xref: /titanic_41/usr/src/uts/common/inet/ipf/ip_nat.c (revision b449fa8aa1f53939091aeb3a5af8a8f3ba9d33dc)
1 /*
2  * Copyright (C) 1995-2004 by Darren Reed.
3  *
4  * See the IPFILTER.LICENCE file for details on licencing.
5  *
6  * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
7  * Use is subject to license terms.
8  */
9 
10 #pragma ident	"%Z%%M%	%I%	%E% SMI"$
11 
12 #if defined(KERNEL) || defined(_KERNEL)
13 # undef KERNEL
14 # undef _KERNEL
15 # define        KERNEL	1
16 # define        _KERNEL	1
17 #endif
18 #include <sys/errno.h>
19 #include <sys/types.h>
20 #include <sys/param.h>
21 #include <sys/time.h>
22 #include <sys/file.h>
23 #if defined(__NetBSD__) && (NetBSD >= 199905) && !defined(IPFILTER_LKM) && \
24     defined(_KERNEL)
25 # include "opt_ipfilter_log.h"
26 #endif
27 #if !defined(_KERNEL)
28 # include <stdio.h>
29 # include <string.h>
30 # include <stdlib.h>
31 # define _KERNEL
32 # ifdef __OpenBSD__
33 struct file;
34 # endif
35 # include <sys/uio.h>
36 # undef _KERNEL
37 #endif
38 #if defined(_KERNEL) && (__FreeBSD_version >= 220000)
39 # include <sys/filio.h>
40 # include <sys/fcntl.h>
41 #else
42 # include <sys/ioctl.h>
43 #endif
44 #if !defined(AIX)
45 # include <sys/fcntl.h>
46 #endif
47 #if !defined(linux)
48 # include <sys/protosw.h>
49 #endif
50 #include <sys/socket.h>
51 #if defined(_KERNEL)
52 # include <sys/systm.h>
53 # if !defined(__SVR4) && !defined(__svr4__)
54 #  include <sys/mbuf.h>
55 # endif
56 #endif
57 #if defined(__SVR4) || defined(__svr4__)
58 # include <sys/filio.h>
59 # include <sys/byteorder.h>
60 # ifdef _KERNEL
61 #  include <sys/dditypes.h>
62 # endif
63 # include <sys/stream.h>
64 # include <sys/kmem.h>
65 #endif
66 #if __FreeBSD_version >= 300000
67 # include <sys/queue.h>
68 #endif
69 #include <net/if.h>
70 #if __FreeBSD_version >= 300000
71 # include <net/if_var.h>
72 # if defined(_KERNEL) && !defined(IPFILTER_LKM)
73 #  include "opt_ipfilter.h"
74 # endif
75 #endif
76 #ifdef sun
77 # include <net/af.h>
78 #endif
79 #include <net/route.h>
80 #include <netinet/in.h>
81 #include <netinet/in_systm.h>
82 #include <netinet/ip.h>
83 
84 #ifdef RFC1825
85 # include <vpn/md5.h>
86 # include <vpn/ipsec.h>
87 extern struct ifnet vpnif;
88 #endif
89 
90 #if !defined(linux)
91 # include <netinet/ip_var.h>
92 #endif
93 #include <netinet/tcp.h>
94 #include <netinet/udp.h>
95 #include <netinet/ip_icmp.h>
96 #include "netinet/ip_compat.h"
97 #include <netinet/tcpip.h>
98 #include "netinet/ip_fil.h"
99 #include "netinet/ip_nat.h"
100 #include "netinet/ip_frag.h"
101 #include "netinet/ip_state.h"
102 #include "netinet/ip_proxy.h"
103 #include "netinet/ipf_stack.h"
104 #ifdef	IPFILTER_SYNC
105 #include "netinet/ip_sync.h"
106 #endif
107 #if (__FreeBSD_version >= 300000)
108 # include <sys/malloc.h>
109 #endif
110 /* END OF INCLUDES */
111 
112 #undef	SOCKADDR_IN
113 #define	SOCKADDR_IN	struct sockaddr_in
114 
115 #if !defined(lint)
116 static const char sccsid[] = "@(#)ip_nat.c	1.11 6/5/96 (C) 1995 Darren Reed";
117 static const char rcsid[] = "@(#)$Id: ip_nat.c,v 2.195.2.42 2005/08/11 19:51:36 darrenr Exp $";
118 #endif
119 
120 
121 /* ======================================================================== */
122 /* How the NAT is organised and works.                                      */
123 /*                                                                          */
124 /* Inside (interface y) NAT       Outside (interface x)                     */
125 /* -------------------- -+- -------------------------------------           */
126 /* Packet going          |   out, processsed by fr_checknatout() for x      */
127 /* ------------>         |   ------------>                                  */
128 /* src=10.1.1.1          |   src=192.1.1.1                                  */
129 /*                       |                                                  */
130 /*                       |   in, processed by fr_checknatin() for x         */
131 /* <------------         |   <------------                                  */
132 /* dst=10.1.1.1          |   dst=192.1.1.1                                  */
133 /* -------------------- -+- -------------------------------------           */
134 /* fr_checknatout() - changes ip_src and if required, sport                 */
135 /*             - creates a new mapping, if required.                        */
136 /* fr_checknatin()  - changes ip_dst and if required, dport                 */
137 /*                                                                          */
138 /* In the NAT table, internal source is recorded as "in" and externally     */
139 /* seen as "out".                                                           */
140 /* ======================================================================== */
141 
142 
143 static	int	nat_flushtable __P((ipf_stack_t *));
144 static	int	nat_clearlist __P((ipf_stack_t *));
145 static	void	nat_addnat __P((struct ipnat *, ipf_stack_t *));
146 static	void	nat_addrdr __P((struct ipnat *, ipf_stack_t *));
147 static	void	nat_delete __P((struct nat *, int, ipf_stack_t *));
148 static	int	fr_natgetent __P((caddr_t, ipf_stack_t *));
149 static	int	fr_natgetsz __P((caddr_t, ipf_stack_t *));
150 static	int	fr_natputent __P((caddr_t, int, ipf_stack_t *));
151 static	void	nat_tabmove __P((nat_t *, ipf_stack_t *));
152 static	int	nat_match __P((fr_info_t *, ipnat_t *));
153 static	INLINE	int nat_newmap __P((fr_info_t *, nat_t *, natinfo_t *));
154 static	INLINE	int nat_newrdr __P((fr_info_t *, nat_t *, natinfo_t *));
155 static	hostmap_t *nat_hostmap __P((ipnat_t *, struct in_addr,
156 				    struct in_addr, struct in_addr, u_32_t,
157 				    ipf_stack_t *));
158 static	INLINE	int nat_icmpquerytype4 __P((int));
159 static	int	nat_ruleaddrinit __P((ipnat_t *));
160 static	int	nat_siocaddnat __P((ipnat_t *, ipnat_t **, int, ipf_stack_t *));
161 static	void	nat_siocdelnat __P((ipnat_t *, ipnat_t **, int, ipf_stack_t *));
162 static	INLINE	int nat_icmperrortype4 __P((int));
163 static	INLINE	int nat_finalise __P((fr_info_t *, nat_t *, natinfo_t *,
164 				      tcphdr_t *, nat_t **, int));
165 static	INLINE	int nat_resolverule __P((ipnat_t *, ipf_stack_t *));
166 static	void	nat_mssclamp __P((tcphdr_t *, u_32_t, u_short *));
167 static	int	nat_getnext __P((ipftoken_t *, ipfgeniter_t *, ipf_stack_t *));
168 static	int	nat_iterator __P((ipftoken_t *, ipfgeniter_t *, ipf_stack_t *));
169 static	int	nat_extraflush __P((int, ipf_stack_t *));
170 static	int	nat_earlydrop __P((ipftq_t *, int, ipf_stack_t *));
171 static	int	nat_flushclosing __P((int, ipf_stack_t *));
172 
173 
174 /*
175  * Below we declare a list of constants used only in the nat_extraflush()
176  * routine.  We are placing it here, instead of in nat_extraflush() itself,
177  * because we want to make it visible to tools such as mdb, nm etc., so the
178  * values can easily be altered during debugging.
179  */
180 static	const int	idletime_tab[] = {
181 	IPF_TTLVAL(30),		/* 30 seconds */
182 	IPF_TTLVAL(1800),	/* 30 minutes */
183 	IPF_TTLVAL(43200),	/* 12 hours */
184 	IPF_TTLVAL(345600),	/* 4 days */
185 };
186 
187 #define NAT_HAS_L4_CHANGED(n)	\
188  	(((n)->nat_flags & (IPN_TCPUDPICMP)) && \
189  	(n)->nat_inport != (n)->nat_outport)
190 
191 /* ------------------------------------------------------------------------ */
192 /* Function:    fr_natinit                                                  */
193 /* Returns:     int - 0 == success, -1 == failure                           */
194 /* Parameters:  Nil                                                         */
195 /*                                                                          */
196 /* Initialise all of the NAT locks, tables and other structures.            */
197 /* ------------------------------------------------------------------------ */
198 int fr_natinit(ifs)
199 ipf_stack_t *ifs;
200 {
201 	int i;
202 
203 	KMALLOCS(ifs->ifs_nat_table[0], nat_t **,
204 		 sizeof(nat_t *) * ifs->ifs_ipf_nattable_sz);
205 	if (ifs->ifs_nat_table[0] != NULL)
206 		bzero((char *)ifs->ifs_nat_table[0],
207 		      ifs->ifs_ipf_nattable_sz * sizeof(nat_t *));
208 	else
209 		return -1;
210 
211 	KMALLOCS(ifs->ifs_nat_table[1], nat_t **,
212 		 sizeof(nat_t *) * ifs->ifs_ipf_nattable_sz);
213 	if (ifs->ifs_nat_table[1] != NULL)
214 		bzero((char *)ifs->ifs_nat_table[1],
215 		      ifs->ifs_ipf_nattable_sz * sizeof(nat_t *));
216 	else
217 		return -2;
218 
219 	KMALLOCS(ifs->ifs_nat_rules, ipnat_t **,
220 		 sizeof(ipnat_t *) * ifs->ifs_ipf_natrules_sz);
221 	if (ifs->ifs_nat_rules != NULL)
222 		bzero((char *)ifs->ifs_nat_rules,
223 		      ifs->ifs_ipf_natrules_sz * sizeof(ipnat_t *));
224 	else
225 		return -3;
226 
227 	KMALLOCS(ifs->ifs_rdr_rules, ipnat_t **,
228 		 sizeof(ipnat_t *) * ifs->ifs_ipf_rdrrules_sz);
229 	if (ifs->ifs_rdr_rules != NULL)
230 		bzero((char *)ifs->ifs_rdr_rules,
231 		      ifs->ifs_ipf_rdrrules_sz * sizeof(ipnat_t *));
232 	else
233 		return -4;
234 
235 	KMALLOCS(ifs->ifs_maptable, hostmap_t **,
236 		 sizeof(hostmap_t *) * ifs->ifs_ipf_hostmap_sz);
237 	if (ifs->ifs_maptable != NULL)
238 		bzero((char *)ifs->ifs_maptable,
239 		      sizeof(hostmap_t *) * ifs->ifs_ipf_hostmap_sz);
240 	else
241 		return -5;
242 
243 	ifs->ifs_ipf_hm_maplist = NULL;
244 
245 	KMALLOCS(ifs->ifs_nat_stats.ns_bucketlen[0], u_long *,
246 		 ifs->ifs_ipf_nattable_sz * sizeof(u_long));
247 	if (ifs->ifs_nat_stats.ns_bucketlen[0] == NULL)
248 		return -1;
249 	bzero((char *)ifs->ifs_nat_stats.ns_bucketlen[0],
250 	      ifs->ifs_ipf_nattable_sz * sizeof(u_long));
251 
252 	KMALLOCS(ifs->ifs_nat_stats.ns_bucketlen[1], u_long *,
253 		 ifs->ifs_ipf_nattable_sz * sizeof(u_long));
254 	if (ifs->ifs_nat_stats.ns_bucketlen[1] == NULL)
255 		return -1;
256 	bzero((char *)ifs->ifs_nat_stats.ns_bucketlen[1],
257 	      ifs->ifs_ipf_nattable_sz * sizeof(u_long));
258 
259 	if (ifs->ifs_fr_nat_maxbucket == 0) {
260 		for (i = ifs->ifs_ipf_nattable_sz; i > 0; i >>= 1)
261 			ifs->ifs_fr_nat_maxbucket++;
262 		ifs->ifs_fr_nat_maxbucket *= 2;
263 	}
264 
265 	fr_sttab_init(ifs->ifs_nat_tqb, ifs);
266 	/*
267 	 * Increase this because we may have "keep state" following this too
268 	 * and packet storms can occur if this is removed too quickly.
269 	 */
270 	ifs->ifs_nat_tqb[IPF_TCPS_CLOSED].ifq_ttl = ifs->ifs_fr_tcplastack;
271 	ifs->ifs_nat_tqb[IPF_TCP_NSTATES - 1].ifq_next = &ifs->ifs_nat_udptq;
272 	ifs->ifs_nat_udptq.ifq_ttl = ifs->ifs_fr_defnatage;
273 	ifs->ifs_nat_udptq.ifq_ref = 1;
274 	ifs->ifs_nat_udptq.ifq_head = NULL;
275 	ifs->ifs_nat_udptq.ifq_tail = &ifs->ifs_nat_udptq.ifq_head;
276 	MUTEX_INIT(&ifs->ifs_nat_udptq.ifq_lock, "nat ipftq udp tab");
277 	ifs->ifs_nat_udptq.ifq_next = &ifs->ifs_nat_icmptq;
278 	ifs->ifs_nat_icmptq.ifq_ttl = ifs->ifs_fr_defnaticmpage;
279 	ifs->ifs_nat_icmptq.ifq_ref = 1;
280 	ifs->ifs_nat_icmptq.ifq_head = NULL;
281 	ifs->ifs_nat_icmptq.ifq_tail = &ifs->ifs_nat_icmptq.ifq_head;
282 	MUTEX_INIT(&ifs->ifs_nat_icmptq.ifq_lock, "nat icmp ipftq tab");
283 	ifs->ifs_nat_icmptq.ifq_next = &ifs->ifs_nat_iptq;
284 	ifs->ifs_nat_iptq.ifq_ttl = ifs->ifs_fr_defnatipage;
285 	ifs->ifs_nat_iptq.ifq_ref = 1;
286 	ifs->ifs_nat_iptq.ifq_head = NULL;
287 	ifs->ifs_nat_iptq.ifq_tail = &ifs->ifs_nat_iptq.ifq_head;
288 	MUTEX_INIT(&ifs->ifs_nat_iptq.ifq_lock, "nat ip ipftq tab");
289 	ifs->ifs_nat_iptq.ifq_next = NULL;
290 
291 	for (i = 0; i < IPF_TCP_NSTATES; i++) {
292 		if (ifs->ifs_nat_tqb[i].ifq_ttl < ifs->ifs_fr_defnaticmpage)
293 			ifs->ifs_nat_tqb[i].ifq_ttl = ifs->ifs_fr_defnaticmpage;
294 #ifdef LARGE_NAT
295 		else if (ifs->ifs_nat_tqb[i].ifq_ttl > ifs->ifs_fr_defnatage)
296 			ifs->ifs_nat_tqb[i].ifq_ttl = ifs->ifs_fr_defnatage;
297 #endif
298 	}
299 
300 	/*
301 	 * Increase this because we may have "keep state" following
302 	 * this too and packet storms can occur if this is removed
303 	 * too quickly.
304 	 */
305 	ifs->ifs_nat_tqb[IPF_TCPS_CLOSED].ifq_ttl =
306 	    ifs->ifs_nat_tqb[IPF_TCPS_LAST_ACK].ifq_ttl;
307 
308 	RWLOCK_INIT(&ifs->ifs_ipf_nat, "ipf IP NAT rwlock");
309 	RWLOCK_INIT(&ifs->ifs_ipf_natfrag, "ipf IP NAT-Frag rwlock");
310 	MUTEX_INIT(&ifs->ifs_ipf_nat_new, "ipf nat new mutex");
311 	MUTEX_INIT(&ifs->ifs_ipf_natio, "ipf nat io mutex");
312 
313 	ifs->ifs_fr_nat_init = 1;
314 
315 	return 0;
316 }
317 
318 
319 /* ------------------------------------------------------------------------ */
320 /* Function:    nat_addrdr                                                  */
321 /* Returns:     Nil                                                         */
322 /* Parameters:  n(I) - pointer to NAT rule to add                           */
323 /*                                                                          */
324 /* Adds a redirect rule to the hash table of redirect rules and the list of */
325 /* loaded NAT rules.  Updates the bitmask indicating which netmasks are in  */
326 /* use by redirect rules.                                                   */
327 /* ------------------------------------------------------------------------ */
328 static void nat_addrdr(n, ifs)
329 ipnat_t *n;
330 ipf_stack_t *ifs;
331 {
332 	ipnat_t **np;
333 	u_32_t j;
334 	u_int hv;
335 	int k;
336 
337 	k = count4bits(n->in_outmsk);
338 	if ((k >= 0) && (k != 32))
339 		ifs->ifs_rdr_masks |= 1 << k;
340 	j = (n->in_outip & n->in_outmsk);
341 	hv = NAT_HASH_FN(j, 0, ifs->ifs_ipf_rdrrules_sz);
342 	np = ifs->ifs_rdr_rules + hv;
343 	while (*np != NULL)
344 		np = &(*np)->in_rnext;
345 	n->in_rnext = NULL;
346 	n->in_prnext = np;
347 	n->in_hv = hv;
348 	*np = n;
349 }
350 
351 
352 /* ------------------------------------------------------------------------ */
353 /* Function:    nat_addnat                                                  */
354 /* Returns:     Nil                                                         */
355 /* Parameters:  n(I) - pointer to NAT rule to add                           */
356 /*                                                                          */
357 /* Adds a NAT map rule to the hash table of rules and the list of  loaded   */
358 /* NAT rules.  Updates the bitmask indicating which netmasks are in use by  */
359 /* redirect rules.                                                          */
360 /* ------------------------------------------------------------------------ */
361 static void nat_addnat(n, ifs)
362 ipnat_t *n;
363 ipf_stack_t *ifs;
364 {
365 	ipnat_t **np;
366 	u_32_t j;
367 	u_int hv;
368 	int k;
369 
370 	k = count4bits(n->in_inmsk);
371 	if ((k >= 0) && (k != 32))
372 		ifs->ifs_nat_masks |= 1 << k;
373 	j = (n->in_inip & n->in_inmsk);
374 	hv = NAT_HASH_FN(j, 0, ifs->ifs_ipf_natrules_sz);
375 	np = ifs->ifs_nat_rules + hv;
376 	while (*np != NULL)
377 		np = &(*np)->in_mnext;
378 	n->in_mnext = NULL;
379 	n->in_pmnext = np;
380 	n->in_hv = hv;
381 	*np = n;
382 }
383 
384 
385 /* ------------------------------------------------------------------------ */
386 /* Function:    nat_delrdr                                                  */
387 /* Returns:     Nil                                                         */
388 /* Parameters:  n(I) - pointer to NAT rule to delete                        */
389 /*                                                                          */
390 /* Removes a redirect rule from the hash table of redirect rules.           */
391 /* ------------------------------------------------------------------------ */
392 void nat_delrdr(n)
393 ipnat_t *n;
394 {
395 	if (n->in_rnext)
396 		n->in_rnext->in_prnext = n->in_prnext;
397 	*n->in_prnext = n->in_rnext;
398 }
399 
400 
401 /* ------------------------------------------------------------------------ */
402 /* Function:    nat_delnat                                                  */
403 /* Returns:     Nil                                                         */
404 /* Parameters:  n(I) - pointer to NAT rule to delete                        */
405 /*                                                                          */
406 /* Removes a NAT map rule from the hash table of NAT map rules.             */
407 /* ------------------------------------------------------------------------ */
408 void nat_delnat(n)
409 ipnat_t *n;
410 {
411 	if (n->in_mnext != NULL)
412 		n->in_mnext->in_pmnext = n->in_pmnext;
413 	*n->in_pmnext = n->in_mnext;
414 }
415 
416 
417 /* ------------------------------------------------------------------------ */
418 /* Function:    nat_hostmap                                                 */
419 /* Returns:     struct hostmap* - NULL if no hostmap could be created,      */
420 /*                                else a pointer to the hostmapping to use  */
421 /* Parameters:  np(I)   - pointer to NAT rule                               */
422 /*              real(I) - real IP address                                   */
423 /*              map(I)  - mapped IP address                                 */
424 /*              port(I) - destination port number                           */
425 /* Write Locks: ipf_nat                                                     */
426 /*                                                                          */
427 /* Check if an ip address has already been allocated for a given mapping    */
428 /* that is not doing port based translation.  If is not yet allocated, then */
429 /* create a new entry if a non-NULL NAT rule pointer has been supplied.     */
430 /* ------------------------------------------------------------------------ */
431 static struct hostmap *nat_hostmap(np, src, dst, map, port, ifs)
432 ipnat_t *np;
433 struct in_addr src;
434 struct in_addr dst;
435 struct in_addr map;
436 u_32_t port;
437 ipf_stack_t *ifs;
438 {
439 	hostmap_t *hm;
440 	u_int hv;
441 
442 	hv = (src.s_addr ^ dst.s_addr);
443 	hv += src.s_addr;
444 	hv += dst.s_addr;
445 	hv %= HOSTMAP_SIZE;
446 	for (hm = ifs->ifs_maptable[hv]; hm; hm = hm->hm_next)
447 		if ((hm->hm_srcip.s_addr == src.s_addr) &&
448 		    (hm->hm_dstip.s_addr == dst.s_addr) &&
449 		    ((np == NULL) || (np == hm->hm_ipnat)) &&
450 		    ((port == 0) || (port == hm->hm_port))) {
451 			hm->hm_ref++;
452 			return hm;
453 		}
454 
455 	if (np == NULL)
456 		return NULL;
457 
458 	KMALLOC(hm, hostmap_t *);
459 	if (hm) {
460 		hm->hm_hnext = ifs->ifs_ipf_hm_maplist;
461 		hm->hm_phnext = &ifs->ifs_ipf_hm_maplist;
462 		if (ifs->ifs_ipf_hm_maplist != NULL)
463 			ifs->ifs_ipf_hm_maplist->hm_phnext = &hm->hm_hnext;
464 		ifs->ifs_ipf_hm_maplist = hm;
465 
466 		hm->hm_next = ifs->ifs_maptable[hv];
467 		hm->hm_pnext = ifs->ifs_maptable + hv;
468 		if (ifs->ifs_maptable[hv] != NULL)
469 			ifs->ifs_maptable[hv]->hm_pnext = &hm->hm_next;
470 		ifs->ifs_maptable[hv] = hm;
471 		hm->hm_ipnat = np;
472 		hm->hm_srcip = src;
473 		hm->hm_dstip = dst;
474 		hm->hm_mapip = map;
475 		hm->hm_ref = 1;
476 		hm->hm_port = port;
477 		hm->hm_v = 4;
478 	}
479 	return hm;
480 }
481 
482 
483 /* ------------------------------------------------------------------------ */
484 /* Function:    fr_hostmapdel                                              */
485 /* Returns:     Nil                                                         */
486 /* Parameters:  hmp(I) - pointer to pointer to hostmap structure            */
487 /* Write Locks: ipf_nat                                                     */
488 /*                                                                          */
489 /* Decrement the references to this hostmap structure by one.  If this      */
490 /* reaches zero then remove it and free it.                                 */
491 /* ------------------------------------------------------------------------ */
492 void fr_hostmapdel(hmp)
493 struct hostmap **hmp;
494 {
495 	struct hostmap *hm;
496 
497 	hm = *hmp;
498 	*hmp = NULL;
499 
500 	hm->hm_ref--;
501 	if (hm->hm_ref == 0) {
502 		if (hm->hm_next)
503 			hm->hm_next->hm_pnext = hm->hm_pnext;
504 		*hm->hm_pnext = hm->hm_next;
505 		if (hm->hm_hnext)
506 			hm->hm_hnext->hm_phnext = hm->hm_phnext;
507 		*hm->hm_phnext = hm->hm_hnext;
508 		KFREE(hm);
509 	}
510 }
511 
512 
513 /* ------------------------------------------------------------------------ */
514 /* Function:    fix_outcksum                                                */
515 /* Returns:     Nil                                                         */
516 /* Parameters:  sp(I)  - location of 16bit checksum to update               */
517 /*              n((I)  - amount to adjust checksum by                       */
518 /*                                                                          */
519 /* Adjusts the 16bit checksum by "n" for packets going out.                 */
520 /* ------------------------------------------------------------------------ */
521 void fix_outcksum(sp, n)
522 u_short *sp;
523 u_32_t n;
524 {
525 	u_short sumshort;
526 	u_32_t sum1;
527 
528 	if (n == 0)
529 		return;
530 
531 	sum1 = (~ntohs(*sp)) & 0xffff;
532 	sum1 += (n);
533 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
534 	/* Again */
535 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
536 	sumshort = ~(u_short)sum1;
537 	*(sp) = htons(sumshort);
538 }
539 
540 
541 /* ------------------------------------------------------------------------ */
542 /* Function:    fix_incksum                                                 */
543 /* Returns:     Nil                                                         */
544 /* Parameters:  sp(I)  - location of 16bit checksum to update               */
545 /*              n((I)  - amount to adjust checksum by                       */
546 /*                                                                          */
547 /* Adjusts the 16bit checksum by "n" for packets going in.                  */
548 /* ------------------------------------------------------------------------ */
549 void fix_incksum(sp, n)
550 u_short *sp;
551 u_32_t n;
552 {
553 	u_short sumshort;
554 	u_32_t sum1;
555 
556 	if (n == 0)
557 		return;
558 
559 	sum1 = (~ntohs(*sp)) & 0xffff;
560 	sum1 += ~(n) & 0xffff;
561 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
562 	/* Again */
563 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
564 	sumshort = ~(u_short)sum1;
565 	*(sp) = htons(sumshort);
566 }
567 
568 
569 /* ------------------------------------------------------------------------ */
570 /* Function:    fix_datacksum                                               */
571 /* Returns:     Nil                                                         */
572 /* Parameters:  sp(I)  - location of 16bit checksum to update               */
573 /*              n((I)  - amount to adjust checksum by                       */
574 /*                                                                          */
575 /* Fix_datacksum is used *only* for the adjustments of checksums in the     */
576 /* data section of an IP packet.                                            */
577 /*                                                                          */
578 /* The only situation in which you need to do this is when NAT'ing an       */
579 /* ICMP error message. Such a message, contains in its body the IP header   */
580 /* of the original IP packet, that causes the error.                        */
581 /*                                                                          */
582 /* You can't use fix_incksum or fix_outcksum in that case, because for the  */
583 /* kernel the data section of the ICMP error is just data, and no special   */
584 /* processing like hardware cksum or ntohs processing have been done by the */
585 /* kernel on the data section.                                              */
586 /* ------------------------------------------------------------------------ */
587 void fix_datacksum(sp, n)
588 u_short *sp;
589 u_32_t n;
590 {
591 	u_short sumshort;
592 	u_32_t sum1;
593 
594 	if (n == 0)
595 		return;
596 
597 	sum1 = (~ntohs(*sp)) & 0xffff;
598 	sum1 += (n);
599 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
600 	/* Again */
601 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
602 	sumshort = ~(u_short)sum1;
603 	*(sp) = htons(sumshort);
604 }
605 
606 
607 /* ------------------------------------------------------------------------ */
608 /* Function:    fr_nat_ioctl                                                */
609 /* Returns:     int - 0 == success, != 0 == failure                         */
610 /* Parameters:  data(I) - pointer to ioctl data                             */
611 /*              cmd(I)  - ioctl command integer                             */
612 /*              mode(I) - file mode bits used with open                     */
613 /*                                                                          */
614 /* Processes an ioctl call made to operate on the IP Filter NAT device.     */
615 /* ------------------------------------------------------------------------ */
616 int fr_nat_ioctl(data, cmd, mode, uid, ctx, ifs)
617 ioctlcmd_t cmd;
618 caddr_t data;
619 int mode, uid;
620 void *ctx;
621 ipf_stack_t *ifs;
622 {
623 	ipnat_t *nat, *nt, *n = NULL, **np = NULL;
624 	int error = 0, ret, arg, getlock;
625 	ipnat_t natd;
626 
627 #if (BSD >= 199306) && defined(_KERNEL)
628 	if ((securelevel >= 2) && (mode & FWRITE))
629 		return EPERM;
630 #endif
631 
632 #if defined(__osf__) && defined(_KERNEL)
633 	getlock = 0;
634 #else
635 	getlock = (mode & NAT_LOCKHELD) ? 0 : 1;
636 #endif
637 
638 	nat = NULL;     /* XXX gcc -Wuninitialized */
639 	if (cmd == (ioctlcmd_t)SIOCADNAT) {
640 		KMALLOC(nt, ipnat_t *);
641 	} else {
642 		nt = NULL;
643 	}
644 
645 	if ((cmd == (ioctlcmd_t)SIOCADNAT) || (cmd == (ioctlcmd_t)SIOCRMNAT)) {
646 		if (mode & NAT_SYSSPACE) {
647 			bcopy(data, (char *)&natd, sizeof(natd));
648 			error = 0;
