xref: /illumos-gate/usr/src/uts/common/inet/ipf/ip_state.c (revision d3edf6ac)
1 /*
2  * Copyright (C) 1995-2003 by Darren Reed.
3  *
4  * See the IPFILTER.LICENCE file for details on licencing.
5  *
6  * Copyright (c) 2003, 2010, Oracle and/or its affiliates. All rights reserved.
7  */
8 
9 #if defined(KERNEL) || defined(_KERNEL)
10 # undef KERNEL
11 # undef _KERNEL
12 # define        KERNEL	1
13 # define        _KERNEL	1
14 #endif
15 #include <sys/errno.h>
16 #include <sys/types.h>
17 #include <sys/param.h>
18 #include <sys/file.h>
19 #if defined(__NetBSD__) && (NetBSD >= 199905) && !defined(IPFILTER_LKM) && \
20     defined(_KERNEL)
21 # include "opt_ipfilter_log.h"
22 #endif
23 #if defined(_KERNEL) && defined(__FreeBSD_version) && \
24     (__FreeBSD_version >= 400000) && !defined(KLD_MODULE)
25 #include "opt_inet6.h"
26 #endif
27 #if !defined(_KERNEL) && !defined(__KERNEL__)
28 # include <stdio.h>
29 # include <stdlib.h>
30 # include <string.h>
31 # define _KERNEL
32 # ifdef __OpenBSD__
33 struct file;
34 # endif
35 # include <sys/uio.h>
36 # undef _KERNEL
37 #endif
38 #if defined(_KERNEL) && (__FreeBSD_version >= 220000)
39 # include <sys/filio.h>
40 # include <sys/fcntl.h>
41 # if (__FreeBSD_version >= 300000) && !defined(IPFILTER_LKM)
42 #  include "opt_ipfilter.h"
43 # endif
44 #else
45 # include <sys/ioctl.h>
46 #endif
47 #include <sys/time.h>
48 #if !defined(linux)
49 # include <sys/protosw.h>
50 #endif
51 #include <sys/socket.h>
52 #if defined(_KERNEL)
53 # include <sys/systm.h>
54 # if !defined(__SVR4) && !defined(__svr4__)
55 #  include <sys/mbuf.h>
56 # endif
57 #endif
58 #if defined(__SVR4) || defined(__svr4__)
59 # include <sys/filio.h>
60 # include <sys/byteorder.h>
61 # ifdef _KERNEL
62 #  include <sys/dditypes.h>
63 # endif
64 # include <sys/stream.h>
65 # include <sys/kmem.h>
66 #endif
67 
68 #include <net/if.h>
69 #ifdef sun
70 # include <net/af.h>
71 #endif
72 #include <net/route.h>
73 #include <netinet/in.h>
74 #include <netinet/in_systm.h>
75 #include <netinet/ip.h>
76 #include <netinet/tcp.h>
77 #if !defined(linux)
78 # include <netinet/ip_var.h>
79 #endif
80 #if !defined(__hpux) && !defined(linux)
81 # include <netinet/tcp_fsm.h>
82 #endif
83 #include <netinet/udp.h>
84 #include <netinet/ip_icmp.h>
85 #include "netinet/ip_compat.h"
86 #include <netinet/tcpip.h>
87 #include "netinet/ip_fil.h"
88 #include "netinet/ip_nat.h"
89 #include "netinet/ip_frag.h"
90 #include "netinet/ip_state.h"
91 #include "netinet/ip_proxy.h"
92 #include "netinet/ipf_stack.h"
93 #ifdef	IPFILTER_SYNC
94 #include "netinet/ip_sync.h"
95 #endif
96 #ifdef	IPFILTER_SCAN
97 #include "netinet/ip_scan.h"
98 #endif
99 #ifdef	USE_INET6
100 #include <netinet/icmp6.h>
101 #endif
102 #if (__FreeBSD_version >= 300000)
103 # include <sys/malloc.h>
104 # if defined(_KERNEL) && !defined(IPFILTER_LKM)
105 #  include <sys/libkern.h>
106 #  include <sys/systm.h>
107 # endif
108 #endif
109 /* END OF INCLUDES */
110 
111 
112 #if !defined(lint)
113 static const char sccsid[] = "@(#)ip_state.c	1.8 6/5/96 (C) 1993-2000 Darren Reed";
114 static const char rcsid[] = "@(#)$Id: ip_state.c,v 2.186.2.36 2005/08/11 19:58:03 darrenr Exp $";
115 #endif
116 
117 #ifdef	USE_INET6
118 static ipstate_t *fr_checkicmp6matchingstate __P((fr_info_t *));
119 #endif
120 static ipstate_t *fr_matchsrcdst __P((fr_info_t *, ipstate_t *, i6addr_t *,
121 				      i6addr_t *, tcphdr_t *, u_32_t));
122 static ipstate_t *fr_checkicmpmatchingstate __P((fr_info_t *));
123 static int fr_state_flush __P((int, int, ipf_stack_t *));
124 static ips_stat_t *fr_statetstats __P((ipf_stack_t *));
125 static int fr_state_remove __P((caddr_t, ipf_stack_t *));
126 static void fr_ipsmove __P((ipstate_t *, u_int, ipf_stack_t *));
127 static int fr_tcpstate __P((fr_info_t *, tcphdr_t *, ipstate_t *));
128 static int fr_tcpoptions __P((fr_info_t *, tcphdr_t *, tcpdata_t *));
129 static ipstate_t *fr_stclone __P((fr_info_t *, tcphdr_t *, ipstate_t *));
130 static void fr_fixinisn __P((fr_info_t *, ipstate_t *));
131 static void fr_fixoutisn __P((fr_info_t *, ipstate_t *));
132 static void fr_checknewisn __P((fr_info_t *, ipstate_t *));
133 static int fr_stateiter __P((ipftoken_t *, ipfgeniter_t *, ipf_stack_t *));
134 
135 int fr_stputent __P((caddr_t, ipf_stack_t *));
136 int fr_stgetent __P((caddr_t, ipf_stack_t *));
137 
138 #define	ONE_DAY		IPF_TTLVAL(1 * 86400)	/* 1 day */
139 #define	FIVE_DAYS	(5 * ONE_DAY)
140 #define	DOUBLE_HASH(x, ifs)	\
141     (((x) + ifs->ifs_ips_seed[(x) % ifs->ifs_fr_statesize]) % ifs->ifs_fr_statesize)
142 
143 
144 /* ------------------------------------------------------------------------ */
145 /* Function:    fr_stateinit                                                */
146 /* Returns:     int - 0 == success, -1 == failure                           */
147 /* Parameters:  ifs - ipf stack instance                                    */
148 /*                                                                          */
149 /* Initialise all the global variables used within the state code.          */
150 /* This action also includes initiailising locks.                           */
151 /* ------------------------------------------------------------------------ */
152 int fr_stateinit(ifs)
153 ipf_stack_t *ifs;
154 {
155 #if defined(NEED_LOCAL_RAND) || !defined(_KERNEL)
156 	struct timeval tv;
157 #endif
158 	int i;
159 
160 	KMALLOCS(ifs->ifs_ips_table, ipstate_t **,
161 		 ifs->ifs_fr_statesize * sizeof(ipstate_t *));
162 	if (ifs->ifs_ips_table == NULL)
163 		return -1;
164 	bzero((char *)ifs->ifs_ips_table,
165 	      ifs->ifs_fr_statesize * sizeof(ipstate_t *));
166 
167 	KMALLOCS(ifs->ifs_ips_seed, u_long *,
168 		 ifs->ifs_fr_statesize * sizeof(*ifs->ifs_ips_seed));
169 	if (ifs->ifs_ips_seed == NULL)
170 		return -2;
171 #if defined(NEED_LOCAL_RAND) || !defined(_KERNEL)
172 	tv.tv_sec = 0;
173 	GETKTIME(&tv);
174 #endif
175 	for (i = 0; i < ifs->ifs_fr_statesize; i++) {
176 		/*
177 		 * XXX - ips_seed[X] should be a random number of sorts.
178 		 */
179 #if !defined(NEED_LOCAL_RAND) && defined(_KERNEL)
180 		ifs->ifs_ips_seed[i] = ipf_random();
181 #else
182 		ifs->ifs_ips_seed[i] = ((u_long)ifs->ifs_ips_seed + i) *
183 		    ifs->ifs_fr_statesize;
184 		ifs->ifs_ips_seed[i] += tv.tv_sec;
185 		ifs->ifs_ips_seed[i] *= (u_long)ifs->ifs_ips_seed;
186 		ifs->ifs_ips_seed[i] ^= 0x5a5aa5a5;
187 		ifs->ifs_ips_seed[i] *= ifs->ifs_fr_statemax;
188 #endif
189 	}
190 
191 	/* fill icmp reply type table */
192 	for (i = 0; i <= ICMP_MAXTYPE; i++)
193 		icmpreplytype4[i] = -1;
194 	icmpreplytype4[ICMP_ECHO] = ICMP_ECHOREPLY;
195 	icmpreplytype4[ICMP_TSTAMP] = ICMP_TSTAMPREPLY;
196 	icmpreplytype4[ICMP_IREQ] = ICMP_IREQREPLY;
197 	icmpreplytype4[ICMP_MASKREQ] = ICMP_MASKREPLY;
198 #ifdef	USE_INET6
199 	/* fill icmp reply type table */
200 	for (i = 0; i <= ICMP6_MAXTYPE; i++)
201 		icmpreplytype6[i] = -1;
202 	icmpreplytype6[ICMP6_ECHO_REQUEST] = ICMP6_ECHO_REPLY;
203 	icmpreplytype6[ICMP6_MEMBERSHIP_QUERY] = ICMP6_MEMBERSHIP_REPORT;
204 	icmpreplytype6[ICMP6_NI_QUERY] = ICMP6_NI_REPLY;
205 	icmpreplytype6[ND_ROUTER_SOLICIT] = ND_ROUTER_ADVERT;
206 	icmpreplytype6[ND_NEIGHBOR_SOLICIT] = ND_NEIGHBOR_ADVERT;
207 #endif
208 
209 	KMALLOCS(ifs->ifs_ips_stats.iss_bucketlen, u_long *,
210 		 ifs->ifs_fr_statesize * sizeof(u_long));
211 	if (ifs->ifs_ips_stats.iss_bucketlen == NULL)
212 		return -1;
213 	bzero((char *)ifs->ifs_ips_stats.iss_bucketlen,
214 	      ifs->ifs_fr_statesize * sizeof(u_long));
215 
216 	if (ifs->ifs_fr_state_maxbucket == 0) {
217 		for (i = ifs->ifs_fr_statesize; i > 0; i >>= 1)
218 			ifs->ifs_fr_state_maxbucket++;
219 		ifs->ifs_fr_state_maxbucket *= 2;
220 	}
221 
222 	fr_sttab_init(ifs->ifs_ips_tqtqb, ifs);
223 	ifs->ifs_ips_tqtqb[IPF_TCP_NSTATES - 1].ifq_next = &ifs->ifs_ips_udptq;
224 	ifs->ifs_ips_udptq.ifq_ttl = (u_long)ifs->ifs_fr_udptimeout;
225 	ifs->ifs_ips_udptq.ifq_ref = 1;
226 	ifs->ifs_ips_udptq.ifq_head = NULL;
227 	ifs->ifs_ips_udptq.ifq_tail = &ifs->ifs_ips_udptq.ifq_head;
228 	MUTEX_INIT(&ifs->ifs_ips_udptq.ifq_lock, "ipftq udp tab");
229 	ifs->ifs_ips_udptq.ifq_next = &ifs->ifs_ips_udpacktq;
230 	ifs->ifs_ips_udpacktq.ifq_ttl = (u_long)ifs->ifs_fr_udpacktimeout;
231 	ifs->ifs_ips_udpacktq.ifq_ref = 1;
232 	ifs->ifs_ips_udpacktq.ifq_head = NULL;
233 	ifs->ifs_ips_udpacktq.ifq_tail = &ifs->ifs_ips_udpacktq.ifq_head;
234 	MUTEX_INIT(&ifs->ifs_ips_udpacktq.ifq_lock, "ipftq udpack tab");
235 	ifs->ifs_ips_udpacktq.ifq_next = &ifs->ifs_ips_icmptq;
236 	ifs->ifs_ips_icmptq.ifq_ttl = (u_long)ifs->ifs_fr_icmptimeout;
237 	ifs->ifs_ips_icmptq.ifq_ref = 1;
238 	ifs->ifs_ips_icmptq.ifq_head = NULL;
239 	ifs->ifs_ips_icmptq.ifq_tail = &ifs->ifs_ips_icmptq.ifq_head;
240 	MUTEX_INIT(&ifs->ifs_ips_icmptq.ifq_lock, "ipftq icmp tab");
241 	ifs->ifs_ips_icmptq.ifq_next = &ifs->ifs_ips_icmpacktq;
242 	ifs->ifs_ips_icmpacktq.ifq_ttl = (u_long)ifs->ifs_fr_icmpacktimeout;
243 	ifs->ifs_ips_icmpacktq.ifq_ref = 1;
244 	ifs->ifs_ips_icmpacktq.ifq_head = NULL;
245 	ifs->ifs_ips_icmpacktq.ifq_tail = &ifs->ifs_ips_icmpacktq.ifq_head;
246 	MUTEX_INIT(&ifs->ifs_ips_icmpacktq.ifq_lock, "ipftq icmpack tab");
247 	ifs->ifs_ips_icmpacktq.ifq_next = &ifs->ifs_ips_iptq;
248 	ifs->ifs_ips_iptq.ifq_ttl = (u_long)ifs->ifs_fr_iptimeout;
249 	ifs->ifs_ips_iptq.ifq_ref = 1;
250 	ifs->ifs_ips_iptq.ifq_head = NULL;
251 	ifs->ifs_ips_iptq.ifq_tail = &ifs->ifs_ips_iptq.ifq_head;
252 	MUTEX_INIT(&ifs->ifs_ips_iptq.ifq_lock, "ipftq ip tab");
253 	ifs->ifs_ips_iptq.ifq_next = &ifs->ifs_ips_deletetq;
254 	/* entry's ttl in deletetq is just 1 tick */
255 	ifs->ifs_ips_deletetq.ifq_ttl = (u_long) 1;
256 	ifs->ifs_ips_deletetq.ifq_ref = 1;
257 	ifs->ifs_ips_deletetq.ifq_head = NULL;
258 	ifs->ifs_ips_deletetq.ifq_tail = &ifs->ifs_ips_deletetq.ifq_head;
259 	MUTEX_INIT(&ifs->ifs_ips_deletetq.ifq_lock, "state delete queue");
260 	ifs->ifs_ips_deletetq.ifq_next = NULL;
261 
262 	RWLOCK_INIT(&ifs->ifs_ipf_state, "ipf IP state rwlock");
263 	MUTEX_INIT(&ifs->ifs_ipf_stinsert, "ipf state insert mutex");
264 	ifs->ifs_fr_state_init = 1;
265 
266 	ifs->ifs_ips_last_force_flush = ifs->ifs_fr_ticks;
267 	return 0;
268 }
269 
270 
271 /* ------------------------------------------------------------------------ */
272 /* Function:    fr_stateunload                                              */
273 /* Returns:     Nil                                                         */
274 /* Parameters:  ifs - ipf stack instance                                    */
275 /*                                                                          */
276 /* Release and destroy any resources acquired or initialised so that        */
277 /* IPFilter can be unloaded or re-initialised.                              */
278 /* ------------------------------------------------------------------------ */
279 void fr_stateunload(ifs)
280 ipf_stack_t *ifs;
281 {
282 	ipftq_t *ifq, *ifqnext;
283 	ipstate_t *is;
284 
285 	while ((is = ifs->ifs_ips_list) != NULL)
286 	    (void) fr_delstate(is, 0, ifs);
287 
288 	/*
289 	 * Proxy timeout queues are not cleaned here because although they
290 	 * exist on the state list, appr_unload is called after fr_stateunload
291 	 * and the proxies actually are responsible for them being created.
292 	 * Should the proxy timeouts have their own list?  There's no real
293 	 * justification as this is the only complicationA
294 	 */
295 	for (ifq = ifs->ifs_ips_utqe; ifq != NULL; ifq = ifqnext) {
296 		ifqnext = ifq->ifq_next;
297 		if (((ifq->ifq_flags & IFQF_PROXY) == 0) &&
298 		    (fr_deletetimeoutqueue(ifq) == 0))
299 			fr_freetimeoutqueue(ifq, ifs);
300 	}
301 
302 	ifs->ifs_ips_stats.iss_inuse = 0;
303 	ifs->ifs_ips_num = 0;
304 
305 	if (ifs->ifs_fr_state_init == 1) {
306 		fr_sttab_destroy(ifs->ifs_ips_tqtqb);
307 		MUTEX_DESTROY(&ifs->ifs_ips_udptq.ifq_lock);
308 		MUTEX_DESTROY(&ifs->ifs_ips_icmptq.ifq_lock);
309 		MUTEX_DESTROY(&ifs->ifs_ips_udpacktq.ifq_lock);
310 		MUTEX_DESTROY(&ifs->ifs_ips_icmpacktq.ifq_lock);
311 		MUTEX_DESTROY(&ifs->ifs_ips_iptq.ifq_lock);
312 		MUTEX_DESTROY(&ifs->ifs_ips_deletetq.ifq_lock);
313 	}
314 
315 	if (ifs->ifs_ips_table != NULL) {
316 		KFREES(ifs->ifs_ips_table,
317 		       ifs->ifs_fr_statesize * sizeof(*ifs->ifs_ips_table));
318 		ifs->ifs_ips_table = NULL;
319 	}
320 
321 	if (ifs->ifs_ips_seed != NULL) {
322 		KFREES(ifs->ifs_ips_seed,
323 		       ifs->ifs_fr_statesize * sizeof(*ifs->ifs_ips_seed));
324 		ifs->ifs_ips_seed = NULL;
325 	}
326 
327 	if (ifs->ifs_ips_stats.iss_bucketlen != NULL) {
328 		KFREES(ifs->ifs_ips_stats.iss_bucketlen,
329 		       ifs->ifs_fr_statesize * sizeof(u_long));
330 		ifs->ifs_ips_stats.iss_bucketlen = NULL;
331 	}
332 
333 	if (ifs->ifs_fr_state_maxbucket_reset == 1)
334 		ifs->ifs_fr_state_maxbucket = 0;
335 
336 	if (ifs->ifs_fr_state_init == 1) {
337 		ifs->ifs_fr_state_init = 0;
338 		RW_DESTROY(&ifs->ifs_ipf_state);
339 		MUTEX_DESTROY(&ifs->ifs_ipf_stinsert);
340 	}
341 }
342 
343 
344 /* ------------------------------------------------------------------------ */
345 /* Function:    fr_statetstats                                              */
346 /* Returns:     ips_state_t* - pointer to state stats structure             */
347 /* Parameters:  Nil                                                         */
348 /*                                                                          */
349 /* Put all the current numbers and pointers into a single struct and return */
350 /* a pointer to it.                                                         */
351 /* ------------------------------------------------------------------------ */
352 static ips_stat_t *fr_statetstats(ifs)
353 ipf_stack_t *ifs;
354 {
355 	ifs->ifs_ips_stats.iss_active = ifs->ifs_ips_num;
356 	ifs->ifs_ips_stats.iss_statesize = ifs->ifs_fr_statesize;
357 	ifs->ifs_ips_stats.iss_statemax = ifs->ifs_fr_statemax;
358 	ifs->ifs_ips_stats.iss_table = ifs->ifs_ips_table;
359 	ifs->ifs_ips_stats.iss_list = ifs->ifs_ips_list;
360 	ifs->ifs_ips_stats.iss_ticks = ifs->ifs_fr_ticks;
361 	return &ifs->ifs_ips_stats;
362 }
363 
364 /* ------------------------------------------------------------------------ */
365 /* Function:    fr_state_remove                                             */
366 /* Returns:     int - 0 == success, != 0 == failure                         */
367 /* Parameters:  data(I) - pointer to state structure to delete from table   */
368 /*              ifs - ipf stack instance                                    */
369 /*                                                                          */
370 /* Search for a state structure that matches the one passed, according to   */
371 /* the IP addresses and other protocol specific information.                */
372 /* ------------------------------------------------------------------------ */
373 static int fr_state_remove(data, ifs)
374 caddr_t data;
375 ipf_stack_t *ifs;
376 {
377 	ipstate_t *sp, st;
378 	int error;
379 
380 	sp = &st;
381 	error = fr_inobj(data, &st, IPFOBJ_IPSTATE);
382 	if (error)
383 		return EFAULT;
384 
385 	WRITE_ENTER(&ifs->ifs_ipf_state);
386 	for (sp = ifs->ifs_ips_list; sp; sp = sp->is_next)
387 		if ((sp->is_p == st.is_p) && (sp->is_v == st.is_v) &&
388 		    !bcmp((caddr_t)&sp->is_src, (caddr_t)&st.is_src,
389 			  sizeof(st.is_src)) &&
390 		    !bcmp((caddr_t)&sp->is_dst, (caddr_t)&st.is_dst,
391 			  sizeof(st.is_dst)) &&
392 		    !bcmp((caddr_t)&sp->is_ps, (caddr_t)&st.is_ps,
393 			  sizeof(st.is_ps))) {
394 			(void) fr_delstate(sp, ISL_REMOVE, ifs);
395 			RWLOCK_EXIT(&ifs->ifs_ipf_state);
396 			return 0;
397 		}
398 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
399 	return ESRCH;
400 }
401 
402 
403 /* ------------------------------------------------------------------------ */
404 /* Function:    fr_state_ioctl                                              */
405 /* Returns:     int - 0 == success, != 0 == failure                         */
406 /* Parameters:  data(I) - pointer to ioctl data                             */
407 /*              cmd(I)  - ioctl command integer                             */
408 /*              mode(I) - file mode bits used with open                     */
409 /*              uid(I)  - uid of caller                                     */
410 /*              ctx(I)  - pointer to give the uid context                   */
411 /*              ifs     - ipf stack instance                                */
412 /*                                                                          */
413 /* Processes an ioctl call made to operate on the IP Filter state device.   */
414 /* ------------------------------------------------------------------------ */
415 int fr_state_ioctl(data, cmd, mode, uid, ctx, ifs)
416 caddr_t data;
417 ioctlcmd_t cmd;
418 int mode, uid;
419 void *ctx;
420 ipf_stack_t *ifs;
421 {
422 	int arg, ret, error = 0;
423 
424 	switch (cmd)
425 	{
426 	/*
427 	 * Delete an entry from the state table.
428 	 */
429 	case SIOCDELST :
430 		error = fr_state_remove(data, ifs);
431 		break;
432 	/*
433 	 * Flush the state table
434 	 */
435 	case SIOCIPFFL :
436 		error = BCOPYIN(data, (char *)&arg, sizeof(arg));
437 		if (error != 0) {
438 			error = EFAULT;
439 		} else {
440 			if (VALID_TABLE_FLUSH_OPT(arg)) {
441 				WRITE_ENTER(&ifs->ifs_ipf_state);
442 				ret = fr_state_flush(arg, 4, ifs);
443 				RWLOCK_EXIT(&ifs->ifs_ipf_state);
444 				error = BCOPYOUT((char *)&ret, data,
445 						sizeof(ret));
446 				if (error != 0)
447 					return EFAULT;
448 			} else {
449 				error = EINVAL;
450 			}
451 		}
452 		break;
453 
454 #ifdef	USE_INET6
455 	case SIOCIPFL6 :
456 		error = BCOPYIN(data, (char *)&arg, sizeof(arg));
457 		if (error != 0) {
458 			error = EFAULT;
459 		} else {
460 			if (VALID_TABLE_FLUSH_OPT(arg)) {
461 				WRITE_ENTER(&ifs->ifs_ipf_state);
462 				ret = fr_state_flush(arg, 6, ifs);
463 				RWLOCK_EXIT(&ifs->ifs_ipf_state);
464 				error = BCOPYOUT((char *)&ret, data,
465 						sizeof(ret));
466 				if (error != 0)
467 					return EFAULT;
468 			} else {
469 				error = EINVAL;
470 			}
471 		}
472 		break;
473 #endif
474 #ifdef	IPFILTER_LOG
475 	/*
476 	 * Flush the state log.
