xref: /netbsd-src/sys/external/bsd/ipf/netinet/ip_nat.c (revision 82ad575716605df31379cf04a2f3efbc97b8a6f5)
1 /*	$NetBSD: ip_nat.c,v 1.6 2012/07/30 19:27:46 pgoyette Exp $	*/
2 
3 /*
4  * Copyright (C) 2012 by Darren Reed.
5  *
6  * See the IPFILTER.LICENCE file for details on licencing.
7  */
8 #if defined(KERNEL) || defined(_KERNEL)
9 # undef KERNEL
10 # undef KERNEL
11 # define        KERNEL	1
12 # define        KERNEL	1
13 #endif
14 #include <sys/errno.h>
15 #include <sys/types.h>
16 #include <sys/param.h>
17 #include <sys/time.h>
18 #include <sys/file.h>
19 #if defined(_KERNEL) && \
20     (defined(__NetBSD_Version) && (__NetBSD_Version >= 399002000))
21 # include <sys/kauth.h>
22 #endif
23 #if !defined(_KERNEL)
24 # include <stdio.h>
25 # include <string.h>
26 # include <stdlib.h>
27 # define KERNEL
28 # ifdef _OpenBSD__
29 struct file;
30 # endif
31 # include <sys/uio.h>
32 # undef KERNEL
33 #endif
34 #if defined(_KERNEL) && \
35     defined(__FreeBSD_version) && (__FreeBSD_version >= 220000)
36 # include <sys/filio.h>
37 # include <sys/fcntl.h>
38 #else
39 # include <sys/ioctl.h>
40 #endif
41 #if !defined(AIX)
42 # include <sys/fcntl.h>
43 #endif
44 #if !defined(linux)
45 # include <sys/protosw.h>
46 #endif
47 #include <sys/socket.h>
48 #if defined(_KERNEL)
49 # include <sys/systm.h>
50 # if !defined(__SVR4) && !defined(__svr4__)
51 #  include <sys/mbuf.h>
52 # endif
53 #endif
54 #if defined(__SVR4) || defined(__svr4__)
55 # include <sys/filio.h>
56 # include <sys/byteorder.h>
57 # ifdef KERNEL
58 #  include <sys/dditypes.h>
59 # endif
60 # include <sys/stream.h>
61 # include <sys/kmem.h>
62 #endif
63 #if _FreeBSD_version >= 300000
64 # include <sys/queue.h>
65 #endif
66 #include <net/if.h>
67 #if _FreeBSD_version >= 300000
68 # include <net/if_var.h>
69 #endif
70 #ifdef sun
71 # include <net/af.h>
72 #endif
73 #include <netinet/in.h>
74 #include <netinet/in_systm.h>
75 #include <netinet/ip.h>
76 
77 #ifdef RFC1825
78 # include <vpn/md5.h>
79 # include <vpn/ipsec.h>
80 extern struct ifnet vpnif;
81 #endif
82 
83 #if !defined(linux)
84 # include <netinet/ip_var.h>
85 #endif
86 #include <netinet/tcp.h>
87 #include <netinet/udp.h>
88 #include <netinet/ip_icmp.h>
89 #include "netinet/ip_compat.h"
90 #include <netinet/tcpip.h>
91 #include "netinet/ipl.h"
92 #include "netinet/ip_fil.h"
93 #include "netinet/ip_nat.h"
94 #include "netinet/ip_frag.h"
95 #include "netinet/ip_state.h"
96 #include "netinet/ip_proxy.h"
97 #include "netinet/ip_lookup.h"
98 #include "netinet/ip_dstlist.h"
99 #include "netinet/ip_sync.h"
100 #if FREEBSD_GE_REV(300000)
101 # include <sys/malloc.h>
102 #endif
103 #ifdef HAS_SYS_MD5_H
104 # include <sys/md5.h>
105 #else
106 # include "md5.h"
107 #endif
108 /* END OF INCLUDES */
109 
110 #undef	SOCKADDR_IN
111 #define	SOCKADDR_IN	struct sockaddr_in
112 
113 #if !defined(lint)
114 #if defined(__NetBSD__)
115 #include <sys/cdefs.h>
116 __KERNEL_RCSID(0, "$NetBSD: ip_nat.c,v 1.6 2012/07/30 19:27:46 pgoyette Exp $");
117 #else
118 static const char sccsid[] = "@(#)ip_nat.c	1.11 6/5/96 (C) 1995 Darren Reed";
119 static const char rcsid[] = "@(#)Id: ip_nat.c,v 1.1.1.2 2012/07/22 13:45:27 darrenr Exp";
120 #endif
121 #endif
122 
123 
124 #define	NATFSUM(n,v,f)	((v) == 4 ? (n)->f.in4.s_addr : (n)->f.i6[0] + \
125 			 (n)->f.i6[1] + (n)->f.i6[2] + (n)->f.i6[3])
126 #define	NBUMP(x)	softn->(x)++
127 #define	NBUMPD(x, y)	do { \
128 				softn->x.y++; \
129 				DT(y); \
130 			} while (0)
131 #define	NBUMPSIDE(y,x)	softn->ipf_nat_stats.ns_side[y].x++
132 #define	NBUMPSIDED(y,x)	do { softn->ipf_nat_stats.ns_side[y].x++; \
133 			     DT(x); } while (0)
134 #define	NBUMPSIDEX(y,x,z) \
135 			do { softn->ipf_nat_stats.ns_side[y].x++; \
136 			     DT(z); } while (0)
137 #define	NBUMPSIDEDF(y,x)do { softn->ipf_nat_stats.ns_side[y].x++; \
138 			     DT1(x, fr_info_t *, fin); } while (0)
139 
140 frentry_t	ipfnatblock;
141 
142 static ipftuneable_t ipf_nat_tuneables[] = {
143 	/* nat */
144 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_lock) },
145 		"nat_lock",	0,	1,
146 		stsizeof(ipf_nat_softc_t, ipf_nat_lock),
147 		IPFT_RDONLY,		NULL,	NULL },
148 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_sz) },
149 		"nat_table_size", 1,	0x7fffffff,
150 		stsizeof(ipf_nat_softc_t, ipf_nat_table_sz),
151 		0,			NULL,	ipf_nat_rehash },
152 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_max) },
153 		"nat_table_max", 1,	0x7fffffff,
154 		stsizeof(ipf_nat_softc_t, ipf_nat_table_max),
155 		0,			NULL,	NULL },
156 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_maprules_sz) },
157 		"nat_rules_size", 1,	0x7fffffff,
158 		stsizeof(ipf_nat_softc_t, ipf_nat_maprules_sz),
159 		0,			NULL,	ipf_nat_rehash_rules },
160 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_rdrrules_sz) },
161 		"rdr_rules_size", 1,	0x7fffffff,
162 		stsizeof(ipf_nat_softc_t, ipf_nat_rdrrules_sz),
163 		0,			NULL,	ipf_nat_rehash_rules },
164 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_hostmap_sz) },
165 		"hostmap_size",	1,	0x7fffffff,
166 		stsizeof(ipf_nat_softc_t, ipf_nat_hostmap_sz),
167 		0,			NULL,	ipf_nat_hostmap_rehash },
168 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_maxbucket) },
169 		"nat_maxbucket",1,	0x7fffffff,
170 		stsizeof(ipf_nat_softc_t, ipf_nat_maxbucket),
171 		0,			NULL,	NULL },
172 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_logging) },
173 		"nat_logging",	0,	1,
174 		stsizeof(ipf_nat_softc_t, ipf_nat_logging),
175 		0,			NULL,	NULL },
176 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_doflush) },
177 		"nat_doflush",	0,	1,
178 		stsizeof(ipf_nat_softc_t, ipf_nat_doflush),
179 		0,			NULL,	NULL },
180 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_wm_low) },
181 		"nat_table_wm_low",	1,	99,
182 		stsizeof(ipf_nat_softc_t, ipf_nat_table_wm_low),
183 		0,			NULL,	NULL },
184 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_wm_high) },
185 		"nat_table_wm_high",	2,	100,
186 		stsizeof(ipf_nat_softc_t, ipf_nat_table_wm_high),
187 		0,			NULL,	NULL },
188 	{ { 0 },
189 		NULL,			0,	0,
190 		0,
191 		0,			NULL,	NULL }
192 };
193 
194 /* ======================================================================== */
195 /* How the NAT is organised and works.                                      */
196 /*                                                                          */
197 /* Inside (interface y) NAT       Outside (interface x)                     */
198 /* -------------------- -+- -------------------------------------           */
199 /* Packet going          |   out, processsed by ipf_nat_checkout() for x    */
200 /* ------------>         |   ------------>                                  */
201 /* src=10.1.1.1          |   src=192.1.1.1                                  */
202 /*                       |                                                  */
203 /*                       |   in, processed by ipf_nat_checkin() for x       */
204 /* <------------         |   <------------                                  */
205 /* dst=10.1.1.1          |   dst=192.1.1.1                                  */
206 /* -------------------- -+- -------------------------------------           */