649 		} else {
650 			error = fr_inobj(data, &natd, IPFOBJ_IPNAT);
651 		}
652 
653 	} else if (cmd == (ioctlcmd_t)SIOCIPFFL) { /* SIOCFLNAT & SIOCCNATL */
654 		BCOPYIN(data, &arg, sizeof(arg));
655 	}
656 
657 	if (error != 0)
658 		goto done;
659 
660 	/*
661 	 * For add/delete, look to see if the NAT entry is already present
662 	 */
663 	if ((cmd == (ioctlcmd_t)SIOCADNAT) || (cmd == (ioctlcmd_t)SIOCRMNAT)) {
664 		nat = &natd;
665 		if (nat->in_v == 0)	/* For backward compat. */
666 			nat->in_v = 4;
667 		nat->in_flags &= IPN_USERFLAGS;
668 		if ((nat->in_redir & NAT_MAPBLK) == 0) {
669 			if ((nat->in_flags & IPN_SPLIT) == 0)
670 				nat->in_inip &= nat->in_inmsk;
671 			if ((nat->in_flags & IPN_IPRANGE) == 0)
672 				nat->in_outip &= nat->in_outmsk;
673 		}
674 		MUTEX_ENTER(&ifs->ifs_ipf_natio);
675 		for (np = &ifs->ifs_nat_list; ((n = *np) != NULL);
676 		     np = &n->in_next)
677 			if (bcmp((char *)&nat->in_flags, (char *)&n->in_flags,
678 			    IPN_CMPSIZ) == 0) {
679 				if (nat->in_redir == NAT_REDIRECT &&
680 				    nat->in_pnext != n->in_pnext)
681 					continue;
682 				break;
683 			}
684 	}
685 
686 	switch (cmd)
687 	{
688 	case SIOCGENITER :
689 	    {
690 		ipfgeniter_t iter;
691 		ipftoken_t *token;
692 
693 		error = fr_inobj(data, &iter, IPFOBJ_GENITER);
694 		if (error != 0)
695 			break;
696 
697 		token = ipf_findtoken(iter.igi_type, uid, ctx, ifs);
698 		if (token != NULL)
699 			error  = nat_iterator(token, &iter, ifs);
700 		else
701 			error = ESRCH;
702 		RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
703 		break;
704 	    }
705 #ifdef  IPFILTER_LOG
706 	case SIOCIPFFB :
707 	{
708 		int tmp;
709 
710 		if (!(mode & FWRITE))
711 			error = EPERM;
712 		else {
713 			tmp = ipflog_clear(IPL_LOGNAT, ifs);
714 			BCOPYOUT((char *)&tmp, (char *)data, sizeof(tmp));
715 		}
716 		break;
717 	}
718 	case SIOCSETLG :
719 		if (!(mode & FWRITE))
720 			error = EPERM;
721 		else {
722 			BCOPYIN((char *)data,
723 				       (char *)&ifs->ifs_nat_logging,
724 				sizeof(ifs->ifs_nat_logging));
725 		}
726 		break;
727 	case SIOCGETLG :
728 		BCOPYOUT((char *)&ifs->ifs_nat_logging, (char *)data,
729 			sizeof(ifs->ifs_nat_logging));
730 		break;
731 	case FIONREAD :
732 		arg = ifs->ifs_iplused[IPL_LOGNAT];
733 		BCOPYOUT(&arg, data, sizeof(arg));
734 		break;
735 #endif
736 	case SIOCADNAT :
737 		if (!(mode & FWRITE)) {
738 			error = EPERM;
739 		} else if (n != NULL) {
740 			error = EEXIST;
741 		} else if (nt == NULL) {
742 			error = ENOMEM;
743 		}
744 		if (error != 0) {
745 			MUTEX_EXIT(&ifs->ifs_ipf_natio);
746 			break;
747 		}
748 		bcopy((char *)nat, (char *)nt, sizeof(*n));
749 		error = nat_siocaddnat(nt, np, getlock, ifs);
750 		MUTEX_EXIT(&ifs->ifs_ipf_natio);
751 		if (error == 0)
752 			nt = NULL;
753 		break;
754 	case SIOCRMNAT :
755 		if (!(mode & FWRITE)) {
756 			error = EPERM;
757 			n = NULL;
758 		} else if (n == NULL) {
759 			error = ESRCH;
760 		}
761 
762 		if (error != 0) {
763 			MUTEX_EXIT(&ifs->ifs_ipf_natio);
764 			break;
765 		}
766 		nat_siocdelnat(n, np, getlock, ifs);
767 
768 		MUTEX_EXIT(&ifs->ifs_ipf_natio);
769 		n = NULL;
770 		break;
771 	case SIOCGNATS :
772 		ifs->ifs_nat_stats.ns_table[0] = ifs->ifs_nat_table[0];
773 		ifs->ifs_nat_stats.ns_table[1] = ifs->ifs_nat_table[1];
774 		ifs->ifs_nat_stats.ns_list = ifs->ifs_nat_list;
775 		ifs->ifs_nat_stats.ns_maptable = ifs->ifs_maptable;
776 		ifs->ifs_nat_stats.ns_maplist = ifs->ifs_ipf_hm_maplist;
777 		ifs->ifs_nat_stats.ns_nattab_max = ifs->ifs_ipf_nattable_max;
778 		ifs->ifs_nat_stats.ns_nattab_sz = ifs->ifs_ipf_nattable_sz;
779 		ifs->ifs_nat_stats.ns_rultab_sz = ifs->ifs_ipf_natrules_sz;
780 		ifs->ifs_nat_stats.ns_rdrtab_sz = ifs->ifs_ipf_rdrrules_sz;
781 		ifs->ifs_nat_stats.ns_hostmap_sz = ifs->ifs_ipf_hostmap_sz;
782 		ifs->ifs_nat_stats.ns_instances = ifs->ifs_nat_instances;
783 		ifs->ifs_nat_stats.ns_apslist = ifs->ifs_ap_sess_list;
784 		error = fr_outobj(data, &ifs->ifs_nat_stats, IPFOBJ_NATSTAT);
785 		break;
786 	case SIOCGNATL :
787 	    {
788 		natlookup_t nl;
789 
790 		if (getlock) {
791 			READ_ENTER(&ifs->ifs_ipf_nat);
792 		}
793 		error = fr_inobj(data, &nl, IPFOBJ_NATLOOKUP);
794 		if (nl.nl_v != 6)
795 			nl.nl_v = 4;
796 		if (error == 0) {
797 			void *ptr;
798 
799 			switch (nl.nl_v)
800 			{
801 			case 4:
802 				ptr = nat_lookupredir(&nl, ifs);
803 				break;
804 #ifdef	USE_INET6
805 			case 6:
806 				ptr = nat6_lookupredir(&nl, ifs);
807 				break;
808 #endif
809 			default:
810 				ptr = NULL;
811 				break;
812 			}
813 
814 			if (ptr != NULL) {
815 				error = fr_outobj(data, &nl, IPFOBJ_NATLOOKUP);
816 			} else {
817 				error = ESRCH;
818 			}
819 		}
820 		if (getlock) {
821 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
822 		}
823 		break;
824 	    }
825 	case SIOCIPFFL :	/* old SIOCFLNAT & SIOCCNATL */
826 		if (!(mode & FWRITE)) {
827 			error = EPERM;
828 			break;
829 		}
830 		if (getlock) {
831 			WRITE_ENTER(&ifs->ifs_ipf_nat);
832 		}
833 		error = 0;
834 		if (arg == 0)
835 			ret = nat_flushtable(ifs);
836 		else if (arg == 1)
837 			ret = nat_clearlist(ifs);
838 		else if (arg >= 2 && arg <= 4)
839 			ret = nat_extraflush(arg - 2, ifs);
840 		else
841 			error = EINVAL;
842 		if (getlock) {
843 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
844 		}
845 		if (error == 0) {
846 			BCOPYOUT(&ret, data, sizeof(ret));
847 		}
848 		break;
849 	case SIOCPROXY :
850 		error = appr_ioctl(data, cmd, mode, ifs);
851 		break;
852 	case SIOCSTLCK :
853 		if (!(mode & FWRITE)) {
854 			error = EPERM;
855 		} else {
856 			fr_lock(data, &ifs->ifs_fr_nat_lock);
857 		}
858 		break;
859 	case SIOCSTPUT :
860 		if ((mode & FWRITE) != 0) {
861 			error = fr_natputent(data, getlock, ifs);
862 		} else {
863 			error = EACCES;
864 		}
865 		break;
866 	case SIOCSTGSZ :
867 		if (ifs->ifs_fr_nat_lock) {
868 			if (getlock) {
869 				READ_ENTER(&ifs->ifs_ipf_nat);
870 			}
871 			error = fr_natgetsz(data, ifs);
872 			if (getlock) {
873 				RWLOCK_EXIT(&ifs->ifs_ipf_nat);
874 			}
875 		} else
876 			error = EACCES;
877 		break;
878 	case SIOCSTGET :
879 		if (ifs->ifs_fr_nat_lock) {
880 			if (getlock) {
881 				READ_ENTER(&ifs->ifs_ipf_nat);
882 			}
883 			error = fr_natgetent(data, ifs);
884 			if (getlock) {
885 				RWLOCK_EXIT(&ifs->ifs_ipf_nat);
886 			}
887 		} else
888 			error = EACCES;
889 		break;
890 	case SIOCIPFDELTOK :
891 		(void) BCOPYIN((caddr_t)data, (caddr_t)&arg, sizeof(arg));
892 		error = ipf_deltoken(arg, uid, ctx, ifs);
893 		break;
894 	default :
895 		error = EINVAL;
896 		break;
897 	}
898 done:
899 	if (nt)
900 		KFREE(nt);
901 	return error;
902 }
903 
904 
905 /* ------------------------------------------------------------------------ */
906 /* Function:    nat_siocaddnat                                              */
907 /* Returns:     int - 0 == success, != 0 == failure                         */
908 /* Parameters:  n(I)       - pointer to new NAT rule                        */
909 /*              np(I)      - pointer to where to insert new NAT rule        */
910 /*              getlock(I) - flag indicating if lock on ipf_nat is held     */
911 /* Mutex Locks: ipf_natio                                                   */
912 /*                                                                          */
913 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
914 /* from information passed to the kernel, then add it  to the appropriate   */
915 /* NAT rule table(s).                                                       */
916 /* ------------------------------------------------------------------------ */
917 static int nat_siocaddnat(n, np, getlock, ifs)
918 ipnat_t *n, **np;
919 int getlock;
920 ipf_stack_t *ifs;
921 {
922 	int error = 0, i, j;
923 
924 	if (nat_resolverule(n, ifs) != 0)
925 		return ENOENT;
926 
927 	if ((n->in_age[0] == 0) && (n->in_age[1] != 0))
928 		return EINVAL;
929 
930 	n->in_use = 0;
931 	if (n->in_redir & NAT_MAPBLK)
932 		n->in_space = USABLE_PORTS * ~ntohl(n->in_outmsk);
933 	else if (n->in_flags & IPN_AUTOPORTMAP)
934 		n->in_space = USABLE_PORTS * ~ntohl(n->in_inmsk);
935 	else if (n->in_flags & IPN_IPRANGE)
936 		n->in_space = ntohl(n->in_outmsk) - ntohl(n->in_outip);
937 	else if (n->in_flags & IPN_SPLIT)
938 		n->in_space = 2;
939 	else if (n->in_outmsk != 0)
940 		n->in_space = ~ntohl(n->in_outmsk);
941 	else
942 		n->in_space = 1;
943 
944 	/*
945 	 * Calculate the number of valid IP addresses in the output
946 	 * mapping range.  In all cases, the range is inclusive of
947 	 * the start and ending IP addresses.
948 	 * If to a CIDR address, lose 2: broadcast + network address
949 	 *                               (so subtract 1)
950 	 * If to a range, add one.
951 	 * If to a single IP address, set to 1.
952 	 */
953 	if (n->in_space) {
954 		if ((n->in_flags & IPN_IPRANGE) != 0)
955 			n->in_space += 1;
956 		else
957 			n->in_space -= 1;
958 	} else
959 		n->in_space = 1;
960 
961 #ifdef	USE_INET6
962 	if (n->in_v == 6 && (n->in_flags & (IPN_IPRANGE|IPN_SPLIT)) == 0 &&
963 	    !IP6_ISONES(&n->in_out[1]) && !IP6_ISZERO(&n->in_out[1]))
964 		IP6_ADD(&n->in_out[0], 1, &n->in_next6)
965 	else if (n->in_v == 6 &&
966 	    (n->in_flags & IPN_SPLIT) && (n->in_redir & NAT_REDIRECT))
967 		n->in_next6 = n->in_in[0];
968 	else if (n->in_v == 6)
969 		n->in_next6 = n->in_out[0];
970 	else
971 #endif
972 	if ((n->in_outmsk != 0xffffffff) && (n->in_outmsk != 0) &&
973 	    ((n->in_flags & (IPN_IPRANGE|IPN_SPLIT)) == 0))
974 		n->in_nip = ntohl(n->in_outip) + 1;
975 	else if ((n->in_flags & IPN_SPLIT) &&
976 		 (n->in_redir & NAT_REDIRECT))
977 		n->in_nip = ntohl(n->in_inip);
978 	else
979 		n->in_nip = ntohl(n->in_outip);
980 
981 	if (n->in_redir & NAT_MAP) {
982 		n->in_pnext = ntohs(n->in_pmin);
983 		/*
984 		 * Multiply by the number of ports made available.
985 		 */
986 		if (ntohs(n->in_pmax) >= ntohs(n->in_pmin)) {
987 			n->in_space *= (ntohs(n->in_pmax) -
988 					ntohs(n->in_pmin) + 1);
989 			/*
990 			 * Because two different sources can map to
991 			 * different destinations but use the same
992 			 * local IP#/port #.
993 			 * If the result is smaller than in_space, then
994 			 * we may have wrapped around 32bits.
995 			 */
996 			i = n->in_inmsk;
997 			if ((i != 0) && (i != 0xffffffff)) {
998 				j = n->in_space * (~ntohl(i) + 1);
999 				if (j >= n->in_space)
1000 					n->in_space = j;
1001 				else
1002 					n->in_space = 0xffffffff;
1003 			}
1004 		}
1005 		/*
1006 		 * If no protocol is specified, multiple by 256 to allow for
1007 		 * at least one IP:IP mapping per protocol.
1008 		 */
1009 		if ((n->in_flags & IPN_TCPUDPICMP) == 0) {
1010 				j = n->in_space * 256;
1011 				if (j >= n->in_space)
1012 					n->in_space = j;
1013 				else
1014 					n->in_space = 0xffffffff;
1015 		}
1016 	}
1017 
1018 	/* Otherwise, these fields are preset */
1019 
1020 	if (getlock) {
1021 		WRITE_ENTER(&ifs->ifs_ipf_nat);
1022 	}
1023 	n->in_next = NULL;
1024 	*np = n;
1025 
1026 	if (n->in_age[0] != 0)
1027 	    n->in_tqehead[0] = fr_addtimeoutqueue(&ifs->ifs_nat_utqe,
1028 						  n->in_age[0], ifs);
1029 
1030 	if (n->in_age[1] != 0)
1031 	    n->in_tqehead[1] = fr_addtimeoutqueue(&ifs->ifs_nat_utqe,
1032 						  n->in_age[1], ifs);
1033 
1034 	if (n->in_redir & NAT_REDIRECT) {
1035 		n->in_flags &= ~IPN_NOTDST;
1036 		switch (n->in_v)
1037 		{
1038 		case 4 :
1039 			nat_addrdr(n, ifs);
1040 			break;
1041 #ifdef	USE_INET6
1042 		case 6 :
1043 			nat6_addrdr(n, ifs);
1044 			break;
1045 #endif
1046 		default :
1047 			break;
1048 		}
1049 	}
1050 	if (n->in_redir & (NAT_MAP|NAT_MAPBLK)) {
1051 		n->in_flags &= ~IPN_NOTSRC;
1052 		switch (n->in_v)
1053 		{
1054 		case 4 :
1055 			nat_addnat(n, ifs);
1056 			break;
1057 #ifdef	USE_INET6
1058 		case 6 :
1059 			nat6_addnat(n, ifs);
1060 			break;
1061 #endif
1062 		default :
1063 			break;
1064 		}
1065 	}
1066 	n = NULL;
1067 	ifs->ifs_nat_stats.ns_rules++;
1068 	if (getlock) {
1069 		RWLOCK_EXIT(&ifs->ifs_ipf_nat);			/* WRITE */
1070 	}
1071 
1072 	return error;
1073 }
1074 
1075 
1076 /* ------------------------------------------------------------------------ */
1077 /* Function:    nat_resolvrule                                              */
1078 /* Returns:     int - 0 == success, -1 == failure                           */
1079 /* Parameters:  n(I)  - pointer to NAT rule                                 */
1080 /*                                                                          */
1081 /* Resolve some of the details inside the NAT rule.  Includes resolving	    */
1082 /* any specified interfaces and proxy labels, and determines whether or not */
1083 /* all proxy labels are correctly specified.				    */
1084 /*									    */
1085 /* Called by nat_siocaddnat() (SIOCADNAT) and fr_natputent (SIOCSTPUT).     */
1086 /* ------------------------------------------------------------------------ */
1087 static int nat_resolverule(n, ifs)
1088 ipnat_t *n;
1089 ipf_stack_t *ifs;
1090 {
1091 	n->in_ifnames[0][LIFNAMSIZ - 1] = '\0';
1092 	n->in_ifps[0] = fr_resolvenic(n->in_ifnames[0], n->in_v, ifs);
1093 
1094 	n->in_ifnames[1][LIFNAMSIZ - 1] = '\0';
1095 	if (n->in_ifnames[1][0] == '\0') {
1096 		(void) strncpy(n->in_ifnames[1], n->in_ifnames[0], LIFNAMSIZ);
1097 		n->in_ifps[1] = n->in_ifps[0];
1098 	} else {
1099 		n->in_ifps[1] = fr_resolvenic(n->in_ifnames[1], n->in_v, ifs);
1100 	}
1101 
1102 	if (n->in_plabel[0] != '\0') {
1103 		n->in_apr = appr_lookup(n->in_p, n->in_plabel, ifs);
1104 		if (n->in_apr == NULL)
1105 			return -1;
1106 	}
1107 	return 0;
1108 }
1109 
1110 
1111 /* ------------------------------------------------------------------------ */
1112 /* Function:    nat_siocdelnat                                              */
1113 /* Returns:     int - 0 == success, != 0 == failure                         */
1114 /* Parameters:  n(I)       - pointer to new NAT rule                        */
1115 /*              np(I)      - pointer to where to insert new NAT rule        */
1116 /*              getlock(I) - flag indicating if lock on ipf_nat is held     */
1117 /* Mutex Locks: ipf_natio                                                   */
1118 /*                                                                          */
1119 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
1120 /* from information passed to the kernel, then add it  to the appropriate   */
1121 /* NAT rule table(s).                                                       */
1122 /* ------------------------------------------------------------------------ */
1123 static void nat_siocdelnat(n, np, getlock, ifs)
1124 ipnat_t *n, **np;
1125 int getlock;
1126 ipf_stack_t *ifs;
1127 {
1128 	int i;
1129 
1130 	if (getlock) {
1131 		WRITE_ENTER(&ifs->ifs_ipf_nat);
1132 	}
1133 	if (n->in_redir & NAT_REDIRECT)
1134 		nat_delrdr(n);
1135 	if (n->in_redir & (NAT_MAPBLK|NAT_MAP))
1136 		nat_delnat(n);
1137 	if (ifs->ifs_nat_list == NULL) {
1138 		ifs->ifs_nat_masks = 0;
1139 		ifs->ifs_rdr_masks = 0;
1140 		for (i = 0; i < 4; i++) {
1141 			ifs->ifs_nat6_masks[i] = 0;
1142 			ifs->ifs_rdr6_masks[i] = 0;
1143 		}
1144 	}
1145 
1146 	if (n->in_tqehead[0] != NULL) {
1147 		if (fr_deletetimeoutqueue(n->in_tqehead[0]) == 0) {
1148 			fr_freetimeoutqueue(n->in_tqehead[0], ifs);
1149 		}
1150 	}
1151 
1152 	if (n->in_tqehead[1] != NULL) {
1153 		if (fr_deletetimeoutqueue(n->in_tqehead[1]) == 0) {
1154 			fr_freetimeoutqueue(n->in_tqehead[1], ifs);
1155 		}
1156 	}
1157 
1158 	*np = n->in_next;
1159 
1160 	if (n->in_use == 0) {
1161 		if (n->in_apr)
1162 			appr_free(n->in_apr);
1163 		KFREE(n);
1164 		ifs->ifs_nat_stats.ns_rules--;
1165 	} else {
1166 		n->in_flags |= IPN_DELETE;
1167 		n->in_next = NULL;
1168 	}
1169 	if (getlock) {
1170 		RWLOCK_EXIT(&ifs->ifs_ipf_nat);			/* READ/WRITE */
1171 	}
1172 }
1173 
1174 
1175 /* ------------------------------------------------------------------------ */
1176 /* Function:    fr_natgetsz                                                 */
1177 /* Returns:     int - 0 == success, != 0 is the error value.                */
1178 /* Parameters:  data(I) - pointer to natget structure with kernel pointer   */
1179 /*                        get the size of.                                  */
1180 /*                                                                          */
1181 /* Handle SIOCSTGSZ.                                                        */
1182 /* Return the size of the nat list entry to be copied back to user space.   */
1183 /* The size of the entry is stored in the ng_sz field and the enture natget */
1184 /* structure is copied back to the user.                                    */
1185 /* ------------------------------------------------------------------------ */
1186 static int fr_natgetsz(data, ifs)
1187 caddr_t data;
1188 ipf_stack_t *ifs;
1189 {
1190 	ap_session_t *aps;
1191 	nat_t *nat, *n;
1192 	natget_t ng;
1193 
1194 	BCOPYIN(data, &ng, sizeof(ng));
1195 
1196 	nat = ng.ng_ptr;
1197 	if (!nat) {
1198 		nat = ifs->ifs_nat_instances;
1199 		ng.ng_sz = 0;
1200 		/*
1201 		 * Empty list so the size returned is 0.  Simple.
1202 		 */
1203 		if (nat == NULL) {
1204 			BCOPYOUT(&ng, data, sizeof(ng));
1205 			return 0;
1206 		}
1207 	} else {
1208 		/*
1209 		 * Make sure the pointer we're copying from exists in the
1210 		 * current list of entries.  Security precaution to prevent
1211 		 * copying of random kernel data.
1212 		 */
1213 		for (n = ifs->ifs_nat_instances; n; n = n->nat_next)
1214 			if (n == nat)
1215 				break;
1216 		if (!n)
1217 			return ESRCH;
1218 	}
1219 
1220 	/*
1221 	 * Incluse any space required for proxy data structures.
1222 	 */
1223 	ng.ng_sz = sizeof(nat_save_t);
1224 	aps = nat->nat_aps;
1225 	if (aps != NULL) {
1226 		ng.ng_sz += sizeof(ap_session_t) - 4;
1227 		if (aps->aps_data != 0)
1228 			ng.ng_sz += aps->aps_psiz;
1229 	}
1230 
1231 	BCOPYOUT(&ng, data, sizeof(ng));
1232 	return 0;
1233 }
1234 
1235 
1236 /* ------------------------------------------------------------------------ */
1237 /* Function:    fr_natgetent                                                */
1238 /* Returns:     int - 0 == success, != 0 is the error value.                */
1239 /* Parameters:  data(I) - pointer to natget structure with kernel pointer   */
1240 /*                        to NAT structure to copy out.                     */
1241 /*                                                                          */
1242 /* Handle SIOCSTGET.                                                        */
1243 /* Copies out NAT entry to user space.  Any additional data held for a      */
1244 /* proxy is also copied, as to is the NAT rule which was responsible for it */
1245 /* ------------------------------------------------------------------------ */
1246 static int fr_natgetent(data, ifs)
1247 caddr_t data;
1248 ipf_stack_t *ifs;
1249 {
1250 	int error, outsize;
1251 	ap_session_t *aps;
1252 	nat_save_t *ipn, ipns;
1253 	nat_t *n, *nat;
1254 
1255 	error = fr_inobj(data, &ipns, IPFOBJ_NATSAVE);
1256 	if (error != 0)
1257 		return error;
1258 
1259 	if ((ipns.ipn_dsize < sizeof(ipns)) || (ipns.ipn_dsize > 81920))
1260 		return EINVAL;
1261 
1262 	KMALLOCS(ipn, nat_save_t *, ipns.ipn_dsize);
1263 	if (ipn == NULL)
1264 		return ENOMEM;
1265 
1266 	ipn->ipn_dsize = ipns.ipn_dsize;
1267 	nat = ipns.ipn_next;
1268 	if (nat == NULL) {
1269 		nat = ifs->ifs_nat_instances;
1270 		if (nat == NULL) {
1271 			if (ifs->ifs_nat_instances == NULL)
1272 				error = ENOENT;
1273 			goto finished;
1274 		}
1275 	} else {
1276 		/*
1277 		 * Make sure the pointer we're copying from exists in the
1278 		 * current list of entries.  Security precaution to prevent
1279 		 * copying of random kernel data.
1280 		 */
1281 		for (n = ifs->ifs_nat_instances; n; n = n->nat_next)
1282 			if (n == nat)
1283 				break;
1284 		if (n == NULL) {
1285 			error = ESRCH;
1286 			goto finished;
1287 		}
1288 	}
1289 	ipn->ipn_next = nat->nat_next;
1290 
1291 	/*
1292 	 * Copy the NAT structure.
1293 	 */
1294 	bcopy((char *)nat, &ipn->ipn_nat, sizeof(*nat));
1295 
1296 	/*
1297 	 * If we have a pointer to the NAT rule it belongs to, save that too.
1298 	 */
1299 	if (nat->nat_ptr != NULL)
1300 		bcopy((char *)nat->nat_ptr, (char *)&ipn->ipn_ipnat,
1301 		      sizeof(ipn->ipn_ipnat));
1302 
1303 	/*
1304 	 * If we also know the NAT entry has an associated filter rule,
1305 	 * save that too.
1306 	 */
1307 	if (nat->nat_fr != NULL)
1308 		bcopy((char *)nat->nat_fr, (char *)&ipn->ipn_fr,
1309 		      sizeof(ipn->ipn_fr));
1310 
1311 	/*
1312 	 * Last but not least, if there is an application proxy session set
1313 	 * up for this NAT entry, then copy that out too, including any
1314 	 * private data saved along side it by the proxy.
1315 	 */
1316 	aps = nat->nat_aps;
1317 	outsize = ipn->ipn_dsize - sizeof(*ipn) + sizeof(ipn->ipn_data);
1318 	if (aps != NULL) {
1319 		char *s;
1320 
1321 		if (outsize < sizeof(*aps)) {
1322 			error = ENOBUFS;
1323 			goto finished;
1324 		}
1325 
1326 		s = ipn->ipn_data;
1327 		bcopy((char *)aps, s, sizeof(*aps));
1328 		s += sizeof(*aps);
1329 		outsize -= sizeof(*aps);
1330 		if ((aps->aps_data != NULL) && (outsize >= aps->aps_psiz))
1331 			bcopy(aps->aps_data, s, aps->aps_psiz);
1332 		else
1333 			error = ENOBUFS;
1334 	}
1335 	if (error == 0) {
1336 		error = fr_outobjsz(data, ipn, IPFOBJ_NATSAVE, ipns.ipn_dsize);
1337 	}
1338 
1339 finished:
1340 	if (ipn != NULL) {
1341 		KFREES(ipn, ipns.ipn_dsize);
1342 	}
1343 	return error;
1344 }
1345 
1346 /* ------------------------------------------------------------------------ */
1347 /* Function:    nat_calc_chksum_diffs					    */
1348 /* Returns:     void							    */
1349 /* Parameters:  nat	-	pointer to NAT table entry		    */
1350 /*                                                                          */
1351 /* Function calculates chksum deltas for IP header (nat_ipsumd) and TCP/UDP */
1352 /* headers (nat_sumd). The things for L4 (UDP/TCP) get complicated when     */
1353 /* we are dealing with partial chksum offload. For these cases we need to   */
1354 /* compute a 'partial chksum delta'. The 'partial chksum delta'is stored    */
1355 /* into nat_sumd[1], while ordinary chksum delta for TCP/UDP is in 	    */
1356 /* nat_sumd[0]. 							    */
1357 /*									    */
1358 /* The function accepts initialized NAT table entry and computes the deltas */
1359 /* from nat_inip/nat_outip members. The function is called right before	    */
1360 /* the new entry is inserted into the table.				    */
1361 /*									    */
1362 /* The ipsumd (IP hedaer chksum delta adjustment) is computed as a chksum   */
1363 /* of delta between original and new IP addresses.			    */
1364 /*									    */
1365 /* the nat_sumd[0] (TCP/UDP header chksum delta adjustment) is computed as  */
1366 /* a chkusm of delta between original an new IP addrress:port tupples.	    */
1367 /*									    */
1368 /* Some facts about chksum, we should remember:				    */
1369 /*	IP header chksum covers IP header only				    */
1370 /*									    */
1371 /*	TCP/UDP chksum covers data payload and so called pseudo header	    */
1372 /*		SRC, DST IP address					    */
1373 /*		SRC, DST Port						    */
1374 /*		length of payload					    */
1375 /*									    */
1376 /* The partial chksum delta (nat_sumd[1] is used to adjust db_ckusm16	    */
1377 /* member of dblk_t structure. The db_ckusm16 member is not part of 	    */
1378 /* IP/UDP/TCP header it is 16 bit value computed by NIC driver with partial */
1379 /* chksum offload capacbility for every inbound packet. The db_cksum16 is   */
1380 /* stored along with other IP packet data in dblk_t structure and used in   */
1381 /* for IP/UDP/TCP chksum validation later in ip.c. 			    */
1382 /*									    */
1383 /* The partial chksum delta (adjustment, nat_sumd[1]) is computed as chksum */
1384 /* of delta between new and orig address. NOTE: the order of operands for   */
1385 /* partial delta operation is swapped compared to computing the IP/TCP/UDP  */
1386 /* header adjustment. It is by design see (IP_CKSUM_RECV() macro in ip.c).  */
1387 /*									    */
1388 /* ------------------------------------------------------------------------ */
1389 void nat_calc_chksum_diffs(nat)
1390 nat_t *nat;
1391 {
1392 	u_32_t	sum_orig = 0;
1393 	u_32_t	sum_changed = 0;
1394 	u_32_t	sumd;
1395 	u_32_t	ipsum_orig = 0;
1396 	u_32_t	ipsum_changed = 0;
1397 
1398 	if (nat->nat_v != 4 && nat->nat_v != 6)
1399 		return;
1400 
1401 	/*
1402 	 * the switch calculates operands for CALC_SUMD(),
1403 	 * which will compute the partial chksum delta.
1404 	 */
1405 	switch (nat->nat_dir)
1406 	{
1407 	case NAT_INBOUND:
1408 		/*
1409 		 * we are dealing with RDR rule (DST address gets
1410 		 * modified on packet from client)
1411 		 */
1412 		if (nat->nat_v == 4) {
1413 			sum_changed = LONG_SUM(ntohl(nat->nat_inip.s_addr));
1414 			sum_orig = LONG_SUM(ntohl(nat->nat_outip.s_addr));
1415 		} else {
1416 			sum_changed = LONG_SUM6(&nat->nat_inip6);
1417 			sum_orig = LONG_SUM6(&nat->nat_outip6);
1418 		}
1419 		break;
1420 	case NAT_OUTBOUND:
1421 		/*
1422 		 * we are dealing with MAP rule (SRC address gets
1423 		 * modified on packet from client)
1424 		 */
1425 		if (nat->nat_v == 4) {
1426 			sum_changed = LONG_SUM(ntohl(nat->nat_outip.s_addr));
1427 			sum_orig = LONG_SUM(ntohl(nat->nat_inip.s_addr));
1428 		} else {
1429 			sum_changed = LONG_SUM6(&nat->nat_outip6);
1430 			sum_orig = LONG_SUM6(&nat->nat_inip6);
1431 		}
1432 		break;
1433 	default: ;
1434 		break;
1435 	}
1436 
1437 	/*
1438 	 * we also preserve CALC_SUMD() operands here, for IP chksum delta
1439 	 * calculation, which happens at the end of function.