477 	 */
478 	case SIOCIPFFB :
479 		if (!(mode & FWRITE))
480 			error = EPERM;
481 		else {
482 			int tmp;
483 
484 			tmp = ipflog_clear(IPL_LOGSTATE, ifs);
485 			error = BCOPYOUT((char *)&tmp, data, sizeof(tmp));
486 			if (error != 0)
487 				error = EFAULT;
488 		}
489 		break;
490 	/*
491 	 * Turn logging of state information on/off.
492 	 */
493 	case SIOCSETLG :
494 		if (!(mode & FWRITE)) {
495 			error = EPERM;
496 		} else {
497 			error = BCOPYIN((char *)data,
498 					(char *)&ifs->ifs_ipstate_logging,
499 					sizeof(ifs->ifs_ipstate_logging));
500 			if (error != 0)
501 				error = EFAULT;
502 		}
503 		break;
504 	/*
505 	 * Return the current state of logging.
506 	 */
507 	case SIOCGETLG :
508 		error = BCOPYOUT((char *)&ifs->ifs_ipstate_logging,
509 				(char *)data,
510 				sizeof(ifs->ifs_ipstate_logging));
511 		if (error != 0)
512 			error = EFAULT;
513 		break;
514 	/*
515 	 * Return the number of bytes currently waiting to be read.
516 	 */
517 	case FIONREAD :
518 		arg = ifs->ifs_iplused[IPL_LOGSTATE]; /* returned in an int */
519 		error = BCOPYOUT((char *)&arg, data, sizeof(arg));
520 		if (error != 0)
521 			error = EFAULT;
522 		break;
523 #endif
524 	/*
525 	 * Get the current state statistics.
526 	 */
527 	case SIOCGETFS :
528 		error = fr_outobj(data, fr_statetstats(ifs), IPFOBJ_STATESTAT);
529 		break;
530 	/*
531 	 * Lock/Unlock the state table.  (Locking prevents any changes, which
532 	 * means no packets match).
533 	 */
534 	case SIOCSTLCK :
535 		if (!(mode & FWRITE)) {
536 			error = EPERM;
537 		} else {
538 			error = fr_lock(data, &ifs->ifs_fr_state_lock);
539 		}
540 		break;
541 	/*
542 	 * Add an entry to the current state table.
543 	 */
544 	case SIOCSTPUT :
545 		if (!ifs->ifs_fr_state_lock || !(mode & FWRITE)) {
546 			error = EACCES;
547 			break;
548 		}
549 		error = fr_stputent(data, ifs);
550 		break;
551 	/*
552 	 * Get a state table entry.
553 	 */
554 	case SIOCSTGET :
555 		if (!ifs->ifs_fr_state_lock) {
556 			error = EACCES;
557 			break;
558 		}
559 		error = fr_stgetent(data, ifs);
560 		break;
561 
562 	case SIOCGENITER :
563 	    {
564 		ipftoken_t *token;
565 		ipfgeniter_t iter;
566 
567 		error = fr_inobj(data, &iter, IPFOBJ_GENITER);
568 		if (error != 0)
569 			break;
570 
571 		token = ipf_findtoken(IPFGENITER_STATE, uid, ctx, ifs);
572 		if (token != NULL)
573 			error = fr_stateiter(token, &iter, ifs);
574 		else
575 			error = ESRCH;
576 		RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
577 		break;
578 	    }
579 
580 	case SIOCIPFDELTOK :
581 		error = BCOPYIN(data, (char *)&arg, sizeof(arg));
582 		if (error != 0) {
583 			error = EFAULT;
584 		} else {
585 			error = ipf_deltoken(arg, uid, ctx, ifs);
586 		}
587 		break;
588 
589 	default :
590 		error = EINVAL;
591 		break;
592 	}
593 	return error;
594 }
595 
596 
597 /* ------------------------------------------------------------------------ */
598 /* Function:    fr_stgetent                                                 */
599 /* Returns:     int - 0 == success, != 0 == failure                         */
600 /* Parameters:  data(I) - pointer to state structure to retrieve from table */
601 /*                                                                          */
602 /* Copy out state information from the kernel to a user space process.  If  */
603 /* there is a filter rule associated with the state entry, copy that out    */
604 /* as well.  The entry to copy out is taken from the value of "ips_next" in */
605 /* the struct passed in and if not null and not found in the list of current*/
606 /* state entries, the retrieval fails.                                      */
607 /* ------------------------------------------------------------------------ */
608 int fr_stgetent(data, ifs)
609 caddr_t data;
610 ipf_stack_t *ifs;
611 {
612 	ipstate_t *is, *isn;
613 	ipstate_save_t ips;
614 	int error;
615 
616 	error = fr_inobj(data, &ips, IPFOBJ_STATESAVE);
617 	if (error)
618 		return EFAULT;
619 
620 	isn = ips.ips_next;
621 	if (isn == NULL) {
622 		isn = ifs->ifs_ips_list;
623 		if (isn == NULL) {
624 			if (ips.ips_next == NULL)
625 				return ENOENT;
626 			return 0;
627 		}
628 	} else {
629 		/*
630 		 * Make sure the pointer we're copying from exists in the
631 		 * current list of entries.  Security precaution to prevent
632 		 * copying of random kernel data.
633 		 */
634 		for (is = ifs->ifs_ips_list; is; is = is->is_next)
635 			if (is == isn)
636 				break;
637 		if (!is)
638 			return ESRCH;
639 	}
640 	ips.ips_next = isn->is_next;
641 	bcopy((char *)isn, (char *)&ips.ips_is, sizeof(ips.ips_is));
642 	ips.ips_rule = isn->is_rule;
643 	if (isn->is_rule != NULL)
644 		bcopy((char *)isn->is_rule, (char *)&ips.ips_fr,
645 		      sizeof(ips.ips_fr));
646 	error = fr_outobj(data, &ips, IPFOBJ_STATESAVE);
647 	if (error)
648 		return EFAULT;
649 	return 0;
650 }
651 
652 
653 /* ------------------------------------------------------------------------ */
654 /* Function:    fr_stputent                                                 */
655 /* Returns:     int - 0 == success, != 0 == failure                         */
656 /* Parameters:  data(I) - pointer to state information struct               */
657 /*              ifs     - ipf stack instance                                */
658 /*                                                                          */
659 /* This function implements the SIOCSTPUT ioctl: insert a state entry into  */
660 /* the state table.  If the state info. includes a pointer to a filter rule */
661 /* then also add in an orphaned rule (will not show up in any "ipfstat -io" */
662 /* output.                                                                  */
663 /* ------------------------------------------------------------------------ */
664 int fr_stputent(data, ifs)
665 caddr_t data;
666 ipf_stack_t *ifs;
667 {
668 	ipstate_t *is, *isn;
669 	ipstate_save_t ips;
670 	int error, i;
671 	frentry_t *fr;
672 	char *name;
673 
674 	error = fr_inobj(data, &ips, IPFOBJ_STATESAVE);
675 	if (error)
676 		return EFAULT;
677 
678 	/*
679 	 * Trigger automatic call to fr_state_flush() if the
680 	 * table has reached capacity specified by hi watermark.
681 	 */
682 	if (ST_TAB_WATER_LEVEL(ifs) > ifs->ifs_state_flush_level_hi)
683 		ifs->ifs_fr_state_doflush = 1;
684 
685 	/*
686 	 * If automatic flushing did not do its job, and the table
687 	 * has filled up, don't try to create a new entry.
688 	 */
689 	if (ifs->ifs_ips_num >= ifs->ifs_fr_statemax) {
690 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_max);
691 		return ENOMEM;
692 	}
693 
694 	KMALLOC(isn, ipstate_t *);
695 	if (isn == NULL)
696 		return ENOMEM;
697 
698 	bcopy((char *)&ips.ips_is, (char *)isn, sizeof(*isn));
699 	bzero((char *)isn, offsetof(struct ipstate, is_pkts));
700 	isn->is_sti.tqe_pnext = NULL;
701 	isn->is_sti.tqe_next = NULL;
702 	isn->is_sti.tqe_ifq = NULL;
703 	isn->is_sti.tqe_parent = isn;
704 	isn->is_ifp[0] = NULL;
705 	isn->is_ifp[1] = NULL;
706 	isn->is_ifp[2] = NULL;
707 	isn->is_ifp[3] = NULL;
708 	isn->is_sync = NULL;
709 	fr = ips.ips_rule;
710 
711 	if (fr == NULL) {
712 		READ_ENTER(&ifs->ifs_ipf_state);
713 		fr_stinsert(isn, 0, ifs);
714 		MUTEX_EXIT(&isn->is_lock);
715 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
716 		return 0;
717 	}
718 
719 	if (isn->is_flags & SI_NEWFR) {
720 		KMALLOC(fr, frentry_t *);
721 		if (fr == NULL) {
722 			KFREE(isn);
723 			return ENOMEM;
724 		}
725 		bcopy((char *)&ips.ips_fr, (char *)fr, sizeof(*fr));
726 		isn->is_rule = fr;
727 		ips.ips_is.is_rule = fr;
728 		MUTEX_NUKE(&fr->fr_lock);
729 		MUTEX_INIT(&fr->fr_lock, "state filter rule lock");
730 
731 		/*
732 		 * Look up all the interface names in the rule.
733 		 */
734 		for (i = 0; i < 4; i++) {
735 			name = fr->fr_ifnames[i];
736 			fr->fr_ifas[i] = fr_resolvenic(name, fr->fr_v, ifs);
737 			name = isn->is_ifname[i];
738 			isn->is_ifp[i] = fr_resolvenic(name, isn->is_v, ifs);
739 		}
740 
741 		fr->fr_ref = 0;
742 		fr->fr_dsize = 0;
743 		fr->fr_data = NULL;
744 		fr->fr_type = FR_T_NONE;
745 
746 		fr_resolvedest(&fr->fr_tif, fr->fr_v, ifs);
747 		fr_resolvedest(&fr->fr_dif, fr->fr_v, ifs);
748 		fr_resolvedest(&fr->fr_rif, fr->fr_v, ifs);
749 
750 		/*
751 		 * send a copy back to userland of what we ended up
752 		 * to allow for verification.
753 		 */
754 		error = fr_outobj(data, &ips, IPFOBJ_STATESAVE);
755 		if (error) {
756 			KFREE(isn);
757 			MUTEX_DESTROY(&fr->fr_lock);
758 			KFREE(fr);
759 			return EFAULT;
760 		}
761 		READ_ENTER(&ifs->ifs_ipf_state);
762 		fr_stinsert(isn, 0, ifs);
763 		MUTEX_EXIT(&isn->is_lock);
764 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
765 
766 	} else {
767 		READ_ENTER(&ifs->ifs_ipf_state);
768 		for (is = ifs->ifs_ips_list; is; is = is->is_next)
769 			if (is->is_rule == fr) {
770 				fr_stinsert(isn, 0, ifs);
771 				MUTEX_EXIT(&isn->is_lock);
772 				break;
773 			}
774 
775 		if (is == NULL) {
776 			KFREE(isn);
777 			isn = NULL;
778 		}
779 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
780 
781 		return (isn == NULL) ? ESRCH : 0;
782 	}
783 
784 	return 0;
785 }
786 
787 
788 /* ------------------------------------------------------------------------ */
789 /* Function:   fr_stinsert                                                  */
790 /* Returns:    Nil                                                          */
791 /* Parameters: is(I)  - pointer to state structure                          */
792 /*             rev(I) - flag indicating forward/reverse direction of packet */
793 /*                                                                          */
794 /* Inserts a state structure into the hash table (for lookups) and the list */
795 /* of state entries (for enumeration).  Resolves all of the interface names */
796 /* to pointers and adjusts running stats for the hash table as appropriate. */
797 /*                                                                          */
798 /* Locking: it is assumed that some kind of lock on ipf_state is held.      */
799 /*          Exits with is_lock initialised and held.                        */
800 /* ------------------------------------------------------------------------ */
801 void fr_stinsert(is, rev, ifs)
802 ipstate_t *is;
803 int rev;
804 ipf_stack_t *ifs;
805 {
806 	frentry_t *fr;
807 	u_int hv;
808 	int i;
809 
810 	MUTEX_INIT(&is->is_lock, "ipf state entry");
811 
812 	fr = is->is_rule;
813 	if (fr != NULL) {
814 		MUTEX_ENTER(&fr->fr_lock);
815 		fr->fr_ref++;
816 		fr->fr_statecnt++;
817 		MUTEX_EXIT(&fr->fr_lock);
818 	}
819 
820 	/*
821 	 * Look up all the interface names in the state entry.
822 	 */
823 	for (i = 0; i < 4; i++) {
824 		if (is->is_ifp[i] != NULL)
825 			continue;
826 		is->is_ifp[i] = fr_resolvenic(is->is_ifname[i], is->is_v, ifs);
827 	}
828 
829 	/*
830 	 * If we could trust is_hv, then the modulous would not be needed, but
831 	 * when running with IPFILTER_SYNC, this stops bad values.
832 	 */
833 	hv = is->is_hv % ifs->ifs_fr_statesize;
834 	is->is_hv = hv;
835 
836 	/*
837 	 * We need to get both of these locks...the first because it is
838 	 * possible that once the insert is complete another packet might
839 	 * come along, match the entry and want to update it.
840 	 */
841 	MUTEX_ENTER(&is->is_lock);
842 	MUTEX_ENTER(&ifs->ifs_ipf_stinsert);
843 
844 	/*
845 	 * add into list table.
846 	 */
847 	if (ifs->ifs_ips_list != NULL)
848 		ifs->ifs_ips_list->is_pnext = &is->is_next;
849 	is->is_pnext = &ifs->ifs_ips_list;
850 	is->is_next = ifs->ifs_ips_list;
851 	ifs->ifs_ips_list = is;
852 
853 	if (ifs->ifs_ips_table[hv] != NULL)
854 		ifs->ifs_ips_table[hv]->is_phnext = &is->is_hnext;
855 	else
856 		ifs->ifs_ips_stats.iss_inuse++;
857 	is->is_phnext = ifs->ifs_ips_table + hv;
858 	is->is_hnext = ifs->ifs_ips_table[hv];
859 	ifs->ifs_ips_table[hv] = is;
860 	ifs->ifs_ips_stats.iss_bucketlen[hv]++;
861 	ifs->ifs_ips_num++;
862 	MUTEX_EXIT(&ifs->ifs_ipf_stinsert);
863 
864 	fr_setstatequeue(is, rev, ifs);
865 }
866 
867 /* ------------------------------------------------------------------------ */
868 /* Function:	fr_match_ipv4addrs					    */
869 /* Returns:	int -	2 strong match (same addresses, same direction)	    */
870 /*			1 weak match (same address, opposite direction)	    */
871 /*			0 no match					    */
872 /*									    */
873 /* Function matches IPv4 addresses.					    */
874 /* ------------------------------------------------------------------------ */
875 static int fr_match_ipv4addrs(is1, is2)
876 ipstate_t *is1;
877 ipstate_t *is2;
878 {
879 	int	rv;
880 
881 	if (is1->is_saddr == is2->is_saddr && is1->is_daddr == is2->is_daddr)
882 		rv = 2;
883 	else if (is1->is_saddr == is2->is_daddr &&
884 	    is1->is_daddr == is2->is_saddr)
885 		rv = 1;
886 	else
887 		rv = 0;
888 
889 	return (rv);
890 }
891 
892 /* ------------------------------------------------------------------------ */
893 /* Function:	fr_match_ipv6addrs					    */
894 /* Returns:	int - 	2 strong match (same addresses, same direction)	    */
895 /*			1 weak match (same addresses, opposite direction)   */
896 /*			0 no match					    */
897 /*									    */
898 /* Function matches IPv6 addresses.					    */
899 /* ------------------------------------------------------------------------ */
900 static int fr_match_ipv6addrs(is1, is2)
901 ipstate_t *is1;
902 ipstate_t *is2;
903 {
904 	int	rv;
905 
906 	if (IP6_EQ(&is1->is_src, &is2->is_src) &&
907 	    IP6_EQ(&is1->is_dst, &is2->is_dst))
908 		rv = 2;
909 	else if (IP6_EQ(&is1->is_src, &is2->is_dst) &&
910 	    IP6_EQ(&is1->is_dst, &is2->is_src)) {
911 		rv = 1;
912 	}
913 	else
914 		rv = 0;
915 
916 	return (rv);
917 }
918 /* ------------------------------------------------------------------------ */
919 /* Function:	fr_match_addresses					    */
920 /* Returns:	int - 	2 strong match (same addresses, same direction)	    */
921 /*			1 weak match (same address, opposite directions)    */
922 /* 			0 no match					    */
923 /* Parameters:	is1, is2 pointers to states we are checking		    */
924 /*									    */
925 /* Matches addresses, function uses fr_match_ipvXaddrs() to deal with IPv4  */
926 /* and IPv6 address format.						    */
927 /* ------------------------------------------------------------------------ */
928 static int fr_match_addresses(is1, is2)
929 ipstate_t *is1;
930 ipstate_t *is2;
931 {
932 	int	rv;
933 
934 	if (is1->is_v == 4) {
935 		rv = fr_match_ipv4addrs(is1, is2);
936 	} else {
937 		rv = fr_match_ipv6addrs(is1, is2);
938 	}
939 
940 	return (rv);
941 }
942 
943 /* ------------------------------------------------------------------------ */
944 /* Function:	fr_match_ppairs						    */
945 /* Returns:	int - 	2 strong match (same ports, same direction)	    */
946 /*			1 weak match (same ports, different direction)	    */
947 /*			0 no match					    */
948 /* Parameters	ppairs1, ppairs - src, dst ports we want to match.	    */
949 /*									    */
950 /* Matches two port_pair_t types (port pairs). Each port pair contains	    */
951 /* src, dst port, which belong to session (state entry).		    */
952 /* ------------------------------------------------------------------------ */
953 static int fr_match_ppairs(ppairs1, ppairs2)
954 port_pair_t *ppairs1;
955 port_pair_t *ppairs2;
956 {
957 	int	rv;
958 
959 	if (ppairs1->pp_sport == ppairs2->pp_sport &&
960 	    ppairs1->pp_dport == ppairs2->pp_dport)
961 		rv = 2;
962 	else if (ppairs1->pp_sport == ppairs2->pp_dport &&
963 		    ppairs1->pp_dport == ppairs2->pp_sport)
964 		rv = 1;
965 	else
966 		rv = 0;
967 
968 	return (rv);
969 }
970 
971 /* ------------------------------------------------------------------------ */
972 /* Function:	fr_match_l4_hdr						    */
973 /* Returns:	int -	0 no match,					    */
974 /*			1 weak match (same ports, different directions)	    */
975 /*			2 strong match (same ports, same direction)	    */
976 /* Parameters	is1, is2 - states we want to match			    */
977 /*									    */
978 /* Function matches L4 header data (source ports for TCP, UDP, CallIds for  */
979 /* GRE protocol).							    */
980 /* ------------------------------------------------------------------------ */
981 static int fr_match_l4_hdr(is1, is2)
982 ipstate_t *is1;
983 ipstate_t *is2;
984 {
985 	int	rv = 0;
986 	port_pair_t	pp1;
987 	port_pair_t	pp2;
988 
989 	if (is1->is_p != is2->is_p)
990 		return (0);
991 
992 	switch (is1->is_p) {
993 		case	IPPROTO_TCP:
994 			pp1.pp_sport = is1->is_ps.is_ts.ts_sport;
995 			pp1.pp_dport = is1->is_ps.is_ts.ts_dport;
996 			pp2.pp_sport = is2->is_ps.is_ts.ts_sport;
997 			pp2.pp_dport = is2->is_ps.is_ts.ts_dport;
998 			rv = fr_match_ppairs(&pp1, &pp2);
999 			break;
1000 		case	IPPROTO_UDP:
1001 			pp1.pp_sport = is1->is_ps.is_us.us_sport;
1002 			pp1.pp_dport = is1->is_ps.is_us.us_dport;
1003 			pp2.pp_sport = is2->is_ps.is_us.us_sport;
1004 			pp2.pp_dport = is2->is_ps.is_us.us_dport;
1005 			rv = fr_match_ppairs(&pp1, &pp2);
1006 			break;
1007 		case	IPPROTO_GRE:
1008 			/* greinfo_t can be also interprted as port pair */
1009 			pp1.pp_sport = is1->is_ps.is_ug.gs_call[0];
1010 			pp1.pp_dport = is1->is_ps.is_ug.gs_call[1];
1011 			pp2.pp_sport = is2->is_ps.is_ug.gs_call[0];
1012 			pp2.pp_dport = is2->is_ps.is_ug.gs_call[1];
1013 			rv = fr_match_ppairs(&pp1, &pp2);
1014 			break;
1015 		case	IPPROTO_ICMP:
1016 		case	IPPROTO_ICMPV6:
1017 			if (bcmp(&is1->is_ps, &is2->is_ps, sizeof (icmpinfo_t)))
1018 				rv = 1;
1019 			else
1020 				rv = 0;
1021 			break;
1022 		default:
1023 			rv = 0;
1024 	}
1025 
1026 	return (rv);
1027 }
1028 
1029 /* ------------------------------------------------------------------------ */
1030 /* Function:	fr_matchstates						    */
1031 /* Returns:	int - nonzero match, zero no match			    */
1032 /* Parameters	is1, is2 - states we want to match			    */
1033 /*									    */
1034 /* The state entries are equal (identical match) if they belong to the same */
1035 /* session. Any time new state entry is being added the fr_addstate()	    */
1036 /* function creates temporal state entry from the data it gets from IP and  */
1037 /* L4 header. The fr_matchstats() must be also aware of packet direction,   */
1038 /* which is also stored within the state entry. We should keep in mind the  */
1039 /* information about packet direction is spread accross L3 (addresses) and  */
1040 /* L4 (ports). There are three possible relationships betwee is1, is2:	    */
1041 /* 		- no match (match(is1, is2) == 0))			    */
1042 /*		- weak match same addresses (ports), but different	    */
1043 /*			directions (1)	(fr_match_xxxx(is1, is2) == 1)	    */
1044 /*		- strong match same addresses (ports) and same directions   */
1045 /*			 (2) (fr_match_xxxx(is1, is2) == 2)		    */
1046 /*									    */
1047 /* There are functions, which match match addresses (L3 header) in is1, is2 */
1048 /* and functions, which are used to compare ports (L4 header) data. We say  */
1049 /* the is1 and is2 are same (identical) if there is a match		    */
1050 /* (fr_match_l4_hdr(is1, is2) != 0) and matchlevels are same for entries    */
1051 /* (fr_match_l3_hdr(is1, is2) == fr_match_l4_hdr(is1, is2)) for is1, is2.   */
1052 /* Such requirement deals with case as follows:				    */
1053 /*	suppose there are two connections between hosts A, B. Connection 1: */
1054 /*			a.a.a.a:12345 <=> b.b.b.b:54321			    */
1055 /*		Connection 2:						    */
1056 /*			a.a.a.a:54321 <=> b.b.b.b:12345			    */
1057 /* since we've introduced match levels into our fr_matchstates(), we are    */
1058 /* able to identify, which packets belong to connection A and which belong  */
1059 /* to connection B.	Assume there are two entries is1, is2. is1 has been */
1060 /* from con. 1 packet, which travelled from A to B:			    */
1061 /*			a.a.a.a:12345 -> b.b.b.b:54321			    */
1062 /* while s2, has been created from packet which belongs to con. 2 and is    */
1063 /* also coming from A to B:						    */
1064 /*			a.a.a.a:54321 -> b.b.b.b:12345			    */
1065 /* fr_match_l3_hdr(is1, is2) == 2 -> strong match, while		    */
1066 /* fr_match_l4_hdr(is1, is2) == 1 -> weak match. Since match levels are	    */
1067 /* different the state entries are not identical -> no match as a final	    */
1068 /* result.								    */
1069 /* ------------------------------------------------------------------------ */
1070 static int fr_matchstates(is1, is2)
1071 ipstate_t *is1;
1072 ipstate_t *is2;
1073 {
1074 	int	rv;
1075 	int	amatch;
1076 	int	pmatch;
1077 
1078 	if (bcmp(&is1->is_pass, &is2->is_pass,
1079 		offsetof(struct ipstate, is_ps) -
1080 		offsetof(struct ipstate, is_pass)) == 0) {
1081 
1082 		pmatch = fr_match_l4_hdr(is1, is2);
1083 		amatch = fr_match_addresses(is1, is2);
1084 		/*
1085 		 * If addresses match (amatch != 0), then 'match levels'
1086 		 * must be same for matching entries. If amatch and pmatch
1087 		 * have different values (different match levels), then
1088 		 * is1 and is2 belong to different sessions.
