */
1573 /* Change timeout depending on whether new packet is a SYN-ACK returning    */
1574 /* for a SYN or a RST or FIN which indicate time to close up shop.          */
1575 /* ------------------------------------------------------------------------ */
1576 static int fr_tcpstate(fin, tcp, is)
1577 fr_info_t *fin;
1578 tcphdr_t *tcp;
1579 ipstate_t *is;
1580 {
1581 	int source, ret = 0, flags;
1582 	tcpdata_t  *fdata, *tdata;
1583 	ipf_stack_t *ifs = fin->fin_ifs;
1584 
1585 	source = !fin->fin_rev;
1586 	if (((is->is_flags & IS_TCPFSM) != 0) && (source == 1) &&
1587 	    (ntohs(is->is_sport) != fin->fin_data[0]))
1588 		source = 0;
1589 	fdata = &is->is_tcp.ts_data[!source];
1590 	tdata = &is->is_tcp.ts_data[source];
1591 
1592 	MUTEX_ENTER(&is->is_lock);
1593 
1594 	/*
1595 	 * If a SYN packet is received for a connection that is in a half
1596 	 * closed state, then move its state entry to deletetq. In such case
1597 	 * the SYN packet will be consequently dropped. This allows new state
1598 	 * entry to be created with a retransmited SYN packet.
1599 	 */
1600 	if ((tcp->th_flags & TH_OPENING) == TH_SYN) {
1601 		if (((is->is_state[source] > IPF_TCPS_ESTABLISHED) ||
1602 		    (is->is_state[source] == IPF_TCPS_CLOSED)) &&
1603 		    ((is->is_state[!source] > IPF_TCPS_ESTABLISHED) ||
1604 		    (is->is_state[!source] == IPF_TCPS_CLOSED))) {
1605 			/*
1606 			 * Do not update is->is_sti.tqe_die in case state entry
1607 			 * is already present in deletetq. It prevents state
1608 			 * entry ttl update by retransmitted SYN packets, which
1609 			 * may arrive before timer tick kicks off. The SYN
1610 			 * packet will be dropped again.
1611 			 */
1612 			if (is->is_sti.tqe_ifq != &ifs->ifs_ips_deletetq)
1613 				fr_movequeue(&is->is_sti, is->is_sti.tqe_ifq,
1614 					&fin->fin_ifs->ifs_ips_deletetq,
1615 					fin->fin_ifs);
1616 
1617 			MUTEX_EXIT(&is->is_lock);
1618 			return 0;
1619 		}
1620 	}
1621 
1622 	if (fr_tcpinwindow(fin, fdata, tdata, tcp, is->is_flags)) {
1623 #ifdef	IPFILTER_SCAN
1624 		if (is->is_flags & (IS_SC_CLIENT|IS_SC_SERVER)) {
1625 			ipsc_packet(fin, is);
1626 			if (FR_ISBLOCK(is->is_pass)) {
1627 				MUTEX_EXIT(&is->is_lock);
1628 				return 1;
1629 			}
1630 		}
1631 #endif
1632 
1633 		/*
1634 		 * Nearing end of connection, start timeout.
1635 		 */
1636 		ret = fr_tcp_age(&is->is_sti, fin, ifs->ifs_ips_tqtqb,
1637 				 is->is_flags);
1638 		if (ret == 0) {
1639 			MUTEX_EXIT(&is->is_lock);
1640 			return 0;
1641 		}
1642 
1643 		/*
1644 		 * set s0's as appropriate.  Use syn-ack packet as it
1645 		 * contains both pieces of required information.
1646 		 */
1647 		/*
1648 		 * Window scale option is only present in SYN/SYN-ACK packet.
1649 		 * Compare with ~TH_FIN to mask out T/TCP setups.
1650 		 */
1651 		flags = tcp->th_flags & ~(TH_FIN|TH_ECNALL);
1652 		if (flags == (TH_SYN|TH_ACK)) {
1653 			is->is_s0[source] = ntohl(tcp->th_ack);
1654 			is->is_s0[!source] = ntohl(tcp->th_seq) + 1;
1655 			if (TCP_OFF(tcp) > (sizeof(tcphdr_t) >> 2)) {
1656 				(void) fr_tcpoptions(fin, tcp, fdata);
1657 			}
1658 			if ((fin->fin_out != 0) && (is->is_pass & FR_NEWISN))
1659 				fr_checknewisn(fin, is);
1660 		} else if (flags == TH_SYN) {
1661 			is->is_s0[source] = ntohl(tcp->th_seq) + 1;
1662 			if ((TCP_OFF(tcp) > (sizeof(tcphdr_t) >> 2)))
1663 				(void) fr_tcpoptions(fin, tcp, tdata);
1664 
1665 			if ((fin->fin_out != 0) && (is->is_pass & FR_NEWISN))
1666 				fr_checknewisn(fin, is);
1667 
1668 		}
1669 		ret = 1;
1670 	} else
1671 		fin->fin_flx |= FI_OOW;
1672 	MUTEX_EXIT(&is->is_lock);
1673 	return ret;
1674 }
1675 
1676 
1677 /* ------------------------------------------------------------------------ */
1678 /* Function:    fr_checknewisn                                              */
1679 /* Returns:     Nil                                                         */
1680 /* Parameters:  fin(I)   - pointer to packet information                    */
1681 /*              is(I)  - pointer to master state structure                  */
1682 /*                                                                          */
1683 /* Check to see if this TCP connection is expecting and needs a new         */
1684 /* sequence number for a particular direction of the connection.            */
1685 /*                                                                          */
1686 /* NOTE: This does not actually change the sequence numbers, only gets new  */
1687 /* one ready.                                                               */
1688 /* ------------------------------------------------------------------------ */
1689 static void fr_checknewisn(fin, is)
1690 fr_info_t *fin;
1691 ipstate_t *is;
1692 {
1693 	u_32_t sumd, old, new;
1694 	tcphdr_t *tcp;
1695 	int i;
1696 
1697 	i = fin->fin_rev;
1698 	tcp = fin->fin_dp;
1699 
1700 	if (((i == 0) && !(is->is_flags & IS_ISNSYN)) ||
1701 	    ((i == 1) && !(is->is_flags & IS_ISNACK))) {
1702 		old = ntohl(tcp->th_seq);
1703 		new = fr_newisn(fin);
1704 		is->is_isninc[i] = new - old;
1705 		CALC_SUMD(old, new, sumd);
1706 		is->is_sumd[i] = (sumd & 0xffff) + (sumd >> 16);
1707 
1708 		is->is_flags |= ((i == 0) ? IS_ISNSYN : IS_ISNACK);
1709 	}
1710 }
1711 
1712 
1713 /* ------------------------------------------------------------------------ */
1714 /* Function:    fr_tcpinwindow                                              */
1715 /* Returns:     int - 1 == packet inside TCP "window", 0 == not inside.     */
1716 /* Parameters:  fin(I)   - pointer to packet information                    */
1717 /*              fdata(I) - pointer to tcp state informatio (forward)        */
1718 /*              tdata(I) - pointer to tcp state informatio (reverse)        */
1719 /*              tcp(I)   - pointer to TCP packet header                     */
1720 /*                                                                          */
1721 /* Given a packet has matched addresses and ports, check to see if it is    */
1722 /* within the TCP data window.  In a show of generosity, allow packets that */
1723 /* are within the window space behind the current sequence # as well.       */
1724 /* ------------------------------------------------------------------------ */
1725 int fr_tcpinwindow(fin, fdata, tdata, tcp, flags)
1726 fr_info_t *fin;
1727 tcpdata_t  *fdata, *tdata;
1728 tcphdr_t *tcp;
1729 int flags;
1730 {
1731 	tcp_seq seq, ack, end;
1732 	int ackskew, tcpflags;
1733 	u_32_t win, maxwin;
1734 	int dsize, inseq;
1735 
1736 	/*
1737 	 * Find difference between last checked packet and this packet.
1738 	 */
1739 	tcpflags = tcp->th_flags;
1740 	seq = ntohl(tcp->th_seq);
1741 	ack = ntohl(tcp->th_ack);
1742 
1743 	if (tcpflags & TH_SYN)
1744 		win = ntohs(tcp->th_win);
1745 	else
1746 		win = ntohs(tcp->th_win) << fdata->td_winscale;
1747 
1748 	/*
1749 	 * win 0 means the receiving endpoint has closed the window, because it
1750 	 * has not enough memory to receive data from sender. In such case we
1751 	 * are pretending window size to be 1 to let TCP probe data through.
1752 	 * TCP probe data can be either 0 or 1 octet of data, the RFC does not
1753 	 * state this accurately, so we have to allow 1 octet (win = 1) even if
1754 	 * the window is closed (win == 0).
1755 	 */
1756 	if (win == 0)
1757 		win = 1;
1758 
1759 	dsize = fin->fin_dlen - (TCP_OFF(tcp) << 2) +
1760 	        ((tcpflags & TH_SYN) ? 1 : 0) + ((tcpflags & TH_FIN) ? 1 : 0);
1761 
1762 	/*
1763 	 * if window scaling is present, the scaling is only allowed
1764 	 * for windows not in the first SYN packet. In that packet the
1765 	 * window is 65535 to specify the largest window possible
1766 	 * for receivers not implementing the window scale option.
1767 	 * Currently, we do not assume TTCP here. That means that
1768 	 * if we see a second packet from a host (after the initial
1769 	 * SYN), we can assume that the receiver of the SYN did
1770 	 * already send back the SYN/ACK (and thus that we know if
1771 	 * the receiver also does window scaling)
1772 	 */
1773 	if (!(tcpflags & TH_SYN) && (fdata->td_winflags & TCP_WSCALE_FIRST)) {
1774 		fdata->td_maxwin = win;
1775 	}
1776 
1777 	end = seq + dsize;
1778 
1779 	if ((fdata->td_end == 0) &&
1780 	    (!(flags & IS_TCPFSM) ||
1781 	     ((tcpflags & TH_OPENING) == TH_OPENING))) {
1782 		/*
1783 		 * Must be a (outgoing) SYN-ACK in reply to a SYN.
1784 		 */
1785 		fdata->td_end = end - 1;
1786 		fdata->td_maxwin = 1;
1787 		fdata->td_maxend = end + win;
1788 	}
1789 
1790 	if (!(tcpflags & TH_ACK)) {  /* Pretend an ack was sent */
1791 		ack = tdata->td_end;
1792 	} else if (((tcpflags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) &&
1793 		   (ack == 0)) {
1794 		/* gross hack to get around certain broken tcp stacks */
1795 		ack = tdata->td_end;
1796 	}
1797 
1798 	maxwin = tdata->td_maxwin;
1799 	ackskew = tdata->td_end - ack;
1800 
1801 	/*
1802 	 * Strict sequencing only allows in-order delivery.
1803 	 */
1804 	if ((flags & IS_STRICT) != 0) {
1805 		if (seq != fdata->td_end) {
1806 			return 0;
1807 		}
1808 	}
1809 
1810 #define	SEQ_GE(a,b)	((int)((a) - (b)) >= 0)
1811 #define	SEQ_GT(a,b)	((int)((a) - (b)) > 0)
1812 	inseq = 0;
1813 	if (
1814 #if defined(_KERNEL)
1815 	    (SEQ_GE(fdata->td_maxend, end)) &&
1816 	    (SEQ_GE(seq, fdata->td_end - maxwin)) &&
1817 #endif
1818 /* XXX what about big packets */
1819 #define MAXACKWINDOW 66000
1820 	    (-ackskew <= (MAXACKWINDOW << fdata->td_winscale)) &&
1821 	    ( ackskew <= (MAXACKWINDOW << fdata->td_winscale))) {
1822 		inseq = 1;
1823 	/*
1824 	 * Microsoft Windows will send the next packet to the right of the
1825 	 * window if SACK is in use.
1826 	 */
1827 	} else if ((seq == fdata->td_maxend) && (ackskew == 0) &&
1828 	    (fdata->td_winflags & TCP_SACK_PERMIT) &&
1829 	    (tdata->td_winflags & TCP_SACK_PERMIT)) {
1830 		inseq = 1;
1831 	/*
1832 	 * RST ACK with SEQ equal to 0 is sent by some OSes (i.e. Solaris) as a
1833 	 * response to initial SYN packet, when  there is no application
1834 	 * listeing to on a port, where the SYN packet has came to.
1835 	 */
1836 	} else if ((seq == 0) && (tcpflags == (TH_RST|TH_ACK)) &&
1837 		   (ackskew >= -1) && (ackskew <= 1)) {
1838 		inseq = 1;
1839 	} else if (!(flags & IS_TCPFSM)) {
1840 
1841 		if (!(fdata->td_winflags &
1842 			    (TCP_WSCALE_SEEN|TCP_WSCALE_FIRST))) {
1843 			/*
1844 			 * No TCPFSM and no window scaling, so make some
1845 			 * extra guesses.
1846 			 */
1847 			if ((seq == fdata->td_maxend) && (ackskew == 0))
1848 				inseq = 1;
1849 			else if (SEQ_GE(seq + maxwin, fdata->td_end - maxwin))
1850 				inseq = 1;
1851 		}
1852 	}
1853 
1854 	if (inseq) {
1855 		/* if ackskew < 0 then this should be due to fragmented
1856 		 * packets. There is no way to know the length of the
1857 		 * total packet in advance.
1858 		 * We do know the total length from the fragment cache though.
1859 		 * Note however that there might be more sessions with
1860 		 * exactly the same source and destination parameters in the
1861 		 * state cache (and source and destination is the only stuff
1862 		 * that is saved in the fragment cache). Note further that
1863 		 * some TCP connections in the state cache are hashed with
1864 		 * sport and dport as well which makes it not worthwhile to
1865 		 * look for them.
1866 		 * Thus, when ackskew is negative but still seems to belong
1867 		 * to this session, we bump up the destinations end value.
