xref: /illumos-gate/usr/src/uts/common/inet/ipf/ip_state.c (revision 55fea89d)
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  * Copyright (c) 2014, Joyent, Inc.  All rights reserved.
9  */
10 
11 #if defined(KERNEL) || defined(_KERNEL)
12 # undef KERNEL
13 # undef _KERNEL
14 # define        KERNEL	1
15 # define        _KERNEL	1
16 #endif
17 #include <sys/errno.h>
18 #include <sys/types.h>
19 #include <sys/param.h>
20 #include <sys/file.h>
21 #if defined(__NetBSD__) && (NetBSD >= 199905) && !defined(IPFILTER_LKM) && \
22     defined(_KERNEL)
23 # include "opt_ipfilter_log.h"
24 #endif
25 #if defined(_KERNEL) && defined(__FreeBSD_version) && \
26     (__FreeBSD_version >= 400000) && !defined(KLD_MODULE)
27 #include "opt_inet6.h"
28 #endif
29 #if !defined(_KERNEL) && !defined(__KERNEL__)
30 # include <stdio.h>
31 # include <stdlib.h>
32 # include <string.h>
33 # define _KERNEL
34 # ifdef __OpenBSD__
35 struct file;
36 # endif
37 # include <sys/uio.h>
38 # undef _KERNEL
39 #endif
40 #if defined(_KERNEL) && (__FreeBSD_version >= 220000)
41 # include <sys/filio.h>
42 # include <sys/fcntl.h>
43 # if (__FreeBSD_version >= 300000) && !defined(IPFILTER_LKM)
44 #  include "opt_ipfilter.h"
45 # endif
46 #else
47 # include <sys/ioctl.h>
48 #endif
49 #include <sys/time.h>
50 #if !defined(linux)
51 # include <sys/protosw.h>
52 #endif
53 #include <sys/socket.h>
54 #if defined(_KERNEL)
55 # include <sys/systm.h>
56 # if !defined(__SVR4) && !defined(__svr4__)
57 #  include <sys/mbuf.h>
58 # endif
59 #endif
60 #if defined(__SVR4) || defined(__svr4__)
61 # include <sys/filio.h>
62 # include <sys/byteorder.h>
63 # ifdef _KERNEL
64 #  include <sys/dditypes.h>
65 # endif
66 # include <sys/stream.h>
67 # include <sys/kmem.h>
68 #endif
69 
70 #include <net/if.h>
71 #ifdef sun
72 # include <net/af.h>
73 #endif
74 #include <net/route.h>
75 #include <netinet/in.h>
76 #include <netinet/in_systm.h>
77 #include <netinet/ip.h>
78 #include <netinet/tcp.h>
79 #if !defined(linux)
80 # include <netinet/ip_var.h>
81 #endif
82 #if !defined(__hpux) && !defined(linux)
83 # include <netinet/tcp_fsm.h>
84 #endif
85 #include <netinet/udp.h>
86 #include <netinet/ip_icmp.h>
87 #include "netinet/ip_compat.h"
88 #include <netinet/tcpip.h>
89 #include "netinet/ip_fil.h"
90 #include "netinet/ip_nat.h"
91 #include "netinet/ip_frag.h"
92 #include "netinet/ip_state.h"
93 #include "netinet/ip_proxy.h"
94 #include "netinet/ipf_stack.h"
95 #ifdef	IPFILTER_SYNC
96 #include "netinet/ip_sync.h"
97 #endif
98 #ifdef	IPFILTER_SCAN
99 #include "netinet/ip_scan.h"
100 #endif
101 #ifdef	USE_INET6
102 #include <netinet/icmp6.h>
103 #endif
104 #if (__FreeBSD_version >= 300000)
105 # include <sys/malloc.h>
106 # if defined(_KERNEL) && !defined(IPFILTER_LKM)
107 #  include <sys/libkern.h>
108 #  include <sys/systm.h>
109 # endif
110 #endif
111 /* END OF INCLUDES */
112 
113 
114 #if !defined(lint)
115 static const char sccsid[] = "@(#)ip_state.c	1.8 6/5/96 (C) 1993-2000 Darren Reed";
116 static const char rcsid[] = "@(#)$Id: ip_state.c,v 2.186.2.36 2005/08/11 19:58:03 darrenr Exp $";
117 #endif
118 
119 #ifdef	USE_INET6
120 static ipstate_t *fr_checkicmp6matchingstate __P((fr_info_t *));
121 #endif
122 static ipstate_t *fr_matchsrcdst __P((fr_info_t *, ipstate_t *, i6addr_t *,
123 				      i6addr_t *, tcphdr_t *, u_32_t));
124 static ipstate_t *fr_checkicmpmatchingstate __P((fr_info_t *));
125 static int fr_state_flush __P((int, int, ipf_stack_t *));
126 static ips_stat_t *fr_statetstats __P((ipf_stack_t *));
127 static int fr_state_remove __P((caddr_t, ipf_stack_t *));
128 static void fr_ipsmove __P((ipstate_t *, u_int, ipf_stack_t *));
129 static int fr_tcpstate __P((fr_info_t *, tcphdr_t *, ipstate_t *));
130 static int fr_tcpoptions __P((fr_info_t *, tcphdr_t *, tcpdata_t *));
131 static ipstate_t *fr_stclone __P((fr_info_t *, tcphdr_t *, ipstate_t *));
132 static void fr_fixinisn __P((fr_info_t *, ipstate_t *));
133 static void fr_fixoutisn __P((fr_info_t *, ipstate_t *));
134 static void fr_checknewisn __P((fr_info_t *, ipstate_t *));
135 static int fr_stateiter __P((ipftoken_t *, ipfgeniter_t *, ipf_stack_t *));
136 
137 int fr_stputent __P((caddr_t, ipf_stack_t *));
138 int fr_stgetent __P((caddr_t, ipf_stack_t *));
139 
140 #define	ONE_DAY		IPF_TTLVAL(1 * 86400)	/* 1 day */
141 #define	FIVE_DAYS	(5 * ONE_DAY)
142 #define	DOUBLE_HASH(x, ifs)	\
143     (((x) + ifs->ifs_ips_seed[(x) % ifs->ifs_fr_statesize]) % ifs->ifs_fr_statesize)
144 
145 
146 /* ------------------------------------------------------------------------ */
147 /* Function:    fr_stateinit                                                */
148 /* Returns:     int - 0 == success, -1 == failure                           */
149 /* Parameters:  ifs - ipf stack instance                                    */
150 /*                                                                          */
151 /* Initialise all the global variables used within the state code.          */
152 /* This action also includes initiailising locks.                           */
153 /* ------------------------------------------------------------------------ */
fr_stateinit(ifs)154 int fr_stateinit(ifs)
155 ipf_stack_t *ifs;
156 {
157 #if defined(NEED_LOCAL_RAND) || !defined(_KERNEL)
158 	struct timeval tv;
159 #endif
160 	int i;
161 
162 	KMALLOCS(ifs->ifs_ips_table, ipstate_t **,
163 		 ifs->ifs_fr_statesize * sizeof(ipstate_t *));
164 	if (ifs->ifs_ips_table == NULL)
165 		return -1;
166 	bzero((char *)ifs->ifs_ips_table,
167 	      ifs->ifs_fr_statesize * sizeof(ipstate_t *));
168 
169 	KMALLOCS(ifs->ifs_ips_seed, u_long *,
170 		 ifs->ifs_fr_statesize * sizeof(*ifs->ifs_ips_seed));
171 	if (ifs->ifs_ips_seed == NULL)
172 		return -2;
173 #if defined(NEED_LOCAL_RAND) || !defined(_KERNEL)
174 	tv.tv_sec = 0;
175 	GETKTIME(&tv);
176 #endif
177 	for (i = 0; i < ifs->ifs_fr_statesize; i++) {
178 		/*
179 		 * XXX - ips_seed[X] should be a random number of sorts.
180 		 */
181 #if !defined(NEED_LOCAL_RAND) && defined(_KERNEL)
182 		ifs->ifs_ips_seed[i] = ipf_random();
183 #else
184 		ifs->ifs_ips_seed[i] = ((u_long)ifs->ifs_ips_seed + i) *
185 		    ifs->ifs_fr_statesize;
186 		ifs->ifs_ips_seed[i] += tv.tv_sec;
187 		ifs->ifs_ips_seed[i] *= (u_long)ifs->ifs_ips_seed;
188 		ifs->ifs_ips_seed[i] ^= 0x5a5aa5a5;
189 		ifs->ifs_ips_seed[i] *= ifs->ifs_fr_statemax;
190 #endif
191 	}
192 
193 	/* fill icmp reply type table */
194 	for (i = 0; i <= ICMP_MAXTYPE; i++)
195 		icmpreplytype4[i] = -1;
196 	icmpreplytype4[ICMP_ECHO] = ICMP_ECHOREPLY;
197 	icmpreplytype4[ICMP_TSTAMP] = ICMP_TSTAMPREPLY;
198 	icmpreplytype4[ICMP_IREQ] = ICMP_IREQREPLY;
199 	icmpreplytype4[ICMP_MASKREQ] = ICMP_MASKREPLY;
200 #ifdef	USE_INET6
201 	/* fill icmp reply type table */
202 	for (i = 0; i <= ICMP6_MAXTYPE; i++)
203 		icmpreplytype6[i] = -1;
204 	icmpreplytype6[ICMP6_ECHO_REQUEST] = ICMP6_ECHO_REPLY;
205 	icmpreplytype6[ICMP6_MEMBERSHIP_QUERY] = ICMP6_MEMBERSHIP_REPORT;
206 	icmpreplytype6[ICMP6_NI_QUERY] = ICMP6_NI_REPLY;
207 	icmpreplytype6[ND_ROUTER_SOLICIT] = ND_ROUTER_ADVERT;
208 	icmpreplytype6[ND_NEIGHBOR_SOLICIT] = ND_NEIGHBOR_ADVERT;
209 #endif
210 
211 	KMALLOCS(ifs->ifs_ips_stats.iss_bucketlen, u_long *,
212 		 ifs->ifs_fr_statesize * sizeof(u_long));
213 	if (ifs->ifs_ips_stats.iss_bucketlen == NULL)
214 		return -1;
215 	bzero((char *)ifs->ifs_ips_stats.iss_bucketlen,
216 	      ifs->ifs_fr_statesize * sizeof(u_long));
217 
218 	if (ifs->ifs_fr_state_maxbucket == 0) {
219 		for (i = ifs->ifs_fr_statesize; i > 0; i >>= 1)
220 			ifs->ifs_fr_state_maxbucket++;
221 		ifs->ifs_fr_state_maxbucket *= 2;
222 	}
223 
224 	fr_sttab_init(ifs->ifs_ips_tqtqb, ifs);
225 	ifs->ifs_ips_tqtqb[IPF_TCP_NSTATES - 1].ifq_next = &ifs->ifs_ips_udptq;
226 	ifs->ifs_ips_udptq.ifq_ttl = (u_long)ifs->ifs_fr_udptimeout;
227 	ifs->ifs_ips_udptq.ifq_ref = 1;
228 	ifs->ifs_ips_udptq.ifq_head = NULL;
229 	ifs->ifs_ips_udptq.ifq_tail = &ifs->ifs_ips_udptq.ifq_head;
230 	MUTEX_INIT(&ifs->ifs_ips_udptq.ifq_lock, "ipftq udp tab");
231 	ifs->ifs_ips_udptq.ifq_next = &ifs->ifs_ips_udpacktq;
232 	ifs->ifs_ips_udpacktq.ifq_ttl = (u_long)ifs->ifs_fr_udpacktimeout;
233 	ifs->ifs_ips_udpacktq.ifq_ref = 1;
234 	ifs->ifs_ips_udpacktq.ifq_head = NULL;
235 	ifs->ifs_ips_udpacktq.ifq_tail = &ifs->ifs_ips_udpacktq.ifq_head;
236 	MUTEX_INIT(&ifs->ifs_ips_udpacktq.ifq_lock, "ipftq udpack tab");
237 	ifs->ifs_ips_udpacktq.ifq_next = &ifs->ifs_ips_icmptq;
238 	ifs->ifs_ips_icmptq.ifq_ttl = (u_long)ifs->ifs_fr_icmptimeout;
239 	ifs->ifs_ips_icmptq.ifq_ref = 1;
240 	ifs->ifs_ips_icmptq.ifq_head = NULL;
241 	ifs->ifs_ips_icmptq.ifq_tail = &ifs->ifs_ips_icmptq.ifq_head;
242 	MUTEX_INIT(&ifs->ifs_ips_icmptq.ifq_lock, "ipftq icmp tab");
243 	ifs->ifs_ips_icmptq.ifq_next = &ifs->ifs_ips_icmpacktq;
244 	ifs->ifs_ips_icmpacktq.ifq_ttl = (u_long)ifs->ifs_fr_icmpacktimeout;
245 	ifs->ifs_ips_icmpacktq.ifq_ref = 1;
246 	ifs->ifs_ips_icmpacktq.ifq_head = NULL;
247 	ifs->ifs_ips_icmpacktq.ifq_tail = &ifs->ifs_ips_icmpacktq.ifq_head;
248 	MUTEX_INIT(&ifs->ifs_ips_icmpacktq.ifq_lock, "ipftq icmpack tab");
249 	ifs->ifs_ips_icmpacktq.ifq_next = &ifs->ifs_ips_iptq;
250 	ifs->ifs_ips_iptq.ifq_ttl = (u_long)ifs->ifs_fr_iptimeout;
251 	ifs->ifs_ips_iptq.ifq_ref = 1;
252 	ifs->ifs_ips_iptq.ifq_head = NULL;
253 	ifs->ifs_ips_iptq.ifq_tail = &ifs->ifs_ips_iptq.ifq_head;
254 	MUTEX_INIT(&ifs->ifs_ips_iptq.ifq_lock, "ipftq ip tab");
255 	ifs->ifs_ips_iptq.ifq_next = &ifs->ifs_ips_deletetq;
256 	/* entry's ttl in deletetq is just 1 tick */
257 	ifs->ifs_ips_deletetq.ifq_ttl = (u_long) 1;
258 	ifs->ifs_ips_deletetq.ifq_ref = 1;
259 	ifs->ifs_ips_deletetq.ifq_head = NULL;
260 	ifs->ifs_ips_deletetq.ifq_tail = &ifs->ifs_ips_deletetq.ifq_head;
261 	MUTEX_INIT(&ifs->ifs_ips_deletetq.ifq_lock, "state delete queue");
262 	ifs->ifs_ips_deletetq.ifq_next = NULL;
263 
264 	RWLOCK_INIT(&ifs->ifs_ipf_state, "ipf IP state rwlock");
265 	MUTEX_INIT(&ifs->ifs_ipf_stinsert, "ipf state insert mutex");
266 	ifs->ifs_fr_state_init = 1;
267 
268 	ifs->ifs_ips_last_force_flush = ifs->ifs_fr_ticks;
269 	return 0;
270 }
271 
272 
273 /* ------------------------------------------------------------------------ */
274 /* Function:    fr_stateunload                                              */
275 /* Returns:     Nil                                                         */
276 /* Parameters:  ifs - ipf stack instance                                    */
277 /*                                                                          */
278 /* Release and destroy any resources acquired or initialised so that        */
279 /* IPFilter can be unloaded or re-initialised.                              */
280 /* ------------------------------------------------------------------------ */
fr_stateunload(ifs)281 void fr_stateunload(ifs)
282 ipf_stack_t *ifs;
283 {
284 	ipftq_t *ifq, *ifqnext;
285 	ipstate_t *is;
286 
287 	while ((is = ifs->ifs_ips_list) != NULL)
288 	    (void) fr_delstate(is, 0, ifs);
289 
290 	/*
291 	 * Proxy timeout queues are not cleaned here because although they
292 	 * exist on the state list, appr_unload is called after fr_stateunload
293 	 * and the proxies actually are responsible for them being created.
294 	 * Should the proxy timeouts have their own list?  There's no real
295 	 * justification as this is the only complicationA
296 	 */
297 	for (ifq = ifs->ifs_ips_utqe; ifq != NULL; ifq = ifqnext) {
298 		ifqnext = ifq->ifq_next;
299 		if (((ifq->ifq_flags & IFQF_PROXY) == 0) &&
300 		    (fr_deletetimeoutqueue(ifq) == 0))
301 			fr_freetimeoutqueue(ifq, ifs);
302 	}
303 
304 	ifs->ifs_ips_stats.iss_inuse = 0;
305 	ifs->ifs_ips_num = 0;
306 
307 	if (ifs->ifs_fr_state_init == 1) {
308 		fr_sttab_destroy(ifs->ifs_ips_tqtqb);
309 		MUTEX_DESTROY(&ifs->ifs_ips_udptq.ifq_lock);
310 		MUTEX_DESTROY(&ifs->ifs_ips_icmptq.ifq_lock);
311 		MUTEX_DESTROY(&ifs->ifs_ips_udpacktq.ifq_lock);
312 		MUTEX_DESTROY(&ifs->ifs_ips_icmpacktq.ifq_lock);
313 		MUTEX_DESTROY(&ifs->ifs_ips_iptq.ifq_lock);
314 		MUTEX_DESTROY(&ifs->ifs_ips_deletetq.ifq_lock);
315 	}
316 
317 	if (ifs->ifs_ips_table != NULL) {
318 		KFREES(ifs->ifs_ips_table,
319 		       ifs->ifs_fr_statesize * sizeof(*ifs->ifs_ips_table));
320 		ifs->ifs_ips_table = NULL;
321 	}
322 
323 	if (ifs->ifs_ips_seed != NULL) {
324 		KFREES(ifs->ifs_ips_seed,
325 		       ifs->ifs_fr_statesize * sizeof(*ifs->ifs_ips_seed));
326 		ifs->ifs_ips_seed = NULL;
327 	}
328 
329 	if (ifs->ifs_ips_stats.iss_bucketlen != NULL) {
330 		KFREES(ifs->ifs_ips_stats.iss_bucketlen,
331 		       ifs->ifs_fr_statesize * sizeof(u_long));
332 		ifs->ifs_ips_stats.iss_bucketlen = NULL;
333 	}
334 
335 	if (ifs->ifs_fr_state_maxbucket_reset == 1)
336 		ifs->ifs_fr_state_maxbucket = 0;
337 
338 	if (ifs->ifs_fr_state_init == 1) {
339 		ifs->ifs_fr_state_init = 0;
340 		RW_DESTROY(&ifs->ifs_ipf_state);
341 		MUTEX_DESTROY(&ifs->ifs_ipf_stinsert);
342 	}
343 }
344 
345 
346 /* ------------------------------------------------------------------------ */
347 /* Function:    fr_statetstats                                              */
348 /* Returns:     ips_state_t* - pointer to state stats structure             */
349 /* Parameters:  Nil                                                         */
350 /*                                                                          */
351 /* Put all the current numbers and pointers into a single struct and return */
352 /* a pointer to it.                                                         */
353 /* ------------------------------------------------------------------------ */
fr_statetstats(ifs)354 static ips_stat_t *fr_statetstats(ifs)
355 ipf_stack_t *ifs;
356 {
357 	ifs->ifs_ips_stats.iss_active = ifs->ifs_ips_num;
358 	ifs->ifs_ips_stats.iss_statesize = ifs->ifs_fr_statesize;
359 	ifs->ifs_ips_stats.iss_statemax = ifs->ifs_fr_statemax;
360 	ifs->ifs_ips_stats.iss_table = ifs->ifs_ips_table;
361 	ifs->ifs_ips_stats.iss_list = ifs->ifs_ips_list;
362 	ifs->ifs_ips_stats.iss_ticks = ifs->ifs_fr_ticks;
363 	return &ifs->ifs_ips_stats;
364 }
365 
366 /* ------------------------------------------------------------------------ */
367 /* Function:    fr_state_remove                                             */
368 /* Returns:     int - 0 == success, != 0 == failure                         */
369 /* Parameters:  data(I) - pointer to state structure to delete from table   */
370 /*              ifs - ipf stack instance                                    */
371 /*                                                                          */
372 /* Search for a state structure that matches the one passed, according to   */
373 /* the IP addresses and other protocol specific information.                */
374 /* ------------------------------------------------------------------------ */
fr_state_remove(data,ifs)375 static int fr_state_remove(data, ifs)
376 caddr_t data;
377 ipf_stack_t *ifs;
378 {
379 	ipstate_t *sp, st;
380 	int error;
381 
382 	sp = &st;
383 	error = fr_inobj(data, &st, IPFOBJ_IPSTATE);
384 	if (error)
385 		return EFAULT;
386 
387 	WRITE_ENTER(&ifs->ifs_ipf_state);
388 	for (sp = ifs->ifs_ips_list; sp; sp = sp->is_next)
389 		if ((sp->is_p == st.is_p) && (sp->is_v == st.is_v) &&
390 		    !bcmp((caddr_t)&sp->is_src, (caddr_t)&st.is_src,
391 			  sizeof(st.is_src)) &&
392 		    !bcmp((caddr_t)&sp->is_dst, (caddr_t)&st.is_dst,
393 			  sizeof(st.is_dst)) &&
394 		    !bcmp((caddr_t)&sp->is_ps, (caddr_t)&st.is_ps,
395 			  sizeof(st.is_ps))) {
396 			(void) fr_delstate(sp, ISL_REMOVE, ifs);
397 			RWLOCK_EXIT(&ifs->ifs_ipf_state);
398 			return 0;
399 		}
400 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
401 	return ESRCH;
402 }
403 
404 
405 /* ------------------------------------------------------------------------ */
406 /* Function:    fr_state_ioctl                                              */
407 /* Returns:     int - 0 == success, != 0 == failure                         */
408 /* Parameters:  data(I) - pointer to ioctl data                             */
409 /*              cmd(I)  - ioctl command integer                             */
410 /*              mode(I) - file mode bits used with open                     */
411 /*              uid(I)  - uid of caller                                     */
412 /*              ctx(I)  - pointer to give the uid context                   */
413 /*              ifs     - ipf stack instance                                */
414 /*                                                                          */
415 /* Processes an ioctl call made to operate on the IP Filter state device.   */
416 /* ------------------------------------------------------------------------ */
fr_state_ioctl(data,cmd,mode,uid,ctx,ifs)417 int fr_state_ioctl(data, cmd, mode, uid, ctx, ifs)
418 caddr_t data;
419 ioctlcmd_t cmd;
420 int mode, uid;
421 void *ctx;
422 ipf_stack_t *ifs;
423 {
424 	int arg, ret, error = 0;
425 
426 	switch (cmd)
427 	{
428 	/*
429 	 * Delete an entry from the state table.
430 	 */
431 	case SIOCDELST :
432 		error = fr_state_remove(data, ifs);
433 		break;
434 	/*
435 	 * Flush the state table
436 	 */
437 	case SIOCIPFFL :
438 		error = BCOPYIN(data, (char *)&arg, sizeof(arg));
439 		if (error != 0) {
440 			error = EFAULT;
441 		} else {
442 			if (VALID_TABLE_FLUSH_OPT(arg)) {
443 				WRITE_ENTER(&ifs->ifs_ipf_state);
444 				ret = fr_state_flush(arg, 4, ifs);
445 				RWLOCK_EXIT(&ifs->ifs_ipf_state);
446 				error = BCOPYOUT((char *)&ret, data,
447 						sizeof(ret));
448 				if (error != 0)
449 					return EFAULT;
450 			} else {
451 				error = EINVAL;
452 			}
453 		}
454 		break;
455 
456 #ifdef	USE_INET6
457 	case SIOCIPFL6 :
458 		error = BCOPYIN(data, (char *)&arg, sizeof(arg));
459 		if (error != 0) {
460 			error = EFAULT;
461 		} else {
462 			if (VALID_TABLE_FLUSH_OPT(arg)) {
463 				WRITE_ENTER(&ifs->ifs_ipf_state);
464 				ret = fr_state_flush(arg, 6, ifs);
465 				RWLOCK_EXIT(&ifs->ifs_ipf_state);
466 				error = BCOPYOUT((char *)&ret, data,
467 						sizeof(ret));
468 				if (error != 0)
469 					return EFAULT;
470 			} else {
471 				error = EINVAL;
472 			}
473 		}
474 		break;
475 #endif
476 #ifdef	IPFILTER_LOG
477 	/*
478 	 * Flush the state log.