207 /* ipf_nat_checkout() - changes ip_src and if required, sport               */
208 /*             - creates a new mapping, if required.                        */
209 /* ipf_nat_checkin()  - changes ip_dst and if required, dport               */
210 /*                                                                          */
211 /* In the NAT table, internal source is recorded as "in" and externally     */
212 /* seen as "out".                                                           */
213 /* ======================================================================== */
214 
215 
216 #if SOLARIS && !defined(INSTANCES)
217 extern	int		pfil_delayed_copy;
218 #endif
219 
220 static	int	ipf_nat_flush_entry(ipf_main_softc_t *, void *);
221 static	int	ipf_nat_getent(ipf_main_softc_t *, void *, int);
222 static	int	ipf_nat_getsz(ipf_main_softc_t *, void *, int);
223 static	int	ipf_nat_putent(ipf_main_softc_t *, void *, int);
224 static	void	ipf_nat_addmap(ipf_nat_softc_t *, ipnat_t *);
225 static	void	ipf_nat_addrdr(ipf_nat_softc_t *, ipnat_t *);
226 static	int	ipf_nat_builddivertmp(ipf_nat_softc_t *, ipnat_t *);
227 static	int	ipf_nat_clearlist(ipf_main_softc_t *, ipf_nat_softc_t *);
228 static	int	ipf_nat_cmp_rules(ipnat_t *, ipnat_t *);
229 static	int	ipf_nat_decap(fr_info_t *, nat_t *);
230 static	void	ipf_nat_delrule(ipf_main_softc_t *, ipf_nat_softc_t *,
231 				     ipnat_t *, int);
232 static	int	ipf_nat_extraflush(ipf_main_softc_t *, ipf_nat_softc_t *, int);
233 static	int	ipf_nat_finalise(fr_info_t *, nat_t *);
234 static	int	ipf_nat_flushtable(ipf_main_softc_t *, ipf_nat_softc_t *);
235 static	int	ipf_nat_getnext(ipf_main_softc_t *, ipftoken_t *,
236 				ipfgeniter_t *, ipfobj_t *);
237 static	int	ipf_nat_gettable(ipf_main_softc_t *, ipf_nat_softc_t *, char *);
238 static	hostmap_t *ipf_nat_hostmap(ipf_nat_softc_t *, ipnat_t *,
239 				   struct in_addr, struct in_addr,
240 				   struct in_addr, u_32_t);
241 static	int	ipf_nat_icmpquerytype(int);
242 static	int	ipf_nat_iterator(ipf_main_softc_t *, ipftoken_t *,
243 				 ipfgeniter_t *, ipfobj_t *);
244 static	int	ipf_nat_match(fr_info_t *, ipnat_t *);
245 static	int	ipf_nat_matcharray(nat_t *, int *, u_long);
246 static	int	ipf_nat_matchflush(ipf_main_softc_t *, ipf_nat_softc_t *,
247 				   void *);
248 static	void	ipf_nat_mssclamp(tcphdr_t *, u_32_t, fr_info_t *, u_short *);
249 static	int	ipf_nat_newmap(fr_info_t *, nat_t *, natinfo_t *);
250 static	int	ipf_nat_newdivert(fr_info_t *, nat_t *, natinfo_t *);
251 static	int	ipf_nat_newrdr(fr_info_t *, nat_t *, natinfo_t *);
252 static	int	ipf_nat_newrewrite(fr_info_t *, nat_t *, natinfo_t *);
253 static	int	ipf_nat_nextaddr(fr_info_t *, nat_addr_t *, u_32_t *, u_32_t *);
254 static	int	ipf_nat_nextaddrinit(ipf_main_softc_t *, char *,
255 				     nat_addr_t *, int, void *);
256 static	int	ipf_nat_resolverule(ipf_main_softc_t *, ipnat_t *);
257 static	int	ipf_nat_ruleaddrinit(ipf_main_softc_t *,
258 				     ipf_nat_softc_t *, ipnat_t *);
259 static	void	ipf_nat_rule_fini(ipf_main_softc_t *, ipnat_t *);
260 static	int	ipf_nat_rule_init(ipf_main_softc_t *, ipf_nat_softc_t *,
261 				       ipnat_t *);
262 static	int	ipf_nat_siocaddnat(ipf_main_softc_t *, ipf_nat_softc_t *,
263 					ipnat_t *, int);
264 static	void	ipf_nat_siocdelnat(ipf_main_softc_t *, ipf_nat_softc_t *,
265 					ipnat_t *, int);
266 static	void	ipf_nat_tabmove(ipf_nat_softc_t *, nat_t *);
267 
268 /* ------------------------------------------------------------------------ */
269 /* Function:    ipf_nat_main_load                                           */
270 /* Returns:     int - 0 == success, -1 == failure                           */
271 /* Parameters:  Nil                                                         */
272 /*                                                                          */
273 /* The only global NAT structure that needs to be initialised is the filter */
274 /* rule that is used with blocking packets.                                 */
275 /* ------------------------------------------------------------------------ */
276 int
277 ipf_nat_main_load(void)
278 {
279 	bzero((char *)&ipfnatblock, sizeof(ipfnatblock));
280 	ipfnatblock.fr_flags = FR_BLOCK|FR_QUICK;
281 	ipfnatblock.fr_ref = 1;
282 
283 	return 0;
284 }
285 
286 
287 /* ------------------------------------------------------------------------ */
288 /* Function:    ipf_nat_main_unload                                         */
289 /* Returns:     int - 0 == success, -1 == failure                           */
290 /* Parameters:  Nil                                                         */
291 /*                                                                          */
292 /* A null-op function that exists as a placeholder so that the flow in      */
293 /* other functions is obvious.                                              */
294 /* ------------------------------------------------------------------------ */
295 int
296 ipf_nat_main_unload(void)
297 {
298 	return 0;
299 }
300 
301 
302 /* ------------------------------------------------------------------------ */
303 /* Function:    ipf_nat_soft_create                                         */
304 /* Returns:     void * - NULL = failure, else pointer to NAT context        */
305 /* Parameters:  softc(I) - pointer to soft context main structure           */
306 /*                                                                          */
307 /* Allocate the initial soft context structure for NAT and populate it with */
308 /* some default values. Creating the tables is left until we call _init so  */
309 /* that sizes can be changed before we get under way.                       */
310 /* ------------------------------------------------------------------------ */
311 void *
312 ipf_nat_soft_create(ipf_main_softc_t *softc)
313 {
314 	ipf_nat_softc_t *softn;
315 
316 	KMALLOC(softn, ipf_nat_softc_t *);
317 	if (softn == NULL)
318 		return NULL;
319 
320 	bzero((char *)softn, sizeof(*softn));
321 
322 	softn->ipf_nat_tune = ipf_tune_array_copy(softn,
323 						  sizeof(ipf_nat_tuneables),
324 						  ipf_nat_tuneables);
325 	if (softn->ipf_nat_tune == NULL) {
326 		ipf_nat_soft_destroy(softc, softn);
327 		return NULL;
328 	}
329 	if (ipf_tune_array_link(softc, softn->ipf_nat_tune) == -1) {
330 		ipf_nat_soft_destroy(softc, softn);
331 		return NULL;
332 	}
333 
334 	softn->ipf_nat_list_tail = &softn->ipf_nat_list;
335 
336 	softn->ipf_nat_table_max = NAT_TABLE_MAX;
337 	softn->ipf_nat_table_sz = NAT_TABLE_SZ;
338 	softn->ipf_nat_maprules_sz = NAT_SIZE;
339 	softn->ipf_nat_rdrrules_sz = RDR_SIZE;
340 	softn->ipf_nat_hostmap_sz = HOSTMAP_SIZE;
341 	softn->ipf_nat_doflush = 0;
342 #ifdef  IPFILTER_LOG
343 	softn->ipf_nat_logging = 1;
344 #else
345 	softn->ipf_nat_logging = 0;
346 #endif
347 
348 	softn->ipf_nat_defage = DEF_NAT_AGE;
349 	softn->ipf_nat_defipage = IPF_TTLVAL(60);
350 	softn->ipf_nat_deficmpage = IPF_TTLVAL(3);
351 	softn->ipf_nat_table_wm_high = 99;
352 	softn->ipf_nat_table_wm_low = 90;
353 
354 	return softn;
355 }
356 
357 /* ------------------------------------------------------------------------ */