1440 	 */
1441 	ipsum_changed = sum_changed;
1442 	ipsum_orig = sum_orig;
1443 	/*
1444 	 * NOTE: the order of operands for partial chksum adjustment
1445 	 * computation has to be swapped!
1446 	 */
1447 	CALC_SUMD(sum_changed, sum_orig, sumd);
1448 	nat->nat_sumd[1] = (sumd & 0xffff) + (sumd >> 16);
1449 
1450 	if (nat->nat_flags & (IPN_TCPUDP | IPN_ICMPQUERY)) {
1451 
1452 		/*
1453 		 * switch calculates operands for CALC_SUMD(), which will
1454 		 * compute the full chksum delta.
1455 		 */
1456 		switch (nat->nat_dir)
1457 		{
1458 		case NAT_INBOUND:
1459 			if (nat->nat_v == 4) {
1460 				sum_changed = LONG_SUM(
1461 				    ntohl(nat->nat_inip.s_addr) +
1462 				    ntohs(nat->nat_inport));
1463 				sum_orig = LONG_SUM(
1464 				    ntohl(nat->nat_outip.s_addr) +
1465 				    ntohs(nat->nat_outport));
1466 			} else {
1467 				sum_changed = LONG_SUM6(&nat->nat_inip6) +
1468 				    ntohs(nat->nat_inport);
1469 				sum_orig = LONG_SUM6(&nat->nat_outip6) +
1470 				    ntohs(nat->nat_outport);
1471 			}
1472 			break;
1473 		case NAT_OUTBOUND:
1474 			if (nat->nat_v == 4) {
1475 				sum_changed = LONG_SUM(
1476 				    ntohl(nat->nat_outip.s_addr) +
1477 				    ntohs(nat->nat_outport));
1478 				sum_orig = LONG_SUM(
1479 				    ntohl(nat->nat_inip.s_addr) +
1480 				    ntohs(nat->nat_inport));
1481 			} else {
1482 				sum_changed = LONG_SUM6(&nat->nat_outip6) +
1483 				    ntohs(nat->nat_outport);
1484 				sum_orig = LONG_SUM6(&nat->nat_inip6) +
1485 				    ntohs(nat->nat_inport);
1486 			}
1487 			break;
1488 		default: ;
1489 			break;
1490 		}
1491 
1492 		CALC_SUMD(sum_orig, sum_changed, sumd);
1493 		nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
1494 
1495 		if (!(nat->nat_flags & IPN_TCPUDP)) {
1496 			/*
1497 			 * partial HW chksum offload works for TCP/UDP headers only,
1498 			 * so we need to enforce full chksum adjustment for ICMP
1499 			 */
1500 			nat->nat_sumd[1] = nat->nat_sumd[0];
1501 		}
1502 	}
1503 	else
1504 		nat->nat_sumd[0] = nat->nat_sumd[1];
1505 
1506 	/*
1507 	 * we may reuse the already computed nat_sumd[0] for IP header chksum
1508 	 * adjustment in case the L4 (TCP/UDP header) is not changed by NAT.
1509 	 */
1510 	if (nat->nat_v == 4) {
1511 		if (NAT_HAS_L4_CHANGED(nat)) {
1512 			/*
1513 			 * bad luck, NAT changes also the L4 header, use IP
1514 			 * addresses to compute chksum adjustment for IP header.
1515 			 */
1516 			CALC_SUMD(ipsum_orig, ipsum_changed, sumd);
1517 			nat->nat_ipsumd = (sumd & 0xffff) + (sumd >> 16);
1518 		} else {
1519 			/*
1520 			 * the NAT does not change L4 hdr -> reuse chksum
1521 			 * adjustment for IP hdr.
1522 			 */
1523 			nat->nat_ipsumd = nat->nat_sumd[0];
1524 
1525 			/*
1526 			 * if L4 header does not use chksum - zero out deltas
1527 			 */
1528 			if (!(nat->nat_flags & IPN_TCPUDP)) {
1529 				nat->nat_sumd[0] = 0;
1530 				nat->nat_sumd[1] = 0;
1531 			}
1532 		}
1533 	}
1534 
1535 	return;
1536 }
1537 
1538 /* ------------------------------------------------------------------------ */
1539 /* Function:    fr_natputent                                                */
1540 /* Returns:     int - 0 == success, != 0 is the error value.                */
1541 /* Parameters:  data(I) -     pointer to natget structure with NAT          */
1542 /*                            structure information to load into the kernel */
1543 /*              getlock(I) - flag indicating whether or not a write lock    */
1544 /*                           on ipf_nat is already held.                    */
1545 /*                                                                          */
1546 /* Handle SIOCSTPUT.                                                        */
1547 /* Loads a NAT table entry from user space, including a NAT rule, proxy and */
1548 /* firewall rule data structures, if pointers to them indicate so.          */
1549 /* ------------------------------------------------------------------------ */
1550 static int fr_natputent(data, getlock, ifs)
1551 caddr_t data;
1552 int getlock;
1553 ipf_stack_t *ifs;
1554 {
1555 	nat_save_t ipn, *ipnn;
1556 	ap_session_t *aps;
1557 	nat_t *n, *nat;
1558 	frentry_t *fr;
1559 	fr_info_t fin;
1560 	ipnat_t *in;
1561 	int error;
1562 
1563 	error = fr_inobj(data, &ipn, IPFOBJ_NATSAVE);
1564 	if (error != 0)
1565 		return error;
1566 
1567 	/*
1568 	 * Trigger automatic call to nat_extraflush() if the
1569 	 * table has reached capcity specified by hi watermark.
1570 	 */
1571 	if (NAT_TAB_WATER_LEVEL(ifs) > ifs->ifs_nat_flush_lvl_hi)
1572 		ifs->ifs_nat_doflush = 1;
1573 
1574 	/*
1575 	 * Initialise early because of code at junkput label.
1576 	 */
1577 	in = NULL;
1578 	aps = NULL;
1579 	nat = NULL;
1580 	ipnn = NULL;
1581 
1582 	/*
1583 	 * New entry, copy in the rest of the NAT entry if it's size is more
1584 	 * than just the nat_t structure.
1585 	 */
1586 	fr = NULL;
1587 	if (ipn.ipn_dsize > sizeof(ipn)) {
1588 		if (ipn.ipn_dsize > 81920) {
1589 			error = ENOMEM;
1590 			goto junkput;
1591 		}
1592 
1593 		KMALLOCS(ipnn, nat_save_t *, ipn.ipn_dsize);
1594 		if (ipnn == NULL)
1595 			return ENOMEM;
1596 
1597 		error = fr_inobjsz(data, ipnn, IPFOBJ_NATSAVE, ipn.ipn_dsize);
1598 		if (error != 0) {
1599 			error = EFAULT;
1600 			goto junkput;
1601 		}
1602 	} else
1603 		ipnn = &ipn;
1604 
1605 	KMALLOC(nat, nat_t *);
1606 	if (nat == NULL) {
1607 		error = ENOMEM;
1608 		goto junkput;
1609 	}
1610 
1611 	bcopy((char *)&ipnn->ipn_nat, (char *)nat, sizeof(*nat));
1612 	/*
1613 	 * Initialize all these so that nat_delete() doesn't cause a crash.
1614 	 */
1615 	bzero((char *)nat, offsetof(struct nat, nat_tqe));
1616 	nat->nat_tqe.tqe_pnext = NULL;
1617 	nat->nat_tqe.tqe_next = NULL;
1618 	nat->nat_tqe.tqe_ifq = NULL;
1619 	nat->nat_tqe.tqe_parent = nat;
1620 
1621 	/*
1622 	 * Restore the rule associated with this nat session
1623 	 */
1624 	in = ipnn->ipn_nat.nat_ptr;
1625 	if (in != NULL) {
1626 		KMALLOC(in, ipnat_t *);
1627 		nat->nat_ptr = in;
1628 		if (in == NULL) {
1629 			error = ENOMEM;
1630 			goto junkput;
1631 		}
1632 		bzero((char *)in, offsetof(struct ipnat, in_next6));
1633 		bcopy((char *)&ipnn->ipn_ipnat, (char *)in, sizeof(*in));
1634 		in->in_use = 1;
1635 		in->in_flags |= IPN_DELETE;
1636 
1637 		ATOMIC_INC(ifs->ifs_nat_stats.ns_rules);
1638 
1639 		if (nat_resolverule(in, ifs) != 0) {
1640 			error = ESRCH;
1641 			goto junkput;
1642 		}
1643 	}
1644 
1645 	/*
1646 	 * Check that the NAT entry doesn't already exist in the kernel.
1647 	 */
1648 	if (nat->nat_v != 6)
1649 		nat->nat_v = 4;
1650 	bzero((char *)&fin, sizeof(fin));
1651 	fin.fin_p = nat->nat_p;
1652 	fin.fin_ifs = ifs;
1653 	if (nat->nat_dir == NAT_OUTBOUND) {
1654 		fin.fin_data[0] = ntohs(nat->nat_oport);
1655 		fin.fin_data[1] = ntohs(nat->nat_outport);
1656 		fin.fin_ifp = nat->nat_ifps[0];
1657 		if (getlock) {
1658 			READ_ENTER(&ifs->ifs_ipf_nat);
1659 		}
1660 
1661 		switch (nat->nat_v)
1662 		{
1663 		case 4:
1664 			fin.fin_v = nat->nat_v;
1665 			n = nat_inlookup(&fin, nat->nat_flags, fin.fin_p,
1666 			    nat->nat_oip, nat->nat_outip);
1667 			break;
1668 #ifdef USE_INET6
1669 		case 6:
1670 			n = nat6_inlookup(&fin, nat->nat_flags, fin.fin_p,
1671 			    &nat->nat_oip6.in6, &nat->nat_outip6.in6);
1672 			break;
1673 #endif
1674 		default:
1675 			n = NULL;
1676 			break;
1677 		}
1678 
1679 		if (getlock) {
1680 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
1681 		}
1682 		if (n != NULL) {
1683 			error = EEXIST;
1684 			goto junkput;
1685 		}
1686 	} else if (nat->nat_dir == NAT_INBOUND) {
1687 		fin.fin_data[0] = ntohs(nat->nat_inport);
1688 		fin.fin_data[1] = ntohs(nat->nat_oport);
1689 		fin.fin_ifp = nat->nat_ifps[1];
1690 		if (getlock) {
1691 			READ_ENTER(&ifs->ifs_ipf_nat);
1692 		}
1693 
1694 		switch (nat->nat_v)
1695 		{
1696 		case 4:
1697 			n = nat_outlookup(&fin, nat->nat_flags, fin.fin_p,
1698 			    nat->nat_inip, nat->nat_oip);
1699 			break;
1700 #ifdef USE_INET6
1701 		case 6:
1702 			n = nat6_outlookup(&fin, nat->nat_flags, fin.fin_p,
1703 			    &nat->nat_inip6.in6, &nat->nat_oip6.in6);
1704 			break;
1705 #endif
1706 		default:
1707 			n = NULL;
1708 			break;
1709 		}
1710 
1711 		if (getlock) {
1712 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
1713 		}
1714 		if (n != NULL) {
1715 			error = EEXIST;
1716 			goto junkput;
1717 		}
1718 	} else {
1719 		error = EINVAL;
1720 		goto junkput;
1721 	}
1722 
1723 	/*
1724 	 * Restore ap_session_t structure.  Include the private data allocated
1725 	 * if it was there.
1726 	 */
1727 	aps = nat->nat_aps;
1728 	if (aps != NULL) {
1729 		KMALLOC(aps, ap_session_t *);
1730 		nat->nat_aps = aps;
1731 		if (aps == NULL) {
1732 			error = ENOMEM;
1733 			goto junkput;
1734 		}
1735 		bcopy(ipnn->ipn_data, (char *)aps, sizeof(*aps));
1736 		if (in != NULL)
1737 			aps->aps_apr = in->in_apr;
1738 		else
1739 			aps->aps_apr = NULL;
1740 		if (aps->aps_psiz != 0) {
1741 			if (aps->aps_psiz > 81920) {
1742 				error = ENOMEM;
1743 				goto junkput;
1744 			}
1745 			KMALLOCS(aps->aps_data, void *, aps->aps_psiz);
1746 			if (aps->aps_data == NULL) {
1747 				error = ENOMEM;
1748 				goto junkput;
1749 			}
1750 			bcopy(ipnn->ipn_data + sizeof(*aps), aps->aps_data,
1751 			      aps->aps_psiz);
1752 		} else {
1753 			aps->aps_psiz = 0;
1754 			aps->aps_data = NULL;
1755 		}
1756 	}
1757 
1758 	/*
1759 	 * If there was a filtering rule associated with this entry then
1760 	 * build up a new one.
1761 	 */
1762 	fr = nat->nat_fr;
1763 	if (fr != NULL) {
1764 		if ((nat->nat_flags & SI_NEWFR) != 0) {
1765 			KMALLOC(fr, frentry_t *);
1766 			nat->nat_fr = fr;
1767 			if (fr == NULL) {
1768 				error = ENOMEM;
1769 				goto junkput;
1770 			}
1771 			ipnn->ipn_nat.nat_fr = fr;
1772 			(void) fr_outobj(data, ipnn, IPFOBJ_NATSAVE);
1773 			bcopy((char *)&ipnn->ipn_fr, (char *)fr, sizeof(*fr));
1774 
1775 			fr->fr_ref = 1;
1776 			fr->fr_dsize = 0;
1777 			fr->fr_data = NULL;
1778 			fr->fr_type = FR_T_NONE;
1779 
1780 			MUTEX_NUKE(&fr->fr_lock);
1781 			MUTEX_INIT(&fr->fr_lock, "nat-filter rule lock");
1782 		} else {
1783 			if (getlock) {
1784 				READ_ENTER(&ifs->ifs_ipf_nat);
1785 			}
1786 			for (n = ifs->ifs_nat_instances; n; n = n->nat_next)
1787 				if (n->nat_fr == fr)
1788 					break;
1789 
1790 			if (n != NULL) {
1791 				MUTEX_ENTER(&fr->fr_lock);
1792 				fr->fr_ref++;
1793 				MUTEX_EXIT(&fr->fr_lock);
1794 			}
1795 			if (getlock) {
1796 				RWLOCK_EXIT(&ifs->ifs_ipf_nat);
1797 			}
1798 			if (!n) {
1799 				error = ESRCH;
1800 				goto junkput;
1801 			}
1802 		}
1803 	}
1804 
1805 	if (ipnn != &ipn) {
1806 		KFREES(ipnn, ipn.ipn_dsize);
1807 		ipnn = NULL;
1808 	}
1809 
1810 	nat_calc_chksum_diffs(nat);
1811 
1812 	if (getlock) {
1813 		WRITE_ENTER(&ifs->ifs_ipf_nat);
1814 	}
1815 
1816 	nat_calc_chksum_diffs(nat);
1817 
1818 	switch (nat->nat_v)
1819 	{
1820 	case 4 :
1821 		error = nat_insert(nat, nat->nat_rev, ifs);
1822 		break;
1823 #ifdef USE_INET6
1824 	case 6 :
1825 		error = nat6_insert(nat, nat->nat_rev, ifs);
1826 		break;
1827 #endif
1828 	default :
1829 		break;
1830 	}
1831 
1832 	if ((error == 0) && (aps != NULL)) {
1833 		aps->aps_next = ifs->ifs_ap_sess_list;
1834 		ifs->ifs_ap_sess_list = aps;
1835 	}
1836 	if (getlock) {
1837 		RWLOCK_EXIT(&ifs->ifs_ipf_nat);
1838 	}
1839 
1840 	if (error == 0)
1841 		return 0;
1842 
1843 	error = ENOMEM;
1844 
1845 junkput:
1846 	if (fr != NULL)
1847 		(void) fr_derefrule(&fr, ifs);
1848 
1849 	if ((ipnn != NULL) && (ipnn != &ipn)) {
1850 		KFREES(ipnn, ipn.ipn_dsize);
1851 	}
1852 	if (nat != NULL) {
1853 		if (aps != NULL) {
1854 			if (aps->aps_data != NULL) {
1855 				KFREES(aps->aps_data, aps->aps_psiz);
1856 			}
1857 			KFREE(aps);
1858 		}
1859 		if (in != NULL) {
1860 			if (in->in_apr)
1861 				appr_free(in->in_apr);
1862 			KFREE(in);
1863 		}
1864 		KFREE(nat);
1865 	}
1866 	return error;
1867 }
1868 
1869 
1870 /* ------------------------------------------------------------------------ */
1871 /* Function:    nat_delete                                                  */
1872 /* Returns:     Nil                                                         */
1873 /* Parameters:  natd(I)    - pointer to NAT structure to delete             */
1874 /*              logtype(I) - type of LOG record to create before deleting   */
1875 /* Write Lock:  ipf_nat                                                     */
1876 /*                                                                          */
1877 /* Delete a nat entry from the various lists and table.  If NAT logging is  */
1878 /* enabled then generate a NAT log record for this event.                   */
1879 /* ------------------------------------------------------------------------ */
1880 static void nat_delete(nat, logtype, ifs)
1881 struct nat *nat;
1882 int logtype;
1883 ipf_stack_t *ifs;
1884 {
1885 	struct ipnat *ipn;
1886 
1887 	if (logtype != 0 && ifs->ifs_nat_logging != 0)
1888 		nat_log(nat, logtype, ifs);
1889 
1890 	/*
1891 	 * Take it as a general indication that all the pointers are set if
1892 	 * nat_pnext is set.
1893 	 */
1894 	if (nat->nat_pnext != NULL) {
1895 		ifs->ifs_nat_stats.ns_bucketlen[0][nat->nat_hv[0]]--;
1896 		ifs->ifs_nat_stats.ns_bucketlen[1][nat->nat_hv[1]]--;
1897 
1898 		*nat->nat_pnext = nat->nat_next;
1899 		if (nat->nat_next != NULL) {
1900 			nat->nat_next->nat_pnext = nat->nat_pnext;
1901 			nat->nat_next = NULL;
1902 		}
1903 		nat->nat_pnext = NULL;
1904 
1905 		*nat->nat_phnext[0] = nat->nat_hnext[0];
1906 		if (nat->nat_hnext[0] != NULL) {
1907 			nat->nat_hnext[0]->nat_phnext[0] = nat->nat_phnext[0];
1908 			nat->nat_hnext[0] = NULL;
1909 		}
1910 		nat->nat_phnext[0] = NULL;
1911 
1912 		*nat->nat_phnext[1] = nat->nat_hnext[1];
1913 		if (nat->nat_hnext[1] != NULL) {
1914 			nat->nat_hnext[1]->nat_phnext[1] = nat->nat_phnext[1];
1915 			nat->nat_hnext[1] = NULL;
1916 		}
1917 		nat->nat_phnext[1] = NULL;
1918 
1919 		if ((nat->nat_flags & SI_WILDP) != 0)
1920 			ifs->ifs_nat_stats.ns_wilds--;
1921 	}
1922 
1923 	if (nat->nat_me != NULL) {
1924 		*nat->nat_me = NULL;
1925 		nat->nat_me = NULL;
1926 	}
1927 
1928 	fr_deletequeueentry(&nat->nat_tqe);
1929 
1930 	MUTEX_ENTER(&nat->nat_lock);
1931 	if (nat->nat_ref > 1) {
1932 		nat->nat_ref--;
1933 		MUTEX_EXIT(&nat->nat_lock);
1934 		return;
1935 	}
1936 	MUTEX_EXIT(&nat->nat_lock);
1937 
1938 	/*
1939 	 * At this point, nat_ref is 1, doing "--" would make it 0..
1940 	 */
1941 	nat->nat_ref = 0;
1942 
1943 #ifdef	IPFILTER_SYNC
1944 	if (nat->nat_sync)
1945 		ipfsync_del(nat->nat_sync);
1946 #endif
1947 
1948 	if (nat->nat_fr != NULL)
1949 		(void)fr_derefrule(&nat->nat_fr, ifs);
1950 
1951 	if (nat->nat_hm != NULL)
1952 		fr_hostmapdel(&nat->nat_hm);
1953 
1954 	/*
1955 	 * If there is an active reference from the nat entry to its parent
1956 	 * rule, decrement the rule's reference count and free it too if no
1957 	 * longer being used.
1958 	 */
1959 	ipn = nat->nat_ptr;
1960 	if (ipn != NULL) {
1961 		ipn->in_space++;
1962 		ipn->in_use--;
1963 		if (ipn->in_use == 0 && (ipn->in_flags & IPN_DELETE)) {
1964 			if (ipn->in_apr)
1965 				appr_free(ipn->in_apr);
1966 			KFREE(ipn);
1967 			ifs->ifs_nat_stats.ns_rules--;
1968 		}
1969 	}
1970 
1971 	MUTEX_DESTROY(&nat->nat_lock);
1972 
1973 	aps_free(nat->nat_aps, ifs);
1974 	ifs->ifs_nat_stats.ns_inuse--;
1975 
1976 	/*
1977 	 * If there's a fragment table entry too for this nat entry, then
1978 	 * dereference that as well.  This is after nat_lock is released
1979 	 * because of Tru64.
1980 	 */
1981 	fr_forgetnat((void *)nat, ifs);
1982 
1983 	KFREE(nat);
1984 }
1985 
1986 
1987 /* ------------------------------------------------------------------------ */
1988 /* Function:    nat_flushtable                                              */
1989 /* Returns:     int - number of NAT rules deleted                           */
1990 /* Parameters:  Nil                                                         */
1991 /*                                                                          */
1992 /* Deletes all currently active NAT sessions.  In deleting each NAT entry a */
1993 /* log record should be emitted in nat_delete() if NAT logging is enabled.  */
1994 /* ------------------------------------------------------------------------ */
1995 /*
1996  * nat_flushtable - clear the NAT table of all mapping entries.
1997  */
1998 static int nat_flushtable(ifs)
1999 ipf_stack_t *ifs;
2000 {
2001 	nat_t *nat;
2002 	int j = 0;
2003 
2004 	/*
2005 	 * ALL NAT mappings deleted, so lets just make the deletions
2006 	 * quicker.
2007 	 */
2008 	if (ifs->ifs_nat_table[0] != NULL)
2009 		bzero((char *)ifs->ifs_nat_table[0],
2010 		      sizeof(ifs->ifs_nat_table[0]) * ifs->ifs_ipf_nattable_sz);
2011 	if (ifs->ifs_nat_table[1] != NULL)
2012 		bzero((char *)ifs->ifs_nat_table[1],
2013 		      sizeof(ifs->ifs_nat_table[1]) * ifs->ifs_ipf_nattable_sz);
2014 
2015 	while ((nat = ifs->ifs_nat_instances) != NULL) {
2016 		nat_delete(nat, NL_FLUSH, ifs);
2017 		j++;
2018 	}
2019 
2020 	return j;
2021 }
2022 
2023 
2024 /* ------------------------------------------------------------------------ */
2025 /* Function:    nat_clearlist                                               */
2026 /* Returns:     int - number of NAT/RDR rules deleted                       */
2027 /* Parameters:  Nil                                                         */
2028 /*                                                                          */
2029 /* Delete all rules in the current list of rules.  There is nothing elegant */
2030 /* about this cleanup: simply free all entries on the list of rules and     */
2031 /* clear out the tables used for hashed NAT rule lookups.                   */
2032 /* ------------------------------------------------------------------------ */
2033 static int nat_clearlist(ifs)
2034 ipf_stack_t *ifs;
2035 {
2036 	ipnat_t *n, **np = &ifs->ifs_nat_list;
2037 	int i = 0;
2038 
2039 	if (ifs->ifs_nat_rules != NULL)
2040 		bzero((char *)ifs->ifs_nat_rules,
2041 		      sizeof(*ifs->ifs_nat_rules) * ifs->ifs_ipf_natrules_sz);
2042 	if (ifs->ifs_rdr_rules != NULL)
2043 		bzero((char *)ifs->ifs_rdr_rules,
2044 		      sizeof(*ifs->ifs_rdr_rules) * ifs->ifs_ipf_rdrrules_sz);
2045 
2046 	while ((n = *np) != NULL) {
2047 		*np = n->in_next;
2048 		if (n->in_use == 0) {
2049 			if (n->in_apr != NULL)
2050 				appr_free(n->in_apr);
2051 			KFREE(n);
2052 			ifs->ifs_nat_stats.ns_rules--;
2053 		} else {
2054 			n->in_flags |= IPN_DELETE;
2055 			n->in_next = NULL;
2056 		}
2057 		i++;
2058 	}
2059 	ifs->ifs_nat_masks = 0;
2060 	ifs->ifs_rdr_masks = 0;
2061 	for (i = 0; i < 4; i++) {
2062 		ifs->ifs_nat6_masks[i] = 0;
2063 		ifs->ifs_rdr6_masks[i] = 0;
2064 	}
2065 	return i;
2066 }
2067 
2068 
2069 /* ------------------------------------------------------------------------ */
2070 /* Function:    nat_newmap                                                  */
2071 /* Returns:     int - -1 == error, 0 == success                             */
2072 /* Parameters:  fin(I) - pointer to packet information                      */
2073 /*              nat(I) - pointer to NAT entry                               */
2074 /*              ni(I)  - pointer to structure with misc. information needed */
2075 /*                       to create new NAT entry.                           */
2076 /*                                                                          */
2077 /* Given an empty NAT structure, populate it with new information about a   */
2078 /* new NAT session, as defined by the matching NAT rule.                    */
2079 /* ni.nai_ip is passed in uninitialised and must be set, in host byte order,*/
2080 /* to the new IP address for the translation.                               */
2081 /* ------------------------------------------------------------------------ */
2082 static INLINE int nat_newmap(fin, nat, ni)
2083 fr_info_t *fin;
2084 nat_t *nat;
2085 natinfo_t *ni;
2086 {
2087 	u_short st_port, dport, sport, port, sp, dp;
2088 	struct in_addr in, inb;
2089 	hostmap_t *hm;
2090 	u_32_t flags;
2091 	u_32_t st_ip;
2092 	ipnat_t *np;
2093 	nat_t *natl;
2094 	int l;
2095 	ipf_stack_t *ifs = fin->fin_ifs;
2096 
2097 	/*
2098 	 * If it's an outbound packet which doesn't match any existing
2099 	 * record, then create a new port
2100 	 */
2101 	l = 0;
2102 	hm = NULL;
2103 	np = ni->nai_np;
2104 	st_ip = np->in_nip;
2105 	st_port = np->in_pnext;
2106 	flags = ni->nai_flags;
2107 	sport = ni->nai_sport;
2108 	dport = ni->nai_dport;
2109 
2110 	/*
2111 	 * Do a loop until we either run out of entries to try or we find
2112 	 * a NAT mapping that isn't currently being used.  This is done
2113 	 * because the change to the source is not (usually) being fixed.
2114 	 */
2115 	do {
2116 		port = 0;
2117 		in.s_addr = htonl(np->in_nip);
2118 		if (l == 0) {
2119 			/*
2120 			 * Check to see if there is an existing NAT
2121 			 * setup for this IP address pair.
2122 			 */
2123 			hm = nat_hostmap(np, fin->fin_src, fin->fin_dst,
2124 					 in, 0, ifs);
2125 			if (hm != NULL)
2126 				in.s_addr = hm->hm_mapip.s_addr;
2127 		} else if ((l == 1) && (hm != NULL)) {
2128 			fr_hostmapdel(&hm);
2129 		}
2130 		in.s_addr = ntohl(in.s_addr);
2131 
2132 		nat->nat_hm = hm;
2133 
2134 		if ((np->in_outmsk == 0xffffffff) && (np->in_pnext == 0)) {
2135 			if (l > 0)
2136 				return -1;
2137 		}
2138 
2139 		if (np->in_redir == NAT_BIMAP &&
2140 		    np->in_inmsk == np->in_outmsk) {
2141 			/*
2142 			 * map the address block in a 1:1 fashion
2143 			 */
2144 			in.s_addr = np->in_outip;
2145 			in.s_addr |= fin->fin_saddr & ~np->in_inmsk;
2146 			in.s_addr = ntohl(in.s_addr);
2147 
2148 		} else if (np->in_redir & NAT_MAPBLK) {
2149 			if ((l >= np->in_ppip) || ((l > 0) &&
2150 			     !(flags & IPN_TCPUDP)))
2151 				return -1;
2152 			/*
2153 			 * map-block - Calculate destination address.
2154 			 */
2155 			in.s_addr = ntohl(fin->fin_saddr);
2156 			in.s_addr &= ntohl(~np->in_inmsk);
2157 			inb.s_addr = in.s_addr;
2158 			in.s_addr /= np->in_ippip;
2159 			in.s_addr &= ntohl(~np->in_outmsk);
2160 			in.s_addr += ntohl(np->in_outip);
2161 			/*
2162 			 * Calculate destination port.
2163 			 */
2164 			if ((flags & IPN_TCPUDP) &&
2165 			    (np->in_ppip != 0)) {
2166 				port = ntohs(sport) + l;
2167 				port %= np->in_ppip;
2168 				port += np->in_ppip *
2169 					(inb.s_addr % np->in_ippip);
2170 				port += MAPBLK_MINPORT;
2171 				port = htons(port);
2172 			}
2173 
2174 		} else if ((np->in_outip == 0) &&
2175 			   (np->in_outmsk == 0xffffffff)) {
2176 			/*
2177 			 * 0/32 - use the interface's IP address.
2178 			 */
2179 			if ((l > 0) ||
2180 			    fr_ifpaddr(4, FRI_NORMAL, fin->fin_ifp,
2181 				       &in, NULL, fin->fin_ifs) == -1)
2182 				return -1;
2183 			in.s_addr = ntohl(in.s_addr);
2184 
2185 		} else if ((np->in_outip == 0) && (np->in_outmsk == 0)) {
2186 			/*
2187 			 * 0/0 - use the original source address/port.