479 	 */
480 	case SIOCIPFFB :
481 		if (!(mode & FWRITE))
482 			error = EPERM;
483 		else {
484 			int tmp;
485 
486 			tmp = ipflog_clear(IPL_LOGSTATE, ifs);
487 			error = BCOPYOUT((char *)&tmp, data, sizeof(tmp));
488 			if (error != 0)
489 				error = EFAULT;
490 		}
491 		break;
492 	/*
493 	 * Turn logging of state information on/off.
494 	 */
495 	case SIOCSETLG :
496 		if (!(mode & FWRITE)) {
497 			error = EPERM;
498 		} else {
499 			error = BCOPYIN((char *)data,
500 					(char *)&ifs->ifs_ipstate_logging,
501 					sizeof(ifs->ifs_ipstate_logging));
502 			if (error != 0)
503 				error = EFAULT;
504 		}
505 		break;
506 	/*
507 	 * Return the current state of logging.
508 	 */
509 	case SIOCGETLG :
510 		error = BCOPYOUT((char *)&ifs->ifs_ipstate_logging,
511 				(char *)data,
512 				sizeof(ifs->ifs_ipstate_logging));
513 		if (error != 0)
514 			error = EFAULT;
515 		break;
516 	/*
517 	 * Return the number of bytes currently waiting to be read.
518 	 */
519 	case FIONREAD :
520 		arg = ifs->ifs_iplused[IPL_LOGSTATE]; /* returned in an int */
521 		error = BCOPYOUT((char *)&arg, data, sizeof(arg));
522 		if (error != 0)
523 			error = EFAULT;
524 		break;
525 #endif
526 	/*
527 	 * Get the current state statistics.
528 	 */
529 	case SIOCGETFS :
530 		error = fr_outobj(data, fr_statetstats(ifs), IPFOBJ_STATESTAT);
531 		break;
532 	/*
533 	 * Lock/Unlock the state table.  (Locking prevents any changes, which
534 	 * means no packets match).
535 	 */
536 	case SIOCSTLCK :
537 		if (!(mode & FWRITE)) {
538 			error = EPERM;
539 		} else {
540 			error = fr_lock(data, &ifs->ifs_fr_state_lock);
541 		}
542 		break;
543 	/*
544 	 * Add an entry to the current state table.
545 	 */
546 	case SIOCSTPUT :
547 		if (!ifs->ifs_fr_state_lock || !(mode & FWRITE)) {
548 			error = EACCES;
549 			break;
550 		}
551 		error = fr_stputent(data, ifs);
552 		break;
553 	/*
554 	 * Get a state table entry.
555 	 */
556 	case SIOCSTGET :
557 		if (!ifs->ifs_fr_state_lock) {
558 			error = EACCES;
559 			break;
560 		}
561 		error = fr_stgetent(data, ifs);
562 		break;
563 
564 	case SIOCGENITER :
565 	    {
566 		ipftoken_t *token;
567 		ipfgeniter_t iter;
568 
569 		error = fr_inobj(data, &iter, IPFOBJ_GENITER);
570 		if (error != 0)
571 			break;
572 
573 		token = ipf_findtoken(IPFGENITER_STATE, uid, ctx, ifs);
574 		if (token != NULL)
575 			error = fr_stateiter(token, &iter, ifs);
576 		else
577 			error = ESRCH;
578 		RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
579 		break;
580 	    }
581 
582 	case SIOCIPFDELTOK :
583 		error = BCOPYIN(data, (char *)&arg, sizeof(arg));
584 		if (error != 0) {
585 			error = EFAULT;
586 		} else {
587 			error = ipf_deltoken(arg, uid, ctx, ifs);
588 		}
589 		break;
590 
591 	default :
592 		error = EINVAL;
593 		break;
594 	}
595 	return error;
596 }
597 
598 
599 /* ------------------------------------------------------------------------ */
600 /* Function:    fr_stgetent                                                 */
601 /* Returns:     int - 0 == success, != 0 == failure                         */
602 /* Parameters:  data(I) - pointer to state structure to retrieve from table */
603 /*                                                                          */
604 /* Copy out state information from the kernel to a user space process.  If  */
605 /* there is a filter rule associated with the state entry, copy that out    */
606 /* as well.  The entry to copy out is taken from the value of "ips_next" in */
607 /* the struct passed in and if not null and not found in the list of current*/
608 /* state entries, the retrieval fails.                                      */
609 /* ------------------------------------------------------------------------ */
fr_stgetent(data,ifs)610 int fr_stgetent(data, ifs)
611 caddr_t data;
612 ipf_stack_t *ifs;
613 {
614 	ipstate_t *is, *isn;
615 	ipstate_save_t ips;
616 	int error;
617 
618 	error = fr_inobj(data, &ips, IPFOBJ_STATESAVE);
619 	if (error)
620 		return EFAULT;
621 
622 	isn = ips.ips_next;
623 	if (isn == NULL) {
624 		isn = ifs->ifs_ips_list;
625 		if (isn == NULL) {
626 			if (ips.ips_next == NULL)
627 				return ENOENT;
628 			return 0;
629 		}
630 	} else {
631 		/*
632 		 * Make sure the pointer we're copying from exists in the
633 		 * current list of entries.  Security precaution to prevent
634 		 * copying of random kernel data.
635 		 */
636 		for (is = ifs->ifs_ips_list; is; is = is->is_next)
637 			if (is == isn)
638 				break;
639 		if (!is)
640 			return ESRCH;
641 	}
642 	ips.ips_next = isn->is_next;
643 	bcopy((char *)isn, (char *)&ips.ips_is, sizeof(ips.ips_is));
644 	ips.ips_rule = isn->is_rule;
645 	if (isn->is_rule != NULL)
646 		bcopy((char *)isn->is_rule, (char *)&ips.ips_fr,
647 		      sizeof(ips.ips_fr));
648 	error = fr_outobj(data, &ips, IPFOBJ_STATESAVE);
649 	if (error)
650 		return EFAULT;
651 	return 0;
652 }
653 
654 
655 /* ------------------------------------------------------------------------ */
656 /* Function:    fr_stputent                                                 */
657 /* Returns:     int - 0 == success, != 0 == failure                         */
658 /* Parameters:  data(I) - pointer to state information struct               */
659 /*              ifs     - ipf stack instance                                */
660 /*                                                                          */
661 /* This function implements the SIOCSTPUT ioctl: insert a state entry into  */
662 /* the state table.  If the state info. includes a pointer to a filter rule */
663 /* then also add in an orphaned rule (will not show up in any "ipfstat -io" */
664 /* output.                                                                  */
665 /* ------------------------------------------------------------------------ */
fr_stputent(data,ifs)666 int fr_stputent(data, ifs)
667 caddr_t data;
668 ipf_stack_t *ifs;
669 {
670 	ipstate_t *is, *isn;
671 	ipstate_save_t ips;
672 	int error, i;
673 	frentry_t *fr;
674 	char *name;
675 
676 	error = fr_inobj(data, &ips, IPFOBJ_STATESAVE);
677 	if (error)
678 		return EFAULT;
679 
680 	/*
681 	 * Trigger automatic call to fr_state_flush() if the
682 	 * table has reached capacity specified by hi watermark.
683 	 */
684 	if (ST_TAB_WATER_LEVEL(ifs) > ifs->ifs_state_flush_level_hi)
685 		ifs->ifs_fr_state_doflush = 1;
686 
687 	/*
688 	 * If automatic flushing did not do its job, and the table
689 	 * has filled up, don't try to create a new entry.
690 	 */
691 	if (ifs->ifs_ips_num >= ifs->ifs_fr_statemax) {
692 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_max);
693 		return ENOMEM;
694 	}
695 
696 	KMALLOC(isn, ipstate_t *);
697 	if (isn == NULL)
698 		return ENOMEM;
699 
700 	bcopy((char *)&ips.ips_is, (char *)isn, sizeof(*isn));
701 	bzero((char *)isn, offsetof(struct ipstate, is_pkts));
702 	isn->is_sti.tqe_pnext = NULL;
703 	isn->is_sti.tqe_next = NULL;
704 	isn->is_sti.tqe_ifq = NULL;
705 	isn->is_sti.tqe_parent = isn;
706 	isn->is_ifp[0] = NULL;
707 	isn->is_ifp[1] = NULL;
708 	isn->is_ifp[2] = NULL;
709 	isn->is_ifp[3] = NULL;
710 	isn->is_sync = NULL;
711 	fr = ips.ips_rule;
712 
713 	if (fr == NULL) {
714 		READ_ENTER(&ifs->ifs_ipf_state);
715 		fr_stinsert(isn, 0, ifs);
716 		MUTEX_EXIT(&isn->is_lock);
717 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
718 		return 0;
719 	}
720 
721 	if (isn->is_flags & SI_NEWFR) {
722 		KMALLOC(fr, frentry_t *);
723 		if (fr == NULL) {
724 			KFREE(isn);
725 			return ENOMEM;
726 		}
727 		bcopy((char *)&ips.ips_fr, (char *)fr, sizeof(*fr));
728 		isn->is_rule = fr;
729 		ips.ips_is.is_rule = fr;
730 		MUTEX_NUKE(&fr->fr_lock);
731 		MUTEX_INIT(&fr->fr_lock, "state filter rule lock");
732 
733 		/*
734 		 * Look up all the interface names in the rule.
735 		 */
736 		for (i = 0; i < 4; i++) {
737 			name = fr->fr_ifnames[i];
738 			fr->fr_ifas[i] = fr_resolvenic(name, fr->fr_v, ifs);
739 			name = isn->is_ifname[i];
740 			isn->is_ifp[i] = fr_resolvenic(name, isn->is_v, ifs);
741 		}
742 
743 		fr->fr_ref = 0;
744 		fr->fr_dsize = 0;
745 		fr->fr_data = NULL;
746 		fr->fr_type = FR_T_NONE;
747 
748 		fr_resolvedest(&fr->fr_tif, fr->fr_v, ifs);
749 		fr_resolvedest(&fr->fr_dif, fr->fr_v, ifs);
750 		fr_resolvedest(&fr->fr_rif, fr->fr_v, ifs);
751 
752 		/*
753 		 * send a copy back to userland of what we ended up
754 		 * to allow for verification.
755 		 */
756 		error = fr_outobj(data, &ips, IPFOBJ_STATESAVE);
757 		if (error) {
758 			KFREE(isn);
759 			MUTEX_DESTROY(&fr->fr_lock);
760 			KFREE(fr);
761 			return EFAULT;
762 		}
763 		READ_ENTER(&ifs->ifs_ipf_state);
764 		fr_stinsert(isn, 0, ifs);
765 		MUTEX_EXIT(&isn->is_lock);
766 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
767 
768 	} else {
769 		READ_ENTER(&ifs->ifs_ipf_state);
770 		for (is = ifs->ifs_ips_list; is; is = is->is_next)
771 			if (is->is_rule == fr) {
772 				fr_stinsert(isn, 0, ifs);
773 				MUTEX_EXIT(&isn->is_lock);
774 				break;
775 			}
776 
777 		if (is == NULL) {
778 			KFREE(isn);
779 			isn = NULL;
780 		}
781 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
782 
783 		return (isn == NULL) ? ESRCH : 0;
784 	}
785 
786 	return 0;
787 }
788 
789 
790 /* ------------------------------------------------------------------------ */
791 /* Function:   fr_stinsert                                                  */
792 /* Returns:    Nil                                                          */
793 /* Parameters: is(I)  - pointer to state structure                          */
794 /*             rev(I) - flag indicating forward/reverse direction of packet */
795 /*                                                                          */
796 /* Inserts a state structure into the hash table (for lookups) and the list */
797 /* of state entries (for enumeration).  Resolves all of the interface names */
798 /* to pointers and adjusts running stats for the hash table as appropriate. */
799 /*                                                                          */
800 /* Locking: it is assumed that some kind of lock on ipf_state is held.      */
801 /*          Exits with is_lock initialised and held.                        */
802 /* ------------------------------------------------------------------------ */
fr_stinsert(is,rev,ifs)803 void fr_stinsert(is, rev, ifs)
804 ipstate_t *is;
805 int rev;
806 ipf_stack_t *ifs;
807 {
808 	frentry_t *fr;
809 	u_int hv;
810 	int i;
811 
812 	MUTEX_INIT(&is->is_lock, "ipf state entry");
813 
814 	fr = is->is_rule;
815 	if (fr != NULL) {
816 		MUTEX_ENTER(&fr->fr_lock);
817 		fr->fr_ref++;
818 		fr->fr_statecnt++;
819 		MUTEX_EXIT(&fr->fr_lock);
820 	}
821 
822 	/*
823 	 * Look up all the interface names in the state entry.
824 	 */
825 	for (i = 0; i < 4; i++) {
826 		if (is->is_ifp[i] != NULL)
827 			continue;
828 		is->is_ifp[i] = fr_resolvenic(is->is_ifname[i], is->is_v, ifs);
829 	}
830 
831 	/*
832 	 * If we could trust is_hv, then the modulous would not be needed, but
833 	 * when running with IPFILTER_SYNC, this stops bad values.
834 	 */
835 	hv = is->is_hv % ifs->ifs_fr_statesize;
836 	is->is_hv = hv;
837 
838 	/*
839 	 * We need to get both of these locks...the first because it is
840 	 * possible that once the insert is complete another packet might
841 	 * come along, match the entry and want to update it.
842 	 */
843 	MUTEX_ENTER(&is->is_lock);
844 	MUTEX_ENTER(&ifs->ifs_ipf_stinsert);
845 
846 	/*
847 	 * add into list table.
848 	 */
849 	if (ifs->ifs_ips_list != NULL)
850 		ifs->ifs_ips_list->is_pnext = &is->is_next;
851 	is->is_pnext = &ifs->ifs_ips_list;
852 	is->is_next = ifs->ifs_ips_list;
853 	ifs->ifs_ips_list = is;
854 
855 	if (ifs->ifs_ips_table[hv] != NULL)
856 		ifs->ifs_ips_table[hv]->is_phnext = &is->is_hnext;
857 	else
858 		ifs->ifs_ips_stats.iss_inuse++;
859 	is->is_phnext = ifs->ifs_ips_table + hv;
860 	is->is_hnext = ifs->ifs_ips_table[hv];
861 	ifs->ifs_ips_table[hv] = is;
862 	ifs->ifs_ips_stats.iss_bucketlen[hv]++;
863 	ifs->ifs_ips_num++;
864 	MUTEX_EXIT(&ifs->ifs_ipf_stinsert);
865 
866 	fr_setstatequeue(is, rev, ifs);
867 }
868 
869 /* ------------------------------------------------------------------------ */
870 /* Function:	fr_match_ipv4addrs					    */
871 /* Returns:	int -	2 strong match (same addresses, same direction)	    */
872 /*			1 weak match (same address, opposite direction)	    */
873 /*			0 no match					    */
874 /*									    */
875 /* Function matches IPv4 addresses.					    */
876 /* ------------------------------------------------------------------------ */
fr_match_ipv4addrs(is1,is2)877 static int fr_match_ipv4addrs(is1, is2)
878 ipstate_t *is1;
879 ipstate_t *is2;
880 {
881 	int	rv;
882 
883 	if (is1->is_saddr == is2->is_saddr && is1->is_daddr == is2->is_daddr)
884 		rv = 2;
885 	else if (is1->is_saddr == is2->is_daddr &&
886 	    is1->is_daddr == is2->is_saddr)
887 		rv = 1;
888 	else
889 		rv = 0;
890 
891 	return (rv);
892 }
893 
894 /* ------------------------------------------------------------------------ */
895 /* Function:	fr_match_ipv6addrs					    */
896 /* Returns:	int - 	2 strong match (same addresses, same direction)	    */
897 /*			1 weak match (same addresses, opposite direction)   */
898 /*			0 no match					    */
899 /*									    */
900 /* Function matches IPv6 addresses.					    */
901 /* ------------------------------------------------------------------------ */
fr_match_ipv6addrs(is1,is2)902 static int fr_match_ipv6addrs(is1, is2)
903 ipstate_t *is1;
904 ipstate_t *is2;
905 {
906 	int	rv;
907 
908 	if (IP6_EQ(&is1->is_src, &is2->is_src) &&
909 	    IP6_EQ(&is1->is_dst, &is2->is_dst))
910 		rv = 2;
911 	else if (IP6_EQ(&is1->is_src, &is2->is_dst) &&
912 	    IP6_EQ(&is1->is_dst, &is2->is_src)) {
913 		rv = 1;
914 	}
915 	else
916 		rv = 0;
917 
918 	return (rv);
919 }
920 /* ------------------------------------------------------------------------ */
921 /* Function:	fr_match_addresses					    */
922 /* Returns:	int - 	2 strong match (same addresses, same direction)	    */
923 /*			1 weak match (same address, opposite directions)    */
924 /* 			0 no match					    */
925 /* Parameters:	is1, is2 pointers to states we are checking		    */
926 /*									    */
927 /* Matches addresses, function uses fr_match_ipvXaddrs() to deal with IPv4  */
928 /* and IPv6 address format.						    */
929 /* ------------------------------------------------------------------------ */
fr_match_addresses(is1,is2)930 static int fr_match_addresses(is1, is2)
931 ipstate_t *is1;
932 ipstate_t *is2;
933 {
934 	int	rv;
935 
936 	if (is1->is_v == 4) {
937 		rv = fr_match_ipv4addrs(is1, is2);
938 	} else {
939 		rv = fr_match_ipv6addrs(is1, is2);
940 	}
941 
942 	return (rv);
943 }
944 
945 /* ------------------------------------------------------------------------ */
946 /* Function:	fr_match_ppairs						    */
947 /* Returns:	int - 	2 strong match (same ports, same direction)	    */
948 /*			1 weak match (same ports, different direction)	    */
949 /*			0 no match					    */
950 /* Parameters	ppairs1, ppairs - src, dst ports we want to match.	    */
951 /*									    */
952 /* Matches two port_pair_t types (port pairs). Each port pair contains	    */
953 /* src, dst port, which belong to session (state entry).		    */
954 /* ------------------------------------------------------------------------ */
fr_match_ppairs(ppairs1,ppairs2)955 static int fr_match_ppairs(ppairs1, ppairs2)
956 port_pair_t *ppairs1;
957 port_pair_t *ppairs2;
958 {
959 	int	rv;
960 
961 	if (ppairs1->pp_sport == ppairs2->pp_sport &&
962 	    ppairs1->pp_dport == ppairs2->pp_dport)
963 		rv = 2;
964 	else if (ppairs1->pp_sport == ppairs2->pp_dport &&
965 		    ppairs1->pp_dport == ppairs2->pp_sport)
966 		rv = 1;
967 	else
968 		rv = 0;
969 
970 	return (rv);
971 }
972 
973 /* ------------------------------------------------------------------------ */
974 /* Function:	fr_match_l4_hdr						    */
975 /* Returns:	int -	0 no match,					    */
976 /*			1 weak match (same ports, different directions)	    */
977 /*			2 strong match (same ports, same direction)	    */
978 /* Parameters	is1, is2 - states we want to match			    */
979 /*									    */
980 /* Function matches L4 header data (source ports for TCP, UDP, CallIds for  */
981 /* GRE protocol).							    */
982 /* ------------------------------------------------------------------------ */
fr_match_l4_hdr(is1,is2)983 static int fr_match_l4_hdr(is1, is2)
984 ipstate_t *is1;
985 ipstate_t *is2;
986 {
987 	int	rv = 0;
988 	port_pair_t	pp1;
989 	port_pair_t	pp2;
990 
991 	if (is1->is_p != is2->is_p)
992 		return (0);
993 
994 	switch (is1->is_p) {
995 		case	IPPROTO_TCP:
996 			pp1.pp_sport = is1->is_ps.is_ts.ts_sport;
997 			pp1.pp_dport = is1->is_ps.is_ts.ts_dport;
998 			pp2.pp_sport = is2->is_ps.is_ts.ts_sport;
999 			pp2.pp_dport = is2->is_ps.is_ts.ts_dport;
1000 			rv = fr_match_ppairs(&pp1, &pp2);
1001 			break;
1002 		case	IPPROTO_UDP:
1003 			pp1.pp_sport = is1->is_ps.is_us.us_sport;
1004 			pp1.pp_dport = is1->is_ps.is_us.us_dport;
1005 			pp2.pp_sport = is2->is_ps.is_us.us_sport;
1006 			pp2.pp_dport = is2->is_ps.is_us.us_dport;
1007 			rv = fr_match_ppairs(&pp1, &pp2);
1008 			break;
1009 		case	IPPROTO_GRE:
1010 			/* greinfo_t can be also interprted as port pair */
1011 			pp1.pp_sport = is1->is_ps.is_ug.gs_call[0];
1012 			pp1.pp_dport = is1->is_ps.is_ug.gs_call[1];
1013 			pp2.pp_sport = is2->is_ps.is_ug.gs_call[0];
1014 			pp2.pp_dport = is2->is_ps.is_ug.gs_call[1];
1015 			rv = fr_match_ppairs(&pp1, &pp2);
1016 			break;
1017 		case	IPPROTO_ICMP:
1018 		case	IPPROTO_ICMPV6:
1019 			if (bcmp(&is1->is_ps, &is2->is_ps, sizeof (icmpinfo_t)))
1020 				rv = 1;
1021 			else
1022 				rv = 0;
1023 			break;
1024 		default:
1025 			rv = 0;
1026 	}
1027 
1028 	return (rv);
1029 }
1030 
1031 /* ------------------------------------------------------------------------ */
1032 /* Function:	fr_matchstates						    */
1033 /* Returns:	int - nonzero match, zero no match			    */
1034 /* Parameters	is1, is2 - states we want to match			    */
1035 /*									    */
1036 /* The state entries are equal (identical match) if they belong to the same */
1037 /* session. Any time new state entry is being added the fr_addstate()	    */
1038 /* function creates temporal state entry from the data it gets from IP and  */
1039 /* L4 header. The fr_matchstats() must be also aware of packet direction,   */
1040 /* which is also stored within the state entry. We should keep in mind the  */
1041 /* information about packet direction is spread accross L3 (addresses) and  */
1042 /* L4 (ports). There are three possible relationships betwee is1, is2:	    */
1043 /* 		- no match (match(is1, is2) == 0))			    */
1044 /*		- weak match same addresses (ports), but different	    */
1045 /*			directions (1)	(fr_match_xxxx(is1, is2) == 1)	    */
1046 /*		- strong match same addresses (ports) and same directions   */
1047 /*			 (2) (fr_match_xxxx(is1, is2) == 2)		    */
1048 /*									    */
1049 /* There are functions, which match match addresses (L3 header) in is1, is2 */
1050 /* and functions, which are used to compare ports (L4 header) data. We say  */
1051 /* the is1 and is2 are same (identical) if there is a match		    */
1052 /* (fr_match_l4_hdr(is1, is2) != 0) and matchlevels are same for entries    */
1053 /* (fr_match_l3_hdr(is1, is2) == fr_match_l4_hdr(is1, is2)) for is1, is2.   */
1054 /* Such requirement deals with case as follows:				    */
1055 /*	suppose there are two connections between hosts A, B. Connection 1: */
1056 /*			a.a.a.a:12345 <=> b.b.b.b:54321			    */
1057 /*		Connection 2:						    */
1058 /*			a.a.a.a:54321 <=> b.b.b.b:12345			    */
1059 /* since we've introduced match levels into our fr_matchstates(), we are    */
1060 /* able to identify, which packets belong to connection A and which belong  */
1061 /* to connection B.	Assume there are two entries is1, is2. is1 has been */
1062 /* from con. 1 packet, which travelled from A to B:			    */
1063 /*			a.a.a.a:12345 -> b.b.b.b:54321			    */
1064 /* while s2, has been created from packet which belongs to con. 2 and is    */
1065 /* also coming from A to B:						    */
1066 /*			a.a.a.a:54321 -> b.b.b.b:12345			    */
1067 /* fr_match_l3_hdr(is1, is2) == 2 -> strong match, while		    */
1068 /* fr_match_l4_hdr(is1, is2) == 1 -> weak match. Since match levels are	    */
1069 /* different the state entries are not identical -> no match as a final	    */
1070 /* result.								    */
1071 /* ------------------------------------------------------------------------ */
fr_matchstates(is1,is2)1072 static int fr_matchstates(is1, is2)
1073 ipstate_t *is1;
1074 ipstate_t *is2;
1075 {
1076 	int	rv;
1077 	int	amatch;
1078 	int	pmatch;
1079 
1080 	if (bcmp(&is1->is_pass, &is2->is_pass,
1081 		offsetof(struct ipstate, is_ps) -
1082 		offsetof(struct ipstate, is_pass)) == 0) {
1083 
1084 		pmatch = fr_match_l4_hdr(is1, is2);
1085 		amatch = fr_match_addresses(is1, is2);
1086 		/*
1087 		 * If addresses match (amatch != 0), then 'match levels'
1088 		 * must be same for matching entries. If amatch and pmatch
1089 		 * have different values (different match levels), then
1090 		 * is1 and is2 belong to different sessions.