1089 		 */
1090 		rv = (amatch != 0) && (amatch == pmatch);
1091 	}
1092 	else
1093 		rv = 0;
1094 
1095 	return (rv);
1096 }
1097 
1098 /* ------------------------------------------------------------------------ */
1099 /* Function:    fr_addstate                                                 */
1100 /* Returns:     ipstate_t* - NULL == failure, else pointer to new state     */
1101 /* Parameters:  fin(I)    - pointer to packet information                   */
1102 /*              stsave(O) - pointer to place to save pointer to created     */
1103 /*                          state structure.                                */
1104 /*              flags(I)  - flags to use when creating the structure        */
1105 /*                                                                          */
1106 /* Creates a new IP state structure from the packet information collected.  */
1107 /* Inserts it into the state table and appends to the bottom of the active  */
1108 /* list.  If the capacity of the table has reached the maximum allowed then */
1109 /* the call will fail and a flush is scheduled for the next timeout call.   */
1110 /* ------------------------------------------------------------------------ */
1111 ipstate_t *fr_addstate(fin, stsave, flags)
1112 fr_info_t *fin;
1113 ipstate_t **stsave;
1114 u_int flags;
1115 {
1116 	ipstate_t *is, ips;
1117 	struct icmp *ic;
1118 	u_int pass, hv;
1119 	frentry_t *fr;
1120 	tcphdr_t *tcp;
1121 	grehdr_t *gre;
1122 	void *ifp;
1123 	int out;
1124 	ipf_stack_t *ifs = fin->fin_ifs;
1125 
1126 	if (ifs->ifs_fr_state_lock ||
1127 	    (fin->fin_flx & (FI_SHORT|FI_STATE|FI_FRAGBODY|FI_BAD)))
1128 		return NULL;
1129 
1130 	if ((fin->fin_flx & FI_OOW) && !(fin->fin_tcpf & TH_SYN))
1131 		return NULL;
1132 
1133 	/*
1134 	 * Trigger automatic call to fr_state_flush() if the
1135 	 * table has reached capacity specified by hi watermark.
1136 	 */
1137 	if (ST_TAB_WATER_LEVEL(ifs) > ifs->ifs_state_flush_level_hi)
1138 		ifs->ifs_fr_state_doflush = 1;
1139 
1140 	/*
1141 	 * If the max number of state entries has been reached, and there is no
1142 	 * limit on the state count for the rule, then do not continue.  In the
1143 	 * case where a limit exists, it's ok allow the entries to be created as
1144 	 * long as specified limit itself has not been reached.
1145 	 *
1146 	 * Note that because the lock isn't held on fr, it is possible to exceed
1147 	 * the specified size of the table.  However, the cost of this is being
1148 	 * ignored here; as the number by which it can go over is a product of
1149 	 * the number of simultaneous threads that could be executing in here.
1150 	 * So, a limit of 100 won't result in 200, but could result in 101 or 102.
1151 	 *
1152 	 * Also note that, since the automatic flush should have been triggered
1153 	 * well before we reach the maximum number of state table entries, the
1154 	 * likelihood of reaching the max (and thus exceedng it) is minimal.
1155 	 */
1156 	fr = fin->fin_fr;
1157 	if (fr != NULL) {
1158 		if ((ifs->ifs_ips_num >= ifs->ifs_fr_statemax) &&
1159 		    (fr->fr_statemax == 0)) {
1160 			ATOMIC_INCL(ifs->ifs_ips_stats.iss_max);
1161 			return NULL;
1162 		}
1163 		if ((fr->fr_statemax != 0) &&
1164 		    (fr->fr_statecnt >= fr->fr_statemax)) {
1165 			ATOMIC_INCL(ifs->ifs_ips_stats.iss_maxref);
1166 			ifs->ifs_fr_state_doflush = 1;
1167 			return NULL;
1168 		}
1169 	}
1170 
1171 	ic = NULL;
1172 	tcp = NULL;
1173 	out = fin->fin_out;
1174 	is = &ips;
1175 	bzero((char *)is, sizeof(*is));
1176 
1177 	if (fr == NULL) {
1178 		pass = ifs->ifs_fr_flags;
1179 		is->is_tag = FR_NOLOGTAG;
1180 	} else {
1181 		pass = fr->fr_flags;
1182 	}
1183 
1184 	is->is_die = 1 + ifs->ifs_fr_ticks;
1185 	/*
1186 	 * We want to check everything that is a property of this packet,
1187 	 * but we don't (automatically) care about it's fragment status as
1188 	 * this may change.
1189 	 */
1190 	is->is_pass = pass;
1191 	is->is_v = fin->fin_v;
1192 	is->is_opt[0] = fin->fin_optmsk;
1193 	is->is_optmsk[0] = 0xffffffff;
1194 	/*
1195 	 * The reverse direction option mask will be set in fr_matchsrcdst(),
1196 	 * when we will see the first packet from the peer. We will leave it
1197 	 * as zero for now.
1198 	 */
1199 	is->is_optmsk[1] = 0x0;
1200 
1201 	if (is->is_v == 6) {
1202 		is->is_opt[0] &= ~0x8;
1203 		is->is_optmsk[0] &= ~0x8;
1204 	}
1205 	is->is_sec = fin->fin_secmsk;
1206 	is->is_secmsk = 0xffff;
1207 	is->is_auth = fin->fin_auth;
1208 	is->is_authmsk = 0xffff;
1209 
1210 	/*
1211 	 * Copy and calculate...
1212 	 */
1213 	hv = (is->is_p = fin->fin_fi.fi_p);
1214 	is->is_src = fin->fin_fi.fi_src;
1215 	hv += is->is_saddr;
1216 	is->is_dst = fin->fin_fi.fi_dst;
1217 	hv += is->is_daddr;
1218 #ifdef	USE_INET6
1219 	if (fin->fin_v == 6) {
1220 		/*
1221 		 * For ICMPv6, we check to see if the destination address is
1222 		 * a multicast address.  If it is, do not include it in the
1223 		 * calculation of the hash because the correct reply will come
1224 		 * back from a real address, not a multicast address.
1225 		 */
1226 		if ((is->is_p == IPPROTO_ICMPV6) &&
1227 		    IN6_IS_ADDR_MULTICAST(&is->is_dst.in6)) {
1228 			/*
1229 			 * So you can do keep state with neighbour discovery.
1230 			 *
1231 			 * Here we could use the address from the neighbour
1232 			 * solicit message to put in the state structure and
1233 			 * we could use that without a wildcard flag too...
1234 			 */
1235 			is->is_flags |= SI_W_DADDR;
1236 			hv -= is->is_daddr;
1237 		} else {
1238 			hv += is->is_dst.i6[1];
1239 			hv += is->is_dst.i6[2];
1240 			hv += is->is_dst.i6[3];
1241 		}
1242 		hv += is->is_src.i6[1];
1243 		hv += is->is_src.i6[2];
1244 		hv += is->is_src.i6[3];
1245 	}
1246 #endif
1247 	if ((fin->fin_v == 4) &&
1248 	    (fin->fin_flx & (FI_MULTICAST|FI_BROADCAST|FI_MBCAST))) {
1249 		if (fin->fin_out == 0) {
1250 			flags |= SI_W_DADDR|SI_CLONE;
1251 			hv -= is->is_daddr;
1252 		} else {
1253 			flags |= SI_W_SADDR|SI_CLONE;
1254 			hv -= is->is_saddr;
1255 		}
1256 	}
1257 
1258 	switch (is->is_p)
1259 	{
1260 #ifdef	USE_INET6
1261 	case IPPROTO_ICMPV6 :
1262 		ic = fin->fin_dp;
1263 
1264 		switch (ic->icmp_type)
1265 		{
1266 		case ICMP6_ECHO_REQUEST :
1267 			is->is_icmp.ici_type = ic->icmp_type;
1268 			hv += (is->is_icmp.ici_id = ic->icmp_id);
1269 			break;
1270 		case ICMP6_MEMBERSHIP_QUERY :
1271 		case ND_ROUTER_SOLICIT :
1272 		case ND_NEIGHBOR_SOLICIT :
1273 		case ICMP6_NI_QUERY :
1274 			is->is_icmp.ici_type = ic->icmp_type;
1275 			break;
1276 		default :
1277 			return NULL;
1278 		}
1279 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_icmp);
1280 		break;
1281 #endif
1282 	case IPPROTO_ICMP :
1283 		ic = fin->fin_dp;
1284 
1285 		switch (ic->icmp_type)
1286 		{
1287 		case ICMP_ECHO :
1288 		case ICMP_TSTAMP :
1289 		case ICMP_IREQ :
1290 		case ICMP_MASKREQ :
1291 			is->is_icmp.ici_type = ic->icmp_type;
1292 			hv += (is->is_icmp.ici_id = ic->icmp_id);
1293 			break;
1294 		default :
1295 			return NULL;
1296 		}
1297 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_icmp);
1298 		break;
1299 
1300 	case IPPROTO_GRE :
1301 		gre = fin->fin_dp;
1302 
1303 		is->is_gre.gs_flags = gre->gr_flags;
1304 		is->is_gre.gs_ptype = gre->gr_ptype;
1305 		if (GRE_REV(is->is_gre.gs_flags) == 1) {
1306 			is->is_call[0] = fin->fin_data[0];
1307 			is->is_call[1] = fin->fin_data[1];
1308 		}
1309 		break;
1310 
1311 	case IPPROTO_TCP :
1312 		tcp = fin->fin_dp;
1313 
1314 		if (tcp->th_flags & TH_RST)
1315 			return NULL;
1316 		/*
1317 		 * The endian of the ports doesn't matter, but the ack and
1318 		 * sequence numbers do as we do mathematics on them later.
1319 		 */
1320 		is->is_sport = htons(fin->fin_data[0]);
1321 		is->is_dport = htons(fin->fin_data[1]);
1322 		if ((flags & (SI_W_DPORT|SI_W_SPORT)) == 0) {
1323 			hv += is->is_sport;
1324 			hv += is->is_dport;
1325 		}
1326 
1327 		/*
1328 		 * If this is a real packet then initialise fields in the
1329 		 * state information structure from the TCP header information.
1330 		 */
1331 
1332 		is->is_maxdwin = 1;
1333 		is->is_maxswin = ntohs(tcp->th_win);
1334 		if (is->is_maxswin == 0)
1335 			is->is_maxswin = 1;
1336 
1337 		if ((fin->fin_flx & FI_IGNORE) == 0) {
1338 			is->is_send = ntohl(tcp->th_seq) + fin->fin_dlen -
1339 				      (TCP_OFF(tcp) << 2) +
1340 				      ((tcp->th_flags & TH_SYN) ? 1 : 0) +
1341 				      ((tcp->th_flags & TH_FIN) ? 1 : 0);
1342 			is->is_maxsend = is->is_send;
1343 
1344 			/*
1345 			 * Window scale option is only present in
1346 			 * SYN/SYN-ACK packet.
1347 			 */
1348 			if ((tcp->th_flags & ~(TH_FIN|TH_ACK|TH_ECNALL)) ==
1349 			    TH_SYN &&
1350 			    (TCP_OFF(tcp) > (sizeof(tcphdr_t) >> 2))) {
1351 				if (fr_tcpoptions(fin, tcp,
1352 					&is->is_tcp.ts_data[0]) == -1) {
1353 					fin->fin_flx |= FI_BAD;
1354 				}
1355 			}
1356 
1357 			if ((fin->fin_out != 0) && (pass & FR_NEWISN) != 0) {
1358 				fr_checknewisn(fin, is);
1359 				fr_fixoutisn(fin, is);
1360 			}
1361 
1362 			if ((tcp->th_flags & TH_OPENING) == TH_SYN)
1363 				flags |= IS_TCPFSM;
1364 			else {
1365 				is->is_maxdwin = is->is_maxswin * 2;
1366 				is->is_dend = ntohl(tcp->th_ack);
1367 				is->is_maxdend = ntohl(tcp->th_ack);
1368 				is->is_maxdwin *= 2;
1369 			}
1370 		}
1371 
1372 		/*
1373 		 * If we're creating state for a starting connection, start the
1374 		 * timer on it as we'll never see an error if it fails to
1375 		 * connect.
1376 		 */
1377 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_tcp);
1378 		break;
1379 
1380 	case IPPROTO_UDP :
1381 		tcp = fin->fin_dp;
1382 
1383 		is->is_sport = htons(fin->fin_data[0]);
1384 		is->is_dport = htons(fin->fin_data[1]);
1385 		if ((flags & (SI_W_DPORT|SI_W_SPORT)) == 0) {
1386 			hv += tcp->th_dport;
1387 			hv += tcp->th_sport;
1388 		}
1389 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_udp);
1390 		break;
1391 
1392 	default :
1393 		break;
1394 	}
1395 	hv = DOUBLE_HASH(hv, ifs);
1396 	is->is_hv = hv;
1397 	is->is_rule = fr;
1398 	is->is_flags = flags & IS_INHERITED;
1399 
1400 	/*
1401 	 * Look for identical state.
1402 	 */
1403 	for (is = ifs->ifs_ips_table[is->is_hv % ifs->ifs_fr_statesize];
1404 	     is != NULL;
1405 	     is = is->is_hnext) {
1406 		if (fr_matchstates(&ips, is) == 1)
1407 			break;
1408 	}
1409 
1410 	/*
1411 	 * we've found a matching state -> state already exists,
1412 	 * we are not going to add a duplicate record.
1413 	 */
1414 	if (is != NULL)
1415 		return NULL;
1416 
1417 	if (ifs->ifs_ips_stats.iss_bucketlen[hv] >= ifs->ifs_fr_state_maxbucket) {
1418 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_bucketfull);
1419 		return NULL;
1420 	}
1421 	KMALLOC(is, ipstate_t *);
1422 	if (is == NULL) {
1423 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_nomem);
1424 		return NULL;
1425 	}
1426 	bcopy((char *)&ips, (char *)is, sizeof(*is));
1427 	/*
1428 	 * Do not do the modulous here, it is done in fr_stinsert().
1429 	 */
1430 	if (fr != NULL) {
1431 		(void) strncpy(is->is_group, fr->fr_group, FR_GROUPLEN);
1432 		if (fr->fr_age[0] != 0) {
1433 			is->is_tqehead[0] =
1434 			    fr_addtimeoutqueue(&ifs->ifs_ips_utqe,
1435 					       fr->fr_age[0], ifs);
1436 			is->is_sti.tqe_flags |= TQE_RULEBASED;
1437 		}
1438 		if (fr->fr_age[1] != 0) {
1439 			is->is_tqehead[1] =
1440 			    fr_addtimeoutqueue(&ifs->ifs_ips_utqe,
1441 					       fr->fr_age[1], ifs);
1442 			is->is_sti.tqe_flags |= TQE_RULEBASED;
1443 		}
1444 		is->is_tag = fr->fr_logtag;
1445 
1446 		is->is_ifp[(out << 1) + 1] = fr->fr_ifas[1];
1447 		is->is_ifp[(1 - out) << 1] = fr->fr_ifas[2];
1448 		is->is_ifp[((1 - out) << 1) + 1] = fr->fr_ifas[3];
1449 
1450 		if (((ifp = fr->fr_ifas[1]) != NULL) &&
1451 		    (ifp != (void *)-1)) {
1452 			COPYIFNAME(ifp, is->is_ifname[(out << 1) + 1], fr->fr_v);
1453 		}
1454 		if (((ifp = fr->fr_ifas[2]) != NULL) &&
1455 		    (ifp != (void *)-1)) {
1456 			COPYIFNAME(ifp, is->is_ifname[(1 - out) << 1], fr->fr_v);
1457 		}
1458 		if (((ifp = fr->fr_ifas[3]) != NULL) &&
1459 		    (ifp != (void *)-1)) {
1460 			COPYIFNAME(ifp, is->is_ifname[((1 - out) << 1) + 1], fr->fr_v);
1461 		}
1462 	}
1463 
1464 	is->is_ifp[out << 1] = fin->fin_ifp;
1465 	if (fin->fin_ifp != NULL) {
1466 		COPYIFNAME(fin->fin_ifp, is->is_ifname[out << 1], fin->fin_v);
1467 	}
1468 
1469 	is->is_ref = 1;
1470 	is->is_pkts[0] = 0, is->is_bytes[0] = 0;
1471 	is->is_pkts[1] = 0, is->is_bytes[1] = 0;
1472 	is->is_pkts[2] = 0, is->is_bytes[2] = 0;
1473 	is->is_pkts[3] = 0, is->is_bytes[3] = 0;
1474 	if ((fin->fin_flx & FI_IGNORE) == 0) {
1475 		is->is_pkts[out] = 1;
1476 		is->is_bytes[out] = fin->fin_plen;
1477 		is->is_flx[out][0] = fin->fin_flx & FI_CMP;
1478 		is->is_flx[out][0] &= ~FI_OOW;
1479 	}
1480 
1481 	if (pass & FR_STSTRICT)
1482 		is->is_flags |= IS_STRICT;
1483 
1484 	if (pass & FR_STATESYNC)
1485 		is->is_flags |= IS_STATESYNC;
1486 
1487 	if (flags & (SI_WILDP|SI_WILDA)) {
1488 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_wild);
1489 	}
1490 	is->is_rulen = fin->fin_rule;
1491 
1492 
1493 	if (pass & FR_LOGFIRST)
1494 		is->is_pass &= ~(FR_LOGFIRST|FR_LOG);
1495 
1496 	READ_ENTER(&ifs->ifs_ipf_state);
1497 	is->is_me = stsave;
1498 
1499 	fr_stinsert(is, fin->fin_rev, ifs);
1500 
1501 	if (fin->fin_p == IPPROTO_TCP) {
1502 		/*
1503 		* If we're creating state for a starting connection, start the
1504 		* timer on it as we'll never see an error if it fails to
1505 		* connect.
1506 		*/
1507 		(void) fr_tcp_age(&is->is_sti, fin, ifs->ifs_ips_tqtqb,
1508 				  is->is_flags);
1509 		MUTEX_EXIT(&is->is_lock);
1510 #ifdef	IPFILTER_SCAN
1511 		if ((is->is_flags & SI_CLONE) == 0)
1512 			(void) ipsc_attachis(is);
1513 #endif
1514 	} else {
1515 		MUTEX_EXIT(&is->is_lock);
1516 	}
1517 #ifdef	IPFILTER_SYNC
1518 	if ((is->is_flags & IS_STATESYNC) && ((is->is_flags & SI_CLONE) == 0))
1519 		is->is_sync = ipfsync_new(SMC_STATE, fin, is);
1520 #endif
1521 	if (ifs->ifs_ipstate_logging)
1522 		ipstate_log(is, ISL_NEW, ifs);
1523 
1524 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
1525 	fin->fin_rev = IP6_NEQ(&is->is_dst, &fin->fin_daddr);
1526 	fin->fin_flx |= FI_STATE;
1527 	if (fin->fin_flx & FI_FRAG)
1528 		(void) fr_newfrag(fin, pass ^ FR_KEEPSTATE);
1529 
1530 	return is;
1531 }
1532 
1533 
1534 /* ------------------------------------------------------------------------ */
1535 /* Function:    fr_tcpoptions                                               */
1536 /* Returns:     int - 1 == packet matches state entry, 0 == it does not     */
1537 /* Parameters:  fin(I) - pointer to packet information                      */
1538 /*              tcp(I) - pointer to TCP packet header                       */
1539 /*              td(I)  - pointer to TCP data held as part of the state      */
1540 /*                                                                          */
1541 /* Look after the TCP header for any options and deal with those that are   */
1542 /* present.  Record details about those that we recogise.                   */
1543 /* ------------------------------------------------------------------------ */
1544 static int fr_tcpoptions(fin, tcp, td)
1545 fr_info_t *fin;
1546 tcphdr_t *tcp;
1547 tcpdata_t *td;
1548 {
1549 	int off, mlen, ol, i, len, retval;
1550 	char buf[64], *s, opt;
1551 	mb_t *m = NULL;
1552 
1553 	len = (TCP_OFF(tcp) << 2);
1554 	if (fin->fin_dlen < len)
1555 		return 0;
1556 	len -= sizeof(*tcp);
1557 
1558 	off = fin->fin_plen - fin->fin_dlen + sizeof(*tcp) + fin->fin_ipoff;
1559 
1560 	m = fin->fin_m;
1561 	mlen = MSGDSIZE(m) - off;
1562 	if (len > mlen) {
1563 		len = mlen;
1564 		retval = 0;
1565 	} else {
1566 		retval = 1;
1567 	}
1568 
1569 	COPYDATA(m, off, len, buf);
1570 
1571 	for (s = buf; len > 0; ) {
1572 		opt = *s;
1573 		if (opt == TCPOPT_EOL)
1574 			break;
1575 		else if (opt == TCPOPT_NOP)
1576 			ol = 1;
1577 		else {
1578 			if (len < 2)
1579 				break;
1580 			ol = (int)*(s + 1);
1581 			if (ol < 2 || ol > len)
1582 				break;
1583 
1584 			/*
1585 			 * Extract the TCP options we are interested in out of
1586 			 * the header and store them in the the tcpdata struct.
1587 			 */
1588 			switch (opt)
1589 			{
1590 			case TCPOPT_WINDOW :
1591 				if (ol == TCPOLEN_WINDOW) {
1592 					i = (int)*(s + 2);
1593 					if (i > TCP_WSCALE_MAX)
1594 						i = TCP_WSCALE_MAX;
1595 					else if (i < 0)
1596 						i = 0;
1597 					td->td_winscale = i;
1598 					td->td_winflags |= TCP_WSCALE_SEEN |
1599 							    TCP_WSCALE_FIRST;
1600 				} else
1601 					retval = -1;
1602 				break;
1603 			case TCPOPT_MAXSEG :
1604 				/*
1605 				 * So, if we wanted to set the TCP MAXSEG,
1606 				 * it should be done here...