2188 			 */
2189 			if (l > 0)
2190 				return -1;
2191 			in.s_addr = ntohl(fin->fin_saddr);
2192 
2193 		} else if ((np->in_outmsk != 0xffffffff) &&
2194 			   (np->in_pnext == 0) && ((l > 0) || (hm == NULL)))
2195 			np->in_nip++;
2196 
2197 		natl = NULL;
2198 
2199 		if ((flags & IPN_TCPUDP) &&
2200 		    ((np->in_redir & NAT_MAPBLK) == 0) &&
2201 		    (np->in_flags & IPN_AUTOPORTMAP)) {
2202 			/*
2203 			 * "ports auto" (without map-block)
2204 			 */
2205 			if ((l > 0) && (l % np->in_ppip == 0)) {
2206 				if (l > np->in_space) {
2207 					return -1;
2208 				} else if ((l > np->in_ppip) &&
2209 					   np->in_outmsk != 0xffffffff)
2210 					np->in_nip++;
2211 			}
2212 			if (np->in_ppip != 0) {
2213 				port = ntohs(sport);
2214 				port += (l % np->in_ppip);
2215 				port %= np->in_ppip;
2216 				port += np->in_ppip *
2217 					(ntohl(fin->fin_saddr) %
2218 					 np->in_ippip);
2219 				port += MAPBLK_MINPORT;
2220 				port = htons(port);
2221 			}
2222 
2223 		} else if (((np->in_redir & NAT_MAPBLK) == 0) &&
2224 			   (flags & IPN_TCPUDPICMP) && (np->in_pnext != 0)) {
2225 			/*
2226 			 * Standard port translation.  Select next port.
2227 			 */
2228 			port = htons(np->in_pnext++);
2229 
2230 			if (np->in_pnext > ntohs(np->in_pmax)) {
2231 				np->in_pnext = ntohs(np->in_pmin);
2232 				if (np->in_outmsk != 0xffffffff)
2233 					np->in_nip++;
2234 			}
2235 		}
2236 
2237 		if (np->in_flags & IPN_IPRANGE) {
2238 			if (np->in_nip > ntohl(np->in_outmsk))
2239 				np->in_nip = ntohl(np->in_outip);
2240 		} else {
2241 			if ((np->in_outmsk != 0xffffffff) &&
2242 			    ((np->in_nip + 1) & ntohl(np->in_outmsk)) >
2243 			    ntohl(np->in_outip))
2244 				np->in_nip = ntohl(np->in_outip) + 1;
2245 		}
2246 
2247 		if ((port == 0) && (flags & (IPN_TCPUDPICMP|IPN_ICMPQUERY)))
2248 			port = sport;
2249 
2250 		/*
2251 		 * Here we do a lookup of the connection as seen from
2252 		 * the outside.  If an IP# pair already exists, try
2253 		 * again.  So if you have A->B becomes C->B, you can
2254 		 * also have D->E become C->E but not D->B causing
2255 		 * another C->B.  Also take protocol and ports into
2256 		 * account when determining whether a pre-existing
2257 		 * NAT setup will cause an external conflict where
2258 		 * this is appropriate.
2259 		 */
2260 		inb.s_addr = htonl(in.s_addr);
2261 		sp = fin->fin_data[0];
2262 		dp = fin->fin_data[1];
2263 		fin->fin_data[0] = fin->fin_data[1];
2264 		fin->fin_data[1] = htons(port);
2265 		natl = nat_inlookup(fin, flags & ~(SI_WILDP|NAT_SEARCH),
2266 				    (u_int)fin->fin_p, fin->fin_dst, inb);
2267 		fin->fin_data[0] = sp;
2268 		fin->fin_data[1] = dp;
2269 
2270 		/*
2271 		 * Has the search wrapped around and come back to the
2272 		 * start ?
2273 		 */
2274 		if ((natl != NULL) &&
2275 		    (np->in_pnext != 0) && (st_port == np->in_pnext) &&
2276 		    (np->in_nip != 0) && (st_ip == np->in_nip))
2277 			return -1;
2278 		l++;
2279 	} while (natl != NULL);
2280 
2281 	if (np->in_space > 0)
2282 		np->in_space--;
2283 
2284 	/* Setup the NAT table */
2285 	nat->nat_inip = fin->fin_src;
2286 	nat->nat_outip.s_addr = htonl(in.s_addr);
2287 	nat->nat_oip = fin->fin_dst;
2288 	if (nat->nat_hm == NULL)
2289 		nat->nat_hm = nat_hostmap(np, fin->fin_src, fin->fin_dst,
2290 					  nat->nat_outip, 0, ifs);
2291 
2292 	if (flags & IPN_TCPUDP) {
2293 		nat->nat_inport = sport;
2294 		nat->nat_outport = port;	/* sport */
2295 		nat->nat_oport = dport;
2296 		((tcphdr_t *)fin->fin_dp)->th_sport = port;
2297 	} else if (flags & IPN_ICMPQUERY) {
2298 		((icmphdr_t *)fin->fin_dp)->icmp_id = port;
2299 		nat->nat_inport = port;
2300 		nat->nat_outport = port;
2301 	}
2302 
2303 	ni->nai_ip.s_addr = in.s_addr;
2304 	ni->nai_port = port;
2305 	ni->nai_nport = dport;
2306 	return 0;
2307 }
2308 
2309 
2310 /* ------------------------------------------------------------------------ */
2311 /* Function:    nat_newrdr                                                  */
2312 /* Returns:     int - -1 == error, 0 == success (no move), 1 == success and */
2313 /*                    allow rule to be moved if IPN_ROUNDR is set.          */
2314 /* Parameters:  fin(I) - pointer to packet information                      */
2315 /*              nat(I) - pointer to NAT entry                               */
2316 /*              ni(I)  - pointer to structure with misc. information needed */
2317 /*                       to create new NAT entry.                           */
2318 /*                                                                          */
2319 /* ni.nai_ip is passed in uninitialised and must be set, in host byte order,*/
2320 /* to the new IP address for the translation.                               */
2321 /* ------------------------------------------------------------------------ */
2322 static INLINE int nat_newrdr(fin, nat, ni)
2323 fr_info_t *fin;
2324 nat_t *nat;
2325 natinfo_t *ni;
2326 {
2327 	u_short nport, dport, sport;
2328 	struct in_addr in, inb;
2329 	u_short sp, dp;
2330 	hostmap_t *hm;
2331 	u_32_t flags;
2332 	ipnat_t *np;
2333 	nat_t *natl;
2334 	int move;
2335 	ipf_stack_t *ifs = fin->fin_ifs;
2336 
2337 	move = 1;
2338 	hm = NULL;
2339 	in.s_addr = 0;
2340 	np = ni->nai_np;
2341 	flags = ni->nai_flags;
2342 	sport = ni->nai_sport;
2343 	dport = ni->nai_dport;
2344 
2345 	/*
2346 	 * If the matching rule has IPN_STICKY set, then we want to have the
2347 	 * same rule kick in as before.  Why would this happen?  If you have
2348 	 * a collection of rdr rules with "round-robin sticky", the current
2349 	 * packet might match a different one to the previous connection but
2350 	 * we want the same destination to be used.
2351 	 */
2352 	if ((np->in_flags & (IPN_ROUNDR|IPN_STICKY)) ==
2353 	    (IPN_ROUNDR|IPN_STICKY)) {
2354 		hm = nat_hostmap(NULL, fin->fin_src, fin->fin_dst, in,
2355 				 (u_32_t)dport, ifs);
2356 		if (hm != NULL) {
2357 			in.s_addr = ntohl(hm->hm_mapip.s_addr);
2358 			np = hm->hm_ipnat;
2359 			ni->nai_np = np;
2360 			move = 0;
2361 		}
2362 	}
2363 
2364 	/*
2365 	 * Otherwise, it's an inbound packet. Most likely, we don't
2366 	 * want to rewrite source ports and source addresses. Instead,
2367 	 * we want to rewrite to a fixed internal address and fixed
2368 	 * internal port.
2369 	 */
2370 	if (np->in_flags & IPN_SPLIT) {
2371 		in.s_addr = np->in_nip;
2372 
2373 		if ((np->in_flags & (IPN_ROUNDR|IPN_STICKY)) == IPN_STICKY) {
2374 			hm = nat_hostmap(np, fin->fin_src, fin->fin_dst,
2375 					 in, (u_32_t)dport, ifs);
2376 			if (hm != NULL) {
2377 				in.s_addr = hm->hm_mapip.s_addr;
2378 				move = 0;
2379 			}
2380 		}
2381 
2382 		if (hm == NULL || hm->hm_ref == 1) {
2383 			if (np->in_inip == htonl(in.s_addr)) {
2384 				np->in_nip = ntohl(np->in_inmsk);
2385 				move = 0;
2386 			} else {
2387 				np->in_nip = ntohl(np->in_inip);
2388 			}
2389 		}
2390 
2391 	} else if ((np->in_inip == 0) && (np->in_inmsk == 0xffffffff)) {
2392 		/*
2393 		 * 0/32 - use the interface's IP address.
2394 		 */
2395 		if (fr_ifpaddr(4, FRI_NORMAL, fin->fin_ifp, &in, NULL,
2396 			   fin->fin_ifs) == -1)
2397 			return -1;
2398 		in.s_addr = ntohl(in.s_addr);
2399 
2400 	} else if ((np->in_inip == 0) && (np->in_inmsk== 0)) {
2401 		/*
2402 		 * 0/0 - use the original destination address/port.
2403 		 */
2404 		in.s_addr = ntohl(fin->fin_daddr);
2405 
2406 	} else if (np->in_redir == NAT_BIMAP &&
2407 		   np->in_inmsk == np->in_outmsk) {
2408 		/*
2409 		 * map the address block in a 1:1 fashion
2410 		 */
2411 		in.s_addr = np->in_inip;
2412 		in.s_addr |= fin->fin_daddr & ~np->in_inmsk;
2413 		in.s_addr = ntohl(in.s_addr);
2414 	} else {
2415 		in.s_addr = ntohl(np->in_inip);
2416 	}
2417 
2418 	if ((np->in_pnext == 0) || ((flags & NAT_NOTRULEPORT) != 0))
2419 		nport = dport;
2420 	else {
2421 		/*
2422 		 * Whilst not optimized for the case where
2423 		 * pmin == pmax, the gain is not significant.
2424 		 */
2425 		if (((np->in_flags & IPN_FIXEDDPORT) == 0) &&
2426 		    (np->in_pmin != np->in_pmax)) {
2427 			nport = ntohs(dport) - ntohs(np->in_pmin) +
2428 				ntohs(np->in_pnext);
2429 			nport = htons(nport);
2430 		} else
2431 			nport = np->in_pnext;
2432 	}
2433 
2434 	/*
2435 	 * When the redirect-to address is set to 0.0.0.0, just
2436 	 * assume a blank `forwarding' of the packet.  We don't
2437 	 * setup any translation for this either.
2438 	 */
2439 	if (in.s_addr == 0) {
2440 		if (nport == dport)
2441 			return -1;
2442 		in.s_addr = ntohl(fin->fin_daddr);
2443 	}
2444 
2445 	/*
2446 	 * Check to see if this redirect mapping already exists and if
2447 	 * it does, return "failure" (allowing it to be created will just
2448 	 * cause one or both of these "connections" to stop working.)
2449 	 */
2450 	inb.s_addr = htonl(in.s_addr);
2451 	sp = fin->fin_data[0];
2452 	dp = fin->fin_data[1];
2453 	fin->fin_data[1] = fin->fin_data[0];
2454 	fin->fin_data[0] = ntohs(nport);
2455 	natl = nat_outlookup(fin, flags & ~(SI_WILDP|NAT_SEARCH),
2456 		    (u_int)fin->fin_p, inb, fin->fin_src);
2457 	fin->fin_data[0] = sp;
2458 	fin->fin_data[1] = dp;
2459 	if (natl != NULL)
2460 		return (-1);
2461 
2462 	nat->nat_inip.s_addr = htonl(in.s_addr);
2463 	nat->nat_outip = fin->fin_dst;
2464 	nat->nat_oip = fin->fin_src;
2465 
2466 	ni->nai_ip.s_addr = in.s_addr;
2467 	ni->nai_nport = nport;
2468 	ni->nai_port = sport;
2469 
2470 	if (flags & IPN_TCPUDP) {
2471 		nat->nat_inport = nport;
2472 		nat->nat_outport = dport;
2473 		nat->nat_oport = sport;
2474 		((tcphdr_t *)fin->fin_dp)->th_dport = nport;
2475 	} else if (flags & IPN_ICMPQUERY) {
2476 		((icmphdr_t *)fin->fin_dp)->icmp_id = nport;
2477 		nat->nat_inport = nport;
2478 		nat->nat_outport = nport;
2479 	}
2480 
2481 	return move;
2482 }
2483 
2484 /* ------------------------------------------------------------------------ */
2485 /* Function:    nat_new                                                     */
2486 /* Returns:     nat_t* - NULL == failure to create new NAT structure,       */
2487 /*                       else pointer to new NAT structure                  */
2488 /* Parameters:  fin(I)       - pointer to packet information                */
2489 /*              np(I)        - pointer to NAT rule                          */
2490 /*              natsave(I)   - pointer to where to store NAT struct pointer */
2491 /*              flags(I)     - flags describing the current packet          */
2492 /*              direction(I) - direction of packet (in/out)                 */
2493 /* Write Lock:  ipf_nat                                                     */
2494 /*                                                                          */
2495 /* Attempts to create a new NAT entry.  Does not actually change the packet */
2496 /* in any way.                                                              */
2497 /*                                                                          */
2498 /* This fucntion is in three main parts: (1) deal with creating a new NAT   */
2499 /* structure for a "MAP" rule (outgoing NAT translation); (2) deal with     */
2500 /* creating a new NAT structure for a "RDR" rule (incoming NAT translation) */
2501 /* and (3) building that structure and putting it into the NAT table(s).    */
2502 /* ------------------------------------------------------------------------ */
2503 nat_t *nat_new(fin, np, natsave, flags, direction)
2504 fr_info_t *fin;
2505 ipnat_t *np;
2506 nat_t **natsave;
2507 u_int flags;
2508 int direction;
2509 {
2510 	tcphdr_t *tcp = NULL;
2511 	hostmap_t *hm = NULL;
2512 	nat_t *nat, *natl;
2513 	u_int nflags;
2514 	natinfo_t ni;
2515 	int move;
2516 	ipf_stack_t *ifs = fin->fin_ifs;
2517 
2518 	/*
2519 	 * Trigger automatic call to nat_extraflush() if the
2520 	 * table has reached capcity specified by hi watermark.
2521 	 */
2522 	if (NAT_TAB_WATER_LEVEL(ifs) > ifs->ifs_nat_flush_lvl_hi)
2523 		ifs->ifs_nat_doflush = 1;
2524 
2525 	if (ifs->ifs_nat_stats.ns_inuse >= ifs->ifs_ipf_nattable_max) {
2526 		ifs->ifs_nat_stats.ns_memfail++;
2527 		return NULL;
2528 	}
2529 
2530 	move = 1;
2531 	nflags = np->in_flags & flags;
2532 	nflags &= NAT_FROMRULE;
2533 
2534 	ni.nai_np = np;
2535 	ni.nai_nflags = nflags;
2536 	ni.nai_flags = flags;
2537 
2538 	/* Give me a new nat */
2539 	KMALLOC(nat, nat_t *);
2540 	if (nat == NULL) {
2541 		ifs->ifs_nat_stats.ns_memfail++;
2542 		/*
2543 		 * Try to automatically tune the max # of entries in the
2544 		 * table allowed to be less than what will cause kmem_alloc()
2545 		 * to fail and try to eliminate panics due to out of memory
2546 		 * conditions arising.
2547 		 */
2548 		if (ifs->ifs_ipf_nattable_max > ifs->ifs_ipf_nattable_sz) {
2549 			ifs->ifs_ipf_nattable_max = ifs->ifs_nat_stats.ns_inuse - 100;
2550 			printf("ipf_nattable_max reduced to %d\n",
2551 				ifs->ifs_ipf_nattable_max);
2552 		}
2553 		return NULL;
2554 	}
2555 
2556 	if (flags & IPN_TCPUDP) {
2557 		tcp = fin->fin_dp;
2558 		ni.nai_sport = htons(fin->fin_sport);
2559 		ni.nai_dport = htons(fin->fin_dport);
2560 	} else if (flags & IPN_ICMPQUERY) {
2561 		/*
2562 		 * In the ICMP query NAT code, we translate the ICMP id fields
2563 		 * to make them unique. This is indepedent of the ICMP type
2564 		 * (e.g. in the unlikely event that a host sends an echo and
2565 		 * an tstamp request with the same id, both packets will have
2566 		 * their ip address/id field changed in the same way).
2567 		 */
2568 		/* The icmp_id field is used by the sender to identify the
2569 		 * process making the icmp request. (the receiver justs
2570 		 * copies it back in its response). So, it closely matches
2571 		 * the concept of source port. We overlay sport, so we can
2572 		 * maximally reuse the existing code.
2573 		 */
2574 		ni.nai_sport = ((icmphdr_t *)fin->fin_dp)->icmp_id;
2575 		ni.nai_dport = ni.nai_sport;
2576 	}
2577 
2578 	bzero((char *)nat, sizeof(*nat));
2579 	nat->nat_flags = flags;
2580 	nat->nat_redir = np->in_redir;
2581 
2582 	if ((flags & NAT_SLAVE) == 0) {
2583 		MUTEX_ENTER(&ifs->ifs_ipf_nat_new);
2584 	}
2585 
2586 	/*
2587 	 * Search the current table for a match.
2588 	 */
2589 	if (direction == NAT_OUTBOUND) {
2590 		/*
2591 		 * We can now arrange to call this for the same connection
2592 		 * because ipf_nat_new doesn't protect the code path into
2593 		 * this function.
2594 		 */
2595 		natl = nat_outlookup(fin, nflags, (u_int)fin->fin_p,
2596 				     fin->fin_src, fin->fin_dst);
2597 		if (natl != NULL) {
2598 			KFREE(nat);
2599 			nat = natl;
2600 			goto done;
2601 		}
2602 
2603 		move = nat_newmap(fin, nat, &ni);
2604 		if (move == -1)
2605 			goto badnat;
2606 
2607 		np = ni.nai_np;
2608 	} else {
2609 		/*
2610 		 * NAT_INBOUND is used only for redirects rules
2611 		 */
2612 		natl = nat_inlookup(fin, nflags, (u_int)fin->fin_p,
2613 				    fin->fin_src, fin->fin_dst);
2614 		if (natl != NULL) {
2615 			KFREE(nat);
2616 			nat = natl;
2617 			goto done;
2618 		}
2619 
2620 		move = nat_newrdr(fin, nat, &ni);
2621 		if (move == -1)
2622 			goto badnat;
2623 
2624 		np = ni.nai_np;
2625 	}
2626 
2627 	if ((move == 1) && (np->in_flags & IPN_ROUNDR)) {
2628 		if (np->in_redir == NAT_REDIRECT) {
2629 			nat_delrdr(np);
2630 			nat_addrdr(np, ifs);
2631 		} else if (np->in_redir == NAT_MAP) {
2632 			nat_delnat(np);
2633 			nat_addnat(np, ifs);
2634 		}
2635 	}
2636 
2637 	if (nat_finalise(fin, nat, &ni, tcp, natsave, direction) == -1) {
2638 		goto badnat;
2639 	}
2640 
2641 	nat_calc_chksum_diffs(nat);
2642 
2643 	if (flags & SI_WILDP)
2644 		ifs->ifs_nat_stats.ns_wilds++;
2645 	goto done;
2646 badnat:
2647 	ifs->ifs_nat_stats.ns_badnat++;
2648 	if ((hm = nat->nat_hm) != NULL)
2649 		fr_hostmapdel(&hm);
2650 	KFREE(nat);
2651 	nat = NULL;
2652 done:
2653 	if ((flags & NAT_SLAVE) == 0) {
2654 		MUTEX_EXIT(&ifs->ifs_ipf_nat_new);
2655 	}
2656 	return nat;
2657 }
2658 
2659 
2660 /* ------------------------------------------------------------------------ */
2661 /* Function:    nat_finalise                                                */
2662 /* Returns:     int - 0 == sucess, -1 == failure                            */
2663 /* Parameters:  fin(I) - pointer to packet information                      */
2664 /*              nat(I) - pointer to NAT entry                               */
2665 /*              ni(I)  - pointer to structure with misc. information needed */
2666 /*                       to create new NAT entry.                           */
2667 /* Write Lock:  ipf_nat                                                     */
2668 /*                                                                          */
2669 /* This is the tail end of constructing a new NAT entry and is the same     */
2670 /* for both IPv4 and IPv6.                                                  */
2671 /* ------------------------------------------------------------------------ */
2672 /*ARGSUSED*/
2673 static INLINE int nat_finalise(fin, nat, ni, tcp, natsave, direction)
2674 fr_info_t *fin;
2675 nat_t *nat;
2676 natinfo_t *ni;
2677 tcphdr_t *tcp;
2678 nat_t **natsave;
2679 int direction;
2680 {
2681 	frentry_t *fr;
2682 	ipnat_t *np;
2683 	ipf_stack_t *ifs = fin->fin_ifs;
2684 
2685 	np = ni->nai_np;
2686 
2687 	COPYIFNAME(fin->fin_ifp, nat->nat_ifnames[0], fin->fin_v);
2688 
2689 #ifdef	IPFILTER_SYNC
2690 	if ((nat->nat_flags & SI_CLONE) == 0)
2691 		nat->nat_sync = ipfsync_new(SMC_NAT, fin, nat);
2692 #endif
2693 
2694 	nat->nat_me = natsave;
2695 	nat->nat_dir = direction;
2696 	nat->nat_ifps[0] = np->in_ifps[0];
2697 	nat->nat_ifps[1] = np->in_ifps[1];
2698 	nat->nat_ptr = np;
2699 	nat->nat_p = fin->fin_p;
2700 	nat->nat_v = fin->fin_v;
2701 	nat->nat_mssclamp = np->in_mssclamp;
2702 	fr = fin->fin_fr;
2703 	nat->nat_fr = fr;
2704 
2705 	if ((np->in_apr != NULL) && ((ni->nai_flags & NAT_SLAVE) == 0))
2706 		if (appr_new(fin, nat) == -1)
2707 			return -1;
2708 
2709 	if (nat_insert(nat, fin->fin_rev, ifs) == 0) {
2710 		if (ifs->ifs_nat_logging)
2711 			nat_log(nat, (u_int)np->in_redir, ifs);
2712 		np->in_use++;
2713 		if (fr != NULL) {
2714 			MUTEX_ENTER(&fr->fr_lock);
2715 			fr->fr_ref++;
2716 			MUTEX_EXIT(&fr->fr_lock);
2717 		}
2718 		return 0;
2719 	}
2720 
2721 	/*
2722 	 * nat_insert failed, so cleanup time...
2723 	 */
2724 	return -1;
2725 }
2726 
2727 
2728 /* ------------------------------------------------------------------------ */
2729 /* Function:   nat_insert                                                   */
2730 /* Returns:    int - 0 == sucess, -1 == failure                             */
2731 /* Parameters: nat(I) - pointer to NAT structure                            */
2732 /*             rev(I) - flag indicating forward/reverse direction of packet */
2733 /* Write Lock: ipf_nat                                                      */
2734 /*                                                                          */
2735 /* Insert a NAT entry into the hash tables for searching and add it to the  */
2736 /* list of active NAT entries.  Adjust global counters when complete.       */
2737 /* ------------------------------------------------------------------------ */
2738 int	nat_insert(nat, rev, ifs)
2739 nat_t	*nat;
2740 int	rev;
2741 ipf_stack_t *ifs;
2742 {
2743 	u_int hv1, hv2;
2744 	nat_t **natp;
2745 
2746 	/*
2747 	 * Try and return an error as early as possible, so calculate the hash
2748 	 * entry numbers first and then proceed.
2749 	 */
2750 	if ((nat->nat_flags & (SI_W_SPORT|SI_W_DPORT)) == 0) {
2751 		hv1 = NAT_HASH_FN(nat->nat_inip.s_addr, nat->nat_inport,
2752 				  0xffffffff);
2753 		hv1 = NAT_HASH_FN(nat->nat_oip.s_addr, hv1 + nat->nat_oport,
2754 				  ifs->ifs_ipf_nattable_sz);
2755 		hv2 = NAT_HASH_FN(nat->nat_outip.s_addr, nat->nat_outport,
2756 				  0xffffffff);
2757 		hv2 = NAT_HASH_FN(nat->nat_oip.s_addr, hv2 + nat->nat_oport,
2758 				  ifs->ifs_ipf_nattable_sz);
2759 	} else {
2760 		hv1 = NAT_HASH_FN(nat->nat_inip.s_addr, 0, 0xffffffff);
2761 		hv1 = NAT_HASH_FN(nat->nat_oip.s_addr, hv1,
2762 				  ifs->ifs_ipf_nattable_sz);
2763 		hv2 = NAT_HASH_FN(nat->nat_outip.s_addr, 0, 0xffffffff);
2764 		hv2 = NAT_HASH_FN(nat->nat_oip.s_addr, hv2,
2765 				  ifs->ifs_ipf_nattable_sz);
2766 	}
2767 
2768 	if (ifs->ifs_nat_stats.ns_bucketlen[0][hv1] >= ifs->ifs_fr_nat_maxbucket ||
2769 	    ifs->ifs_nat_stats.ns_bucketlen[1][hv2] >= ifs->ifs_fr_nat_maxbucket) {
2770 		return -1;
2771 	}
2772 
2773 	nat->nat_hv[0] = hv1;
2774 	nat->nat_hv[1] = hv2;
2775 
2776 	MUTEX_INIT(&nat->nat_lock, "nat entry lock");
2777 
2778 	nat->nat_rev = rev;
2779 	nat->nat_ref = 1;
2780 	nat->nat_bytes[0] = 0;
2781 	nat->nat_pkts[0] = 0;
2782 	nat->nat_bytes[1] = 0;
2783 	nat->nat_pkts[1] = 0;
2784 
2785 	nat->nat_ifnames[0][LIFNAMSIZ - 1] = '\0';
2786 	nat->nat_ifps[0] = fr_resolvenic(nat->nat_ifnames[0], 4, ifs);
2787 
2788 	if (nat->nat_ifnames[1][0] !='\0') {
2789 		nat->nat_ifnames[1][LIFNAMSIZ - 1] = '\0';
2790 		nat->nat_ifps[1] = fr_resolvenic(nat->nat_ifnames[1], 4, ifs);
2791 	} else {
2792 		(void) strncpy(nat->nat_ifnames[1], nat->nat_ifnames[0],
2793 			       LIFNAMSIZ);
2794 		nat->nat_ifnames[1][LIFNAMSIZ - 1] = '\0';
2795 		nat->nat_ifps[1] = nat->nat_ifps[0];
2796 	}
2797 
2798 	nat->nat_next = ifs->ifs_nat_instances;
2799 	nat->nat_pnext = &ifs->ifs_nat_instances;
2800 	if (ifs->ifs_nat_instances)
2801 		ifs->ifs_nat_instances->nat_pnext = &nat->nat_next;
2802 	ifs->ifs_nat_instances = nat;
2803 
2804 	natp = &ifs->ifs_nat_table[0][hv1];
2805 	if (*natp)
2806 		(*natp)->nat_phnext[0] = &nat->nat_hnext[0];
2807 	nat->nat_phnext[0] = natp;
2808 	nat->nat_hnext[0] = *natp;
2809 	*natp = nat;
2810 	ifs->ifs_nat_stats.ns_bucketlen[0][hv1]++;
2811 
2812 	natp = &ifs->ifs_nat_table[1][hv2];
2813 	if (*natp)
2814 		(*natp)->nat_phnext[1] = &nat->nat_hnext[1];
2815 	nat->nat_phnext[1] = natp;
2816 	nat->nat_hnext[1] = *natp;
2817 	*natp = nat;
2818 	ifs->ifs_nat_stats.ns_bucketlen[1][hv2]++;
2819 
2820 	fr_setnatqueue(nat, rev, ifs);
2821 
2822 	ifs->ifs_nat_stats.ns_added++;
2823 	ifs->ifs_nat_stats.ns_inuse++;
2824 	return 0;
2825 }
2826 
2827 
2828 /* ------------------------------------------------------------------------ */
2829 /* Function:    nat_icmperrorlookup                                         */
2830 /* Returns:     nat_t* - point to matching NAT structure                    */
2831 /* Parameters:  fin(I) - pointer to packet information                      */
2832 /*              dir(I) - direction of packet (in/out)                       */
2833 /*                                                                          */
2834 /* Check if the ICMP error message is related to an existing TCP, UDP or    */
2835 /* ICMP query nat entry.  It is assumed that the packet is already of the   */
2836 /* the required length.                                                     */
2837 /* ------------------------------------------------------------------------ */
2838 nat_t *nat_icmperrorlookup(fin, dir)
2839 fr_info_t *fin;
2840 int dir;
2841 {
2842 	int flags = 0, minlen;
2843 	icmphdr_t *orgicmp;
2844 	tcphdr_t *tcp = NULL;
2845 	u_short data[2];
2846 	nat_t *nat;
2847 	ip_t *oip;
2848 	u_int p;
2849 
2850 	/*
2851 	 * Does it at least have the return (basic) IP header ?
2852 	 * Only a basic IP header (no options) should be with an ICMP error
2853 	 * header.  Also, if it's not an error type, then return.
2854 	 */
2855 	if ((fin->fin_hlen != sizeof(ip_t)) || !(fin->fin_flx & FI_ICMPERR))
2856 		return NULL;
2857 
2858 	/*
2859 	 * Check packet size
2860 	 */
2861 	oip = (ip_t *)((char *)fin->fin_dp + 8);
2862 	minlen = IP_HL(oip) << 2;
2863 	if ((minlen < sizeof(ip_t)) ||
2864 	    (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen))
2865 		return NULL;
2866 	/*
2867 	 * Is the buffer big enough for all of it ?  It's the size of the IP
2868 	 * header claimed in the encapsulated part which is of concern.  It
2869 	 * may be too big to be in this buffer but not so big that it's
2870 	 * outside the ICMP packet, leading to TCP deref's causing problems.
2871 	 * This is possible because we don't know how big oip_hl is when we
2872 	 * do the pullup early in fr_check() and thus can't gaurantee it is
2873 	 * all here now.