1868 		 */
1869 		if (ackskew < 0)
1870 			tdata->td_end = ack;
1871 
1872 		/* update max window seen */
1873 		if (fdata->td_maxwin < win)
1874 			fdata->td_maxwin = win;
1875 		if (SEQ_GT(end, fdata->td_end))
1876 			fdata->td_end = end;
1877 		if (SEQ_GE(ack + win, tdata->td_maxend))
1878 			tdata->td_maxend = ack + win;
1879 		return 1;
1880 	}
1881 	fin->fin_flx |= FI_OOW;
1882 	return 0;
1883 }
1884 
1885 
1886 /* ------------------------------------------------------------------------ */
1887 /* Function:    fr_stclone                                                  */
1888 /* Returns:     ipstate_t* - NULL == cloning failed,                        */
1889 /*                           else pointer to new state structure            */
1890 /* Parameters:  fin(I) - pointer to packet information                      */
1891 /*              tcp(I) - pointer to TCP/UDP header                          */
1892 /*              is(I)  - pointer to master state structure                  */
1893 /*                                                                          */
1894 /* Create a "duplcate" state table entry from the master.                   */
1895 /* ------------------------------------------------------------------------ */
1896 static ipstate_t *fr_stclone(fin, tcp, is)
1897 fr_info_t *fin;
1898 tcphdr_t *tcp;
1899 ipstate_t *is;
1900 {
1901 	ipstate_t *clone;
1902 	u_32_t send;
1903 	ipf_stack_t *ifs = fin->fin_ifs;
1904 
1905 	if (ifs->ifs_ips_num == ifs->ifs_fr_statemax) {
1906 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_max);
1907 		ifs->ifs_fr_state_doflush = 1;
1908 		return NULL;
1909 	}
1910 	KMALLOC(clone, ipstate_t *);
1911 	if (clone == NULL)
1912 		return NULL;
1913 	bcopy((char *)is, (char *)clone, sizeof(*clone));
1914 
1915 	MUTEX_NUKE(&clone->is_lock);
1916 
1917 	clone->is_die = ONE_DAY + ifs->ifs_fr_ticks;
1918 	clone->is_state[0] = 0;
1919 	clone->is_state[1] = 0;
1920 	send = ntohl(tcp->th_seq) + fin->fin_dlen - (TCP_OFF(tcp) << 2) +
1921 		((tcp->th_flags & TH_SYN) ? 1 : 0) +
1922 		((tcp->th_flags & TH_FIN) ? 1 : 0);
1923 
1924 	if (fin->fin_rev == 1) {
1925 		clone->is_dend = send;
1926 		clone->is_maxdend = send;
1927 		clone->is_send = 0;
1928 		clone->is_maxswin = 1;
1929 		clone->is_maxdwin = ntohs(tcp->th_win);
1930 		if (clone->is_maxdwin == 0)
1931 			clone->is_maxdwin = 1;
1932 	} else {
1933 		clone->is_send = send;
1934 		clone->is_maxsend = send;
1935 		clone->is_dend = 0;
1936 		clone->is_maxdwin = 1;
1937 		clone->is_maxswin = ntohs(tcp->th_win);
1938 		if (clone->is_maxswin == 0)
1939 			clone->is_maxswin = 1;
1940 	}
1941 
1942 	clone->is_flags &= ~SI_CLONE;
1943 	clone->is_flags |= SI_CLONED;
1944 	fr_stinsert(clone, fin->fin_rev, ifs);
1945 	clone->is_ref = 2;
1946 	if (clone->is_p == IPPROTO_TCP) {
1947 		(void) fr_tcp_age(&clone->is_sti, fin, ifs->ifs_ips_tqtqb,
1948 				  clone->is_flags);
1949 	}
1950 	MUTEX_EXIT(&clone->is_lock);
1951 #ifdef	IPFILTER_SCAN
1952 	(void) ipsc_attachis(is);
1953 #endif
1954 #ifdef	IPFILTER_SYNC
1955 	if (is->is_flags & IS_STATESYNC)
1956 		clone->is_sync = ipfsync_new(SMC_STATE, fin, clone);
1957 #endif
1958 	return clone;
1959 }
1960 
1961 
1962 /* ------------------------------------------------------------------------ */
1963 /* Function:    fr_matchsrcdst                                              */
1964 /* Returns:     Nil                                                         */
1965 /* Parameters:  fin(I) - pointer to packet information                      */
1966 /*              is(I)  - pointer to state structure                         */
1967 /*              src(I) - pointer to source address                          */
1968 /*              dst(I) - pointer to destination address                     */
1969 /*              tcp(I) - pointer to TCP/UDP header                          */
1970 /*                                                                          */
1971 /* Match a state table entry against an IP packet.  The logic below is that */
1972 /* ret gets set to one if the match succeeds, else remains 0.  If it is     */
1973 /* still 0 after the test. no match.                                        */
1974 /* ------------------------------------------------------------------------ */
1975 static ipstate_t *fr_matchsrcdst(fin, is, src, dst, tcp, cmask)
1976 fr_info_t *fin;
1977 ipstate_t *is;
1978 i6addr_t *src, *dst;
1979 tcphdr_t *tcp;
1980 u_32_t cmask;
1981 {
1982 	int ret = 0, rev, out, flags, flx = 0, idx;
1983 	u_short sp, dp;
1984 	u_32_t cflx;
1985 	void *ifp;
1986 	ipf_stack_t *ifs = fin->fin_ifs;
1987 
1988 	rev = IP6_NEQ(&is->is_dst, dst);
1989 	ifp = fin->fin_ifp;
1990 	out = fin->fin_out;
1991 	flags = is->is_flags;
1992 	sp = 0;
1993 	dp = 0;
1994 
1995 	if (tcp != NULL) {
1996 		sp = htons(fin->fin_sport);
1997 		dp = ntohs(fin->fin_dport);
1998 	}
1999 	if (!rev) {
2000 		if (tcp != NULL) {
2001 			if (!(flags & SI_W_SPORT) && (sp != is->is_sport))
2002 				rev = 1;
2003 			else if (!(flags & SI_W_DPORT) && (dp != is->is_dport))
2004 				rev = 1;
2005 		}
2006 	}
2007 
2008 	idx = (out << 1) + rev;
2009 
2010 	/*
2011 	 * If the interface for this 'direction' is set, make sure it matches.
2012 	 * An interface name that is not set matches any, as does a name of *.
2013 	 */
2014 	if ((is->is_ifp[idx] == NULL &&
2015 	    (*is->is_ifname[idx] == '\0' || *is->is_ifname[idx] == '*')) ||
2016 	    is->is_ifp[idx] == ifp)
2017 		ret = 1;
2018 
2019 	if (ret == 0)
2020 		return NULL;
2021 	ret = 0;
2022 
2023 	/*
2024 	 * Match addresses and ports.
2025 	 */
2026 	if (rev == 0) {
2027 		if ((IP6_EQ(&is->is_dst, dst) || (flags & SI_W_DADDR)) &&
2028 		    (IP6_EQ(&is->is_src, src) || (flags & SI_W_SADDR))) {
2029 			if (tcp) {
2030 				if ((sp == is->is_sport || flags & SI_W_SPORT)&&
2031 				    (dp == is->is_dport || flags & SI_W_DPORT))
2032 					ret = 1;
2033 			} else {
2034 				ret = 1;
2035 			}
2036 		}
2037 	} else {
2038 		if ((IP6_EQ(&is->is_dst, src) || (flags & SI_W_DADDR)) &&
2039 		    (IP6_EQ(&is->is_src, dst) || (flags & SI_W_SADDR))) {
2040 			if (tcp) {
2041 				if ((dp == is->is_sport || flags & SI_W_SPORT)&&
2042 				    (sp == is->is_dport || flags & SI_W_DPORT))
2043 					ret = 1;
2044 			} else {
2045 				ret = 1;
2046 			}
2047 		}
2048 	}
2049 
2050 	if (ret == 0)
2051 		return NULL;
2052 
2053 	/*
2054 	 * Whether or not this should be here, is questionable, but the aim
2055 	 * is to get this out of the main line.
2056 	 */
2057 	if (tcp == NULL)
2058 		flags = is->is_flags & ~(SI_WILDP|SI_NEWFR|SI_CLONE|SI_CLONED);
2059 
2060 	/*
2061 	 * Only one of the source or destination address can be flaged as a
2062 	 * wildcard.  Fill in the missing address, if set.
2063 	 * For IPv6, if the address being copied in is multicast, then
2064 	 * don't reset the wild flag - multicast causes it to be set in the
2065 	 * first place!
2066 	 */
2067 	if ((flags & (SI_W_SADDR|SI_W_DADDR))) {
2068 		fr_ip_t *fi = &fin->fin_fi;
2069 
2070 		if ((flags & SI_W_SADDR) != 0) {
2071 			if (rev == 0) {
2072 #ifdef USE_INET6
2073 				if (is->is_v == 6 &&
2074 				    IN6_IS_ADDR_MULTICAST(&fi->fi_src.in6))
2075 					/*EMPTY*/;
2076 				else
2077 #endif
2078 				{
2079 					is->is_src = fi->fi_src;
2080 					is->is_flags &= ~SI_W_SADDR;
2081 				}
2082 			} else {
2083 #ifdef USE_INET6
2084 				if (is->is_v == 6 &&
2085 				    IN6_IS_ADDR_MULTICAST(&fi->fi_dst.in6))
2086 					/*EMPTY*/;
2087 				else
2088 #endif
2089 				{
2090 					is->is_src = fi->fi_dst;
2091 					is->is_flags &= ~SI_W_SADDR;
2092 				}
2093 			}
2094 		} else if ((flags & SI_W_DADDR) != 0) {
2095 			if (rev == 0) {
2096 #ifdef USE_INET6
2097 				if (is->is_v == 6 &&
2098 				    IN6_IS_ADDR_MULTICAST(&fi->fi_dst.in6))
2099 					/*EMPTY*/;
2100 				else
2101 #endif
2102 				{
2103 					is->is_dst = fi->fi_dst;
2104 					is->is_flags &= ~SI_W_DADDR;
2105 				}
2106 			} else {
2107 #ifdef USE_INET6
2108 				if (is->is_v == 6 &&
2109 				    IN6_IS_ADDR_MULTICAST(&fi->fi_src.in6))
2110 					/*EMPTY*/;
2111 				else
2112 #endif
2113 				{
2114 					is->is_dst = fi->fi_src;
2115 					is->is_flags &= ~SI_W_DADDR;
2116 				}
2117 			}
2118 		}
2119 		if ((is->is_flags & (SI_WILDA|SI_WILDP)) == 0) {
2120 			ATOMIC_DECL(ifs->ifs_ips_stats.iss_wild);
2121 		}
2122 	}
2123 
2124 	flx = fin->fin_flx & cmask;
2125 	cflx = is->is_flx[out][rev];
2126 
2127 	/*
2128 	 * Match up any flags set from IP options.
2129 	 */
2130 	if ((cflx && (flx != (cflx & cmask))) ||
2131 	    ((fin->fin_optmsk & is->is_optmsk[rev]) != is->is_opt[rev]) ||
2132 	    ((fin->fin_secmsk & is->is_secmsk) != is->is_sec) ||
2133 	    ((fin->fin_auth & is->is_authmsk) != is->is_auth))
2134 		return NULL;
2135 
2136 	/*
2137 	 * Only one of the source or destination port can be flagged as a
2138 	 * wildcard.  When filling it in, fill in a copy of the matched entry
2139 	 * if it has the cloning flag set.
2140 	 */
2141 	if ((fin->fin_flx & FI_IGNORE) != 0) {
2142 		fin->fin_rev = rev;
2143 		return is;
2144 	}
2145 
2146 	if ((flags & (SI_W_SPORT|SI_W_DPORT))) {
2147 		if ((flags & SI_CLONE) != 0) {
2148 			ipstate_t *clone;
2149 
2150 			clone = fr_stclone(fin, tcp, is);
2151 			if (clone == NULL)
2152 				return NULL;
2153 			is = clone;
2154 		} else {
2155 			ATOMIC_DECL(ifs->ifs_ips_stats.iss_wild);
2156 		}
2157 
2158 		if ((flags & SI_W_SPORT) != 0) {
2159 			if (rev == 0) {
2160 				is->is_sport = sp;
2161 				is->is_send = ntohl(tcp->th_seq);
2162 			} else {
2163 				is->is_sport = dp;
2164 				is->is_send = ntohl(tcp->th_ack);
2165 			}
2166 			is->is_maxsend = is->is_send + 1;
2167 		} else if ((flags & SI_W_DPORT) != 0) {
2168 			if (rev == 0) {
2169 				is->is_dport = dp;
2170 				is->is_dend = ntohl(tcp->th_ack);
2171 			} else {
2172 				is->is_dport = sp;
2173 				is->is_dend = ntohl(tcp->th_seq);
2174 			}
2175 			is->is_maxdend = is->is_dend + 1;
2176 		}
2177 		is->is_flags &= ~(SI_W_SPORT|SI_W_DPORT);
2178 		if ((flags & SI_CLONED) && ifs->ifs_ipstate_logging)
2179 			ipstate_log(is, ISL_CLONE, ifs);
2180 	}
2181 
2182 	ret = -1;
2183 
2184 	if (is->is_flx[out][rev] == 0) {
2185 		is->is_flx[out][rev] = flx;
2186 		is->is_opt[rev] = fin->fin_optmsk;
2187 		if (is->is_v == 6) {
2188 			is->is_opt[rev] &= ~0x8;
2189 			is->is_optmsk[rev] &= ~0x8;
2190 		}
2191 	}
2192 
2193 	/*
2194 	 * Check if the interface name for this "direction" is set and if not,
2195 	 * fill it in.
2196 	 */
2197 	if (is->is_ifp[idx] == NULL &&
2198 	    (*is->is_ifname[idx] == '\0' || *is->is_ifname[idx] == '*')) {
2199 		is->is_ifp[idx] = ifp;
2200 		COPYIFNAME(ifp, is->is_ifname[idx], fin->fin_v);
2201 	}
2202 	fin->fin_rev = rev;
2203 	return is;
2204 }
2205 
2206 
2207 /* ------------------------------------------------------------------------ */
2208 /* Function:    fr_checkicmpmatchingstate                                   */
2209 /* Returns:     Nil                                                         */
2210 /* Parameters:  fin(I) - pointer to packet information                      */
2211 /*                                                                          */
2212 /* If we've got an ICMP error message, using the information stored in the  */
2213 /* ICMP packet, look for a matching state table entry.                      */
2214 /*                                                                          */
2215 /* If we return NULL then no lock on ipf_state is held.                     */
2216 /* If we return non-null then a read-lock on ipf_state is held.             */
2217 /* ------------------------------------------------------------------------ */
2218 static ipstate_t *fr_checkicmpmatchingstate(fin)
2219 fr_info_t *fin;
2220 {
2221 	ipstate_t *is, **isp;
2222 	u_short sport, dport;
2223 	u_char	pr;
2224 	int backward, i, oi;
2225 	i6addr_t dst, src;
2226 	struct icmp *ic;
2227 	u_short savelen;
2228 	icmphdr_t *icmp;
2229 	fr_info_t ofin;
2230 	tcphdr_t *tcp;
2231 	int len;
2232 	ip_t *oip;
2233 	u_int hv;
2234 	ipf_stack_t *ifs = fin->fin_ifs;
2235 
2236 	/*
2237 	 * Does it at least have the return (basic) IP header ?