1607 				 */
1608 				if (ol == TCPOLEN_MAXSEG) {
1609 					i = (int)*(s + 2);
1610 					i <<= 8;
1611 					i += (int)*(s + 3);
1612 					td->td_maxseg = i;
1613 				} else
1614 					retval = -1;
1615 				break;
1616 			case TCPOPT_SACK_PERMITTED :
1617 				if (ol == TCPOLEN_SACK_PERMITTED)
1618 					td->td_winflags |= TCP_SACK_PERMIT;
1619 				else
1620 					retval = -1;
1621 				break;
1622 			}
1623 		}
1624 		len -= ol;
1625 		s += ol;
1626 	}
1627 	return retval;
1628 }
1629 
1630 
1631 /* ------------------------------------------------------------------------ */
1632 /* Function:    fr_tcpstate                                                 */
1633 /* Returns:     int - 1 == packet matches state entry, 0 == it does not     */
1634 /* Parameters:  fin(I)   - pointer to packet information                    */
1635 /*              tcp(I)   - pointer to TCP packet header                     */
1636 /*              is(I)  - pointer to master state structure                  */
1637 /*                                                                          */
1638 /* Check to see if a packet with TCP headers fits within the TCP window.    */
1639 /* Change timeout depending on whether new packet is a SYN-ACK returning    */
1640 /* for a SYN or a RST or FIN which indicate time to close up shop.          */
1641 /* ------------------------------------------------------------------------ */
1642 static int fr_tcpstate(fin, tcp, is)
1643 fr_info_t *fin;
1644 tcphdr_t *tcp;
1645 ipstate_t *is;
1646 {
1647 	int source, ret = 0, flags;
1648 	tcpdata_t  *fdata, *tdata;
1649 	ipf_stack_t *ifs = fin->fin_ifs;
1650 
1651 	source = !fin->fin_rev;
1652 	if (((is->is_flags & IS_TCPFSM) != 0) && (source == 1) &&
1653 	    (ntohs(is->is_sport) != fin->fin_data[0]))
1654 		source = 0;
1655 	fdata = &is->is_tcp.ts_data[!source];
1656 	tdata = &is->is_tcp.ts_data[source];
1657 
1658 	MUTEX_ENTER(&is->is_lock);
1659 
1660 	/*
1661 	 * If a SYN packet is received for a connection that is in a half
1662 	 * closed state, then move its state entry to deletetq. In such case
1663 	 * the SYN packet will be consequently dropped. This allows new state
1664 	 * entry to be created with a retransmited SYN packet.
1665 	 */
1666 	if ((tcp->th_flags & TH_OPENING) == TH_SYN) {
1667 		if ((is->is_state[source] > IPF_TCPS_ESTABLISHED) &&
1668 		    (is->is_state[!source] > IPF_TCPS_ESTABLISHED)) {
1669 			is->is_state[source] = IPF_TCPS_CLOSED;
1670 			is->is_state[!source] = IPF_TCPS_CLOSED;
1671 			/*
1672 			 * Do not update is->is_sti.tqe_die in case state entry
1673 			 * is already present in deletetq. It prevents state
1674 			 * entry ttl update by retransmitted SYN packets, which
1675 			 * may arrive before timer tick kicks off. The SYN
1676 			 * packet will be dropped again.
1677 			 */
1678 			if (is->is_sti.tqe_ifq != &ifs->ifs_ips_deletetq)
1679 				fr_movequeue(&is->is_sti, is->is_sti.tqe_ifq,
1680 					&fin->fin_ifs->ifs_ips_deletetq,
1681 					fin->fin_ifs);
1682 
1683 			MUTEX_EXIT(&is->is_lock);
1684 			return 0;
1685 		}
1686 	}
1687 
1688 	if (fr_tcpinwindow(fin, fdata, tdata, tcp, is->is_flags)) {
1689 #ifdef	IPFILTER_SCAN
1690 		if (is->is_flags & (IS_SC_CLIENT|IS_SC_SERVER)) {
1691 			ipsc_packet(fin, is);
1692 			if (FR_ISBLOCK(is->is_pass)) {
1693 				MUTEX_EXIT(&is->is_lock);
1694 				return 1;
1695 			}
1696 		}
1697 #endif
1698 
1699 		/*
1700 		 * Nearing end of connection, start timeout.
1701 		 */
1702 		ret = fr_tcp_age(&is->is_sti, fin, ifs->ifs_ips_tqtqb,
1703 				 is->is_flags);
1704 		if (ret == 0) {
1705 			MUTEX_EXIT(&is->is_lock);
1706 			return 0;
1707 		}
1708 
1709 		/*
1710 		 * set s0's as appropriate.  Use syn-ack packet as it
1711 		 * contains both pieces of required information.
1712 		 */
1713 		/*
1714 		 * Window scale option is only present in SYN/SYN-ACK packet.
1715 		 * Compare with ~TH_FIN to mask out T/TCP setups.
1716 		 */
1717 		flags = tcp->th_flags & ~(TH_FIN|TH_ECNALL);
1718 		if (flags == (TH_SYN|TH_ACK)) {
1719 			is->is_s0[source] = ntohl(tcp->th_ack);
1720 			is->is_s0[!source] = ntohl(tcp->th_seq) + 1;
1721 			if (TCP_OFF(tcp) > (sizeof (tcphdr_t) >> 2)) {
1722 				(void) fr_tcpoptions(fin, tcp, fdata);
1723 			}
1724 			if ((fin->fin_out != 0) && (is->is_pass & FR_NEWISN))
1725 				fr_checknewisn(fin, is);
1726 		} else if (flags == TH_SYN) {
1727 			is->is_s0[source] = ntohl(tcp->th_seq) + 1;
1728 			if ((TCP_OFF(tcp) > (sizeof(tcphdr_t) >> 2)))
1729 				(void) fr_tcpoptions(fin, tcp, tdata);
1730 
1731 			if ((fin->fin_out != 0) && (is->is_pass & FR_NEWISN))
1732 				fr_checknewisn(fin, is);
1733 
1734 		}
1735 		ret = 1;
1736 	} else
1737 		fin->fin_flx |= FI_OOW;
1738 	MUTEX_EXIT(&is->is_lock);
1739 	return ret;
1740 }
1741 
1742 
1743 /* ------------------------------------------------------------------------ */
1744 /* Function:    fr_checknewisn                                              */
1745 /* Returns:     Nil                                                         */
1746 /* Parameters:  fin(I)   - pointer to packet information                    */
1747 /*              is(I)  - pointer to master state structure                  */
1748 /*                                                                          */
1749 /* Check to see if this TCP connection is expecting and needs a new         */
1750 /* sequence number for a particular direction of the connection.            */
1751 /*                                                                          */
1752 /* NOTE: This does not actually change the sequence numbers, only gets new  */
1753 /* one ready.                                                               */
1754 /* ------------------------------------------------------------------------ */
1755 static void fr_checknewisn(fin, is)
1756 fr_info_t *fin;
1757 ipstate_t *is;
1758 {
1759 	u_32_t sumd, old, new;
1760 	tcphdr_t *tcp;
1761 	int i;
1762 
1763 	i = fin->fin_rev;
1764 	tcp = fin->fin_dp;
1765 
1766 	if (((i == 0) && !(is->is_flags & IS_ISNSYN)) ||
1767 	    ((i == 1) && !(is->is_flags & IS_ISNACK))) {
1768 		old = ntohl(tcp->th_seq);
1769 		new = fr_newisn(fin);
1770 		is->is_isninc[i] = new - old;
1771 		CALC_SUMD(old, new, sumd);
1772 		is->is_sumd[i] = (sumd & 0xffff) + (sumd >> 16);
1773 
1774 		is->is_flags |= ((i == 0) ? IS_ISNSYN : IS_ISNACK);
1775 	}
1776 }
1777 
1778 
1779 /* ------------------------------------------------------------------------ */
1780 /* Function:    fr_tcpinwindow                                              */
1781 /* Returns:     int - 1 == packet inside TCP "window", 0 == not inside.     */
1782 /* Parameters:  fin(I)   - pointer to packet information                    */
1783 /*              fdata(I) - pointer to tcp state informatio (forward)        */
1784 /*              tdata(I) - pointer to tcp state informatio (reverse)        */
1785 /*              tcp(I)   - pointer to TCP packet header                     */
1786 /*                                                                          */
1787 /* Given a packet has matched addresses and ports, check to see if it is    */
1788 /* within the TCP data window.  In a show of generosity, allow packets that */
1789 /* are within the window space behind the current sequence # as well.       */
1790 /* ------------------------------------------------------------------------ */
1791 int fr_tcpinwindow(fin, fdata, tdata, tcp, flags)
1792 fr_info_t *fin;
1793 tcpdata_t  *fdata, *tdata;
1794 tcphdr_t *tcp;
1795 int flags;
1796 {
1797 	tcp_seq seq, ack, end;
1798 	int ackskew, tcpflags;
1799 	u_32_t win, maxwin;
1800 	int dsize, inseq;
1801 
1802 	/*
1803 	 * Find difference between last checked packet and this packet.
1804 	 */
1805 	tcpflags = tcp->th_flags;
1806 	seq = ntohl(tcp->th_seq);
1807 	ack = ntohl(tcp->th_ack);
1808 
1809 	if (tcpflags & TH_SYN)
1810 		win = ntohs(tcp->th_win);
1811 	else
1812 		win = ntohs(tcp->th_win) << fdata->td_winscale;
1813 
1814 	/*
1815 	 * win 0 means the receiving endpoint has closed the window, because it
1816 	 * has not enough memory to receive data from sender. In such case we
1817 	 * are pretending window size to be 1 to let TCP probe data through.
1818 	 * TCP probe data can be either 0 or 1 octet of data, the RFC does not
1819 	 * state this accurately, so we have to allow 1 octet (win = 1) even if
1820 	 * the window is closed (win == 0).
1821 	 */
1822 	if (win == 0)
1823 		win = 1;
1824 
1825 	dsize = fin->fin_dlen - (TCP_OFF(tcp) << 2) +
1826 		((tcpflags & TH_SYN) ? 1 : 0) + ((tcpflags & TH_FIN) ? 1 : 0);
1827 
1828 	/*
1829 	 * if window scaling is present, the scaling is only allowed
1830 	 * for windows not in the first SYN packet. In that packet the
1831 	 * window is 65535 to specify the largest window possible
1832 	 * for receivers not implementing the window scale option.
1833 	 * Currently, we do not assume TTCP here. That means that
1834 	 * if we see a second packet from a host (after the initial
1835 	 * SYN), we can assume that the receiver of the SYN did
1836 	 * already send back the SYN/ACK (and thus that we know if
1837 	 * the receiver also does window scaling)
1838 	 */
1839 	if (!(tcpflags & TH_SYN) && (fdata->td_winflags & TCP_WSCALE_FIRST)) {
1840 		fdata->td_maxwin = win;
1841 	}
1842 
1843 	end = seq + dsize;
1844 
1845 	if ((fdata->td_end == 0) &&
1846 	    (!(flags & IS_TCPFSM) ||
1847 	     ((tcpflags & TH_OPENING) == TH_OPENING))) {
1848 		/*
1849 		 * Must be a (outgoing) SYN-ACK in reply to a SYN.
1850 		 */
1851 		fdata->td_end = end - 1;
1852 		fdata->td_maxwin = 1;
1853 		fdata->td_maxend = end + win;
1854 	}
1855 
1856 	if (!(tcpflags & TH_ACK)) {  /* Pretend an ack was sent */
1857 		ack = tdata->td_end;
1858 	} else if (((tcpflags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) &&
1859 		   (ack == 0)) {
1860 		/* gross hack to get around certain broken tcp stacks */
1861 		ack = tdata->td_end;
1862 	}
1863 
1864 	maxwin = tdata->td_maxwin;
1865 	ackskew = tdata->td_end - ack;
1866 
1867 	/*
1868 	 * Strict sequencing only allows in-order delivery.
1869 	 */
1870 	if ((flags & IS_STRICT) != 0) {
1871 		if (seq != fdata->td_end) {
1872 			DTRACE_PROBE(strict_check);
1873 			return 0;
1874 		}
1875 	}
1876 
1877 #define	SEQ_GE(a,b)	((int)((a) - (b)) >= 0)
1878 #define	SEQ_GT(a,b)	((int)((a) - (b)) > 0)
1879 	inseq = 0;
1880 	DTRACE_PROBE4(
1881 		dyn_params,
1882 		int, dsize,
1883 		int, ackskew,
1884 		int, maxwin,
1885 		int, win
1886 	);
1887 	if (
1888 #if defined(_KERNEL)
1889 		/*
1890 		 * end <-> s + n
1891 		 * maxend <-> ack + win
1892 		 * this is upperbound check
1893 		 */
1894 	    (SEQ_GE(fdata->td_maxend, end)) &&
1895 		/*
1896 		 * this is lowerbound check
1897 		 */
1898 	    (SEQ_GE(seq, fdata->td_end - maxwin)) &&
1899 #endif
1900 /* XXX what about big packets */
1901 #define MAXACKWINDOW 66000
1902 	    (-ackskew <= (MAXACKWINDOW << fdata->td_winscale)) &&
1903 	    ( ackskew <= (MAXACKWINDOW << fdata->td_winscale))) {
1904 		inseq = 1;
1905 	/*
1906 	 * Microsoft Windows will send the next packet to the right of the
1907 	 * window if SACK is in use.
1908 	 */
1909 	} else if ((seq == fdata->td_maxend) && (ackskew == 0) &&
1910 	    (fdata->td_winflags & TCP_SACK_PERMIT) &&
1911 	    (tdata->td_winflags & TCP_SACK_PERMIT)) {
1912 		inseq = 1;
1913 	/*
1914 	 * RST ACK with SEQ equal to 0 is sent by some OSes (i.e. Solaris) as a
1915 	 * response to initial SYN packet, when  there is no application
1916 	 * listeing to on a port, where the SYN packet has came to.
1917 	 */
1918 	} else if ((seq == 0) && (tcpflags == (TH_RST|TH_ACK)) &&
1919 			(ackskew >= -1) && (ackskew <= 1)) {
1920 		inseq = 1;
1921 	} else if (!(flags & IS_TCPFSM)) {
1922 
1923 		if (!(fdata->td_winflags &
1924 			    (TCP_WSCALE_SEEN|TCP_WSCALE_FIRST))) {
1925 			/*
1926 			 * No TCPFSM and no window scaling, so make some
1927 			 * extra guesses.
1928 			 */
1929 			if ((seq == fdata->td_maxend) && (ackskew == 0))
1930 				inseq = 1;
1931 			else if (SEQ_GE(seq + maxwin, fdata->td_end - maxwin))
1932 				inseq = 1;
1933 		}
1934 	}
1935 
1936 	if (inseq) {
1937 		/* if ackskew < 0 then this should be due to fragmented
1938 		 * packets. There is no way to know the length of the
1939 		 * total packet in advance.
1940 		 * We do know the total length from the fragment cache though.
1941 		 * Note however that there might be more sessions with
1942 		 * exactly the same source and destination parameters in the
1943 		 * state cache (and source and destination is the only stuff
1944 		 * that is saved in the fragment cache). Note further that
1945 		 * some TCP connections in the state cache are hashed with
1946 		 * sport and dport as well which makes it not worthwhile to
1947 		 * look for them.
1948 		 * Thus, when ackskew is negative but still seems to belong
1949 		 * to this session, we bump up the destinations end value.
1950 		 */
1951 		if (ackskew < 0) {
1952 			DTRACE_PROBE2(end_update_td,
1953 				int, tdata->td_end,
1954 				int, ack
1955 			);
1956 			tdata->td_end = ack;
1957 		}
1958 
1959 		/* update max window seen */
1960 		if (fdata->td_maxwin < win) {
1961 			DTRACE_PROBE2(win_update_fd,
1962 				int, fdata->td_maxwin,
1963 				int, win
1964 			);
1965 			fdata->td_maxwin = win;
1966 		}
1967 
1968 		if (SEQ_GT(end, fdata->td_end)) {
1969 			DTRACE_PROBE2(end_update_fd,
1970 				int, fdata->td_end,
1971 				int, end
1972 			);
1973 			fdata->td_end = end;
1974 		}
1975 
1976 		if (SEQ_GE(ack + win, tdata->td_maxend)) {
1977 			DTRACE_PROBE2(max_end_update_td,
1978 				int, tdata->td_maxend,
1979 				int, ack + win
1980 			);
1981 			tdata->td_maxend = ack + win;
1982 		}
1983 
1984 		return 1;
1985 	}
1986 	fin->fin_flx |= FI_OOW;
1987 
1988 #if defined(_KERNEL)
1989 	if (!(SEQ_GE(seq, fdata->td_end - maxwin)))
1990 		fin->fin_flx |= FI_NEG_OOW;
1991 #endif
1992 
1993 	return 0;
1994 }
1995 
1996 
1997 /* ------------------------------------------------------------------------ */
1998 /* Function:    fr_stclone                                                  */
1999 /* Returns:     ipstate_t* - NULL == cloning failed,                        */
2000 /*                           else pointer to new state structure            */
2001 /* Parameters:  fin(I) - pointer to packet information                      */
2002 /*              tcp(I) - pointer to TCP/UDP header                          */
2003 /*              is(I)  - pointer to master state structure                  */
2004 /*                                                                          */
2005 /* Create a "duplcate" state table entry from the master.                   */
2006 /* ------------------------------------------------------------------------ */
2007 static ipstate_t *fr_stclone(fin, tcp, is)
2008 fr_info_t *fin;
2009 tcphdr_t *tcp;
2010 ipstate_t *is;
2011 {
2012 	ipstate_t *clone;
2013 	u_32_t send;
2014 	ipf_stack_t *ifs = fin->fin_ifs;
2015 
2016 	/*
2017 	 * Trigger automatic call to fr_state_flush() if the
2018 	 * table has reached capacity specified by hi watermark.
2019 	 */
2020 	if (ST_TAB_WATER_LEVEL(ifs) > ifs->ifs_state_flush_level_hi)
2021 		ifs->ifs_fr_state_doflush = 1;
2022 
2023 	/*
2024 	 * If automatic flushing did not do its job, and the table
2025 	 * has filled up, don't try to create a new entry.  A NULL
2026 	 * return will indicate that the cloning has failed.
2027 	 */
2028 	if (ifs->ifs_ips_num >= ifs->ifs_fr_statemax) {
2029 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_max);
2030 		return NULL;
2031 	}
2032 
2033 	KMALLOC(clone, ipstate_t *);
2034 	if (clone == NULL)
2035 		return NULL;
2036 	bcopy((char *)is, (char *)clone, sizeof(*clone));
2037 
2038 	MUTEX_NUKE(&clone->is_lock);
2039 
2040 	clone->is_die = ONE_DAY + ifs->ifs_fr_ticks;
2041 	clone->is_state[0] = 0;
2042 	clone->is_state[1] = 0;
2043 	send = ntohl(tcp->th_seq) + fin->fin_dlen - (TCP_OFF(tcp) << 2) +
2044 		((tcp->th_flags & TH_SYN) ? 1 : 0) +
2045 		((tcp->th_flags & TH_FIN) ? 1 : 0);
2046 
2047 	if (fin->fin_rev == 1) {
2048 		clone->is_dend = send;
2049 		clone->is_maxdend = send;
2050 		clone->is_send = 0;
2051 		clone->is_maxswin = 1;
2052 		clone->is_maxdwin = ntohs(tcp->th_win);
2053 		if (clone->is_maxdwin == 0)
2054 			clone->is_maxdwin = 1;
2055 	} else {
2056 		clone->is_send = send;
2057 		clone->is_maxsend = send;
2058 		clone->is_dend = 0;
2059 		clone->is_maxdwin = 1;
2060 		clone->is_maxswin = ntohs(tcp->th_win);
2061 		if (clone->is_maxswin == 0)
2062 			clone->is_maxswin = 1;
2063 	}
2064 
2065 	clone->is_flags &= ~SI_CLONE;
2066 	clone->is_flags |= SI_CLONED;
2067 	fr_stinsert(clone, fin->fin_rev, ifs);
2068 	clone->is_ref = 1;
2069 	if (clone->is_p == IPPROTO_TCP) {
2070 		(void) fr_tcp_age(&clone->is_sti, fin, ifs->ifs_ips_tqtqb,
2071 				  clone->is_flags);
2072 	}
2073 	MUTEX_EXIT(&clone->is_lock);
2074 #ifdef	IPFILTER_SCAN
2075 	(void) ipsc_attachis(is);
2076 #endif
2077 #ifdef	IPFILTER_SYNC
2078 	if (is->is_flags & IS_STATESYNC)
2079 		clone->is_sync = ipfsync_new(SMC_STATE, fin, clone);
2080 #endif
2081 	return clone;
2082 }
2083 
2084 
2085 /* ------------------------------------------------------------------------ */
2086 /* Function:    fr_matchsrcdst                                              */
2087 /* Returns:     Nil                                                         */
2088 /* Parameters:  fin(I) - pointer to packet information                      */
2089 /*              is(I)  - pointer to state structure                         */
2090 /*              src(I) - pointer to source address                          */
2091 /*              dst(I) - pointer to destination address                     */
2092 /*              tcp(I) - pointer to TCP/UDP header                          */
2093 /*                                                                          */
2094 /* Match a state table entry against an IP packet.  The logic below is that */
2095 /* ret gets set to one if the match succeeds, else remains 0.  If it is     */
2096 /* still 0 after the test. no match.                                        */
2097 /* ------------------------------------------------------------------------ */
2098 static ipstate_t *fr_matchsrcdst(fin, is, src, dst, tcp, cmask)
2099 fr_info_t *fin;
2100 ipstate_t *is;
2101 i6addr_t *src, *dst;
2102 tcphdr_t *tcp;
2103 u_32_t cmask;
2104 {
2105 	int ret = 0, rev, out, flags, flx = 0, idx;
2106 	u_short sp, dp;
2107 	u_32_t cflx;
2108 	void *ifp;
2109 	ipf_stack_t *ifs = fin->fin_ifs;
2110 
2111 	rev = IP6_NEQ(&is->is_dst, dst);
2112 	ifp = fin->fin_ifp;
2113 	out = fin->fin_out;
2114 	flags = is->is_flags;
2115 	sp = 0;
2116 	dp = 0;
2117 
2118 	if (tcp != NULL) {
2119 		sp = htons(fin->fin_sport);
2120 		dp = ntohs(fin->fin_dport);
2121 	}
2122 	if (!rev) {
2123 		if (tcp != NULL) {
2124 			if (!(flags & SI_W_SPORT) && (sp != is->is_sport))
2125 				rev = 1;
2126 			else if (!(flags & SI_W_DPORT) && (dp != is->is_dport))
2127 				rev = 1;
2128 		}
2129 	}
2130 
2131 	idx = (out << 1) + rev;
2132 
2133 	/*
2134 	 * If the interface for this 'direction' is set, make sure it matches.
2135 	 * An interface name that is not set matches any, as does a name of *.
2136 	 */
2137 	if ((is->is_ifp[idx] == NULL &&
2138 	    (*is->is_ifname[idx] == '\0' || *is->is_ifname[idx] == '*')) ||
2139 	    is->is_ifp[idx] == ifp)
2140 		ret = 1;
2141 
2142 	if (ret == 0) {
2143 		DTRACE_PROBE(no_match_on_iface);
2144 		return NULL;
2145 	}
2146 	ret = 0;
2147 
2148 	/*
2149 	 * Match addresses and ports.
2150 	 */
2151 	if (rev == 0) {
2152 		if ((IP6_EQ(&is->is_dst, dst) || (flags & SI_W_DADDR)) &&
2153 		    (IP6_EQ(&is->is_src, src) || (flags & SI_W_SADDR))) {
2154 			if (tcp) {
2155 				if ((sp == is->is_sport || flags & SI_W_SPORT)&&
2156 				    (dp == is->is_dport || flags & SI_W_DPORT))
2157 					ret = 1;
2158 			} else {
2159 				ret = 1;
2160 			}
2161 		}
2162 	} else {
2163 		if ((IP6_EQ(&is->is_dst, src) || (flags & SI_W_DADDR)) &&
2164 		    (IP6_EQ(&is->is_src, dst) || (flags & SI_W_SADDR))) {
2165 			if (tcp) {
2166 				if ((dp == is->is_sport || flags & SI_W_SPORT)&&
2167 				    (sp == is->is_dport || flags & SI_W_DPORT))
2168 					ret = 1;
2169 			} else {
2170 				ret = 1;
2171 			}
2172 		}
2173 	}
2174 
2175 	if (ret == 0) {
2176 		DTRACE_PROBE(no_match_on_addrs);
2177 		return NULL;
2178 	}
2179 	/*
2180 	 * Whether or not this should be here, is questionable, but the aim
2181 	 * is to get this out of the main line.