2874 	 */
2875 #ifdef  _KERNEL
2876 	{
2877 	mb_t *m;
2878 
2879 	m = fin->fin_m;
2880 # if defined(MENTAT)
2881 	if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN > (char *)m->b_wptr)
2882 		return NULL;
2883 # else
2884 	if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN >
2885 	    (char *)fin->fin_ip + M_LEN(m))
2886 		return NULL;
2887 # endif
2888 	}
2889 #endif
2890 
2891 	if (fin->fin_daddr != oip->ip_src.s_addr)
2892 		return NULL;
2893 
2894 	p = oip->ip_p;
2895 	if (p == IPPROTO_TCP)
2896 		flags = IPN_TCP;
2897 	else if (p == IPPROTO_UDP)
2898 		flags = IPN_UDP;
2899 	else if (p == IPPROTO_ICMP) {
2900 		orgicmp = (icmphdr_t *)((char *)oip + (IP_HL(oip) << 2));
2901 
2902 		/* see if this is related to an ICMP query */
2903 		if (nat_icmpquerytype4(orgicmp->icmp_type)) {
2904 			data[0] = fin->fin_data[0];
2905 			data[1] = fin->fin_data[1];
2906 			fin->fin_data[0] = 0;
2907 			fin->fin_data[1] = orgicmp->icmp_id;
2908 
2909 			flags = IPN_ICMPERR|IPN_ICMPQUERY;
2910 			/*
2911 			 * NOTE : dir refers to the direction of the original
2912 			 *        ip packet. By definition the icmp error
2913 			 *        message flows in the opposite direction.
2914 			 */
2915 			if (dir == NAT_INBOUND)
2916 				nat = nat_inlookup(fin, flags, p, oip->ip_dst,
2917 						   oip->ip_src);
2918 			else
2919 				nat = nat_outlookup(fin, flags, p, oip->ip_dst,
2920 						    oip->ip_src);
2921 			fin->fin_data[0] = data[0];
2922 			fin->fin_data[1] = data[1];
2923 			return nat;
2924 		}
2925 	}
2926 
2927 	if (flags & IPN_TCPUDP) {
2928 		minlen += 8;		/* + 64bits of data to get ports */
2929 		if (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen)
2930 			return NULL;
2931 
2932 		data[0] = fin->fin_data[0];
2933 		data[1] = fin->fin_data[1];
2934 		tcp = (tcphdr_t *)((char *)oip + (IP_HL(oip) << 2));
2935 		fin->fin_data[0] = ntohs(tcp->th_dport);
2936 		fin->fin_data[1] = ntohs(tcp->th_sport);
2937 
2938 		if (dir == NAT_INBOUND) {
2939 			nat = nat_inlookup(fin, flags, p, oip->ip_dst,
2940 					   oip->ip_src);
2941 		} else {
2942 			nat = nat_outlookup(fin, flags, p, oip->ip_dst,
2943 					    oip->ip_src);
2944 		}
2945 		fin->fin_data[0] = data[0];
2946 		fin->fin_data[1] = data[1];
2947 		return nat;
2948 	}
2949 	if (dir == NAT_INBOUND)
2950 		return nat_inlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
2951 	else
2952 		return nat_outlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
2953 }
2954 
2955 
2956 /* ------------------------------------------------------------------------ */
2957 /* Function:    nat_icmperror                                               */
2958 /* Returns:     nat_t* - point to matching NAT structure                    */
2959 /* Parameters:  fin(I)    - pointer to packet information                   */
2960 /*              nflags(I) - NAT flags for this packet                       */
2961 /*              dir(I)    - direction of packet (in/out)                    */
2962 /*                                                                          */
2963 /* Fix up an ICMP packet which is an error message for an existing NAT      */
2964 /* session.  This will correct both packet header data and checksums.       */
2965 /*                                                                          */
2966 /* This should *ONLY* be used for incoming ICMP error packets to make sure  */
2967 /* a NAT'd ICMP packet gets correctly recognised.                           */
2968 /* ------------------------------------------------------------------------ */
2969 nat_t *nat_icmperror(fin, nflags, dir)
2970 fr_info_t *fin;
2971 u_int *nflags;
2972 int dir;
2973 {
2974 	u_32_t sum1, sum2, sumd, psum1, psum2, psumd, sumd2;
2975 	struct in_addr in;
2976 	icmphdr_t *icmp, *orgicmp;
2977 	int dlen;
2978 	udphdr_t *udp;
2979 	tcphdr_t *tcp;
2980 	nat_t *nat;
2981 	ip_t *oip;
2982 	if ((fin->fin_flx & (FI_SHORT|FI_FRAGBODY)))
2983 		return NULL;
2984 
2985 	/*
2986 	 * nat_icmperrorlookup() looks up nat entry associated with the
2987 	 * offending IP packet and returns pointer to the entry, or NULL
2988 	 * if packet wasn't natted or for `defective' packets.
2989 	 */
2990 
2991 	if ((fin->fin_v != 4) || !(nat = nat_icmperrorlookup(fin, dir)))
2992 		return NULL;
2993 
2994 	sumd2 = 0;
2995 	*nflags = IPN_ICMPERR;
2996 	icmp = fin->fin_dp;
2997 	oip = (ip_t *)&icmp->icmp_ip;
2998 	udp = (udphdr_t *)((((char *)oip) + (IP_HL(oip) << 2)));
2999 	tcp = (tcphdr_t *)udp;
3000 	dlen = fin->fin_plen - ((char *)udp - (char *)fin->fin_ip);
3001 
3002 	/*
3003 	 * Need to adjust ICMP header to include the real IP#'s and
3004 	 * port #'s.  There are three steps required.
3005 	 *
3006 	 * Step 1
3007 	 * Fix the IP addresses in the offending IP packet and update
3008 	 * ip header checksum to compensate for the change.
3009 	 *
3010 	 * No update needed here for icmp_cksum because the ICMP checksum
3011 	 * is calculated over the complete ICMP packet, which includes the
3012 	 * changed oip IP addresses and oip->ip_sum.  These two changes
3013 	 * cancel each other out (if the delta for the IP address is x,
3014 	 * then the delta for ip_sum is minus x).
3015 	 */
3016 
3017 	if (oip->ip_dst.s_addr == nat->nat_oip.s_addr) {
3018 		sum1 = LONG_SUM(ntohl(oip->ip_src.s_addr));
3019 		in = nat->nat_inip;
3020 		oip->ip_src = in;
3021 	} else {
3022 		sum1 = LONG_SUM(ntohl(oip->ip_dst.s_addr));
3023 		in = nat->nat_outip;
3024 		oip->ip_dst = in;
3025 	}
3026 
3027 	sum2 = LONG_SUM(ntohl(in.s_addr));
3028 	CALC_SUMD(sum1, sum2, sumd);
3029 	fix_datacksum(&oip->ip_sum, sumd);
3030 
3031 	/*
3032 	 * Step 2
3033 	 * Perform other adjustments based on protocol of offending packet.
3034 	 */
3035 
3036 	switch (oip->ip_p) {
3037 		case IPPROTO_TCP :
3038 		case IPPROTO_UDP :
3039 
3040 			/*
3041 			* For offending TCP/UDP IP packets, translate the ports
3042 			* based on the NAT specification.
3043 			*
3044 			* Advance notice : Now it becomes complicated :-)
3045 			*
3046 			* Since the port and IP addresse fields are both part
3047 			* of the TCP/UDP checksum of the offending IP packet,
3048 			* we need to adjust that checksum as well.
3049 			*
3050 			* To further complicate things, the TCP/UDP checksum
3051 			* may not be present.  We must check to see if the
3052 			* length of the data portion is big enough to hold
3053 			* the checksum.  In the UDP case, a test to determine
3054 			* if the checksum is even set is also required.
3055 			*
3056 			* Any changes to an IP address, port or checksum within
3057 			* the ICMP packet requires a change to icmp_cksum.
3058 			*
3059 			* Be extremely careful here ... The change is dependent
3060 			* upon whether or not the TCP/UPD checksum is present.
3061 			*
3062 			* If TCP/UPD checksum is present, the icmp_cksum must
3063 			* compensate for checksum modification resulting from
3064 			* IP address change only.  Port change and resulting
3065 			* data checksum adjustments cancel each other out.
3066 			*
3067 			* If TCP/UDP checksum is not present, icmp_cksum must
3068 			* compensate for port change only.  The IP address
3069 			* change does not modify anything else in this case.
3070 			*/
3071 
3072 			psum1 = 0;
3073 			psum2 = 0;
3074 			psumd = 0;
3075 
3076 			if ((tcp->th_dport == nat->nat_oport) &&
3077 			    (tcp->th_sport != nat->nat_inport)) {
3078 
3079 				/*
3080 				 * Translate the source port.
3081 				 */
3082 
3083 				psum1 = ntohs(tcp->th_sport);
3084 				psum2 = ntohs(nat->nat_inport);
3085 				tcp->th_sport = nat->nat_inport;
3086 
3087 			} else if ((tcp->th_sport == nat->nat_oport) &&
3088 				    (tcp->th_dport != nat->nat_outport)) {
3089 
3090 				/*
3091 				 * Translate the destination port.
3092 				 */
3093 
3094 				psum1 = ntohs(tcp->th_dport);
3095 				psum2 = ntohs(nat->nat_outport);
3096 				tcp->th_dport = nat->nat_outport;
3097 			}
3098 
3099 			if ((oip->ip_p == IPPROTO_TCP) && (dlen >= 18)) {
3100 
3101 				/*
3102 				 * TCP checksum present.
3103 				 *
3104 				 * Adjust data checksum and icmp checksum to
3105 				 * compensate for any IP address change.
3106 				 */
3107 
3108 				sum1 = ntohs(tcp->th_sum);
3109 				fix_datacksum(&tcp->th_sum, sumd);
3110 				sum2 = ntohs(tcp->th_sum);
3111 				sumd2 = sumd << 1;
3112 				CALC_SUMD(sum1, sum2, sumd);
3113 				sumd2 += sumd;
3114 
3115 				/*
3116 				 * Also make data checksum adjustment to
3117 				 * compensate for any port change.
3118 				 */
3119 
3120 				if (psum1 != psum2) {
3121 					CALC_SUMD(psum1, psum2, psumd);
3122 					fix_datacksum(&tcp->th_sum, psumd);
3123 				}
3124 
3125 			} else if ((oip->ip_p == IPPROTO_UDP) &&
3126 				   (dlen >= 8) && (udp->uh_sum != 0)) {
3127 
3128 				/*
3129 				 * The UDP checksum is present and set.
3130 				 *
3131 				 * Adjust data checksum and icmp checksum to
3132 				 * compensate for any IP address change.
3133 				 */
3134 
3135 				sum1 = ntohs(udp->uh_sum);
3136 				fix_datacksum(&udp->uh_sum, sumd);
3137 				sum2 = ntohs(udp->uh_sum);
3138 				sumd2 = sumd << 1;
3139 				CALC_SUMD(sum1, sum2, sumd);
3140 				sumd2 += sumd;
3141 
3142 				/*
3143 				 * Also make data checksum adjustment to
3144 				 * compensate for any port change.
3145 				 */
3146 
3147 				if (psum1 != psum2) {
3148 					CALC_SUMD(psum1, psum2, psumd);
3149 					fix_datacksum(&udp->uh_sum, psumd);
3150 				}
3151 
3152 			} else {
3153 
3154 				/*
3155 				 * Data checksum was not present.
3156 				 *
3157 				 * Compensate for any port change.
3158 				 */
3159 
3160 				CALC_SUMD(psum2, psum1, psumd);
3161 				sumd2 += psumd;
3162 			}
3163 			break;
3164 
3165 		case IPPROTO_ICMP :
3166 
3167 			orgicmp = (icmphdr_t *)udp;
3168 
3169 			if ((nat->nat_dir == NAT_OUTBOUND) &&
3170 			    (orgicmp->icmp_id != nat->nat_inport) &&
3171 			    (dlen >= 8)) {
3172 
3173 				/*
3174 				 * Fix ICMP checksum (of the offening ICMP
3175 				 * query packet) to compensate the change
3176 				 * in the ICMP id of the offending ICMP
3177 				 * packet.
3178 				 *
3179 				 * Since you modify orgicmp->icmp_id with
3180 				 * a delta (say x) and you compensate that
3181 				 * in origicmp->icmp_cksum with a delta
3182 				 * minus x, you don't have to adjust the
3183 				 * overall icmp->icmp_cksum
3184 				 */
3185 
3186 				sum1 = ntohs(orgicmp->icmp_id);
3187 				sum2 = ntohs(nat->nat_inport);
3188 				CALC_SUMD(sum1, sum2, sumd);
3189 				orgicmp->icmp_id = nat->nat_inport;
3190 				fix_datacksum(&orgicmp->icmp_cksum, sumd);
3191 
3192 			} /* nat_dir can't be NAT_INBOUND for icmp queries */
3193 
3194 			break;
3195 
3196 		default :
3197 
3198 			break;
3199 
3200 	} /* switch (oip->ip_p) */
3201 
3202 	/*
3203 	 * Step 3
3204 	 * Make the adjustments to icmp checksum.
3205 	 */
3206 
3207 	if (sumd2 != 0) {
3208 		sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3209 		sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3210 		fix_incksum(&icmp->icmp_cksum, sumd2);
3211 	}
3212 	return nat;
3213 }
3214 
3215 
3216 /*
3217  * NB: these lookups don't lock access to the list, it assumed that it has
3218  * already been done!
3219  */
3220 
3221 /* ------------------------------------------------------------------------ */
3222 /* Function:    nat_inlookup                                                */
3223 /* Returns:     nat_t* - NULL == no match,                                  */
3224 /*                       else pointer to matching NAT entry                 */
3225 /* Parameters:  fin(I)    - pointer to packet information                   */
3226 /*              flags(I)  - NAT flags for this packet                       */
3227 /*              p(I)      - protocol for this packet                        */
3228 /*              src(I)    - source IP address                               */
3229 /*              mapdst(I) - destination IP address                          */
3230 /*                                                                          */
3231 /* Lookup a nat entry based on the mapped destination ip address/port and   */
3232 /* real source address/port.  We use this lookup when receiving a packet,   */
3233 /* we're looking for a table entry, based on the destination address.       */
3234 /*                                                                          */
3235 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY.         */
3236 /*                                                                          */
3237 /* NOTE: IT IS ASSUMED THAT ipf_nat IS ONLY HELD WITH A READ LOCK WHEN      */
3238 /*       THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags.             */
3239 /*                                                                          */
3240 /* flags   -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if   */
3241 /*            the packet is of said protocol                                */
3242 /* ------------------------------------------------------------------------ */
3243 nat_t *nat_inlookup(fin, flags, p, src, mapdst)
3244 fr_info_t *fin;
3245 u_int flags, p;
3246 struct in_addr src , mapdst;
3247 {
3248 	u_short sport, dport;
3249 	ipnat_t *ipn;
3250 	u_int sflags;
3251 	nat_t *nat;
3252 	int nflags;
3253 	u_32_t dst;
3254 	void *ifp;
3255 	u_int hv;
3256 	ipf_stack_t *ifs = fin->fin_ifs;
3257 
3258 	if (fin != NULL)
3259 		ifp = fin->fin_ifp;
3260 	else
3261 		ifp = NULL;
3262 	sport = 0;
3263 	dport = 0;
3264 	dst = mapdst.s_addr;
3265 	sflags = flags & NAT_TCPUDPICMP;
3266 
3267 	switch (p)
3268 	{
3269 	case IPPROTO_TCP :
3270 	case IPPROTO_UDP :
3271 		sport = htons(fin->fin_data[0]);
3272 		dport = htons(fin->fin_data[1]);
3273 		break;
3274 	case IPPROTO_ICMP :
3275 		if (flags & IPN_ICMPERR)
3276 			sport = fin->fin_data[1];
3277 		else
3278 			dport = fin->fin_data[1];
3279 		break;
3280 	default :
3281 		break;
3282 	}
3283 
3284 
3285 	if ((flags & SI_WILDP) != 0)
3286 		goto find_in_wild_ports;
3287 
3288 	hv = NAT_HASH_FN(dst, dport, 0xffffffff);
3289 	hv = NAT_HASH_FN(src.s_addr, hv + sport, ifs->ifs_ipf_nattable_sz);
3290 	nat = ifs->ifs_nat_table[1][hv];
3291 	for (; nat; nat = nat->nat_hnext[1]) {
3292 		if (nat->nat_v != 4)
3293 			continue;
3294 
3295 		if (nat->nat_ifps[0] != NULL) {
3296 			if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
3297 				continue;
3298 		} else if (ifp != NULL)
3299 			nat->nat_ifps[0] = ifp;
3300 
3301 		nflags = nat->nat_flags;
3302 
3303 		if (nat->nat_oip.s_addr == src.s_addr &&
3304 		    nat->nat_outip.s_addr == dst &&
3305 		    (((p == 0) &&
3306 		      (sflags == (nat->nat_flags & IPN_TCPUDPICMP)))
3307 		     || (p == nat->nat_p))) {
3308 			switch (p)
3309 			{
3310 #if 0
3311 			case IPPROTO_GRE :
3312 				if (nat->nat_call[1] != fin->fin_data[0])
3313 					continue;
3314 				break;
3315 #endif
3316 			case IPPROTO_ICMP :
3317 				if ((flags & IPN_ICMPERR) != 0) {
3318 					if (nat->nat_outport != sport)
3319 						continue;
3320 				} else {
3321 					if (nat->nat_outport != dport)
3322 						continue;
3323 				}
3324 				break;
3325 			case IPPROTO_TCP :
3326 			case IPPROTO_UDP :
3327 				if (nat->nat_oport != sport)
3328 					continue;
3329 				if (nat->nat_outport != dport)
3330 					continue;
3331 				break;
3332 			default :
3333 				break;
3334 			}
3335 
3336 			ipn = nat->nat_ptr;
3337 			if ((ipn != NULL) && (nat->nat_aps != NULL))
3338 				if (appr_match(fin, nat) != 0)
3339 					continue;
3340 			return nat;
3341 		}
3342 	}
3343 
3344 	/*
3345 	 * So if we didn't find it but there are wildcard members in the hash
3346 	 * table, go back and look for them.  We do this search and update here
3347 	 * because it is modifying the NAT table and we want to do this only
3348 	 * for the first packet that matches.  The exception, of course, is
3349 	 * for "dummy" (FI_IGNORE) lookups.
3350 	 */
3351 find_in_wild_ports:
3352 	if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH))
3353 		return NULL;
3354 	if (ifs->ifs_nat_stats.ns_wilds == 0)
3355 		return NULL;
3356 
3357 	RWLOCK_EXIT(&ifs->ifs_ipf_nat);
3358 
3359 	hv = NAT_HASH_FN(dst, 0, 0xffffffff);
3360 	hv = NAT_HASH_FN(src.s_addr, hv, ifs->ifs_ipf_nattable_sz);
3361 
3362 	WRITE_ENTER(&ifs->ifs_ipf_nat);
3363 
3364 	nat = ifs->ifs_nat_table[1][hv];
3365 	for (; nat; nat = nat->nat_hnext[1]) {
3366 		if (nat->nat_v != 4)
3367 			continue;
3368 
3369 		if (nat->nat_ifps[0] != NULL) {
3370 			if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
3371 				continue;
3372 		} else if (ifp != NULL)
3373 			nat->nat_ifps[0] = ifp;
3374 
3375 		if (nat->nat_p != fin->fin_p)
3376 			continue;
3377 		if (nat->nat_oip.s_addr != src.s_addr ||
3378 		    nat->nat_outip.s_addr != dst)
3379 			continue;
3380 
3381 		nflags = nat->nat_flags;
3382 		if (!(nflags & (NAT_TCPUDP|SI_WILDP)))
3383 			continue;
3384 
3385 		if (nat_wildok(nat, (int)sport, (int)dport, nflags,
3386 			       NAT_INBOUND) == 1) {
3387 			if ((fin->fin_flx & FI_IGNORE) != 0)
3388 				break;
3389 			if ((nflags & SI_CLONE) != 0) {
3390 				nat = fr_natclone(fin, nat);
3391 				if (nat == NULL)
3392 					break;
3393 			} else {
3394 				MUTEX_ENTER(&ifs->ifs_ipf_nat_new);
3395 				ifs->ifs_nat_stats.ns_wilds--;
3396 				MUTEX_EXIT(&ifs->ifs_ipf_nat_new);
3397 			}
3398 			nat->nat_oport = sport;
3399 			nat->nat_outport = dport;
3400 			nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
3401 			nat_tabmove(nat, ifs);
3402 			break;
3403 		}
3404 	}
3405 
3406 	MUTEX_DOWNGRADE(&ifs->ifs_ipf_nat);
3407 
3408 	return nat;
3409 }
3410 
3411 
3412 /* ------------------------------------------------------------------------ */
3413 /* Function:    nat_tabmove                                                 */
3414 /* Returns:     Nil                                                         */
3415 /* Parameters:  nat(I) - pointer to NAT structure                           */
3416 /* Write Lock:  ipf_nat                                                     */
3417 /*                                                                          */
3418 /* This function is only called for TCP/UDP NAT table entries where the     */
3419 /* original was placed in the table without hashing on the ports and we now */
3420 /* want to include hashing on port numbers.                                 */
3421 /* ------------------------------------------------------------------------ */
3422 static void nat_tabmove(nat, ifs)
3423 nat_t *nat;
3424 ipf_stack_t *ifs;
3425 {
3426 	nat_t **natp;
3427 	u_int hv;
3428 
3429 	if (nat->nat_flags & SI_CLONE)
3430 		return;
3431 
3432 	/*
3433 	 * Remove the NAT entry from the old location
3434 	 */
3435 	if (nat->nat_hnext[0])
3436 		nat->nat_hnext[0]->nat_phnext[0] = nat->nat_phnext[0];
3437 	*nat->nat_phnext[0] = nat->nat_hnext[0];
3438 	ifs->ifs_nat_stats.ns_bucketlen[0][nat->nat_hv[0]]--;
3439 
3440 	if (nat->nat_hnext[1])
3441 		nat->nat_hnext[1]->nat_phnext[1] = nat->nat_phnext[1];
3442 	*nat->nat_phnext[1] = nat->nat_hnext[1];
3443 	ifs->ifs_nat_stats.ns_bucketlen[1][nat->nat_hv[1]]--;
3444 
3445 	/*
3446 	 * Add into the NAT table in the new position
3447 	 */
3448 	hv = NAT_HASH_FN(nat->nat_inip.s_addr, nat->nat_inport, 0xffffffff);
3449 	hv = NAT_HASH_FN(nat->nat_oip.s_addr, hv + nat->nat_oport,
3450 			 ifs->ifs_ipf_nattable_sz);
3451 	nat->nat_hv[0] = hv;
3452 	natp = &ifs->ifs_nat_table[0][hv];
3453 	if (*natp)
3454 		(*natp)->nat_phnext[0] = &nat->nat_hnext[0];
3455 	nat->nat_phnext[0] = natp;
3456 	nat->nat_hnext[0] = *natp;
3457 	*natp = nat;
3458 	ifs->ifs_nat_stats.ns_bucketlen[0][hv]++;
3459 
3460 	hv = NAT_HASH_FN(nat->nat_outip.s_addr, nat->nat_outport, 0xffffffff);
3461 	hv = NAT_HASH_FN(nat->nat_oip.s_addr, hv + nat->nat_oport,
3462 			 ifs->ifs_ipf_nattable_sz);
3463 	nat->nat_hv[1] = hv;
3464 	natp = &ifs->ifs_nat_table[1][hv];
3465 	if (*natp)
3466 		(*natp)->nat_phnext[1] = &nat->nat_hnext[1];
3467 	nat->nat_phnext[1] = natp;
3468 	nat->nat_hnext[1] = *natp;
3469 	*natp = nat;
3470 	ifs->ifs_nat_stats.ns_bucketlen[1][hv]++;
3471 }
3472 
3473 
3474 /* ------------------------------------------------------------------------ */
3475 /* Function:    nat_outlookup                                               */
3476 /* Returns:     nat_t* - NULL == no match,                                  */
3477 /*                       else pointer to matching NAT entry                 */
3478 /* Parameters:  fin(I)   - pointer to packet information                    */
3479 /*              flags(I) - NAT flags for this packet                        */
3480 /*              p(I)     - protocol for this packet                         */
3481 /*              src(I)   - source IP address                                */
3482 /*              dst(I)   - destination IP address                           */
3483 /*              rw(I)    - 1 == write lock on ipf_nat held, 0 == read lock. */
3484 /*                                                                          */
3485 /* Lookup a nat entry based on the source 'real' ip address/port and        */
3486 /* destination address/port.  We use this lookup when sending a packet out, */
3487 /* we're looking for a table entry, based on the source address.            */
3488 /*                                                                          */
3489 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY.         */
3490 /*                                                                          */
3491 /* NOTE: IT IS ASSUMED THAT ipf_nat IS ONLY HELD WITH A READ LOCK WHEN      */
3492 /*       THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags.             */
3493 /*                                                                          */
3494 /* flags   -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if   */
3495 /*            the packet is of said protocol                                */
3496 /* ------------------------------------------------------------------------ */
3497 nat_t *nat_outlookup(fin, flags, p, src, dst)
3498 fr_info_t *fin;
3499 u_int flags, p;
3500 struct in_addr src , dst;
3501 {
3502 	u_short sport, dport;
3503 	u_int sflags;
3504 	ipnat_t *ipn;
3505 	u_32_t srcip;
3506 	nat_t *nat;
3507 	int nflags;
3508 	void *ifp;
3509 	u_int hv;
3510 	ipf_stack_t *ifs = fin->fin_ifs;
3511 
3512 	ifp = fin->fin_ifp;
3513 
3514 	srcip = src.s_addr;
3515 	sflags = flags & IPN_TCPUDPICMP;
3516 	sport = 0;
3517 	dport = 0;
3518 
3519 	switch (p)
3520 	{
3521 	case IPPROTO_TCP :
3522 	case IPPROTO_UDP :
3523 		sport = htons(fin->fin_data[0]);
3524 		dport = htons(fin->fin_data[1]);
3525 		break;
3526 	case IPPROTO_ICMP :
3527 		if (flags & IPN_ICMPERR)
3528 			sport = fin->fin_data[1];
3529 		else
3530 			dport = fin->fin_data[1];
3531 		break;
3532 	default :
3533 		break;
3534 	}
3535 
3536 	if ((flags & SI_WILDP) != 0)
3537 		goto find_out_wild_ports;
3538 
3539 	hv = NAT_HASH_FN(srcip, sport, 0xffffffff);
3540 	hv = NAT_HASH_FN(dst.s_addr, hv + dport, ifs->ifs_ipf_nattable_sz);
3541 	nat = ifs->ifs_nat_table[0][hv];
3542 	for (; nat; nat = nat->nat_hnext[0]) {
3543 		if (nat->nat_v != 4)
3544 			continue;
3545 
3546 		if (nat->nat_ifps[1] != NULL) {
3547 			if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
3548 				continue;
3549 		} else if (ifp != NULL)
3550 			nat->nat_ifps[1] = ifp;
3551 
3552 		nflags = nat->nat_flags;
3553 
3554 		if (nat->nat_inip.s_addr == srcip &&
3555 		    nat->nat_oip.s_addr == dst.s_addr &&
3556 		    (((p == 0) && (sflags == (nflags & NAT_TCPUDPICMP)))
3557 		     || (p == nat->nat_p))) {
3558 			switch (p)
3559 			{
3560 #if 0
3561 			case IPPROTO_GRE :
3562 				if (nat->nat_call[1] != fin->fin_data[0])
3563 					continue;
3564 				break;
3565 #endif
3566 			case IPPROTO_TCP :
3567 			case IPPROTO_UDP :
3568 				if (nat->nat_oport != dport)
3569 					continue;
3570 				if (nat->nat_inport != sport)
3571 					continue;
3572 				break;
3573 			default :
3574 				break;
3575 			}
3576 
3577 			ipn = nat->nat_ptr;
3578 			if ((ipn != NULL) && (nat->nat_aps != NULL))
3579 				if (appr_match(fin, nat) != 0)
3580 					continue;
3581 			return nat;
3582 		}
3583 	}
3584 
3585 	/*
3586 	 * So if we didn't find it but there are wildcard members in the hash
3587 	 * table, go back and look for them.  We do this search and update here
3588 	 * because it is modifying the NAT table and we want to do this only
3589 	 * for the first packet that matches.  The exception, of course, is
3590 	 * for "dummy" (FI_IGNORE) lookups.
3591 	 */
3592 find_out_wild_ports:
3593 	if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH))
3594 		return NULL;
3595 	if (ifs->ifs_nat_stats.ns_wilds == 0)
3596 		return NULL;
3597 
3598 	RWLOCK_EXIT(&ifs->ifs_ipf_nat);
3599 
3600 	hv = NAT_HASH_FN(srcip, 0, 0xffffffff);
3601 	hv = NAT_HASH_FN(dst.s_addr, hv, ifs->ifs_ipf_nattable_sz);
3602 
3603 	WRITE_ENTER(&ifs->ifs_ipf_nat);
3604 
3605 	nat = ifs->ifs_nat_table[0][hv];
3606 	for (; nat; nat = nat->nat_hnext[0]) {
3607 		if (nat->nat_v != 4)
3608 			continue;
3609 
3610 		if (nat->nat_ifps[1] != NULL) {
3611 			if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
3612 				continue;
3613 		} else if (ifp != NULL)
3614 			nat->nat_ifps[1] = ifp;
3615 
3616 		if (nat->nat_p != fin->fin_p)
3617 			continue;
3618 		if ((nat->nat_inip.s_addr != srcip) ||
3619 		    (nat->nat_oip.s_addr != dst.s_addr))
3620 			continue;
3621 
3622 		nflags = nat->nat_flags;
3623 		if (!(nflags & (NAT_TCPUDP|SI_WILDP)))
3624 			continue;
3625 
3626 		if (nat_wildok(nat, (int)sport, (int)dport, nflags,
3627 			       NAT_OUTBOUND) == 1) {
3628 			if ((fin->fin_flx & FI_IGNORE) != 0)
3629 				break;
3630 			if ((nflags & SI_CLONE) != 0) {
3631 				nat = fr_natclone(fin, nat);
3632 				if (nat == NULL)
3633 					break;
3634 			} else {
3635 				MUTEX_ENTER(&ifs->ifs_ipf_nat_new);
3636 				ifs->ifs_nat_stats.ns_wilds--;
3637 				MUTEX_EXIT(&ifs->ifs_ipf_nat_new);
3638 			}
3639 			nat->nat_inport = sport;
3640 			nat->nat_oport = dport;
3641 			if (nat->nat_outport == 0)
3642 				nat->nat_outport = sport;
3643 			nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
3644 			nat_tabmove(nat, ifs);
3645 			break;
3646 		}
3647 	}
3648 
3649 	MUTEX_DOWNGRADE(&ifs->ifs_ipf_nat);
3650 
3651 	return nat;
3652 }
3653 
3654 
3655 /* ------------------------------------------------------------------------ */
3656 /* Function:    nat_lookupredir                                             */
3657 /* Returns:     nat_t* - NULL == no match,                                  */
3658 /*                       else pointer to matching NAT entry                 */
3659 /* Parameters:  np(I) - pointer to description of packet to find NAT table  */
3660 /*                      entry for.                                          */
3661 /*                                                                          */
3662 /* Lookup the NAT tables to search for a matching redirect                  */
3663 /* ------------------------------------------------------------------------ */
3664 nat_t *nat_lookupredir(np, ifs)
3665 natlookup_t *np;
3666 ipf_stack_t *ifs;
3667 {
3668 	fr_info_t fi;
3669 	nat_t *nat;
3670 
3671 	bzero((char *)&fi, sizeof(fi));
3672 	if (np->nl_flags & IPN_IN) {
3673 		fi.fin_data[0] = ntohs(np->nl_realport);
3674 		fi.fin_data[1] = ntohs(np->nl_outport);
3675 	} else {
3676 		fi.fin_data[0] = ntohs(np->nl_inport);
3677 		fi.fin_data[1] = ntohs(np->nl_outport);
3678 	}
3679 	if (np->nl_flags & IPN_TCP)
3680 		fi.fin_p = IPPROTO_TCP;
3681 	else if (np->nl_flags & IPN_UDP)
3682 		fi.fin_p = IPPROTO_UDP;
3683 	else if (np->nl_flags & (IPN_ICMPERR|IPN_ICMPQUERY))
3684 		fi.fin_p = IPPROTO_ICMP;
3685 
3686 	fi.fin_ifs = ifs;
3687 	/*
3688 	 * We can do two sorts of lookups:
3689 	 * - IPN_IN: we have the `real' and `out' address, look for `in'.