2238 	 * Is it an actual recognised ICMP error type?
2239 	 * Only a basic IP header (no options) should be with
2240 	 * an ICMP error header.
2241 	 */
2242 	if ((fin->fin_v != 4) || (fin->fin_hlen != sizeof(ip_t)) ||
2243 	    (fin->fin_plen < ICMPERR_MINPKTLEN) ||
2244 	    !(fin->fin_flx & FI_ICMPERR))
2245 		return NULL;
2246 	ic = fin->fin_dp;
2247 
2248 	oip = (ip_t *)((char *)ic + ICMPERR_ICMPHLEN);
2249 	/*
2250 	 * Check if the at least the old IP header (with options) and
2251 	 * 8 bytes of payload is present.
2252 	 */
2253 	if (fin->fin_plen < ICMPERR_MAXPKTLEN + ((IP_HL(oip) - 5) << 2))
2254 		return NULL;
2255 
2256 	/*
2257 	 * Sanity Checks.
2258 	 */
2259 	len = fin->fin_dlen - ICMPERR_ICMPHLEN;
2260 	if ((len <= 0) || ((IP_HL(oip) << 2) > len))
2261 		return NULL;
2262 
2263 	/*
2264 	 * Is the buffer big enough for all of it ?  It's the size of the IP
2265 	 * header claimed in the encapsulated part which is of concern.  It
2266 	 * may be too big to be in this buffer but not so big that it's
2267 	 * outside the ICMP packet, leading to TCP deref's causing problems.
2268 	 * This is possible because we don't know how big oip_hl is when we
2269 	 * do the pullup early in fr_check() and thus can't guarantee it is
2270 	 * all here now.
2271 	 */
2272 #ifdef  _KERNEL
2273 	{
2274 	mb_t *m;
2275 
2276 	m = fin->fin_m;
2277 # if defined(MENTAT)
2278 	if ((char *)oip + len > (char *)m->b_wptr)
2279 		return NULL;
2280 # else
2281 	if ((char *)oip + len > (char *)fin->fin_ip + m->m_len)
2282 		return NULL;
2283 # endif
2284 	}
2285 #endif
2286 	bcopy((char *)fin, (char *)&ofin, sizeof(*fin));
2287 
2288 	/*
2289 	 * in the IPv4 case we must zero the i6addr union otherwise
2290 	 * the IP6_EQ and IP6_NEQ macros produce the wrong results because
2291 	 * of the 'junk' in the unused part of the union
2292 	 */
2293 	bzero((char *)&src, sizeof(src));
2294 	bzero((char *)&dst, sizeof(dst));
2295 
2296 	/*
2297 	 * we make an fin entry to be able to feed it to
2298 	 * matchsrcdst note that not all fields are encessary
2299 	 * but this is the cleanest way. Note further we fill
2300 	 * in fin_mp such that if someone uses it we'll get
2301 	 * a kernel panic. fr_matchsrcdst does not use this.
2302 	 *
2303 	 * watch out here, as ip is in host order and oip in network
2304 	 * order. Any change we make must be undone afterwards, like
2305 	 * oip->ip_off - it is still in network byte order so fix it.
2306 	 */
2307 	savelen = oip->ip_len;
2308 	oip->ip_len = len;
2309 	oip->ip_off = ntohs(oip->ip_off);
2310 
2311 	ofin.fin_flx = FI_NOCKSUM;
2312 	ofin.fin_v = 4;
2313 	ofin.fin_ip = oip;
2314 	ofin.fin_m = NULL;	/* if dereferenced, panic XXX */
2315 	ofin.fin_mp = NULL;	/* if dereferenced, panic XXX */
2316 	ofin.fin_plen = fin->fin_dlen - ICMPERR_ICMPHLEN;
2317 	(void) fr_makefrip(IP_HL(oip) << 2, oip, &ofin);
2318 	ofin.fin_ifp = fin->fin_ifp;
2319 	ofin.fin_out = !fin->fin_out;
2320 	/*
2321 	 * Reset the short and bad flag here because in fr_matchsrcdst()
2322 	 * the flags for the current packet (fin_flx) are compared against
2323 	 * those for the existing session.
2324 	 */
2325 	ofin.fin_flx &= ~(FI_BAD|FI_SHORT);
2326 
2327 	/*
2328 	 * Put old values of ip_len and ip_off back as we don't know
2329 	 * if we have to forward the packet (or process it again.
2330 	 */
2331 	oip->ip_len = savelen;
2332 	oip->ip_off = htons(oip->ip_off);
2333 
2334 	switch (oip->ip_p)
2335 	{
2336 	case IPPROTO_ICMP :
2337 		/*
2338 		 * an ICMP error can only be generated as a result of an
2339 		 * ICMP query, not as the response on an ICMP error
2340 		 *
2341 		 * XXX theoretically ICMP_ECHOREP and the other reply's are
2342 		 * ICMP query's as well, but adding them here seems strange XXX
2343 		 */
2344 		if ((ofin.fin_flx & FI_ICMPERR) != 0)
2345 		    	return NULL;
2346 
2347 		/*
2348 		 * perform a lookup of the ICMP packet in the state table
2349 		 */
2350 		icmp = (icmphdr_t *)((char *)oip + (IP_HL(oip) << 2));
2351 		hv = (pr = oip->ip_p);
2352 		src.in4 = oip->ip_src;
2353 		hv += src.in4.s_addr;
2354 		dst.in4 = oip->ip_dst;
2355 		hv += dst.in4.s_addr;
2356 		hv += icmp->icmp_id;
2357 		hv = DOUBLE_HASH(hv, ifs);
2358 
2359 		READ_ENTER(&ifs->ifs_ipf_state);
2360 		for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
2361 			isp = &is->is_hnext;
2362 			if ((is->is_p != pr) || (is->is_v != 4))
2363 				continue;
2364 			if (is->is_pass & FR_NOICMPERR)
2365 				continue;
2366 			is = fr_matchsrcdst(&ofin, is, &src, &dst,
2367 					    NULL, FI_ICMPCMP);
2368 			if (is != NULL) {
2369 				if ((is->is_pass & FR_NOICMPERR) != 0) {
2370 					RWLOCK_EXIT(&ifs->ifs_ipf_state);
2371 					return NULL;
2372 				}
2373 				/*
2374 				 * i  : the index of this packet (the icmp
2375 				 *      unreachable)
2376 				 * oi : the index of the original packet found
2377 				 *      in the icmp header (i.e. the packet
2378 				 *      causing this icmp)
2379 				 * backward : original packet was backward
2380 				 *      compared to the state
2381 				 */
2382 				backward = IP6_NEQ(&is->is_src, &src);
2383 				fin->fin_rev = !backward;
2384 				i = (!backward << 1) + fin->fin_out;
2385 				oi = (backward << 1) + ofin.fin_out;
2386 				if (is->is_icmppkts[i] > is->is_pkts[oi])
2387 					continue;
2388 				ifs->ifs_ips_stats.iss_hits++;
2389 				is->is_icmppkts[i]++;
2390 				return is;
2391 			}
2392 		}
2393 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
2394 		return NULL;
2395 	case IPPROTO_TCP :
2396 	case IPPROTO_UDP :
2397 		break;
2398 	default :
2399 		return NULL;
2400 	}
2401 
2402 	tcp = (tcphdr_t *)((char *)oip + (IP_HL(oip) << 2));
2403 	dport = tcp->th_dport;
2404 	sport = tcp->th_sport;
2405 
2406 	hv = (pr = oip->ip_p);
2407 	src.in4 = oip->ip_src;
2408 	hv += src.in4.s_addr;
2409 	dst.in4 = oip->ip_dst;
2410 	hv += dst.in4.s_addr;
2411 	hv += dport;
2412 	hv += sport;
2413 	hv = DOUBLE_HASH(hv, ifs);
2414 
2415 	READ_ENTER(&ifs->ifs_ipf_state);
2416 	for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
2417 		isp = &is->is_hnext;
2418 		/*
2419 		 * Only allow this icmp though if the
2420 		 * encapsulated packet was allowed through the
2421 		 * other way around. Note that the minimal amount
2422 		 * of info present does not allow for checking against
2423 		 * tcp internals such as seq and ack numbers.   Only the
2424 		 * ports are known to be present and can be even if the
2425 		 * short flag is set.
2426 		 */
2427 		if ((is->is_p == pr) && (is->is_v == 4) &&
2428 		    (is = fr_matchsrcdst(&ofin, is, &src, &dst,
2429 					 tcp, FI_ICMPCMP))) {
2430 			/*
2431 			 * i  : the index of this packet (the icmp unreachable)
2432 			 * oi : the index of the original packet found in the
2433 			 *      icmp header (i.e. the packet causing this icmp)
2434 			 * backward : original packet was backward compared to
2435 			 *            the state
2436 			 */
2437 			backward = IP6_NEQ(&is->is_src, &src);
2438 			fin->fin_rev = !backward;
2439 			i = (!backward << 1) + fin->fin_out;
2440 			oi = (backward << 1) + ofin.fin_out;
2441 
2442 			if (((is->is_pass & FR_NOICMPERR) != 0) ||
2443 			    (is->is_icmppkts[i] > is->is_pkts[oi]))
2444 				break;
2445 			ifs->ifs_ips_stats.iss_hits++;
2446 			is->is_icmppkts[i]++;
2447 			/*
2448 			 * we deliberately do not touch the timeouts
2449 			 * for the accompanying state table entry.
2450 			 * It remains to be seen if that is correct. XXX
2451 			 */
2452 			return is;
2453 		}
2454 	}
2455 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
2456 	return NULL;
2457 }
2458 
2459 
2460 /* ------------------------------------------------------------------------ */
2461 /* Function:    fr_ipsmove                                                  */
2462 /* Returns:     Nil                                                         */
2463 /* Parameters:  is(I) - pointer to state table entry                        */
2464 /*              hv(I) - new hash value for state table entry                */
2465 /* Write Locks: ipf_state                                                   */
2466 /*                                                                          */
2467 /* Move a state entry from one position in the hash table to another.       */
2468 /* ------------------------------------------------------------------------ */
2469 static void fr_ipsmove(is, hv, ifs)
2470 ipstate_t *is;
2471 u_int hv;
2472 ipf_stack_t *ifs;
2473 {
2474 	ipstate_t **isp;
2475 	u_int hvm;
2476 
2477 	ASSERT(rw_read_locked(&ifs->ifs_ipf_state.ipf_lk) == 0);
2478 
2479 	hvm = is->is_hv;
2480 	/*
2481 	 * Remove the hash from the old location...
2482 	 */
2483 	isp = is->is_phnext;
2484 	if (is->is_hnext)
2485 		is->is_hnext->is_phnext = isp;
2486 	*isp = is->is_hnext;
2487 	if (ifs->ifs_ips_table[hvm] == NULL)
2488 		ifs->ifs_ips_stats.iss_inuse--;
2489 	ifs->ifs_ips_stats.iss_bucketlen[hvm]--;
2490 
2491 	/*
2492 	 * ...and put the hash in the new one.
2493 	 */
2494 	hvm = DOUBLE_HASH(hv, ifs);
2495 	is->is_hv = hvm;
2496 	isp = &ifs->ifs_ips_table[hvm];
2497 	if (*isp)
2498 		(*isp)->is_phnext = &is->is_hnext;
2499 	else
2500 		ifs->ifs_ips_stats.iss_inuse++;
2501 	ifs->ifs_ips_stats.iss_bucketlen[hvm]++;
2502 	is->is_phnext = isp;
2503 	is->is_hnext = *isp;
2504 	*isp = is;
2505 }
2506 
2507 
2508 /* ------------------------------------------------------------------------ */
2509 /* Function:    fr_stlookup                                                 */
2510 /* Returns:     ipstate_t* - NULL == no matching state found,               */
2511 /*                           else pointer to state information is returned  */
2512 /* Parameters:  fin(I) - pointer to packet information                      */
2513 /*              tcp(I) - pointer to TCP/UDP header.                         */
2514 /*                                                                          */
2515 /* Search the state table for a matching entry to the packet described by   */
2516 /* the contents of *fin.                                                    */
2517 /*                                                                          */
2518 /* If we return NULL then no lock on ipf_state is held.                     */
2519 /* If we return non-null then a read-lock on ipf_state is held.             */
2520 /* ------------------------------------------------------------------------ */
2521 ipstate_t *fr_stlookup(fin, tcp, ifqp)
2522 fr_info_t *fin;
2523 tcphdr_t *tcp;
2524 ipftq_t **ifqp;
2525 {
2526 	u_int hv, hvm, pr, v, tryagain;
2527 	ipstate_t *is, **isp;
2528 	u_short dport, sport;
2529 	i6addr_t src, dst;
2530 	struct icmp *ic;
2531 	ipftq_t *ifq;
2532 	int oow;
2533 	ipf_stack_t *ifs = fin->fin_ifs;
2534 
2535 	is = NULL;
2536 	ifq = NULL;
2537 	tcp = fin->fin_dp;
2538 	ic = (struct icmp *)tcp;
2539 	hv = (pr = fin->fin_fi.fi_p);
2540 	src = fin->fin_fi.fi_src;
2541 	dst = fin->fin_fi.fi_dst;
2542 	hv += src.in4.s_addr;
2543 	hv += dst.in4.s_addr;
2544 
2545 	v = fin->fin_fi.fi_v;
2546 #ifdef	USE_INET6
2547 	if (v == 6) {
2548 		hv  += fin->fin_fi.fi_src.i6[1];
2549 		hv  += fin->fin_fi.fi_src.i6[2];
2550 		hv  += fin->fin_fi.fi_src.i6[3];
2551 
2552 		if ((fin->fin_p == IPPROTO_ICMPV6) &&
2553 		    IN6_IS_ADDR_MULTICAST(&fin->fin_fi.fi_dst.in6)) {
2554 			hv -= dst.in4.s_addr;
2555 		} else {
2556 			hv += fin->fin_fi.fi_dst.i6[1];
2557 			hv += fin->fin_fi.fi_dst.i6[2];
2558 			hv += fin->fin_fi.fi_dst.i6[3];
2559 		}
2560 	}
2561 #endif
2562 
2563 	/*
2564 	 * Search the hash table for matching packet header info.