358 /* Function:    ipf_nat_soft_destroy                                        */
359 /* Returns:     Nil                                                         */
360 /* Parameters:  softc(I) - pointer to soft context main structure           */
361 /*                                                                          */
362 /* ------------------------------------------------------------------------ */
363 void
364 ipf_nat_soft_destroy(ipf_main_softc_t *softc, void *arg)
365 {
366 	ipf_nat_softc_t *softn = arg;
367 
368 	if (softn->ipf_nat_tune != NULL) {
369 		ipf_tune_array_unlink(softc, softn->ipf_nat_tune);
370 		KFREES(softn->ipf_nat_tune, sizeof(ipf_nat_tuneables));
371 		softn->ipf_nat_tune = NULL;
372 	}
373 
374 	KFREE(softn);
375 }
376 
377 
378 /* ------------------------------------------------------------------------ */
379 /* Function:    ipf_nat_init                                                */
380 /* Returns:     int - 0 == success, -1 == failure                           */
381 /* Parameters:  softc(I) - pointer to soft context main structure           */
382 /*                                                                          */
383 /* Initialise all of the NAT locks, tables and other structures.            */
384 /* ------------------------------------------------------------------------ */
385 int
386 ipf_nat_soft_init(ipf_main_softc_t *softc, void *arg)
387 {
388 	ipf_nat_softc_t *softn = arg;
389 	ipftq_t *tq;
390 	int i;
391 
392 	KMALLOCS(softn->ipf_nat_table[0], nat_t **, \
393 		 sizeof(nat_t *) * softn->ipf_nat_table_sz);
394 
395 	if (softn->ipf_nat_table[0] != NULL) {
396 		bzero((char *)softn->ipf_nat_table[0],
397 		      softn->ipf_nat_table_sz * sizeof(nat_t *));
398 	} else {
399 		return -1;
400 	}
401 
402 	KMALLOCS(softn->ipf_nat_table[1], nat_t **, \
403 		 sizeof(nat_t *) * softn->ipf_nat_table_sz);
404 
405 	if (softn->ipf_nat_table[1] != NULL) {
406 		bzero((char *)softn->ipf_nat_table[1],
407 		      softn->ipf_nat_table_sz * sizeof(nat_t *));
408 	} else {
409 		return -2;
410 	}
411 
412 	KMALLOCS(softn->ipf_nat_map_rules, ipnat_t **, \
413 		 sizeof(ipnat_t *) * softn->ipf_nat_maprules_sz);
414 
415 	if (softn->ipf_nat_map_rules != NULL) {
416 		bzero((char *)softn->ipf_nat_map_rules,
417 		      softn->ipf_nat_maprules_sz * sizeof(ipnat_t *));
418 	} else {
419 		return -3;
420 	}
421 
422 	KMALLOCS(softn->ipf_nat_rdr_rules, ipnat_t **, \
423 		 sizeof(ipnat_t *) * softn->ipf_nat_rdrrules_sz);
424 
425 	if (softn->ipf_nat_rdr_rules != NULL) {
426 		bzero((char *)softn->ipf_nat_rdr_rules,
427 		      softn->ipf_nat_rdrrules_sz * sizeof(ipnat_t *));
428 	} else {
429 		return -4;
430 	}
431 
432 	KMALLOCS(softn->ipf_hm_maptable, hostmap_t **, \
433 		 sizeof(hostmap_t *) * softn->ipf_nat_hostmap_sz);
434 
435 	if (softn->ipf_hm_maptable != NULL) {
436 		bzero((char *)softn->ipf_hm_maptable,
437 		      sizeof(hostmap_t *) * softn->ipf_nat_hostmap_sz);
438 	} else {
439 		return -5;
440 	}
441 	softn->ipf_hm_maplist = NULL;
442 
443 	KMALLOCS(softn->ipf_nat_stats.ns_side[0].ns_bucketlen, u_int *,
444 		 softn->ipf_nat_table_sz * sizeof(u_int));
445 
446 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen == NULL) {
447 		return -6;
448 	}
449 	bzero((char *)softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
450 	      softn->ipf_nat_table_sz * sizeof(u_int));
451 
452 	KMALLOCS(softn->ipf_nat_stats.ns_side[1].ns_bucketlen, u_int *,
453 		 softn->ipf_nat_table_sz * sizeof(u_int));
454 
455 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen == NULL) {
456 		return -7;
457 	}
458 
459 	bzero((char *)softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
460 	      softn->ipf_nat_table_sz * sizeof(u_int));
461 
462 	if (softn->ipf_nat_maxbucket == 0) {
463 		for (i = softn->ipf_nat_table_sz; i > 0; i >>= 1)
464 			softn->ipf_nat_maxbucket++;
465 		softn->ipf_nat_maxbucket *= 2;
466 	}
467 
468 	ipf_sttab_init(softc, softn->ipf_nat_tcptq);
469 	/*
470 	 * Increase this because we may have "keep state" following this too
471 	 * and packet storms can occur if this is removed too quickly.
472 	 */
473 	softn->ipf_nat_tcptq[IPF_TCPS_CLOSED].ifq_ttl = softc->ipf_tcplastack;
474 	softn->ipf_nat_tcptq[IPF_TCP_NSTATES - 1].ifq_next =
475 							&softn->ipf_nat_udptq;
476 
477 	IPFTQ_INIT(&softn->ipf_nat_udptq, softn->ipf_nat_defage,
478 		   "nat ipftq udp tab");
479 	softn->ipf_nat_udptq.ifq_next = &softn->ipf_nat_udpacktq;
480 
481 	IPFTQ_INIT(&softn->ipf_nat_udpacktq, softn->ipf_nat_defage,
482 		   "nat ipftq udpack tab");
483 	softn->ipf_nat_udpacktq.ifq_next = &softn->ipf_nat_icmptq;
484 
485 	IPFTQ_INIT(&softn->ipf_nat_icmptq, softn->ipf_nat_deficmpage,
486 		   "nat icmp ipftq tab");
487 	softn->ipf_nat_icmptq.ifq_next = &softn->ipf_nat_icmpacktq;
488 
489 	IPFTQ_INIT(&softn->ipf_nat_icmpacktq, softn->ipf_nat_defage,
490 		   "nat icmpack ipftq tab");
491 	softn->ipf_nat_icmpacktq.ifq_next = &softn->ipf_nat_iptq;
492 
493 	IPFTQ_INIT(&softn->ipf_nat_iptq, softn->ipf_nat_defipage,
494 		   "nat ip ipftq tab");
495 	softn->ipf_nat_iptq.ifq_next = &softn->ipf_nat_pending;
496 
497 	IPFTQ_INIT(&softn->ipf_nat_pending, 1, "nat pending ipftq tab");
498 	softn->ipf_nat_pending.ifq_next = NULL;
499 
500 	for (i = 0, tq = softn->ipf_nat_tcptq; i < IPF_TCP_NSTATES; i++, tq++) {
501 		if (tq->ifq_ttl < softn->ipf_nat_deficmpage)
502 			tq->ifq_ttl = softn->ipf_nat_deficmpage;
503 #ifdef LARGE_NAT
504 		else if (tq->ifq_ttl > softn->ipf_nat_defage)
505 			tq->ifq_ttl = softn->ipf_nat_defage;
506 #endif
507 	}
508 
509 	/*
510 	 * Increase this because we may have "keep state" following
511 	 * this too and packet storms can occur if this is removed
512 	 * too quickly.
513 	 */
514 	softn->ipf_nat_tcptq[IPF_TCPS_CLOSED].ifq_ttl = softc->ipf_tcplastack;
515 
516 	MUTEX_INIT(&softn->ipf_nat_new, "ipf nat new mutex");
517 	MUTEX_INIT(&softn->ipf_nat_io, "ipf nat io mutex");
518 
519 	softn->ipf_nat_inited = 1;
520 
521 	return 0;
522 }
523 
524 
525 /* ------------------------------------------------------------------------ */
526 /* Function:    ipf_nat_soft_fini                                           */
527 /* Returns:     Nil                                                         */
528 /* Parameters:  softc(I) - pointer to soft context main structure           */
529 /*                                                                          */
530 /* Free all memory used by NAT structures allocated at runtime.             */
531 /* ------------------------------------------------------------------------ */
532 int
533 ipf_nat_soft_fini(ipf_main_softc_t *softc, void *arg)
534 {
535 	ipf_nat_softc_t *softn = arg;
536 	ipftq_t *ifq, *ifqnext;
537 
538 	(void) ipf_nat_clearlist(softc, softn);
539 	(void) ipf_nat_flushtable(softc, softn);
540 
541 	/*
542 	 * Proxy timeout queues are not cleaned here because although they
543 	 * exist on the NAT list, ipf_proxy_unload is called after unload
544 	 * and the proxies actually are responsible for them being created.
545 	 * Should the proxy timeouts have their own list?  There's no real
546 	 * justification as this is the only complication.