3690 	 * - default: we have the `in' and `out' address, look for `real'.
3691 	 */
3692 	if (np->nl_flags & IPN_IN) {
3693 		if ((nat = nat_inlookup(&fi, np->nl_flags, fi.fin_p,
3694 					np->nl_realip, np->nl_outip))) {
3695 			np->nl_inip = nat->nat_inip;
3696 			np->nl_inport = nat->nat_inport;
3697 		}
3698 	} else {
3699 		/*
3700 		 * If nl_inip is non null, this is a lookup based on the real
3701 		 * ip address. Else, we use the fake.
3702 		 */
3703 		if ((nat = nat_outlookup(&fi, np->nl_flags, fi.fin_p,
3704 					 np->nl_inip, np->nl_outip))) {
3705 
3706 			if ((np->nl_flags & IPN_FINDFORWARD) != 0) {
3707 				fr_info_t fin;
3708 				bzero((char *)&fin, sizeof(fin));
3709 				fin.fin_p = nat->nat_p;
3710 				fin.fin_data[0] = ntohs(nat->nat_outport);
3711 				fin.fin_data[1] = ntohs(nat->nat_oport);
3712 				fin.fin_ifs = ifs;
3713 				if (nat_inlookup(&fin, np->nl_flags, fin.fin_p,
3714 						 nat->nat_outip,
3715 						 nat->nat_oip) != NULL) {
3716 					np->nl_flags &= ~IPN_FINDFORWARD;
3717 				}
3718 			}
3719 
3720 			np->nl_realip = nat->nat_outip;
3721 			np->nl_realport = nat->nat_outport;
3722 		}
3723  	}
3724 
3725 	return nat;
3726 }
3727 
3728 
3729 /* ------------------------------------------------------------------------ */
3730 /* Function:    nat_match                                                   */
3731 /* Returns:     int - 0 == no match, 1 == match                             */
3732 /* Parameters:  fin(I)   - pointer to packet information                    */
3733 /*              np(I)    - pointer to NAT rule                              */
3734 /*                                                                          */
3735 /* Pull the matching of a packet against a NAT rule out of that complex     */
3736 /* loop inside fr_checknatin() and lay it out properly in its own function. */
3737 /* ------------------------------------------------------------------------ */
3738 static int nat_match(fin, np)
3739 fr_info_t *fin;
3740 ipnat_t *np;
3741 {
3742 	frtuc_t *ft;
3743 
3744 	if (fin->fin_v != 4)
3745 		return 0;
3746 
3747 	if (np->in_p && fin->fin_p != np->in_p)
3748 		return 0;
3749 
3750 	if (fin->fin_out) {
3751 		if (!(np->in_redir & (NAT_MAP|NAT_MAPBLK)))
3752 			return 0;
3753 		if (((fin->fin_fi.fi_saddr & np->in_inmsk) != np->in_inip)
3754 		    ^ ((np->in_flags & IPN_NOTSRC) != 0))
3755 			return 0;
3756 		if (((fin->fin_fi.fi_daddr & np->in_srcmsk) != np->in_srcip)
3757 		    ^ ((np->in_flags & IPN_NOTDST) != 0))
3758 			return 0;
3759 	} else {
3760 		if (!(np->in_redir & NAT_REDIRECT))
3761 			return 0;
3762 		if (((fin->fin_fi.fi_saddr & np->in_srcmsk) != np->in_srcip)
3763 		    ^ ((np->in_flags & IPN_NOTSRC) != 0))
3764 			return 0;
3765 		if (((fin->fin_fi.fi_daddr & np->in_outmsk) != np->in_outip)
3766 		    ^ ((np->in_flags & IPN_NOTDST) != 0))
3767 			return 0;
3768 	}
3769 
3770 	ft = &np->in_tuc;
3771 	if (!(fin->fin_flx & FI_TCPUDP) ||
3772 	    (fin->fin_flx & (FI_SHORT|FI_FRAGBODY))) {
3773 		if (ft->ftu_scmp || ft->ftu_dcmp)
3774 			return 0;
3775 		return 1;
3776 	}
3777 
3778 	return fr_tcpudpchk(fin, ft);
3779 }
3780 
3781 
3782 /* ------------------------------------------------------------------------ */
3783 /* Function:    nat_update                                                  */
3784 /* Returns:     Nil                                                         */
3785 /* Parameters:  nat(I)    - pointer to NAT structure                        */
3786 /*              np(I)     - pointer to NAT rule                             */
3787 /*                                                                          */
3788 /* Updates the lifetime of a NAT table entry for non-TCP packets.  Must be  */
3789 /* called with fin_rev updated - i.e. after calling nat_proto().            */
3790 /* ------------------------------------------------------------------------ */
3791 void nat_update(fin, nat, np)
3792 fr_info_t *fin;
3793 nat_t *nat;
3794 ipnat_t *np;
3795 {
3796 	ipftq_t *ifq, *ifq2;
3797 	ipftqent_t *tqe;
3798 	ipf_stack_t *ifs = fin->fin_ifs;
3799 
3800 	MUTEX_ENTER(&nat->nat_lock);
3801 	tqe = &nat->nat_tqe;
3802 	ifq = tqe->tqe_ifq;
3803 
3804 	/*
3805 	 * We allow over-riding of NAT timeouts from NAT rules, even for
3806 	 * TCP, however, if it is TCP and there is no rule timeout set,
3807 	 * then do not update the timeout here.
3808 	 */
3809 	if (np != NULL)
3810 		ifq2 = np->in_tqehead[fin->fin_rev];
3811 	else
3812 		ifq2 = NULL;
3813 
3814 	if (nat->nat_p == IPPROTO_TCP && ifq2 == NULL) {
3815 		(void) fr_tcp_age(&nat->nat_tqe, fin, ifs->ifs_nat_tqb, 0);
3816 	} else {
3817 		if (ifq2 == NULL) {
3818 			if (nat->nat_p == IPPROTO_UDP)
3819 				ifq2 = &ifs->ifs_nat_udptq;
3820 			else if (nat->nat_p == IPPROTO_ICMP)
3821 				ifq2 = &ifs->ifs_nat_icmptq;
3822 			else
3823 				ifq2 = &ifs->ifs_nat_iptq;
3824 		}
3825 
3826 		fr_movequeue(tqe, ifq, ifq2, ifs);
3827 	}
3828 	MUTEX_EXIT(&nat->nat_lock);
3829 }
3830 
3831 
3832 /* ------------------------------------------------------------------------ */
3833 /* Function:    fr_checknatout                                              */
3834 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
3835 /*                     0 == no packet translation occurred,                 */
3836 /*                     1 == packet was successfully translated.             */
3837 /* Parameters:  fin(I)   - pointer to packet information                    */
3838 /*              passp(I) - pointer to filtering result flags                */
3839 /*                                                                          */
3840 /* Check to see if an outcoming packet should be changed.  ICMP packets are */
3841 /* first checked to see if they match an existing entry (if an error),      */
3842 /* otherwise a search of the current NAT table is made.  If neither results */
3843 /* in a match then a search for a matching NAT rule is made.  Create a new  */
3844 /* NAT entry if a we matched a NAT rule.  Lastly, actually change the       */
3845 /* packet header(s) as required.                                            */
3846 /* ------------------------------------------------------------------------ */
3847 int fr_checknatout(fin, passp)
3848 fr_info_t *fin;
3849 u_32_t *passp;
3850 {
3851 	ipnat_t *np = NULL, *npnext;
3852 	struct ifnet *ifp, *sifp;
3853 	icmphdr_t *icmp = NULL;
3854 	tcphdr_t *tcp = NULL;
3855 	int rval, natfailed;
3856 	u_int nflags = 0;
3857 	u_32_t ipa, iph;
3858 	int natadd = 1;
3859 	frentry_t *fr;
3860 	nat_t *nat;
3861 	ipf_stack_t *ifs = fin->fin_ifs;
3862 
3863 	if (ifs->ifs_nat_stats.ns_rules == 0 || ifs->ifs_fr_nat_lock != 0)
3864 		return 0;
3865 
3866 	natfailed = 0;
3867 	fr = fin->fin_fr;
3868 	sifp = fin->fin_ifp;
3869 	if ((fr != NULL) && !(fr->fr_flags & FR_DUP) &&
3870 	    fr->fr_tifs[fin->fin_rev].fd_ifp &&
3871 	    fr->fr_tifs[fin->fin_rev].fd_ifp != (void *)-1)
3872 		fin->fin_ifp = fr->fr_tifs[fin->fin_rev].fd_ifp;
3873 	ifp = fin->fin_ifp;
3874 
3875 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
3876 		switch (fin->fin_p)
3877 		{
3878 		case IPPROTO_TCP :
3879 			nflags = IPN_TCP;
3880 			break;
3881 		case IPPROTO_UDP :
3882 			nflags = IPN_UDP;
3883 			break;
3884 		case IPPROTO_ICMP :
3885 			icmp = fin->fin_dp;
3886 
3887 			/*
3888 			 * This is an incoming packet, so the destination is
3889 			 * the icmp_id and the source port equals 0
3890 			 */
3891 			if (nat_icmpquerytype4(icmp->icmp_type))
3892 				nflags = IPN_ICMPQUERY;
3893 			break;
3894 		default :
3895 			break;
3896 		}
3897 
3898 		if ((nflags & IPN_TCPUDP))
3899 			tcp = fin->fin_dp;
3900 	}
3901 
3902 	ipa = fin->fin_saddr;
3903 
3904 	READ_ENTER(&ifs->ifs_ipf_nat);
3905 
3906 	if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
3907 	    (nat = nat_icmperror(fin, &nflags, NAT_OUTBOUND)))
3908 		/*EMPTY*/;
3909 	else if ((fin->fin_flx & FI_FRAG) && (nat = fr_nat_knownfrag(fin)))
3910 		natadd = 0;
3911 	else if ((nat = nat_outlookup(fin, nflags|NAT_SEARCH, (u_int)fin->fin_p,
3912 				      fin->fin_src, fin->fin_dst))) {
3913 		nflags = nat->nat_flags;
3914 	} else {
3915 		u_32_t hv, msk, nmsk;
3916 
3917 		/*
3918 		 * If there is no current entry in the nat table for this IP#,
3919 		 * create one for it (if there is a matching rule).
3920 		 */
3921 		msk = 0xffffffff;
3922 		nmsk = ifs->ifs_nat_masks;
3923 maskloop:
3924 		iph = ipa & htonl(msk);
3925 		hv = NAT_HASH_FN(iph, 0, ifs->ifs_ipf_natrules_sz);
3926 		for (np = ifs->ifs_nat_rules[hv]; np; np = npnext) {
3927 			npnext = np->in_mnext;
3928 			if ((np->in_ifps[1] && (np->in_ifps[1] != ifp)))
3929 				continue;
3930 			if (np->in_v != fin->fin_v)
3931 				continue;
3932 			if (np->in_p && (np->in_p != fin->fin_p))
3933 				continue;
3934 			if ((np->in_flags & IPN_RF) && !(np->in_flags & nflags))
3935 				continue;
3936 			if (np->in_flags & IPN_FILTER) {
3937 				if (!nat_match(fin, np))
3938 					continue;
3939 			} else if ((ipa & np->in_inmsk) != np->in_inip)
3940 				continue;
3941 
3942 			if ((fr != NULL) &&
3943 			    !fr_matchtag(&np->in_tag, &fr->fr_nattag))
3944 				continue;
3945 
3946 			if (*np->in_plabel != '\0') {
3947 				if (((np->in_flags & IPN_FILTER) == 0) &&
3948 				    (np->in_dport != tcp->th_dport))
3949 					continue;
3950 				if (appr_ok(fin, tcp, np) == 0)
3951 					continue;
3952 			}
3953 
3954 			ATOMIC_INC32(np->in_use);
3955 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
3956 			WRITE_ENTER(&ifs->ifs_ipf_nat);
3957 			nat = nat_new(fin, np, NULL, nflags, NAT_OUTBOUND);
3958 			if (nat != NULL) {
3959 				np->in_use--;
3960 				np->in_hits++;
3961 				MUTEX_DOWNGRADE(&ifs->ifs_ipf_nat);
3962 				break;
3963 			}
3964 			natfailed = -1;
3965 			npnext = np->in_mnext;
3966 			fr_ipnatderef(&np, ifs);
3967 			MUTEX_DOWNGRADE(&ifs->ifs_ipf_nat);
3968 		}
3969 		if ((np == NULL) && (nmsk != 0)) {
3970 			while (nmsk) {
3971 				msk <<= 1;
3972 				if (nmsk & 0x80000000)
3973 					break;
3974 				nmsk <<= 1;
3975 			}
3976 			if (nmsk != 0) {
3977 				nmsk <<= 1;
3978 				goto maskloop;
3979 			}
3980 		}
3981 	}
3982 
3983 	if (nat != NULL) {
3984 		rval = fr_natout(fin, nat, natadd, nflags);
3985 		if (rval == 1) {
3986 			MUTEX_ENTER(&nat->nat_lock);
3987 			nat->nat_ref++;
3988 			MUTEX_EXIT(&nat->nat_lock);
3989 			nat->nat_touched = ifs->ifs_fr_ticks;
3990 			fin->fin_nat = nat;
3991 		}
3992 	} else
3993 		rval = natfailed;
3994 	RWLOCK_EXIT(&ifs->ifs_ipf_nat);
3995 
3996 	if (rval == -1) {
3997 		if (passp != NULL)
3998 			*passp = FR_BLOCK;
3999 		fin->fin_flx |= FI_BADNAT;
4000 	}
4001 	fin->fin_ifp = sifp;
4002 	return rval;
4003 }
4004 
4005 /* ------------------------------------------------------------------------ */
4006 /* Function:    fr_natout                                                   */
4007 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
4008 /*                     1 == packet was successfully translated.             */
4009 /* Parameters:  fin(I)    - pointer to packet information                   */
4010 /*              nat(I)    - pointer to NAT structure                        */
4011 /*              natadd(I) - flag indicating if it is safe to add frag cache */
4012 /*              nflags(I) - NAT flags set for this packet                   */
4013 /*                                                                          */
4014 /* Translate a packet coming "out" on an interface.                         */
4015 /* ------------------------------------------------------------------------ */
4016 int fr_natout(fin, nat, natadd, nflags)
4017 fr_info_t *fin;
4018 nat_t *nat;
4019 int natadd;
4020 u_32_t nflags;
4021 {
4022 	icmphdr_t *icmp;
4023 	u_short *csump;
4024 	u_32_t sumd;
4025 	tcphdr_t *tcp;
4026 	ipnat_t *np;
4027 	int i;
4028 	ipf_stack_t *ifs = fin->fin_ifs;
4029 
4030 	if (fin->fin_v == 6) {
4031 #ifdef	USE_INET6
4032 		return fr_nat6out(fin, nat, natadd, nflags);
4033 #else
4034 		return NULL;
4035 #endif
4036 	}
4037 
4038 #if SOLARIS && defined(_KERNEL)
4039 	net_data_t net_data_p = ifs->ifs_ipf_ipv4;
4040 #endif
4041 
4042 	tcp = NULL;
4043 	icmp = NULL;
4044 	csump = NULL;
4045 	np = nat->nat_ptr;
4046 
4047 	if ((natadd != 0) && (fin->fin_flx & FI_FRAG))
4048 		(void) fr_nat_newfrag(fin, 0, nat);
4049 
4050 	MUTEX_ENTER(&nat->nat_lock);
4051 	nat->nat_bytes[1] += fin->fin_plen;
4052 	nat->nat_pkts[1]++;
4053 	MUTEX_EXIT(&nat->nat_lock);
4054 
4055 	/*
4056 	 * Fix up checksums, not by recalculating them, but
4057 	 * simply computing adjustments.
4058 	 * This is only done for STREAMS based IP implementations where the
4059 	 * checksum has already been calculated by IP.  In all other cases,
4060 	 * IPFilter is called before the checksum needs calculating so there
4061 	 * is no call to modify whatever is in the header now.
4062 	 */
4063 	ASSERT(fin->fin_m != NULL);
4064 	if (fin->fin_v == 4 && !NET_IS_HCK_L3_FULL(net_data_p, fin->fin_m)) {
4065 		if (nflags == IPN_ICMPERR) {
4066 			u_32_t s1, s2;
4067 
4068 			s1 = LONG_SUM(ntohl(fin->fin_saddr));
4069 			s2 = LONG_SUM(ntohl(nat->nat_outip.s_addr));
4070 			CALC_SUMD(s1, s2, sumd);
4071 
4072 			fix_outcksum(&fin->fin_ip->ip_sum, sumd);
4073 		}
4074 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
4075     defined(linux) || defined(BRIDGE_IPF)
4076 		else {
4077 			/*
4078 			 * Strictly speaking, this isn't necessary on BSD
4079 			 * kernels because they do checksum calculation after
4080 			 * this code has run BUT if ipfilter is being used
4081 			 * to do NAT as a bridge, that code doesn't exist.
4082 			 */
4083 			if (nat->nat_dir == NAT_OUTBOUND)
4084 				fix_outcksum(&fin->fin_ip->ip_sum,
4085 					    nat->nat_ipsumd);
4086 			else
4087 				fix_incksum(&fin->fin_ip->ip_sum,
4088 				 	   nat->nat_ipsumd);
4089 		}
4090 #endif
4091 	}
4092 
4093 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
4094 		if ((nat->nat_outport != 0) && (nflags & IPN_TCPUDP)) {
4095 			tcp = fin->fin_dp;
4096 
4097 			tcp->th_sport = nat->nat_outport;
4098 			fin->fin_data[0] = ntohs(nat->nat_outport);
4099 		}
4100 
4101 		if ((nat->nat_outport != 0) && (nflags & IPN_ICMPQUERY)) {
4102 			icmp = fin->fin_dp;
4103 			icmp->icmp_id = nat->nat_outport;
4104 		}
4105 
4106 		csump = nat_proto(fin, nat, nflags);
4107 	}
4108 
4109 	fin->fin_ip->ip_src = nat->nat_outip;
4110 
4111 	nat_update(fin, nat, np);
4112 
4113 	/*
4114 	 * The above comments do not hold for layer 4 (or higher) checksums...
4115 	 */
4116 	if (csump != NULL && !NET_IS_HCK_L4_FULL(net_data_p, fin->fin_m)) {
4117 		if (nflags & IPN_TCPUDP &&
4118 	   	    NET_IS_HCK_L4_PART(net_data_p, fin->fin_m))
4119 			sumd = nat->nat_sumd[1];
4120 		else
4121 			sumd = nat->nat_sumd[0];
4122 
4123 		if (nat->nat_dir == NAT_OUTBOUND)
4124 			fix_outcksum(csump, sumd);
4125 		else
4126 			fix_incksum(csump, sumd);
4127 	}
4128 #ifdef	IPFILTER_SYNC
4129 	ipfsync_update(SMC_NAT, fin, nat->nat_sync);
4130 #endif
4131 	/* ------------------------------------------------------------- */
4132 	/* A few quick notes:						 */
4133 	/*	Following are test conditions prior to calling the 	 */
4134 	/*	appr_check routine.					 */
4135 	/*								 */
4136 	/* 	A NULL tcp indicates a non TCP/UDP packet.  When dealing */
4137 	/*	with a redirect rule, we attempt to match the packet's	 */
4138 	/*	source port against in_dport, otherwise	we'd compare the */
4139 	/*	packet's destination.			 		 */
4140 	/* ------------------------------------------------------------- */
4141 	if ((np != NULL) && (np->in_apr != NULL)) {
4142 		i = appr_check(fin, nat);
4143 		if (i == 0)
4144 			i = 1;
4145 	} else
4146 		i = 1;
4147 	ifs->ifs_nat_stats.ns_mapped[1]++;
4148 	fin->fin_flx |= FI_NATED;
4149 	return i;
4150 }
4151 
4152 
4153 /* ------------------------------------------------------------------------ */
4154 /* Function:    fr_checknatin                                               */
4155 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
4156 /*                     0 == no packet translation occurred,                 */
4157 /*                     1 == packet was successfully translated.             */
4158 /* Parameters:  fin(I)   - pointer to packet information                    */
4159 /*              passp(I) - pointer to filtering result flags                */
4160 /*                                                                          */
4161 /* Check to see if an incoming packet should be changed.  ICMP packets are  */
4162 /* first checked to see if they match an existing entry (if an error),      */
4163 /* otherwise a search of the current NAT table is made.  If neither results */
4164 /* in a match then a search for a matching NAT rule is made.  Create a new  */
4165 /* NAT entry if a we matched a NAT rule.  Lastly, actually change the       */
4166 /* packet header(s) as required.                                            */
4167 /* ------------------------------------------------------------------------ */
4168 int fr_checknatin(fin, passp)
4169 fr_info_t *fin;
4170 u_32_t *passp;
4171 {
4172 	u_int nflags, natadd;
4173 	ipnat_t *np, *npnext;
4174 	int rval, natfailed;
4175 	struct ifnet *ifp;
4176 	struct in_addr in;
4177 	icmphdr_t *icmp;
4178 	tcphdr_t *tcp;
4179 	u_short dport;
4180 	nat_t *nat;
4181 	u_32_t iph;
4182 	ipf_stack_t *ifs = fin->fin_ifs;
4183 
4184 	if (ifs->ifs_nat_stats.ns_rules == 0 || ifs->ifs_fr_nat_lock != 0)
4185 		return 0;
4186 
4187 	tcp = NULL;
4188 	icmp = NULL;
4189 	dport = 0;
4190 	natadd = 1;
4191 	nflags = 0;
4192 	natfailed = 0;
4193 	ifp = fin->fin_ifp;
4194 
4195 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
4196 		switch (fin->fin_p)
4197 		{
4198 		case IPPROTO_TCP :
4199 			nflags = IPN_TCP;
4200 			break;
4201 		case IPPROTO_UDP :
4202 			nflags = IPN_UDP;
4203 			break;
4204 		case IPPROTO_ICMP :
4205 			icmp = fin->fin_dp;
4206 
4207 			/*
4208 			 * This is an incoming packet, so the destination is
4209 			 * the icmp_id and the source port equals 0
4210 			 */
4211 			if (nat_icmpquerytype4(icmp->icmp_type)) {
4212 				nflags = IPN_ICMPQUERY;
4213 				dport = icmp->icmp_id;
4214 			} break;
4215 		default :
4216 			break;
4217 		}
4218 
4219 		if ((nflags & IPN_TCPUDP)) {
4220 			tcp = fin->fin_dp;
4221 			dport = tcp->th_dport;
4222 		}
4223 	}
4224 
4225 	in = fin->fin_dst;
4226 
4227 	READ_ENTER(&ifs->ifs_ipf_nat);
4228 
4229 	if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
4230 	    (nat = nat_icmperror(fin, &nflags, NAT_INBOUND)))
4231 		/*EMPTY*/;
4232 	else if ((fin->fin_flx & FI_FRAG) && (nat = fr_nat_knownfrag(fin)))
4233 		natadd = 0;
4234 	else if ((nat = nat_inlookup(fin, nflags|NAT_SEARCH, (u_int)fin->fin_p,
4235 				     fin->fin_src, in))) {
4236 		nflags = nat->nat_flags;
4237 	} else {
4238 		u_32_t hv, msk, rmsk;
4239 
4240 		rmsk = ifs->ifs_rdr_masks;
4241 		msk = 0xffffffff;
4242 		/*
4243 		 * If there is no current entry in the nat table for this IP#,
4244 		 * create one for it (if there is a matching rule).
4245 		 */
4246 maskloop:
4247 		iph = in.s_addr & htonl(msk);
4248 		hv = NAT_HASH_FN(iph, 0, ifs->ifs_ipf_rdrrules_sz);
4249 		for (np = ifs->ifs_rdr_rules[hv]; np; np = npnext) {
4250 			npnext = np->in_rnext;
4251 			if (np->in_ifps[0] && (np->in_ifps[0] != ifp))
4252 				continue;
4253 			if (np->in_v != fin->fin_v)
4254 				continue;
4255 			if (np->in_p && (np->in_p != fin->fin_p))
4256 				continue;
4257 			if ((np->in_flags & IPN_RF) && !(np->in_flags & nflags))
4258 				continue;
4259 			if (np->in_flags & IPN_FILTER) {
4260 				if (!nat_match(fin, np))
4261 					continue;
4262 			} else {
4263 				if ((in.s_addr & np->in_outmsk) != np->in_outip)
4264 					continue;
4265 				if (np->in_pmin &&
4266 				    ((ntohs(np->in_pmax) < ntohs(dport)) ||
4267 				     (ntohs(dport) < ntohs(np->in_pmin))))
4268 					continue;
4269 			}
4270 
4271 			if (*np->in_plabel != '\0') {
4272 				if (!appr_ok(fin, tcp, np)) {
4273 					continue;
4274 				}
4275 			}
4276 
4277 			ATOMIC_INC32(np->in_use);
4278 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
4279 			WRITE_ENTER(&ifs->ifs_ipf_nat);
4280 			nat = nat_new(fin, np, NULL, nflags, NAT_INBOUND);
4281 			if (nat != NULL) {
4282 				np->in_use--;
4283 				np->in_hits++;
4284 				MUTEX_DOWNGRADE(&ifs->ifs_ipf_nat);
4285 				break;
4286 			}
4287 			natfailed = -1;
4288 			npnext = np->in_rnext;
4289 			fr_ipnatderef(&np, ifs);
4290 			MUTEX_DOWNGRADE(&ifs->ifs_ipf_nat);
4291 		}
4292 
4293 		if ((np == NULL) && (rmsk != 0)) {
4294 			while (rmsk) {
4295 				msk <<= 1;
4296 				if (rmsk & 0x80000000)
4297 					break;
4298 				rmsk <<= 1;
4299 			}
4300 			if (rmsk != 0) {
4301 				rmsk <<= 1;
4302 				goto maskloop;
4303 			}
4304 		}
4305 	}
4306 	if (nat != NULL) {
4307 		rval = fr_natin(fin, nat, natadd, nflags);
4308 		if (rval == 1) {
4309 			MUTEX_ENTER(&nat->nat_lock);
4310 			nat->nat_ref++;
4311 			MUTEX_EXIT(&nat->nat_lock);
4312 			nat->nat_touched = ifs->ifs_fr_ticks;
4313 			fin->fin_nat = nat;
4314 			fin->fin_state = nat->nat_state;
4315 		}
4316 	} else
4317 		rval = natfailed;
4318 	RWLOCK_EXIT(&ifs->ifs_ipf_nat);
4319 
4320 	if (rval == -1) {
4321 		if (passp != NULL)
4322 			*passp = FR_BLOCK;
4323 		fin->fin_flx |= FI_BADNAT;
4324 	}
4325 	return rval;
4326 }
4327 
4328 
4329 /* ------------------------------------------------------------------------ */
4330 /* Function:    fr_natin                                                    */
4331 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
4332 /*                     1 == packet was successfully translated.             */
4333 /* Parameters:  fin(I)    - pointer to packet information                   */
4334 /*              nat(I)    - pointer to NAT structure                        */
4335 /*              natadd(I) - flag indicating if it is safe to add frag cache */
4336 /*              nflags(I) - NAT flags set for this packet                   */
4337 /* Locks Held:  ipf_nat (READ)                                              */
4338 /*                                                                          */
4339 /* Translate a packet coming "in" on an interface.                          */
4340 /* ------------------------------------------------------------------------ */
4341 int fr_natin(fin, nat, natadd, nflags)
4342 fr_info_t *fin;
4343 nat_t *nat;
4344 int natadd;
4345 u_32_t nflags;
4346 {
4347 	icmphdr_t *icmp;
4348 	u_short *csump;
4349 	tcphdr_t *tcp;
4350 	ipnat_t *np;
4351 	int i;
4352 	ipf_stack_t *ifs = fin->fin_ifs;
4353 
4354 	if (fin->fin_v == 6) {
4355 #ifdef	USE_INET6
4356 		return fr_nat6in(fin, nat, natadd, nflags);
4357 #else
4358 		return NULL;
4359 #endif
4360 	}
4361 
4362 #if SOLARIS && defined(_KERNEL)
4363 	net_data_t net_data_p = ifs->ifs_ipf_ipv4;
4364 #endif
4365 
4366 	tcp = NULL;
4367 	csump = NULL;
4368 	np = nat->nat_ptr;
4369 	fin->fin_fr = nat->nat_fr;
4370 
4371 	if ((natadd != 0) && (fin->fin_flx & FI_FRAG))
4372 		(void) fr_nat_newfrag(fin, 0, nat);
4373 
4374 	if (np != NULL) {
4375 
4376 	/* ------------------------------------------------------------- */
4377 	/* A few quick notes:						 */
4378 	/*	Following are test conditions prior to calling the 	 */
4379 	/*	appr_check routine.					 */
4380 	/*								 */
4381 	/* 	A NULL tcp indicates a non TCP/UDP packet.  When dealing */
4382 	/*	with a map rule, we attempt to match the packet's	 */
4383 	/*	source port against in_dport, otherwise	we'd compare the */
4384 	/*	packet's destination.			 		 */
4385 	/* ------------------------------------------------------------- */
4386 		if (np->in_apr != NULL) {
4387 			i = appr_check(fin, nat);
4388 			if (i == -1) {
4389 				return -1;
4390 			}
4391 		}
4392 	}
4393 
4394 #ifdef	IPFILTER_SYNC
4395 	ipfsync_update(SMC_NAT, fin, nat->nat_sync);
4396 #endif
4397 
4398 	MUTEX_ENTER(&nat->nat_lock);
4399 	nat->nat_bytes[0] += fin->fin_plen;
4400 	nat->nat_pkts[0]++;
4401 	MUTEX_EXIT(&nat->nat_lock);
4402 
4403 	fin->fin_ip->ip_dst = nat->nat_inip;
4404 	fin->fin_fi.fi_daddr = nat->nat_inip.s_addr;
4405 	if (nflags & IPN_TCPUDP)
4406 		tcp = fin->fin_dp;
4407 
4408 	/*
4409 	 * Fix up checksums, not by recalculating them, but
4410 	 * simply computing adjustments.