2182 	 */
2183 	if (tcp == NULL)
2184 		flags = is->is_flags & ~(SI_WILDP|SI_NEWFR|SI_CLONE|SI_CLONED);
2185 
2186 	/*
2187 	 * Only one of the source or destination address can be flaged as a
2188 	 * wildcard.  Fill in the missing address, if set.
2189 	 * For IPv6, if the address being copied in is multicast, then
2190 	 * don't reset the wild flag - multicast causes it to be set in the
2191 	 * first place!
2192 	 */
2193 	if ((flags & (SI_W_SADDR|SI_W_DADDR))) {
2194 		fr_ip_t *fi = &fin->fin_fi;
2195 
2196 		if ((flags & SI_W_SADDR) != 0) {
2197 			if (rev == 0) {
2198 #ifdef USE_INET6
2199 				if (is->is_v == 6 &&
2200 				    IN6_IS_ADDR_MULTICAST(&fi->fi_src.in6))
2201 					/*EMPTY*/;
2202 				else
2203 #endif
2204 				{
2205 					is->is_src = fi->fi_src;
2206 					is->is_flags &= ~SI_W_SADDR;
2207 				}
2208 			} else {
2209 #ifdef USE_INET6
2210 				if (is->is_v == 6 &&
2211 				    IN6_IS_ADDR_MULTICAST(&fi->fi_dst.in6))
2212 					/*EMPTY*/;
2213 				else
2214 #endif
2215 				{
2216 					is->is_src = fi->fi_dst;
2217 					is->is_flags &= ~SI_W_SADDR;
2218 				}
2219 			}
2220 		} else if ((flags & SI_W_DADDR) != 0) {
2221 			if (rev == 0) {
2222 #ifdef USE_INET6
2223 				if (is->is_v == 6 &&
2224 				    IN6_IS_ADDR_MULTICAST(&fi->fi_dst.in6))
2225 					/*EMPTY*/;
2226 				else
2227 #endif
2228 				{
2229 					is->is_dst = fi->fi_dst;
2230 					is->is_flags &= ~SI_W_DADDR;
2231 				}
2232 			} else {
2233 #ifdef USE_INET6
2234 				if (is->is_v == 6 &&
2235 				    IN6_IS_ADDR_MULTICAST(&fi->fi_src.in6))
2236 					/*EMPTY*/;
2237 				else
2238 #endif
2239 				{
2240 					is->is_dst = fi->fi_src;
2241 					is->is_flags &= ~SI_W_DADDR;
2242 				}
2243 			}
2244 		}
2245 		if ((is->is_flags & (SI_WILDA|SI_WILDP)) == 0) {
2246 			ATOMIC_DECL(ifs->ifs_ips_stats.iss_wild);
2247 		}
2248 	}
2249 
2250 	flx = fin->fin_flx & cmask;
2251 	cflx = is->is_flx[out][rev];
2252 
2253 	/*
2254 	 * Match up any flags set from IP options.
2255 	 */
2256 	if ((cflx && (flx != (cflx & cmask))) ||
2257 	    ((fin->fin_optmsk & is->is_optmsk[rev]) != is->is_opt[rev]) ||
2258 	    ((fin->fin_secmsk & is->is_secmsk) != is->is_sec) ||
2259 	    ((fin->fin_auth & is->is_authmsk) != is->is_auth)) {
2260 		DTRACE_PROBE4(no_match_on_flags,
2261 		    int, (cflx && (flx != (cflx & cmask))),
2262 		    int,
2263 		    ((fin->fin_optmsk & is->is_optmsk[rev]) != is->is_opt[rev]),
2264 		    int, ((fin->fin_secmsk & is->is_secmsk) != is->is_sec),
2265 		    int, ((fin->fin_auth & is->is_authmsk) != is->is_auth)
2266 		);
2267 		return NULL;
2268 	}
2269 	/*
2270 	 * Only one of the source or destination port can be flagged as a
2271 	 * wildcard.  When filling it in, fill in a copy of the matched entry
2272 	 * if it has the cloning flag set.
2273 	 */
2274 	if ((fin->fin_flx & FI_IGNORE) != 0) {
2275 		fin->fin_rev = rev;
2276 		return is;
2277 	}
2278 
2279 	if ((flags & (SI_W_SPORT|SI_W_DPORT))) {
2280 		if ((flags & SI_CLONE) != 0) {
2281 			ipstate_t *clone;
2282 
2283 			clone = fr_stclone(fin, tcp, is);
2284 			if (clone == NULL)
2285 				return NULL;
2286 			is = clone;
2287 		} else {
2288 			ATOMIC_DECL(ifs->ifs_ips_stats.iss_wild);
2289 		}
2290 
2291 		if ((flags & SI_W_SPORT) != 0) {
2292 			if (rev == 0) {
2293 				is->is_sport = sp;
2294 				is->is_send = ntohl(tcp->th_seq);
2295 			} else {
2296 				is->is_sport = dp;
2297 				is->is_send = ntohl(tcp->th_ack);
2298 			}
2299 			is->is_maxsend = is->is_send + 1;
2300 		} else if ((flags & SI_W_DPORT) != 0) {
2301 			if (rev == 0) {
2302 				is->is_dport = dp;
2303 				is->is_dend = ntohl(tcp->th_ack);
2304 			} else {
2305 				is->is_dport = sp;
2306 				is->is_dend = ntohl(tcp->th_seq);
2307 			}
2308 			is->is_maxdend = is->is_dend + 1;
2309 		}
2310 		is->is_flags &= ~(SI_W_SPORT|SI_W_DPORT);
2311 		if ((flags & SI_CLONED) && ifs->ifs_ipstate_logging)
2312 			ipstate_log(is, ISL_CLONE, ifs);
2313 	}
2314 
2315 	ret = -1;
2316 
2317 	if (is->is_flx[out][rev] == 0) {
2318 		is->is_flx[out][rev] = flx;
2319 		/*
2320 		 * If we are dealing with the first packet coming in reverse
2321 		 * direction (sent by peer), then we have to set options into
2322 		 * state.
2323 		 */
2324 		if (rev == 1 && is->is_optmsk[1] == 0x0) {
2325 			is->is_optmsk[1] = 0xffffffff;
2326 			is->is_opt[1] = fin->fin_optmsk;
2327 			DTRACE_PROBE(set_rev_opts);
2328 		}
2329 		if (is->is_v == 6) {
2330 			is->is_opt[rev] &= ~0x8;
2331 			is->is_optmsk[rev] &= ~0x8;
2332 		}
2333 	}
2334 
2335 	/*
2336 	 * Check if the interface name for this "direction" is set and if not,
2337 	 * fill it in.
2338 	 */
2339 	if (is->is_ifp[idx] == NULL &&
2340 	    (*is->is_ifname[idx] == '\0' || *is->is_ifname[idx] == '*')) {
2341 		is->is_ifp[idx] = ifp;
2342 		COPYIFNAME(ifp, is->is_ifname[idx], fin->fin_v);
2343 	}
2344 	fin->fin_rev = rev;
2345 	return is;
2346 }
2347 
2348 
2349 /* ------------------------------------------------------------------------ */
2350 /* Function:    fr_checkicmpmatchingstate                                   */
2351 /* Returns:     Nil                                                         */
2352 /* Parameters:  fin(I) - pointer to packet information                      */
2353 /*                                                                          */
2354 /* If we've got an ICMP error message, using the information stored in the  */
2355 /* ICMP packet, look for a matching state table entry.                      */
2356 /*                                                                          */
2357 /* If we return NULL then no lock on ipf_state is held.                     */
2358 /* If we return non-null then a read-lock on ipf_state is held.             */
2359 /* ------------------------------------------------------------------------ */
2360 static ipstate_t *fr_checkicmpmatchingstate(fin)
2361 fr_info_t *fin;
2362 {
2363 	ipstate_t *is, **isp;
2364 	u_short sport, dport;
2365 	u_char	pr;
2366 	int backward, i, oi;
2367 	i6addr_t dst, src;
2368 	struct icmp *ic;
2369 	u_short savelen;
2370 	icmphdr_t *icmp;
2371 	fr_info_t ofin;
2372 	tcphdr_t *tcp;
2373 	int len;
2374 	ip_t *oip;
2375 	u_int hv;
2376 	ipf_stack_t *ifs = fin->fin_ifs;
2377 
2378 	/*
2379 	 * Does it at least have the return (basic) IP header ?
2380 	 * Is it an actual recognised ICMP error type?
2381 	 * Only a basic IP header (no options) should be with
2382 	 * an ICMP error header.
2383 	 */
2384 	if ((fin->fin_v != 4) || (fin->fin_hlen != sizeof(ip_t)) ||
2385 	    (fin->fin_plen < ICMPERR_MINPKTLEN) ||
2386 	    !(fin->fin_flx & FI_ICMPERR))
2387 		return NULL;
2388 	ic = fin->fin_dp;
2389 
2390 	oip = (ip_t *)((char *)ic + ICMPERR_ICMPHLEN);
2391 	/*
2392 	 * Check if the at least the old IP header (with options) and
2393 	 * 8 bytes of payload is present.
2394 	 */
2395 	if (fin->fin_plen < ICMPERR_MAXPKTLEN + ((IP_HL(oip) - 5) << 2))
2396 		return NULL;
2397 
2398 	/*
2399 	 * Sanity Checks.
2400 	 */
2401 	len = fin->fin_dlen - ICMPERR_ICMPHLEN;
2402 	if ((len <= 0) || ((IP_HL(oip) << 2) > len))
2403 		return NULL;
2404 
2405 	/*
2406 	 * Is the buffer big enough for all of it ?  It's the size of the IP
2407 	 * header claimed in the encapsulated part which is of concern.  It
2408 	 * may be too big to be in this buffer but not so big that it's
2409 	 * outside the ICMP packet, leading to TCP deref's causing problems.
2410 	 * This is possible because we don't know how big oip_hl is when we
2411 	 * do the pullup early in fr_check() and thus can't guarantee it is
2412 	 * all here now.
2413 	 */
2414 #ifdef  _KERNEL
2415 	{
2416 	mb_t *m;
2417 
2418 	m = fin->fin_m;
2419 # if defined(MENTAT)
2420 	if ((char *)oip + len > (char *)m->b_wptr)
2421 		return NULL;
2422 # else
2423 	if ((char *)oip + len > (char *)fin->fin_ip + m->m_len)
2424 		return NULL;
2425 # endif
2426 	}
2427 #endif
2428 	bcopy((char *)fin, (char *)&ofin, sizeof(*fin));
2429 
2430 	/*
2431 	 * in the IPv4 case we must zero the i6addr union otherwise
2432 	 * the IP6_EQ and IP6_NEQ macros produce the wrong results because
2433 	 * of the 'junk' in the unused part of the union
2434 	 */
2435 	bzero((char *)&src, sizeof(src));
2436 	bzero((char *)&dst, sizeof(dst));
2437 
2438 	/*
2439 	 * we make an fin entry to be able to feed it to
2440 	 * matchsrcdst note that not all fields are encessary
2441 	 * but this is the cleanest way. Note further we fill
2442 	 * in fin_mp such that if someone uses it we'll get
2443 	 * a kernel panic. fr_matchsrcdst does not use this.
2444 	 *
2445 	 * watch out here, as ip is in host order and oip in network
2446 	 * order. Any change we make must be undone afterwards, like
2447 	 * oip->ip_off - it is still in network byte order so fix it.
2448 	 */
2449 	savelen = oip->ip_len;
2450 	oip->ip_len = len;
2451 	oip->ip_off = ntohs(oip->ip_off);
2452 
2453 	ofin.fin_flx = FI_NOCKSUM;
2454 	ofin.fin_v = 4;
2455 	ofin.fin_ip = oip;
2456 	ofin.fin_m = NULL;	/* if dereferenced, panic XXX */
2457 	ofin.fin_mp = NULL;	/* if dereferenced, panic XXX */
2458 	ofin.fin_plen = fin->fin_dlen - ICMPERR_ICMPHLEN;
2459 	(void) fr_makefrip(IP_HL(oip) << 2, oip, &ofin);
2460 	ofin.fin_ifp = fin->fin_ifp;
2461 	ofin.fin_out = !fin->fin_out;
2462 	/*
2463 	 * Reset the short and bad flag here because in fr_matchsrcdst()
2464 	 * the flags for the current packet (fin_flx) are compared against
2465 	 * those for the existing session.
2466 	 */
2467 	ofin.fin_flx &= ~(FI_BAD|FI_SHORT);
2468 
2469 	/*
2470 	 * Put old values of ip_len and ip_off back as we don't know
2471 	 * if we have to forward the packet (or process it again.
2472 	 */
2473 	oip->ip_len = savelen;
2474 	oip->ip_off = htons(oip->ip_off);
2475 
2476 	switch (oip->ip_p)
2477 	{
2478 	case IPPROTO_ICMP :
2479 		/*
2480 		 * an ICMP error can only be generated as a result of an
2481 		 * ICMP query, not as the response on an ICMP error
2482 		 *
2483 		 * XXX theoretically ICMP_ECHOREP and the other reply's are
2484 		 * ICMP query's as well, but adding them here seems strange XXX
2485 		 */
2486 		if ((ofin.fin_flx & FI_ICMPERR) != 0)
2487 		    	return NULL;
2488 
2489 		/*
2490 		 * perform a lookup of the ICMP packet in the state table
2491 		 */
2492 		icmp = (icmphdr_t *)((char *)oip + (IP_HL(oip) << 2));
2493 		hv = (pr = oip->ip_p);
2494 		src.in4 = oip->ip_src;
2495 		hv += src.in4.s_addr;
2496 		dst.in4 = oip->ip_dst;
2497 		hv += dst.in4.s_addr;
2498 		hv += icmp->icmp_id;
2499 		hv = DOUBLE_HASH(hv, ifs);
2500 
2501 		READ_ENTER(&ifs->ifs_ipf_state);
2502 		for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
2503 			isp = &is->is_hnext;
2504 			if ((is->is_p != pr) || (is->is_v != 4))
2505 				continue;
2506 			if (is->is_pass & FR_NOICMPERR)
2507 				continue;
2508 			is = fr_matchsrcdst(&ofin, is, &src, &dst,
2509 					    NULL, FI_ICMPCMP);
2510 			if (is != NULL) {
2511 				if ((is->is_pass & FR_NOICMPERR) != 0) {
2512 					RWLOCK_EXIT(&ifs->ifs_ipf_state);
2513 					return NULL;
2514 				}
2515 				/*
2516 				 * i  : the index of this packet (the icmp
2517 				 *      unreachable)
2518 				 * oi : the index of the original packet found
2519 				 *      in the icmp header (i.e. the packet
2520 				 *      causing this icmp)
2521 				 * backward : original packet was backward
2522 				 *      compared to the state
2523 				 */
2524 				backward = IP6_NEQ(&is->is_src, &src);
2525 				fin->fin_rev = !backward;
2526 				i = (!backward << 1) + fin->fin_out;
2527 				oi = (backward << 1) + ofin.fin_out;
2528 				if (is->is_icmppkts[i] > is->is_pkts[oi])
2529 					continue;
2530 				ifs->ifs_ips_stats.iss_hits++;
2531 				is->is_icmppkts[i]++;
2532 				return is;
2533 			}
2534 		}
2535 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
2536 		return NULL;
2537 	case IPPROTO_TCP :
2538 	case IPPROTO_UDP :
2539 		break;
2540 	default :
2541 		return NULL;
2542 	}
2543 
2544 	tcp = (tcphdr_t *)((char *)oip + (IP_HL(oip) << 2));
2545 	dport = tcp->th_dport;
2546 	sport = tcp->th_sport;
2547 
2548 	hv = (pr = oip->ip_p);
2549 	src.in4 = oip->ip_src;
2550 	hv += src.in4.s_addr;
2551 	dst.in4 = oip->ip_dst;
2552 	hv += dst.in4.s_addr;
2553 	hv += dport;
2554 	hv += sport;
2555 	hv = DOUBLE_HASH(hv, ifs);
2556 
2557 	READ_ENTER(&ifs->ifs_ipf_state);
2558 	for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
2559 		isp = &is->is_hnext;
2560 		/*
2561 		 * Only allow this icmp though if the
2562 		 * encapsulated packet was allowed through the
2563 		 * other way around. Note that the minimal amount
2564 		 * of info present does not allow for checking against
2565 		 * tcp internals such as seq and ack numbers.   Only the
2566 		 * ports are known to be present and can be even if the
2567 		 * short flag is set.
2568 		 */
2569 		if ((is->is_p == pr) && (is->is_v == 4) &&
2570 		    (is = fr_matchsrcdst(&ofin, is, &src, &dst,
2571 					 tcp, FI_ICMPCMP))) {
2572 			/*
2573 			 * i  : the index of this packet (the icmp unreachable)
2574 			 * oi : the index of the original packet found in the
2575 			 *      icmp header (i.e. the packet causing this icmp)
2576 			 * backward : original packet was backward compared to
2577 			 *            the state
2578 			 */
2579 			backward = IP6_NEQ(&is->is_src, &src);
2580 			fin->fin_rev = !backward;
2581 			i = (!backward << 1) + fin->fin_out;
2582 			oi = (backward << 1) + ofin.fin_out;
2583 
2584 			if (((is->is_pass & FR_NOICMPERR) != 0) ||
2585 			    (is->is_icmppkts[i] > is->is_pkts[oi]))
2586 				break;
2587 			ifs->ifs_ips_stats.iss_hits++;
2588 			is->is_icmppkts[i]++;
2589 			/*
2590 			 * we deliberately do not touch the timeouts
2591 			 * for the accompanying state table entry.
2592 			 * It remains to be seen if that is correct. XXX
2593 			 */
2594 			return is;
2595 		}
2596 	}
2597 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
2598 	return NULL;
2599 }
2600 
2601 
2602 /* ------------------------------------------------------------------------ */
2603 /* Function:    fr_ipsmove                                                  */
2604 /* Returns:     Nil                                                         */
2605 /* Parameters:  is(I) - pointer to state table entry                        */
2606 /*              hv(I) - new hash value for state table entry                */
2607 /* Write Locks: ipf_state                                                   */
2608 /*                                                                          */
2609 /* Move a state entry from one position in the hash table to another.       */
2610 /* ------------------------------------------------------------------------ */
2611 static void fr_ipsmove(is, hv, ifs)
2612 ipstate_t *is;
2613 u_int hv;
2614 ipf_stack_t *ifs;
2615 {
2616 	ipstate_t **isp;
2617 	u_int hvm;
2618 
2619 	ASSERT(rw_read_locked(&ifs->ifs_ipf_state.ipf_lk) == 0);
2620 
2621 	hvm = is->is_hv;
2622 	/*
2623 	 * Remove the hash from the old location...
2624 	 */
2625 	isp = is->is_phnext;
2626 	if (is->is_hnext)
2627 		is->is_hnext->is_phnext = isp;
2628 	*isp = is->is_hnext;
2629 	if (ifs->ifs_ips_table[hvm] == NULL)
2630 		ifs->ifs_ips_stats.iss_inuse--;
2631 	ifs->ifs_ips_stats.iss_bucketlen[hvm]--;
2632 
2633 	/*
2634 	 * ...and put the hash in the new one.
2635 	 */
2636 	hvm = DOUBLE_HASH(hv, ifs);
2637 	is->is_hv = hvm;
2638 	isp = &ifs->ifs_ips_table[hvm];
2639 	if (*isp)
2640 		(*isp)->is_phnext = &is->is_hnext;
2641 	else
2642 		ifs->ifs_ips_stats.iss_inuse++;
2643 	ifs->ifs_ips_stats.iss_bucketlen[hvm]++;
2644 	is->is_phnext = isp;
2645 	is->is_hnext = *isp;
2646 	*isp = is;
2647 }
2648 
2649 
2650 /* ------------------------------------------------------------------------ */
2651 /* Function:    fr_stlookup                                                 */
2652 /* Returns:     ipstate_t* - NULL == no matching state found,               */
2653 /*                           else pointer to state information is returned  */
2654 /* Parameters:  fin(I) - pointer to packet information                      */
2655 /*              tcp(I) - pointer to TCP/UDP header.                         */
2656 /*                                                                          */
2657 /* Search the state table for a matching entry to the packet described by   */
2658 /* the contents of *fin.                                                    */
2659 /*                                                                          */
2660 /* If we return NULL then no lock on ipf_state is held.                     */
2661 /* If we return non-null then a read-lock on ipf_state is held.             */
2662 /* ------------------------------------------------------------------------ */
2663 ipstate_t *fr_stlookup(fin, tcp, ifqp)
2664 fr_info_t *fin;
2665 tcphdr_t *tcp;
2666 ipftq_t **ifqp;
2667 {
2668 	u_int hv, hvm, pr, v, tryagain;
2669 	ipstate_t *is, **isp;
2670 	u_short dport, sport;
2671 	i6addr_t src, dst;
2672 	struct icmp *ic;
2673 	ipftq_t *ifq;
2674 	int oow;
2675 	ipf_stack_t *ifs = fin->fin_ifs;
2676 
2677 	is = NULL;
2678 	ifq = NULL;
2679 	tcp = fin->fin_dp;
2680 	ic = (struct icmp *)tcp;
2681 	hv = (pr = fin->fin_fi.fi_p);
2682 	src = fin->fin_fi.fi_src;
2683 	dst = fin->fin_fi.fi_dst;
2684 	hv += src.in4.s_addr;
2685 	hv += dst.in4.s_addr;
2686 
2687 	v = fin->fin_fi.fi_v;
2688 #ifdef	USE_INET6
2689 	if (v == 6) {
2690 		hv  += fin->fin_fi.fi_src.i6[1];
2691 		hv  += fin->fin_fi.fi_src.i6[2];
2692 		hv  += fin->fin_fi.fi_src.i6[3];
2693 
2694 		if ((fin->fin_p == IPPROTO_ICMPV6) &&
2695 		    IN6_IS_ADDR_MULTICAST(&fin->fin_fi.fi_dst.in6)) {
2696 			hv -= dst.in4.s_addr;
2697 		} else {
2698 			hv += fin->fin_fi.fi_dst.i6[1];
2699 			hv += fin->fin_fi.fi_dst.i6[2];
2700 			hv += fin->fin_fi.fi_dst.i6[3];
2701 		}
2702 	}
2703 #endif
2704 	if ((v == 4) &&
2705 	    (fin->fin_flx & (FI_MULTICAST|FI_BROADCAST|FI_MBCAST))) {
2706 		if (fin->fin_out == 0) {
2707 			hv -= src.in4.s_addr;
2708 		} else {
2709 			hv -= dst.in4.s_addr;
2710 		}
2711 	}
2712 
2713 	/*
2714 	 * Search the hash table for matching packet header info.