547 	 */
548 	for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifqnext) {
549 		ifqnext = ifq->ifq_next;
550 		if (ipf_deletetimeoutqueue(ifq) == 0)
551 			ipf_freetimeoutqueue(softc, ifq);
552 	}
553 
554 	if (softn->ipf_nat_table[0] != NULL) {
555 		KFREES(softn->ipf_nat_table[0],
556 		       sizeof(nat_t *) * softn->ipf_nat_table_sz);
557 		softn->ipf_nat_table[0] = NULL;
558 	}
559 	if (softn->ipf_nat_table[1] != NULL) {
560 		KFREES(softn->ipf_nat_table[1],
561 		       sizeof(nat_t *) * softn->ipf_nat_table_sz);
562 		softn->ipf_nat_table[1] = NULL;
563 	}
564 	if (softn->ipf_nat_map_rules != NULL) {
565 		KFREES(softn->ipf_nat_map_rules,
566 		       sizeof(ipnat_t *) * softn->ipf_nat_maprules_sz);
567 		softn->ipf_nat_map_rules = NULL;
568 	}
569 	if (softn->ipf_nat_rdr_rules != NULL) {
570 		KFREES(softn->ipf_nat_rdr_rules,
571 		       sizeof(ipnat_t *) * softn->ipf_nat_rdrrules_sz);
572 		softn->ipf_nat_rdr_rules = NULL;
573 	}
574 	if (softn->ipf_hm_maptable != NULL) {
575 		KFREES(softn->ipf_hm_maptable,
576 		       sizeof(hostmap_t *) * softn->ipf_nat_hostmap_sz);
577 		softn->ipf_hm_maptable = NULL;
578 	}
579 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen != NULL) {
580 		KFREES(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
581 		       sizeof(u_int) * softn->ipf_nat_table_sz);
582 		softn->ipf_nat_stats.ns_side[0].ns_bucketlen = NULL;
583 	}
584 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen != NULL) {
585 		KFREES(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
586 		       sizeof(u_int) * softn->ipf_nat_table_sz);
587 		softn->ipf_nat_stats.ns_side[1].ns_bucketlen = NULL;
588 	}
589 
590 	if (softn->ipf_nat_inited == 1) {
591 		softn->ipf_nat_inited = 0;
592 		ipf_sttab_destroy(softn->ipf_nat_tcptq);
593 
594 		MUTEX_DESTROY(&softn->ipf_nat_new);
595 		MUTEX_DESTROY(&softn->ipf_nat_io);
596 
597 		MUTEX_DESTROY(&softn->ipf_nat_udptq.ifq_lock);
598 		MUTEX_DESTROY(&softn->ipf_nat_udpacktq.ifq_lock);
599 		MUTEX_DESTROY(&softn->ipf_nat_icmptq.ifq_lock);
600 		MUTEX_DESTROY(&softn->ipf_nat_icmpacktq.ifq_lock);
601 		MUTEX_DESTROY(&softn->ipf_nat_iptq.ifq_lock);
602 		MUTEX_DESTROY(&softn->ipf_nat_pending.ifq_lock);
603 	}
604 
605 	return 0;
606 }
607 
608 
609 /* ------------------------------------------------------------------------ */
610 /* Function:    ipf_nat_setlock                                             */
611 /* Returns:     Nil                                                         */
612 /* Parameters:  arg(I) - pointer to soft state information                  */
613 /*              tmp(I) - new lock value                                     */
614 /*                                                                          */
615 /* Set the "lock status" of NAT to the value in tmp.                        */
616 /* ------------------------------------------------------------------------ */
617 void
618 ipf_nat_setlock(void *arg, int tmp)
619 {
620 	ipf_nat_softc_t *softn = arg;
621 
622 	softn->ipf_nat_lock = tmp;
623 }
624 
625 
626 /* ------------------------------------------------------------------------ */
627 /* Function:    ipf_nat_addrdr                                              */
628 /* Returns:     Nil                                                         */
629 /* Parameters:  n(I) - pointer to NAT rule to add                           */
630 /*                                                                          */
631 /* Adds a redirect rule to the hash table of redirect rules and the list of */
632 /* loaded NAT rules.  Updates the bitmask indicating which netmasks are in  */
633 /* use by redirect rules.                                                   */
634 /* ------------------------------------------------------------------------ */
635 static void
636 ipf_nat_addrdr(ipf_nat_softc_t *softn, ipnat_t *n)
637 {
638 	ipnat_t **np;
639 	u_32_t j;
640 	u_int hv;
641 	u_int rhv;
642 	int k;
643 
644 	if (n->in_odstatype == FRI_NORMAL) {
645 		k = count4bits(n->in_odstmsk);
646 		ipf_inet_mask_add(k, &softn->ipf_nat_rdr_mask);
647 		j = (n->in_odstaddr & n->in_odstmsk);
648 		rhv = NAT_HASH_FN(j, 0, 0xffffffff);
649 	} else {
650 		ipf_inet_mask_add(0, &softn->ipf_nat_rdr_mask);
651 		j = 0;
652 		rhv = 0;
653 	}
654 	hv = rhv % softn->ipf_nat_rdrrules_sz;
655 	np = softn->ipf_nat_rdr_rules + hv;
656 	while (*np != NULL)
657 		np = &(*np)->in_rnext;
658 	n->in_rnext = NULL;
659 	n->in_prnext = np;
660 	n->in_hv[0] = hv;
661 	n->in_use++;
662 	*np = n;
663 }
664 
665 
666 /* ------------------------------------------------------------------------ */
667 /* Function:    ipf_nat_addmap                                              */
668 /* Returns:     Nil                                                         */
669 /* Parameters:  n(I) - pointer to NAT rule to add                           */
670 /*                                                                          */
671 /* Adds a NAT map rule to the hash table of rules and the list of  loaded   */
672 /* NAT rules.  Updates the bitmask indicating which netmasks are in use by  */
673 /* redirect rules.                                                          */
674 /* ------------------------------------------------------------------------ */
675 static void
676 ipf_nat_addmap(ipf_nat_softc_t *softn, ipnat_t *n)
677 {
678 	ipnat_t **np;
679 	u_32_t j;
680 	u_int hv;
681 	u_int rhv;
682 	int k;
683 
684 	if (n->in_osrcatype == FRI_NORMAL) {
685 		k = count4bits(n->in_osrcmsk);
686 		ipf_inet_mask_add(k, &softn->ipf_nat_map_mask);
687 		j = (n->in_osrcaddr & n->in_osrcmsk);
688 		rhv = NAT_HASH_FN(j, 0, 0xffffffff);
689 	} else {
690 		ipf_inet_mask_add(0, &softn->ipf_nat_map_mask);
691 		j = 0;
692 		rhv = 0;
693 	}
694 	hv = rhv % softn->ipf_nat_maprules_sz;
695 	np = softn->ipf_nat_map_rules + hv;
696 	while (*np != NULL)
697 		np = &(*np)->in_mnext;
698 	n->in_mnext = NULL;
699 	n->in_pmnext = np;
700 	n->in_hv[1] = rhv;
701 	n->in_use++;
702 	*np = n;
703 }
704 
705 
706 /* ------------------------------------------------------------------------ */
707 /* Function:    ipf_nat_delrdr                                              */
708 /* Returns:     Nil                                                         */
709 /* Parameters:  n(I) - pointer to NAT rule to delete                        */
710 /*                                                                          */
711 /* Removes a redirect rule from the hash table of redirect rules.           */
712 /* ------------------------------------------------------------------------ */
713 void
714 ipf_nat_delrdr(ipf_nat_softc_t *softn, ipnat_t *n)
715 {
716 	if (n->in_odstatype == FRI_NORMAL) {
717 		int k = count4bits(n->in_odstmsk);
718 		ipf_inet_mask_del(k, &softn->ipf_nat_rdr_mask);
719 	} else {
720 		ipf_inet_mask_del(0, &softn->ipf_nat_rdr_mask);
721 	}
722 	if (n->in_rnext)
723 		n->in_rnext->in_prnext = n->in_prnext;
724 	*n->in_prnext = n->in_rnext;
725 	n->in_use--;
726 }
727 
728 
729 /* ------------------------------------------------------------------------ */
730 /* Function:    ipf_nat_delmap                                              */
731 /* Returns:     Nil                                                         */
732 /* Parameters:  n(I) - pointer to NAT rule to delete                        */
733 /*                                                                          */
734 /* Removes a NAT map rule from the hash table of NAT map rules.             */
735 /* ------------------------------------------------------------------------ */
736 void
737 ipf_nat_delmap(ipf_nat_softc_t *softn, ipnat_t *n)
738 {
739 	if (n->in_osrcatype == FRI_NORMAL) {
740 		int k = count4bits(n->in_osrcmsk);
741 		ipf_inet_mask_del(k, &softn->ipf_nat_map_mask);
742 	} else {
743 		ipf_inet_mask_del(0, &softn->ipf_nat_map_mask);
744 	}
745 	if (n->in_mnext != NULL)
746 		n->in_mnext->in_pmnext = n->in_pmnext;
747 	*n->in_pmnext = n->in_mnext;
748 	n->in_use--;
749 }
750 
751 
752 /* ------------------------------------------------------------------------ */
753 /* Function:    ipf_nat_hostmap                                             */
754 /* Returns:     struct hostmap* - NULL if no hostmap could be created,      */
755 /*                                else a pointer to the hostmapping to use  */
756 /* Parameters:  np(I)   - pointer to NAT rule                               */