1091 		 */
1092 		rv = (amatch != 0) && (amatch == pmatch);
1093 	}
1094 	else
1095 		rv = 0;
1096 
1097 	return (rv);
1098 }
1099 
1100 /* ------------------------------------------------------------------------ */
1101 /* Function:    fr_addstate                                                 */
1102 /* Returns:     ipstate_t* - NULL == failure, else pointer to new state     */
1103 /* Parameters:  fin(I)    - pointer to packet information                   */
1104 /*              stsave(O) - pointer to place to save pointer to created     */
1105 /*                          state structure.                                */
1106 /*              flags(I)  - flags to use when creating the structure        */
1107 /*                                                                          */
1108 /* Creates a new IP state structure from the packet information collected.  */
1109 /* Inserts it into the state table and appends to the bottom of the active  */
1110 /* list.  If the capacity of the table has reached the maximum allowed then */
1111 /* the call will fail and a flush is scheduled for the next timeout call.   */
1112 /* ------------------------------------------------------------------------ */
fr_addstate(fin,stsave,flags)1113 ipstate_t *fr_addstate(fin, stsave, flags)
1114 fr_info_t *fin;
1115 ipstate_t **stsave;
1116 u_int flags;
1117 {
1118 	ipstate_t *is, ips;
1119 	struct icmp *ic;
1120 	u_int pass, hv;
1121 	frentry_t *fr;
1122 	tcphdr_t *tcp;
1123 	grehdr_t *gre;
1124 	void *ifp;
1125 	int out;
1126 	ipf_stack_t *ifs = fin->fin_ifs;
1127 
1128 	if (ifs->ifs_fr_state_lock ||
1129 	    (fin->fin_flx & (FI_SHORT|FI_STATE|FI_FRAGBODY|FI_BAD)))
1130 		return NULL;
1131 
1132 	if ((fin->fin_flx & FI_OOW) && !(fin->fin_tcpf & TH_SYN))
1133 		return NULL;
1134 
1135 	/*
1136 	 * Trigger automatic call to fr_state_flush() if the
1137 	 * table has reached capacity specified by hi watermark.
1138 	 */
1139 	if (ST_TAB_WATER_LEVEL(ifs) > ifs->ifs_state_flush_level_hi)
1140 		ifs->ifs_fr_state_doflush = 1;
1141 
1142 	/*
1143 	 * If the max number of state entries has been reached, and there is no
1144 	 * limit on the state count for the rule, then do not continue.  In the
1145 	 * case where a limit exists, it's ok allow the entries to be created as
1146 	 * long as specified limit itself has not been reached.
1147 	 *
1148 	 * Note that because the lock isn't held on fr, it is possible to exceed
1149 	 * the specified size of the table.  However, the cost of this is being
1150 	 * ignored here; as the number by which it can go over is a product of
1151 	 * the number of simultaneous threads that could be executing in here.
1152 	 * So, a limit of 100 won't result in 200, but could result in 101 or 102.
1153 	 *
1154 	 * Also note that, since the automatic flush should have been triggered
1155 	 * well before we reach the maximum number of state table entries, the
1156 	 * likelihood of reaching the max (and thus exceedng it) is minimal.
1157 	 */
1158 	fr = fin->fin_fr;
1159 	if (fr != NULL) {
1160 		if ((ifs->ifs_ips_num >= ifs->ifs_fr_statemax) &&
1161 		    (fr->fr_statemax == 0)) {
1162 			ATOMIC_INCL(ifs->ifs_ips_stats.iss_max);
1163 			return NULL;
1164 		}
1165 		if ((fr->fr_statemax != 0) &&
1166 		    (fr->fr_statecnt >= fr->fr_statemax)) {
1167 			ATOMIC_INCL(ifs->ifs_ips_stats.iss_maxref);
1168 			ifs->ifs_fr_state_doflush = 1;
1169 			return NULL;
1170 		}
1171 	}
1172 
1173 	ic = NULL;
1174 	tcp = NULL;
1175 	out = fin->fin_out;
1176 	is = &ips;
1177 	bzero((char *)is, sizeof(*is));
1178 
1179 	if (fr == NULL) {
1180 		pass = ifs->ifs_fr_flags;
1181 		is->is_tag = FR_NOLOGTAG;
1182 	} else {
1183 		pass = fr->fr_flags;
1184 	}
1185 
1186 	is->is_die = 1 + ifs->ifs_fr_ticks;
1187 	/*
1188 	 * We want to check everything that is a property of this packet,
1189 	 * but we don't (automatically) care about it's fragment status as
1190 	 * this may change.
1191 	 */
1192 	is->is_pass = pass;
1193 	is->is_v = fin->fin_v;
1194 	is->is_opt[0] = fin->fin_optmsk;
1195 	is->is_optmsk[0] = 0xffffffff;
1196 	/*
1197 	 * The reverse direction option mask will be set in fr_matchsrcdst(),
1198 	 * when we will see the first packet from the peer. We will leave it
1199 	 * as zero for now.
1200 	 */
1201 	is->is_optmsk[1] = 0x0;
1202 
1203 	if (is->is_v == 6) {
1204 		is->is_opt[0] &= ~0x8;
1205 		is->is_optmsk[0] &= ~0x8;
1206 	}
1207 	is->is_sec = fin->fin_secmsk;
1208 	is->is_secmsk = 0xffff;
1209 	is->is_auth = fin->fin_auth;
1210 	is->is_authmsk = 0xffff;
1211 
1212 	/*
1213 	 * Copy and calculate...
1214 	 */
1215 	hv = (is->is_p = fin->fin_fi.fi_p);
1216 	is->is_src = fin->fin_fi.fi_src;
1217 	hv += is->is_saddr;
1218 	is->is_dst = fin->fin_fi.fi_dst;
1219 	hv += is->is_daddr;
1220 #ifdef	USE_INET6
1221 	if (fin->fin_v == 6) {
1222 		/*
1223 		 * For ICMPv6, we check to see if the destination address is
1224 		 * a multicast address.  If it is, do not include it in the
1225 		 * calculation of the hash because the correct reply will come
1226 		 * back from a real address, not a multicast address.
1227 		 */
1228 		if ((is->is_p == IPPROTO_ICMPV6) &&
1229 		    IN6_IS_ADDR_MULTICAST(&is->is_dst.in6)) {
1230 			/*
1231 			 * So you can do keep state with neighbour discovery.
1232 			 *
1233 			 * Here we could use the address from the neighbour
1234 			 * solicit message to put in the state structure and
1235 			 * we could use that without a wildcard flag too...
1236 			 */
1237 			is->is_flags |= SI_W_DADDR;
1238 			hv -= is->is_daddr;
1239 		} else {
1240 			hv += is->is_dst.i6[1];
1241 			hv += is->is_dst.i6[2];
1242 			hv += is->is_dst.i6[3];
1243 		}
1244 		hv += is->is_src.i6[1];
1245 		hv += is->is_src.i6[2];
1246 		hv += is->is_src.i6[3];
1247 	}
1248 #endif
1249 	if ((fin->fin_v == 4) &&
1250 	    (fin->fin_flx & (FI_MULTICAST|FI_BROADCAST|FI_MBCAST))) {
1251 		if (fin->fin_out == 0) {
1252 			flags |= SI_W_DADDR|SI_CLONE;
1253 			hv -= is->is_daddr;
1254 		} else {
1255 			flags |= SI_W_SADDR|SI_CLONE;
1256 			hv -= is->is_saddr;
1257 		}
1258 	}
1259 
1260 	switch (is->is_p)
1261 	{
1262 #ifdef	USE_INET6
1263 	case IPPROTO_ICMPV6 :
1264 		ic = fin->fin_dp;
1265 
1266 		switch (ic->icmp_type)
1267 		{
1268 		case ICMP6_ECHO_REQUEST :
1269 			is->is_icmp.ici_type = ic->icmp_type;
1270 			hv += (is->is_icmp.ici_id = ic->icmp_id);
1271 			break;
1272 		case ICMP6_MEMBERSHIP_QUERY :
1273 		case ND_ROUTER_SOLICIT :
1274 		case ND_NEIGHBOR_SOLICIT :
1275 		case ICMP6_NI_QUERY :
1276 			is->is_icmp.ici_type = ic->icmp_type;
1277 			break;
1278 		default :
1279 			return NULL;
1280 		}
1281 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_icmp);
1282 		break;
1283 #endif
1284 	case IPPROTO_ICMP :
1285 		ic = fin->fin_dp;
1286 
1287 		switch (ic->icmp_type)
1288 		{
1289 		case ICMP_ECHO :
1290 		case ICMP_ECHOREPLY :
1291 		case ICMP_TSTAMP :
1292 		case ICMP_IREQ :
1293 		case ICMP_MASKREQ :
1294 			is->is_icmp.ici_type = ic->icmp_type;
1295 			hv += (is->is_icmp.ici_id = ic->icmp_id);
1296 			break;
1297 		default :
1298 			return NULL;
1299 		}
1300 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_icmp);
1301 		break;
1302 
1303 	case IPPROTO_GRE :
1304 		gre = fin->fin_dp;
1305 
1306 		is->is_gre.gs_flags = gre->gr_flags;
1307 		is->is_gre.gs_ptype = gre->gr_ptype;
1308 		if (GRE_REV(is->is_gre.gs_flags) == 1) {
1309 			is->is_call[0] = fin->fin_data[0];
1310 			is->is_call[1] = fin->fin_data[1];
1311 		}
1312 		break;
1313 
1314 	case IPPROTO_TCP :
1315 		tcp = fin->fin_dp;
1316 
1317 		if (tcp->th_flags & TH_RST)
1318 			return NULL;
1319 		/*
1320 		 * The endian of the ports doesn't matter, but the ack and
1321 		 * sequence numbers do as we do mathematics on them later.
1322 		 */
1323 		is->is_sport = htons(fin->fin_data[0]);
1324 		is->is_dport = htons(fin->fin_data[1]);
1325 		if ((flags & (SI_W_DPORT|SI_W_SPORT)) == 0) {
1326 			hv += is->is_sport;
1327 			hv += is->is_dport;
1328 		}
1329 
1330 		/*
1331 		 * If this is a real packet then initialise fields in the
1332 		 * state information structure from the TCP header information.
1333 		 */
1334 
1335 		is->is_maxdwin = 1;
1336 		is->is_maxswin = ntohs(tcp->th_win);
1337 		if (is->is_maxswin == 0)
1338 			is->is_maxswin = 1;
1339 
1340 		if ((fin->fin_flx & FI_IGNORE) == 0) {
1341 			is->is_send = ntohl(tcp->th_seq) + fin->fin_dlen -
1342 				      (TCP_OFF(tcp) << 2) +
1343 				      ((tcp->th_flags & TH_SYN) ? 1 : 0) +
1344 				      ((tcp->th_flags & TH_FIN) ? 1 : 0);
1345 			is->is_maxsend = is->is_send;
1346 
1347 			/*
1348 			 * Window scale option is only present in
1349 			 * SYN/SYN-ACK packet.
1350 			 */
1351 			if ((tcp->th_flags & ~(TH_FIN|TH_ACK|TH_ECNALL)) ==
1352 			    TH_SYN &&
1353 			    (TCP_OFF(tcp) > (sizeof(tcphdr_t) >> 2))) {
1354 				if (fr_tcpoptions(fin, tcp,
1355 					&is->is_tcp.ts_data[0]) == -1) {
1356 					fin->fin_flx |= FI_BAD;
1357 				}
1358 			}
1359 
1360 			if ((fin->fin_out != 0) && (pass & FR_NEWISN) != 0) {
1361 				fr_checknewisn(fin, is);
1362 				fr_fixoutisn(fin, is);
1363 			}
1364 
1365 			if ((tcp->th_flags & TH_OPENING) == TH_SYN)
1366 				flags |= IS_TCPFSM;
1367 			else {
1368 				is->is_maxdwin = is->is_maxswin * 2;
1369 				is->is_dend = ntohl(tcp->th_ack);
1370 				is->is_maxdend = ntohl(tcp->th_ack);
1371 				is->is_maxdwin *= 2;
1372 			}
1373 		}
1374 
1375 		/*
1376 		 * If we're creating state for a starting connection, start the
1377 		 * timer on it as we'll never see an error if it fails to
1378 		 * connect.
1379 		 */
1380 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_tcp);
1381 		break;
1382 
1383 	case IPPROTO_UDP :
1384 		tcp = fin->fin_dp;
1385 
1386 		is->is_sport = htons(fin->fin_data[0]);
1387 		is->is_dport = htons(fin->fin_data[1]);
1388 		if ((flags & (SI_W_DPORT|SI_W_SPORT)) == 0) {
1389 			hv += tcp->th_dport;
1390 			hv += tcp->th_sport;
1391 		}
1392 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_udp);
1393 		break;
1394 
1395 	default :
1396 		break;
1397 	}
1398 	hv = DOUBLE_HASH(hv, ifs);
1399 	is->is_hv = hv;
1400 	is->is_rule = fr;
1401 	is->is_flags = flags & IS_INHERITED;
1402 
1403 	/*
1404 	 * Look for identical state.
1405 	 */
1406 	for (is = ifs->ifs_ips_table[is->is_hv % ifs->ifs_fr_statesize];
1407 	     is != NULL;
1408 	     is = is->is_hnext) {
1409 		if (fr_matchstates(&ips, is) == 1)
1410 			break;
1411 	}
1412 
1413 	/*
1414 	 * we've found a matching state -> state already exists,
1415 	 * we are not going to add a duplicate record.
1416 	 */
1417 	if (is != NULL)
1418 		return NULL;
1419 
1420 	if (ifs->ifs_ips_stats.iss_bucketlen[hv] >= ifs->ifs_fr_state_maxbucket) {
1421 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_bucketfull);
1422 		return NULL;
1423 	}
1424 	KMALLOC(is, ipstate_t *);
1425 	if (is == NULL) {
1426 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_nomem);
1427 		return NULL;
1428 	}
1429 	bcopy((char *)&ips, (char *)is, sizeof(*is));
1430 	/*
1431 	 * Do not do the modulous here, it is done in fr_stinsert().
1432 	 */
1433 	if (fr != NULL) {
1434 		(void) strncpy(is->is_group, fr->fr_group, FR_GROUPLEN);
1435 		if (fr->fr_age[0] != 0) {
1436 			is->is_tqehead[0] =
1437 			    fr_addtimeoutqueue(&ifs->ifs_ips_utqe,
1438 					       fr->fr_age[0], ifs);
1439 			is->is_sti.tqe_flags |= TQE_RULEBASED;
1440 		}
1441 		if (fr->fr_age[1] != 0) {
1442 			is->is_tqehead[1] =
1443 			    fr_addtimeoutqueue(&ifs->ifs_ips_utqe,
1444 					       fr->fr_age[1], ifs);
1445 			is->is_sti.tqe_flags |= TQE_RULEBASED;
1446 		}
1447 		is->is_tag = fr->fr_logtag;
1448 
1449 		is->is_ifp[(out << 1) + 1] = fr->fr_ifas[1];
1450 		is->is_ifp[(1 - out) << 1] = fr->fr_ifas[2];
1451 		is->is_ifp[((1 - out) << 1) + 1] = fr->fr_ifas[3];
1452 
1453 		if (((ifp = fr->fr_ifas[1]) != NULL) &&
1454 		    (ifp != (void *)-1)) {
1455 			COPYIFNAME(ifp, is->is_ifname[(out << 1) + 1], fr->fr_v);
1456 		}
1457 		if (((ifp = fr->fr_ifas[2]) != NULL) &&
1458 		    (ifp != (void *)-1)) {
1459 			COPYIFNAME(ifp, is->is_ifname[(1 - out) << 1], fr->fr_v);
1460 		}
1461 		if (((ifp = fr->fr_ifas[3]) != NULL) &&
1462 		    (ifp != (void *)-1)) {
1463 			COPYIFNAME(ifp, is->is_ifname[((1 - out) << 1) + 1], fr->fr_v);
1464 		}
1465 	}
1466 
1467 	is->is_ifp[out << 1] = fin->fin_ifp;
1468 	if (fin->fin_ifp != NULL) {
1469 		COPYIFNAME(fin->fin_ifp, is->is_ifname[out << 1], fin->fin_v);
1470 	}
1471 
1472 	is->is_ref = 1;
1473 	is->is_pkts[0] = 0, is->is_bytes[0] = 0;
1474 	is->is_pkts[1] = 0, is->is_bytes[1] = 0;
1475 	is->is_pkts[2] = 0, is->is_bytes[2] = 0;
1476 	is->is_pkts[3] = 0, is->is_bytes[3] = 0;
1477 	if ((fin->fin_flx & FI_IGNORE) == 0) {
1478 		is->is_pkts[out] = 1;
1479 		is->is_bytes[out] = fin->fin_plen;
1480 		is->is_flx[out][0] = fin->fin_flx & FI_CMP;
1481 		is->is_flx[out][0] &= ~FI_OOW;
1482 	}
1483 
1484 	if (pass & FR_STSTRICT)
1485 		is->is_flags |= IS_STRICT;
1486 
1487 	if (pass & FR_STATESYNC)
1488 		is->is_flags |= IS_STATESYNC;
1489 
1490 	if (flags & (SI_WILDP|SI_WILDA)) {
1491 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_wild);
1492 	}
1493 	is->is_rulen = fin->fin_rule;
1494 
1495 
1496 	if (pass & FR_LOGFIRST)
1497 		is->is_pass &= ~(FR_LOGFIRST|FR_LOG);
1498 
1499 	READ_ENTER(&ifs->ifs_ipf_state);
1500 	is->is_me = stsave;
1501 
1502 	fr_stinsert(is, fin->fin_rev, ifs);
1503 
1504 	if (fin->fin_p == IPPROTO_TCP) {
1505 		/*
1506 		* If we're creating state for a starting connection, start the
1507 		* timer on it as we'll never see an error if it fails to
1508 		* connect.
1509 		*/
1510 		(void) fr_tcp_age(&is->is_sti, fin, ifs->ifs_ips_tqtqb,
1511 				  is->is_flags);
1512 		MUTEX_EXIT(&is->is_lock);
1513 #ifdef	IPFILTER_SCAN
1514 		if ((is->is_flags & SI_CLONE) == 0)
1515 			(void) ipsc_attachis(is);
1516 #endif
1517 	} else {
1518 		MUTEX_EXIT(&is->is_lock);
1519 	}
1520 #ifdef	IPFILTER_SYNC
1521 	if ((is->is_flags & IS_STATESYNC) && ((is->is_flags & SI_CLONE) == 0))
1522 		is->is_sync = ipfsync_new(SMC_STATE, fin, is);
1523 #endif
1524 	if (ifs->ifs_ipstate_logging)
1525 		ipstate_log(is, ISL_NEW, ifs);
1526 
1527 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
1528 	fin->fin_rev = IP6_NEQ(&is->is_dst, &fin->fin_daddr);
1529 	fin->fin_flx |= FI_STATE;
1530 	if (fin->fin_flx & FI_FRAG)
1531 		(void) fr_newfrag(fin, pass ^ FR_KEEPSTATE);
1532 
1533 	return is;
1534 }
1535 
1536 
1537 /* ------------------------------------------------------------------------ */
1538 /* Function:    fr_tcpoptions                                               */
1539 /* Returns:     int - 1 == packet matches state entry, 0 == it does not     */
1540 /* Parameters:  fin(I) - pointer to packet information                      */
1541 /*              tcp(I) - pointer to TCP packet header                       */
1542 /*              td(I)  - pointer to TCP data held as part of the state      */
1543 /*                                                                          */
1544 /* Look after the TCP header for any options and deal with those that are   */
1545 /* present.  Record details about those that we recogise.                   */
1546 /* ------------------------------------------------------------------------ */
fr_tcpoptions(fin,tcp,td)1547 static int fr_tcpoptions(fin, tcp, td)
1548 fr_info_t *fin;
1549 tcphdr_t *tcp;
1550 tcpdata_t *td;
1551 {
1552 	int off, mlen, ol, i, len, retval;
1553 	char buf[64], *s, opt;
1554 	mb_t *m = NULL;
1555 
1556 	len = (TCP_OFF(tcp) << 2);
1557 	if (fin->fin_dlen < len)
1558 		return 0;
1559 	len -= sizeof(*tcp);
1560 
1561 	off = fin->fin_plen - fin->fin_dlen + sizeof(*tcp) + fin->fin_ipoff;
1562 
1563 	m = fin->fin_m;
1564 	mlen = MSGDSIZE(m) - off;
1565 	if (len > mlen) {
1566 		len = mlen;
1567 		retval = 0;
1568 	} else {
1569 		retval = 1;
1570 	}
1571 
1572 	COPYDATA(m, off, len, buf);
1573 
1574 	for (s = buf; len > 0; ) {
1575 		opt = *s;
1576 		if (opt == TCPOPT_EOL)
1577 			break;
1578 		else if (opt == TCPOPT_NOP)
1579 			ol = 1;
1580 		else {
1581 			if (len < 2)
1582 				break;
1583 			ol = (int)*(s + 1);
1584 			if (ol < 2 || ol > len)
1585 				break;
1586 
1587 			/*
1588 			 * Extract the TCP options we are interested in out of
1589 			 * the header and store them in the the tcpdata struct.
1590 			 */
1591 			switch (opt)
1592 			{
1593 			case TCPOPT_WINDOW :
1594 				if (ol == TCPOLEN_WINDOW) {
1595 					i = (int)*(s + 2);
1596 					if (i > TCP_WSCALE_MAX)
1597 						i = TCP_WSCALE_MAX;
1598 					else if (i < 0)
1599 						i = 0;
1600 					td->td_winscale = i;
1601 					td->td_winflags |= TCP_WSCALE_SEEN |
1602 							    TCP_WSCALE_FIRST;
1603 				} else
1604 					retval = -1;
1605 				break;
1606 			case TCPOPT_MAXSEG :
1607 				/*
1608 				 * So, if we wanted to set the TCP MAXSEG,
1609 				 * it should be done here...