2715 	 */
2716 	switch (pr)
2717 	{
2718 #ifdef	USE_INET6
2719 	case IPPROTO_ICMPV6 :
2720 		tryagain = 0;
2721 		if (v == 6) {
2722 			if ((ic->icmp_type == ICMP6_ECHO_REQUEST) ||
2723 			    (ic->icmp_type == ICMP6_ECHO_REPLY)) {
2724 				hv += ic->icmp_id;
2725 			}
2726 		}
2727 		READ_ENTER(&ifs->ifs_ipf_state);
2728 icmp6again:
2729 		hvm = DOUBLE_HASH(hv, ifs);
2730 		for (isp = &ifs->ifs_ips_table[hvm]; ((is = *isp) != NULL); ) {
2731 			isp = &is->is_hnext;
2732 			if ((is->is_p != pr) || (is->is_v != v))
2733 				continue;
2734 			is = fr_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP);
2735 			if (is != NULL &&
2736 			    fr_matchicmpqueryreply(v, &is->is_icmp,
2737 						   ic, fin->fin_rev)) {
2738 				if (fin->fin_rev)
2739 					ifq = &ifs->ifs_ips_icmpacktq;
2740 				else
2741 					ifq = &ifs->ifs_ips_icmptq;
2742 				break;
2743 			}
2744 		}
2745 
2746 		if (is != NULL) {
2747 			if ((tryagain != 0) && !(is->is_flags & SI_W_DADDR)) {
2748 				hv += fin->fin_fi.fi_src.i6[0];
2749 				hv += fin->fin_fi.fi_src.i6[1];
2750 				hv += fin->fin_fi.fi_src.i6[2];
2751 				hv += fin->fin_fi.fi_src.i6[3];
2752 				fr_ipsmove(is, hv, ifs);
2753 				MUTEX_DOWNGRADE(&ifs->ifs_ipf_state);
2754 			}
2755 			break;
2756 		}
2757 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
2758 
2759 		/*
2760 		 * No matching icmp state entry. Perhaps this is a
2761 		 * response to another state entry.
2762 		 *
2763 		 * XXX With some ICMP6 packets, the "other" address is already
2764 		 * in the packet, after the ICMP6 header, and this could be
2765 		 * used in place of the multicast address.  However, taking
2766 		 * advantage of this requires some significant code changes
2767 		 * to handle the specific types where that is the case.
2768 		 */
2769 		if ((ifs->ifs_ips_stats.iss_wild != 0) && (v == 6) && (tryagain == 0) &&
2770 		    !IN6_IS_ADDR_MULTICAST(&fin->fin_fi.fi_src.in6)) {
2771 			hv -= fin->fin_fi.fi_src.i6[0];
2772 			hv -= fin->fin_fi.fi_src.i6[1];
2773 			hv -= fin->fin_fi.fi_src.i6[2];
2774 			hv -= fin->fin_fi.fi_src.i6[3];
2775 			tryagain = 1;
2776 			WRITE_ENTER(&ifs->ifs_ipf_state);
2777 			goto icmp6again;
2778 		}
2779 
2780 		is = fr_checkicmp6matchingstate(fin);
2781 		if (is != NULL)
2782 			return is;
2783 		break;
2784 #endif
2785 
2786 	case IPPROTO_ICMP :
2787 		if (v == 4) {
2788 			hv += ic->icmp_id;
2789 		}
2790 		hv = DOUBLE_HASH(hv, ifs);
2791 		READ_ENTER(&ifs->ifs_ipf_state);
2792 		for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
2793 			isp = &is->is_hnext;
2794 			if ((is->is_p != pr) || (is->is_v != v))
2795 				continue;
2796 			is = fr_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP);
2797 			if (is != NULL &&
2798 			    fr_matchicmpqueryreply(v, &is->is_icmp,
2799 						   ic, fin->fin_rev)) {
2800 				if (fin->fin_rev)
2801 					ifq = &ifs->ifs_ips_icmpacktq;
2802 				else
2803 					ifq = &ifs->ifs_ips_icmptq;
2804 				break;
2805 			}
2806 		}
2807 		if (is == NULL) {
2808 			RWLOCK_EXIT(&ifs->ifs_ipf_state);
2809 		}
2810 		break;
2811 
2812 	case IPPROTO_TCP :
2813 	case IPPROTO_UDP :
2814 		ifqp = NULL;
2815 		sport = htons(fin->fin_data[0]);
2816 		hv += sport;
2817 		dport = htons(fin->fin_data[1]);
2818 		hv += dport;
2819 		oow = 0;
2820 		tryagain = 0;
2821 		READ_ENTER(&ifs->ifs_ipf_state);
2822 retry_tcpudp:
2823 		hvm = DOUBLE_HASH(hv, ifs);
2824 		for (isp = &ifs->ifs_ips_table[hvm]; ((is = *isp) != NULL); ) {
2825 			isp = &is->is_hnext;
2826 			if ((is->is_p != pr) || (is->is_v != v))
2827 				continue;
2828 			fin->fin_flx &= ~FI_OOW;
2829 			is = fr_matchsrcdst(fin, is, &src, &dst, tcp, FI_CMP);
2830 			if (is != NULL) {
2831 				if (pr == IPPROTO_TCP) {
2832 					if (!fr_tcpstate(fin, tcp, is)) {
2833 						oow |= fin->fin_flx & FI_OOW;
2834 						continue;
2835 					}
2836 				}
2837 				break;
2838 			}
2839 		}
2840 		if (is != NULL) {
2841 			if (tryagain &&
2842 			    !(is->is_flags & (SI_CLONE|SI_WILDP|SI_WILDA))) {
2843 				hv += dport;
2844 				hv += sport;
2845 				fr_ipsmove(is, hv, ifs);
2846 				MUTEX_DOWNGRADE(&ifs->ifs_ipf_state);
2847 			}
2848 			break;
2849 		}
2850 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
2851 
2852 		if (ifs->ifs_ips_stats.iss_wild) {
2853 			if (tryagain == 0) {
2854 				hv -= dport;
2855 				hv -= sport;
2856 			} else if (tryagain == 1) {
2857 				hv = fin->fin_fi.fi_p;
2858 				/*
2859 				 * If we try to pretend this is a reply to a
2860 				 * multicast/broadcast packet then we need to
2861 				 * exclude part of the address from the hash
2862 				 * calculation.
2863 				 */
2864 				if (fin->fin_out == 0) {
2865 					hv += src.in4.s_addr;
2866 				} else {
2867 					hv += dst.in4.s_addr;
2868 				}
2869 				hv += dport;
2870 				hv += sport;
2871 			}
2872 			tryagain++;
2873 			if (tryagain <= 2) {
2874 				WRITE_ENTER(&ifs->ifs_ipf_state);
2875 				goto retry_tcpudp;
2876 			}
2877 		}
2878 		fin->fin_flx |= oow;
2879 		break;
2880 
2881 #if 0
2882 	case IPPROTO_GRE :
2883 		gre = fin->fin_dp;
2884 		if (GRE_REV(gre->gr_flags) == 1) {
2885 			hv += gre->gr_call;
2886 		}
2887 		/* FALLTHROUGH */
2888 #endif
2889 	default :
2890 		ifqp = NULL;
2891 		hvm = DOUBLE_HASH(hv, ifs);
2892 		READ_ENTER(&ifs->ifs_ipf_state);
2893 		for (isp = &ifs->ifs_ips_table[hvm]; ((is = *isp) != NULL); ) {
2894 			isp = &is->is_hnext;
2895 			if ((is->is_p != pr) || (is->is_v != v))
2896 				continue;
2897 			is = fr_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP);
2898 			if (is != NULL) {
2899 				ifq = &ifs->ifs_ips_iptq;
2900 				break;
2901 			}
2902 		}
2903 		if (is == NULL) {
2904 			RWLOCK_EXIT(&ifs->ifs_ipf_state);
2905 		}
2906 		break;
2907 	}
2908 
2909 	if ((is != NULL) && ((is->is_sti.tqe_flags & TQE_RULEBASED) != 0) &&
2910 	    (is->is_tqehead[fin->fin_rev] != NULL))
2911 		ifq = is->is_tqehead[fin->fin_rev];
2912 	if (ifq != NULL && ifqp != NULL)
2913 		*ifqp = ifq;
2914 	return is;
2915 }
2916 
2917 
2918 /* ------------------------------------------------------------------------ */
2919 /* Function:    fr_updatestate                                              */
2920 /* Returns:     Nil                                                         */
2921 /* Parameters:  fin(I) - pointer to packet information                      */
2922 /*              is(I)  - pointer to state table entry                       */
2923 /* Read Locks:  ipf_state                                                   */
2924 /*                                                                          */
2925 /* Updates packet and byte counters for a newly received packet.  Seeds the */
2926 /* fragment cache with a new entry as required.                             */
2927 /* ------------------------------------------------------------------------ */
2928 void fr_updatestate(fin, is, ifq)
2929 fr_info_t *fin;
2930 ipstate_t *is;
2931 ipftq_t *ifq;
2932 {
2933 	ipftqent_t *tqe;
2934 	int i, pass;
2935 	ipf_stack_t *ifs = fin->fin_ifs;
2936 
2937 	i = (fin->fin_rev << 1) + fin->fin_out;
2938 
2939 	/*
2940 	 * For TCP packets, ifq == NULL.  For all others, check if this new
2941 	 * queue is different to the last one it was on and move it if so.
2942 	 */
2943 	tqe = &is->is_sti;
2944 	MUTEX_ENTER(&is->is_lock);
2945 	if ((tqe->tqe_flags & TQE_RULEBASED) != 0)
2946 		ifq = is->is_tqehead[fin->fin_rev];
2947 
2948 	if (ifq != NULL)
2949 		fr_movequeue(tqe, tqe->tqe_ifq, ifq, ifs);
2950 
2951 	is->is_pkts[i]++;
2952 	fin->fin_pktnum = is->is_pkts[i] + is->is_icmppkts[i];
2953 	is->is_bytes[i] += fin->fin_plen;
2954 	MUTEX_EXIT(&is->is_lock);
2955 
2956 #ifdef	IPFILTER_SYNC
2957 	if (is->is_flags & IS_STATESYNC)
2958 		ipfsync_update(SMC_STATE, fin, is->is_sync);
2959 #endif
2960 
2961 	ATOMIC_INCL(ifs->ifs_ips_stats.iss_hits);
2962 
2963 	fin->fin_fr = is->is_rule;
2964 
2965 	/*
2966 	 * If this packet is a fragment and the rule says to track fragments,
2967 	 * then create a new fragment cache entry.
2968 	 */
2969 	pass = is->is_pass;
2970 	if ((fin->fin_flx & FI_FRAG) && FR_ISPASS(pass))
2971 		(void) fr_newfrag(fin, pass ^ FR_KEEPSTATE);
2972 }
2973 
2974 
2975 /* ------------------------------------------------------------------------ */
2976 /* Function:    fr_checkstate                                               */
2977 /* Returns:     frentry_t* - NULL == search failed,                         */
2978 /*                           else pointer to rule for matching state        */
2979 /* Parameters:  ifp(I)   - pointer to interface                             */
2980 /*              passp(I) - pointer to filtering result flags                */
2981 /*                                                                          */
2982 /* Check if a packet is associated with an entry in the state table.        */
2983 /* ------------------------------------------------------------------------ */
2984 frentry_t *fr_checkstate(fin, passp)
2985 fr_info_t *fin;
2986 u_32_t *passp;
2987 {
2988 	ipstate_t *is;
2989 	frentry_t *fr;
2990 	tcphdr_t *tcp;
2991 	ipftq_t *ifq;
2992 	u_int pass;
2993 	ipf_stack_t *ifs = fin->fin_ifs;
2994 
2995 	if (ifs->ifs_fr_state_lock || (ifs->ifs_ips_list == NULL) ||
2996 	    (fin->fin_flx & (FI_SHORT|FI_STATE|FI_FRAGBODY|FI_BAD)))
2997 		return NULL;
2998 
2999 	is = NULL;
3000 	if ((fin->fin_flx & FI_TCPUDP) ||
3001 	    (fin->fin_fi.fi_p == IPPROTO_ICMP)
3002 #ifdef	USE_INET6
3003 	    || (fin->fin_fi.fi_p == IPPROTO_ICMPV6)
3004 #endif
3005 	    )
3006 		tcp = fin->fin_dp;
3007 	else
3008 		tcp = NULL;
3009 
3010 	/*
3011 	 * Search the hash table for matching packet header info.
3012 	 */
3013 	ifq = NULL;
3014 	is = fr_stlookup(fin, tcp, &ifq);
3015 	switch (fin->fin_p)
3016 	{
3017 #ifdef	USE_INET6
3018 	case IPPROTO_ICMPV6 :
3019 		if (is != NULL)
3020 			break;
3021 		if (fin->fin_v == 6) {
3022 			is = fr_checkicmp6matchingstate(fin);
3023 			if (is != NULL)
3024 				goto matched;
3025 		}
3026 		break;
3027 #endif
3028 	case IPPROTO_ICMP :
3029 		if (is != NULL)
3030 			break;
3031 		/*
3032 		 * No matching icmp state entry. Perhaps this is a
3033 		 * response to another state entry.
3034 		 */
3035 		is = fr_checkicmpmatchingstate(fin);
3036 		if (is != NULL)
3037 			goto matched;
3038 		break;
3039 	case IPPROTO_TCP :
3040 		if (is == NULL)
3041 			break;
3042 
3043 		if (is->is_pass & FR_NEWISN) {
3044 			if (fin->fin_out == 0)
3045 				fr_fixinisn(fin, is);
3046 			else if (fin->fin_out == 1)
3047 				fr_fixoutisn(fin, is);
3048 		}
3049 		break;
3050 	default :
3051 		if (fin->fin_rev)
3052 			ifq = &ifs->ifs_ips_udpacktq;
3053 		else
3054 			ifq = &ifs->ifs_ips_udptq;
3055 		break;
3056 	}
3057 	if (is == NULL) {
3058 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_miss);
3059 		return NULL;
3060 	}
3061 
3062 matched:
3063 	fr = is->is_rule;
3064 	if (fr != NULL) {
3065 		if ((fin->fin_out == 0) && (fr->fr_nattag.ipt_num[0] != 0)) {
3066 			if (fin->fin_nattag == NULL) {
3067 				RWLOCK_EXIT(&ifs->ifs_ipf_state);
3068 				return NULL;
3069 			}
3070 			if (fr_matchtag(&fr->fr_nattag, fin->fin_nattag) != 0) {
3071 				RWLOCK_EXIT(&ifs->ifs_ipf_state);
3072 				return NULL;
3073 			}
3074 		}
3075 		(void) strncpy(fin->fin_group, fr->fr_group, FR_GROUPLEN);
3076 		fin->fin_icode = fr->fr_icode;
3077 	}
3078 
3079 	fin->fin_rule = is->is_rulen;
3080 	pass = is->is_pass;
3081 	fr_updatestate(fin, is, ifq);
3082 
3083 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
3084 	fin->fin_flx |= FI_STATE;
3085 	if ((pass & FR_LOGFIRST) != 0)
3086 		pass &= ~(FR_LOGFIRST|FR_LOG);
3087 	*passp = pass;
3088 	return fr;
3089 }
3090 
3091 
3092 /* ------------------------------------------------------------------------ */
3093 /* Function:    fr_fixoutisn                                                */
3094 /* Returns:     Nil                                                         */
3095 /* Parameters:  fin(I)   - pointer to packet information                    */
3096 /*              is(I)  - pointer to master state structure                  */
3097 /*                                                                          */
3098 /* Called only for outbound packets, adjusts the sequence number and the    */
3099 /* TCP checksum to match that change.                                       */
3100 /* ------------------------------------------------------------------------ */
3101 static void fr_fixoutisn(fin, is)
3102 fr_info_t *fin;
3103 ipstate_t *is;
3104 {
3105 	tcphdr_t *tcp;
3106 	int rev;
3107 	u_32_t seq;
3108 
3109 	tcp = fin->fin_dp;
3110 	rev = fin->fin_rev;
3111 	if ((is->is_flags & IS_ISNSYN) != 0) {
3112 		if (rev == 0) {
3113 			seq = ntohl(tcp->th_seq);
3114 			seq += is->is_isninc[0];
3115 			tcp->th_seq = htonl(seq);
3116 			fix_outcksum(&tcp->th_sum, is->is_sumd[0]);
3117 		}
3118 	}
3119 	if ((is->is_flags & IS_ISNACK) != 0) {
3120 		if (rev == 1) {
3121 			seq = ntohl(tcp->th_seq);
3122 			seq += is->is_isninc[1];
3123 			tcp->th_seq = htonl(seq);
3124 			fix_outcksum(&tcp->th_sum, is->is_sumd[1]);
3125 		}
3126 	}
3127 }
3128 
3129 
3130 /* ------------------------------------------------------------------------ */
3131 /* Function:    fr_fixinisn                                                 */
3132 /* Returns:     Nil                                                         */
3133 /* Parameters:  fin(I)   - pointer to packet information                    */
3134 /*              is(I)  - pointer to master state structure                  */
3135 /*                                                                          */
3136 /* Called only for inbound packets, adjusts the acknowledge number and the  */
3137 /* TCP checksum to match that change.                                       */
3138 /* ------------------------------------------------------------------------ */
3139 static void fr_fixinisn(fin, is)
3140 fr_info_t *fin;
3141 ipstate_t *is;
3142 {
3143 	tcphdr_t *tcp;
3144 	int rev;
3145 	u_32_t ack;
3146 
3147 	tcp = fin->fin_dp;
3148 	rev = fin->fin_rev;
3149 	if ((is->is_flags & IS_ISNSYN) != 0) {
3150 		if (rev == 1) {
3151 			ack = ntohl(tcp->th_ack);
3152 			ack -= is->is_isninc[0];
3153 			tcp->th_ack = htonl(ack);
3154 			fix_incksum(&tcp->th_sum, is->is_sumd[0]);
3155 		}
3156 	}
3157 	if ((is->is_flags & IS_ISNACK) != 0) {
3158 		if (rev == 0) {
3159 			ack = ntohl(tcp->th_ack);
3160 			ack -= is->is_isninc[1];
3161 			tcp->th_ack = htonl(ack);
3162 			fix_incksum(&tcp->th_sum, is->is_sumd[1]);
3163 		}
3164 	}
3165 }
3166 
3167 
3168 /* ------------------------------------------------------------------------ */
3169 /* Function:    fr_statesync                                                */
3170 /* Returns:     Nil                                                         */
3171 /* Parameters:  action(I) - type of synchronisation to do                   */
3172 /*              v(I)      - IP version being sync'd (v4 or v6)              */
3173 /*              ifp(I)    - interface identifier associated with action     */
3174 /*              name(I)   - name associated with ifp parameter              */
3175 /*                                                                          */
3176 /* Walk through all state entries and if an interface pointer match is      */
3177 /* found then look it up again, based on its name in case the pointer has   */
3178 /* changed since last time.                                                 */
3179 /*                                                                          */
3180 /* If ifp is passed in as being non-null then we are only doing updates for */
3181 /* existing, matching, uses of it.                                          */
3182 /* ------------------------------------------------------------------------ */
3183 void fr_statesync(action, v, ifp, name, ifs)
3184 int action, v;
3185 void *ifp;
3186 char *name;
3187 ipf_stack_t *ifs;
3188 {
3189 	ipstate_t *is;
3190 	int i;
3191 
3192 	if (ifs->ifs_fr_running <= 0)
3193 		return;
3194 
3195 	WRITE_ENTER(&ifs->ifs_ipf_state);
3196 
3197 	if (ifs->ifs_fr_running <= 0) {
3198 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
3199 		return;
3200 	}
3201 
3202 	switch (action)
3203 	{
3204 	case IPFSYNC_RESYNC :
3205 		for (is = ifs->ifs_ips_list; is; is = is->is_next) {
3206 			if (v != 0 && is->is_v != v)
3207 				continue;
3208 			/*
3209 			 * Look up all the interface names in the state entry.
3210 			 */
3211 			for (i = 0; i < 4; i++) {
3212 				is->is_ifp[i] = fr_resolvenic(is->is_ifname[i],
3213 							      is->is_v, ifs);
3214 			}
3215 		}
3216 		break;
3217 	case IPFSYNC_NEWIFP :
3218 		for (is = ifs->ifs_ips_list; is; is = is->is_next) {
3219 			if (v != 0 && is->is_v != v)
3220 				continue;
3221 			/*
3222 			 * Look up all the interface names in the state entry.
3223 			 */
3224 			for (i = 0; i < 4; i++) {
3225 				if (!strncmp(is->is_ifname[i], name,
3226 					     sizeof(is->is_ifname[i])))
3227 					is->is_ifp[i] = ifp;
3228 			}
3229 		}
3230 		break;
3231 	case IPFSYNC_OLDIFP :
3232 		for (is = ifs->ifs_ips_list; is; is = is->is_next) {
3233 			if (v != 0 && is->is_v != v)
3234 				continue;
3235 			/*
3236 			 * Look up all the interface names in the state entry.
3237 			 */
3238 			for (i = 0; i < 4; i++) {
3239 				if (is->is_ifp[i] == ifp)
3240 					is->is_ifp[i] = (void *)-1;
3241 			}
3242 		}
3243 		break;
3244 	}
3245 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
3246 }
3247 
3248 
3249 #if SOLARIS2 >= 10
3250 /* ------------------------------------------------------------------------ */
3251 /* Function:    fr_stateifindexsync					    */
3252 /* Returns:     void							    */
3253 /* Parameters:	ifp	- current network interface descriptor (ifindex)    */
3254 /*              newifp	- new interface descriptor (new ifindex)	    */
3255 /*		ifs	- pointer to IPF stack				    */
3256 /*									    */
3257 /* Write Locks: assumes ipf_mutex is locked				    */
3258 /*                                                                          */
3259 /* Updates all interface indeces matching to ifp with new interface index   */
3260 /* value.								    */
3261 /* ------------------------------------------------------------------------ */
3262 void fr_stateifindexsync(ifp, newifp, ifs)
3263 void *ifp;
3264 void *newifp;
3265 ipf_stack_t *ifs;
3266 {
3267 	ipstate_t *is;
3268 	int i;
3269 
3270 	WRITE_ENTER(&ifs->ifs_ipf_state);
3271 
3272 	for (is = ifs->ifs_ips_list; is != NULL; is = is->is_next) {
3273 
3274 		for (i = 0; i < 4; i++) {
3275 			if (is->is_ifp[i] == ifp)
3276 				is->is_ifp[i] = newifp;
3277 		}
3278 	}
3279 
3280 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
3281 }
3282 #endif
3283 
3284 /* ------------------------------------------------------------------------ */
3285 /* Function:    fr_delstate                                                 */
3286 /* Returns:     int - 0 = entry deleted, else ref count on entry            */
3287 /* Parameters:  is(I)  - pointer to state structure to delete               */
3288 /*              why(I) - if not 0, log reason why it was deleted            */
3289 /*              ifs    - ipf stack instance                                 */
3290 /* Write Locks: ipf_state/ipf_global                                        */
3291 /*                                                                          */
3292 /* Deletes a state entry from the enumerated list as well as the hash table */
3293 /* and timeout queue lists.  Make adjustments to hash table statistics and  */
3294 /* global counters as required.                                             */
3295 /* ------------------------------------------------------------------------ */
3296 int fr_delstate(is, why, ifs)
3297 ipstate_t *is;
3298 int why;
3299 ipf_stack_t *ifs;
3300 {
3301 	int removed = 0;
3302 
3303 	ASSERT(rw_write_held(&ifs->ifs_ipf_global.ipf_lk) == 0 ||
3304 		rw_write_held(&ifs->ifs_ipf_state.ipf_lk) == 0);
3305 
3306 	/*
3307 	 * Start by removing the entry from the hash table of state entries
3308 	 * so it will not be "used" again.
3309 	 *
3310 	 * It will remain in the "list" of state entries until all references
3311 	 * have been accounted for.
3312 	 */
3313 	if (is->is_phnext != NULL) {
3314 		removed = 1;
3315 		*is->is_phnext = is->is_hnext;
3316 		if (is->is_hnext != NULL)
3317 			is->is_hnext->is_phnext = is->is_phnext;
3318 		if (ifs->ifs_ips_table[is->is_hv] == NULL)
3319 			ifs->ifs_ips_stats.iss_inuse--;
3320 		ifs->ifs_ips_stats.iss_bucketlen[is->is_hv]--;
3321 
3322 		is->is_phnext = NULL;
3323 		is->is_hnext = NULL;
3324 	}
3325 
3326 	/*
3327 	 * Because ifs->ifs_ips_stats.iss_wild is a count of entries in the state
3328 	 * table that have wildcard flags set, only decerement it once
3329 	 * and do it here.