2565 	 */
2566 	switch (pr)
2567 	{
2568 #ifdef	USE_INET6
2569 	case IPPROTO_ICMPV6 :
2570 		tryagain = 0;
2571 		if (v == 6) {
2572 			if ((ic->icmp_type == ICMP6_ECHO_REQUEST) ||
2573 			    (ic->icmp_type == ICMP6_ECHO_REPLY)) {
2574 				hv += ic->icmp_id;
2575 			}
2576 		}
2577 		READ_ENTER(&ifs->ifs_ipf_state);
2578 icmp6again:
2579 		hvm = DOUBLE_HASH(hv, ifs);
2580 		for (isp = &ifs->ifs_ips_table[hvm]; ((is = *isp) != NULL); ) {
2581 			isp = &is->is_hnext;
2582 			if ((is->is_p != pr) || (is->is_v != v))
2583 				continue;
2584 			is = fr_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP);
2585 			if (is != NULL &&
2586 			    fr_matchicmpqueryreply(v, &is->is_icmp,
2587 						   ic, fin->fin_rev)) {
2588 				if (fin->fin_rev)
2589 					ifq = &ifs->ifs_ips_icmpacktq;
2590 				else
2591 					ifq = &ifs->ifs_ips_icmptq;
2592 				break;
2593 			}
2594 		}
2595 
2596 		if (is != NULL) {
2597 			if ((tryagain != 0) && !(is->is_flags & SI_W_DADDR)) {
2598 				hv += fin->fin_fi.fi_src.i6[0];
2599 				hv += fin->fin_fi.fi_src.i6[1];
2600 				hv += fin->fin_fi.fi_src.i6[2];
2601 				hv += fin->fin_fi.fi_src.i6[3];
2602 				fr_ipsmove(is, hv, ifs);
2603 				MUTEX_DOWNGRADE(&ifs->ifs_ipf_state);
2604 			}
2605 			break;
2606 		}
2607 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
2608 
2609 		/*
2610 		 * No matching icmp state entry. Perhaps this is a
2611 		 * response to another state entry.
2612 		 *
2613 		 * XXX With some ICMP6 packets, the "other" address is already
2614 		 * in the packet, after the ICMP6 header, and this could be
2615 		 * used in place of the multicast address.  However, taking
2616 		 * advantage of this requires some significant code changes
2617 		 * to handle the specific types where that is the case.
2618 		 */
2619 		if ((ifs->ifs_ips_stats.iss_wild != 0) && (v == 6) && (tryagain == 0) &&
2620 		    !IN6_IS_ADDR_MULTICAST(&fin->fin_fi.fi_src.in6)) {
2621 			hv -= fin->fin_fi.fi_src.i6[0];
2622 			hv -= fin->fin_fi.fi_src.i6[1];
2623 			hv -= fin->fin_fi.fi_src.i6[2];
2624 			hv -= fin->fin_fi.fi_src.i6[3];
2625 			tryagain = 1;
2626 			WRITE_ENTER(&ifs->ifs_ipf_state);
2627 			goto icmp6again;
2628 		}
2629 
2630 		is = fr_checkicmp6matchingstate(fin);
2631 		if (is != NULL)
2632 			return is;
2633 		break;
2634 #endif
2635 
2636 	case IPPROTO_ICMP :
2637 		if (v == 4) {
2638 			hv += ic->icmp_id;
2639 		}
2640 		hv = DOUBLE_HASH(hv, ifs);
2641 		READ_ENTER(&ifs->ifs_ipf_state);
2642 		for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
2643 			isp = &is->is_hnext;
2644 			if ((is->is_p != pr) || (is->is_v != v))
2645 				continue;
2646 			is = fr_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP);
2647 			if (is != NULL &&
2648 			    fr_matchicmpqueryreply(v, &is->is_icmp,
2649 						   ic, fin->fin_rev)) {
2650 				if (fin->fin_rev)
2651 					ifq = &ifs->ifs_ips_icmpacktq;
2652 				else
2653 					ifq = &ifs->ifs_ips_icmptq;
2654 				break;
2655 			}
2656 		}
2657 		if (is == NULL) {
2658 			RWLOCK_EXIT(&ifs->ifs_ipf_state);
2659 		}
2660 		break;
2661 
2662 	case IPPROTO_TCP :
2663 	case IPPROTO_UDP :
2664 		ifqp = NULL;
2665 		sport = htons(fin->fin_data[0]);
2666 		hv += sport;
2667 		dport = htons(fin->fin_data[1]);
2668 		hv += dport;
2669 		oow = 0;
2670 		tryagain = 0;
2671 		READ_ENTER(&ifs->ifs_ipf_state);
2672 retry_tcpudp:
2673 		hvm = DOUBLE_HASH(hv, ifs);
2674 		for (isp = &ifs->ifs_ips_table[hvm]; ((is = *isp) != NULL); ) {
2675 			isp = &is->is_hnext;
2676 			if ((is->is_p != pr) || (is->is_v != v))
2677 				continue;
2678 			fin->fin_flx &= ~FI_OOW;
2679 			is = fr_matchsrcdst(fin, is, &src, &dst, tcp, FI_CMP);
2680 			if (is != NULL) {
2681 				if (pr == IPPROTO_TCP) {
2682 					if (!fr_tcpstate(fin, tcp, is)) {
2683 						oow |= fin->fin_flx & FI_OOW;
2684 						continue;
2685 					}
2686 				}
2687 				break;
2688 			}
2689 		}
2690 		if (is != NULL) {
2691 			if (tryagain &&
2692 			    !(is->is_flags & (SI_CLONE|SI_WILDP|SI_WILDA))) {
2693 				hv += dport;
2694 				hv += sport;
2695 				fr_ipsmove(is, hv, ifs);
2696 				MUTEX_DOWNGRADE(&ifs->ifs_ipf_state);
2697 			}
2698 			break;
2699 		}
2700 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
2701 
2702 		if (!tryagain && ifs->ifs_ips_stats.iss_wild) {
2703 			hv -= dport;
2704 			hv -= sport;
2705 			tryagain = 1;
2706 			WRITE_ENTER(&ifs->ifs_ipf_state);
2707 			goto retry_tcpudp;
2708 		}
2709 		fin->fin_flx |= oow;
2710 		break;
2711 
2712 #if 0
2713 	case IPPROTO_GRE :
2714 		gre = fin->fin_dp;
2715 		if (GRE_REV(gre->gr_flags) == 1) {
2716 			hv += gre->gr_call;
2717 		}
2718 		/* FALLTHROUGH */
2719 #endif
2720 	default :
2721 		ifqp = NULL;
2722 		hvm = DOUBLE_HASH(hv, ifs);
2723 		READ_ENTER(&ifs->ifs_ipf_state);
2724 		for (isp = &ifs->ifs_ips_table[hvm]; ((is = *isp) != NULL); ) {
2725 			isp = &is->is_hnext;
2726 			if ((is->is_p != pr) || (is->is_v != v))
2727 				continue;
2728 			is = fr_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP);
2729 			if (is != NULL) {
2730 				ifq = &ifs->ifs_ips_iptq;
2731 				break;
2732 			}
2733 		}
2734 		if (is == NULL) {
2735 			RWLOCK_EXIT(&ifs->ifs_ipf_state);
2736 		}
2737 		break;
2738 	}
2739 
2740 	if ((is != NULL) && ((is->is_sti.tqe_flags & TQE_RULEBASED) != 0) &&
2741 	    (is->is_tqehead[fin->fin_rev] != NULL))
2742 		ifq = is->is_tqehead[fin->fin_rev];
2743 	if (ifq != NULL && ifqp != NULL)
2744 		*ifqp = ifq;
2745 	return is;
2746 }
2747 
2748 
2749 /* ------------------------------------------------------------------------ */
2750 /* Function:    fr_updatestate                                              */
2751 /* Returns:     Nil                                                         */
2752 /* Parameters:  fin(I) - pointer to packet information                      */
2753 /*              is(I)  - pointer to state table entry                       */
2754 /* Read Locks:  ipf_state                                                   */
2755 /*                                                                          */
2756 /* Updates packet and byte counters for a newly received packet.  Seeds the */
2757 /* fragment cache with a new entry as required.                             */
2758 /* ------------------------------------------------------------------------ */
2759 void fr_updatestate(fin, is, ifq)
2760 fr_info_t *fin;
2761 ipstate_t *is;
2762 ipftq_t *ifq;
2763 {
2764 	ipftqent_t *tqe;
2765 	int i, pass;
2766 	ipf_stack_t *ifs = fin->fin_ifs;
2767 
2768 	i = (fin->fin_rev << 1) + fin->fin_out;
2769 
2770 	/*
2771 	 * For TCP packets, ifq == NULL.  For all others, check if this new
2772 	 * queue is different to the last one it was on and move it if so.
2773 	 */
2774 	tqe = &is->is_sti;
2775 	MUTEX_ENTER(&is->is_lock);
2776 	if ((tqe->tqe_flags & TQE_RULEBASED) != 0)
2777 		ifq = is->is_tqehead[fin->fin_rev];
2778 
2779 	if (ifq != NULL)
2780 		fr_movequeue(tqe, tqe->tqe_ifq, ifq, ifs);
2781 
2782 	is->is_pkts[i]++;
2783 	is->is_bytes[i] += fin->fin_plen;
2784 	MUTEX_EXIT(&is->is_lock);
2785 
2786 #ifdef	IPFILTER_SYNC
2787 	if (is->is_flags & IS_STATESYNC)
2788 		ipfsync_update(SMC_STATE, fin, is->is_sync);
2789 #endif
2790 
2791 	ATOMIC_INCL(ifs->ifs_ips_stats.iss_hits);
2792 
2793 	fin->fin_fr = is->is_rule;
2794 
2795 	/*
2796 	 * If this packet is a fragment and the rule says to track fragments,
2797 	 * then create a new fragment cache entry.
2798 	 */
2799 	pass = is->is_pass;
2800 	if ((fin->fin_flx & FI_FRAG) && FR_ISPASS(pass))
2801 		(void) fr_newfrag(fin, pass ^ FR_KEEPSTATE);
2802 }
2803 
2804 
2805 /* ------------------------------------------------------------------------ */
2806 /* Function:    fr_checkstate                                               */
2807 /* Returns:     frentry_t* - NULL == search failed,                         */
2808 /*                           else pointer to rule for matching state        */
2809 /* Parameters:  ifp(I)   - pointer to interface                             */
2810 /*              passp(I) - pointer to filtering result flags                */
2811 /*                                                                          */
2812 /* Check if a packet is associated with an entry in the state table.        */
2813 /* ------------------------------------------------------------------------ */
2814 frentry_t *fr_checkstate(fin, passp)
2815 fr_info_t *fin;
2816 u_32_t *passp;
2817 {
2818 	ipstate_t *is;
2819 	frentry_t *fr;
2820 	tcphdr_t *tcp;
2821 	ipftq_t *ifq;
2822 	u_int pass;
2823 	ipf_stack_t *ifs = fin->fin_ifs;
2824 
2825 	if (ifs->ifs_fr_state_lock || (ifs->ifs_ips_list == NULL) ||
2826 	    (fin->fin_flx & (FI_SHORT|FI_STATE|FI_FRAGBODY|FI_BAD)))
2827 		return NULL;
2828 
2829 	is = NULL;
2830 	if ((fin->fin_flx & FI_TCPUDP) ||
2831 	    (fin->fin_fi.fi_p == IPPROTO_ICMP)
2832 #ifdef	USE_INET6
2833 	    || (fin->fin_fi.fi_p == IPPROTO_ICMPV6)
2834 #endif
2835 	    )
2836 		tcp = fin->fin_dp;
2837 	else
2838 		tcp = NULL;
2839 
2840 	/*
2841 	 * Search the hash table for matching packet header info.
2842 	 */
2843 	ifq = NULL;
2844 	is = fin->fin_state;
2845 	if (is == NULL)
2846 		is = fr_stlookup(fin, tcp, &ifq);
2847 	switch (fin->fin_p)
2848 	{
2849 #ifdef	USE_INET6
2850 	case IPPROTO_ICMPV6 :
2851 		if (is != NULL)
2852 			break;
2853 		if (fin->fin_v == 6) {
2854 			is = fr_checkicmp6matchingstate(fin);
2855 			if (is != NULL)
2856 				goto matched;
2857 		}
2858 		break;
2859 #endif
2860 	case IPPROTO_ICMP :
2861 		if (is != NULL)
2862 			break;
2863 		/*
2864 		 * No matching icmp state entry. Perhaps this is a
2865 		 * response to another state entry.