757 /*              real(I) - real IP address                                   */
758 /*              map(I)  - mapped IP address                                 */
759 /*              port(I) - destination port number                           */
760 /* Write Locks: ipf_nat                                                     */
761 /*                                                                          */
762 /* Check if an ip address has already been allocated for a given mapping    */
763 /* that is not doing port based translation.  If is not yet allocated, then */
764 /* create a new entry if a non-NULL NAT rule pointer has been supplied.     */
765 /* ------------------------------------------------------------------------ */
766 static struct hostmap *
767 ipf_nat_hostmap(ipf_nat_softc_t *softn, ipnat_t *np, struct in_addr src,
768     struct in_addr dst, struct in_addr map, u_32_t port)
769 {
770 	hostmap_t *hm;
771 	u_int hv, rhv;
772 
773 	hv = (src.s_addr ^ dst.s_addr);
774 	hv += src.s_addr;
775 	hv += dst.s_addr;
776 	rhv = hv;
777 	hv %= softn->ipf_nat_hostmap_sz;
778 	for (hm = softn->ipf_hm_maptable[hv]; hm; hm = hm->hm_hnext)
779 		if ((hm->hm_osrcip.s_addr == src.s_addr) &&
780 		    (hm->hm_odstip.s_addr == dst.s_addr) &&
781 		    ((np == NULL) || (np == hm->hm_ipnat)) &&
782 		    ((port == 0) || (port == hm->hm_port))) {
783 			softn->ipf_nat_stats.ns_hm_addref++;
784 			hm->hm_ref++;
785 			return hm;
786 		}
787 
788 	if (np == NULL) {
789 		softn->ipf_nat_stats.ns_hm_nullnp++;
790 		return NULL;
791 	}
792 
793 	KMALLOC(hm, hostmap_t *);
794 	if (hm) {
795 		hm->hm_next = softn->ipf_hm_maplist;
796 		hm->hm_pnext = &softn->ipf_hm_maplist;
797 		if (softn->ipf_hm_maplist != NULL)
798 			softn->ipf_hm_maplist->hm_pnext = &hm->hm_next;
799 		softn->ipf_hm_maplist = hm;
800 		hm->hm_hnext = softn->ipf_hm_maptable[hv];
801 		hm->hm_phnext = softn->ipf_hm_maptable + hv;
802 		if (softn->ipf_hm_maptable[hv] != NULL)
803 			softn->ipf_hm_maptable[hv]->hm_phnext = &hm->hm_hnext;
804 		softn->ipf_hm_maptable[hv] = hm;
805 		hm->hm_ipnat = np;
806 		np->in_use++;
807 		hm->hm_osrcip = src;
808 		hm->hm_odstip = dst;
809 		hm->hm_nsrcip = map;
810 		hm->hm_ndstip.s_addr = 0;
811 		hm->hm_ref = 1;
812 		hm->hm_port = port;
813 		hm->hm_hv = rhv;
814 		hm->hm_v = 4;
815 		softn->ipf_nat_stats.ns_hm_new++;
816 	} else {
817 		softn->ipf_nat_stats.ns_hm_newfail++;
818 	}
819 	return hm;
820 }
821 
822 
823 /* ------------------------------------------------------------------------ */
824 /* Function:    ipf_nat_hostmapdel                                          */
825 /* Returns:     Nil                                                         */
826 /* Parameters:  hmp(I) - pointer to hostmap structure pointer               */
827 /* Write Locks: ipf_nat                                                     */
828 /*                                                                          */
829 /* Decrement the references to this hostmap structure by one.  If this      */
830 /* reaches zero then remove it and free it.                                 */
831 /* ------------------------------------------------------------------------ */
832 void
833 ipf_nat_hostmapdel(ipf_main_softc_t *softc, struct hostmap **hmp)
834 {
835 	struct hostmap *hm;
836 
837 	hm = *hmp;
838 	*hmp = NULL;
839 
840 	hm->hm_ref--;
841 	if (hm->hm_ref == 0) {
842 		ipf_nat_rule_deref(softc, &hm->hm_ipnat);
843 		if (hm->hm_hnext)
844 			hm->hm_hnext->hm_phnext = hm->hm_phnext;
845 		*hm->hm_phnext = hm->hm_hnext;
846 		if (hm->hm_next)
847 			hm->hm_next->hm_pnext = hm->hm_pnext;
848 		*hm->hm_pnext = hm->hm_next;
849 		KFREE(hm);
850 	}
851 }
852 
853 
854 /* ------------------------------------------------------------------------ */
855 /* Function:    ipf_fix_outcksum                                            */
856 /* Returns:     Nil                                                         */
857 /* Parameters:  fin(I) - pointer to packet information                      */
858 /*              sp(I)  - location of 16bit checksum to update               */
859 /*              n((I)  - amount to adjust checksum by                       */
860 /*                                                                          */
861 /* Adjusts the 16bit checksum by "n" for packets going out.                 */
862 /* ------------------------------------------------------------------------ */
863 void
864 ipf_fix_outcksum(int cksum, u_short *sp, u_32_t n, u_32_t partial)
865 {
866 	u_short sumshort;
867 	u_32_t sum1;
868 
869 	if (n == 0)
870 		return;
871 
872 	if (cksum == 4) {
873 		*sp = 0;
874 		return;
875 	}
876 	if (cksum == 2) {
877 		sum1 = partial;
878 		sum1 = (sum1 & 0xffff) + (sum1 >> 16);
879 		*sp = htons(sum1);
880 		return;
881 	}
882 	sum1 = (~ntohs(*sp)) & 0xffff;
883 	sum1 += (n);
884 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
885 	/* Again */
886 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
887 	sumshort = ~(u_short)sum1;
888 	*(sp) = htons(sumshort);
889 }
890 
891 
892 /* ------------------------------------------------------------------------ */
893 /* Function:    ipf_fix_incksum                                             */
894 /* Returns:     Nil                                                         */
895 /* Parameters:  fin(I) - pointer to packet information                      */
896 /*              sp(I)  - location of 16bit checksum to update               */
897 /*              n((I)  - amount to adjust checksum by                       */
898 /*                                                                          */
899 /* Adjusts the 16bit checksum by "n" for packets going in.                  */
900 /* ------------------------------------------------------------------------ */
901 void
902 ipf_fix_incksum(int cksum, u_short *sp, u_32_t n, u_32_t partial)
903 {
904 	u_short sumshort;
905 	u_32_t sum1;
906 
907 	if (n == 0)
908 		return;
909 
910 	if (cksum == 4) {
911 		*sp = 0;
912 		return;
913 	}
914 	if (cksum == 2) {
915 		sum1 = partial;
916 		sum1 = (sum1 & 0xffff) + (sum1 >> 16);
917 		*sp = htons(sum1);
918 		return;
919 	}
920 
921 	sum1 = (~ntohs(*sp)) & 0xffff;
922 	sum1 += ~(n) & 0xffff;
923 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
924 	/* Again */
925 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
926 	sumshort = ~(u_short)sum1;
927 	*(sp) = htons(sumshort);
928 }
929 
930 
931 /* ------------------------------------------------------------------------ */
932 /* Function:    ipf_fix_datacksum                                           */
933 /* Returns:     Nil                                                         */
934 /* Parameters:  sp(I)  - location of 16bit checksum to update               */
935 /*              n((I)  - amount to adjust checksum by                       */
936 /*                                                                          */
937 /* Fix_datacksum is used *only* for the adjustments of checksums in the     */
938 /* data section of an IP packet.                                            */
939 /*                                                                          */
940 /* The only situation in which you need to do this is when NAT'ing an       */
941 /* ICMP error message. Such a message, contains in its body the IP header   */
942 /* of the original IP packet, that causes the error.                        */
943 /*                                                                          */
944 /* You can't use fix_incksum or fix_outcksum in that case, because for the  */
945 /* kernel the data section of the ICMP error is just data, and no special   */
946 /* processing like hardware cksum or ntohs processing have been done by the */
947 /* kernel on the data section.                                              */
948 /* ------------------------------------------------------------------------ */
949 void
950 ipf_fix_datacksum(u_short *sp, u_32_t n)
951 {
952 	u_short sumshort;
953 	u_32_t sum1;
954 
955 	if (n == 0)
956 		return;
957 
958 	sum1 = (~ntohs(*sp)) & 0xffff;
959 	sum1 += (n);
960 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
961 	/* Again */
962 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
963 	sumshort = ~(u_short)sum1;
964 	*(sp) = htons(sumshort);
965 }
966 
967 
968 /* ------------------------------------------------------------------------ */