1610 				 */
1611 				if (ol == TCPOLEN_MAXSEG) {
1612 					i = (int)*(s + 2);
1613 					i <<= 8;
1614 					i += (int)*(s + 3);
1615 					td->td_maxseg = i;
1616 				} else
1617 					retval = -1;
1618 				break;
1619 			case TCPOPT_SACK_PERMITTED :
1620 				if (ol == TCPOLEN_SACK_PERMITTED)
1621 					td->td_winflags |= TCP_SACK_PERMIT;
1622 				else
1623 					retval = -1;
1624 				break;
1625 			}
1626 		}
1627 		len -= ol;
1628 		s += ol;
1629 	}
1630 	return retval;
1631 }
1632 
1633 
1634 /* ------------------------------------------------------------------------ */
1635 /* Function:    fr_tcpstate                                                 */
1636 /* Returns:     int - 1 == packet matches state entry, 0 == it does not     */
1637 /* Parameters:  fin(I)   - pointer to packet information                    */
1638 /*              tcp(I)   - pointer to TCP packet header                     */
1639 /*              is(I)  - pointer to master state structure                  */
1640 /*                                                                          */
1641 /* Check to see if a packet with TCP headers fits within the TCP window.    */
1642 /* Change timeout depending on whether new packet is a SYN-ACK returning    */
1643 /* for a SYN or a RST or FIN which indicate time to close up shop.          */
1644 /* ------------------------------------------------------------------------ */
fr_tcpstate(fin,tcp,is)1645 static int fr_tcpstate(fin, tcp, is)
1646 fr_info_t *fin;
1647 tcphdr_t *tcp;
1648 ipstate_t *is;
1649 {
1650 	int source, ret = 0, flags;
1651 	tcpdata_t  *fdata, *tdata;
1652 	ipf_stack_t *ifs = fin->fin_ifs;
1653 
1654 	source = !fin->fin_rev;
1655 	if (((is->is_flags & IS_TCPFSM) != 0) && (source == 1) &&
1656 	    (ntohs(is->is_sport) != fin->fin_data[0]))
1657 		source = 0;
1658 	fdata = &is->is_tcp.ts_data[!source];
1659 	tdata = &is->is_tcp.ts_data[source];
1660 
1661 	MUTEX_ENTER(&is->is_lock);
1662 
1663 	/*
1664 	 * If a SYN packet is received for a connection that is in a half
1665 	 * closed state, then move its state entry to deletetq. In such case
1666 	 * the SYN packet will be consequently dropped. This allows new state
1667 	 * entry to be created with a retransmited SYN packet.
1668 	 */
1669 	if ((tcp->th_flags & TH_OPENING) == TH_SYN) {
1670 		if ((is->is_state[source] > IPF_TCPS_ESTABLISHED) &&
1671 		    (is->is_state[!source] > IPF_TCPS_ESTABLISHED)) {
1672 			is->is_state[source] = IPF_TCPS_CLOSED;
1673 			is->is_state[!source] = IPF_TCPS_CLOSED;
1674 			/*
1675 			 * Do not update is->is_sti.tqe_die in case state entry
1676 			 * is already present in deletetq. It prevents state
1677 			 * entry ttl update by retransmitted SYN packets, which
1678 			 * may arrive before timer tick kicks off. The SYN
1679 			 * packet will be dropped again.
1680 			 */
1681 			if (is->is_sti.tqe_ifq != &ifs->ifs_ips_deletetq)
1682 				fr_movequeue(&is->is_sti, is->is_sti.tqe_ifq,
1683 					&fin->fin_ifs->ifs_ips_deletetq,
1684 					fin->fin_ifs);
1685 
1686 			MUTEX_EXIT(&is->is_lock);
1687 			return 0;
1688 		}
1689 	}
1690 
1691 	if (fr_tcpinwindow(fin, fdata, tdata, tcp, is->is_flags)) {
1692 #ifdef	IPFILTER_SCAN
1693 		if (is->is_flags & (IS_SC_CLIENT|IS_SC_SERVER)) {
1694 			ipsc_packet(fin, is);
1695 			if (FR_ISBLOCK(is->is_pass)) {
1696 				MUTEX_EXIT(&is->is_lock);
1697 				return 1;
1698 			}
1699 		}
1700 #endif
1701 
1702 		/*
1703 		 * Nearing end of connection, start timeout.
1704 		 */
1705 		ret = fr_tcp_age(&is->is_sti, fin, ifs->ifs_ips_tqtqb,
1706 				 is->is_flags);
1707 		if (ret == 0) {
1708 			MUTEX_EXIT(&is->is_lock);
1709 			return 0;
1710 		}
1711 
1712 		/*
1713 		 * set s0's as appropriate.  Use syn-ack packet as it
1714 		 * contains both pieces of required information.
1715 		 */
1716 		/*
1717 		 * Window scale option is only present in SYN/SYN-ACK packet.
1718 		 * Compare with ~TH_FIN to mask out T/TCP setups.
1719 		 */
1720 		flags = tcp->th_flags & ~(TH_FIN|TH_ECNALL);
1721 		if (flags == (TH_SYN|TH_ACK)) {
1722 			is->is_s0[source] = ntohl(tcp->th_ack);
1723 			is->is_s0[!source] = ntohl(tcp->th_seq) + 1;
1724 			if (TCP_OFF(tcp) > (sizeof (tcphdr_t) >> 2)) {
1725 				(void) fr_tcpoptions(fin, tcp, fdata);
1726 			}
1727 			if ((fin->fin_out != 0) && (is->is_pass & FR_NEWISN))
1728 				fr_checknewisn(fin, is);
1729 		} else if (flags == TH_SYN) {
1730 			is->is_s0[source] = ntohl(tcp->th_seq) + 1;
1731 			if ((TCP_OFF(tcp) > (sizeof(tcphdr_t) >> 2)))
1732 				(void) fr_tcpoptions(fin, tcp, fdata);
1733 
1734 			if ((fin->fin_out != 0) && (is->is_pass & FR_NEWISN))
1735 				fr_checknewisn(fin, is);
1736 
1737 		}
1738 		ret = 1;
1739 	} else
1740 		fin->fin_flx |= FI_OOW;
1741 	MUTEX_EXIT(&is->is_lock);
1742 	return ret;
1743 }
1744 
1745 
1746 /* ------------------------------------------------------------------------ */
1747 /* Function:    fr_checknewisn                                              */
1748 /* Returns:     Nil                                                         */
1749 /* Parameters:  fin(I)   - pointer to packet information                    */
1750 /*              is(I)  - pointer to master state structure                  */
1751 /*                                                                          */
1752 /* Check to see if this TCP connection is expecting and needs a new         */
1753 /* sequence number for a particular direction of the connection.            */
1754 /*                                                                          */
1755 /* NOTE: This does not actually change the sequence numbers, only gets new  */
1756 /* one ready.                                                               */
1757 /* ------------------------------------------------------------------------ */
fr_checknewisn(fin,is)1758 static void fr_checknewisn(fin, is)
1759 fr_info_t *fin;
1760 ipstate_t *is;
1761 {
1762 	u_32_t sumd, old, new;
1763 	tcphdr_t *tcp;
1764 	int i;
1765 
1766 	i = fin->fin_rev;
1767 	tcp = fin->fin_dp;
1768 
1769 	if (((i == 0) && !(is->is_flags & IS_ISNSYN)) ||
1770 	    ((i == 1) && !(is->is_flags & IS_ISNACK))) {
1771 		old = ntohl(tcp->th_seq);
1772 		new = fr_newisn(fin);
1773 		is->is_isninc[i] = new - old;
1774 		CALC_SUMD(old, new, sumd);
1775 		is->is_sumd[i] = (sumd & 0xffff) + (sumd >> 16);
1776 
1777 		is->is_flags |= ((i == 0) ? IS_ISNSYN : IS_ISNACK);
1778 	}
1779 }
1780 
1781 
1782 /* ------------------------------------------------------------------------ */
1783 /* Function:    fr_tcpinwindow                                              */
1784 /* Returns:     int - 1 == packet inside TCP "window", 0 == not inside.     */
1785 /* Parameters:  fin(I)   - pointer to packet information                    */
1786 /*              fdata(I) - pointer to tcp state informatio (forward)        */
1787 /*              tdata(I) - pointer to tcp state informatio (reverse)        */
1788 /*              tcp(I)   - pointer to TCP packet header                     */
1789 /*                                                                          */
1790 /* Given a packet has matched addresses and ports, check to see if it is    */
1791 /* within the TCP data window.  In a show of generosity, allow packets that */
1792 /* are within the window space behind the current sequence # as well.       */
1793 /* ------------------------------------------------------------------------ */
fr_tcpinwindow(fin,fdata,tdata,tcp,flags)1794 int fr_tcpinwindow(fin, fdata, tdata, tcp, flags)
1795 fr_info_t *fin;
1796 tcpdata_t  *fdata, *tdata;
1797 tcphdr_t *tcp;
1798 int flags;
1799 {
1800 	tcp_seq seq, ack, end;
1801 	int ackskew, tcpflags;
1802 	u_32_t win, maxwin;
1803 	int dsize, inseq;
1804 
1805 	/*
1806 	 * Find difference between last checked packet and this packet.
1807 	 */
1808 	tcpflags = tcp->th_flags;
1809 	seq = ntohl(tcp->th_seq);
1810 	ack = ntohl(tcp->th_ack);
1811 
1812 	if (tcpflags & TH_SYN)
1813 		win = ntohs(tcp->th_win);
1814 	else
1815 		win = ntohs(tcp->th_win) << fdata->td_winscale;
1816 
1817 	/*
1818 	 * win 0 means the receiving endpoint has closed the window, because it
1819 	 * has not enough memory to receive data from sender. In such case we
1820 	 * are pretending window size to be 1 to let TCP probe data through.
1821 	 * TCP probe data can be either 0 or 1 octet of data, the RFC does not
1822 	 * state this accurately, so we have to allow 1 octet (win = 1) even if
1823 	 * the window is closed (win == 0).
1824 	 */
1825 	if (win == 0)
1826 		win = 1;
1827 
1828 	dsize = fin->fin_dlen - (TCP_OFF(tcp) << 2) +
1829 		((tcpflags & TH_SYN) ? 1 : 0) + ((tcpflags & TH_FIN) ? 1 : 0);
1830 
1831 	/*
1832 	 * if window scaling is present, the scaling is only allowed
1833 	 * for windows not in the first SYN packet. In that packet the
1834 	 * window is 65535 to specify the largest window possible
1835 	 * for receivers not implementing the window scale option.
1836 	 * Currently, we do not assume TTCP here. That means that
1837 	 * if we see a second packet from a host (after the initial
1838 	 * SYN), we can assume that the receiver of the SYN did
1839 	 * already send back the SYN/ACK (and thus that we know if
1840 	 * the receiver also does window scaling)
1841 	 */
1842 	if (!(tcpflags & TH_SYN) && (fdata->td_winflags & TCP_WSCALE_FIRST)) {
1843 		fdata->td_winflags &= ~TCP_WSCALE_FIRST;
1844 		fdata->td_maxwin = win;
1845 	}
1846 
1847 	end = seq + dsize;
1848 
1849 	if ((fdata->td_end == 0) &&
1850 	    (!(flags & IS_TCPFSM) ||
1851 	     ((tcpflags & TH_OPENING) == TH_OPENING))) {
1852 		/*
1853 		 * Must be a (outgoing) SYN-ACK in reply to a SYN.
1854 		 */
1855 		fdata->td_end = end - 1;
1856 		fdata->td_maxwin = 1;
1857 		fdata->td_maxend = end + win;
1858 	}
1859 
1860 	if (!(tcpflags & TH_ACK)) {  /* Pretend an ack was sent */
1861 		ack = tdata->td_end;
1862 	} else if (((tcpflags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) &&
1863 		   (ack == 0)) {
1864 		/* gross hack to get around certain broken tcp stacks */
1865 		ack = tdata->td_end;
1866 	}
1867 
1868 	maxwin = tdata->td_maxwin;
1869 	ackskew = tdata->td_end - ack;
1870 
1871 	/*
1872 	 * Strict sequencing only allows in-order delivery.
1873 	 */
1874 	if ((flags & IS_STRICT) != 0) {
1875 		if (seq != fdata->td_end) {
1876 			DTRACE_PROBE(strict_check);
1877 			return 0;
1878 		}
1879 	}
1880 
1881 #define	SEQ_GE(a,b)	((int)((a) - (b)) >= 0)
1882 #define	SEQ_GT(a,b)	((int)((a) - (b)) > 0)
1883 	inseq = 0;
1884 	DTRACE_PROBE4(
1885 		dyn_params,
1886 		int, dsize,
1887 		int, ackskew,
1888 		int, maxwin,
1889 		int, win
1890 	);
1891 	if (
1892 #if defined(_KERNEL)
1893 		/*
1894 		 * end <-> s + n
1895 		 * maxend <-> ack + win
1896 		 * this is upperbound check
1897 		 */
1898 	    (SEQ_GE(fdata->td_maxend, end)) &&
1899 		/*
1900 		 * this is lowerbound check
1901 		 */
1902 	    (SEQ_GE(seq, fdata->td_end - maxwin)) &&
1903 #endif
1904 /* XXX what about big packets */
1905 #define MAXACKWINDOW 66000
1906 	    (-ackskew <= (MAXACKWINDOW)) &&
1907 	    ( ackskew <= (MAXACKWINDOW << fdata->td_winscale))) {
1908 		inseq = 1;
1909 	/*
1910 	 * Microsoft Windows will send the next packet to the right of the
1911 	 * window if SACK is in use.
1912 	 */
1913 	} else if ((seq == fdata->td_maxend) && (ackskew == 0) &&
1914 	    (fdata->td_winflags & TCP_SACK_PERMIT) &&
1915 	    (tdata->td_winflags & TCP_SACK_PERMIT)) {
1916 		inseq = 1;
1917 	/*
1918 	 * RST ACK with SEQ equal to 0 is sent by some OSes (i.e. Solaris) as a
1919 	 * response to initial SYN packet, when  there is no application
1920 	 * listeing to on a port, where the SYN packet has came to.
1921 	 */
1922 	} else if ((seq == 0) && (tcpflags == (TH_RST|TH_ACK)) &&
1923 			(ackskew >= -1) && (ackskew <= 1)) {
1924 		inseq = 1;
1925 	} else if (!(flags & IS_TCPFSM)) {
1926 
1927 		if (!(fdata->td_winflags &
1928 			    (TCP_WSCALE_SEEN|TCP_WSCALE_FIRST))) {
1929 			/*
1930 			 * No TCPFSM and no window scaling, so make some
1931 			 * extra guesses.
1932 			 */
1933 			if ((seq == fdata->td_maxend) && (ackskew == 0))
1934 				inseq = 1;
1935 			else if (SEQ_GE(seq + maxwin, fdata->td_end - maxwin))
1936 				inseq = 1;
1937 		}
1938 	}
1939 
1940 	if (inseq) {
1941 		/* if ackskew < 0 then this should be due to fragmented
1942 		 * packets. There is no way to know the length of the
1943 		 * total packet in advance.
1944 		 * We do know the total length from the fragment cache though.
1945 		 * Note however that there might be more sessions with
1946 		 * exactly the same source and destination parameters in the
1947 		 * state cache (and source and destination is the only stuff
1948 		 * that is saved in the fragment cache). Note further that
1949 		 * some TCP connections in the state cache are hashed with
1950 		 * sport and dport as well which makes it not worthwhile to
1951 		 * look for them.
1952 		 * Thus, when ackskew is negative but still seems to belong
1953 		 * to this session, we bump up the destinations end value.
1954 		 */
1955 		if (ackskew < 0) {
1956 			DTRACE_PROBE2(end_update_td,
1957 				int, tdata->td_end,
1958 				int, ack
1959 			);
1960 			tdata->td_end = ack;
1961 		}
1962 
1963 		/* update max window seen */
1964 		if (fdata->td_maxwin < win) {
1965 			DTRACE_PROBE2(win_update_fd,
1966 				int, fdata->td_maxwin,
1967 				int, win
1968 			);
1969 			fdata->td_maxwin = win;
1970 		}
1971 
1972 		if (SEQ_GT(end, fdata->td_end)) {
1973 			DTRACE_PROBE2(end_update_fd,
1974 				int, fdata->td_end,
1975 				int, end
1976 			);
1977 			fdata->td_end = end;
1978 		}
1979 
1980 		if (SEQ_GE(ack + win, tdata->td_maxend)) {
1981 			DTRACE_PROBE2(max_end_update_td,
1982 				int, tdata->td_maxend,
1983 				int, ack + win
1984 			);
1985 			tdata->td_maxend = ack + win;
1986 		}
1987 
1988 		return 1;
1989 	}
1990 	fin->fin_flx |= FI_OOW;
1991 
1992 #if defined(_KERNEL)
1993 	if (!(SEQ_GE(seq, fdata->td_end - maxwin)))
1994 		fin->fin_flx |= FI_NEG_OOW;
1995 #endif
1996 
1997 	return 0;
1998 }
1999 
2000 
2001 /* ------------------------------------------------------------------------ */
2002 /* Function:    fr_stclone                                                  */
2003 /* Returns:     ipstate_t* - NULL == cloning failed,                        */
2004 /*                           else pointer to new state structure            */
2005 /* Parameters:  fin(I) - pointer to packet information                      */
2006 /*              tcp(I) - pointer to TCP/UDP header                          */
2007 /*              is(I)  - pointer to master state structure                  */
2008 /*                                                                          */
2009 /* Create a "duplcate" state table entry from the master.                   */
2010 /* ------------------------------------------------------------------------ */
fr_stclone(fin,tcp,is)2011 static ipstate_t *fr_stclone(fin, tcp, is)
2012 fr_info_t *fin;
2013 tcphdr_t *tcp;
2014 ipstate_t *is;
2015 {
2016 	ipstate_t *clone;
2017 	u_32_t send;
2018 	ipf_stack_t *ifs = fin->fin_ifs;
2019 
2020 	/*
2021 	 * Trigger automatic call to fr_state_flush() if the
2022 	 * table has reached capacity specified by hi watermark.
2023 	 */
2024 	if (ST_TAB_WATER_LEVEL(ifs) > ifs->ifs_state_flush_level_hi)
2025 		ifs->ifs_fr_state_doflush = 1;
2026 
2027 	/*
2028 	 * If automatic flushing did not do its job, and the table
2029 	 * has filled up, don't try to create a new entry.  A NULL
2030 	 * return will indicate that the cloning has failed.
2031 	 */
2032 	if (ifs->ifs_ips_num >= ifs->ifs_fr_statemax) {
2033 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_max);
2034 		return NULL;
2035 	}
2036 
2037 	KMALLOC(clone, ipstate_t *);
2038 	if (clone == NULL)
2039 		return NULL;
2040 	bcopy((char *)is, (char *)clone, sizeof(*clone));
2041 
2042 	MUTEX_NUKE(&clone->is_lock);
2043 
2044 	clone->is_die = ONE_DAY + ifs->ifs_fr_ticks;
2045 	clone->is_state[0] = 0;
2046 	clone->is_state[1] = 0;
2047 	send = ntohl(tcp->th_seq) + fin->fin_dlen - (TCP_OFF(tcp) << 2) +
2048 		((tcp->th_flags & TH_SYN) ? 1 : 0) +
2049 		((tcp->th_flags & TH_FIN) ? 1 : 0);
2050 
2051 	if (fin->fin_rev == 1) {
2052 		clone->is_dend = send;
2053 		clone->is_maxdend = send;
2054 		clone->is_send = 0;
2055 		clone->is_maxswin = 1;
2056 		clone->is_maxdwin = ntohs(tcp->th_win);
2057 		if (clone->is_maxdwin == 0)
2058 			clone->is_maxdwin = 1;
2059 	} else {
2060 		clone->is_send = send;
2061 		clone->is_maxsend = send;
2062 		clone->is_dend = 0;
2063 		clone->is_maxdwin = 1;
2064 		clone->is_maxswin = ntohs(tcp->th_win);
2065 		if (clone->is_maxswin == 0)
2066 			clone->is_maxswin = 1;
2067 	}
2068 
2069 	clone->is_flags &= ~SI_CLONE;
2070 	clone->is_flags |= SI_CLONED;
2071 	fr_stinsert(clone, fin->fin_rev, ifs);
2072 	clone->is_ref = 1;
2073 	if (clone->is_p == IPPROTO_TCP) {
2074 		(void) fr_tcp_age(&clone->is_sti, fin, ifs->ifs_ips_tqtqb,
2075 				  clone->is_flags);
2076 	}
2077 	MUTEX_EXIT(&clone->is_lock);
2078 #ifdef	IPFILTER_SCAN
2079 	(void) ipsc_attachis(is);
2080 #endif
2081 #ifdef	IPFILTER_SYNC
2082 	if (is->is_flags & IS_STATESYNC)
2083 		clone->is_sync = ipfsync_new(SMC_STATE, fin, clone);
2084 #endif
2085 	return clone;
2086 }
2087 
2088 
2089 /* ------------------------------------------------------------------------ */
2090 /* Function:    fr_matchsrcdst                                              */
2091 /* Returns:     Nil                                                         */
2092 /* Parameters:  fin(I) - pointer to packet information                      */
2093 /*              is(I)  - pointer to state structure                         */
2094 /*              src(I) - pointer to source address                          */
2095 /*              dst(I) - pointer to destination address                     */
2096 /*              tcp(I) - pointer to TCP/UDP header                          */
2097 /*                                                                          */
2098 /* Match a state table entry against an IP packet.  The logic below is that */
2099 /* ret gets set to one if the match succeeds, else remains 0.  If it is     */
2100 /* still 0 after the test. no match.                                        */
2101 /* ------------------------------------------------------------------------ */
fr_matchsrcdst(fin,is,src,dst,tcp,cmask)2102 static ipstate_t *fr_matchsrcdst(fin, is, src, dst, tcp, cmask)
2103 fr_info_t *fin;
2104 ipstate_t *is;
2105 i6addr_t *src, *dst;
2106 tcphdr_t *tcp;
2107 u_32_t cmask;
2108 {
2109 	int ret = 0, rev, out, flags, flx = 0, idx;
2110 	u_short sp, dp;
2111 	u_32_t cflx;
2112 	void *ifp;
2113 	ipf_stack_t *ifs = fin->fin_ifs;
2114 
2115 	rev = IP6_NEQ(&is->is_dst, dst);
2116 	ifp = fin->fin_ifp;
2117 	out = fin->fin_out;
2118 	flags = is->is_flags;
2119 	sp = 0;
2120 	dp = 0;
2121 
2122 	if (tcp != NULL) {
2123 		sp = htons(fin->fin_sport);
2124 		dp = ntohs(fin->fin_dport);
2125 	}
2126 	if (!rev) {
2127 		if (tcp != NULL) {
2128 			if (!(flags & SI_W_SPORT) && (sp != is->is_sport))
2129 				rev = 1;
2130 			else if (!(flags & SI_W_DPORT) && (dp != is->is_dport))
2131 				rev = 1;
2132 		}
2133 	}
2134 
2135 	idx = (out << 1) + rev;
2136 
2137 	/*
2138 	 * If the interface for this 'direction' is set, make sure it matches.
2139 	 * An interface name that is not set matches any, as does a name of *.
2140 	 */
2141 	if ((is->is_ifp[idx] == NULL &&
2142 	    (*is->is_ifname[idx] == '\0' || *is->is_ifname[idx] == '*')) ||
2143 	    is->is_ifp[idx] == ifp)
2144 		ret = 1;
2145 
2146 	if (ret == 0) {
2147 		DTRACE_PROBE(no_match_on_iface);
2148 		return NULL;
2149 	}
2150 	ret = 0;
2151 
2152 	/*
2153 	 * Match addresses and ports.
2154 	 */
2155 	if (rev == 0) {
2156 		if ((IP6_EQ(&is->is_dst, dst) || (flags & SI_W_DADDR)) &&
2157 		    (IP6_EQ(&is->is_src, src) || (flags & SI_W_SADDR))) {
2158 			if (tcp) {
2159 				if ((sp == is->is_sport || flags & SI_W_SPORT)&&
2160 				    (dp == is->is_dport || flags & SI_W_DPORT))
2161 					ret = 1;
2162 			} else {
2163 				ret = 1;
2164 			}
2165 		}
2166 	} else {
2167 		if ((IP6_EQ(&is->is_dst, src) || (flags & SI_W_DADDR)) &&
2168 		    (IP6_EQ(&is->is_src, dst) || (flags & SI_W_SADDR))) {
2169 			if (tcp) {
2170 				if ((dp == is->is_sport || flags & SI_W_SPORT)&&
2171 				    (sp == is->is_dport || flags & SI_W_DPORT))
2172 					ret = 1;
2173 			} else {
2174 				ret = 1;
2175 			}
2176 		}
2177 	}
2178 
2179 	if (ret == 0) {
2180 		DTRACE_PROBE(no_match_on_addrs);
2181 		return NULL;
2182 	}
2183 	/*
2184 	 * Whether or not this should be here, is questionable, but the aim
2185 	 * is to get this out of the main line.