2866 		 */
2867 		is = fr_checkicmpmatchingstate(fin);
2868 		if (is != NULL)
2869 			goto matched;
2870 		break;
2871 	case IPPROTO_TCP :
2872 		if (is == NULL)
2873 			break;
2874 
2875 		if (is->is_pass & FR_NEWISN) {
2876 			if (fin->fin_out == 0)
2877 				fr_fixinisn(fin, is);
2878 			else if (fin->fin_out == 1)
2879 				fr_fixoutisn(fin, is);
2880 		}
2881 		break;
2882 	default :
2883 		if (fin->fin_rev)
2884 			ifq = &ifs->ifs_ips_udpacktq;
2885 		else
2886 			ifq = &ifs->ifs_ips_udptq;
2887 		break;
2888 	}
2889 	if (is == NULL) {
2890 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_miss);
2891 		return NULL;
2892 	}
2893 
2894 matched:
2895 	fr = is->is_rule;
2896 	if (fr != NULL) {
2897 		if ((fin->fin_out == 0) && (fr->fr_nattag.ipt_num[0] != 0)) {
2898 			if (fin->fin_nattag == NULL)
2899 				return NULL;
2900 			if (fr_matchtag(&fr->fr_nattag, fin->fin_nattag) != 0)
2901 				return NULL;
2902 		}
2903 		(void) strncpy(fin->fin_group, fr->fr_group, FR_GROUPLEN);
2904 		fin->fin_icode = fr->fr_icode;
2905 	}
2906 
2907 	fin->fin_rule = is->is_rulen;
2908 	pass = is->is_pass;
2909 	fr_updatestate(fin, is, ifq);
2910 	if (fin->fin_out == 1)
2911 		fin->fin_nat = is->is_nat[fin->fin_rev];
2912 
2913 	fin->fin_state = is;
2914 	is->is_touched = ifs->ifs_fr_ticks;
2915 	MUTEX_ENTER(&is->is_lock);
2916 	is->is_ref++;
2917 	MUTEX_EXIT(&is->is_lock);
2918 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
2919 	fin->fin_flx |= FI_STATE;
2920 	if ((pass & FR_LOGFIRST) != 0)
2921 		pass &= ~(FR_LOGFIRST|FR_LOG);
2922 	*passp = pass;
2923 	return fr;
2924 }
2925 
2926 
2927 /* ------------------------------------------------------------------------ */
2928 /* Function:    fr_fixoutisn                                                */
2929 /* Returns:     Nil                                                         */
2930 /* Parameters:  fin(I)   - pointer to packet information                    */
2931 /*              is(I)  - pointer to master state structure                  */
2932 /*                                                                          */
2933 /* Called only for outbound packets, adjusts the sequence number and the    */
2934 /* TCP checksum to match that change.                                       */
2935 /* ------------------------------------------------------------------------ */
2936 static void fr_fixoutisn(fin, is)
2937 fr_info_t *fin;
2938 ipstate_t *is;
2939 {
2940 	tcphdr_t *tcp;
2941 	int rev;
2942 	u_32_t seq;
2943 
2944 	tcp = fin->fin_dp;
2945 	rev = fin->fin_rev;
2946 	if ((is->is_flags & IS_ISNSYN) != 0) {
2947 		if (rev == 0) {
2948 			seq = ntohl(tcp->th_seq);
2949 			seq += is->is_isninc[0];
2950 			tcp->th_seq = htonl(seq);
2951 			fix_outcksum(&tcp->th_sum, is->is_sumd[0]);
2952 		}
2953 	}
2954 	if ((is->is_flags & IS_ISNACK) != 0) {
2955 		if (rev == 1) {
2956 			seq = ntohl(tcp->th_seq);
2957 			seq += is->is_isninc[1];
2958 			tcp->th_seq = htonl(seq);
2959 			fix_outcksum(&tcp->th_sum, is->is_sumd[1]);
2960 		}
2961 	}
2962 }
2963 
2964 
2965 /* ------------------------------------------------------------------------ */
2966 /* Function:    fr_fixinisn                                                 */
2967 /* Returns:     Nil                                                         */
2968 /* Parameters:  fin(I)   - pointer to packet information                    */
2969 /*              is(I)  - pointer to master state structure                  */
2970 /*                                                                          */
2971 /* Called only for inbound packets, adjusts the acknowledge number and the  */
2972 /* TCP checksum to match that change.                                       */
2973 /* ------------------------------------------------------------------------ */
2974 static void fr_fixinisn(fin, is)
2975 fr_info_t *fin;
2976 ipstate_t *is;
2977 {
2978 	tcphdr_t *tcp;
2979 	int rev;
2980 	u_32_t ack;
2981 
2982 	tcp = fin->fin_dp;
2983 	rev = fin->fin_rev;
2984 	if ((is->is_flags & IS_ISNSYN) != 0) {
2985 		if (rev == 1) {
2986 			ack = ntohl(tcp->th_ack);
2987 			ack -= is->is_isninc[0];
2988 			tcp->th_ack = htonl(ack);
2989 			fix_incksum(&tcp->th_sum, is->is_sumd[0]);
2990 		}
2991 	}
2992 	if ((is->is_flags & IS_ISNACK) != 0) {
2993 		if (rev == 0) {
2994 			ack = ntohl(tcp->th_ack);
2995 			ack -= is->is_isninc[1];
2996 			tcp->th_ack = htonl(ack);
2997 			fix_incksum(&tcp->th_sum, is->is_sumd[1]);
2998 		}
2999 	}
3000 }
3001 
3002 
3003 /* ------------------------------------------------------------------------ */
3004 /* Function:    fr_statesync                                                */
3005 /* Returns:     Nil                                                         */
3006 /* Parameters:  action(I) - type of synchronisation to do                   */
3007 /*              v(I)      - IP version being sync'd (v4 or v6)              */
3008 /*              ifp(I)    - interface identifier associated with action     */
3009 /*              name(I)   - name associated with ifp parameter              */
3010 /*                                                                          */
3011 /* Walk through all state entries and if an interface pointer match is      */
3012 /* found then look it up again, based on its name in case the pointer has   */
3013 /* changed since last time.                                                 */
3014 /*                                                                          */
3015 /* If ifp is passed in as being non-null then we are only doing updates for */
3016 /* existing, matching, uses of it.                                          */
3017 /* ------------------------------------------------------------------------ */
3018 void fr_statesync(action, v, ifp, name, ifs)
3019 int action, v;
3020 void *ifp;
3021 char *name;
3022 ipf_stack_t *ifs;
3023 {
3024 	ipstate_t *is;
3025 	int i;
3026 
3027 	if (ifs->ifs_fr_running <= 0)
3028 		return;
3029 
3030 	WRITE_ENTER(&ifs->ifs_ipf_state);
3031 
3032 	if (ifs->ifs_fr_running <= 0) {
3033 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
3034 		return;
3035 	}
3036 
3037 	switch (action)
3038 	{
3039 	case IPFSYNC_RESYNC :
3040 		for (is = ifs->ifs_ips_list; is; is = is->is_next) {
3041 			if (v != 0 && is->is_v != v)
3042 				continue;
3043 			/*
3044 			 * Look up all the interface names in the state entry.
3045 			 */
3046 			for (i = 0; i < 4; i++) {
3047 				is->is_ifp[i] = fr_resolvenic(is->is_ifname[i],
3048 							      is->is_v, ifs);
3049 			}
3050 		}
3051 		break;
3052 	case IPFSYNC_NEWIFP :
3053 		for (is = ifs->ifs_ips_list; is; is = is->is_next) {
3054 			if (v != 0 && is->is_v != v)
3055 				continue;
3056 			/*
3057 			 * Look up all the interface names in the state entry.
3058 			 */
3059 			for (i = 0; i < 4; i++) {
3060 				if (!strncmp(is->is_ifname[i], name,
3061 					     sizeof(is->is_ifname[i])))
3062 					is->is_ifp[i] = ifp;
3063 			}
3064 		}
3065 		break;
3066 	case IPFSYNC_OLDIFP :
3067 		for (is = ifs->ifs_ips_list; is; is = is->is_next) {
3068 			if (v != 0 && is->is_v != v)
3069 				continue;
3070 			/*
3071 			 * Look up all the interface names in the state entry.
3072 			 */
3073 			for (i = 0; i < 4; i++) {
3074 				if (is->is_ifp[i] == ifp)
3075 					is->is_ifp[i] = (void *)-1;
3076 			}
3077 		}
3078 		break;
3079 	}
3080 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
3081 }
3082 
3083 
3084 /* ------------------------------------------------------------------------ */
3085 /* Function:    fr_delstate                                                 */
3086 /* Returns:     Nil                                                         */
3087 /* Parameters:  is(I)  - pointer to state structure to delete               */
3088 /*              why(I) - if not 0, log reason why it was deleted            */
3089 /* Write Locks: ipf_state/ipf_global                                        */
3090 /*                                                                          */
3091 /* Deletes a state entry from the enumerated list as well as the hash table */
3092 /* and timeout queue lists.  Make adjustments to hash table statistics and  */
3093 /* global counters as required.                                             */
3094 /* ------------------------------------------------------------------------ */
3095 static void fr_delstate(is, why, ifs)
3096 ipstate_t *is;
3097 int why;
3098 ipf_stack_t *ifs;
3099 {
3100 
3101 	ASSERT(rw_write_held(&ifs->ifs_ipf_global.ipf_lk) == 0 ||
3102 		rw_write_held(&ifs->ifs_ipf_state.ipf_lk) == 0);
3103 
3104 	/*
3105 	 * Since we want to delete this, remove it from the state table,
3106 	 * where it can be found & used, first.
3107 	 */
3108 	if (is->is_pnext != NULL) {
3109 		*is->is_pnext = is->is_next;
3110 
3111 		if (is->is_next != NULL)
3112 			is->is_next->is_pnext = is->is_pnext;
3113 
3114 		is->is_pnext = NULL;
3115 		is->is_next = NULL;
3116 	}
3117 
3118 	if (is->is_phnext != NULL) {
3119 		*is->is_phnext = is->is_hnext;
3120 		if (is->is_hnext != NULL)
3121 			is->is_hnext->is_phnext = is->is_phnext;
3122 		if (ifs->ifs_ips_table[is->is_hv] == NULL)
3123 			ifs->ifs_ips_stats.iss_inuse--;
3124 		ifs->ifs_ips_stats.iss_bucketlen[is->is_hv]--;
3125 
3126 		is->is_phnext = NULL;
3127 		is->is_hnext = NULL;
3128 	}
3129 
3130 	/*
3131 	 * Because ifs->ifs_ips_stats.iss_wild is a count of entries in the state
3132 	 * table that have wildcard flags set, only decerement it once
3133 	 * and do it here.
3134 	 */
3135 	if (is->is_flags & (SI_WILDP|SI_WILDA)) {
3136 		if (!(is->is_flags & SI_CLONED)) {
3137 			ATOMIC_DECL(ifs->ifs_ips_stats.iss_wild);
3138 		}
3139 		is->is_flags &= ~(SI_WILDP|SI_WILDA);
3140 	}
3141 
3142 	/*
3143 	 * Next, remove it from the timeout queue it is in.
3144 	 */
3145 	fr_deletequeueentry(&is->is_sti);
3146 
3147 	is->is_me = NULL;
3148 
3149 	/*
3150 	 * If it is still in use by something else, do not go any further,
3151 	 * but note that at this point it is now an orphan.
3152 	 */
3153 	MUTEX_ENTER(&is->is_lock);
3154 	if (is->is_ref > 1) {
3155 		is->is_ref--;
3156 		MUTEX_EXIT(&is->is_lock);
3157 		return;
3158 	}
3159 	MUTEX_EXIT(&is->is_lock);
3160 
3161 	is->is_ref = 0;
3162 
3163 	if (is->is_tqehead[0] != NULL)
3164 		(void) fr_deletetimeoutqueue(is->is_tqehead[0]);
3165 
3166 	if (is->is_tqehead[1] != NULL)
3167 		(void) fr_deletetimeoutqueue(is->is_tqehead[1]);
3168 
3169 #ifdef	IPFILTER_SYNC
3170 	if (is->is_sync)
3171 		ipfsync_del(is->is_sync);
3172 #endif
3173 #ifdef	IPFILTER_SCAN
3174 	(void) ipsc_detachis(is);
3175 #endif
3176 
3177 	if (ifs->ifs_ipstate_logging != 0 && why != 0)
3178 		ipstate_log(is, why, ifs);
3179 
3180 	if (is->is_rule != NULL) {
3181 		is->is_rule->fr_statecnt--;
3182 		(void)fr_derefrule(&is->is_rule, ifs);
3183 	}
3184 
3185 	MUTEX_DESTROY(&is->is_lock);
3186 	KFREE(is);
3187 	ifs->ifs_ips_num--;
3188 }
3189 
3190 
3191 /* ------------------------------------------------------------------------ */
3192 /* Function:    fr_timeoutstate                                             */
3193 /* Returns:     Nil                                                         */
3194 /* Parameters:  Nil                                                         */
3195 /*                                                                          */
3196 /* Slowly expire held state for thingslike UDP and ICMP.  The algorithm     */
3197 /* used here is to keep the queue sorted with the oldest things at the top  */
3198 /* and the youngest at the bottom.  So if the top one doesn't need to be    */
3199 /* expired then neither will any under it.                                  */
3200 /* ------------------------------------------------------------------------ */
3201 void fr_timeoutstate(ifs)
3202 ipf_stack_t *ifs;
3203 {
3204 	ipftq_t *ifq, *ifqnext;
3205 	ipftqent_t *tqe, *tqn;
3206 	ipstate_t *is;
3207 	SPL_INT(s);
3208 
3209 	SPL_NET(s);
3210 	WRITE_ENTER(&ifs->ifs_ipf_state);
3211 	for (ifq = ifs->ifs_ips_tqtqb; ifq != NULL; ifq = ifq->ifq_next)
3212 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); ) {
3213 			if (tqe->tqe_die > ifs->ifs_fr_ticks)
3214 				break;
3215 			tqn = tqe->tqe_next;
3216 			is = tqe->tqe_parent;
3217 			fr_delstate(is, ISL_EXPIRE, ifs);
3218 		}
3219 
3220 	for (ifq = ifs->ifs_ips_utqe; ifq != NULL; ifq = ifq->ifq_next) {
3221 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); ) {
3222 			if (tqe->tqe_die > ifs->ifs_fr_ticks)
3223 				break;
3224 			tqn = tqe->tqe_next;
3225 			is = tqe->tqe_parent;
3226 			fr_delstate(is, ISL_EXPIRE, ifs);
3227 		}
3228 	}
3229 
3230 	for (ifq = ifs->ifs_ips_utqe; ifq != NULL; ifq = ifqnext) {
3231 		ifqnext = ifq->ifq_next;
3232 
3233 		if (((ifq->ifq_flags & IFQF_DELETE) != 0) &&
3234 		    (ifq->ifq_ref == 0)) {
3235 			fr_freetimeoutqueue(ifq, ifs);
3236 		}
3237 	}
3238 
3239 	if (ifs->ifs_fr_state_doflush) {
3240 		(void) fr_state_flush(2, 0, ifs);
3241 		ifs->ifs_fr_state_doflush = 0;
3242 	}
3243 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
3244 	SPL_X(s);
3245 }
3246 
3247 
3248 /* ------------------------------------------------------------------------ */
3249 /* Function:    fr_state_flush                                              */
3250 /* Returns:     int - 0 == success, -1 == failure                           */
3251 /* Parameters:  Nil                                                         */
3252 /* Write Locks: ipf_state                                                   */
3253 /*                                                                          */
3254 /* Flush state tables.  Three actions currently defined:                    */
3255 /* which == 0 : flush all state table entries                               */
3256 /* which == 1 : flush TCP connections which have started to close but are   */
3257 /*	      stuck for some reason.                                        */
3258 /* which == 2 : flush TCP connections which have been idle for a long time, */
3259 /*	      starting at > 4 days idle and working back in successive half-*/
3260 /*	      days to at most 12 hours old.  If this fails to free enough   */
3261 /*            slots then work backwards in half hour slots to 30 minutes.   */
3262 /*            If that too fails, then work backwards in 30 second intervals */
3263 /*            for the last 30 minutes to at worst 30 seconds idle.          */
3264 /* ------------------------------------------------------------------------ */
3265 static int fr_state_flush(which, proto, ifs)
3266 int which, proto;
3267 ipf_stack_t *ifs;
3268 {
3269 	ipftq_t *ifq, *ifqnext;
3270 	ipftqent_t *tqe, *tqn;
3271 	ipstate_t *is, **isp;
3272 	int delete, removed;
3273 	long try, maxtick;
3274 	u_long interval;
3275 	SPL_INT(s);
3276 
3277 	removed = 0;
3278 
3279 	SPL_NET(s);
3280 	for (isp = &ifs->ifs_ips_list; ((is = *isp) != NULL); ) {
3281 		delete = 0;
3282 
3283 		if ((proto != 0) && (is->is_v != proto)) {
3284 			isp = &is->is_next;
3285 			continue;
3286 		}
3287 
3288 		switch (which)
3289 		{
3290 		case 0 :
3291 			delete = 1;
3292 			break;
3293 		case 1 :
3294 		case 2 :
3295 			if (is->is_p != IPPROTO_TCP)
3296 				break;
3297 			if ((is->is_state[0] != IPF_TCPS_ESTABLISHED) ||
3298 			    (is->is_state[1] != IPF_TCPS_ESTABLISHED))
3299 				delete = 1;
3300 			break;
3301 		}
3302 
3303 		if (delete) {
3304 			if (is->is_p == IPPROTO_TCP)
3305 				ifs->ifs_ips_stats.iss_fin++;
3306 			else
3307 				ifs->ifs_ips_stats.iss_expire++;
3308 			fr_delstate(is, ISL_FLUSH, ifs);
3309 			removed++;
3310 		} else
3311 			isp = &is->is_next;
3312 	}
3313 
3314 	if (which != 2) {
3315 		SPL_X(s);
3316 		return removed;
3317 	}
3318 
3319 	/*
3320 	 * Asked to remove inactive entries because the table is full, try
3321 	 * again, 3 times, if first attempt failed with a different criteria
3322 	 * each time.  The order tried in must be in decreasing age.