969 /* Function:    ipf_nat_ioctl                                               */
970 /* Returns:     int - 0 == success, != 0 == failure                         */
971 /* Parameters:  softc(I) - pointer to soft context main structure           */
972 /*              data(I)  - pointer to ioctl data                            */
973 /*              cmd(I)   - ioctl command integer                            */
974 /*              mode(I)  - file mode bits used with open                    */
975 /*              uid(I)   - uid of calling process                           */
976 /*              ctx(I)   - pointer used as key for finding context          */
977 /*                                                                          */
978 /* Processes an ioctl call made to operate on the IP Filter NAT device.     */
979 /* ------------------------------------------------------------------------ */
980 int
981 ipf_nat_ioctl(ipf_main_softc_t *softc, void *data, ioctlcmd_t cmd, int mode,
982    int uid, void *ctx)
983 {
984 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
985 	int error = 0, ret, arg, getlock;
986 	ipnat_t *nat, *nt, *n;
987 	ipnat_t natd;
988 	SPL_INT(s);
989 
990 #if BSD_GE_YEAR(199306) && defined(_KERNEL)
991 # if NETBSD_GE_REV(399002000)
992 	if ((mode & FWRITE) &&
993 	     kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_FIREWALL,
994 				     KAUTH_REQ_NETWORK_FIREWALL_FW,
995 				     NULL, NULL, NULL))
996 # else
997 #  if defined(__FreeBSD_version) && (__FreeBSD_version >= 500034)
998 	if (securelevel_ge(curthread->td_ucred, 3) && (mode & FWRITE))
999 #  else
1000 	if ((securelevel >= 3) && (mode & FWRITE))
1001 #  endif
1002 # endif
1003 	{
1004 		IPFERROR(60001);
1005 		return EPERM;
1006 	}
1007 #endif
1008 
1009 #if defined(__osf__) && defined(_KERNEL)
1010 	getlock = 0;
1011 #else
1012 	getlock = (mode & NAT_LOCKHELD) ? 0 : 1;
1013 #endif
1014 
1015 	n = NULL;
1016 	nt = NULL;
1017 	nat = NULL;
1018 
1019 	if ((cmd == (ioctlcmd_t)SIOCADNAT) || (cmd == (ioctlcmd_t)SIOCRMNAT) ||
1020 	    (cmd == (ioctlcmd_t)SIOCPURGENAT)) {
1021 		if (mode & NAT_SYSSPACE) {
1022 			bcopy(data, (char *)&natd, sizeof(natd));
1023 			nat = &natd;
1024 			error = 0;
1025 		} else {
1026 			bzero(&natd, sizeof(natd));
1027 			error = ipf_inobj(softc, data, NULL, &natd,
1028 					  IPFOBJ_IPNAT);
1029 			if (error != 0)
1030 				goto done;
1031 
1032 			if (natd.in_size < sizeof(ipnat_t)) {
1033 				error = EINVAL;
1034 				goto done;
1035 			}
1036 			KMALLOCS(nt, ipnat_t *, natd.in_size);
1037 			if (nt == NULL) {
1038 				IPFERROR(60070);
1039 				error = ENOMEM;
1040 				goto done;
1041 			}
1042 			bzero(nt, natd.in_size);
1043 			error = ipf_inobjsz(softc, data, nt, IPFOBJ_IPNAT,
1044 					    natd.in_size);
1045 			if (error)
1046 				goto done;
1047 			nat = nt;
1048 		}
1049 
1050 		/*
1051 		 * For add/delete, look to see if the NAT entry is
1052 		 * already present
1053 		 */
1054 		nat->in_flags &= IPN_USERFLAGS;
1055 		if ((nat->in_redir & NAT_MAPBLK) == 0) {
1056 			if (nat->in_osrcatype == FRI_NORMAL ||
1057 			    nat->in_osrcatype == FRI_NONE)
1058 				nat->in_osrcaddr &= nat->in_osrcmsk;
1059 			if (nat->in_odstatype == FRI_NORMAL ||
1060 			    nat->in_odstatype == FRI_NONE)
1061 				nat->in_odstaddr &= nat->in_odstmsk;
1062 			if ((nat->in_flags & (IPN_SPLIT|IPN_SIPRANGE)) == 0) {
1063 				if (nat->in_nsrcatype == FRI_NORMAL)
1064 					nat->in_nsrcaddr &= nat->in_nsrcmsk;
1065 				if (nat->in_ndstatype == FRI_NORMAL)
1066 					nat->in_ndstaddr &= nat->in_ndstmsk;
1067 			}
1068 		}
1069 
1070 		error = ipf_nat_rule_init(softc, softn, nat);
1071 		if (error != 0)
1072 			goto done;
1073 
1074 		MUTEX_ENTER(&softn->ipf_nat_io);
1075 		for (n = softn->ipf_nat_list; n != NULL; n = n->in_next)
1076 			if (ipf_nat_cmp_rules(nat, n) == 0)
1077 				break;
1078 	}
1079 
1080 	switch (cmd)
1081 	{
1082 #ifdef  IPFILTER_LOG
1083 	case SIOCIPFFB :
1084 	{
1085 		int tmp;
1086 
1087 		if (!(mode & FWRITE)) {
1088 			IPFERROR(60002);
1089 			error = EPERM;
1090 		} else {
1091 			tmp = ipf_log_clear(softc, IPL_LOGNAT);
1092 			error = BCOPYOUT(&tmp, data, sizeof(tmp));
1093 			if (error != 0) {
1094 				IPFERROR(60057);
1095 				error = EFAULT;
1096 			}
1097 		}
1098 		break;
1099 	}
1100 
1101 	case SIOCSETLG :
1102 		if (!(mode & FWRITE)) {
1103 			IPFERROR(60003);
1104 			error = EPERM;
1105 		} else {
1106 			error = BCOPYIN(data, &softn->ipf_nat_logging,
1107 					sizeof(softn->ipf_nat_logging));
1108 			if (error != 0)
1109 				error = EFAULT;
1110 		}
1111 		break;
1112 
1113 	case SIOCGETLG :
1114 		error = BCOPYOUT(&softn->ipf_nat_logging, data,
1115 				 sizeof(softn->ipf_nat_logging));
1116 		if (error != 0) {
1117 			IPFERROR(60004);
1118 			error = EFAULT;
1119 		}
1120 		break;
1121 
1122 	case FIONREAD :
1123 		arg = ipf_log_bytesused(softc, IPL_LOGNAT);
1124 		error = BCOPYOUT(&arg, data, sizeof(arg));
1125 		if (error != 0) {
1126 			IPFERROR(60005);
1127 			error = EFAULT;
1128 		}
1129 		break;
1130 #endif
1131 	case SIOCADNAT :
1132 		if (!(mode & FWRITE)) {
1133 			IPFERROR(60006);
1134 			error = EPERM;
1135 		} else if (n != NULL) {
1136 			natd.in_flineno = n->in_flineno;
1137 			(void) ipf_outobj(softc, data, &natd, IPFOBJ_IPNAT);
1138 			IPFERROR(60007);
1139 			error = EEXIST;
1140 		} else if (nt == NULL) {
1141 			IPFERROR(60008);
1142 			error = ENOMEM;
1143 		}
1144 		if (error != 0) {
1145 			MUTEX_EXIT(&softn->ipf_nat_io);
1146 			break;
1147 		}
1148 		if (nat != nt)
1149 			bcopy((char *)nat, (char *)nt, sizeof(*n));
1150 		error = ipf_nat_siocaddnat(softc, softn, nt, getlock);
1151 		MUTEX_EXIT(&softn->ipf_nat_io);
1152 		if (error == 0) {
1153 			nat = NULL;
1154 			nt = NULL;
1155 		}
1156 		break;
1157 
1158 	case SIOCRMNAT :
1159 	case SIOCPURGENAT :
1160 		if (!(mode & FWRITE)) {
1161 			IPFERROR(60009);
1162 			error = EPERM;
1163 			n = NULL;
1164 		} else if (n == NULL) {
1165 			IPFERROR(60010);
1166 			error = ESRCH;
1167 		}
1168 
1169 		if (error != 0) {
1170 			MUTEX_EXIT(&softn->ipf_nat_io);
1171 			break;
1172 		}
1173 		if (cmd == (ioctlcmd_t)SIOCPURGENAT) {
1174 			error = ipf_outobjsz(softc, data, n, IPFOBJ_IPNAT,
1175 					     n->in_size);
1176 			if (error) {
1177 				MUTEX_EXIT(&softn->ipf_nat_io);
1178 				goto done;
1179 			}
1180 			n->in_flags |= IPN_PURGE;
1181 		}
1182 		ipf_nat_siocdelnat(softc, softn, n, getlock);
1183 
1184 		MUTEX_EXIT(&softn->ipf_nat_io);
1185 		n = NULL;
1186 		break;
1187 
1188 	case SIOCGNATS :
1189 	    {
1190 		natstat_t *nsp = &softn->ipf_nat_stats;
1191 
1192 		nsp->ns_side[0].ns_table = softn->ipf_nat_table[0];
1193 		nsp->ns_side[1].ns_table = softn->ipf_nat_table[1];
1194 		nsp->ns_list = softn->ipf_nat_list;
1195 		nsp->ns_maptable = softn->ipf_hm_maptable;
1196 		nsp->ns_maplist = softn->ipf_hm_maplist;
1197 		nsp->ns_nattab_sz = softn->ipf_nat_table_sz;
1198 		nsp->ns_nattab_max = softn->ipf_nat_table_max;
1199 		nsp->ns_rultab_sz = softn->ipf_nat_maprules_sz;
1200 		nsp->ns_rdrtab_sz = softn->ipf_nat_rdrrules_sz;
1201 		nsp->ns_hostmap_sz = softn->ipf_nat_hostmap_sz;
1202 		nsp->ns_instances = softn->ipf_nat_instances;
1203 		nsp->ns_ticks = softc->ipf_ticks;
1204 #ifdef IPFILTER_LOGGING
1205 		nsp->ns_log_ok = ipf_log_logok(softc, IPF_LOGNAT);
1206 		nsp->ns_log_fail = ipf_log_failures(softc, IPF_LOGNAT);
1207 #else
1208 		nsp->ns_log_ok = 0;
1209 		nsp->ns_log_fail = 0;
1210 #endif
1211 		error = ipf_outobj(softc, data, nsp, IPFOBJ_NATSTAT);
1212 		break;
1213 	    }
1214 
1215 	case SIOCGNATL :
1216 	    {
1217 		natlookup_t nl;
1218 
1219 		error = ipf_inobj(softc, data, NULL, &nl, IPFOBJ_NATLOOKUP);
1220 		if (error == 0) {
1221 			void *ptr;
1222 
1223 			if (getlock) {
1224 				READ_ENTER(&softc->ipf_nat);
1225 			}
1226 
1227 			switch (nl.nl_v)
1228 			{
1229 			case 4 :
1230 				ptr = ipf_nat_lookupredir(&nl);
1231 				break;
1232 #ifdef USE_INET6
1233 			case 6 :
1234 				ptr = ipf_nat6_lookupredir(&nl);
1235 				break;
1236 #endif
1237 			default:
1238 				ptr = NULL;
1239 				break;
1240 			}
1241 
1242 			if (getlock) {
1243 				RWLOCK_EXIT(&softc->ipf_nat);
1244 			}
1245 			if (ptr != NULL) {
1246 				error = ipf_outobj(softc, data, &nl,
1247 						   IPFOBJ_NATLOOKUP);
1248 			} else {
1249 				IPFERROR(60011);
1250 				error = ESRCH;
1251 			}
1252 		}
1253 		break;
1254 	    }
1255 