2186 	 */
2187 	if (tcp == NULL)
2188 		flags = is->is_flags & ~(SI_WILDP|SI_NEWFR|SI_CLONE|SI_CLONED);
2189 
2190 	/*
2191 	 * Only one of the source or destination address can be flaged as a
2192 	 * wildcard.  Fill in the missing address, if set.
2193 	 * For IPv6, if the address being copied in is multicast, then
2194 	 * don't reset the wild flag - multicast causes it to be set in the
2195 	 * first place!
2196 	 */
2197 	if ((flags & (SI_W_SADDR|SI_W_DADDR))) {
2198 		fr_ip_t *fi = &fin->fin_fi;
2199 
2200 		if ((flags & SI_W_SADDR) != 0) {
2201 			if (rev == 0) {
2202 #ifdef USE_INET6
2203 				if (is->is_v == 6 &&
2204 				    IN6_IS_ADDR_MULTICAST(&fi->fi_src.in6))
2205 					/*EMPTY*/;
2206 				else
2207 #endif
2208 				{
2209 					is->is_src = fi->fi_src;
2210 					is->is_flags &= ~SI_W_SADDR;
2211 				}
2212 			} else {
2213 #ifdef USE_INET6
2214 				if (is->is_v == 6 &&
2215 				    IN6_IS_ADDR_MULTICAST(&fi->fi_dst.in6))
2216 					/*EMPTY*/;
2217 				else
2218 #endif
2219 				{
2220 					is->is_src = fi->fi_dst;
2221 					is->is_flags &= ~SI_W_SADDR;
2222 				}
2223 			}
2224 		} else if ((flags & SI_W_DADDR) != 0) {
2225 			if (rev == 0) {
2226 #ifdef USE_INET6
2227 				if (is->is_v == 6 &&
2228 				    IN6_IS_ADDR_MULTICAST(&fi->fi_dst.in6))
2229 					/*EMPTY*/;
2230 				else
2231 #endif
2232 				{
2233 					is->is_dst = fi->fi_dst;
2234 					is->is_flags &= ~SI_W_DADDR;
2235 				}
2236 			} else {
2237 #ifdef USE_INET6
2238 				if (is->is_v == 6 &&
2239 				    IN6_IS_ADDR_MULTICAST(&fi->fi_src.in6))
2240 					/*EMPTY*/;
2241 				else
2242 #endif
2243 				{
2244 					is->is_dst = fi->fi_src;
2245 					is->is_flags &= ~SI_W_DADDR;
2246 				}
2247 			}
2248 		}
2249 		if ((is->is_flags & (SI_WILDA|SI_WILDP)) == 0) {
2250 			ATOMIC_DECL(ifs->ifs_ips_stats.iss_wild);
2251 		}
2252 	}
2253 
2254 	flx = fin->fin_flx & cmask;
2255 	cflx = is->is_flx[out][rev];
2256 
2257 	/*
2258 	 * Match up any flags set from IP options.
2259 	 */
2260 	if ((cflx && (flx != (cflx & cmask))) ||
2261 	    ((fin->fin_optmsk & is->is_optmsk[rev]) != is->is_opt[rev]) ||
2262 	    ((fin->fin_secmsk & is->is_secmsk) != is->is_sec) ||
2263 	    ((fin->fin_auth & is->is_authmsk) != is->is_auth)) {
2264 		DTRACE_PROBE4(no_match_on_flags,
2265 		    int, (cflx && (flx != (cflx & cmask))),
2266 		    int,
2267 		    ((fin->fin_optmsk & is->is_optmsk[rev]) != is->is_opt[rev]),
2268 		    int, ((fin->fin_secmsk & is->is_secmsk) != is->is_sec),
2269 		    int, ((fin->fin_auth & is->is_authmsk) != is->is_auth)
2270 		);
2271 		return NULL;
2272 	}
2273 	/*
2274 	 * Only one of the source or destination port can be flagged as a
2275 	 * wildcard.  When filling it in, fill in a copy of the matched entry
2276 	 * if it has the cloning flag set.
2277 	 */
2278 	if ((fin->fin_flx & FI_IGNORE) != 0) {
2279 		fin->fin_rev = rev;
2280 		return is;
2281 	}
2282 
2283 	if ((flags & (SI_W_SPORT|SI_W_DPORT))) {
2284 		if ((flags & SI_CLONE) != 0) {
2285 			ipstate_t *clone;
2286 
2287 			clone = fr_stclone(fin, tcp, is);
2288 			if (clone == NULL)
2289 				return NULL;
2290 			is = clone;
2291 		} else {
2292 			ATOMIC_DECL(ifs->ifs_ips_stats.iss_wild);
2293 		}
2294 
2295 		if ((flags & SI_W_SPORT) != 0) {
2296 			if (rev == 0) {
2297 				is->is_sport = sp;
2298 				is->is_send = ntohl(tcp->th_seq);
2299 			} else {
2300 				is->is_sport = dp;
2301 				is->is_send = ntohl(tcp->th_ack);
2302 			}
2303 			is->is_maxsend = is->is_send + 1;
2304 		} else if ((flags & SI_W_DPORT) != 0) {
2305 			if (rev == 0) {
2306 				is->is_dport = dp;
2307 				is->is_dend = ntohl(tcp->th_ack);
2308 			} else {
2309 				is->is_dport = sp;
2310 				is->is_dend = ntohl(tcp->th_seq);
2311 			}
2312 			is->is_maxdend = is->is_dend + 1;
2313 		}
2314 		is->is_flags &= ~(SI_W_SPORT|SI_W_DPORT);
2315 		if ((flags & SI_CLONED) && ifs->ifs_ipstate_logging)
2316 			ipstate_log(is, ISL_CLONE, ifs);
2317 	}
2318 
2319 	ret = -1;
2320 
2321 	if (is->is_flx[out][rev] == 0) {
2322 		is->is_flx[out][rev] = flx;
2323 		/*
2324 		 * If we are dealing with the first packet coming in reverse
2325 		 * direction (sent by peer), then we have to set options into
2326 		 * state.
2327 		 */
2328 		if (rev == 1 && is->is_optmsk[1] == 0x0) {
2329 			is->is_optmsk[1] = 0xffffffff;
2330 			is->is_opt[1] = fin->fin_optmsk;
2331 			DTRACE_PROBE(set_rev_opts);
2332 		}
2333 		if (is->is_v == 6) {
2334 			is->is_opt[rev] &= ~0x8;
2335 			is->is_optmsk[rev] &= ~0x8;
2336 		}
2337 	}
2338 
2339 	/*
2340 	 * Check if the interface name for this "direction" is set and if not,
2341 	 * fill it in.
2342 	 */
2343 	if (is->is_ifp[idx] == NULL &&
2344 	    (*is->is_ifname[idx] == '\0' || *is->is_ifname[idx] == '*')) {
2345 		is->is_ifp[idx] = ifp;
2346 		COPYIFNAME(ifp, is->is_ifname[idx], fin->fin_v);
2347 	}
2348 	fin->fin_rev = rev;
2349 	return is;
2350 }
2351 
2352 
2353 /* ------------------------------------------------------------------------ */
2354 /* Function:    fr_checkicmpmatchingstate                                   */
2355 /* Returns:     Nil                                                         */
2356 /* Parameters:  fin(I) - pointer to packet information                      */
2357 /*                                                                          */
2358 /* If we've got an ICMP error message, using the information stored in the  */
2359 /* ICMP packet, look for a matching state table entry.                      */
2360 /*                                                                          */
2361 /* If we return NULL then no lock on ipf_state is held.                     */
2362 /* If we return non-null then a read-lock on ipf_state is held.             */
2363 /* ------------------------------------------------------------------------ */
fr_checkicmpmatchingstate(fin)2364 static ipstate_t *fr_checkicmpmatchingstate(fin)
2365 fr_info_t *fin;
2366 {
2367 	ipstate_t *is, **isp;
2368 	u_short sport, dport;
2369 	u_char	pr;
2370 	int backward, i, oi;
2371 	i6addr_t dst, src;
2372 	struct icmp *ic;
2373 	u_short savelen;
2374 	icmphdr_t *icmp;
2375 	fr_info_t ofin;
2376 	tcphdr_t *tcp;
2377 	int len;
2378 	ip_t *oip;
2379 	u_int hv;
2380 	ipf_stack_t *ifs = fin->fin_ifs;
2381 
2382 	/*
2383 	 * Does it at least have the return (basic) IP header ?
2384 	 * Is it an actual recognised ICMP error type?
2385 	 * Only a basic IP header (no options) should be with
2386 	 * an ICMP error header.
2387 	 */
2388 	if ((fin->fin_v != 4) || (fin->fin_hlen != sizeof(ip_t)) ||
2389 	    (fin->fin_plen < ICMPERR_MINPKTLEN) ||
2390 	    !(fin->fin_flx & FI_ICMPERR))
2391 		return NULL;
2392 	ic = fin->fin_dp;
2393 
2394 	oip = (ip_t *)((char *)ic + ICMPERR_ICMPHLEN);
2395 	/*
2396 	 * Check if the at least the old IP header (with options) and
2397 	 * 8 bytes of payload is present.
2398 	 */
2399 	if (fin->fin_plen < ICMPERR_MAXPKTLEN + ((IP_HL(oip) - 5) << 2))
2400 		return NULL;
2401 
2402 	/*
2403 	 * Sanity Checks.
2404 	 */
2405 	len = fin->fin_dlen - ICMPERR_ICMPHLEN;
2406 	if ((len <= 0) || ((IP_HL(oip) << 2) > len))
2407 		return NULL;
2408 
2409 	/*
2410 	 * Is the buffer big enough for all of it ?  It's the size of the IP
2411 	 * header claimed in the encapsulated part which is of concern.  It
2412 	 * may be too big to be in this buffer but not so big that it's
2413 	 * outside the ICMP packet, leading to TCP deref's causing problems.
2414 	 * This is possible because we don't know how big oip_hl is when we
2415 	 * do the pullup early in fr_check() and thus can't guarantee it is
2416 	 * all here now.
2417 	 */
2418 #ifdef  _KERNEL
2419 	{
2420 	mb_t *m;
2421 
2422 	m = fin->fin_m;
2423 # if defined(MENTAT)
2424 	if ((char *)oip + len > (char *)m->b_wptr)
2425 		return NULL;
2426 # else
2427 	if ((char *)oip + len > (char *)fin->fin_ip + m->m_len)
2428 		return NULL;
2429 # endif
2430 	}
2431 #endif
2432 	bcopy((char *)fin, (char *)&ofin, sizeof(*fin));
2433 
2434 	/*
2435 	 * in the IPv4 case we must zero the i6addr union otherwise
2436 	 * the IP6_EQ and IP6_NEQ macros produce the wrong results because
2437 	 * of the 'junk' in the unused part of the union
2438 	 */
2439 	bzero((char *)&src, sizeof(src));
2440 	bzero((char *)&dst, sizeof(dst));
2441 
2442 	/*
2443 	 * we make an fin entry to be able to feed it to
2444 	 * matchsrcdst note that not all fields are encessary
2445 	 * but this is the cleanest way. Note further we fill
2446 	 * in fin_mp such that if someone uses it we'll get
2447 	 * a kernel panic. fr_matchsrcdst does not use this.
2448 	 *
2449 	 * watch out here, as ip is in host order and oip in network
2450 	 * order. Any change we make must be undone afterwards, like
2451 	 * oip->ip_off - it is still in network byte order so fix it.
2452 	 */
2453 	savelen = oip->ip_len;
2454 	oip->ip_len = len;
2455 	oip->ip_off = ntohs(oip->ip_off);
2456 
2457 	ofin.fin_flx = FI_NOCKSUM;
2458 	ofin.fin_v = 4;
2459 	ofin.fin_ip = oip;
2460 	ofin.fin_m = NULL;	/* if dereferenced, panic XXX */
2461 	ofin.fin_mp = NULL;	/* if dereferenced, panic XXX */
2462 	ofin.fin_plen = fin->fin_dlen - ICMPERR_ICMPHLEN;
2463 	(void) fr_makefrip(IP_HL(oip) << 2, oip, &ofin);
2464 	ofin.fin_ifp = fin->fin_ifp;
2465 	ofin.fin_out = !fin->fin_out;
2466 	/*
2467 	 * Reset the short and bad flag here because in fr_matchsrcdst()
2468 	 * the flags for the current packet (fin_flx) are compared against
2469 	 * those for the existing session.
2470 	 */
2471 	ofin.fin_flx &= ~(FI_BAD|FI_SHORT);
2472 
2473 	/*
2474 	 * Put old values of ip_len and ip_off back as we don't know
2475 	 * if we have to forward the packet (or process it again.
2476 	 */
2477 	oip->ip_len = savelen;
2478 	oip->ip_off = htons(oip->ip_off);
2479 
2480 	switch (oip->ip_p)
2481 	{
2482 	case IPPROTO_ICMP :
2483 		/*
2484 		 * an ICMP error can only be generated as a result of an
2485 		 * ICMP query, not as the response on an ICMP error
2486 		 *
2487 		 * XXX theoretically ICMP_ECHOREP and the other reply's are
2488 		 * ICMP query's as well, but adding them here seems strange XXX
2489 		 */
2490 		if ((ofin.fin_flx & FI_ICMPERR) != 0)
2491 		    	return NULL;
2492 
2493 		/*
2494 		 * perform a lookup of the ICMP packet in the state table
2495 		 */
2496 		icmp = (icmphdr_t *)((char *)oip + (IP_HL(oip) << 2));
2497 		hv = (pr = oip->ip_p);
2498 		src.in4 = oip->ip_src;
2499 		hv += src.in4.s_addr;
2500 		dst.in4 = oip->ip_dst;
2501 		hv += dst.in4.s_addr;
2502 		hv += icmp->icmp_id;
2503 		hv = DOUBLE_HASH(hv, ifs);
2504 
2505 		READ_ENTER(&ifs->ifs_ipf_state);
2506 		for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
2507 			isp = &is->is_hnext;
2508 			if ((is->is_p != pr) || (is->is_v != 4))
2509 				continue;
2510 			if (is->is_pass & FR_NOICMPERR)
2511 				continue;
2512 			is = fr_matchsrcdst(&ofin, is, &src, &dst,
2513 					    NULL, FI_ICMPCMP);
2514 			if (is != NULL) {
2515 				if ((is->is_pass & FR_NOICMPERR) != 0) {
2516 					RWLOCK_EXIT(&ifs->ifs_ipf_state);
2517 					return NULL;
2518 				}
2519 				/*
2520 				 * i  : the index of this packet (the icmp
2521 				 *      unreachable)
2522 				 * oi : the index of the original packet found
2523 				 *      in the icmp header (i.e. the packet
2524 				 *      causing this icmp)
2525 				 * backward : original packet was backward
2526 				 *      compared to the state
2527 				 */
2528 				backward = IP6_NEQ(&is->is_src, &src);
2529 				fin->fin_rev = !backward;
2530 				i = (!backward << 1) + fin->fin_out;
2531 				oi = (backward << 1) + ofin.fin_out;
2532 				if (is->is_icmppkts[i] > is->is_pkts[oi])
2533 					continue;
2534 				ifs->ifs_ips_stats.iss_hits++;
2535 				is->is_icmppkts[i]++;
2536 				return is;
2537 			}
2538 		}
2539 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
2540 		return NULL;
2541 	case IPPROTO_TCP :
2542 	case IPPROTO_UDP :
2543 		break;
2544 	default :
2545 		return NULL;
2546 	}
2547 
2548 	tcp = (tcphdr_t *)((char *)oip + (IP_HL(oip) << 2));
2549 	dport = tcp->th_dport;
2550 	sport = tcp->th_sport;
2551 
2552 	hv = (pr = oip->ip_p);
2553 	src.in4 = oip->ip_src;
2554 	hv += src.in4.s_addr;
2555 	dst.in4 = oip->ip_dst;
2556 	hv += dst.in4.s_addr;
2557 	hv += dport;
2558 	hv += sport;
2559 	hv = DOUBLE_HASH(hv, ifs);
2560 
2561 	READ_ENTER(&ifs->ifs_ipf_state);
2562 	for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
2563 		isp = &is->is_hnext;
2564 		/*
2565 		 * Only allow this icmp though if the
2566 		 * encapsulated packet was allowed through the
2567 		 * other way around. Note that the minimal amount
2568 		 * of info present does not allow for checking against
2569 		 * tcp internals such as seq and ack numbers.   Only the
2570 		 * ports are known to be present and can be even if the
2571 		 * short flag is set.
2572 		 */
2573 		if ((is->is_p == pr) && (is->is_v == 4) &&
2574 		    (is = fr_matchsrcdst(&ofin, is, &src, &dst,
2575 					 tcp, FI_ICMPCMP))) {
2576 			/*
2577 			 * i  : the index of this packet (the icmp unreachable)
2578 			 * oi : the index of the original packet found in the
2579 			 *      icmp header (i.e. the packet causing this icmp)
2580 			 * backward : original packet was backward compared to
2581 			 *            the state
2582 			 */
2583 			backward = IP6_NEQ(&is->is_src, &src);
2584 			fin->fin_rev = !backward;
2585 			i = (!backward << 1) + fin->fin_out;
2586 			oi = (backward << 1) + ofin.fin_out;
2587 
2588 			if (((is->is_pass & FR_NOICMPERR) != 0) ||
2589 			    (is->is_icmppkts[i] > is->is_pkts[oi]))
2590 				break;
2591 			ifs->ifs_ips_stats.iss_hits++;
2592 			is->is_icmppkts[i]++;
2593 			/*
2594 			 * we deliberately do not touch the timeouts
2595 			 * for the accompanying state table entry.
2596 			 * It remains to be seen if that is correct. XXX
2597 			 */
2598 			return is;
2599 		}
2600 	}
2601 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
2602 	return NULL;
2603 }
2604 
2605 
2606 /* ------------------------------------------------------------------------ */
2607 /* Function:    fr_ipsmove                                                  */
2608 /* Returns:     Nil                                                         */
2609 /* Parameters:  is(I) - pointer to state table entry                        */
2610 /*              hv(I) - new hash value for state table entry                */
2611 /* Write Locks: ipf_state                                                   */
2612 /*                                                                          */
2613 /* Move a state entry from one position in the hash table to another.       */
2614 /* ------------------------------------------------------------------------ */
fr_ipsmove(is,hv,ifs)2615 static void fr_ipsmove(is, hv, ifs)
2616 ipstate_t *is;
2617 u_int hv;
2618 ipf_stack_t *ifs;
2619 {
2620 	ipstate_t **isp;
2621 	u_int hvm;
2622 
2623 	ASSERT(rw_read_locked(&ifs->ifs_ipf_state.ipf_lk) == 0);
2624 
2625 	hvm = is->is_hv;
2626 	/*
2627 	 * Remove the hash from the old location...
2628 	 */
2629 	isp = is->is_phnext;
2630 	if (is->is_hnext)
2631 		is->is_hnext->is_phnext = isp;
2632 	*isp = is->is_hnext;
2633 	if (ifs->ifs_ips_table[hvm] == NULL)
2634 		ifs->ifs_ips_stats.iss_inuse--;
2635 	ifs->ifs_ips_stats.iss_bucketlen[hvm]--;
2636 
2637 	/*
2638 	 * ...and put the hash in the new one.
2639 	 */
2640 	hvm = DOUBLE_HASH(hv, ifs);
2641 	is->is_hv = hvm;
2642 	isp = &ifs->ifs_ips_table[hvm];
2643 	if (*isp)
2644 		(*isp)->is_phnext = &is->is_hnext;
2645 	else
2646 		ifs->ifs_ips_stats.iss_inuse++;
2647 	ifs->ifs_ips_stats.iss_bucketlen[hvm]++;
2648 	is->is_phnext = isp;
2649 	is->is_hnext = *isp;
2650 	*isp = is;
2651 }
2652 
2653 
2654 /* ------------------------------------------------------------------------ */
2655 /* Function:    fr_stlookup                                                 */
2656 /* Returns:     ipstate_t* - NULL == no matching state found,               */
2657 /*                           else pointer to state information is returned  */
2658 /* Parameters:  fin(I) - pointer to packet information                      */
2659 /*              tcp(I) - pointer to TCP/UDP header.                         */
2660 /*                                                                          */
2661 /* Search the state table for a matching entry to the packet described by   */
2662 /* the contents of *fin.                                                    */
2663 /*                                                                          */
2664 /* If we return NULL then no lock on ipf_state is held.                     */
2665 /* If we return non-null then a read-lock on ipf_state is held.             */
2666 /* ------------------------------------------------------------------------ */
fr_stlookup(fin,tcp,ifqp)2667 ipstate_t *fr_stlookup(fin, tcp, ifqp)
2668 fr_info_t *fin;
2669 tcphdr_t *tcp;
2670 ipftq_t **ifqp;
2671 {
2672 	u_int hv, hvm, pr, v, tryagain;
2673 	ipstate_t *is, **isp;
2674 	u_short dport, sport;
2675 	i6addr_t src, dst;
2676 	struct icmp *ic;
2677 	ipftq_t *ifq;
2678 	int oow;
2679 	ipf_stack_t *ifs = fin->fin_ifs;
2680 
2681 	is = NULL;
2682 	ifq = NULL;
2683 	tcp = fin->fin_dp;
2684 	ic = (struct icmp *)tcp;
2685 	hv = (pr = fin->fin_fi.fi_p);
2686 	src = fin->fin_fi.fi_src;
2687 	dst = fin->fin_fi.fi_dst;
2688 	hv += src.in4.s_addr;
2689 	hv += dst.in4.s_addr;
2690 
2691 	v = fin->fin_fi.fi_v;
2692 #ifdef	USE_INET6
2693 	if (v == 6) {
2694 		hv  += fin->fin_fi.fi_src.i6[1];
2695 		hv  += fin->fin_fi.fi_src.i6[2];
2696 		hv  += fin->fin_fi.fi_src.i6[3];
2697 
2698 		if ((fin->fin_p == IPPROTO_ICMPV6) &&
2699 		    IN6_IS_ADDR_MULTICAST(&fin->fin_fi.fi_dst.in6)) {
2700 			hv -= dst.in4.s_addr;
2701 		} else {
2702 			hv += fin->fin_fi.fi_dst.i6[1];
2703 			hv += fin->fin_fi.fi_dst.i6[2];
2704 			hv += fin->fin_fi.fi_dst.i6[3];
2705 		}
2706 	}
2707 #endif
2708 	if ((v == 4) &&
2709 	    (fin->fin_flx & (FI_MULTICAST|FI_BROADCAST|FI_MBCAST))) {
2710 		if (fin->fin_out == 0) {
2711 			hv -= src.in4.s_addr;
2712 		} else {
2713 			hv -= dst.in4.s_addr;
2714 		}
2715 	}
2716 
2717 	/*
2718 	 * Search the hash table for matching packet header info.