3330 	 */
3331 	if (is->is_flags & (SI_WILDP|SI_WILDA)) {
3332 		if (!(is->is_flags & SI_CLONED)) {
3333 			ATOMIC_DECL(ifs->ifs_ips_stats.iss_wild);
3334 		}
3335 		is->is_flags &= ~(SI_WILDP|SI_WILDA);
3336 	}
3337 
3338 	/*
3339 	 * Next, remove it from the timeout queue it is in.
3340 	 */
3341 	fr_deletequeueentry(&is->is_sti);
3342 
3343 	is->is_me = NULL;
3344 
3345 	/*
3346 	 * If it is still in use by something else, do not go any further,
3347 	 * but note that at this point it is now an orphan.
3348 	 */
3349 	MUTEX_ENTER(&is->is_lock);
3350 	if (is->is_ref > 1) {
3351 		is->is_ref--;
3352 		MUTEX_EXIT(&is->is_lock);
3353 		if (removed)
3354 			ifs->ifs_ips_stats.iss_orphans++;
3355 		return (is->is_ref);
3356 	}
3357 	MUTEX_EXIT(&is->is_lock);
3358 
3359 	is->is_ref = 0;
3360 
3361 	/*
3362 	 * If entry has already been removed from table,
3363 	 * it means we're simply cleaning up an orphan.
3364 	 */
3365 	if (!removed)
3366 		ifs->ifs_ips_stats.iss_orphans--;
3367 
3368 	if (is->is_tqehead[0] != NULL)
3369 		(void) fr_deletetimeoutqueue(is->is_tqehead[0]);
3370 
3371 	if (is->is_tqehead[1] != NULL)
3372 		(void) fr_deletetimeoutqueue(is->is_tqehead[1]);
3373 
3374 #ifdef	IPFILTER_SYNC
3375 	if (is->is_sync)
3376 		ipfsync_del(is->is_sync);
3377 #endif
3378 #ifdef	IPFILTER_SCAN
3379 	(void) ipsc_detachis(is);
3380 #endif
3381 
3382 	/*
3383 	 * Now remove it from master list of state table entries.
3384 	 */
3385 	if (is->is_pnext != NULL) {
3386 		*is->is_pnext = is->is_next;
3387 		if (is->is_next != NULL) {
3388 			is->is_next->is_pnext = is->is_pnext;
3389 			is->is_next = NULL;
3390 		}
3391 		is->is_pnext = NULL;
3392 	}
3393 
3394 	if (ifs->ifs_ipstate_logging != 0 && why != 0)
3395 		ipstate_log(is, why, ifs);
3396 
3397 	if (is->is_rule != NULL) {
3398 		is->is_rule->fr_statecnt--;
3399 		(void)fr_derefrule(&is->is_rule, ifs);
3400 	}
3401 
3402 	MUTEX_DESTROY(&is->is_lock);
3403 	KFREE(is);
3404 	ifs->ifs_ips_num--;
3405 
3406 	return (0);
3407 }
3408 
3409 
3410 /* ------------------------------------------------------------------------ */
3411 /* Function:    fr_timeoutstate                                             */
3412 /* Returns:     Nil                                                         */
3413 /* Parameters:  ifs - ipf stack instance                                    */
3414 /*                                                                          */
3415 /* Slowly expire held state for thingslike UDP and ICMP.  The algorithm     */
3416 /* used here is to keep the queue sorted with the oldest things at the top  */
3417 /* and the youngest at the bottom.  So if the top one doesn't need to be    */
3418 /* expired then neither will any under it.                                  */
3419 /* ------------------------------------------------------------------------ */
3420 void fr_timeoutstate(ifs)
3421 ipf_stack_t *ifs;
3422 {
3423 	ipftq_t *ifq, *ifqnext;
3424 	ipftqent_t *tqe, *tqn;
3425 	ipstate_t *is;
3426 	SPL_INT(s);
3427 
3428 	SPL_NET(s);
3429 	WRITE_ENTER(&ifs->ifs_ipf_state);
3430 	for (ifq = ifs->ifs_ips_tqtqb; ifq != NULL; ifq = ifq->ifq_next)
3431 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); ) {
3432 			if (tqe->tqe_die > ifs->ifs_fr_ticks)
3433 				break;
3434 			tqn = tqe->tqe_next;
3435 			is = tqe->tqe_parent;
3436 			(void) fr_delstate(is, ISL_EXPIRE, ifs);
3437 		}
3438 
3439 	for (ifq = ifs->ifs_ips_utqe; ifq != NULL; ifq = ifq->ifq_next) {
3440 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); ) {
3441 			if (tqe->tqe_die > ifs->ifs_fr_ticks)
3442 				break;
3443 			tqn = tqe->tqe_next;
3444 			is = tqe->tqe_parent;
3445 			(void) fr_delstate(is, ISL_EXPIRE, ifs);
3446 		}
3447 	}
3448 
3449 	for (ifq = ifs->ifs_ips_utqe; ifq != NULL; ifq = ifqnext) {
3450 		ifqnext = ifq->ifq_next;
3451 
3452 		if (((ifq->ifq_flags & IFQF_DELETE) != 0) &&
3453 		    (ifq->ifq_ref == 0)) {
3454 			fr_freetimeoutqueue(ifq, ifs);
3455 		}
3456 	}
3457 
3458 	if (ifs->ifs_fr_state_doflush) {
3459 		(void) fr_state_flush(FLUSH_TABLE_EXTRA, 0, ifs);
3460 		ifs->ifs_fr_state_doflush = 0;
3461 	}
3462 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
3463 	SPL_X(s);
3464 }
3465 
3466 
3467 /* ---------------------------------------------------------------------- */
3468 /* Function:    fr_state_flush                                            */
3469 /* Returns:     int - 0 == success, -1 == failure                         */
3470 /* Parameters:  flush_option - how to flush the active State table	  */
3471 /*              proto    - IP version to flush (4, 6, or both)            */
3472 /*              ifs      - ipf stack instance                             */
3473 /* Write Locks: ipf_state                                                 */
3474 /*                                                                        */
3475 /* Flush state tables.  Three possible flush options currently defined:	  */
3476 /*                                                                        */
3477 /* FLUSH_TABLE_ALL	: Flush all state table entries			  */
3478 /*                                                                        */
3479 /* FLUSH_TABLE_CLOSING	: Flush entries with TCP connections which	  */
3480 /*			  have started to close on both ends using	  */
3481 /*			  ipf_flushclosing().				  */
3482 /*                                                                        */
3483 /* FLUSH_TABLE_EXTRA	: First, flush entries which are "almost" closed. */
3484 /*			  Then, if needed, flush entries with TCP	  */
3485 /*			  connections which have been idle for a long	  */
3486 /*			  time with ipf_extraflush().			  */
3487 /* ---------------------------------------------------------------------- */
3488 static int fr_state_flush(flush_option, proto, ifs)
3489 int flush_option, proto;
3490 ipf_stack_t *ifs;
3491 {
3492 	ipstate_t *is, *isn;
3493 	int removed;
3494 	SPL_INT(s);
3495 
3496 	removed = 0;
3497 
3498 	SPL_NET(s);
3499 	switch (flush_option)
3500 	{
3501 	case FLUSH_TABLE_ALL:
3502 		isn = ifs->ifs_ips_list;
3503 		while ((is = isn) != NULL) {
3504 			isn = is->is_next;
3505 			if ((proto != 0) && (is->is_v != proto))
3506 				continue;
3507 			if (fr_delstate(is, ISL_FLUSH, ifs) == 0)
3508 				removed++;
3509 		}
3510 		break;
3511 
3512 	case FLUSH_TABLE_CLOSING:
3513 		removed = ipf_flushclosing(STATE_FLUSH,
3514 					   IPF_TCPS_CLOSE_WAIT,
3515 					   ifs->ifs_ips_tqtqb,
3516 					   ifs->ifs_ips_utqe,
3517 					   ifs);
3518 		break;
3519 
3520 	case FLUSH_TABLE_EXTRA:
3521 		removed = ipf_flushclosing(STATE_FLUSH,
3522 					   IPF_TCPS_FIN_WAIT_2,
3523 					   ifs->ifs_ips_tqtqb,
3524 					   ifs->ifs_ips_utqe,
3525 					   ifs);
3526 
3527 		/*
3528 		 * Be sure we haven't done this in the last 10 seconds.
3529 		 */
3530 		if (ifs->ifs_fr_ticks - ifs->ifs_ips_last_force_flush <
3531 		    IPF_TTLVAL(10))
3532 			break;
3533 		ifs->ifs_ips_last_force_flush = ifs->ifs_fr_ticks;
3534                 removed += ipf_extraflush(STATE_FLUSH,
3535 					  &ifs->ifs_ips_tqtqb[IPF_TCPS_ESTABLISHED],
3536 					  ifs->ifs_ips_utqe,
3537 					  ifs);
3538 		break;
3539 
3540 	default: /* Flush Nothing */
3541 		break;
3542 	}
3543 
3544 	SPL_X(s);
3545 	return (removed);
3546 }
3547 
3548 
3549 /* ------------------------------------------------------------------------ */
3550 /* Function:    fr_tcp_age                                                  */
3551 /* Returns:     int - 1 == state transition made, 0 == no change (rejected) */
3552 /* Parameters:  tq(I)    - pointer to timeout queue information             */
3553 /*              fin(I)   - pointer to packet information                    */
3554 /*              tqtab(I) - TCP timeout queue table this is in               */
3555 /*              flags(I) - flags from state/NAT entry                       */
3556 /*                                                                          */
3557 /* Rewritten by Arjan de Vet <Arjan.deVet@adv.iae.nl>, 2000-07-29:          */
3558 /*                                                                          */
3559 /* - (try to) base state transitions on real evidence only,                 */
3560 /*   i.e. packets that are sent and have been received by ipfilter;         */
3561 /*   diagram 18.12 of TCP/IP volume 1 by W. Richard Stevens was used.       */
3562 /*                                                                          */
3563 /* - deal with half-closed connections correctly;                           */
3564 /*                                                                          */
3565 /* - store the state of the source in state[0] such that ipfstat            */
3566 /*   displays the state as source/dest instead of dest/source; the calls    */
3567 /*   to fr_tcp_age have been changed accordingly.                           */
3568 /*                                                                          */
3569 /* Internal Parameters:                                                     */
3570 /*                                                                          */
3571 /*    state[0] = state of source (host that initiated connection)           */
3572 /*    state[1] = state of dest   (host that accepted the connection)        */
3573 /*                                                                          */
3574 /*    dir == 0 : a packet from source to dest                               */
3575 /*    dir == 1 : a packet from dest to source                               */
3576 /*                                                                          */
3577 /* Locking: it is assumed that the parent of the tqe structure is locked.   */
3578 /* ------------------------------------------------------------------------ */
3579 int fr_tcp_age(tqe, fin, tqtab, flags)
3580 ipftqent_t *tqe;
3581 fr_info_t *fin;
3582 ipftq_t *tqtab;
3583 int flags;
3584 {
3585 	int dlen, ostate, nstate, rval, dir;
3586 	u_char tcpflags;
3587 	tcphdr_t *tcp;
3588 	ipf_stack_t *ifs = fin->fin_ifs;
3589 
3590 	tcp = fin->fin_dp;
3591 
3592 	rval = 0;
3593 	dir = fin->fin_rev;
3594 	tcpflags = tcp->th_flags;
3595 	dlen = fin->fin_dlen - (TCP_OFF(tcp) << 2);
3596 
3597 	ostate = tqe->tqe_state[1 - dir];
3598 	nstate = tqe->tqe_state[dir];
3599 
3600 	DTRACE_PROBE4(
3601 		indata,
3602 		fr_info_t *, fin,
3603 		int, ostate,
3604 		int, nstate,
3605 		u_char, tcpflags
3606 	);
3607 
3608 	if (tcpflags & TH_RST) {
3609 		if (!(tcpflags & TH_PUSH) && !dlen)
3610 			nstate = IPF_TCPS_CLOSED;
3611 		else
3612 			nstate = IPF_TCPS_CLOSE_WAIT;
3613 
3614 		/*
3615 		 * Once RST is received, we must advance peer's state to
3616 		 * CLOSE_WAIT.
3617 		 */
3618 		if (ostate <= IPF_TCPS_ESTABLISHED) {
3619 			tqe->tqe_state[1 - dir] = IPF_TCPS_CLOSE_WAIT;
3620 		}
3621 		rval = 1;
3622 	} else {
3623 
3624 		switch (nstate)
3625 		{
3626 		case IPF_TCPS_LISTEN: /* 0 */
3627 			if ((tcpflags & TH_OPENING) == TH_OPENING) {
3628 				/*
3629 				 * 'dir' received an S and sends SA in
3630 				 * response, CLOSED -> SYN_RECEIVED
3631 				 */
3632 				nstate = IPF_TCPS_SYN_RECEIVED;
3633 				rval = 1;
3634 			} else if ((tcpflags & TH_OPENING) == TH_SYN) {
3635 				/* 'dir' sent S, CLOSED -> SYN_SENT */
3636 				nstate = IPF_TCPS_SYN_SENT;
3637 				rval = 1;
3638 			}
3639 			/*
3640 			 * the next piece of code makes it possible to get
3641 			 * already established connections into the state table
3642 			 * after a restart or reload of the filter rules; this
3643 			 * does not work when a strict 'flags S keep state' is
3644 			 * used for tcp connections of course
3645 			 */
3646 			if (((flags & IS_TCPFSM) == 0) &&
3647 			    ((tcpflags & TH_ACKMASK) == TH_ACK)) {
3648 				/*
3649 				 * we saw an A, guess 'dir' is in ESTABLISHED
3650 				 * mode
3651 				 */
3652 				switch (ostate)
3653 				{
3654 				case IPF_TCPS_LISTEN :
3655 				case IPF_TCPS_SYN_RECEIVED :
3656 					nstate = IPF_TCPS_HALF_ESTAB;
3657 					rval = 1;
3658 					break;
3659 				case IPF_TCPS_HALF_ESTAB :
3660 				case IPF_TCPS_ESTABLISHED :
3661 					nstate = IPF_TCPS_ESTABLISHED;
3662 					rval = 1;
3663 					break;
3664 				default :
3665 					break;
3666 				}
3667 			}
3668 			/*
3669 			 * TODO: besides regular ACK packets we can have other
3670 			 * packets as well; it is yet to be determined how we
3671 			 * should initialize the states in those cases
3672 			 */
3673 			break;
3674 
3675 		case IPF_TCPS_SYN_SENT: /* 1 */
3676 			if ((tcpflags & ~(TH_ECN|TH_CWR)) == TH_SYN) {
3677 				/*
3678 				 * A retransmitted SYN packet.  We do not reset
3679 				 * the timeout here to fr_tcptimeout because a
3680 				 * connection connect timeout does not renew
3681 				 * after every packet that is sent.  We need to
3682 				 * set rval so as to indicate the packet has
3683 				 * passed the check for its flags being valid
3684 				 * in the TCP FSM.  Setting rval to 2 has the
3685 				 * result of not resetting the timeout.
3686 				 */
3687 				rval = 2;
3688 			} else if ((tcpflags & (TH_SYN|TH_FIN|TH_ACK)) ==
3689 				   TH_ACK) {
3690 				/*
3691 				 * we see an A from 'dir' which is in SYN_SENT
3692 				 * state: 'dir' sent an A in response to an SA
3693 				 * which it received, SYN_SENT -> ESTABLISHED
3694 				 */
3695 				nstate = IPF_TCPS_ESTABLISHED;
3696 				rval = 1;
3697 			} else if (tcpflags & TH_FIN) {
3698 				/*
3699 				 * we see an F from 'dir' which is in SYN_SENT
3700 				 * state and wants to close its side of the
3701 				 * connection; SYN_SENT -> FIN_WAIT_1
3702 				 */
3703 				nstate = IPF_TCPS_FIN_WAIT_1;
3704 				rval = 1;
3705 			} else if ((tcpflags & TH_OPENING) == TH_OPENING) {
3706 				/*
3707 				 * we see an SA from 'dir' which is already in
3708 				 * SYN_SENT state, this means we have a
3709 				 * simultaneous open; SYN_SENT -> SYN_RECEIVED
3710 				 */
3711 				nstate = IPF_TCPS_SYN_RECEIVED;
3712 				rval = 1;
3713 			}
3714 			break;
3715 
3716 		case IPF_TCPS_SYN_RECEIVED: /* 2 */
3717 			if ((tcpflags & (TH_SYN|TH_FIN|TH_ACK)) == TH_ACK) {
3718 				/*
3719 				 * we see an A from 'dir' which was in
3720 				 * SYN_RECEIVED state so it must now be in
3721 				 * established state, SYN_RECEIVED ->
3722 				 * ESTABLISHED
3723 				 */
3724 				nstate = IPF_TCPS_ESTABLISHED;
3725 				rval = 1;
3726 			} else if ((tcpflags & ~(TH_ECN|TH_CWR)) ==
3727 				   TH_OPENING) {
3728 				/*
3729 				 * We see an SA from 'dir' which is already in
3730 				 * SYN_RECEIVED state.
3731 				 */
3732 				rval = 2;
3733 			} else if (tcpflags & TH_FIN) {
3734 				/*
3735 				 * we see an F from 'dir' which is in
3736 				 * SYN_RECEIVED state and wants to close its
3737 				 * side of the connection; SYN_RECEIVED ->
3738 				 * FIN_WAIT_1
3739 				 */
3740 				nstate = IPF_TCPS_FIN_WAIT_1;
3741 				rval = 1;
3742 			}
3743 			break;
3744 
3745 		case IPF_TCPS_HALF_ESTAB: /* 3 */
3746 			if (tcpflags & TH_FIN) {
3747 				nstate = IPF_TCPS_FIN_WAIT_1;
3748 				rval = 1;
3749 			} else if ((tcpflags & TH_ACKMASK) == TH_ACK) {
3750 				/*
3751 				 * If we've picked up a connection in mid
3752 				 * flight, we could be looking at a follow on
3753 				 * packet from the same direction as the one
3754 				 * that created this state.  Recognise it but
3755 				 * do not advance the entire connection's
3756 				 * state.
3757 				 */
3758 				switch (ostate)
3759 				{
3760 				case IPF_TCPS_LISTEN :
3761 				case IPF_TCPS_SYN_SENT :
3762 				case IPF_TCPS_SYN_RECEIVED :
3763 					rval = 1;
3764 					break;
3765 				case IPF_TCPS_HALF_ESTAB :
3766 				case IPF_TCPS_ESTABLISHED :
3767 					nstate = IPF_TCPS_ESTABLISHED;
3768 					rval = 1;
3769 					break;
3770 				default :
3771 					break;
3772 				}
3773 			}
3774 			break;
3775 
3776 		case IPF_TCPS_ESTABLISHED: /* 4 */
3777 			rval = 1;
3778 			if (tcpflags & TH_FIN) {
3779 				/*
3780 				 * 'dir' closed its side of the connection;
3781 				 * this gives us a half-closed connection;
3782 				 * ESTABLISHED -> FIN_WAIT_1
3783 				 */
3784 				if (ostate == IPF_TCPS_FIN_WAIT_1) {
3785 					nstate = IPF_TCPS_CLOSING;
3786 				} else {
3787 					nstate = IPF_TCPS_FIN_WAIT_1;
3788 				}
3789 			} else if (tcpflags & TH_ACK) {
3790 				/*
3791 				 * an ACK, should we exclude other flags here?
3792 				 */
3793 				if (ostate == IPF_TCPS_FIN_WAIT_1) {
3794 					/*
3795 					 * We know the other side did an active
3796 					 * close, so we are ACKing the recvd
3797 					 * FIN packet (does the window matching
3798 					 * code guarantee this?) and go into
3799 					 * CLOSE_WAIT state; this gives us a
3800 					 * half-closed connection
3801 					 */
3802 					nstate = IPF_TCPS_CLOSE_WAIT;
3803 				} else if (ostate < IPF_TCPS_CLOSE_WAIT) {
3804 					/*
3805 					 * still a fully established
3806 					 * connection reset timeout
3807 					 */
3808 					nstate = IPF_TCPS_ESTABLISHED;
3809 				}
3810 			}
3811 			break;
3812 
3813 		case IPF_TCPS_CLOSE_WAIT: /* 5 */
3814 			rval = 1;
3815 			if (tcpflags & TH_FIN) {
3816 				/*
3817 				 * application closed and 'dir' sent a FIN,
3818 				 * we're now going into LAST_ACK state
3819 				 */
3820 				nstate = IPF_TCPS_LAST_ACK;
3821 			} else {
3822 				/*
3823 				 * we remain in CLOSE_WAIT because the other
3824 				 * side has closed already and we did not
3825 				 * close our side yet; reset timeout
3826 				 */
3827 				nstate = IPF_TCPS_CLOSE_WAIT;
3828 			}
3829 			break;
3830 
3831 		case IPF_TCPS_FIN_WAIT_1: /* 6 */
3832 			rval = 1;
3833 			if ((tcpflags & TH_ACK) &&
3834 			    ostate > IPF_TCPS_CLOSE_WAIT) {
3835 				/*
3836 				 * if the other side is not active anymore
3837 				 * it has sent us a FIN packet that we are
3838 				 * ack'ing now with an ACK; this means both
3839 				 * sides have now closed the connection and
3840 				 * we go into LAST_ACK
3841 				 */
3842 				/*
3843 				 * XXX: how do we know we really are ACKing
3844 				 * the FIN packet here? does the window code
3845 				 * guarantee that?
3846 				 */
3847 				nstate = IPF_TCPS_LAST_ACK;
3848 			} else {
3849 				/*
3850 				 * we closed our side of the connection
3851 				 * already but the other side is still active
3852 				 * (ESTABLISHED/CLOSE_WAIT); continue with
3853 				 * this half-closed connection
3854 				 */
3855 				nstate = IPF_TCPS_FIN_WAIT_1;
3856 			}
3857 			break;
3858 
3859 		case IPF_TCPS_CLOSING: /* 7 */
3860 			if ((tcpflags & (TH_FIN|TH_ACK)) == TH_ACK) {
3861 				nstate = IPF_TCPS_TIME_WAIT;
3862 			}
3863 			rval = 1;
3864 			break;
3865 
3866 		case IPF_TCPS_LAST_ACK: /* 8 */
3867 			/*
3868 			 * We want to reset timer here to keep state in table.
3869 			 * If we would allow the state to time out here, while
3870 			 * there would still be packets being retransmitted, we
3871 			 * would cut off line between the two peers preventing
3872 			 * them to close connection properly.
3873 			 */
3874 			rval = 1;
3875 			break;
3876 
3877 		case IPF_TCPS_FIN_WAIT_2: /* 9 */
3878 			/* NOT USED */
3879 			break;
3880 
3881 		case IPF_TCPS_TIME_WAIT: /* 10 */
3882 			/* we're in 2MSL timeout now */
3883 			if (ostate == IPF_TCPS_LAST_ACK) {
3884 				nstate = IPF_TCPS_CLOSED;
3885 				rval = 1;
3886 			} else {
3887 				rval = 2;
3888 			}
3889 			break;
3890 
3891 		case IPF_TCPS_CLOSED: /* 11 */
3892 			rval = 2;
3893 			break;
3894 
3895 		default :
3896 #if defined(_KERNEL)
3897 			ASSERT(nstate >= IPF_TCPS_LISTEN &&
3898 			    nstate <= IPF_TCPS_CLOSED);
3899 #else
3900 			abort();
3901 #endif
3902 			break;
3903 		}
3904 	}
3905 
3906 	/*
3907 	 * If rval == 2 then do not update the queue position, but treat the
3908 	 * packet as being ok.