1256 	case SIOCIPFFL :	/* old SIOCFLNAT & SIOCCNATL */
1257 		if (!(mode & FWRITE)) {
1258 			IPFERROR(60012);
1259 			error = EPERM;
1260 			break;
1261 		}
1262 		if (getlock) {
1263 			WRITE_ENTER(&softc->ipf_nat);
1264 		}
1265 
1266 		error = BCOPYIN(data, &arg, sizeof(arg));
1267 		if (error != 0) {
1268 			IPFERROR(60013);
1269 			error = EFAULT;
1270 		} else {
1271 			if (arg == 0)
1272 				ret = ipf_nat_flushtable(softc, softn);
1273 			else if (arg == 1)
1274 				ret = ipf_nat_clearlist(softc, softn);
1275 			else
1276 				ret = ipf_nat_extraflush(softc, softn, arg);
1277 			ipf_proxy_flush(softc->ipf_proxy_soft, arg);
1278 		}
1279 
1280 		if (getlock) {
1281 			RWLOCK_EXIT(&softc->ipf_nat);
1282 		}
1283 		if (error == 0) {
1284 			error = BCOPYOUT(&ret, data, sizeof(ret));
1285 		}
1286 		break;
1287 
1288 	case SIOCMATCHFLUSH :
1289 		if (!(mode & FWRITE)) {
1290 			IPFERROR(60014);
1291 			error = EPERM;
1292 			break;
1293 		}
1294 		if (getlock) {
1295 			WRITE_ENTER(&softc->ipf_nat);
1296 		}
1297 
1298 		error = ipf_nat_matchflush(softc, softn, data);
1299 
1300 		if (getlock) {
1301 			RWLOCK_EXIT(&softc->ipf_nat);
1302 		}
1303 		break;
1304 
1305 	case SIOCPROXY :
1306 		error = ipf_proxy_ioctl(softc, data, cmd, mode, ctx);
1307 		break;
1308 
1309 	case SIOCSTLCK :
1310 		if (!(mode & FWRITE)) {
1311 			IPFERROR(60015);
1312 			error = EPERM;
1313 		} else {
1314 			error = ipf_lock(data, &softn->ipf_nat_lock);
1315 		}
1316 		break;
1317 
1318 	case SIOCSTPUT :
1319 		if ((mode & FWRITE) != 0) {
1320 			error = ipf_nat_putent(softc, data, getlock);
1321 		} else {
1322 			IPFERROR(60016);
1323 			error = EACCES;
1324 		}
1325 		break;
1326 
1327 	case SIOCSTGSZ :
1328 		if (softn->ipf_nat_lock) {
1329 			error = ipf_nat_getsz(softc, data, getlock);
1330 		} else {
1331 			IPFERROR(60017);
1332 			error = EACCES;
1333 		}
1334 		break;
1335 
1336 	case SIOCSTGET :
1337 		if (softn->ipf_nat_lock) {
1338 			error = ipf_nat_getent(softc, data, getlock);
1339 		} else {
1340 			IPFERROR(60018);
1341 			error = EACCES;
1342 		}
1343 		break;
1344 
1345 	case SIOCGENITER :
1346 	    {
1347 		ipfgeniter_t iter;
1348 		ipftoken_t *token;
1349 		ipfobj_t obj;
1350 
1351 		error = ipf_inobj(softc, data, &obj, &iter, IPFOBJ_GENITER);
1352 		if (error != 0)
1353 			break;
1354 
1355 		SPL_SCHED(s);
1356 		token = ipf_token_find(softc, iter.igi_type, uid, ctx);
1357 		if (token != NULL) {
1358 			error  = ipf_nat_iterator(softc, token, &iter, &obj);
1359 			WRITE_ENTER(&softc->ipf_tokens);
1360 			ipf_token_deref(softc, token);
1361 			RWLOCK_EXIT(&softc->ipf_tokens);
1362 		}
1363 		SPL_X(s);
1364 		break;
1365 	    }
1366 
1367 	case SIOCIPFDELTOK :
1368 		error = BCOPYIN(data, &arg, sizeof(arg));
1369 		if (error == 0) {
1370 			SPL_SCHED(s);
1371 			error = ipf_token_del(softc, arg, uid, ctx);
1372 			SPL_X(s);
1373 		} else {
1374 			IPFERROR(60019);
1375 			error = EFAULT;
1376 		}
1377 		break;
1378 
1379 	case SIOCGTQTAB :
1380 		error = ipf_outobj(softc, data, softn->ipf_nat_tcptq,
1381 				   IPFOBJ_STATETQTAB);
1382 		break;
1383 
1384 	case SIOCGTABL :
1385 		error = ipf_nat_gettable(softc, softn, data);
1386 		break;
1387 
1388 	default :
1389 		IPFERROR(60020);
1390 		error = EINVAL;
1391 		break;
1392 	}
1393 done:
1394 	if (nat != NULL)
1395 		ipf_nat_rule_fini(softc, nat);
1396 	if (nt != NULL)
1397 		KFREES(nt, nt->in_size);
1398 	return error;
1399 }
1400 
1401 
1402 /* ------------------------------------------------------------------------ */
1403 /* Function:    ipf_nat_siocaddnat                                          */
1404 /* Returns:     int - 0 == success, != 0 == failure                         */
1405 /* Parameters:  softc(I) - pointer to soft context main structure           */
1406 /*              softn(I) - pointer to NAT context structure                 */
1407 /*              n(I)       - pointer to new NAT rule                        */
1408 /*              np(I)      - pointer to where to insert new NAT rule        */
1409 /*              getlock(I) - flag indicating if lock on  is held            */
1410 /* Mutex Locks: ipf_nat_io                                                  */
1411 /*                                                                          */
1412 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
1413 /* from information passed to the kernel, then add it  to the appropriate   */
1414 /* NAT rule table(s).                                                       */
1415 /* ------------------------------------------------------------------------ */
1416 static int
1417 ipf_nat_siocaddnat(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *n,
1418 	int getlock)
1419 {
1420 	int error = 0;
1421 
1422 	if (ipf_nat_resolverule(softc, n) != 0) {
1423 		IPFERROR(60022);
1424 		return ENOENT;
1425 	}
1426 
1427 	if ((n->in_age[0] == 0) && (n->in_age[1] != 0)) {
1428 		IPFERROR(60023);
1429 		return EINVAL;
1430 	}
1431 
1432 	if (n->in_redir == (NAT_DIVERTUDP|NAT_MAP)) {
1433 		/*
1434 		 * Prerecord whether or not the destination of the divert
1435 		 * is local or not to the interface the packet is going
1436 		 * to be sent out.
1437 		 */
1438 		n->in_dlocal = ipf_deliverlocal(softc, n->in_v[1],
1439 						n->in_ifps[1], &n->in_ndstip6);
1440 	}
1441 
1442 	if (getlock) {
1443 		WRITE_ENTER(&softc->ipf_nat);
1444 	}
1445 	n->in_next = NULL;
1446 	n->in_pnext = softn->ipf_nat_list_tail;
1447 	*n->in_pnext = n;
1448 	softn->ipf_nat_list_tail = &n->in_next;
1449 	n->in_use++;
1450 
1451 	if (n->in_redir & NAT_REDIRECT) {
1452 		n->in_flags &= ~IPN_NOTDST;
1453 		switch (n->in_v[0])
1454 		{
1455 		case 4 :
1456 			ipf_nat_addrdr(softn, n);
1457 			break;
1458 #ifdef USE_INET6
1459 		case 6 :
1460 			ipf_nat6_addrdr(softn, n);
1461 			break;
1462 #endif
1463 		default :
1464 			break;
1465 		}
1466 		ATOMIC_INC32(softn->ipf_nat_stats.ns_rules_rdr);
1467 	}
1468 
1469 	if (n->in_redir & (NAT_MAP|NAT_MAPBLK)) {
1470 		n->in_flags &= ~IPN_NOTSRC;
1471 		switch (n->in_v[0])
1472 		{
1473 		case 4 :
1474 			ipf_nat_addmap(softn, n);
1475 			break;
1476 #ifdef USE_INET6
1477 		case 6 :
1478 			ipf_nat6_addmap(softn, n);
1479 			break;
1480 #endif
1481 		default :
1482 			break;
1483 		}
1484 		ATOMIC_INC32(softn->ipf_nat_stats.ns_rules_map);
1485 	}
1486 
1487 	if (n->in_age[0] != 0)
1488 		n->in_tqehead[0] = ipf_addtimeoutqueue(softc,
1489 						       &softn->ipf_nat_utqe,
1490 						       n->in_age[0]);
1491 
1492 	if (n->in_age[1] != 0)
1493 		n->in_tqehead[1] = ipf_addtimeoutqueue(softc,
1494 						       &softn->ipf_nat_utqe,
1495 						       n->in_age[1]);
1496 
1497 	MUTEX_INIT(&n->in_lock, "ipnat rule lock");
1498 
1499 	n = NULL;
1500 	ATOMIC_INC32(softn->ipf_nat_stats.ns_rules);
1501 #if SOLARIS && !defined(INSTANCES)
1502 	pfil_delayed_copy = 0;
1503 #endif
1504 	if (getlock) {
1505 		RWLOCK_EXIT(&softc->ipf_nat);			/* WRITE */
1506 	}
1507 
1508 	return error;
1509 }
1510 
1511 
1512 /* ------------------------------------------------------------------------ */
1513 /* Function:    ipf_nat_ruleaddrinit                                        */
1514 /* Parameters:  softc(I) - pointer to soft context main structure           */
1515 /*              softn(I) - pointer to NAT context structure                 */
1516 /*              n(I)     - pointer to NAT rule                              */
1517 /*                                                                          */
1518 /* Initialise all of the NAT address structures in a NAT rule.              */
1519 /* ------------------------------------------------------------------------ */
1520 static int
1521 ipf_nat_ruleaddrinit(ipf_main_softc_t *softc, ipf_nat_softc_t *softn,
1522     ipnat_t *n)
1523 {
1524 	int idx, error;
1525 
1526 	if ((n->in_ndst.na_atype == FRI_LOOKUP) &&
1527 	    (n->in_ndst.na_type != IPLT_DSTLIST)) {
1528 		IPFERROR(60071);
1529 		return EINVAL;
1530 	}
1531 	if ((n->in_nsrc.na_atype == FRI_LOOKUP) &&
1532 	    (n->in_nsrc.na_type != IPLT_DSTLIST)) {
1533 		IPFERROR(60069);
1534 		return EINVAL;
1535 	}
1536 
1537 	if (n->in_redir == NAT_BIMAP) {
1538 		n->in_ndstaddr = n->in_osrcaddr;
1539 		n->in_ndstmsk = n->in_osrcmsk;
1540 		n->in_odstaddr = n->in_nsrcaddr;
1541 		n->in_odstmsk = n->in_nsrcmsk;
1542 
1543 	}
1544 
1545 	if (n->in_redir & NAT_REDIRECT)
1546 		idx = 1;
1547 	else
1548 		idx = 0;
1549 	/*
1550 	 * Initialise all of the address fields.