3323 	 * Another alternative is to implement random drop and drop N entries
3324 	 * at random until N have been freed up.
3325 	 */
3326 	if (ifs->ifs_fr_ticks - ifs->ifs_ips_last_force_flush < IPF_TTLVAL(5))
3327 		goto force_flush_skipped;
3328 	ifs->ifs_ips_last_force_flush = ifs->ifs_fr_ticks;
3329 
3330 	if (ifs->ifs_fr_ticks > IPF_TTLVAL(43200))
3331 		interval = IPF_TTLVAL(43200);
3332 	else if (ifs->ifs_fr_ticks > IPF_TTLVAL(1800))
3333 		interval = IPF_TTLVAL(1800);
3334 	else if (ifs->ifs_fr_ticks > IPF_TTLVAL(30))
3335 		interval = IPF_TTLVAL(30);
3336 	else
3337 		interval = IPF_TTLVAL(10);
3338 	try = ifs->ifs_fr_ticks - (ifs->ifs_fr_ticks - interval);
3339 	if (try < 0)
3340 		goto force_flush_skipped;
3341 
3342 	while (removed == 0) {
3343 		maxtick = ifs->ifs_fr_ticks - interval;
3344 		if (maxtick < 0)
3345 			break;
3346 
3347 		while (try < maxtick) {
3348 			for (ifq = ifs->ifs_ips_tqtqb; ifq != NULL;
3349 			     ifq = ifq->ifq_next) {
3350 				for (tqn = ifq->ifq_head;
3351 				     ((tqe = tqn) != NULL); ) {
3352 					if (tqe->tqe_die > try)
3353 						break;
3354 					tqn = tqe->tqe_next;
3355 					is = tqe->tqe_parent;
3356 					fr_delstate(is, ISL_EXPIRE, ifs);
3357 					removed++;
3358 				}
3359 			}
3360 
3361 			for (ifq = ifs->ifs_ips_utqe; ifq != NULL; ifq = ifqnext) {
3362 				ifqnext = ifq->ifq_next;
3363 
3364 				for (tqn = ifq->ifq_head;
3365 				     ((tqe = tqn) != NULL); ) {
3366 					if (tqe->tqe_die > try)
3367 						break;
3368 					tqn = tqe->tqe_next;
3369 					is = tqe->tqe_parent;
3370 					fr_delstate(is, ISL_EXPIRE, ifs);
3371 					removed++;
3372 				}
3373 			}
3374 			if (try + interval > maxtick)
3375 				break;
3376 			try += interval;
3377 		}
3378 
3379 		if (removed == 0) {
3380 			if (interval == IPF_TTLVAL(43200)) {
3381 				interval = IPF_TTLVAL(1800);
3382 			} else if (interval == IPF_TTLVAL(1800)) {
3383 				interval = IPF_TTLVAL(30);
3384 			} else if (interval == IPF_TTLVAL(30)) {
3385 				interval = IPF_TTLVAL(10);
3386 			} else {
3387 				break;
3388 			}
3389 		}
3390 	}
3391 force_flush_skipped:
3392 	SPL_X(s);
3393 	return removed;
3394 }
3395 
3396 
3397 
3398 /* ------------------------------------------------------------------------ */
3399 /* Function:    fr_tcp_age                                                  */
3400 /* Returns:     int - 1 == state transition made, 0 == no change (rejected) */
3401 /* Parameters:  tq(I)    - pointer to timeout queue information             */
3402 /*              fin(I)   - pointer to packet information                    */
3403 /*              tqtab(I) - TCP timeout queue table this is in               */
3404 /*              flags(I) - flags from state/NAT entry                       */
3405 /*                                                                          */
3406 /* Rewritten by Arjan de Vet <Arjan.deVet@adv.iae.nl>, 2000-07-29:          */
3407 /*                                                                          */
3408 /* - (try to) base state transitions on real evidence only,                 */
3409 /*   i.e. packets that are sent and have been received by ipfilter;         */
3410 /*   diagram 18.12 of TCP/IP volume 1 by W. Richard Stevens was used.       */
3411 /*                                                                          */
3412 /* - deal with half-closed connections correctly;                           */
3413 /*                                                                          */
3414 /* - store the state of the source in state[0] such that ipfstat            */
3415 /*   displays the state as source/dest instead of dest/source; the calls    */
3416 /*   to fr_tcp_age have been changed accordingly.                           */
3417 /*                                                                          */
3418 /* Internal Parameters:                                                     */
3419 /*                                                                          */
3420 /*    state[0] = state of source (host that initiated connection)           */
3421 /*    state[1] = state of dest   (host that accepted the connection)        */
3422 /*                                                                          */
3423 /*    dir == 0 : a packet from source to dest                               */
3424 /*    dir == 1 : a packet from dest to source                               */
3425 /*                                                                          */
3426 /* Locking: it is assumed that the parent of the tqe structure is locked.   */
3427 /* ------------------------------------------------------------------------ */
3428 int fr_tcp_age(tqe, fin, tqtab, flags)
3429 ipftqent_t *tqe;
3430 fr_info_t *fin;
3431 ipftq_t *tqtab;
3432 int flags;
3433 {
3434 	int dlen, ostate, nstate, rval, dir;
3435 	u_char tcpflags;
3436 	tcphdr_t *tcp;
3437 	ipf_stack_t *ifs = fin->fin_ifs;
3438 
3439 	tcp = fin->fin_dp;
3440 
3441 	rval = 0;
3442 	dir = fin->fin_rev;
3443 	tcpflags = tcp->th_flags;
3444 	dlen = fin->fin_dlen - (TCP_OFF(tcp) << 2);
3445 
3446 	if (tcpflags & TH_RST) {
3447 		if (!(tcpflags & TH_PUSH) && !dlen)
3448 			nstate = IPF_TCPS_CLOSED;
3449 		else
3450 			nstate = IPF_TCPS_CLOSE_WAIT;
3451 		rval = 1;
3452 	} else {
3453 		ostate = tqe->tqe_state[1 - dir];
3454 		nstate = tqe->tqe_state[dir];
3455 
3456 		switch (nstate)
3457 		{
3458 		case IPF_TCPS_CLOSED: /* 0 */
3459 			if ((tcpflags & TH_OPENING) == TH_OPENING) {
3460 				/*
3461 				 * 'dir' received an S and sends SA in
3462 				 * response, CLOSED -> SYN_RECEIVED
3463 				 */
3464 				nstate = IPF_TCPS_SYN_RECEIVED;
3465 				rval = 1;
3466 			} else if ((tcpflags & TH_OPENING) == TH_SYN) {
3467 				/* 'dir' sent S, CLOSED -> SYN_SENT */
3468 				nstate = IPF_TCPS_SYN_SENT;
3469 				rval = 1;
3470 			}
3471 			/*
3472 			 * the next piece of code makes it possible to get
3473 			 * already established connections into the state table
3474 			 * after a restart or reload of the filter rules; this
3475 			 * does not work when a strict 'flags S keep state' is
3476 			 * used for tcp connections of course
3477 			 */
3478 			if (((flags & IS_TCPFSM) == 0) &&
3479 			    ((tcpflags & TH_ACKMASK) == TH_ACK)) {
3480 				/*
3481 				 * we saw an A, guess 'dir' is in ESTABLISHED
3482 				 * mode
3483 				 */
3484 				switch (ostate)
3485 				{
3486 				case IPF_TCPS_CLOSED :
3487 				case IPF_TCPS_SYN_RECEIVED :
3488 					nstate = IPF_TCPS_HALF_ESTAB;
3489 					rval = 1;
3490 					break;
3491 				case IPF_TCPS_HALF_ESTAB :
3492 				case IPF_TCPS_ESTABLISHED :
3493 					nstate = IPF_TCPS_ESTABLISHED;
3494 					rval = 1;
3495 					break;
3496 				default :
3497 					break;
3498 				}
3499 			}
3500 			/*
3501 			 * TODO: besides regular ACK packets we can have other
3502 			 * packets as well; it is yet to be determined how we
3503 			 * should initialize the states in those cases
3504 			 */
3505 			break;
3506 
3507 		case IPF_TCPS_LISTEN: /* 1 */
3508 			/* NOT USED */
3509 			break;
3510 
3511 		case IPF_TCPS_SYN_SENT: /* 2 */
3512 			if ((tcpflags & ~(TH_ECN|TH_CWR)) == TH_SYN) {
3513 				/*
3514 				 * A retransmitted SYN packet.  We do not reset
3515 				 * the timeout here to fr_tcptimeout because a
3516 				 * connection connect timeout does not renew
3517 				 * after every packet that is sent.  We need to
3518 				 * set rval so as to indicate the packet has
3519 				 * passed the check for its flags being valid
3520 				 * in the TCP FSM.  Setting rval to 2 has the
3521 				 * result of not resetting the timeout.
3522 				 */
3523 				rval = 2;
3524 			} else if ((tcpflags & (TH_SYN|TH_FIN|TH_ACK)) ==
3525 				   TH_ACK) {
3526 				/*
3527 				 * we see an A from 'dir' which is in SYN_SENT
3528 				 * state: 'dir' sent an A in response to an SA
3529 				 * which it received, SYN_SENT -> ESTABLISHED
3530 				 */
3531 				nstate = IPF_TCPS_ESTABLISHED;
3532 				rval = 1;
3533 			} else if (tcpflags & TH_FIN) {
3534 				/*
3535 				 * we see an F from 'dir' which is in SYN_SENT
3536 				 * state and wants to close its side of the
3537 				 * connection; SYN_SENT -> FIN_WAIT_1
3538 				 */
3539 				nstate = IPF_TCPS_FIN_WAIT_1;
3540 				rval = 1;
3541 			} else if ((tcpflags & TH_OPENING) == TH_OPENING) {
3542 				/*
3543 				 * we see an SA from 'dir' which is already in
3544 				 * SYN_SENT state, this means we have a
3545 				 * simultaneous open; SYN_SENT -> SYN_RECEIVED
3546 				 */
3547 				nstate = IPF_TCPS_SYN_RECEIVED;
3548 				rval = 1;
3549 			}
3550 			break;
3551 
3552 		case IPF_TCPS_SYN_RECEIVED: /* 3 */
3553 			if ((tcpflags & (TH_SYN|TH_FIN|TH_ACK)) == TH_ACK) {
3554 				/*
3555 				 * we see an A from 'dir' which was in
3556 				 * SYN_RECEIVED state so it must now be in
3557 				 * established state, SYN_RECEIVED ->
3558 				 * ESTABLISHED
3559 				 */
3560 				nstate = IPF_TCPS_ESTABLISHED;
3561 				rval = 1;
3562 			} else if ((tcpflags & ~(TH_ECN|TH_CWR)) ==
3563 				   TH_OPENING) {
3564 				/*
3565 				 * We see an SA from 'dir' which is already in
3566 				 * SYN_RECEIVED state.