3909 	 */
3910 	if (rval == 2) {
3911 		DTRACE_PROBE1(state_keeping_timer, int, nstate);
3912 		rval = 1;
3913 	}
3914 	else if (rval == 1) {
3915 		tqe->tqe_state[dir] = nstate;
3916 		/*
3917 		 * The nstate can either advance to a new state, or remain
3918 		 * unchanged, resetting the timer by moving to the bottom of
3919 		 * the queue.
3920 		 */
3921 		DTRACE_PROBE1(state_done, int, nstate);
3922 
3923 		if ((tqe->tqe_flags & TQE_RULEBASED) == 0)
3924 			fr_movequeue(tqe, tqe->tqe_ifq, tqtab + nstate, ifs);
3925 	}
3926 
3927 	return rval;
3928 }
3929 
3930 
3931 /* ------------------------------------------------------------------------ */
3932 /* Function:    ipstate_log                                                 */
3933 /* Returns:     Nil                                                         */
3934 /* Parameters:  is(I)   - pointer to state structure                        */
3935 /*              type(I) - type of log entry to create                       */
3936 /*                                                                          */
3937 /* Creates a state table log entry using the state structure and type info. */
3938 /* passed in.  Log packet/byte counts, source/destination address and other */
3939 /* protocol specific information.                                           */
3940 /* ------------------------------------------------------------------------ */
3941 void ipstate_log(is, type, ifs)
3942 struct ipstate *is;
3943 u_int type;
3944 ipf_stack_t *ifs;
3945 {
3946 #ifdef	IPFILTER_LOG
3947 	struct	ipslog	ipsl;
3948 	size_t sizes[1];
3949 	void *items[1];
3950 	int types[1];
3951 
3952 	/*
3953 	 * Copy information out of the ipstate_t structure and into the
3954 	 * structure used for logging.
3955 	 */
3956 	ipsl.isl_type = type;
3957 	ipsl.isl_pkts[0] = is->is_pkts[0] + is->is_icmppkts[0];
3958 	ipsl.isl_bytes[0] = is->is_bytes[0];
3959 	ipsl.isl_pkts[1] = is->is_pkts[1] + is->is_icmppkts[1];
3960 	ipsl.isl_bytes[1] = is->is_bytes[1];
3961 	ipsl.isl_pkts[2] = is->is_pkts[2] + is->is_icmppkts[2];
3962 	ipsl.isl_bytes[2] = is->is_bytes[2];
3963 	ipsl.isl_pkts[3] = is->is_pkts[3] + is->is_icmppkts[3];
3964 	ipsl.isl_bytes[3] = is->is_bytes[3];
3965 	ipsl.isl_src = is->is_src;
3966 	ipsl.isl_dst = is->is_dst;
3967 	ipsl.isl_p = is->is_p;
3968 	ipsl.isl_v = is->is_v;
3969 	ipsl.isl_flags = is->is_flags;
3970 	ipsl.isl_tag = is->is_tag;
3971 	ipsl.isl_rulen = is->is_rulen;
3972 	(void) strncpy(ipsl.isl_group, is->is_group, FR_GROUPLEN);
3973 
3974 	if (ipsl.isl_p == IPPROTO_TCP || ipsl.isl_p == IPPROTO_UDP) {
3975 		ipsl.isl_sport = is->is_sport;
3976 		ipsl.isl_dport = is->is_dport;
3977 		if (ipsl.isl_p == IPPROTO_TCP) {
3978 			ipsl.isl_state[0] = is->is_state[0];
3979 			ipsl.isl_state[1] = is->is_state[1];
3980 		}
3981 	} else if (ipsl.isl_p == IPPROTO_ICMP) {
3982 		ipsl.isl_itype = is->is_icmp.ici_type;
3983 	} else if (ipsl.isl_p == IPPROTO_ICMPV6) {
3984 		ipsl.isl_itype = is->is_icmp.ici_type;
3985 	} else {
3986 		ipsl.isl_ps.isl_filler[0] = 0;
3987 		ipsl.isl_ps.isl_filler[1] = 0;
3988 	}
3989 
3990 	items[0] = &ipsl;
3991 	sizes[0] = sizeof(ipsl);
3992 	types[0] = 0;
3993 
3994 	if (ipllog(IPL_LOGSTATE, NULL, items, sizes, types, 1, ifs)) {
3995 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_logged);
3996 	} else {
3997 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_logfail);
3998 	}
3999 #endif
4000 }
4001 
4002 
4003 #ifdef	USE_INET6
4004 /* ------------------------------------------------------------------------ */
4005 /* Function:    fr_checkicmp6matchingstate                                  */
4006 /* Returns:     ipstate_t* - NULL == no match found,                        */
4007 /*                           else  pointer to matching state entry          */
4008 /* Parameters:  fin(I) - pointer to packet information                      */
4009 /* Locks:       NULL == no locks, else Read Lock on ipf_state               */
4010 /*                                                                          */
4011 /* If we've got an ICMPv6 error message, using the information stored in    */
4012 /* the ICMPv6 packet, look for a matching state table entry.                */
4013 /* ------------------------------------------------------------------------ */
4014 static ipstate_t *fr_checkicmp6matchingstate(fin)
4015 fr_info_t *fin;
4016 {
4017 	struct icmp6_hdr *ic6, *oic;
4018 	int backward, i;
4019 	ipstate_t *is, **isp;
4020 	u_short sport, dport;
4021 	i6addr_t dst, src;
4022 	u_short savelen;
4023 	icmpinfo_t *ic;
4024 	fr_info_t ofin;
4025 	tcphdr_t *tcp;
4026 	ip6_t *oip6;
4027 	u_char	pr;
4028 	u_int hv;
4029 	ipf_stack_t *ifs = fin->fin_ifs;
4030 
4031 	/*
4032 	 * Does it at least have the return (basic) IP header ?
4033 	 * Is it an actual recognised ICMP error type?
4034 	 * Only a basic IP header (no options) should be with
4035 	 * an ICMP error header.
4036 	 */
4037 	if ((fin->fin_v != 6) || (fin->fin_plen < ICMP6ERR_MINPKTLEN) ||
4038 	    !(fin->fin_flx & FI_ICMPERR))
4039 		return NULL;
4040 
4041 	ic6 = fin->fin_dp;
4042 
4043 	oip6 = (ip6_t *)((char *)ic6 + ICMPERR_ICMPHLEN);
4044 	if (fin->fin_plen < sizeof(*oip6))
4045 		return NULL;
4046 
4047 	bcopy((char *)fin, (char *)&ofin, sizeof(*fin));
4048 	ofin.fin_v = 6;
4049 	ofin.fin_ifp = fin->fin_ifp;
4050 	ofin.fin_out = !fin->fin_out;
4051 	ofin.fin_m = NULL;	/* if dereferenced, panic XXX */
4052 	ofin.fin_mp = NULL;	/* if dereferenced, panic XXX */
4053 
4054 	/*
4055 	 * We make a fin entry to be able to feed it to
4056 	 * matchsrcdst. Note that not all fields are necessary
4057 	 * but this is the cleanest way. Note further we fill
4058 	 * in fin_mp such that if someone uses it we'll get
4059 	 * a kernel panic. fr_matchsrcdst does not use this.
4060 	 *
4061 	 * watch out here, as ip is in host order and oip6 in network
4062 	 * order. Any change we make must be undone afterwards.
4063 	 */
4064 	savelen = oip6->ip6_plen;
4065 	oip6->ip6_plen = fin->fin_dlen - ICMPERR_ICMPHLEN;
4066 	ofin.fin_flx = FI_NOCKSUM;
4067 	ofin.fin_ip = (ip_t *)oip6;
4068 	ofin.fin_plen = oip6->ip6_plen;
4069 	(void) fr_makefrip(sizeof(*oip6), (ip_t *)oip6, &ofin);
4070 	ofin.fin_flx &= ~(FI_BAD|FI_SHORT);
4071 	oip6->ip6_plen = savelen;
4072 
4073 	if (oip6->ip6_nxt == IPPROTO_ICMPV6) {
4074 		oic = (struct icmp6_hdr *)(oip6 + 1);
4075 		/*
4076 		 * an ICMP error can only be generated as a result of an
4077 		 * ICMP query, not as the response on an ICMP error
4078 		 *
4079 		 * XXX theoretically ICMP_ECHOREP and the other reply's are
4080 		 * ICMP query's as well, but adding them here seems strange XXX
4081 		 */
4082 		 if (!(oic->icmp6_type & ICMP6_INFOMSG_MASK))
4083 		    	return NULL;
4084 
4085 		/*
4086 		 * perform a lookup of the ICMP packet in the state table
4087 		 */
4088 		hv = (pr = oip6->ip6_nxt);
4089 		src.in6 = oip6->ip6_src;
4090 		hv += src.in4.s_addr;
4091 		dst.in6 = oip6->ip6_dst;
4092 		hv += dst.in4.s_addr;
4093 		hv += oic->icmp6_id;
4094 		hv += oic->icmp6_seq;
4095 		hv = DOUBLE_HASH(hv, ifs);
4096 
4097 		READ_ENTER(&ifs->ifs_ipf_state);
4098 		for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
4099 			ic = &is->is_icmp;
4100 			isp = &is->is_hnext;
4101 			if ((is->is_p == pr) &&
4102 			    !(is->is_pass & FR_NOICMPERR) &&
4103 			    (oic->icmp6_id == ic->ici_id) &&
4104 			    (oic->icmp6_seq == ic->ici_seq) &&
4105 			    (is = fr_matchsrcdst(&ofin, is, &src,
4106 						 &dst, NULL, FI_ICMPCMP))) {
4107 			    	/*
4108 			    	 * in the state table ICMP query's are stored
4109 			    	 * with the type of the corresponding ICMP
4110 			    	 * response. Correct here
4111 			    	 */
4112 				if (((ic->ici_type == ICMP6_ECHO_REPLY) &&
4113 				     (oic->icmp6_type == ICMP6_ECHO_REQUEST)) ||
4114 				     (ic->ici_type - 1 == oic->icmp6_type )) {
4115 				    	ifs->ifs_ips_stats.iss_hits++;
4116 					backward = IP6_NEQ(&is->is_dst, &src);
4117 					fin->fin_rev = !backward;
4118 					i = (backward << 1) + fin->fin_out;
4119     					is->is_icmppkts[i]++;
4120 					return is;
4121 				}
4122 			}
4123 		}
4124 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
4125 		return NULL;
4126 	}
4127 
4128 	hv = (pr = oip6->ip6_nxt);
4129 	src.in6 = oip6->ip6_src;
4130 	hv += src.i6[0];
4131 	hv += src.i6[1];
4132 	hv += src.i6[2];
4133 	hv += src.i6[3];
4134 	dst.in6 = oip6->ip6_dst;
4135 	hv += dst.i6[0];
4136 	hv += dst.i6[1];
4137 	hv += dst.i6[2];
4138 	hv += dst.i6[3];
4139 
4140 	if ((oip6->ip6_nxt == IPPROTO_TCP) || (oip6->ip6_nxt == IPPROTO_UDP)) {
4141 		tcp = (tcphdr_t *)(oip6 + 1);
4142 		dport = tcp->th_dport;
4143 		sport = tcp->th_sport;
4144 		hv += dport;
4145 		hv += sport;
4146 	} else
4147 		tcp = NULL;
4148 	hv = DOUBLE_HASH(hv, ifs);
4149 
4150 	READ_ENTER(&ifs->ifs_ipf_state);
4151 	for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
4152 		isp = &is->is_hnext;
4153 		/*
4154 		 * Only allow this icmp though if the
4155 		 * encapsulated packet was allowed through the
4156 		 * other way around. Note that the minimal amount
4157 		 * of info present does not allow for checking against
4158 		 * tcp internals such as seq and ack numbers.
4159 		 */
4160 		if ((is->is_p != pr) || (is->is_v != 6) ||
4161 		    (is->is_pass & FR_NOICMPERR))
4162 			continue;
4163 		is = fr_matchsrcdst(&ofin, is, &src, &dst, tcp, FI_ICMPCMP);
4164 		if (is != NULL) {
4165 			ifs->ifs_ips_stats.iss_hits++;
4166 			backward = IP6_NEQ(&is->is_dst, &src);
4167 			fin->fin_rev = !backward;
4168 			i = (backward << 1) + fin->fin_out;
4169 			is->is_icmppkts[i]++;
4170 			/*
4171 			 * we deliberately do not touch the timeouts
4172 			 * for the accompanying state table entry.
4173 			 * It remains to be seen if that is correct. XXX
4174 			 */
4175 			return is;
4176 		}
4177 	}
4178 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
4179 	return NULL;
4180 }
4181 #endif
4182 
4183 
4184 /* ------------------------------------------------------------------------ */
4185 /* Function:    fr_sttab_init                                               */
4186 /* Returns:     Nil                                                         */
4187 /* Parameters:  tqp(I) - pointer to an array of timeout queues for TCP      */
4188 /*                                                                          */
4189 /* Initialise the array of timeout queues for TCP.                          */
4190 /* ------------------------------------------------------------------------ */
4191 void fr_sttab_init(tqp, ifs)
4192 ipftq_t *tqp;
4193 ipf_stack_t *ifs;
4194 {
4195 	int i;
4196 
4197 	for (i = IPF_TCP_NSTATES - 1; i >= 0; i--) {
4198 		tqp[i].ifq_ttl = 0;
4199 		tqp[i].ifq_ref = 1;
4200 		tqp[i].ifq_head = NULL;
4201 		tqp[i].ifq_tail = &tqp[i].ifq_head;
4202 		tqp[i].ifq_next = tqp + i + 1;
4203 		MUTEX_INIT(&tqp[i].ifq_lock, "ipftq tcp tab");
4204 	}
4205 	tqp[IPF_TCP_NSTATES - 1].ifq_next = NULL;
4206 	tqp[IPF_TCPS_CLOSED].ifq_ttl = ifs->ifs_fr_tcpclosed;
4207 	tqp[IPF_TCPS_LISTEN].ifq_ttl = ifs->ifs_fr_tcptimeout;
4208 	tqp[IPF_TCPS_SYN_SENT].ifq_ttl = ifs->ifs_fr_tcptimeout;
4209 	tqp[IPF_TCPS_SYN_RECEIVED].ifq_ttl = ifs->ifs_fr_tcptimeout;
4210 	tqp[IPF_TCPS_ESTABLISHED].ifq_ttl = ifs->ifs_fr_tcpidletimeout;
4211 	tqp[IPF_TCPS_CLOSE_WAIT].ifq_ttl = ifs->ifs_fr_tcphalfclosed;
4212 	tqp[IPF_TCPS_FIN_WAIT_1].ifq_ttl = ifs->ifs_fr_tcphalfclosed;
4213 	tqp[IPF_TCPS_CLOSING].ifq_ttl = ifs->ifs_fr_tcptimeout;
4214 	tqp[IPF_TCPS_LAST_ACK].ifq_ttl = ifs->ifs_fr_tcplastack;
4215 	tqp[IPF_TCPS_FIN_WAIT_2].ifq_ttl = ifs->ifs_fr_tcpclosewait;
4216 	tqp[IPF_TCPS_TIME_WAIT].ifq_ttl = ifs->ifs_fr_tcptimeout;
4217 	tqp[IPF_TCPS_HALF_ESTAB].ifq_ttl = ifs->ifs_fr_tcptimeout;
4218 }
4219 
4220 
4221 /* ------------------------------------------------------------------------ */
4222 /* Function:    fr_sttab_destroy                                            */
4223 /* Returns:     Nil                                                         */
4224 /* Parameters:  tqp(I) - pointer to an array of timeout queues for TCP      */
4225 /*                                                                          */
4226 /* Do whatever is necessary to "destroy" each of the entries in the array   */
4227 /* of timeout queues for TCP.                                               */
4228 /* ------------------------------------------------------------------------ */
4229 void fr_sttab_destroy(tqp)
4230 ipftq_t *tqp;
4231 {
4232 	int i;
4233 
4234 	for (i = IPF_TCP_NSTATES - 1; i >= 0; i--)
4235 		MUTEX_DESTROY(&tqp[i].ifq_lock);
4236 }
4237 
4238 
4239 /* ------------------------------------------------------------------------ */
4240 /* Function:    fr_statederef                                               */
4241 /* Returns:     Nil                                                         */
4242 /* Parameters:  isp(I) - pointer to pointer to state table entry            */
4243 /*              ifs - ipf stack instance                                    */
4244 /*                                                                          */
4245 /* Decrement the reference counter for this state table entry and free it   */
4246 /* if there are no more things using it.                                    */
4247 /*                                                                          */
4248 /* Internal parameters:                                                     */
4249 /*    state[0] = state of source (host that initiated connection)           */
4250 /*    state[1] = state of dest   (host that accepted the connection)        */
4251 /* ------------------------------------------------------------------------ */
4252 void fr_statederef(isp, ifs)
4253 ipstate_t **isp;
4254 ipf_stack_t *ifs;
4255 {
4256 	ipstate_t *is;
4257 
4258 	is = *isp;
4259 	*isp = NULL;
4260 
4261 	MUTEX_ENTER(&is->is_lock);
4262 	if (is->is_ref > 1) {
4263 		is->is_ref--;
4264 		MUTEX_EXIT(&is->is_lock);
4265 #ifndef	_KERNEL
4266 		if ((is->is_sti.tqe_state[0] > IPF_TCPS_ESTABLISHED) ||
4267 		   (is->is_sti.tqe_state[1] > IPF_TCPS_ESTABLISHED)) {
4268 			(void) fr_delstate(is, ISL_ORPHAN, ifs);
4269 		}
4270 #endif
4271 		return;
4272 	}
4273 	MUTEX_EXIT(&is->is_lock);
4274 
4275 	WRITE_ENTER(&ifs->ifs_ipf_state);
4276 	(void) fr_delstate(is, ISL_EXPIRE, ifs);
4277 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
4278 }
4279 
4280 
4281 /* ------------------------------------------------------------------------ */
4282 /* Function:    fr_setstatequeue                                            */
4283 /* Returns:     Nil                                                         */
4284 /* Parameters:  is(I) - pointer to state structure                          */
4285 /*              rev(I) - forward(0) or reverse(1) direction                 */
4286 /* Locks:       ipf_state (read or write)                                   */
4287 /*                                                                          */
4288 /* Put the state entry on its default queue entry, using rev as a helped in */
4289 /* determining which queue it should be placed on.                          */
4290 /* ------------------------------------------------------------------------ */
4291 void fr_setstatequeue(is, rev, ifs)
4292 ipstate_t *is;
4293 int rev;
4294 ipf_stack_t *ifs;
4295 {
4296 	ipftq_t *oifq, *nifq;
4297 
4298 
4299 	if ((is->is_sti.tqe_flags & TQE_RULEBASED) != 0)
4300 		nifq = is->is_tqehead[rev];
4301 	else
4302 		nifq = NULL;
4303 
4304 	if (nifq == NULL) {
4305 		switch (is->is_p)
4306 		{
4307 #ifdef USE_INET6
4308 		case IPPROTO_ICMPV6 :
4309 			if (rev == 1)
4310 				nifq = &ifs->ifs_ips_icmpacktq;
4311 			else
4312 				nifq = &ifs->ifs_ips_icmptq;
4313 			break;
4314 #endif
4315 		case IPPROTO_ICMP :
4316 			if (rev == 1)
4317 				nifq = &ifs->ifs_ips_icmpacktq;
4318 			else
4319 				nifq = &ifs->ifs_ips_icmptq;
4320 			break;
4321 		case IPPROTO_TCP :
4322 			nifq = ifs->ifs_ips_tqtqb + is->is_state[rev];
4323 			break;
4324 
4325 		case IPPROTO_UDP :
4326 			if (rev == 1)
4327 				nifq = &ifs->ifs_ips_udpacktq;
4328 			else
4329 				nifq = &ifs->ifs_ips_udptq;
4330 			break;
4331 
4332 		default :
4333 			nifq = &ifs->ifs_ips_iptq;
4334 			break;
4335 		}
4336 	}
4337 
4338 	oifq = is->is_sti.tqe_ifq;
4339 	/*
4340 	 * If it's currently on a timeout queue, move it from one queue to
4341 	 * another, else put it on the end of the newly determined queue.
4342 	 */
4343 	if (oifq != NULL)
4344 		fr_movequeue(&is->is_sti, oifq, nifq, ifs);
4345 	else
4346 		fr_queueappend(&is->is_sti, nifq, is, ifs);
4347 	return;
4348 }
4349 
4350 
4351 /* ------------------------------------------------------------------------ */
4352 /* Function:    fr_stateiter                                                */
4353 /* Returns:     int - 0 == success, else error                              */
4354 /* Parameters:  token(I) - pointer to ipftoken structure                    */
4355 /*              itp(I)   - pointer to ipfgeniter structure                  */
4356 /*                                                                          */
4357 /* This function handles the SIOCGENITER ioctl for the state tables and     */
4358 /* walks through the list of entries in the state table list (ips_list.)    */
4359 /* ------------------------------------------------------------------------ */
4360 static int fr_stateiter(token, itp, ifs)
4361 ipftoken_t *token;
4362 ipfgeniter_t *itp;
4363 ipf_stack_t *ifs;
4364 {
4365 	ipstate_t *is, *next, zero;
4366 	int error, count;
4367 	char *dst;
4368 
4369 	if (itp->igi_data == NULL)
4370 		return EFAULT;
4371 
4372 	if (itp->igi_nitems == 0)
4373 		return EINVAL;
4374 
4375 	if (itp->igi_type != IPFGENITER_STATE)
4376 		return EINVAL;
4377 
4378 	error = 0;
4379 
4380 	READ_ENTER(&ifs->ifs_ipf_state);
4381 
4382 	/*
4383 	 * Get "previous" entry from the token and find the next entry.
4384 	 */
4385 	is = token->ipt_data;
4386 	if (is == NULL) {
4387 		next = ifs->ifs_ips_list;
4388 	} else {
4389 		next = is->is_next;
4390 	}
4391 
4392 	dst = itp->igi_data;
4393 	for (count = itp->igi_nitems; count > 0; count--) {
4394 		/*
4395 		 * If we found an entry, add a reference to it and update the token.
4396 		 * Otherwise, zero out data to be returned and NULL out token.
4397 		 */
4398 		if (next != NULL) {
4399 			MUTEX_ENTER(&next->is_lock);
4400 			next->is_ref++;
4401 			MUTEX_EXIT(&next->is_lock);
4402 			token->ipt_data = next;
4403 		} else {
4404 			bzero(&zero, sizeof(zero));
4405 			next = &zero;
4406 			token->ipt_data = NULL;
4407 		}
4408 
4409 		/*
4410 		 * Safe to release lock now the we have a reference.
4411 		 */
4412 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
4413 
4414 		/*
4415 		 * Copy out data and clean up references and tokens.
4416 		 */
4417 		error = COPYOUT(next, dst, sizeof(*next));
4418 		if (error != 0)
4419 			error = EFAULT;
4420 		if (token->ipt_data == NULL) {
4421 			ipf_freetoken(token, ifs);
4422 			break;
4423 		} else {
4424 			if (is != NULL)
4425 				fr_statederef(&is, ifs);
4426 			if (next->is_next == NULL) {
4427 				ipf_freetoken(token, ifs);
4428 				break;
4429 			}
4430 		}
4431 
4432 		if ((count == 1) || (error != 0))
4433 			break;
4434 
4435 		READ_ENTER(&ifs->ifs_ipf_state);
4436 		dst += sizeof(*next);
4437 		is = next;
4438 		next = is->is_next;
4439 	}
4440 
4441 	return error;
4442 }
4443