2719 	 */
2720 	switch (pr)
2721 	{
2722 #ifdef	USE_INET6
2723 	case IPPROTO_ICMPV6 :
2724 		tryagain = 0;
2725 		if (v == 6) {
2726 			if ((ic->icmp_type == ICMP6_ECHO_REQUEST) ||
2727 			    (ic->icmp_type == ICMP6_ECHO_REPLY)) {
2728 				hv += ic->icmp_id;
2729 			}
2730 		}
2731 		READ_ENTER(&ifs->ifs_ipf_state);
2732 icmp6again:
2733 		hvm = DOUBLE_HASH(hv, ifs);
2734 		for (isp = &ifs->ifs_ips_table[hvm]; ((is = *isp) != NULL); ) {
2735 			isp = &is->is_hnext;
2736 			if ((is->is_p != pr) || (is->is_v != v))
2737 				continue;
2738 			is = fr_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP);
2739 			if (is != NULL &&
2740 			    fr_matchicmpqueryreply(v, &is->is_icmp,
2741 						   ic, fin->fin_rev)) {
2742 				if (fin->fin_rev)
2743 					ifq = &ifs->ifs_ips_icmpacktq;
2744 				else
2745 					ifq = &ifs->ifs_ips_icmptq;
2746 				break;
2747 			}
2748 		}
2749 
2750 		if (is != NULL) {
2751 			if ((tryagain != 0) && !(is->is_flags & SI_W_DADDR)) {
2752 				hv += fin->fin_fi.fi_src.i6[0];
2753 				hv += fin->fin_fi.fi_src.i6[1];
2754 				hv += fin->fin_fi.fi_src.i6[2];
2755 				hv += fin->fin_fi.fi_src.i6[3];
2756 				fr_ipsmove(is, hv, ifs);
2757 				MUTEX_DOWNGRADE(&ifs->ifs_ipf_state);
2758 			}
2759 			break;
2760 		}
2761 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
2762 
2763 		/*
2764 		 * No matching icmp state entry. Perhaps this is a
2765 		 * response to another state entry.
2766 		 *
2767 		 * XXX With some ICMP6 packets, the "other" address is already
2768 		 * in the packet, after the ICMP6 header, and this could be
2769 		 * used in place of the multicast address.  However, taking
2770 		 * advantage of this requires some significant code changes
2771 		 * to handle the specific types where that is the case.
2772 		 */
2773 		if ((ifs->ifs_ips_stats.iss_wild != 0) && (v == 6) && (tryagain == 0) &&
2774 		    !IN6_IS_ADDR_MULTICAST(&fin->fin_fi.fi_src.in6)) {
2775 			hv -= fin->fin_fi.fi_src.i6[0];
2776 			hv -= fin->fin_fi.fi_src.i6[1];
2777 			hv -= fin->fin_fi.fi_src.i6[2];
2778 			hv -= fin->fin_fi.fi_src.i6[3];
2779 			tryagain = 1;
2780 			WRITE_ENTER(&ifs->ifs_ipf_state);
2781 			goto icmp6again;
2782 		}
2783 
2784 		is = fr_checkicmp6matchingstate(fin);
2785 		if (is != NULL)
2786 			return is;
2787 		break;
2788 #endif
2789 
2790 	case IPPROTO_ICMP :
2791 		if (v == 4) {
2792 			hv += ic->icmp_id;
2793 		}
2794 		hv = DOUBLE_HASH(hv, ifs);
2795 		READ_ENTER(&ifs->ifs_ipf_state);
2796 		for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
2797 			isp = &is->is_hnext;
2798 			if ((is->is_p != pr) || (is->is_v != v))
2799 				continue;
2800 			is = fr_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP);
2801 			if (is != NULL &&
2802 			    fr_matchicmpqueryreply(v, &is->is_icmp,
2803 						   ic, fin->fin_rev)) {
2804 				if (fin->fin_rev)
2805 					ifq = &ifs->ifs_ips_icmpacktq;
2806 				else
2807 					ifq = &ifs->ifs_ips_icmptq;
2808 				break;
2809 			}
2810 		}
2811 		if (is == NULL) {
2812 			RWLOCK_EXIT(&ifs->ifs_ipf_state);
2813 		}
2814 		break;
2815 
2816 	case IPPROTO_TCP :
2817 	case IPPROTO_UDP :
2818 		ifqp = NULL;
2819 		sport = htons(fin->fin_data[0]);
2820 		hv += sport;
2821 		dport = htons(fin->fin_data[1]);
2822 		hv += dport;
2823 		oow = 0;
2824 		tryagain = 0;
2825 		READ_ENTER(&ifs->ifs_ipf_state);
2826 retry_tcpudp:
2827 		hvm = DOUBLE_HASH(hv, ifs);
2828 		for (isp = &ifs->ifs_ips_table[hvm]; ((is = *isp) != NULL); ) {
2829 			isp = &is->is_hnext;
2830 			if ((is->is_p != pr) || (is->is_v != v))
2831 				continue;
2832 			fin->fin_flx &= ~FI_OOW;
2833 			is = fr_matchsrcdst(fin, is, &src, &dst, tcp, FI_CMP);
2834 			if (is != NULL) {
2835 				if (pr == IPPROTO_TCP) {
2836 					if (!fr_tcpstate(fin, tcp, is)) {
2837 						oow |= fin->fin_flx & FI_OOW;
2838 						continue;
2839 					}
2840 				}
2841 				break;
2842 			}
2843 		}
2844 		if (is != NULL) {
2845 			if (tryagain &&
2846 			    !(is->is_flags & (SI_CLONE|SI_WILDP|SI_WILDA))) {
2847 				hv += dport;
2848 				hv += sport;
2849 				fr_ipsmove(is, hv, ifs);
2850 				MUTEX_DOWNGRADE(&ifs->ifs_ipf_state);
2851 			}
2852 			break;
2853 		}
2854 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
2855 
2856 		if (ifs->ifs_ips_stats.iss_wild) {
2857 			if (tryagain == 0) {
2858 				hv -= dport;
2859 				hv -= sport;
2860 			} else if (tryagain == 1) {
2861 				hv = fin->fin_fi.fi_p;
2862 				/*
2863 				 * If we try to pretend this is a reply to a
2864 				 * multicast/broadcast packet then we need to
2865 				 * exclude part of the address from the hash
2866 				 * calculation.
2867 				 */
2868 				if (fin->fin_out == 0) {
2869 					hv += src.in4.s_addr;
2870 				} else {
2871 					hv += dst.in4.s_addr;
2872 				}
2873 				hv += dport;
2874 				hv += sport;
2875 			}
2876 			tryagain++;
2877 			if (tryagain <= 2) {
2878 				WRITE_ENTER(&ifs->ifs_ipf_state);
2879 				goto retry_tcpudp;
2880 			}
2881 		}
2882 		fin->fin_flx |= oow;
2883 		break;
2884 
2885 #if 0
2886 	case IPPROTO_GRE :
2887 		gre = fin->fin_dp;
2888 		if (GRE_REV(gre->gr_flags) == 1) {
2889 			hv += gre->gr_call;
2890 		}
2891 		/* FALLTHROUGH */
2892 #endif
2893 	default :
2894 		ifqp = NULL;
2895 		hvm = DOUBLE_HASH(hv, ifs);
2896 		READ_ENTER(&ifs->ifs_ipf_state);
2897 		for (isp = &ifs->ifs_ips_table[hvm]; ((is = *isp) != NULL); ) {
2898 			isp = &is->is_hnext;
2899 			if ((is->is_p != pr) || (is->is_v != v))
2900 				continue;
2901 			is = fr_matchsrcdst(fin, is, &src, &dst, NULL, FI_CMP);
2902 			if (is != NULL) {
2903 				ifq = &ifs->ifs_ips_iptq;
2904 				break;
2905 			}
2906 		}
2907 		if (is == NULL) {
2908 			RWLOCK_EXIT(&ifs->ifs_ipf_state);
2909 		}
2910 		break;
2911 	}
2912 
2913 	if ((is != NULL) && ((is->is_sti.tqe_flags & TQE_RULEBASED) != 0) &&
2914 	    (is->is_tqehead[fin->fin_rev] != NULL))
2915 		ifq = is->is_tqehead[fin->fin_rev];
2916 	if (ifq != NULL && ifqp != NULL)
2917 		*ifqp = ifq;
2918 	return is;
2919 }
2920 
2921 
2922 /* ------------------------------------------------------------------------ */
2923 /* Function:    fr_updatestate                                              */
2924 /* Returns:     Nil                                                         */
2925 /* Parameters:  fin(I) - pointer to packet information                      */
2926 /*              is(I)  - pointer to state table entry                       */
2927 /* Read Locks:  ipf_state                                                   */
2928 /*                                                                          */
2929 /* Updates packet and byte counters for a newly received packet.  Seeds the */
2930 /* fragment cache with a new entry as required.                             */
2931 /* ------------------------------------------------------------------------ */
fr_updatestate(fin,is,ifq)2932 void fr_updatestate(fin, is, ifq)
2933 fr_info_t *fin;
2934 ipstate_t *is;
2935 ipftq_t *ifq;
2936 {
2937 	ipftqent_t *tqe;
2938 	int i, pass;
2939 	ipf_stack_t *ifs = fin->fin_ifs;
2940 
2941 	i = (fin->fin_rev << 1) + fin->fin_out;
2942 
2943 	/*
2944 	 * For TCP packets, ifq == NULL.  For all others, check if this new
2945 	 * queue is different to the last one it was on and move it if so.
2946 	 */
2947 	tqe = &is->is_sti;
2948 	MUTEX_ENTER(&is->is_lock);
2949 	if ((tqe->tqe_flags & TQE_RULEBASED) != 0)
2950 		ifq = is->is_tqehead[fin->fin_rev];
2951 
2952 	if (ifq != NULL)
2953 		fr_movequeue(tqe, tqe->tqe_ifq, ifq, ifs);
2954 
2955 	is->is_pkts[i]++;
2956 	fin->fin_pktnum = is->is_pkts[i] + is->is_icmppkts[i];
2957 	is->is_bytes[i] += fin->fin_plen;
2958 	MUTEX_EXIT(&is->is_lock);
2959 
2960 #ifdef	IPFILTER_SYNC
2961 	if (is->is_flags & IS_STATESYNC)
2962 		ipfsync_update(SMC_STATE, fin, is->is_sync);
2963 #endif
2964 
2965 	ATOMIC_INCL(ifs->ifs_ips_stats.iss_hits);
2966 
2967 	fin->fin_fr = is->is_rule;
2968 
2969 	/*
2970 	 * If this packet is a fragment and the rule says to track fragments,
2971 	 * then create a new fragment cache entry.
2972 	 */
2973 	pass = is->is_pass;
2974 	if ((fin->fin_flx & FI_FRAG) && FR_ISPASS(pass))
2975 		(void) fr_newfrag(fin, pass ^ FR_KEEPSTATE);
2976 }
2977 
2978 
2979 /* ------------------------------------------------------------------------ */
2980 /* Function:    fr_checkstate                                               */
2981 /* Returns:     frentry_t* - NULL == search failed,                         */
2982 /*                           else pointer to rule for matching state        */
2983 /* Parameters:  ifp(I)   - pointer to interface                             */
2984 /*              passp(I) - pointer to filtering result flags                */
2985 /*                                                                          */
2986 /* Check if a packet is associated with an entry in the state table.        */
2987 /* ------------------------------------------------------------------------ */
fr_checkstate(fin,passp)2988 frentry_t *fr_checkstate(fin, passp)
2989 fr_info_t *fin;
2990 u_32_t *passp;
2991 {
2992 	ipstate_t *is;
2993 	frentry_t *fr;
2994 	tcphdr_t *tcp;
2995 	ipftq_t *ifq;
2996 	u_int pass;
2997 	ipf_stack_t *ifs = fin->fin_ifs;
2998 
2999 	if (ifs->ifs_fr_state_lock || (ifs->ifs_ips_list == NULL) ||
3000 	    (fin->fin_flx & (FI_SHORT|FI_STATE|FI_FRAGBODY|FI_BAD)))
3001 		return NULL;
3002 
3003 	is = NULL;
3004 	if ((fin->fin_flx & FI_TCPUDP) ||
3005 	    (fin->fin_fi.fi_p == IPPROTO_ICMP)
3006 #ifdef	USE_INET6
3007 	    || (fin->fin_fi.fi_p == IPPROTO_ICMPV6)
3008 #endif
3009 	    )
3010 		tcp = fin->fin_dp;
3011 	else
3012 		tcp = NULL;
3013 
3014 	/*
3015 	 * Search the hash table for matching packet header info.
3016 	 */
3017 	ifq = NULL;
3018 	is = fr_stlookup(fin, tcp, &ifq);
3019 	switch (fin->fin_p)
3020 	{
3021 #ifdef	USE_INET6
3022 	case IPPROTO_ICMPV6 :
3023 		if (is != NULL)
3024 			break;
3025 		if (fin->fin_v == 6) {
3026 			is = fr_checkicmp6matchingstate(fin);
3027 			if (is != NULL)
3028 				goto matched;
3029 		}
3030 		break;
3031 #endif
3032 	case IPPROTO_ICMP :
3033 		if (is != NULL)
3034 			break;
3035 		/*
3036 		 * No matching icmp state entry. Perhaps this is a
3037 		 * response to another state entry.
3038 		 */
3039 		is = fr_checkicmpmatchingstate(fin);
3040 		if (is != NULL)
3041 			goto matched;
3042 		break;
3043 	case IPPROTO_TCP :
3044 		if (is == NULL)
3045 			break;
3046 
3047 		if (is->is_pass & FR_NEWISN) {
3048 			if (fin->fin_out == 0)
3049 				fr_fixinisn(fin, is);
3050 			else if (fin->fin_out == 1)
3051 				fr_fixoutisn(fin, is);
3052 		}
3053 		break;
3054 	default :
3055 		if (fin->fin_rev)
3056 			ifq = &ifs->ifs_ips_udpacktq;
3057 		else
3058 			ifq = &ifs->ifs_ips_udptq;
3059 		break;
3060 	}
3061 	if (is == NULL) {
3062 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_miss);
3063 		return NULL;
3064 	}
3065 
3066 matched:
3067 	fr = is->is_rule;
3068 	if (fr != NULL) {
3069 		if ((fin->fin_out == 0) && (fr->fr_nattag.ipt_num[0] != 0)) {
3070 			if (fin->fin_nattag == NULL) {
3071 				RWLOCK_EXIT(&ifs->ifs_ipf_state);
3072 				return NULL;
3073 			}
3074 			if (fr_matchtag(&fr->fr_nattag, fin->fin_nattag) != 0) {
3075 				RWLOCK_EXIT(&ifs->ifs_ipf_state);
3076 				return NULL;
3077 			}
3078 		}
3079 		(void) strncpy(fin->fin_group, fr->fr_group, FR_GROUPLEN);
3080 		fin->fin_icode = fr->fr_icode;
3081 	}
3082 
3083 	fin->fin_rule = is->is_rulen;
3084 	pass = is->is_pass;
3085 	fr_updatestate(fin, is, ifq);
3086 
3087 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
3088 	fin->fin_flx |= FI_STATE;
3089 	if ((pass & FR_LOGFIRST) != 0)
3090 		pass &= ~(FR_LOGFIRST|FR_LOG);
3091 	*passp = pass;
3092 	return fr;
3093 }
3094 
3095 
3096 /* ------------------------------------------------------------------------ */
3097 /* Function:    fr_fixoutisn                                                */
3098 /* Returns:     Nil                                                         */
3099 /* Parameters:  fin(I)   - pointer to packet information                    */
3100 /*              is(I)  - pointer to master state structure                  */
3101 /*                                                                          */
3102 /* Called only for outbound packets, adjusts the sequence number and the    */
3103 /* TCP checksum to match that change.                                       */
3104 /* ------------------------------------------------------------------------ */
fr_fixoutisn(fin,is)3105 static void fr_fixoutisn(fin, is)
3106 fr_info_t *fin;
3107 ipstate_t *is;
3108 {
3109 	tcphdr_t *tcp;
3110 	int rev;
3111 	u_32_t seq;
3112 
3113 	tcp = fin->fin_dp;
3114 	rev = fin->fin_rev;
3115 	if ((is->is_flags & IS_ISNSYN) != 0) {
3116 		if (rev == 0) {
3117 			seq = ntohl(tcp->th_seq);
3118 			seq += is->is_isninc[0];
3119 			tcp->th_seq = htonl(seq);
3120 			fix_outcksum(&tcp->th_sum, is->is_sumd[0]);
3121 		}
3122 	}
3123 	if ((is->is_flags & IS_ISNACK) != 0) {
3124 		if (rev == 1) {
3125 			seq = ntohl(tcp->th_seq);
3126 			seq += is->is_isninc[1];
3127 			tcp->th_seq = htonl(seq);
3128 			fix_outcksum(&tcp->th_sum, is->is_sumd[1]);
3129 		}
3130 	}
3131 }
3132 
3133 
3134 /* ------------------------------------------------------------------------ */
3135 /* Function:    fr_fixinisn                                                 */
3136 /* Returns:     Nil                                                         */
3137 /* Parameters:  fin(I)   - pointer to packet information                    */
3138 /*              is(I)  - pointer to master state structure                  */
3139 /*                                                                          */
3140 /* Called only for inbound packets, adjusts the acknowledge number and the  */
3141 /* TCP checksum to match that change.                                       */
3142 /* ------------------------------------------------------------------------ */
fr_fixinisn(fin,is)3143 static void fr_fixinisn(fin, is)
3144 fr_info_t *fin;
3145 ipstate_t *is;
3146 {
3147 	tcphdr_t *tcp;
3148 	int rev;
3149 	u_32_t ack;
3150 
3151 	tcp = fin->fin_dp;
3152 	rev = fin->fin_rev;
3153 	if ((is->is_flags & IS_ISNSYN) != 0) {
3154 		if (rev == 1) {
3155 			ack = ntohl(tcp->th_ack);
3156 			ack -= is->is_isninc[0];
3157 			tcp->th_ack = htonl(ack);
3158 			fix_incksum(&tcp->th_sum, is->is_sumd[0]);
3159 		}
3160 	}
3161 	if ((is->is_flags & IS_ISNACK) != 0) {
3162 		if (rev == 0) {
3163 			ack = ntohl(tcp->th_ack);
3164 			ack -= is->is_isninc[1];
3165 			tcp->th_ack = htonl(ack);
3166 			fix_incksum(&tcp->th_sum, is->is_sumd[1]);
3167 		}
3168 	}
3169 }
3170 
3171 
3172 /* ------------------------------------------------------------------------ */
3173 /* Function:    fr_statesync                                                */
3174 /* Returns:     Nil                                                         */
3175 /* Parameters:  action(I) - type of synchronisation to do                   */
3176 /*              v(I)      - IP version being sync'd (v4 or v6)              */
3177 /*              ifp(I)    - interface identifier associated with action     */
3178 /*              name(I)   - name associated with ifp parameter              */
3179 /*                                                                          */
3180 /* Walk through all state entries and if an interface pointer match is      */
3181 /* found then look it up again, based on its name in case the pointer has   */
3182 /* changed since last time.                                                 */
3183 /*                                                                          */
3184 /* If ifp is passed in as being non-null then we are only doing updates for */
3185 /* existing, matching, uses of it.                                          */
3186 /* ------------------------------------------------------------------------ */
fr_statesync(action,v,ifp,name,ifs)3187 void fr_statesync(action, v, ifp, name, ifs)
3188 int action, v;
3189 void *ifp;
3190 char *name;
3191 ipf_stack_t *ifs;
3192 {
3193 	ipstate_t *is;
3194 	int i;
3195 
3196 	if (ifs->ifs_fr_running <= 0)
3197 		return;
3198 
3199 	WRITE_ENTER(&ifs->ifs_ipf_state);
3200 
3201 	if (ifs->ifs_fr_running <= 0) {
3202 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
3203 		return;
3204 	}
3205 
3206 	switch (action)
3207 	{
3208 	case IPFSYNC_RESYNC :
3209 		for (is = ifs->ifs_ips_list; is; is = is->is_next) {
3210 			if (v != 0 && is->is_v != v)
3211 				continue;
3212 			/*
3213 			 * Look up all the interface names in the state entry.
3214 			 */
3215 			for (i = 0; i < 4; i++) {
3216 				is->is_ifp[i] = fr_resolvenic(is->is_ifname[i],
3217 							      is->is_v, ifs);
3218 			}
3219 		}
3220 		break;
3221 	case IPFSYNC_NEWIFP :
3222 		for (is = ifs->ifs_ips_list; is; is = is->is_next) {
3223 			if (v != 0 && is->is_v != v)
3224 				continue;
3225 			/*
3226 			 * Look up all the interface names in the state entry.
3227 			 */
3228 			for (i = 0; i < 4; i++) {
3229 				if (!strncmp(is->is_ifname[i], name,
3230 					     sizeof(is->is_ifname[i])))
3231 					is->is_ifp[i] = ifp;
3232 			}
3233 		}
3234 		break;
3235 	case IPFSYNC_OLDIFP :
3236 		for (is = ifs->ifs_ips_list; is; is = is->is_next) {
3237 			if (v != 0 && is->is_v != v)
3238 				continue;
3239 			/*
3240 			 * Look up all the interface names in the state entry.
3241 			 */
3242 			for (i = 0; i < 4; i++) {
3243 				if (is->is_ifp[i] == ifp)
3244 					is->is_ifp[i] = (void *)-1;
3245 			}
3246 		}
3247 		break;
3248 	}
3249 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
3250 }
3251 
3252 
3253 #if SOLARIS2 >= 10
3254 /* ------------------------------------------------------------------------ */
3255 /* Function:    fr_stateifindexsync					    */
3256 /* Returns:     void							    */
3257 /* Parameters:	ifp	- current network interface descriptor (ifindex)    */
3258 /*              newifp	- new interface descriptor (new ifindex)	    */
3259 /*		ifs	- pointer to IPF stack				    */
3260 /*									    */
3261 /* Write Locks: assumes ipf_mutex is locked				    */
3262 /*                                                                          */
3263 /* Updates all interface indeces matching to ifp with new interface index   */
3264 /* value.								    */
3265 /* ------------------------------------------------------------------------ */
fr_stateifindexsync(ifp,newifp,ifs)3266 void fr_stateifindexsync(ifp, newifp, ifs)
3267 void *ifp;
3268 void *newifp;
3269 ipf_stack_t *ifs;
3270 {
3271 	ipstate_t *is;
3272 	int i;
3273 
3274 	WRITE_ENTER(&ifs->ifs_ipf_state);
3275 
3276 	for (is = ifs->ifs_ips_list; is != NULL; is = is->is_next) {
3277 
3278 		for (i = 0; i < 4; i++) {
3279 			if (is->is_ifp[i] == ifp)
3280 				is->is_ifp[i] = newifp;
3281 		}
3282 	}
3283 
3284 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
3285 }
3286 #endif
3287 
3288 /* ------------------------------------------------------------------------ */
3289 /* Function:    fr_delstate                                                 */
3290 /* Returns:     int - 0 = entry deleted, else ref count on entry            */
3291 /* Parameters:  is(I)  - pointer to state structure to delete               */
3292 /*              why(I) - if not 0, log reason why it was deleted            */
3293 /*              ifs    - ipf stack instance                                 */
3294 /* Write Locks: ipf_state/ipf_global                                        */
3295 /*                                                                          */
3296 /* Deletes a state entry from the enumerated list as well as the hash table */
3297 /* and timeout queue lists.  Make adjustments to hash table statistics and  */
3298 /* global counters as required.                                             */
3299 /* ------------------------------------------------------------------------ */
fr_delstate(is,why,ifs)3300 int fr_delstate(is, why, ifs)
3301 ipstate_t *is;
3302 int why;
3303 ipf_stack_t *ifs;
3304 {
3305 	int removed = 0;
3306 
3307 	ASSERT(rw_write_held(&ifs->ifs_ipf_global.ipf_lk) == 0 ||
3308 		rw_write_held(&ifs->ifs_ipf_state.ipf_lk) == 0);
3309 
3310 	/*
3311 	 * Start by removing the entry from the hash table of state entries
3312 	 * so it will not be "used" again.
3313 	 *
3314 	 * It will remain in the "list" of state entries until all references
3315 	 * have been accounted for.
3316 	 */
3317 	if (is->is_phnext != NULL) {
3318 		removed = 1;
3319 		*is->is_phnext = is->is_hnext;
3320 		if (is->is_hnext != NULL)
3321 			is->is_hnext->is_phnext = is->is_phnext;
3322 		if (ifs->ifs_ips_table[is->is_hv] == NULL)
3323 			ifs->ifs_ips_stats.iss_inuse--;
3324 		ifs->ifs_ips_stats.iss_bucketlen[is->is_hv]--;
3325 
3326 		is->is_phnext = NULL;
3327 		is->is_hnext = NULL;
3328 	}
3329 
3330 	/*
3331 	 * Because ifs->ifs_ips_stats.iss_wild is a count of entries in the state
3332 	 * table that have wildcard flags set, only decerement it once
3333 	 * and do it here.