4411 	 * Why only do this for some platforms on inbound packets ?
4412 	 * Because for those that it is done, IP processing is yet to happen
4413 	 * and so the IPv4 header checksum has not yet been evaluated.
4414 	 * Perhaps it should always be done for the benefit of things like
4415 	 * fast forwarding (so that it doesn't need to be recomputed) but with
4416 	 * header checksum offloading, perhaps it is a moot point.
4417 	 */
4418 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
4419      defined(__osf__) || defined(linux)
4420 	if (nat->nat_dir == NAT_OUTBOUND)
4421 		fix_incksum(&fin->fin_ip->ip_sum, nat->nat_ipsumd);
4422 	else
4423 		fix_outcksum(&fin->fin_ip->ip_sum, nat->nat_ipsumd);
4424 #endif
4425 
4426 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
4427 		if ((nat->nat_inport != 0) && (nflags & IPN_TCPUDP)) {
4428 			tcp->th_dport = nat->nat_inport;
4429 			fin->fin_data[1] = ntohs(nat->nat_inport);
4430 		}
4431 
4432 
4433 		if ((nat->nat_inport != 0) && (nflags & IPN_ICMPQUERY)) {
4434 			icmp = fin->fin_dp;
4435 
4436 			icmp->icmp_id = nat->nat_inport;
4437 		}
4438 
4439 		csump = nat_proto(fin, nat, nflags);
4440 	}
4441 
4442 	nat_update(fin, nat, np);
4443 
4444 	/*
4445 	 * In case they are being forwarded, inbound packets always need to have
4446 	 * their checksum adjusted even if hardware checksum validation said OK.
4447 	 */
4448 	if (csump != NULL) {
4449 		if (nat->nat_dir == NAT_OUTBOUND)
4450 			fix_incksum(csump, nat->nat_sumd[0]);
4451 		else
4452 			fix_outcksum(csump, nat->nat_sumd[0]);
4453 	}
4454 
4455 #if SOLARIS && defined(_KERNEL)
4456 	if (nflags & IPN_TCPUDP &&
4457 	    NET_IS_HCK_L4_PART(net_data_p, fin->fin_m)) {
4458 		/*
4459 		 * Need to adjust the partial checksum result stored in
4460 		 * db_cksum16, which will be used for validation in IP.
4461 		 * See IP_CKSUM_RECV().
4462 		 * Adjustment data should be the inverse of the IP address
4463 		 * changes, because db_cksum16 is supposed to be the complement
4464 		 * of the pesudo header.
4465 		 */
4466 		csump = &fin->fin_m->b_datap->db_cksum16;
4467 		if (nat->nat_dir == NAT_OUTBOUND)
4468 			fix_outcksum(csump, nat->nat_sumd[1]);
4469 		else
4470 			fix_incksum(csump, nat->nat_sumd[1]);
4471 	}
4472 #endif
4473 
4474 	ifs->ifs_nat_stats.ns_mapped[0]++;
4475 	fin->fin_flx |= FI_NATED;
4476 	if (np != NULL && np->in_tag.ipt_num[0] != 0)
4477 		fin->fin_nattag = &np->in_tag;
4478 	return 1;
4479 }
4480 
4481 
4482 /* ------------------------------------------------------------------------ */
4483 /* Function:    nat_proto                                                   */
4484 /* Returns:     u_short* - pointer to transport header checksum to update,  */
4485 /*                         NULL if the transport protocol is not recognised */
4486 /*                         as needing a checksum update.                    */
4487 /* Parameters:  fin(I)    - pointer to packet information                   */
4488 /*              nat(I)    - pointer to NAT structure                        */
4489 /*              nflags(I) - NAT flags set for this packet                   */
4490 /*                                                                          */
4491 /* Return the pointer to the checksum field for each protocol so understood.*/
4492 /* If support for making other changes to a protocol header is required,    */
4493 /* that is not strictly 'address' translation, such as clamping the MSS in  */
4494 /* TCP down to a specific value, then do it from here.                      */
4495 /* ------------------------------------------------------------------------ */
4496 u_short *nat_proto(fin, nat, nflags)
4497 fr_info_t *fin;
4498 nat_t *nat;
4499 u_int nflags;
4500 {
4501 	icmphdr_t *icmp;
4502 	struct icmp6_hdr *icmp6;
4503 	u_short *csump;
4504 	tcphdr_t *tcp;
4505 	udphdr_t *udp;
4506 
4507 	csump = NULL;
4508 	if (fin->fin_out == 0) {
4509 		fin->fin_rev = (nat->nat_dir == NAT_OUTBOUND);
4510 	} else {
4511 		fin->fin_rev = (nat->nat_dir == NAT_INBOUND);
4512 	}
4513 
4514 	switch (fin->fin_p)
4515 	{
4516 	case IPPROTO_TCP :
4517 		tcp = fin->fin_dp;
4518 
4519 		csump = &tcp->th_sum;
4520 
4521 		/*
4522 		 * Do a MSS CLAMPING on a SYN packet,
4523 		 * only deal IPv4 for now.
4524 		 */
4525 		if ((nat->nat_mssclamp != 0) && (tcp->th_flags & TH_SYN) != 0)
4526 			nat_mssclamp(tcp, nat->nat_mssclamp, csump);
4527 
4528 		break;
4529 
4530 	case IPPROTO_UDP :
4531 		udp = fin->fin_dp;
4532 
4533 		if (udp->uh_sum)
4534 			csump = &udp->uh_sum;
4535 		break;
4536 
4537 	case IPPROTO_ICMP :
4538 		icmp = fin->fin_dp;
4539 
4540 		if ((nflags & IPN_ICMPQUERY) != 0) {
4541 			if (icmp->icmp_cksum != 0)
4542 				csump = &icmp->icmp_cksum;
4543 		}
4544 		break;
4545 
4546 	case IPPROTO_ICMPV6 :
4547 		icmp6 = fin->fin_dp;
4548 
4549 		if ((nflags & IPN_ICMPQUERY) != 0) {
4550 			if (icmp6->icmp6_cksum != 0)
4551 				csump = &icmp6->icmp6_cksum;
4552 		}
4553 		break;
4554 	}
4555 	return csump;
4556 }
4557 
4558 
4559 /* ------------------------------------------------------------------------ */
4560 /* Function:    fr_natunload                                                */
4561 /* Returns:     Nil                                                         */
4562 /* Parameters:  Nil                                                         */
4563 /*                                                                          */
4564 /* Free all memory used by NAT structures allocated at runtime.             */
4565 /* ------------------------------------------------------------------------ */
4566 void fr_natunload(ifs)
4567 ipf_stack_t *ifs;
4568 {
4569 	ipftq_t *ifq, *ifqnext;
4570 
4571 	(void) nat_clearlist(ifs);
4572 	(void) nat_flushtable(ifs);
4573 
4574 	/*
4575 	 * Proxy timeout queues are not cleaned here because although they
4576 	 * exist on the NAT list, appr_unload is called after fr_natunload
4577 	 * and the proxies actually are responsible for them being created.
4578 	 * Should the proxy timeouts have their own list?  There's no real
4579 	 * justification as this is the only complication.
4580 	 */
4581 	for (ifq = ifs->ifs_nat_utqe; ifq != NULL; ifq = ifqnext) {
4582 		ifqnext = ifq->ifq_next;
4583 		if (((ifq->ifq_flags & IFQF_PROXY) == 0) &&
4584 		    (fr_deletetimeoutqueue(ifq) == 0))
4585 			fr_freetimeoutqueue(ifq, ifs);
4586 	}
4587 
4588 	if (ifs->ifs_nat_table[0] != NULL) {
4589 		KFREES(ifs->ifs_nat_table[0],
4590 		       sizeof(nat_t *) * ifs->ifs_ipf_nattable_sz);
4591 		ifs->ifs_nat_table[0] = NULL;
4592 	}
4593 	if (ifs->ifs_nat_table[1] != NULL) {
4594 		KFREES(ifs->ifs_nat_table[1],
4595 		       sizeof(nat_t *) * ifs->ifs_ipf_nattable_sz);
4596 		ifs->ifs_nat_table[1] = NULL;
4597 	}
4598 	if (ifs->ifs_nat_rules != NULL) {
4599 		KFREES(ifs->ifs_nat_rules,
4600 		       sizeof(ipnat_t *) * ifs->ifs_ipf_natrules_sz);
4601 		ifs->ifs_nat_rules = NULL;
4602 	}
4603 	if (ifs->ifs_rdr_rules != NULL) {
4604 		KFREES(ifs->ifs_rdr_rules,
4605 		       sizeof(ipnat_t *) * ifs->ifs_ipf_rdrrules_sz);
4606 		ifs->ifs_rdr_rules = NULL;
4607 	}
4608 	if (ifs->ifs_maptable != NULL) {
4609 		KFREES(ifs->ifs_maptable,
4610 		       sizeof(hostmap_t *) * ifs->ifs_ipf_hostmap_sz);
4611 		ifs->ifs_maptable = NULL;
4612 	}
4613 	if (ifs->ifs_nat_stats.ns_bucketlen[0] != NULL) {
4614 		KFREES(ifs->ifs_nat_stats.ns_bucketlen[0],
4615 		       sizeof(u_long *) * ifs->ifs_ipf_nattable_sz);
4616 		ifs->ifs_nat_stats.ns_bucketlen[0] = NULL;
4617 	}
4618 	if (ifs->ifs_nat_stats.ns_bucketlen[1] != NULL) {
4619 		KFREES(ifs->ifs_nat_stats.ns_bucketlen[1],
4620 		       sizeof(u_long *) * ifs->ifs_ipf_nattable_sz);
4621 		ifs->ifs_nat_stats.ns_bucketlen[1] = NULL;
4622 	}
4623 
4624 	if (ifs->ifs_fr_nat_maxbucket_reset == 1)
4625 		ifs->ifs_fr_nat_maxbucket = 0;
4626 
4627 	if (ifs->ifs_fr_nat_init == 1) {
4628 		ifs->ifs_fr_nat_init = 0;
4629 		fr_sttab_destroy(ifs->ifs_nat_tqb);
4630 
4631 		RW_DESTROY(&ifs->ifs_ipf_natfrag);
4632 		RW_DESTROY(&ifs->ifs_ipf_nat);
4633 
4634 		MUTEX_DESTROY(&ifs->ifs_ipf_nat_new);
4635 		MUTEX_DESTROY(&ifs->ifs_ipf_natio);
4636 
4637 		MUTEX_DESTROY(&ifs->ifs_nat_udptq.ifq_lock);
4638 		MUTEX_DESTROY(&ifs->ifs_nat_icmptq.ifq_lock);
4639 		MUTEX_DESTROY(&ifs->ifs_nat_iptq.ifq_lock);
4640 	}
4641 }
4642 
4643 
4644 /* ------------------------------------------------------------------------ */
4645 /* Function:    fr_natexpire                                                */
4646 /* Returns:     Nil                                                         */
4647 /* Parameters:  Nil                                                         */
4648 /*                                                                          */
4649 /* Check all of the timeout queues for entries at the top which need to be  */
4650 /* expired.                                                                 */
4651 /* ------------------------------------------------------------------------ */
4652 void fr_natexpire(ifs)
4653 ipf_stack_t *ifs;
4654 {
4655 	ipftq_t *ifq, *ifqnext;
4656 	ipftqent_t *tqe, *tqn;
4657 	int i;
4658 	SPL_INT(s);
4659 
4660 	SPL_NET(s);
4661 	WRITE_ENTER(&ifs->ifs_ipf_nat);
4662 	for (ifq = ifs->ifs_nat_tqb, i = 0; ifq != NULL; ifq = ifq->ifq_next) {
4663 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
4664 			if (tqe->tqe_die > ifs->ifs_fr_ticks)
4665 				break;
4666 			tqn = tqe->tqe_next;
4667 			nat_delete(tqe->tqe_parent, NL_EXPIRE, ifs);
4668 		}
4669 	}
4670 
4671 	for (ifq = ifs->ifs_nat_utqe; ifq != NULL; ifq = ifqnext) {
4672 		ifqnext = ifq->ifq_next;
4673 
4674 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
4675 			if (tqe->tqe_die > ifs->ifs_fr_ticks)
4676 				break;
4677 			tqn = tqe->tqe_next;
4678 			nat_delete(tqe->tqe_parent, NL_EXPIRE, ifs);
4679 		}
4680 	}
4681 
4682 	for (ifq = ifs->ifs_nat_utqe; ifq != NULL; ifq = ifqnext) {
4683 		ifqnext = ifq->ifq_next;
4684 
4685 		if (((ifq->ifq_flags & IFQF_DELETE) != 0) &&
4686 		    (ifq->ifq_ref == 0)) {
4687 			fr_freetimeoutqueue(ifq, ifs);
4688 		}
4689 	}
4690 
4691 	if (ifs->ifs_nat_doflush != 0) {
4692 		(void) nat_extraflush(2, ifs);
4693 		ifs->ifs_nat_doflush = 0;
4694 	}
4695 
4696 	RWLOCK_EXIT(&ifs->ifs_ipf_nat);
4697 	SPL_X(s);
4698 }
4699 
4700 
4701 /* ------------------------------------------------------------------------ */
4702 /* Function:    fr_nataddrsync                                              */
4703 /* Returns:     Nil                                                         */
4704 /* Parameters:  ifp(I) -  pointer to network interface                      */
4705 /*              addr(I) - pointer to new network address                    */
4706 /*                                                                          */
4707 /* Walk through all of the currently active NAT sessions, looking for those */
4708 /* which need to have their translated address updated (where the interface */
4709 /* matches the one passed in) and change it, recalculating the checksum sum */
4710 /* difference too.                                                          */
4711 /* ------------------------------------------------------------------------ */
4712 void fr_nataddrsync(v, ifp, addr, ifs)
4713 int v;
4714 void *ifp;
4715 void *addr;
4716 ipf_stack_t *ifs;
4717 {
4718 	u_32_t sum1, sum2, sumd;
4719 	nat_t *nat;
4720 	ipnat_t *np;
4721 	SPL_INT(s);
4722 
4723 	if (ifs->ifs_fr_running <= 0)
4724 		return;
4725 
4726 	SPL_NET(s);
4727 	WRITE_ENTER(&ifs->ifs_ipf_nat);
4728 
4729 	if (ifs->ifs_fr_running <= 0) {
4730 		RWLOCK_EXIT(&ifs->ifs_ipf_nat);
4731 		return;
4732 	}
4733 
4734 	/*
4735 	 * Change IP addresses for NAT sessions for any protocol except TCP
4736 	 * since it will break the TCP connection anyway.  The only rules
4737 	 * which will get changed are those which are "map ... -> 0/32",
4738 	 * where the rule specifies the address is taken from the interface.
4739 	 */
4740 	for (nat = ifs->ifs_nat_instances; nat; nat = nat->nat_next) {
4741 		if (addr != NULL) {
4742 			if (((ifp != NULL) && ifp != (nat->nat_ifps[0])) ||
4743 			    ((nat->nat_flags & IPN_TCP) != 0))
4744 				continue;
4745 			if ((np = nat->nat_ptr) == NULL)
4746 				continue;
4747 			if (v == 4 && np->in_v == 4) {
4748 				if (np->in_nip || np->in_outmsk != 0xffffffff)
4749 					continue;
4750 				/*
4751 				 * Change the map-to address to be the same as
4752 				 * the new one.
4753 				 */
4754 				sum1 = nat->nat_outip.s_addr;
4755 				nat->nat_outip = *(struct in_addr *)addr;
4756 				sum2 = nat->nat_outip.s_addr;
4757 			} else if (v == 6 && np->in_v == 6) {
4758 				if (!IP6_ISZERO(&np->in_next6.in6) ||
4759 				    !IP6_ISONES(&np->in_out[1].in6))
4760 					continue;
4761 				/*
4762 				 * Change the map-to address to be the same as
4763 				 * the new one.
4764 				 */
4765 				nat->nat_outip6.in6 = *(struct in6_addr *)addr;
4766 			} else
4767 				continue;
4768 
4769 		} else if (((ifp == NULL) || (ifp == nat->nat_ifps[0])) &&
4770 		    !(nat->nat_flags & IPN_TCP) && (np = nat->nat_ptr)) {
4771 			if (np->in_v == 4 && (v == 4 || v == 0)) {
4772 				struct in_addr in;
4773 				if (np->in_outmsk != 0xffffffff || np->in_nip)
4774 					continue;
4775 				/*
4776 				 * Change the map-to address to be the same as
4777 				 * the new one.
4778 				 */
4779 				sum1 = nat->nat_outip.s_addr;
4780 				if (fr_ifpaddr(4, FRI_NORMAL, nat->nat_ifps[0],
4781 					       &in, NULL, ifs) != -1)
4782 					nat->nat_outip = in;
4783 				sum2 = nat->nat_outip.s_addr;
4784 			} else if (np->in_v == 6 && (v == 6 || v == 0)) {
4785 				struct in6_addr in6;
4786 				if (!IP6_ISZERO(&np->in_next6.in6) ||
4787 				    !IP6_ISONES(&np->in_out[1].in6))
4788 					continue;
4789 				/*
4790 				 * Change the map-to address to be the same as
4791 				 * the new one.
4792 				 */
4793 				if (fr_ifpaddr(6, FRI_NORMAL, nat->nat_ifps[0],
4794 					       (void *)&in6, NULL, ifs) != -1)
4795 					nat->nat_outip6.in6 = in6;
4796 			} else
4797 				continue;
4798 		} else {
4799 			continue;
4800 		}
4801 
4802 		if (sum1 == sum2)
4803 			continue;
4804 		/*
4805 		 * Readjust the checksum adjustment to take into
4806 		 * account the new IP#.
4807 		 */
4808 		CALC_SUMD(sum1, sum2, sumd);
4809 		/* XXX - dont change for TCP when solaris does
4810 		 * hardware checksumming.
4811 		 */
4812 		sumd += nat->nat_sumd[0];
4813 		nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
4814 		nat->nat_sumd[1] = nat->nat_sumd[0];
4815 	}
4816 
4817 	RWLOCK_EXIT(&ifs->ifs_ipf_nat);
4818 	SPL_X(s);
4819 }
4820 
4821 
4822 /* ------------------------------------------------------------------------ */
4823 /* Function:    fr_natifpsync                                               */
4824 /* Returns:     Nil                                                         */
4825 /* Parameters:  action(I) - how we are syncing                              */
4826 /*              ifp(I)    - pointer to network interface                    */
4827 /*              name(I)   - name of interface to sync to                    */
4828 /*                                                                          */
4829 /* This function is used to resync the mapping of interface names and their */
4830 /* respective 'pointers'.  For "action == IPFSYNC_RESYNC", resync all       */
4831 /* interfaces by doing a new lookup of name to 'pointer'.  For "action ==   */
4832 /* IPFSYNC_NEWIFP", treat ifp as the new pointer value associated with      */
4833 /* "name" and for "action == IPFSYNC_OLDIFP", ifp is a pointer for which    */
4834 /* there is no longer any interface associated with it.                     */
4835 /* ------------------------------------------------------------------------ */
4836 void fr_natifpsync(action, v, ifp, name, ifs)
4837 int action, v;
4838 void *ifp;
4839 char *name;
4840 ipf_stack_t *ifs;
4841 {
4842 #if defined(_KERNEL) && !defined(MENTAT) && defined(USE_SPL)
4843 	int s;
4844 #endif
4845 	nat_t *nat;
4846 	ipnat_t *n;
4847 	int nv;
4848 
4849 	if (ifs->ifs_fr_running <= 0)
4850 		return;
4851 
4852 	SPL_NET(s);
4853 	WRITE_ENTER(&ifs->ifs_ipf_nat);
4854 
4855 	if (ifs->ifs_fr_running <= 0) {
4856 		RWLOCK_EXIT(&ifs->ifs_ipf_nat);
4857 		return;
4858 	}
4859 
4860 	switch (action)
4861 	{
4862 	case IPFSYNC_RESYNC :
4863 		for (nat = ifs->ifs_nat_instances; nat; nat = nat->nat_next) {
4864 			nv = (v == 0) ? nat->nat_v : v;
4865 			if (nat->nat_v != nv)
4866 				continue;
4867 			if ((ifp == nat->nat_ifps[0]) ||
4868 			    (nat->nat_ifps[0] == (void *)-1)) {
4869 				nat->nat_ifps[0] =
4870 				    fr_resolvenic(nat->nat_ifnames[0], nv, ifs);
4871 			}
4872 
4873 			if ((ifp == nat->nat_ifps[1]) ||
4874 			    (nat->nat_ifps[1] == (void *)-1)) {
4875 				nat->nat_ifps[1] =
4876 				    fr_resolvenic(nat->nat_ifnames[1], nv, ifs);
4877 			}
4878 		}
4879 
4880 		for (n = ifs->ifs_nat_list; (n != NULL); n = n->in_next) {
4881 			nv = (v == 0) ? (int)n->in_v : v;
4882 			if ((int)n->in_v != nv)
4883 				continue;
4884 			if (n->in_ifps[0] == ifp ||
4885 			    n->in_ifps[0] == (void *)-1) {
4886 				n->in_ifps[0] =
4887 				    fr_resolvenic(n->in_ifnames[0], nv, ifs);
4888 			}
4889 			if (n->in_ifps[1] == ifp ||
4890 			    n->in_ifps[1] == (void *)-1) {
4891 				n->in_ifps[1] =
4892 				    fr_resolvenic(n->in_ifnames[1], nv, ifs);
4893 			}
4894 		}
4895 		break;
4896 	case IPFSYNC_NEWIFP :
4897 		for (nat = ifs->ifs_nat_instances; nat; nat = nat->nat_next) {
4898 			if (nat->nat_v != v)
4899 				continue;
4900 			if (!strncmp(name, nat->nat_ifnames[0],
4901 				     sizeof(nat->nat_ifnames[0])))
4902 				nat->nat_ifps[0] = ifp;
4903 			if (!strncmp(name, nat->nat_ifnames[1],
4904 				     sizeof(nat->nat_ifnames[1])))
4905 				nat->nat_ifps[1] = ifp;
4906 		}
4907 		for (n = ifs->ifs_nat_list; (n != NULL); n = n->in_next) {
4908 			if ((int)n->in_v != v)
4909 				continue;
4910 			if (!strncmp(name, n->in_ifnames[0],
4911 				     sizeof(n->in_ifnames[0])))
4912 				n->in_ifps[0] = ifp;
4913 			if (!strncmp(name, n->in_ifnames[1],
4914 				     sizeof(n->in_ifnames[1])))
4915 				n->in_ifps[1] = ifp;
4916 		}
4917 		break;
4918 	case IPFSYNC_OLDIFP :
4919 		for (nat = ifs->ifs_nat_instances; nat; nat = nat->nat_next) {
4920 			if (nat->nat_v != v)
4921 				continue;
4922 			if (ifp == nat->nat_ifps[0])
4923 				nat->nat_ifps[0] = (void *)-1;
4924 			if (ifp == nat->nat_ifps[1])
4925 				nat->nat_ifps[1] = (void *)-1;
4926 		}
4927 		for (n = ifs->ifs_nat_list; (n != NULL); n = n->in_next) {
4928 			if ((int)n->in_v != v)
4929 				continue;
4930 			if (n->in_ifps[0] == ifp)
4931 				n->in_ifps[0] = (void *)-1;
4932 			if (n->in_ifps[1] == ifp)
4933 				n->in_ifps[1] = (void *)-1;
4934 		}
4935 		break;
4936 	}
4937 	RWLOCK_EXIT(&ifs->ifs_ipf_nat);
4938 	SPL_X(s);
4939 }
4940 
4941 
4942 /* ------------------------------------------------------------------------ */
4943 /* Function:    nat_icmpquerytype4                                          */
4944 /* Returns:     int - 1 == success, 0 == failure                            */
4945 /* Parameters:  icmptype(I) - ICMP type number                              */
4946 /*                                                                          */
4947 /* Tests to see if the ICMP type number passed is a query/response type or  */
4948 /* not.                                                                     */
4949 /* ------------------------------------------------------------------------ */
4950 static INLINE int nat_icmpquerytype4(icmptype)
4951 int icmptype;
4952 {
4953 
4954 	/*
4955 	 * For the ICMP query NAT code, it is essential that both the query
4956 	 * and the reply match on the NAT rule. Because the NAT structure
4957 	 * does not keep track of the icmptype, and a single NAT structure
4958 	 * is used for all icmp types with the same src, dest and id, we
4959 	 * simply define the replies as queries as well. The funny thing is,
4960 	 * altough it seems silly to call a reply a query, this is exactly
4961 	 * as it is defined in the IPv4 specification
4962 	 */
4963 
4964 	switch (icmptype)
4965 	{
4966 
4967 	case ICMP_ECHOREPLY:
4968 	case ICMP_ECHO:
4969 	/* route aedvertisement/solliciation is currently unsupported: */
4970 	/* it would require rewriting the ICMP data section            */
4971 	case ICMP_TSTAMP:
4972 	case ICMP_TSTAMPREPLY:
4973 	case ICMP_IREQ:
4974 	case ICMP_IREQREPLY:
4975 	case ICMP_MASKREQ:
4976 	case ICMP_MASKREPLY:
4977 		return 1;
4978 	default:
4979 		return 0;
4980 	}
4981 }
4982 
4983 
4984 /* ------------------------------------------------------------------------ */
4985 /* Function:    nat_log                                                     */
4986 /* Returns:     Nil                                                         */
4987 /* Parameters:  nat(I)  - pointer to NAT structure                          */
4988 /*              type(I) - type of log entry to create                       */
4989 /*                                                                          */
4990 /* Creates a NAT log entry.                                                 */
4991 /* ------------------------------------------------------------------------ */
4992 void nat_log(nat, type, ifs)
4993 struct nat *nat;
4994 u_int type;
4995 ipf_stack_t *ifs;
4996 {
4997 #ifdef	IPFILTER_LOG
4998 # ifndef LARGE_NAT
4999 	struct ipnat *np;
5000 	int rulen;
5001 # endif
5002 	struct natlog natl;
5003 	void *items[1];
5004 	size_t sizes[1];
5005 	int types[1];
5006 
5007 	natl.nlg_inip = nat->nat_inip6;
5008 	natl.nlg_outip = nat->nat_outip6;
5009 	natl.nlg_origip = nat->nat_oip6;
5010 	natl.nlg_bytes[0] = nat->nat_bytes[0];
5011 	natl.nlg_bytes[1] = nat->nat_bytes[1];
5012 	natl.nlg_pkts[0] = nat->nat_pkts[0];
5013 	natl.nlg_pkts[1] = nat->nat_pkts[1];
5014 	natl.nlg_origport = nat->nat_oport;
5015 	natl.nlg_inport = nat->nat_inport;
5016 	natl.nlg_outport = nat->nat_outport;
5017 	natl.nlg_p = nat->nat_p;
5018 	natl.nlg_type = type;
5019 	natl.nlg_rule = -1;
5020 	natl.nlg_v = nat->nat_v;
5021 # ifndef LARGE_NAT
5022 	if (nat->nat_ptr != NULL) {
5023 		for (rulen = 0, np = ifs->ifs_nat_list; np;
5024 		     np = np->in_next, rulen++)
5025 			if (np == nat->nat_ptr) {
5026 				natl.nlg_rule = rulen;
5027 				break;
5028 			}
5029 	}
5030 # endif
5031 	items[0] = &natl;
5032 	sizes[0] = sizeof(natl);
5033 	types[0] = 0;
5034 
5035 	(void) ipllog(IPL_LOGNAT, NULL, items, sizes, types, 1, ifs);
5036 #endif
5037 }
5038 
5039 
5040 #if defined(__OpenBSD__)
5041 /* ------------------------------------------------------------------------ */
5042 /* Function:    nat_ifdetach                                                */
5043 /* Returns:     Nil                                                         */
5044 /* Parameters:  ifp(I) - pointer to network interface                       */
5045 /*                                                                          */
5046 /* Compatibility interface for OpenBSD to trigger the correct updating of   */
5047 /* interface references within IPFilter.                                    */
5048 /* ------------------------------------------------------------------------ */
5049 void nat_ifdetach(ifp, ifs)
5050 void *ifp;
5051 ipf_stack_t *ifs;
5052 {
5053 	frsync(ifp, ifs);
5054 	return;
5055 }
5056 #endif
5057 
5058 
5059 /* ------------------------------------------------------------------------ */
5060 /* Function:    fr_ipnatderef                                               */
5061 /* Returns:     Nil                                                         */
5062 /* Parameters:  inp(I) - pointer to pointer to NAT rule                     */
5063 /* Write Locks: ipf_nat                                                     */
5064 /*                                                                          */
5065 /* ------------------------------------------------------------------------ */
5066 void fr_ipnatderef(inp, ifs)
5067 ipnat_t **inp;
5068 ipf_stack_t *ifs;
5069 {
5070 	ipnat_t *in;
5071 
5072 	in = *inp;
5073 	*inp = NULL;
5074 	in->in_use--;
5075 	if (in->in_use == 0 && (in->in_flags & IPN_DELETE)) {
5076 		if (in->in_apr)
5077 			appr_free(in->in_apr);
5078 		KFREE(in);
5079 		ifs->ifs_nat_stats.ns_rules--;
5080 #ifdef notdef
5081 #if SOLARIS
5082 		if (ifs->ifs_nat_stats.ns_rules == 0)
5083 			ifs->ifs_pfil_delayed_copy = 1;
5084 #endif
5085 #endif
5086 	}
5087 }
5088 
5089 
5090 /* ------------------------------------------------------------------------ */
5091 /* Function:    fr_natderef                                                 */
5092 /* Returns:     Nil                                                         */
5093 /* Parameters:  isp(I) - pointer to pointer to NAT table entry              */
5094 /*                                                                          */
5095 /* Decrement the reference counter for this NAT table entry and free it if  */
5096 /* there are no more things using it.                                       */
5097 /*                                                                          */
5098 /* IF nat_ref == 1 when this function is called, then we have an orphan nat */
5099 /* structure *because* it only gets called on paths _after_ nat_ref has been*/
5100 /* incremented.  If nat_ref == 1 then we shouldn't decrement it here        */
5101 /* because nat_delete() will do that and send nat_ref to -1.                */
5102 /*                                                                          */
5103 /* Holding the lock on nat_lock is required to serialise nat_delete() being */
5104 /* called from a NAT flush ioctl with a deref happening because of a packet.*/
5105 /* ------------------------------------------------------------------------ */
5106 void fr_natderef(natp, ifs)
5107 nat_t **natp;
5108 ipf_stack_t *ifs;
5109 {
5110 	nat_t *nat;
5111 
5112 	nat = *natp;
5113 	*natp = NULL;
5114 
5115 	MUTEX_ENTER(&nat->nat_lock);
5116 	if (nat->nat_ref > 1) {
5117 		nat->nat_ref--;
5118 		MUTEX_EXIT(&nat->nat_lock);
5119 		return;
5120 	}
5121 	MUTEX_EXIT(&nat->nat_lock);
5122 
5123 	WRITE_ENTER(&ifs->ifs_ipf_nat);
5124 	nat_delete(nat, NL_EXPIRE, ifs);
5125 	RWLOCK_EXIT(&ifs->ifs_ipf_nat);
5126 }
5127 
5128 
5129 /* ------------------------------------------------------------------------ */
5130 /* Function:    fr_natclone                                                 */
5131 /* Returns:     ipstate_t* - NULL == cloning failed,                        */
5132 /*                           else pointer to new state structure            */
5133 /* Parameters:  fin(I) - pointer to packet information                      */
5134 /*              is(I)  - pointer to master state structure                  */
5135 /* Write Lock:  ipf_nat                                                     */
5136 /*                                                                          */
5137 /* Create a "duplcate" state table entry from the master.                   */
5138 /* ------------------------------------------------------------------------ */
5139 nat_t *fr_natclone(fin, nat)
5140 fr_info_t *fin;
5141 nat_t *nat;
5142 {
5143 	frentry_t *fr;
5144 	nat_t *clone;
5145 	ipnat_t *np;
5146 	ipf_stack_t *ifs = fin->fin_ifs;
5147 
5148 	KMALLOC(clone, nat_t *);
5149 	if (clone == NULL)
5150 		return NULL;
5151 	bcopy((char *)nat, (char *)clone, sizeof(*clone));
5152 
5153 	MUTEX_NUKE(&clone->nat_lock);
5154 
5155 	clone->nat_aps = NULL;
5156 	/*
5157 	 * Initialize all these so that nat_delete() doesn't cause a crash.