3567 				 */
3568 				rval = 2;
3569 			} else if (tcpflags & TH_FIN) {
3570 				/*
3571 				 * we see an F from 'dir' which is in
3572 				 * SYN_RECEIVED state and wants to close its
3573 				 * side of the connection; SYN_RECEIVED ->
3574 				 * FIN_WAIT_1
3575 				 */
3576 				nstate = IPF_TCPS_FIN_WAIT_1;
3577 				rval = 1;
3578 			}
3579 			break;
3580 
3581 		case IPF_TCPS_HALF_ESTAB: /* 4 */
3582 			if (ostate >= IPF_TCPS_HALF_ESTAB) {
3583 				if ((tcpflags & TH_ACKMASK) == TH_ACK) {
3584 					nstate = IPF_TCPS_ESTABLISHED;
3585 				}
3586 			}
3587 			rval = 1;
3588 
3589 			break;
3590 
3591 		case IPF_TCPS_ESTABLISHED: /* 5 */
3592 			rval = 1;
3593 			if (tcpflags & TH_FIN) {
3594 				/*
3595 				 * 'dir' closed its side of the connection;
3596 				 * this gives us a half-closed connection;
3597 				 * ESTABLISHED -> FIN_WAIT_1
3598 				 */
3599 				nstate = IPF_TCPS_FIN_WAIT_1;
3600 			} else if (tcpflags & TH_ACK) {
3601 				/*
3602 				 * an ACK, should we exclude other flags here?
3603 				 */
3604 				if (ostate == IPF_TCPS_FIN_WAIT_1) {
3605 					/*
3606 					 * We know the other side did an active
3607 					 * close, so we are ACKing the recvd
3608 					 * FIN packet (does the window matching
3609 					 * code guarantee this?) and go into
3610 					 * CLOSE_WAIT state; this gives us a
3611 					 * half-closed connection
3612 					 */
3613 					nstate = IPF_TCPS_CLOSE_WAIT;
3614 				} else if (ostate < IPF_TCPS_CLOSE_WAIT) {
3615 					/*
3616 					 * still a fully established
3617 					 * connection reset timeout
3618 					 */
3619 					nstate = IPF_TCPS_ESTABLISHED;
3620 				}
3621 			}
3622 			break;
3623 
3624 		case IPF_TCPS_CLOSE_WAIT: /* 6 */
3625 			rval = 1;
3626 			if (tcpflags & TH_FIN) {
3627 				/*
3628 				 * application closed and 'dir' sent a FIN,
3629 				 * we're now going into LAST_ACK state
3630 				 */
3631 				nstate = IPF_TCPS_LAST_ACK;
3632 			} else {
3633 				/*
3634 				 * we remain in CLOSE_WAIT because the other
3635 				 * side has closed already and we did not
3636 				 * close our side yet; reset timeout
3637 				 */
3638 				nstate = IPF_TCPS_CLOSE_WAIT;
3639 			}
3640 			break;
3641 
3642 		case IPF_TCPS_FIN_WAIT_1: /* 7 */
3643 			rval = 1;
3644 			if ((tcpflags & TH_ACK) &&
3645 			    ostate > IPF_TCPS_CLOSE_WAIT) {
3646 				/*
3647 				 * if the other side is not active anymore
3648 				 * it has sent us a FIN packet that we are
3649 				 * ack'ing now with an ACK; this means both
3650 				 * sides have now closed the connection and
3651 				 * we go into TIME_WAIT
3652 				 */
3653 				/*
3654 				 * XXX: how do we know we really are ACKing
3655 				 * the FIN packet here? does the window code
3656 				 * guarantee that?
3657 				 */
3658 				nstate = IPF_TCPS_TIME_WAIT;
3659 			} else {
3660 				/*
3661 				 * we closed our side of the connection
3662 				 * already but the other side is still active
3663 				 * (ESTABLISHED/CLOSE_WAIT); continue with
3664 				 * this half-closed connection
3665 				 */
3666 				nstate = IPF_TCPS_FIN_WAIT_1;
3667 			}
3668 			break;
3669 
3670 		case IPF_TCPS_CLOSING: /* 8 */
3671 			/* NOT USED */
3672 			break;
3673 
3674 		case IPF_TCPS_LAST_ACK: /* 9 */
3675 			/*
3676 			 * We want to reset timer here to keep state in table.
3677 			 * If we would allow the state to time out here, while
3678 			 * there would still be packets being retransmitted, we
3679 			 * would cut off line between the two peers preventing
3680 			 * them to close connection properly.
3681 			 */
3682 			rval = 1;
3683 			break;
3684 
3685 		case IPF_TCPS_FIN_WAIT_2: /* 10 */
3686 			rval = 1;
3687 			if ((tcpflags & TH_OPENING) == TH_OPENING)
3688 				nstate = IPF_TCPS_SYN_RECEIVED;
3689 			else if (tcpflags & TH_SYN)
3690 				nstate = IPF_TCPS_SYN_SENT;
3691 			break;
3692 
3693 		case IPF_TCPS_TIME_WAIT: /* 11 */
3694 			/* we're in 2MSL timeout now */
3695 			rval = 1;
3696 			break;
3697 
3698 		default :
3699 #if defined(_KERNEL)
3700 # if SOLARIS
3701 			cmn_err(CE_NOTE,
3702 				"tcp %lx flags %x si %lx nstate %d ostate %d\n",
3703 				(u_long)tcp, tcpflags, (u_long)tqe,
3704 				nstate, ostate);
3705 # else
3706 			printf("tcp %lx flags %x si %lx nstate %d ostate %d\n",
3707 				(u_long)tcp, tcpflags, (u_long)tqe,
3708 				nstate, ostate);
3709 # endif
3710 #else
3711 			abort();
3712 #endif
3713 			break;
3714 		}
3715 	}
3716 
3717 	/*
3718 	 * If rval == 2 then do not update the queue position, but treat the
3719 	 * packet as being ok.
3720 	 */
3721 	if (rval == 2)
3722 		rval = 1;
3723 	else if (rval == 1) {
3724 		tqe->tqe_state[dir] = nstate;
3725 		if ((tqe->tqe_flags & TQE_RULEBASED) == 0)
3726 			fr_movequeue(tqe, tqe->tqe_ifq, tqtab + nstate, ifs);
3727 	}
3728 
3729 	return rval;
3730 }
3731 
3732 
3733 /* ------------------------------------------------------------------------ */
3734 /* Function:    ipstate_log                                                 */
3735 /* Returns:     Nil                                                         */
3736 /* Parameters:  is(I)   - pointer to state structure                        */
3737 /*              type(I) - type of log entry to create                       */
3738 /*                                                                          */
3739 /* Creates a state table log entry using the state structure and type info. */
3740 /* passed in.  Log packet/byte counts, source/destination address and other */
3741 /* protocol specific information.                                           */
3742 /* ------------------------------------------------------------------------ */
3743 void ipstate_log(is, type, ifs)
3744 struct ipstate *is;
3745 u_int type;
3746 ipf_stack_t *ifs;
3747 {
3748 #ifdef	IPFILTER_LOG
3749 	struct	ipslog	ipsl;
3750 	size_t sizes[1];
3751 	void *items[1];
3752 	int types[1];
3753 
3754 	/*
3755 	 * Copy information out of the ipstate_t structure and into the
3756 	 * structure used for logging.
3757 	 */
3758 	ipsl.isl_type = type;
3759 	ipsl.isl_pkts[0] = is->is_pkts[0] + is->is_icmppkts[0];
3760 	ipsl.isl_bytes[0] = is->is_bytes[0];
3761 	ipsl.isl_pkts[1] = is->is_pkts[1] + is->is_icmppkts[1];
3762 	ipsl.isl_bytes[1] = is->is_bytes[1];
3763 	ipsl.isl_pkts[2] = is->is_pkts[2] + is->is_icmppkts[2];
3764 	ipsl.isl_bytes[2] = is->is_bytes[2];
3765 	ipsl.isl_pkts[3] = is->is_pkts[3] + is->is_icmppkts[3];
3766 	ipsl.isl_bytes[3] = is->is_bytes[3];
3767 	ipsl.isl_src = is->is_src;
3768 	ipsl.isl_dst = is->is_dst;
3769 	ipsl.isl_p = is->is_p;
3770 	ipsl.isl_v = is->is_v;
3771 	ipsl.isl_flags = is->is_flags;
3772 	ipsl.isl_tag = is->is_tag;
3773 	ipsl.isl_rulen = is->is_rulen;
3774 	(void) strncpy(ipsl.isl_group, is->is_group, FR_GROUPLEN);
3775 
3776 	if (ipsl.isl_p == IPPROTO_TCP || ipsl.isl_p == IPPROTO_UDP) {
3777 		ipsl.isl_sport = is->is_sport;
3778 		ipsl.isl_dport = is->is_dport;
3779 		if (ipsl.isl_p == IPPROTO_TCP) {
3780 			ipsl.isl_state[0] = is->is_state[0];
3781 			ipsl.isl_state[1] = is->is_state[1];
3782 		}
3783 	} else if (ipsl.isl_p == IPPROTO_ICMP) {
3784 		ipsl.isl_itype = is->is_icmp.ici_type;
3785 	} else if (ipsl.isl_p == IPPROTO_ICMPV6) {
3786 		ipsl.isl_itype = is->is_icmp.ici_type;
3787 	} else {
3788 		ipsl.isl_ps.isl_filler[0] = 0;
3789 		ipsl.isl_ps.isl_filler[1] = 0;
3790 	}
3791 
3792 	items[0] = &ipsl;
3793 	sizes[0] = sizeof(ipsl);
3794 	types[0] = 0;
3795 
3796 	if (ipllog(IPL_LOGSTATE, NULL, items, sizes, types, 1, ifs)) {
3797 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_logged);
3798 	} else {
3799 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_logfail);
3800 	}
3801 #endif
3802 }
3803 
3804 
3805 #ifdef	USE_INET6
3806 /* ------------------------------------------------------------------------ */
3807 /* Function:    fr_checkicmp6matchingstate                                  */
3808 /* Returns:     ipstate_t* - NULL == no match found,                        */
3809 /*                           else  pointer to matching state entry          */
3810 /* Parameters:  fin(I) - pointer to packet information                      */
3811 /* Locks:       NULL == no locks, else Read Lock on ipf_state               */
3812 /*                                                                          */
3813 /* If we've got an ICMPv6 error message, using the information stored in    */
3814 /* the ICMPv6 packet, look for a matching state table entry.                */
3815 /* ------------------------------------------------------------------------ */
3816 static ipstate_t *fr_checkicmp6matchingstate(fin)
3817 fr_info_t *fin;
3818 {
3819 	struct icmp6_hdr *ic6, *oic;
3820 	int backward, i;
3821 	ipstate_t *is, **isp;
3822 	u_short sport, dport;
3823 	i6addr_t dst, src;
3824 	u_short savelen;
3825 	icmpinfo_t *ic;
3826 	fr_info_t ofin;
3827 	tcphdr_t *tcp;
3828 	ip6_t *oip6;
3829 	u_char	pr;
3830 	u_int hv;
3831 	ipf_stack_t *ifs = fin->fin_ifs;
3832 
3833 	/*
3834 	 * Does it at least have the return (basic) IP header ?
3835 	 * Is it an actual recognised ICMP error type?
3836 	 * Only a basic IP header (no options) should be with
3837 	 * an ICMP error header.
3838 	 */
3839 	if ((fin->fin_v != 6) || (fin->fin_plen < ICMP6ERR_MINPKTLEN) ||
3840 	    !(fin->fin_flx & FI_ICMPERR))
3841 		return NULL;
3842 
3843 	ic6 = fin->fin_dp;
3844 
3845 	oip6 = (ip6_t *)((char *)ic6 + ICMPERR_ICMPHLEN);
3846 	if (fin->fin_plen < sizeof(*oip6))
3847 		return NULL;
3848 
3849 	bcopy((char *)fin, (char *)&ofin, sizeof(*fin));
3850 	ofin.fin_v = 6;
3851 	ofin.fin_ifp = fin->fin_ifp;
3852 	ofin.fin_out = !fin->fin_out;
3853 	ofin.fin_m = NULL;	/* if dereferenced, panic XXX */
3854 	ofin.fin_mp = NULL;	/* if dereferenced, panic XXX */
3855 
3856 	/*
3857 	 * We make a fin entry to be able to feed it to
3858 	 * matchsrcdst. Note that not all fields are necessary
3859 	 * but this is the cleanest way. Note further we fill
3860 	 * in fin_mp such that if someone uses it we'll get
3861 	 * a kernel panic. fr_matchsrcdst does not use this.
3862 	 *
3863 	 * watch out here, as ip is in host order and oip6 in network
3864 	 * order. Any change we make must be undone afterwards.