3334 	 */
3335 	if (is->is_flags & (SI_WILDP|SI_WILDA)) {
3336 		if (!(is->is_flags & SI_CLONED)) {
3337 			ATOMIC_DECL(ifs->ifs_ips_stats.iss_wild);
3338 		}
3339 		is->is_flags &= ~(SI_WILDP|SI_WILDA);
3340 	}
3341 
3342 	/*
3343 	 * Next, remove it from the timeout queue it is in.
3344 	 */
3345 	fr_deletequeueentry(&is->is_sti);
3346 
3347 	is->is_me = NULL;
3348 
3349 	/*
3350 	 * If it is still in use by something else, do not go any further,
3351 	 * but note that at this point it is now an orphan.
3352 	 */
3353 	MUTEX_ENTER(&is->is_lock);
3354 	if (is->is_ref > 1) {
3355 		is->is_ref--;
3356 		MUTEX_EXIT(&is->is_lock);
3357 		if (removed)
3358 			ifs->ifs_ips_stats.iss_orphans++;
3359 		return (is->is_ref);
3360 	}
3361 	MUTEX_EXIT(&is->is_lock);
3362 
3363 	is->is_ref = 0;
3364 
3365 	/*
3366 	 * If entry has already been removed from table,
3367 	 * it means we're simply cleaning up an orphan.
3368 	 */
3369 	if (!removed)
3370 		ifs->ifs_ips_stats.iss_orphans--;
3371 
3372 	if (is->is_tqehead[0] != NULL)
3373 		(void) fr_deletetimeoutqueue(is->is_tqehead[0]);
3374 
3375 	if (is->is_tqehead[1] != NULL)
3376 		(void) fr_deletetimeoutqueue(is->is_tqehead[1]);
3377 
3378 #ifdef	IPFILTER_SYNC
3379 	if (is->is_sync)
3380 		ipfsync_del(is->is_sync);
3381 #endif
3382 #ifdef	IPFILTER_SCAN
3383 	(void) ipsc_detachis(is);
3384 #endif
3385 
3386 	/*
3387 	 * Now remove it from master list of state table entries.
3388 	 */
3389 	if (is->is_pnext != NULL) {
3390 		*is->is_pnext = is->is_next;
3391 		if (is->is_next != NULL) {
3392 			is->is_next->is_pnext = is->is_pnext;
3393 			is->is_next = NULL;
3394 		}
3395 		is->is_pnext = NULL;
3396 	}
3397 
3398 	if (ifs->ifs_ipstate_logging != 0 && why != 0)
3399 		ipstate_log(is, why, ifs);
3400 
3401 	if (is->is_rule != NULL) {
3402 		is->is_rule->fr_statecnt--;
3403 		(void)fr_derefrule(&is->is_rule, ifs);
3404 	}
3405 
3406 	MUTEX_DESTROY(&is->is_lock);
3407 	KFREE(is);
3408 	ifs->ifs_ips_num--;
3409 
3410 	return (0);
3411 }
3412 
3413 
3414 /* ------------------------------------------------------------------------ */
3415 /* Function:    fr_timeoutstate                                             */
3416 /* Returns:     Nil                                                         */
3417 /* Parameters:  ifs - ipf stack instance                                    */
3418 /*                                                                          */
3419 /* Slowly expire held state for thingslike UDP and ICMP.  The algorithm     */
3420 /* used here is to keep the queue sorted with the oldest things at the top  */
3421 /* and the youngest at the bottom.  So if the top one doesn't need to be    */
3422 /* expired then neither will any under it.                                  */
3423 /* ------------------------------------------------------------------------ */
fr_timeoutstate(ifs)3424 void fr_timeoutstate(ifs)
3425 ipf_stack_t *ifs;
3426 {
3427 	ipftq_t *ifq, *ifqnext;
3428 	ipftqent_t *tqe, *tqn;
3429 	ipstate_t *is;
3430 	SPL_INT(s);
3431 
3432 	SPL_NET(s);
3433 	WRITE_ENTER(&ifs->ifs_ipf_state);
3434 	for (ifq = ifs->ifs_ips_tqtqb; ifq != NULL; ifq = ifq->ifq_next)
3435 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); ) {
3436 			if (tqe->tqe_die > ifs->ifs_fr_ticks)
3437 				break;
3438 			tqn = tqe->tqe_next;
3439 			is = tqe->tqe_parent;
3440 			(void) fr_delstate(is, ISL_EXPIRE, ifs);
3441 		}
3442 
3443 	for (ifq = ifs->ifs_ips_utqe; ifq != NULL; ifq = ifq->ifq_next) {
3444 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); ) {
3445 			if (tqe->tqe_die > ifs->ifs_fr_ticks)
3446 				break;
3447 			tqn = tqe->tqe_next;
3448 			is = tqe->tqe_parent;
3449 			(void) fr_delstate(is, ISL_EXPIRE, ifs);
3450 		}
3451 	}
3452 
3453 	for (ifq = ifs->ifs_ips_utqe; ifq != NULL; ifq = ifqnext) {
3454 		ifqnext = ifq->ifq_next;
3455 
3456 		if (((ifq->ifq_flags & IFQF_DELETE) != 0) &&
3457 		    (ifq->ifq_ref == 0)) {
3458 			fr_freetimeoutqueue(ifq, ifs);
3459 		}
3460 	}
3461 
3462 	if (ifs->ifs_fr_state_doflush) {
3463 		(void) fr_state_flush(FLUSH_TABLE_EXTRA, 0, ifs);
3464 		ifs->ifs_fr_state_doflush = 0;
3465 	}
3466 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
3467 	SPL_X(s);
3468 }
3469 
3470 
3471 /* ---------------------------------------------------------------------- */
3472 /* Function:    fr_state_flush                                            */
3473 /* Returns:     int - 0 == success, -1 == failure                         */
3474 /* Parameters:  flush_option - how to flush the active State table	  */
3475 /*              proto    - IP version to flush (4, 6, or both)            */
3476 /*              ifs      - ipf stack instance                             */
3477 /* Write Locks: ipf_state                                                 */
3478 /*                                                                        */
3479 /* Flush state tables.  Three possible flush options currently defined:	  */
3480 /*                                                                        */
3481 /* FLUSH_TABLE_ALL	: Flush all state table entries			  */
3482 /*                                                                        */
3483 /* FLUSH_TABLE_CLOSING	: Flush entries with TCP connections which	  */
3484 /*			  have started to close on both ends using	  */
3485 /*			  ipf_flushclosing().				  */
3486 /*                                                                        */
3487 /* FLUSH_TABLE_EXTRA	: First, flush entries which are "almost" closed. */
3488 /*			  Then, if needed, flush entries with TCP	  */
3489 /*			  connections which have been idle for a long	  */
3490 /*			  time with ipf_extraflush().			  */
3491 /* ---------------------------------------------------------------------- */
fr_state_flush(flush_option,proto,ifs)3492 static int fr_state_flush(flush_option, proto, ifs)
3493 int flush_option, proto;
3494 ipf_stack_t *ifs;
3495 {
3496 	ipstate_t *is, *isn;
3497 	int removed;
3498 	SPL_INT(s);
3499 
3500 	removed = 0;
3501 
3502 	SPL_NET(s);
3503 	switch (flush_option)
3504 	{
3505 	case FLUSH_TABLE_ALL:
3506 		isn = ifs->ifs_ips_list;
3507 		while ((is = isn) != NULL) {
3508 			isn = is->is_next;
3509 			if ((proto != 0) && (is->is_v != proto))
3510 				continue;
3511 			if (fr_delstate(is, ISL_FLUSH, ifs) == 0)
3512 				removed++;
3513 		}
3514 		break;
3515 
3516 	case FLUSH_TABLE_CLOSING:
3517 		removed = ipf_flushclosing(STATE_FLUSH,
3518 					   IPF_TCPS_CLOSE_WAIT,
3519 					   ifs->ifs_ips_tqtqb,
3520 					   ifs->ifs_ips_utqe,
3521 					   ifs);
3522 		break;
3523 
3524 	case FLUSH_TABLE_EXTRA:
3525 		removed = ipf_flushclosing(STATE_FLUSH,
3526 					   IPF_TCPS_FIN_WAIT_2,
3527 					   ifs->ifs_ips_tqtqb,
3528 					   ifs->ifs_ips_utqe,
3529 					   ifs);
3530 
3531 		/*
3532 		 * Be sure we haven't done this in the last 10 seconds.
3533 		 */
3534 		if (ifs->ifs_fr_ticks - ifs->ifs_ips_last_force_flush <
3535 		    IPF_TTLVAL(10))
3536 			break;
3537 		ifs->ifs_ips_last_force_flush = ifs->ifs_fr_ticks;
3538                 removed += ipf_extraflush(STATE_FLUSH,
3539 					  &ifs->ifs_ips_tqtqb[IPF_TCPS_ESTABLISHED],
3540 					  ifs->ifs_ips_utqe,
3541 					  ifs);
3542 		break;
3543 
3544 	default: /* Flush Nothing */
3545 		break;
3546 	}
3547 
3548 	SPL_X(s);
3549 	return (removed);
3550 }
3551 
3552 
3553 /* ------------------------------------------------------------------------ */
3554 /* Function:    fr_tcp_age                                                  */
3555 /* Returns:     int - 1 == state transition made, 0 == no change (rejected) */
3556 /* Parameters:  tq(I)    - pointer to timeout queue information             */
3557 /*              fin(I)   - pointer to packet information                    */
3558 /*              tqtab(I) - TCP timeout queue table this is in               */
3559 /*              flags(I) - flags from state/NAT entry                       */
3560 /*                                                                          */
3561 /* Rewritten by Arjan de Vet <Arjan.deVet@adv.iae.nl>, 2000-07-29:          */
3562 /*                                                                          */
3563 /* - (try to) base state transitions on real evidence only,                 */
3564 /*   i.e. packets that are sent and have been received by ipfilter;         */
3565 /*   diagram 18.12 of TCP/IP volume 1 by W. Richard Stevens was used.       */
3566 /*                                                                          */
3567 /* - deal with half-closed connections correctly;                           */
3568 /*                                                                          */
3569 /* - store the state of the source in state[0] such that ipfstat            */
3570 /*   displays the state as source/dest instead of dest/source; the calls    */
3571 /*   to fr_tcp_age have been changed accordingly.                           */
3572 /*                                                                          */
3573 /* Internal Parameters:                                                     */
3574 /*                                                                          */
3575 /*    state[0] = state of source (host that initiated connection)           */
3576 /*    state[1] = state of dest   (host that accepted the connection)        */
3577 /*                                                                          */
3578 /*    dir == 0 : a packet from source to dest                               */
3579 /*    dir == 1 : a packet from dest to source                               */
3580 /*                                                                          */
3581 /* Locking: it is assumed that the parent of the tqe structure is locked.   */
3582 /* ------------------------------------------------------------------------ */
fr_tcp_age(tqe,fin,tqtab,flags)3583 int fr_tcp_age(tqe, fin, tqtab, flags)
3584 ipftqent_t *tqe;
3585 fr_info_t *fin;
3586 ipftq_t *tqtab;
3587 int flags;
3588 {
3589 	int dlen, ostate, nstate, rval, dir;
3590 	u_char tcpflags;
3591 	tcphdr_t *tcp;
3592 	ipf_stack_t *ifs = fin->fin_ifs;
3593 
3594 	tcp = fin->fin_dp;
3595 
3596 	rval = 0;
3597 	dir = fin->fin_rev;
3598 	tcpflags = tcp->th_flags;
3599 	dlen = fin->fin_dlen - (TCP_OFF(tcp) << 2);
3600 
3601 	ostate = tqe->tqe_state[1 - dir];
3602 	nstate = tqe->tqe_state[dir];
3603 
3604 	DTRACE_PROBE4(
3605 		indata,
3606 		fr_info_t *, fin,
3607 		int, ostate,
3608 		int, nstate,
3609 		u_char, tcpflags
3610 	);
3611 
3612 	if (tcpflags & TH_RST) {
3613 		if (!(tcpflags & TH_PUSH) && !dlen)
3614 			nstate = IPF_TCPS_CLOSED;
3615 		else
3616 			nstate = IPF_TCPS_CLOSE_WAIT;
3617 
3618 		/*
3619 		 * Once RST is received, we must advance peer's state to
3620 		 * CLOSE_WAIT.
3621 		 */
3622 		if (ostate <= IPF_TCPS_ESTABLISHED) {
3623 			tqe->tqe_state[1 - dir] = IPF_TCPS_CLOSE_WAIT;
3624 		}
3625 		rval = 1;
3626 	} else {
3627 
3628 		switch (nstate)
3629 		{
3630 		case IPF_TCPS_LISTEN: /* 0 */
3631 			if ((tcpflags & TH_OPENING) == TH_OPENING) {
3632 				/*
3633 				 * 'dir' received an S and sends SA in
3634 				 * response, CLOSED -> SYN_RECEIVED
3635 				 */
3636 				nstate = IPF_TCPS_SYN_RECEIVED;
3637 				rval = 1;
3638 			} else if ((tcpflags & TH_OPENING) == TH_SYN) {
3639 				/* 'dir' sent S, CLOSED -> SYN_SENT */
3640 				nstate = IPF_TCPS_SYN_SENT;
3641 				rval = 1;
3642 			}
3643 			/*
3644 			 * the next piece of code makes it possible to get
3645 			 * already established connections into the state table
3646 			 * after a restart or reload of the filter rules; this
3647 			 * does not work when a strict 'flags S keep state' is
3648 			 * used for tcp connections of course
3649 			 */
3650 			if (((flags & IS_TCPFSM) == 0) &&
3651 			    ((tcpflags & TH_ACKMASK) == TH_ACK)) {
3652 				/*
3653 				 * we saw an A, guess 'dir' is in ESTABLISHED
3654 				 * mode
3655 				 */
3656 				switch (ostate)
3657 				{
3658 				case IPF_TCPS_LISTEN :
3659 				case IPF_TCPS_SYN_RECEIVED :
3660 					nstate = IPF_TCPS_HALF_ESTAB;
3661 					rval = 1;
3662 					break;
3663 				case IPF_TCPS_HALF_ESTAB :
3664 				case IPF_TCPS_ESTABLISHED :
3665 					nstate = IPF_TCPS_ESTABLISHED;
3666 					rval = 1;
3667 					break;
3668 				default :
3669 					break;
3670 				}
3671 			}
3672 			/*
3673 			 * TODO: besides regular ACK packets we can have other
3674 			 * packets as well; it is yet to be determined how we
3675 			 * should initialize the states in those cases
3676 			 */
3677 			break;
3678 
3679 		case IPF_TCPS_SYN_SENT: /* 1 */
3680 			if ((tcpflags & ~(TH_ECN|TH_CWR)) == TH_SYN) {
3681 				/*
3682 				 * A retransmitted SYN packet.  We do not reset
3683 				 * the timeout here to fr_tcptimeout because a
3684 				 * connection connect timeout does not renew
3685 				 * after every packet that is sent.  We need to
3686 				 * set rval so as to indicate the packet has
3687 				 * passed the check for its flags being valid
3688 				 * in the TCP FSM.  Setting rval to 2 has the
3689 				 * result of not resetting the timeout.
3690 				 */
3691 				rval = 2;
3692 			} else if ((tcpflags & (TH_SYN|TH_FIN|TH_ACK)) ==
3693 				   TH_ACK) {
3694 				/*
3695 				 * we see an A from 'dir' which is in SYN_SENT
3696 				 * state: 'dir' sent an A in response to an SA
3697 				 * which it received, SYN_SENT -> ESTABLISHED
3698 				 */
3699 				nstate = IPF_TCPS_ESTABLISHED;
3700 				rval = 1;
3701 			} else if (tcpflags & TH_FIN) {
3702 				/*
3703 				 * we see an F from 'dir' which is in SYN_SENT
3704 				 * state and wants to close its side of the
3705 				 * connection; SYN_SENT -> FIN_WAIT_1
3706 				 */
3707 				nstate = IPF_TCPS_FIN_WAIT_1;
3708 				rval = 1;
3709 			} else if ((tcpflags & TH_OPENING) == TH_OPENING) {
3710 				/*
3711 				 * we see an SA from 'dir' which is already in
3712 				 * SYN_SENT state, this means we have a
3713 				 * simultaneous open; SYN_SENT -> SYN_RECEIVED
3714 				 */
3715 				nstate = IPF_TCPS_SYN_RECEIVED;
3716 				rval = 1;
3717 			}
3718 			break;
3719 
3720 		case IPF_TCPS_SYN_RECEIVED: /* 2 */
3721 			if ((tcpflags & (TH_SYN|TH_FIN|TH_ACK)) == TH_ACK) {
3722 				/*
3723 				 * we see an A from 'dir' which was in
3724 				 * SYN_RECEIVED state so it must now be in
3725 				 * established state, SYN_RECEIVED ->
3726 				 * ESTABLISHED
3727 				 */
3728 				nstate = IPF_TCPS_ESTABLISHED;
3729 				rval = 1;
3730 			} else if ((tcpflags & ~(TH_ECN|TH_CWR)) ==
3731 				   TH_OPENING) {
3732 				/*
3733 				 * We see an SA from 'dir' which is already in
3734 				 * SYN_RECEIVED state.
3735 				 */
3736 				rval = 2;
3737 			} else if (tcpflags & TH_FIN) {
3738 				/*
3739 				 * we see an F from 'dir' which is in
3740 				 * SYN_RECEIVED state and wants to close its
3741 				 * side of the connection; SYN_RECEIVED ->
3742 				 * FIN_WAIT_1
3743 				 */
3744 				nstate = IPF_TCPS_FIN_WAIT_1;
3745 				rval = 1;
3746 			}
3747 			break;
3748 
3749 		case IPF_TCPS_HALF_ESTAB: /* 3 */
3750 			if (tcpflags & TH_FIN) {
3751 				nstate = IPF_TCPS_FIN_WAIT_1;
3752 				rval = 1;
3753 			} else if ((tcpflags & TH_ACKMASK) == TH_ACK) {
3754 				/*
3755 				 * If we've picked up a connection in mid
3756 				 * flight, we could be looking at a follow on
3757 				 * packet from the same direction as the one
3758 				 * that created this state.  Recognise it but
3759 				 * do not advance the entire connection's
3760 				 * state.
3761 				 */
3762 				switch (ostate)
3763 				{
3764 				case IPF_TCPS_LISTEN :
3765 				case IPF_TCPS_SYN_SENT :
3766 				case IPF_TCPS_SYN_RECEIVED :
3767 					rval = 1;
3768 					break;
3769 				case IPF_TCPS_HALF_ESTAB :
3770 				case IPF_TCPS_ESTABLISHED :
3771 					nstate = IPF_TCPS_ESTABLISHED;
3772 					rval = 1;
3773 					break;
3774 				default :
3775 					break;
3776 				}
3777 			}
3778 			break;
3779 
3780 		case IPF_TCPS_ESTABLISHED: /* 4 */
3781 			rval = 1;
3782 			if (tcpflags & TH_FIN) {
3783 				/*
3784 				 * 'dir' closed its side of the connection;
3785 				 * this gives us a half-closed connection;
3786 				 * ESTABLISHED -> FIN_WAIT_1
3787 				 */
3788 				if (ostate == IPF_TCPS_FIN_WAIT_1) {
3789 					nstate = IPF_TCPS_CLOSING;
3790 				} else {
3791 					nstate = IPF_TCPS_FIN_WAIT_1;
3792 				}
3793 			} else if (tcpflags & TH_ACK) {
3794 				/*
3795 				 * an ACK, should we exclude other flags here?
3796 				 */
3797 				if (ostate == IPF_TCPS_FIN_WAIT_1) {
3798 					/*
3799 					 * We know the other side did an active
3800 					 * close, so we are ACKing the recvd
3801 					 * FIN packet (does the window matching
3802 					 * code guarantee this?) and go into
3803 					 * CLOSE_WAIT state; this gives us a
3804 					 * half-closed connection
3805 					 */
3806 					nstate = IPF_TCPS_CLOSE_WAIT;
3807 				} else if (ostate < IPF_TCPS_CLOSE_WAIT) {
3808 					/*
3809 					 * still a fully established
3810 					 * connection reset timeout
3811 					 */
3812 					nstate = IPF_TCPS_ESTABLISHED;
3813 				}
3814 			}
3815 			break;
3816 
3817 		case IPF_TCPS_CLOSE_WAIT: /* 5 */
3818 			rval = 1;
3819 			if (tcpflags & TH_FIN) {
3820 				/*
3821 				 * application closed and 'dir' sent a FIN,
3822 				 * we're now going into LAST_ACK state
3823 				 */
3824 				nstate = IPF_TCPS_LAST_ACK;
3825 			} else {
3826 				/*
3827 				 * we remain in CLOSE_WAIT because the other
3828 				 * side has closed already and we did not
3829 				 * close our side yet; reset timeout
3830 				 */
3831 				nstate = IPF_TCPS_CLOSE_WAIT;
3832 			}
3833 			break;
3834 
3835 		case IPF_TCPS_FIN_WAIT_1: /* 6 */
3836 			rval = 1;
3837 			if ((tcpflags & TH_ACK) &&
3838 			    ostate > IPF_TCPS_CLOSE_WAIT) {
3839 				/*
3840 				 * if the other side is not active anymore
3841 				 * it has sent us a FIN packet that we are
3842 				 * ack'ing now with an ACK; this means both
3843 				 * sides have now closed the connection and
3844 				 * we go into LAST_ACK
3845 				 */
3846 				/*
3847 				 * XXX: how do we know we really are ACKing
3848 				 * the FIN packet here? does the window code
3849 				 * guarantee that?
3850 				 */
3851 				nstate = IPF_TCPS_LAST_ACK;
3852 			} else {
3853 				/*
3854 				 * we closed our side of the connection
3855 				 * already but the other side is still active
3856 				 * (ESTABLISHED/CLOSE_WAIT); continue with
3857 				 * this half-closed connection
3858 				 */
3859 				nstate = IPF_TCPS_FIN_WAIT_1;
3860 			}
3861 			break;
3862 
3863 		case IPF_TCPS_CLOSING: /* 7 */
3864 			if ((tcpflags & (TH_FIN|TH_ACK)) == TH_ACK) {
3865 				nstate = IPF_TCPS_TIME_WAIT;
3866 			}
3867 			rval = 1;
3868 			break;
3869 
3870 		case IPF_TCPS_LAST_ACK: /* 8 */
3871 			/*
3872 			 * We want to reset timer here to keep state in table.
3873 			 * If we would allow the state to time out here, while
3874 			 * there would still be packets being retransmitted, we
3875 			 * would cut off line between the two peers preventing
3876 			 * them to close connection properly.
3877 			 */
3878 			rval = 1;
3879 			break;
3880 
3881 		case IPF_TCPS_FIN_WAIT_2: /* 9 */
3882 			/* NOT USED */
3883 			break;
3884 
3885 		case IPF_TCPS_TIME_WAIT: /* 10 */
3886 			/* we're in 2MSL timeout now */
3887 			if (ostate == IPF_TCPS_LAST_ACK) {
3888 				nstate = IPF_TCPS_CLOSED;
3889 				rval = 1;
3890 			} else {
3891 				rval = 2;
3892 			}
3893 			break;
3894 
3895 		case IPF_TCPS_CLOSED: /* 11 */
3896 			rval = 2;
3897 			break;
3898 
3899 		default :
3900 #if defined(_KERNEL)
3901 			ASSERT(nstate >= IPF_TCPS_LISTEN &&
3902 			    nstate <= IPF_TCPS_CLOSED);
3903 #else
3904 			abort();
3905 #endif
3906 			break;
3907 		}
3908 	}
3909 
3910 	/*
3911 	 * If rval == 2 then do not update the queue position, but treat the
3912 	 * packet as being ok.