5158 	 */
5159 	clone->nat_tqe.tqe_pnext = NULL;
5160 	clone->nat_tqe.tqe_next = NULL;
5161 	clone->nat_tqe.tqe_ifq = NULL;
5162 	clone->nat_tqe.tqe_parent = clone;
5163 
5164 	clone->nat_flags &= ~SI_CLONE;
5165 	clone->nat_flags |= SI_CLONED;
5166 
5167 	if (clone->nat_hm)
5168 		clone->nat_hm->hm_ref++;
5169 
5170 	if (nat_insert(clone, fin->fin_rev, ifs) == -1) {
5171 		KFREE(clone);
5172 		return NULL;
5173 	}
5174 	np = clone->nat_ptr;
5175 	if (np != NULL) {
5176 		if (ifs->ifs_nat_logging)
5177 			nat_log(clone, (u_int)np->in_redir, ifs);
5178 		np->in_use++;
5179 	}
5180 	fr = clone->nat_fr;
5181 	if (fr != NULL) {
5182 		MUTEX_ENTER(&fr->fr_lock);
5183 		fr->fr_ref++;
5184 		MUTEX_EXIT(&fr->fr_lock);
5185 	}
5186 
5187 	/*
5188 	 * Because the clone is created outside the normal loop of things and
5189 	 * TCP has special needs in terms of state, initialise the timeout
5190 	 * state of the new NAT from here.
5191 	 */
5192 	if (clone->nat_p == IPPROTO_TCP) {
5193 		(void) fr_tcp_age(&clone->nat_tqe, fin, ifs->ifs_nat_tqb,
5194 				  clone->nat_flags);
5195 	}
5196 #ifdef	IPFILTER_SYNC
5197 	clone->nat_sync = ipfsync_new(SMC_NAT, fin, clone);
5198 #endif
5199 	if (ifs->ifs_nat_logging)
5200 		nat_log(clone, NL_CLONE, ifs);
5201 	return clone;
5202 }
5203 
5204 
5205 /* ------------------------------------------------------------------------ */
5206 /* Function:   nat_wildok                                                   */
5207 /* Returns:    int - 1 == packet's ports match wildcards                    */
5208 /*                   0 == packet's ports don't match wildcards              */
5209 /* Parameters: nat(I)   - NAT entry                                         */
5210 /*             sport(I) - source port                                       */
5211 /*             dport(I) - destination port                                  */
5212 /*             flags(I) - wildcard flags                                    */
5213 /*             dir(I)   - packet direction                                  */
5214 /*                                                                          */
5215 /* Use NAT entry and packet direction to determine which combination of     */
5216 /* wildcard flags should be used.                                           */
5217 /* ------------------------------------------------------------------------ */
5218 int nat_wildok(nat, sport, dport, flags, dir)
5219 nat_t *nat;
5220 int sport;
5221 int dport;
5222 int flags;
5223 int dir;
5224 {
5225 	/*
5226 	 * When called by       dir is set to
5227 	 * nat_inlookup         NAT_INBOUND (0)
5228 	 * nat_outlookup        NAT_OUTBOUND (1)
5229 	 *
5230 	 * We simply combine the packet's direction in dir with the original
5231 	 * "intended" direction of that NAT entry in nat->nat_dir to decide
5232 	 * which combination of wildcard flags to allow.
5233 	 */
5234 
5235 	switch ((dir << 1) | nat->nat_dir)
5236 	{
5237 	case 3: /* outbound packet / outbound entry */
5238 		if (((nat->nat_inport == sport) ||
5239 		    (flags & SI_W_SPORT)) &&
5240 		    ((nat->nat_oport == dport) ||
5241 		    (flags & SI_W_DPORT)))
5242 			return 1;
5243 		break;
5244 	case 2: /* outbound packet / inbound entry */
5245 		if (((nat->nat_outport == sport) ||
5246 		    (flags & SI_W_DPORT)) &&
5247 		    ((nat->nat_oport == dport) ||
5248 		    (flags & SI_W_SPORT)))
5249 			return 1;
5250 		break;
5251 	case 1: /* inbound packet / outbound entry */
5252 		if (((nat->nat_oport == sport) ||
5253 		    (flags & SI_W_DPORT)) &&
5254 		    ((nat->nat_outport == dport) ||
5255 		    (flags & SI_W_SPORT)))
5256 			return 1;
5257 		break;
5258 	case 0: /* inbound packet / inbound entry */
5259 		if (((nat->nat_oport == sport) ||
5260 		    (flags & SI_W_SPORT)) &&
5261 		    ((nat->nat_outport == dport) ||
5262 		    (flags & SI_W_DPORT)))
5263 			return 1;
5264 		break;
5265 	default:
5266 		break;
5267 	}
5268 
5269 	return(0);
5270 }
5271 
5272 
5273 /* ------------------------------------------------------------------------ */
5274 /* Function:    nat_mssclamp                                                */
5275 /* Returns:     Nil                                                         */
5276 /* Parameters:  tcp(I)    - pointer to TCP header                           */
5277 /*              maxmss(I) - value to clamp the TCP MSS to                   */
5278 /*              csump(I)  - pointer to TCP checksum                         */
5279 /*                                                                          */
5280 /* Check for MSS option and clamp it if necessary.  If found and changed,   */
5281 /* then the TCP header checksum will be updated to reflect the change in    */
5282 /* the MSS.                                                                 */
5283 /* ------------------------------------------------------------------------ */
5284 static void nat_mssclamp(tcp, maxmss, csump)
5285 tcphdr_t *tcp;
5286 u_32_t maxmss;
5287 u_short *csump;
5288 {
5289 	u_char *cp, *ep, opt;
5290 	int hlen, advance;
5291 	u_32_t mss, sumd;
5292 
5293 	hlen = TCP_OFF(tcp) << 2;
5294 	if (hlen > sizeof(*tcp)) {
5295 		cp = (u_char *)tcp + sizeof(*tcp);
5296 		ep = (u_char *)tcp + hlen;
5297 
5298 		while (cp < ep) {
5299 			opt = cp[0];
5300 			if (opt == TCPOPT_EOL)
5301 				break;
5302 			else if (opt == TCPOPT_NOP) {
5303 				cp++;
5304 				continue;
5305 			}
5306 
5307 			if (cp + 1 >= ep)
5308 				break;
5309 			advance = cp[1];
5310 			if ((cp + advance > ep) || (advance <= 0))
5311 				break;
5312 			switch (opt)
5313 			{
5314 			case TCPOPT_MAXSEG:
5315 				if (advance != 4)
5316 					break;
5317 				mss = cp[2] * 256 + cp[3];
5318 				if (mss > maxmss) {
5319 					cp[2] = maxmss / 256;
5320 					cp[3] = maxmss & 0xff;
5321 					CALC_SUMD(mss, maxmss, sumd);
5322 					fix_outcksum(csump, sumd);
5323 				}
5324 				break;
5325 			default:
5326 				/* ignore unknown options */
5327 				break;
5328 			}
5329 
5330 			cp += advance;
5331 		}
5332 	}
5333 }
5334 
5335 
5336 /* ------------------------------------------------------------------------ */
5337 /* Function:    fr_setnatqueue                                              */
5338 /* Returns:     Nil                                                         */
5339 /* Parameters:  nat(I)- pointer to NAT structure                            */
5340 /*              rev(I) - forward(0) or reverse(1) direction                 */
5341 /* Locks:       ipf_nat (read or write)                                     */
5342 /*                                                                          */
5343 /* Put the NAT entry on its default queue entry, using rev as a helped in   */
5344 /* determining which queue it should be placed on.                          */
5345 /* ------------------------------------------------------------------------ */
5346 void fr_setnatqueue(nat, rev, ifs)
5347 nat_t *nat;
5348 int rev;
5349 ipf_stack_t *ifs;
5350 {
5351 	ipftq_t *oifq, *nifq;
5352 
5353 	if (nat->nat_ptr != NULL)
5354 		nifq = nat->nat_ptr->in_tqehead[rev];
5355 	else
5356 		nifq = NULL;
5357 
5358 	if (nifq == NULL) {
5359 		switch (nat->nat_p)
5360 		{
5361 		case IPPROTO_UDP :
5362 			nifq = &ifs->ifs_nat_udptq;
5363 			break;
5364 		case IPPROTO_ICMP :
5365 			nifq = &ifs->ifs_nat_icmptq;
5366 			break;
5367 		case IPPROTO_TCP :
5368 			nifq = ifs->ifs_nat_tqb + nat->nat_tqe.tqe_state[rev];
5369 			break;
5370 		default :
5371 			nifq = &ifs->ifs_nat_iptq;
5372 			break;
5373 		}
5374 	}
5375 
5376 	oifq = nat->nat_tqe.tqe_ifq;
5377 	/*
5378 	 * If it's currently on a timeout queue, move it from one queue to
5379 	 * another, else put it on the end of the newly determined queue.
5380 	 */
5381 	if (oifq != NULL)
5382 		fr_movequeue(&nat->nat_tqe, oifq, nifq, ifs);
5383 	else
5384 		fr_queueappend(&nat->nat_tqe, nifq, nat, ifs);
5385 	return;
5386 }
5387 
5388 /* ------------------------------------------------------------------------ */
5389 /* Function:    nat_getnext                                                 */
5390 /* Returns:     int - 0 == ok, else error                                   */
5391 /* Parameters:  t(I)   - pointer to ipftoken structure                      */
5392 /*              itp(I) - pointer to ipfgeniter_t structure                  */
5393 /*              ifs - ipf stack instance                                    */
5394 /*                                                                          */
5395 /* Fetch the next nat/ipnat/hostmap structure pointer from the linked list  */
5396 /* and copy it out to the storage space pointed to by itp.  The next item   */
5397 /* in the list to look at is put back in the ipftoken struture.             */
5398 /* ------------------------------------------------------------------------ */
5399 static int nat_getnext(t, itp, ifs)
5400 ipftoken_t *t;
5401 ipfgeniter_t *itp;
5402 ipf_stack_t *ifs;
5403 {
5404 	hostmap_t *hm, *nexthm = NULL, zerohm;
5405 	ipnat_t *ipn, *nextipnat = NULL, zeroipn;
5406 	nat_t *nat, *nextnat = NULL, zeronat;
5407 	int error = 0, count;
5408 	char *dst;
5409 
5410 	if (itp->igi_nitems == 0)
5411 		return EINVAL;
5412 
5413 	READ_ENTER(&ifs->ifs_ipf_nat);
5414 
5415 	/*
5416 	 * Get "previous" entry from the token and find the next entry.
5417 	 */
5418 	switch (itp->igi_type)
5419 	{
5420 	case IPFGENITER_HOSTMAP :
5421 		hm = t->ipt_data;
5422 		if (hm == NULL) {
5423 			nexthm = ifs->ifs_ipf_hm_maplist;
5424 		} else {
5425 			nexthm = hm->hm_next;
5426 		}
5427 		break;
5428 
5429 	case IPFGENITER_IPNAT :
5430 		ipn = t->ipt_data;
5431 		if (ipn == NULL) {
5432 			nextipnat = ifs->ifs_nat_list;
5433 		} else {
5434 			nextipnat = ipn->in_next;
5435 		}
5436 		break;
5437 
5438 	case IPFGENITER_NAT :
5439 		nat = t->ipt_data;
5440 		if (nat == NULL) {
5441 			nextnat = ifs->ifs_nat_instances;
5442 		} else {
5443 			nextnat = nat->nat_next;
5444 		}
5445 		break;
5446 	default :
5447 		RWLOCK_EXIT(&ifs->ifs_ipf_nat);
5448 		return EINVAL;
5449 	}
5450 
5451 	dst = itp->igi_data;
5452 	for (count = itp->igi_nitems; count > 0; count--) {
5453 		/*
5454 		 * If we found an entry, add a reference to it and update the token.
5455 		 * Otherwise, zero out data to be returned and NULL out token.
5456 		 */
5457 		switch (itp->igi_type)
5458 		{
5459 		case IPFGENITER_HOSTMAP :
5460 			if (nexthm != NULL) {
5461 				ATOMIC_INC32(nexthm->hm_ref);
5462 				t->ipt_data = nexthm;
5463 			} else {
5464 				bzero(&zerohm, sizeof(zerohm));
5465 				nexthm = &zerohm;
5466 				t->ipt_data = NULL;
5467 			}
5468 			break;
5469 		case IPFGENITER_IPNAT :
5470 			if (nextipnat != NULL) {
5471 				ATOMIC_INC32(nextipnat->in_use);
5472 				t->ipt_data = nextipnat;
5473 			} else {
5474 				bzero(&zeroipn, sizeof(zeroipn));
5475 				nextipnat = &zeroipn;
5476 				t->ipt_data = NULL;
5477 			}
5478 			break;
5479 		case IPFGENITER_NAT :
5480 			if (nextnat != NULL) {
5481 				MUTEX_ENTER(&nextnat->nat_lock);
5482 				nextnat->nat_ref++;
5483 				MUTEX_EXIT(&nextnat->nat_lock);
5484 				t->ipt_data = nextnat;
5485 			} else {
5486 				bzero(&zeronat, sizeof(zeronat));
5487 				nextnat = &zeronat;
5488 				t->ipt_data = NULL;
5489 			}
5490 			break;
5491 		default :
5492 			break;
5493 		}
5494 
5495 		/*
5496 		 * Now that we have ref, it's save to give up lock.
5497 		 */
5498 		RWLOCK_EXIT(&ifs->ifs_ipf_nat);
5499 
5500 		/*
5501 		 * Copy out data and clean up references and token as needed.
5502 		 */
5503 		switch (itp->igi_type)
5504 		{
5505 		case IPFGENITER_HOSTMAP :
5506 			error = COPYOUT(nexthm, dst, sizeof(*nexthm));
5507 			if (error != 0)
5508 				error = EFAULT;
5509 			if (t->ipt_data == NULL) {
5510 				ipf_freetoken(t, ifs);
5511 				break;
5512 			} else {
5513 				if (hm != NULL) {
5514 					WRITE_ENTER(&ifs->ifs_ipf_nat);
5515 					fr_hostmapdel(&hm);
5516 					RWLOCK_EXIT(&ifs->ifs_ipf_nat);
5517 				}
5518 				if (nexthm->hm_next == NULL) {
5519 					ipf_freetoken(t, ifs);
5520 					break;
5521 				}
5522 				dst += sizeof(*nexthm);
5523 				hm = nexthm;
5524 				nexthm = nexthm->hm_next;
5525 			}
5526 			break;
5527 
5528 		case IPFGENITER_IPNAT :
5529 			error = COPYOUT(nextipnat, dst, sizeof(*nextipnat));
5530 			if (error != 0)
5531 				error = EFAULT;
5532 			if (t->ipt_data == NULL) {
5533 				ipf_freetoken(t, ifs);
5534 				break;
5535 			} else {
5536 				if (ipn != NULL) {
5537 					WRITE_ENTER(&ifs->ifs_ipf_nat);
5538 					fr_ipnatderef(&ipn, ifs);
5539 					RWLOCK_EXIT(&ifs->ifs_ipf_nat);
5540 				}
5541 				if (nextipnat->in_next == NULL) {
5542 					ipf_freetoken(t, ifs);
5543 					break;
5544 				}
5545 				dst += sizeof(*nextipnat);
5546 				ipn = nextipnat;
5547 				nextipnat = nextipnat->in_next;
5548 			}
5549 			break;
5550 
5551 		case IPFGENITER_NAT :
5552 			error = COPYOUT(nextnat, dst, sizeof(*nextnat));
5553 			if (error != 0)
5554 				error = EFAULT;
5555 			if (t->ipt_data == NULL) {
5556 				ipf_freetoken(t, ifs);
5557 				break;
5558 			} else {
5559 				if (nat != NULL)
5560 					fr_natderef(&nat, ifs);
5561 				if (nextnat->nat_next == NULL) {
5562 					ipf_freetoken(t, ifs);
5563 					break;
5564 				}
5565 				dst += sizeof(*nextnat);
5566 				nat = nextnat;
5567 				nextnat = nextnat->nat_next;
5568 			}
5569 			break;
5570 		default :
5571 			break;
5572 		}
5573 
5574 		if ((count == 1) || (error != 0))
5575 			break;
5576 
5577 		READ_ENTER(&ifs->ifs_ipf_nat);
5578 	}
5579 
5580 	return error;
5581 }
5582 
5583 
5584 /* ------------------------------------------------------------------------ */
5585 /* Function:    nat_iterator                                                */
5586 /* Returns:     int - 0 == ok, else error                                   */
5587 /* Parameters:  token(I) - pointer to ipftoken structure                    */
5588 /*              itp(I) - pointer to ipfgeniter_t structure                  */
5589 /*                                                                          */
5590 /* This function acts as a handler for the SIOCGENITER ioctls that use a    */
5591 /* generic structure to iterate through a list.  There are three different  */
5592 /* linked lists of NAT related information to go through: NAT rules, active */
5593 /* NAT mappings and the NAT fragment cache.                                 */
5594 /* ------------------------------------------------------------------------ */
5595 static int nat_iterator(token, itp, ifs)
5596 ipftoken_t *token;
5597 ipfgeniter_t *itp;
5598 ipf_stack_t *ifs;
5599 {
5600 	int error;
5601 
5602 	if (itp->igi_data == NULL)
5603 		return EFAULT;
5604 
5605 	token->ipt_subtype = itp->igi_type;
5606 
5607 	switch (itp->igi_type)
5608 	{
5609 	case IPFGENITER_HOSTMAP :
5610 	case IPFGENITER_IPNAT :
5611 	case IPFGENITER_NAT :
5612 		error = nat_getnext(token, itp, ifs);
5613 		break;
5614 	case IPFGENITER_NATFRAG :
5615 		error = fr_nextfrag(token, itp, &ifs->ifs_ipfr_natlist,
5616 				    &ifs->ifs_ipfr_nattail,
5617 				    &ifs->ifs_ipf_natfrag, ifs);
5618 		break;
5619 	default :
5620 		error = EINVAL;
5621 		break;
5622 	}
5623 
5624 	return error;
5625 }
5626 
5627 
5628 /* -------------------------------------------------------------------- */
5629 /* Function:	nat_earlydrop						*/
5630 /* Returns:	number of dropped/removed entries from the queue	*/
5631 /* Parameters:	ifq - pointer to queue with entries to be processed	*/
5632 /*		maxidle - entry must be idle this long to be dropped	*/
5633 /*		ifs - ipf stack instance				*/
5634 /*									*/
5635 /* Function is invoked from nat_extraflush() only.  Removes entries	*/
5636 /* form specified timeout queue, based on how long they've sat idle,	*/
5637 /* without waiting for it to happen on its own.				*/
5638 /* -------------------------------------------------------------------- */
5639 static int nat_earlydrop(ifq, maxidle, ifs)
5640 ipftq_t *ifq;
5641 int maxidle;
5642 ipf_stack_t *ifs;
5643 {
5644 	ipftqent_t *tqe, *tqn;
5645 	nat_t *nat;
5646 	unsigned int dropped;
5647 	int droptick;
5648 
5649 	if (ifq == NULL)
5650 		return (0);
5651 
5652 	dropped = 0;
5653 
5654 	/*
5655 	 * Determine the tick representing the idle time we're interested
5656 	 * in.  If an entry exists in the queue, and it was touched before
5657 	 * that tick, then it's been idle longer than maxidle ... remove it.
5658 	 */
5659 	droptick = ifs->ifs_fr_ticks - maxidle;
5660 	tqn = ifq->ifq_head;
5661 	while ((tqe = tqn) != NULL && tqe->tqe_touched < droptick) {
5662 		tqn = tqe->tqe_next;
5663 		nat = tqe->tqe_parent;
5664 		nat_delete(nat, ISL_EXPIRE, ifs);
5665 		dropped++;
5666 	}
5667 	return (dropped);
5668 }
5669 
5670 
5671 /* --------------------------------------------------------------------- */
5672 /* Function:	nat_flushclosing					 */
5673 /* Returns:	int - number of NAT entries deleted			 */
5674 /* Parameters:	stateval(I) - State at which to start removing entries	 */
5675 /*		ifs - ipf stack instance				 */
5676 /*									 */
5677 /* Remove nat table entries for TCP connections which are in the process */
5678 /* of closing, and are in (or "beyond") state specified by 'stateval'.	 */
5679 /* --------------------------------------------------------------------- */
5680 static int nat_flushclosing(stateval, ifs)
5681 int stateval;
5682 ipf_stack_t *ifs;
5683 {
5684 	ipftq_t *ifq, *ifqn;
5685 	ipftqent_t *tqe, *tqn;
5686 	nat_t *nat;
5687 	int dropped;
5688 
5689 	dropped = 0;
5690 
5691 	/*
5692 	 * Start by deleting any entries in specific timeout queues.
5693 	 */
5694 	ifqn = &ifs->ifs_nat_tqb[stateval];
5695 	while ((ifq = ifqn) != NULL) {
5696 		ifqn = ifq->ifq_next;
5697 		dropped += nat_earlydrop(ifq, (int)0, ifs);
5698 	}
5699 
5700 	/*
5701 	 * Next, look through user defined queues for closing entries.
5702 	 */
5703 	ifqn = ifs->ifs_nat_utqe;
5704 	while ((ifq = ifqn) != NULL) {
5705 		ifqn = ifq->ifq_next;
5706 		tqn = ifq->ifq_head;
5707 		while ((tqe = tqn) != NULL) {
5708 			tqn = tqe->tqe_next;
5709 			nat = tqe->tqe_parent;
5710 			if (nat->nat_p != IPPROTO_TCP)
5711 				continue;
5712 			if ((nat->nat_tcpstate[0] >= stateval) &&
5713 			    (nat->nat_tcpstate[1] >= stateval)) {
5714 				nat_delete(nat, NL_EXPIRE, ifs);
5715 				dropped++;
5716 			}
5717 		}
5718 	}
5719 	return (dropped);
5720 }
5721 
5722 
5723 /* --------------------------------------------------------------------- */
5724 /* Function:	nat_extraflush						 */
5725 /* Returns:	int - number of NAT entries deleted			 */
5726 /* Parameters:	which(I) - how to flush the active NAT table		 */
5727 /*		ifs - ipf stack instance				 */
5728 /* Write Locks:	ipf_nat							 */
5729 /*									 */
5730 /* Flush nat tables.  Three actions currently defined:			 */
5731 /*									 */
5732 /* which == 0 :	Flush all nat table entries.				 */
5733 /*									 */
5734 /* which == 1 :	Flush entries with TCP connections which have started	 */
5735 /*		to close on both ends.					 */
5736 /*									 */
5737 /* which == 2 :	First, flush entries which are "almost" closed.  If that */
5738 /*		does not take us below specified threshold in the table, */
5739 /*		we want to flush entries with TCP connections which have */
5740 /*		been idle for a long time.  Start with connections idle	 */
5741 /*		over 12 hours,  and then work backwards in half hour	 */
5742 /*		increments to at most 30 minutes idle, and finally work	 */
5743 /*		back in 30 second increments to at most 30 seconds.	 */
5744 /* --------------------------------------------------------------------- */
5745 static int nat_extraflush(which, ifs)
5746 int which;
5747 ipf_stack_t *ifs;
5748 {
5749 	ipftq_t *ifq, *ifqn;
5750 	nat_t *nat, **natp;
5751 	int idletime, removed, idle_idx;
5752 	SPL_INT(s);
5753 
5754 	removed = 0;
5755 
5756 	SPL_NET(s);
5757 	switch (which)
5758 	{
5759 	case 0:
5760 		natp = &ifs->ifs_nat_instances;
5761 		while ((nat = *natp) != NULL) {
5762 			natp = &nat->nat_next;
5763 			nat_delete(nat, ISL_FLUSH, ifs);
5764 			removed++;
5765 		}
5766 		break;
5767 
5768 	case 1:
5769 		removed = nat_flushclosing(IPF_TCPS_CLOSE_WAIT, ifs);
5770 		break;
5771 
5772 	case 2:
5773 		removed = nat_flushclosing(IPF_TCPS_FIN_WAIT_2, ifs);
5774 
5775 		/*
5776 		 * Be sure we haven't done this in the last 10 seconds.
5777 		 */
5778 		if (ifs->ifs_fr_ticks - ifs->ifs_nat_last_force_flush <
5779 		    IPF_TTLVAL(10))
5780 			break;
5781 		ifs->ifs_nat_last_force_flush = ifs->ifs_fr_ticks;
5782 
5783 		/*
5784 		 * Determine initial threshold for minimum idle time based on
5785 		 * how long ipfilter has been running.  Ipfilter needs to have
5786 		 * been up as long as the smallest interval to continue on.
5787 		 *
5788 		 * Minimum idle times stored in idletime_tab and indexed by
5789 		 * idle_idx.  Start at upper end of array and work backwards.
5790 		 *
5791 		 * Once the index is found, set the initial idle time to the
5792 		 * first interval before the current ipfilter run time.
5793 		 */
5794 		if (ifs->ifs_fr_ticks < idletime_tab[0])
5795 			break;  /* switch */
5796 		idle_idx = (sizeof (idletime_tab) / sizeof (int)) - 1;
5797 		if (ifs->ifs_fr_ticks > idletime_tab[idle_idx]) {
5798 			idletime = idletime_tab[idle_idx];
5799 		} else {
5800 			while ((idle_idx > 0) &&
5801 			    (ifs->ifs_fr_ticks < idletime_tab[idle_idx]))
5802 				idle_idx--;
5803 			idletime = (ifs->ifs_fr_ticks /
5804 				    idletime_tab[idle_idx]) *
5805 				    idletime_tab[idle_idx];
5806 		}
5807 
5808 		while ((idle_idx >= 0) &&
5809 		    (NAT_TAB_WATER_LEVEL(ifs) > ifs->ifs_nat_flush_lvl_lo)) {
5810 			/*
5811 			 * Start with appropriate timeout queue.
5812 			 */
5813 			removed += nat_earlydrop(
5814 					&ifs->ifs_nat_tqb[IPF_TCPS_ESTABLISHED],
5815 					idletime, ifs);
5816 
5817 			/*
5818 			 * Make sure we haven't already deleted enough
5819 			 * entries before checking the user defined queues.
5820 			 */
5821 			if (NAT_TAB_WATER_LEVEL(ifs) <=
5822 			    ifs->ifs_nat_flush_lvl_lo)
5823 				break;
5824 
5825 			/*
5826 			 * Next, look through the user defined queues.
5827 			 */
5828 			ifqn = ifs->ifs_nat_utqe;
5829 			while ((ifq = ifqn) != NULL) {
5830 				ifqn = ifq->ifq_next;
5831 				removed += nat_earlydrop(ifq, idletime, ifs);
5832 			}
5833 
5834 			/*
5835 			 * Adjust the granularity of idle time.
5836 			 *
5837 			 * If we reach an interval boundary, we need to
5838 			 * either adjust the idle time accordingly or exit
5839 			 * the loop altogether (if this is very last check).
5840 			 */
5841 			idletime -= idletime_tab[idle_idx];
5842 			if (idletime < idletime_tab[idle_idx]) {
5843 				if (idle_idx != 0) {
5844 					idletime = idletime_tab[idle_idx] -
5845 					    idletime_tab[idle_idx - 1];
5846 					idle_idx--;
5847 				} else {
5848 					break;  /* while */
5849 				}
5850 			}
5851 		}
5852 		break;
5853 	default:
5854 		break;
5855 	}
5856 
5857 	SPL_X(s);
5858 	return (removed);
5859 }
5860