3865 	 */
3866 	savelen = oip6->ip6_plen;
3867 	oip6->ip6_plen = fin->fin_dlen - ICMPERR_ICMPHLEN;
3868 	ofin.fin_flx = FI_NOCKSUM;
3869 	ofin.fin_ip = (ip_t *)oip6;
3870 	ofin.fin_plen = oip6->ip6_plen;
3871 	(void) fr_makefrip(sizeof(*oip6), (ip_t *)oip6, &ofin);
3872 	ofin.fin_flx &= ~(FI_BAD|FI_SHORT);
3873 	oip6->ip6_plen = savelen;
3874 
3875 	if (oip6->ip6_nxt == IPPROTO_ICMPV6) {
3876 		oic = (struct icmp6_hdr *)(oip6 + 1);
3877 		/*
3878 		 * an ICMP error can only be generated as a result of an
3879 		 * ICMP query, not as the response on an ICMP error
3880 		 *
3881 		 * XXX theoretically ICMP_ECHOREP and the other reply's are
3882 		 * ICMP query's as well, but adding them here seems strange XXX
3883 		 */
3884 		 if (!(oic->icmp6_type & ICMP6_INFOMSG_MASK))
3885 		    	return NULL;
3886 
3887 		/*
3888 		 * perform a lookup of the ICMP packet in the state table
3889 		 */
3890 		hv = (pr = oip6->ip6_nxt);
3891 		src.in6 = oip6->ip6_src;
3892 		hv += src.in4.s_addr;
3893 		dst.in6 = oip6->ip6_dst;
3894 		hv += dst.in4.s_addr;
3895 		hv += oic->icmp6_id;
3896 		hv += oic->icmp6_seq;
3897 		hv = DOUBLE_HASH(hv, ifs);
3898 
3899 		READ_ENTER(&ifs->ifs_ipf_state);
3900 		for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
3901 			ic = &is->is_icmp;
3902 			isp = &is->is_hnext;
3903 			if ((is->is_p == pr) &&
3904 			    !(is->is_pass & FR_NOICMPERR) &&
3905 			    (oic->icmp6_id == ic->ici_id) &&
3906 			    (oic->icmp6_seq == ic->ici_seq) &&
3907 			    (is = fr_matchsrcdst(&ofin, is, &src,
3908 						 &dst, NULL, FI_ICMPCMP))) {
3909 			    	/*
3910 			    	 * in the state table ICMP query's are stored
3911 			    	 * with the type of the corresponding ICMP
3912 			    	 * response. Correct here
3913 			    	 */
3914 				if (((ic->ici_type == ICMP6_ECHO_REPLY) &&
3915 				     (oic->icmp6_type == ICMP6_ECHO_REQUEST)) ||
3916 				     (ic->ici_type - 1 == oic->icmp6_type )) {
3917 				    	ifs->ifs_ips_stats.iss_hits++;
3918 					backward = IP6_NEQ(&is->is_dst, &src);
3919 					fin->fin_rev = !backward;
3920 					i = (backward << 1) + fin->fin_out;
3921     					is->is_icmppkts[i]++;
3922 					return is;
3923 				}
3924 			}
3925 		}
3926 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
3927 		return NULL;
3928 	}
3929 
3930 	hv = (pr = oip6->ip6_nxt);
3931 	src.in6 = oip6->ip6_src;
3932 	hv += src.i6[0];
3933 	hv += src.i6[1];
3934 	hv += src.i6[2];
3935 	hv += src.i6[3];
3936 	dst.in6 = oip6->ip6_dst;
3937 	hv += dst.i6[0];
3938 	hv += dst.i6[1];
3939 	hv += dst.i6[2];
3940 	hv += dst.i6[3];
3941 
3942 	if ((oip6->ip6_nxt == IPPROTO_TCP) || (oip6->ip6_nxt == IPPROTO_UDP)) {
3943 		tcp = (tcphdr_t *)(oip6 + 1);
3944 		dport = tcp->th_dport;
3945 		sport = tcp->th_sport;
3946 		hv += dport;
3947 		hv += sport;
3948 	} else
3949 		tcp = NULL;
3950 	hv = DOUBLE_HASH(hv, ifs);
3951 
3952 	READ_ENTER(&ifs->ifs_ipf_state);
3953 	for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
3954 		isp = &is->is_hnext;
3955 		/*
3956 		 * Only allow this icmp though if the
3957 		 * encapsulated packet was allowed through the
3958 		 * other way around. Note that the minimal amount
3959 		 * of info present does not allow for checking against
3960 		 * tcp internals such as seq and ack numbers.
3961 		 */
3962 		if ((is->is_p != pr) || (is->is_v != 6) ||
3963 		    (is->is_pass & FR_NOICMPERR))
3964 			continue;
3965 		is = fr_matchsrcdst(&ofin, is, &src, &dst, tcp, FI_ICMPCMP);
3966 		if (is != NULL) {
3967 			ifs->ifs_ips_stats.iss_hits++;
3968 			backward = IP6_NEQ(&is->is_dst, &src);
3969 			fin->fin_rev = !backward;
3970 			i = (backward << 1) + fin->fin_out;
3971 			is->is_icmppkts[i]++;
3972 			/*
3973 			 * we deliberately do not touch the timeouts
3974 			 * for the accompanying state table entry.
3975 			 * It remains to be seen if that is correct. XXX
3976 			 */
3977 			return is;
3978 		}
3979 	}
3980 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
3981 	return NULL;
3982 }
3983 #endif
3984 
3985 
3986 /* ------------------------------------------------------------------------ */
3987 /* Function:    fr_sttab_init                                               */
3988 /* Returns:     Nil                                                         */
3989 /* Parameters:  tqp(I) - pointer to an array of timeout queues for TCP      */
3990 /*                                                                          */
3991 /* Initialise the array of timeout queues for TCP.                          */
3992 /* ------------------------------------------------------------------------ */
3993 void fr_sttab_init(tqp, ifs)
3994 ipftq_t *tqp;
3995 ipf_stack_t *ifs;
3996 {
3997 	int i;
3998 
3999 	for (i = IPF_TCP_NSTATES - 1; i >= 0; i--) {
4000 		tqp[i].ifq_ttl = 0;
4001 		tqp[i].ifq_ref = 1;
4002 		tqp[i].ifq_head = NULL;
4003 		tqp[i].ifq_tail = &tqp[i].ifq_head;
4004 		tqp[i].ifq_next = tqp + i + 1;
4005 		MUTEX_INIT(&tqp[i].ifq_lock, "ipftq tcp tab");
4006 	}
4007 	tqp[IPF_TCP_NSTATES - 1].ifq_next = NULL;
4008 	tqp[IPF_TCPS_CLOSED].ifq_ttl = ifs->ifs_fr_tcpclosed;
4009 	tqp[IPF_TCPS_LISTEN].ifq_ttl = ifs->ifs_fr_tcptimeout;
4010 	tqp[IPF_TCPS_SYN_SENT].ifq_ttl = ifs->ifs_fr_tcptimeout;
4011 	tqp[IPF_TCPS_SYN_RECEIVED].ifq_ttl = ifs->ifs_fr_tcptimeout;
4012 	tqp[IPF_TCPS_ESTABLISHED].ifq_ttl = ifs->ifs_fr_tcpidletimeout;
4013 	tqp[IPF_TCPS_CLOSE_WAIT].ifq_ttl = ifs->ifs_fr_tcphalfclosed;
4014 	tqp[IPF_TCPS_FIN_WAIT_1].ifq_ttl = ifs->ifs_fr_tcphalfclosed;
4015 	tqp[IPF_TCPS_CLOSING].ifq_ttl = ifs->ifs_fr_tcptimeout;
4016 	tqp[IPF_TCPS_LAST_ACK].ifq_ttl = ifs->ifs_fr_tcplastack;
4017 	tqp[IPF_TCPS_FIN_WAIT_2].ifq_ttl = ifs->ifs_fr_tcpclosewait;
4018 	tqp[IPF_TCPS_TIME_WAIT].ifq_ttl = ifs->ifs_fr_tcptimeout;
4019 	tqp[IPF_TCPS_HALF_ESTAB].ifq_ttl = ifs->ifs_fr_tcptimeout;
4020 }
4021 
4022 
4023 /* ------------------------------------------------------------------------ */
4024 /* Function:    fr_sttab_destroy                                            */
4025 /* Returns:     Nil                                                         */
4026 /* Parameters:  tqp(I) - pointer to an array of timeout queues for TCP      */
4027 /*                                                                          */
4028 /* Do whatever is necessary to "destroy" each of the entries in the array   */
4029 /* of timeout queues for TCP.                                               */
4030 /* ------------------------------------------------------------------------ */
4031 void fr_sttab_destroy(tqp)
4032 ipftq_t *tqp;
4033 {
4034 	int i;
4035 
4036 	for (i = IPF_TCP_NSTATES - 1; i >= 0; i--)
4037 		MUTEX_DESTROY(&tqp[i].ifq_lock);
4038 }
4039 
4040 
4041 /* ------------------------------------------------------------------------ */
4042 /* Function:    fr_statederef                                               */
4043 /* Returns:     Nil                                                         */
4044 /* Parameters:  isp(I) - pointer to pointer to state table entry            */
4045 /*              ifs - ipf stack instance                                    */
4046 /*                                                                          */
4047 /* Decrement the reference counter for this state table entry and free it   */
4048 /* if there are no more things using it.                                    */
4049 /*                                                                          */
4050 /* Internal parameters:                                                     */
4051 /*    state[0] = state of source (host that initiated connection)           */
4052 /*    state[1] = state of dest   (host that accepted the connection)        */
4053 /* ------------------------------------------------------------------------ */
4054 void fr_statederef(isp, ifs)
4055 ipstate_t **isp;
4056 ipf_stack_t *ifs;
4057 {
4058 	ipstate_t *is;
4059 
4060 	is = *isp;
4061 	*isp = NULL;
4062 
4063 	MUTEX_ENTER(&is->is_lock);
4064 	if (is->is_ref > 1) {
4065 		is->is_ref--;
4066 		MUTEX_EXIT(&is->is_lock);
4067 #ifndef	_KERNEL
4068 		if ((is->is_sti.tqe_state[0] > IPF_TCPS_ESTABLISHED) ||
4069 		   (is->is_sti.tqe_state[1] > IPF_TCPS_ESTABLISHED)) {
4070 			fr_delstate(is, ISL_ORPHAN, ifs);
4071 		}
4072 #endif
4073 		return;
4074 	}
4075 	MUTEX_EXIT(&is->is_lock);
4076 
4077 	WRITE_ENTER(&ifs->ifs_ipf_state);
4078 	fr_delstate(is, ISL_EXPIRE, ifs);
4079 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
4080 }
4081 
4082 
4083 /* ------------------------------------------------------------------------ */
4084 /* Function:    fr_setstatequeue                                            */
4085 /* Returns:     Nil                                                         */
4086 /* Parameters:  is(I) - pointer to state structure                          */
4087 /*              rev(I) - forward(0) or reverse(1) direction                 */
4088 /* Locks:       ipf_state (read or write)                                   */
4089 /*                                                                          */
4090 /* Put the state entry on its default queue entry, using rev as a helped in */
4091 /* determining which queue it should be placed on.                          */
4092 /* ------------------------------------------------------------------------ */
4093 void fr_setstatequeue(is, rev, ifs)
4094 ipstate_t *is;
4095 int rev;
4096 ipf_stack_t *ifs;
4097 {
4098 	ipftq_t *oifq, *nifq;
4099 
4100 
4101 	if ((is->is_sti.tqe_flags & TQE_RULEBASED) != 0)
4102 		nifq = is->is_tqehead[rev];
4103 	else
4104 		nifq = NULL;
4105 
4106 	if (nifq == NULL) {
4107 		switch (is->is_p)
4108 		{
4109 #ifdef USE_INET6
4110 		case IPPROTO_ICMPV6 :
4111 			if (rev == 1)
4112 				nifq = &ifs->ifs_ips_icmpacktq;
4113 			else
4114 				nifq = &ifs->ifs_ips_icmptq;
4115 			break;
4116 #endif
4117 		case IPPROTO_ICMP :
4118 			if (rev == 1)
4119 				nifq = &ifs->ifs_ips_icmpacktq;
4120 			else
4121 				nifq = &ifs->ifs_ips_icmptq;
4122 			break;
4123 		case IPPROTO_TCP :
4124 			nifq = ifs->ifs_ips_tqtqb + is->is_state[rev];
4125 			break;
4126 
4127 		case IPPROTO_UDP :
4128 			if (rev == 1)
4129 				nifq = &ifs->ifs_ips_udpacktq;
4130 			else
4131 				nifq = &ifs->ifs_ips_udptq;
4132 			break;
4133 
4134 		default :
4135 			nifq = &ifs->ifs_ips_iptq;
4136 			break;
4137 		}
4138 	}
4139 
4140 	oifq = is->is_sti.tqe_ifq;
4141 	/*
4142 	 * If it's currently on a timeout queue, move it from one queue to
4143 	 * another, else put it on the end of the newly determined queue.
4144 	 */
4145 	if (oifq != NULL)
4146 		fr_movequeue(&is->is_sti, oifq, nifq, ifs);
4147 	else
4148 		fr_queueappend(&is->is_sti, nifq, is, ifs);
4149 	return;
4150 }
4151 
4152 
4153 /* ------------------------------------------------------------------------ */
4154 /* Function:    fr_stateiter                                                */
4155 /* Returns:     int - 0 == success, else error                              */
4156 /* Parameters:  token(I) - pointer to ipftoken structure                    */
4157 /*              itp(I)   - pointer to ipfgeniter structure                  */
4158 /*                                                                          */
4159 /* This function handles the SIOCGENITER ioctl for the state tables and     */
4160 /* walks through the list of entries in the state table list (ips_list.)    */
4161 /* ------------------------------------------------------------------------ */
4162 static int fr_stateiter(token, itp, ifs)
4163 ipftoken_t *token;
4164 ipfgeniter_t *itp;
4165 ipf_stack_t *ifs;
4166 {
4167 	ipstate_t *is, *next, zero;
4168 	int error, count;
4169 	char *dst;
4170 
4171 	if (itp->igi_data == NULL)
4172 		return EFAULT;
4173 
4174 	if (itp->igi_nitems == 0)
4175 		return EINVAL;
4176 
4177 	if (itp->igi_type != IPFGENITER_STATE)
4178 		return EINVAL;
4179 
4180 	error = 0;
4181 
4182 	READ_ENTER(&ifs->ifs_ipf_state);
4183 
4184 	/*
4185 	 * Get "previous" entry from the token and find the next entry.
4186 	 */
4187 	is = token->ipt_data;
4188 	if (is == NULL) {
4189 		next = ifs->ifs_ips_list;
4190 	} else {
4191 		next = is->is_next;
4192 	}
4193 
4194 	dst = itp->igi_data;
4195 	for (count = itp->igi_nitems; count > 0; count--) {
4196 		/*
4197 		 * If we found an entry, add a reference to it and update the token.
4198 		 * Otherwise, zero out data to be returned and NULL out token.
4199 		 */
4200 		if (next != NULL) {
4201 			MUTEX_ENTER(&next->is_lock);
4202 			next->is_ref++;
4203 			MUTEX_EXIT(&next->is_lock);
4204 			token->ipt_data = next;
4205 		} else {
4206 			bzero(&zero, sizeof(zero));
4207 			next = &zero;
4208 			token->ipt_data = NULL;
4209 		}
4210 
4211 		/*
4212 		 * Safe to release lock now the we have a reference.
4213 		 */
4214 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
4215 
4216 		/*
4217 		 * Copy out data and clean up references and tokens.
4218 		 */
4219 		error = COPYOUT(next, dst, sizeof(*next));
4220 		if (error != 0)
4221 			error = EFAULT;
4222 		if (token->ipt_data == NULL) {
4223 			ipf_freetoken(token, ifs);
4224 			break;
4225 		} else {
4226 			if (is != NULL)
4227 				fr_statederef(&is, ifs);
4228 			if (next->is_next == NULL) {
4229 				ipf_freetoken(token, ifs);
4230 				break;
4231 			}
4232 		}
4233 
4234 		if ((count == 1) || (error != 0))
4235 			break;
4236 
4237 		READ_ENTER(&ifs->ifs_ipf_state);
4238 		dst += sizeof(*next);
4239 		is = next;
4240 		next = is->is_next;
4241 	}
4242 
4243 	return error;
4244 }
4245