3913 	 */
3914 	if (rval == 2) {
3915 		DTRACE_PROBE1(state_keeping_timer, int, nstate);
3916 		rval = 1;
3917 	}
3918 	else if (rval == 1) {
3919 		tqe->tqe_state[dir] = nstate;
3920 		/*
3921 		 * The nstate can either advance to a new state, or remain
3922 		 * unchanged, resetting the timer by moving to the bottom of
3923 		 * the queue.
3924 		 */
3925 		DTRACE_PROBE1(state_done, int, nstate);
3926 
3927 		if ((tqe->tqe_flags & TQE_RULEBASED) == 0)
3928 			fr_movequeue(tqe, tqe->tqe_ifq, tqtab + nstate, ifs);
3929 	}
3930 
3931 	return rval;
3932 }
3933 
3934 
3935 /* ------------------------------------------------------------------------ */
3936 /* Function:    ipstate_log                                                 */
3937 /* Returns:     Nil                                                         */
3938 /* Parameters:  is(I)   - pointer to state structure                        */
3939 /*              type(I) - type of log entry to create                       */
3940 /*                                                                          */
3941 /* Creates a state table log entry using the state structure and type info. */
3942 /* passed in.  Log packet/byte counts, source/destination address and other */
3943 /* protocol specific information.                                           */
3944 /* ------------------------------------------------------------------------ */
ipstate_log(is,type,ifs)3945 void ipstate_log(is, type, ifs)
3946 struct ipstate *is;
3947 u_int type;
3948 ipf_stack_t *ifs;
3949 {
3950 #ifdef	IPFILTER_LOG
3951 	struct	ipslog	ipsl;
3952 	size_t sizes[1];
3953 	void *items[1];
3954 	int types[1];
3955 
3956 	/*
3957 	 * Copy information out of the ipstate_t structure and into the
3958 	 * structure used for logging.
3959 	 */
3960 	ipsl.isl_type = type;
3961 	ipsl.isl_pkts[0] = is->is_pkts[0] + is->is_icmppkts[0];
3962 	ipsl.isl_bytes[0] = is->is_bytes[0];
3963 	ipsl.isl_pkts[1] = is->is_pkts[1] + is->is_icmppkts[1];
3964 	ipsl.isl_bytes[1] = is->is_bytes[1];
3965 	ipsl.isl_pkts[2] = is->is_pkts[2] + is->is_icmppkts[2];
3966 	ipsl.isl_bytes[2] = is->is_bytes[2];
3967 	ipsl.isl_pkts[3] = is->is_pkts[3] + is->is_icmppkts[3];
3968 	ipsl.isl_bytes[3] = is->is_bytes[3];
3969 	ipsl.isl_src = is->is_src;
3970 	ipsl.isl_dst = is->is_dst;
3971 	ipsl.isl_p = is->is_p;
3972 	ipsl.isl_v = is->is_v;
3973 	ipsl.isl_flags = is->is_flags;
3974 	ipsl.isl_tag = is->is_tag;
3975 	ipsl.isl_rulen = is->is_rulen;
3976 	(void) strncpy(ipsl.isl_group, is->is_group, FR_GROUPLEN);
3977 
3978 	if (ipsl.isl_p == IPPROTO_TCP || ipsl.isl_p == IPPROTO_UDP) {
3979 		ipsl.isl_sport = is->is_sport;
3980 		ipsl.isl_dport = is->is_dport;
3981 		if (ipsl.isl_p == IPPROTO_TCP) {
3982 			ipsl.isl_state[0] = is->is_state[0];
3983 			ipsl.isl_state[1] = is->is_state[1];
3984 		}
3985 	} else if (ipsl.isl_p == IPPROTO_ICMP) {
3986 		ipsl.isl_itype = is->is_icmp.ici_type;
3987 	} else if (ipsl.isl_p == IPPROTO_ICMPV6) {
3988 		ipsl.isl_itype = is->is_icmp.ici_type;
3989 	} else {
3990 		ipsl.isl_ps.isl_filler[0] = 0;
3991 		ipsl.isl_ps.isl_filler[1] = 0;
3992 	}
3993 
3994 	items[0] = &ipsl;
3995 	sizes[0] = sizeof(ipsl);
3996 	types[0] = 0;
3997 
3998 	if (ipllog(IPL_LOGSTATE, NULL, items, sizes, types, 1, ifs)) {
3999 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_logged);
4000 	} else {
4001 		ATOMIC_INCL(ifs->ifs_ips_stats.iss_logfail);
4002 	}
4003 #endif
4004 }
4005 
4006 
4007 #ifdef	USE_INET6
4008 /* ------------------------------------------------------------------------ */
4009 /* Function:    fr_checkicmp6matchingstate                                  */
4010 /* Returns:     ipstate_t* - NULL == no match found,                        */
4011 /*                           else  pointer to matching state entry          */
4012 /* Parameters:  fin(I) - pointer to packet information                      */
4013 /* Locks:       NULL == no locks, else Read Lock on ipf_state               */
4014 /*                                                                          */
4015 /* If we've got an ICMPv6 error message, using the information stored in    */
4016 /* the ICMPv6 packet, look for a matching state table entry.                */
4017 /* ------------------------------------------------------------------------ */
fr_checkicmp6matchingstate(fin)4018 static ipstate_t *fr_checkicmp6matchingstate(fin)
4019 fr_info_t *fin;
4020 {
4021 	struct icmp6_hdr *ic6, *oic;
4022 	int backward, i;
4023 	ipstate_t *is, **isp;
4024 	u_short sport, dport;
4025 	i6addr_t dst, src;
4026 	u_short savelen;
4027 	icmpinfo_t *ic;
4028 	fr_info_t ofin;
4029 	tcphdr_t *tcp;
4030 	ip6_t *oip6;
4031 	u_char	pr;
4032 	u_int hv;
4033 	ipf_stack_t *ifs = fin->fin_ifs;
4034 
4035 	/*
4036 	 * Does it at least have the return (basic) IP header ?
4037 	 * Is it an actual recognised ICMP error type?
4038 	 * Only a basic IP header (no options) should be with
4039 	 * an ICMP error header.
4040 	 */
4041 	if ((fin->fin_v != 6) || (fin->fin_plen < ICMP6ERR_MINPKTLEN) ||
4042 	    !(fin->fin_flx & FI_ICMPERR))
4043 		return NULL;
4044 
4045 	ic6 = fin->fin_dp;
4046 
4047 	oip6 = (ip6_t *)((char *)ic6 + ICMPERR_ICMPHLEN);
4048 	if (fin->fin_plen < sizeof(*oip6))
4049 		return NULL;
4050 
4051 	bcopy((char *)fin, (char *)&ofin, sizeof(*fin));
4052 	ofin.fin_v = 6;
4053 	ofin.fin_ifp = fin->fin_ifp;
4054 	ofin.fin_out = !fin->fin_out;
4055 	ofin.fin_m = NULL;	/* if dereferenced, panic XXX */
4056 	ofin.fin_mp = NULL;	/* if dereferenced, panic XXX */
4057 
4058 	/*
4059 	 * We make a fin entry to be able to feed it to
4060 	 * matchsrcdst. Note that not all fields are necessary
4061 	 * but this is the cleanest way. Note further we fill
4062 	 * in fin_mp such that if someone uses it we'll get
4063 	 * a kernel panic. fr_matchsrcdst does not use this.
4064 	 *
4065 	 * watch out here, as ip is in host order and oip6 in network
4066 	 * order. Any change we make must be undone afterwards.
4067 	 */
4068 	savelen = oip6->ip6_plen;
4069 	oip6->ip6_plen = fin->fin_dlen - ICMPERR_ICMPHLEN;
4070 	ofin.fin_flx = FI_NOCKSUM;
4071 	ofin.fin_ip = (ip_t *)oip6;
4072 	ofin.fin_plen = oip6->ip6_plen;
4073 	(void) fr_makefrip(sizeof(*oip6), (ip_t *)oip6, &ofin);
4074 	ofin.fin_flx &= ~(FI_BAD|FI_SHORT);
4075 	oip6->ip6_plen = savelen;
4076 
4077 	if (oip6->ip6_nxt == IPPROTO_ICMPV6) {
4078 		oic = (struct icmp6_hdr *)(oip6 + 1);
4079 		/*
4080 		 * an ICMP error can only be generated as a result of an
4081 		 * ICMP query, not as the response on an ICMP error
4082 		 *
4083 		 * XXX theoretically ICMP_ECHOREP and the other reply's are
4084 		 * ICMP query's as well, but adding them here seems strange XXX
4085 		 */
4086 		 if (!(oic->icmp6_type & ICMP6_INFOMSG_MASK))
4087 		    	return NULL;
4088 
4089 		/*
4090 		 * perform a lookup of the ICMP packet in the state table
4091 		 */
4092 		hv = (pr = oip6->ip6_nxt);
4093 		src.in6 = oip6->ip6_src;
4094 		hv += src.in4.s_addr;
4095 		dst.in6 = oip6->ip6_dst;
4096 		hv += dst.in4.s_addr;
4097 		hv += oic->icmp6_id;
4098 		hv += oic->icmp6_seq;
4099 		hv = DOUBLE_HASH(hv, ifs);
4100 
4101 		READ_ENTER(&ifs->ifs_ipf_state);
4102 		for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
4103 			ic = &is->is_icmp;
4104 			isp = &is->is_hnext;
4105 			if ((is->is_p == pr) &&
4106 			    !(is->is_pass & FR_NOICMPERR) &&
4107 			    (oic->icmp6_id == ic->ici_id) &&
4108 			    (oic->icmp6_seq == ic->ici_seq) &&
4109 			    (is = fr_matchsrcdst(&ofin, is, &src,
4110 						 &dst, NULL, FI_ICMPCMP))) {
4111 			    	/*
4112 			    	 * in the state table ICMP query's are stored
4113 			    	 * with the type of the corresponding ICMP
4114 			    	 * response. Correct here
4115 			    	 */
4116 				if (((ic->ici_type == ICMP6_ECHO_REPLY) &&
4117 				     (oic->icmp6_type == ICMP6_ECHO_REQUEST)) ||
4118 				     (ic->ici_type - 1 == oic->icmp6_type )) {
4119 				    	ifs->ifs_ips_stats.iss_hits++;
4120 					backward = IP6_NEQ(&is->is_dst, &src);
4121 					fin->fin_rev = !backward;
4122 					i = (backward << 1) + fin->fin_out;
4123     					is->is_icmppkts[i]++;
4124 					return is;
4125 				}
4126 			}
4127 		}
4128 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
4129 		return NULL;
4130 	}
4131 
4132 	hv = (pr = oip6->ip6_nxt);
4133 	src.in6 = oip6->ip6_src;
4134 	hv += src.i6[0];
4135 	hv += src.i6[1];
4136 	hv += src.i6[2];
4137 	hv += src.i6[3];
4138 	dst.in6 = oip6->ip6_dst;
4139 	hv += dst.i6[0];
4140 	hv += dst.i6[1];
4141 	hv += dst.i6[2];
4142 	hv += dst.i6[3];
4143 
4144 	if ((oip6->ip6_nxt == IPPROTO_TCP) || (oip6->ip6_nxt == IPPROTO_UDP)) {
4145 		tcp = (tcphdr_t *)(oip6 + 1);
4146 		dport = tcp->th_dport;
4147 		sport = tcp->th_sport;
4148 		hv += dport;
4149 		hv += sport;
4150 	} else
4151 		tcp = NULL;
4152 	hv = DOUBLE_HASH(hv, ifs);
4153 
4154 	READ_ENTER(&ifs->ifs_ipf_state);
4155 	for (isp = &ifs->ifs_ips_table[hv]; ((is = *isp) != NULL); ) {
4156 		isp = &is->is_hnext;
4157 		/*
4158 		 * Only allow this icmp though if the
4159 		 * encapsulated packet was allowed through the
4160 		 * other way around. Note that the minimal amount
4161 		 * of info present does not allow for checking against
4162 		 * tcp internals such as seq and ack numbers.
4163 		 */
4164 		if ((is->is_p != pr) || (is->is_v != 6) ||
4165 		    (is->is_pass & FR_NOICMPERR))
4166 			continue;
4167 		is = fr_matchsrcdst(&ofin, is, &src, &dst, tcp, FI_ICMPCMP);
4168 		if (is != NULL) {
4169 			ifs->ifs_ips_stats.iss_hits++;
4170 			backward = IP6_NEQ(&is->is_dst, &src);
4171 			fin->fin_rev = !backward;
4172 			i = (backward << 1) + fin->fin_out;
4173 			is->is_icmppkts[i]++;
4174 			/*
4175 			 * we deliberately do not touch the timeouts
4176 			 * for the accompanying state table entry.
4177 			 * It remains to be seen if that is correct. XXX
4178 			 */
4179 			return is;
4180 		}
4181 	}
4182 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
4183 	return NULL;
4184 }
4185 #endif
4186 
4187 
4188 /* ------------------------------------------------------------------------ */
4189 /* Function:    fr_sttab_init                                               */
4190 /* Returns:     Nil                                                         */
4191 /* Parameters:  tqp(I) - pointer to an array of timeout queues for TCP      */
4192 /*                                                                          */
4193 /* Initialise the array of timeout queues for TCP.                          */
4194 /* ------------------------------------------------------------------------ */
fr_sttab_init(tqp,ifs)4195 void fr_sttab_init(tqp, ifs)
4196 ipftq_t *tqp;
4197 ipf_stack_t *ifs;
4198 {
4199 	int i;
4200 
4201 	for (i = IPF_TCP_NSTATES - 1; i >= 0; i--) {
4202 		tqp[i].ifq_ttl = 0;
4203 		tqp[i].ifq_ref = 1;
4204 		tqp[i].ifq_head = NULL;
4205 		tqp[i].ifq_tail = &tqp[i].ifq_head;
4206 		tqp[i].ifq_next = tqp + i + 1;
4207 		MUTEX_INIT(&tqp[i].ifq_lock, "ipftq tcp tab");
4208 	}
4209 	tqp[IPF_TCP_NSTATES - 1].ifq_next = NULL;
4210 	tqp[IPF_TCPS_CLOSED].ifq_ttl = ifs->ifs_fr_tcpclosed;
4211 	tqp[IPF_TCPS_LISTEN].ifq_ttl = ifs->ifs_fr_tcptimeout;
4212 	tqp[IPF_TCPS_SYN_SENT].ifq_ttl = ifs->ifs_fr_tcptimeout;
4213 	tqp[IPF_TCPS_SYN_RECEIVED].ifq_ttl = ifs->ifs_fr_tcptimeout;
4214 	tqp[IPF_TCPS_ESTABLISHED].ifq_ttl = ifs->ifs_fr_tcpidletimeout;
4215 	tqp[IPF_TCPS_CLOSE_WAIT].ifq_ttl = ifs->ifs_fr_tcphalfclosed;
4216 	tqp[IPF_TCPS_FIN_WAIT_1].ifq_ttl = ifs->ifs_fr_tcphalfclosed;
4217 	tqp[IPF_TCPS_CLOSING].ifq_ttl = ifs->ifs_fr_tcptimeout;
4218 	tqp[IPF_TCPS_LAST_ACK].ifq_ttl = ifs->ifs_fr_tcplastack;
4219 	tqp[IPF_TCPS_FIN_WAIT_2].ifq_ttl = ifs->ifs_fr_tcpclosewait;
4220 	tqp[IPF_TCPS_TIME_WAIT].ifq_ttl = ifs->ifs_fr_tcptimeout;
4221 	tqp[IPF_TCPS_HALF_ESTAB].ifq_ttl = ifs->ifs_fr_tcptimeout;
4222 }
4223 
4224 
4225 /* ------------------------------------------------------------------------ */
4226 /* Function:    fr_sttab_destroy                                            */
4227 /* Returns:     Nil                                                         */
4228 /* Parameters:  tqp(I) - pointer to an array of timeout queues for TCP      */
4229 /*                                                                          */
4230 /* Do whatever is necessary to "destroy" each of the entries in the array   */
4231 /* of timeout queues for TCP.                                               */
4232 /* ------------------------------------------------------------------------ */
fr_sttab_destroy(tqp)4233 void fr_sttab_destroy(tqp)
4234 ipftq_t *tqp;
4235 {
4236 	int i;
4237 
4238 	for (i = IPF_TCP_NSTATES - 1; i >= 0; i--)
4239 		MUTEX_DESTROY(&tqp[i].ifq_lock);
4240 }
4241 
4242 
4243 /* ------------------------------------------------------------------------ */
4244 /* Function:    fr_statederef                                               */
4245 /* Returns:     Nil                                                         */
4246 /* Parameters:  isp(I) - pointer to pointer to state table entry            */
4247 /*              ifs - ipf stack instance                                    */
4248 /*                                                                          */
4249 /* Decrement the reference counter for this state table entry and free it   */
4250 /* if there are no more things using it.                                    */
4251 /*                                                                          */
4252 /* Internal parameters:                                                     */
4253 /*    state[0] = state of source (host that initiated connection)           */
4254 /*    state[1] = state of dest   (host that accepted the connection)        */
4255 /* ------------------------------------------------------------------------ */
fr_statederef(isp,ifs)4256 void fr_statederef(isp, ifs)
4257 ipstate_t **isp;
4258 ipf_stack_t *ifs;
4259 {
4260 	ipstate_t *is;
4261 
4262 	is = *isp;
4263 	*isp = NULL;
4264 
4265 	MUTEX_ENTER(&is->is_lock);
4266 	if (is->is_ref > 1) {
4267 		is->is_ref--;
4268 		MUTEX_EXIT(&is->is_lock);
4269 #ifndef	_KERNEL
4270 		if ((is->is_sti.tqe_state[0] > IPF_TCPS_ESTABLISHED) ||
4271 		   (is->is_sti.tqe_state[1] > IPF_TCPS_ESTABLISHED)) {
4272 			(void) fr_delstate(is, ISL_ORPHAN, ifs);
4273 		}
4274 #endif
4275 		return;
4276 	}
4277 	MUTEX_EXIT(&is->is_lock);
4278 
4279 	WRITE_ENTER(&ifs->ifs_ipf_state);
4280 	(void) fr_delstate(is, ISL_EXPIRE, ifs);
4281 	RWLOCK_EXIT(&ifs->ifs_ipf_state);
4282 }
4283 
4284 
4285 /* ------------------------------------------------------------------------ */
4286 /* Function:    fr_setstatequeue                                            */
4287 /* Returns:     Nil                                                         */
4288 /* Parameters:  is(I) - pointer to state structure                          */
4289 /*              rev(I) - forward(0) or reverse(1) direction                 */
4290 /* Locks:       ipf_state (read or write)                                   */
4291 /*                                                                          */
4292 /* Put the state entry on its default queue entry, using rev as a helped in */
4293 /* determining which queue it should be placed on.                          */
4294 /* ------------------------------------------------------------------------ */
fr_setstatequeue(is,rev,ifs)4295 void fr_setstatequeue(is, rev, ifs)
4296 ipstate_t *is;
4297 int rev;
4298 ipf_stack_t *ifs;
4299 {
4300 	ipftq_t *oifq, *nifq;
4301 
4302 
4303 	if ((is->is_sti.tqe_flags & TQE_RULEBASED) != 0)
4304 		nifq = is->is_tqehead[rev];
4305 	else
4306 		nifq = NULL;
4307 
4308 	if (nifq == NULL) {
4309 		switch (is->is_p)
4310 		{
4311 #ifdef USE_INET6
4312 		case IPPROTO_ICMPV6 :
4313 			if (rev == 1)
4314 				nifq = &ifs->ifs_ips_icmpacktq;
4315 			else
4316 				nifq = &ifs->ifs_ips_icmptq;
4317 			break;
4318 #endif
4319 		case IPPROTO_ICMP :
4320 			if (rev == 1)
4321 				nifq = &ifs->ifs_ips_icmpacktq;
4322 			else
4323 				nifq = &ifs->ifs_ips_icmptq;
4324 			break;
4325 		case IPPROTO_TCP :
4326 			nifq = ifs->ifs_ips_tqtqb + is->is_state[rev];
4327 			break;
4328 
4329 		case IPPROTO_UDP :
4330 			if (rev == 1)
4331 				nifq = &ifs->ifs_ips_udpacktq;
4332 			else
4333 				nifq = &ifs->ifs_ips_udptq;
4334 			break;
4335 
4336 		default :
4337 			nifq = &ifs->ifs_ips_iptq;
4338 			break;
4339 		}
4340 	}
4341 
4342 	oifq = is->is_sti.tqe_ifq;
4343 	/*
4344 	 * If it's currently on a timeout queue, move it from one queue to
4345 	 * another, else put it on the end of the newly determined queue.
4346 	 */
4347 	if (oifq != NULL)
4348 		fr_movequeue(&is->is_sti, oifq, nifq, ifs);
4349 	else
4350 		fr_queueappend(&is->is_sti, nifq, is, ifs);
4351 	return;
4352 }
4353 
4354 
4355 /* ------------------------------------------------------------------------ */
4356 /* Function:    fr_stateiter                                                */
4357 /* Returns:     int - 0 == success, else error                              */
4358 /* Parameters:  token(I) - pointer to ipftoken structure                    */
4359 /*              itp(I)   - pointer to ipfgeniter structure                  */
4360 /*                                                                          */
4361 /* This function handles the SIOCGENITER ioctl for the state tables and     */
4362 /* walks through the list of entries in the state table list (ips_list.)    */
4363 /* ------------------------------------------------------------------------ */
fr_stateiter(token,itp,ifs)4364 static int fr_stateiter(token, itp, ifs)
4365 ipftoken_t *token;
4366 ipfgeniter_t *itp;
4367 ipf_stack_t *ifs;
4368 {
4369 	ipstate_t *is, *next, zero;
4370 	int error, count;
4371 	char *dst;
4372 
4373 	if (itp->igi_data == NULL)
4374 		return EFAULT;
4375 
4376 	if (itp->igi_nitems == 0)
4377 		return EINVAL;
4378 
4379 	if (itp->igi_type != IPFGENITER_STATE)
4380 		return EINVAL;
4381 
4382 	error = 0;
4383 
4384 	READ_ENTER(&ifs->ifs_ipf_state);
4385 
4386 	/*
4387 	 * Get "previous" entry from the token and find the next entry.
4388 	 */
4389 	is = token->ipt_data;
4390 	if (is == NULL) {
4391 		next = ifs->ifs_ips_list;
4392 	} else {
4393 		next = is->is_next;
4394 	}
4395 
4396 	dst = itp->igi_data;
4397 	for (count = itp->igi_nitems; count > 0; count--) {
4398 		/*
4399 		 * If we found an entry, add a reference to it and update the token.
4400 		 * Otherwise, zero out data to be returned and NULL out token.
4401 		 */
4402 		if (next != NULL) {
4403 			MUTEX_ENTER(&next->is_lock);
4404 			next->is_ref++;
4405 			MUTEX_EXIT(&next->is_lock);
4406 			token->ipt_data = next;
4407 		} else {
4408 			bzero(&zero, sizeof(zero));
4409 			next = &zero;
4410 			token->ipt_data = NULL;
4411 		}
4412 
4413 		/*
4414 		 * Safe to release lock now the we have a reference.
4415 		 */
4416 		RWLOCK_EXIT(&ifs->ifs_ipf_state);
4417 
4418 		/*
4419 		 * Copy out data and clean up references and tokens.
4420 		 */
4421 		error = COPYOUT(next, dst, sizeof(*next));
4422 		if (error != 0)
4423 			error = EFAULT;
4424 		if (token->ipt_data == NULL) {
4425 			ipf_freetoken(token, ifs);
4426 			break;
4427 		} else {
4428 			if (is != NULL)
4429 				fr_statederef(&is, ifs);
4430 			if (next->is_next == NULL) {
4431 				ipf_freetoken(token, ifs);
4432 				break;
4433 			}
4434 		}
4435 
4436 		if ((count == 1) || (error != 0))
4437 			break;
4438 
4439 		READ_ENTER(&ifs->ifs_ipf_state);
4440 		dst += sizeof(*next);
4441 		is = next;
4442 		next = is->is_next;
4443 	}
4444 
4445 	return error;
4446 }
4447