1551 	 */
1552 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_osrc, 1,
1553 				     n->in_ifps[idx]);
1554 	if (error != 0)
1555 		return error;
1556 
1557 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_odst, 1,
1558 				     n->in_ifps[idx]);
1559 	if (error != 0)
1560 		return error;
1561 
1562 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_nsrc, 1,
1563 				     n->in_ifps[idx]);
1564 	if (error != 0)
1565 		return error;
1566 
1567 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_ndst, 1,
1568 				     n->in_ifps[idx]);
1569 	if (error != 0)
1570 		return error;
1571 
1572 	if (n->in_redir & NAT_DIVERTUDP)
1573 		ipf_nat_builddivertmp(softn, n);
1574 
1575 	return 0;
1576 }
1577 
1578 
1579 /* ------------------------------------------------------------------------ */
1580 /* Function:    ipf_nat_resolvrule                                          */
1581 /* Returns:     Nil                                                         */
1582 /* Parameters:  softc(I) - pointer to soft context main structure           */
1583 /*              n(I)     - pointer to NAT rule                              */
1584 /*                                                                          */
1585 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
1586 /* from information passed to the kernel, then add it  to the appropriate   */
1587 /* NAT rule table(s).                                                       */
1588 /* ------------------------------------------------------------------------ */
1589 static int
1590 ipf_nat_resolverule(ipf_main_softc_t *softc, ipnat_t *n)
1591 {
1592 	char *base;
1593 
1594 	base = n->in_names;
1595 
1596 	n->in_ifps[0] = ipf_resolvenic(softc, base + n->in_ifnames[0],
1597 				       n->in_v[0]);
1598 
1599 	if (n->in_ifnames[1] == -1) {
1600 		n->in_ifnames[1] = n->in_ifnames[0];
1601 		n->in_ifps[1] = n->in_ifps[0];
1602 	} else {
1603 		n->in_ifps[1] = ipf_resolvenic(softc, base + n->in_ifnames[1],
1604 					       n->in_v[1]);
1605 	}
1606 
1607 	if (n->in_plabel != -1) {
1608 		if (n->in_redir & NAT_REDIRECT)
1609 			n->in_apr = ipf_proxy_lookup(softc->ipf_proxy_soft,
1610 						     n->in_pr[0],
1611 						     base + n->in_plabel);
1612 		else
1613 			n->in_apr = ipf_proxy_lookup(softc->ipf_proxy_soft,
1614 						     n->in_pr[1],
1615 						     base + n->in_plabel);
1616 		if (n->in_apr == NULL)
1617 			return -1;
1618 	}
1619 	return 0;
1620 }
1621 
1622 
1623 /* ------------------------------------------------------------------------ */
1624 /* Function:    ipf_nat_siocdelnat                                          */
1625 /* Returns:     int - 0 == success, != 0 == failure                         */
1626 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1627 /*              softn(I)   - pointer to NAT context structure               */
1628 /*              n(I)       - pointer to new NAT rule                        */
1629 /*              getlock(I) - flag indicating if lock on  is held            */
1630 /* Mutex Locks: ipf_nat_io                                                  */
1631 /*                                                                          */
1632 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
1633 /* from information passed to the kernel, then add it  to the appropriate   */
1634 /* NAT rule table(s).                                                       */
1635 /* ------------------------------------------------------------------------ */
1636 static void
1637 ipf_nat_siocdelnat(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *n,
1638 	int getlock)
1639 {
1640 #ifdef IPF_NAT6
1641 	int i;
1642 #endif
1643 
1644 	if (getlock) {
1645 		WRITE_ENTER(&softc->ipf_nat);
1646 	}
1647 
1648 	ipf_nat_delrule(softc, softn, n, 1);
1649 
1650 	if (getlock) {
1651 		RWLOCK_EXIT(&softc->ipf_nat);			/* READ/WRITE */
1652 	}
1653 }
1654 
1655 
1656 /* ------------------------------------------------------------------------ */
1657 /* Function:    ipf_nat_getsz                                               */
1658 /* Returns:     int - 0 == success, != 0 is the error value.                */
1659 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1660 /*              data(I)    - pointer to natget structure with kernel        */
1661 /*                           pointer get the size of.                       */
1662 /*              getlock(I) - flag indicating whether or not the caller      */
1663 /*                           holds a lock on ipf_nat                        */
1664 /*                                                                          */
1665 /* Handle SIOCSTGSZ.                                                        */
1666 /* Return the size of the nat list entry to be copied back to user space.   */
1667 /* The size of the entry is stored in the ng_sz field and the enture natget */
1668 /* structure is copied back to the user.                                    */
1669 /* ------------------------------------------------------------------------ */
1670 static int
1671 ipf_nat_getsz(ipf_main_softc_t *softc, void *data, int getlock)
1672 {
1673 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1674 	ap_session_t *aps;
1675 	nat_t *nat, *n;
1676 	natget_t ng;
1677 	int error;
1678 
1679 	error = BCOPYIN(data, &ng, sizeof(ng));
1680 	if (error != 0) {
1681 		IPFERROR(60024);
1682 		return EFAULT;
1683 	}
1684 
1685 	if (getlock) {
1686 		READ_ENTER(&softc->ipf_nat);
1687 	}
1688 
1689 	nat = ng.ng_ptr;
1690 	if (!nat) {
1691 		nat = softn->ipf_nat_instances;
1692 		ng.ng_sz = 0;
1693 		/*
1694 		 * Empty list so the size returned is 0.  Simple.
1695 		 */
1696 		if (nat == NULL) {
1697 			if (getlock) {
1698 				RWLOCK_EXIT(&softc->ipf_nat);
1699 			}
1700 			error = BCOPYOUT(&ng, data, sizeof(ng));
1701 			if (error != 0) {
1702 				IPFERROR(60025);
1703 				return EFAULT;
1704 			}
1705 			return 0;
1706 		}
1707 	} else {
1708 		/*
1709 		 * Make sure the pointer we're copying from exists in the
1710 		 * current list of entries.  Security precaution to prevent
1711 		 * copying of random kernel data.
1712 		 */
1713 		for (n = softn->ipf_nat_instances; n; n = n->nat_next)
1714 			if (n == nat)
1715 				break;
1716 		if (n == NULL) {
1717 			if (getlock) {
1718 				RWLOCK_EXIT(&softc->ipf_nat);
1719 			}
1720 			IPFERROR(60026);
1721 			return ESRCH;
1722 		}
1723 	}
1724 
1725 	/*
1726 	 * Incluse any space required for proxy data structures.
1727 	 */
1728 	ng.ng_sz = sizeof(nat_save_t);
1729 	aps = nat->nat_aps;
1730 	if (aps != NULL) {
1731 		ng.ng_sz += sizeof(ap_session_t) - 4;
1732 		if (aps->aps_data != 0)
1733 			ng.ng_sz += aps->aps_psiz;
1734 	}
1735 	if (getlock) {
1736 		RWLOCK_EXIT(&softc->ipf_nat);
1737 	}
1738 
1739 	error = BCOPYOUT(&ng, data, sizeof(ng));
1740 	if (error != 0) {
1741 		IPFERROR(60027);
1742 		return EFAULT;
1743 	}
1744 	return 0;
1745 }
1746 
1747 
1748 /* ------------------------------------------------------------------------ */
1749 /* Function:    ipf_nat_getent                                              */
1750 /* Returns:     int - 0 == success, != 0 is the error value.                */
1751 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1752 /*              data(I)    - pointer to natget structure with kernel pointer*/
1753 /*                           to NAT structure to copy out.                  */
1754 /*              getlock(I) - flag indicating whether or not the caller      */
1755 /*                           holds a lock on ipf_nat                        */
1756 /*                                                                          */
1757 /* Handle SIOCSTGET.                                                        */
1758 /* Copies out NAT entry to user space.  Any additional data held for a      */
1759 /* proxy is also copied, as to is the NAT rule which was responsible for it */
1760 /* ------------------------------------------------------------------------ */
1761 static int
1762 ipf_nat_getent(ipf_main_softc_t *softc, void *data, int getlock)
1763 {
1764 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1765 	int error, outsize;
1766 	ap_session_t *aps;
1767 	nat_save_t *ipn, ipns;
1768 	nat_t *n, *nat;
1769 
1770 	error = ipf_inobj(softc, data, NULL, &ipns, IPFOBJ_NATSAVE);
1771 	if (error != 0)
1772 		return error;
1773 
1774 	if ((ipns.ipn_dsize < sizeof(ipns)) || (ipns.ipn_dsize > 81920)) {
1775 		IPFERROR(60028);
1776 		return EINVAL;
1777 	}
1778 
1779 	KMALLOCS(ipn, nat_save_t *, ipns.ipn_dsize);
1780 	if (ipn == NULL) {
1781 		IPFERROR(60029);
1782 		return ENOMEM;
1783 	}
1784 
1785 	if (getlock) {
1786 		READ_ENTER(&softc->ipf_nat);
1787 	}
1788 
1789 	ipn->ipn_dsize = ipns.ipn_dsize;
1790 	nat = ipns.ipn_next;
1791 	if (nat == NULL) {
1792 		nat = softn->ipf_nat_instances;
1793 		if (nat == NULL) {
1794 			if (softn->ipf_nat_instances == NULL) {
1795 				IPFERROR(60030);
1796 				error = ENOENT;
1797 			}
1798 			goto finished;
1799 		}
1800 	} else {
1801 		/*
1802 		 * Make sure the pointer we're copying from exists in the
1803 		 * current list of entries.  Security precaution to prevent
1804 		 * copying of random kernel data.
1805 		 */
1806 		for (n = softn->ipf_nat_instances; n; n = n->nat_next)
1807 			if (n == nat)
1808 				break;
1809 		if (n == NULL) {
1810 			IPFERROR(60031);
1811 			error = ESRCH;
1812 			goto finished;
1813 		}
1814 	}
1815 	ipn->ipn_next = nat->nat_next;
1816 
1817 	/*
1818 	 * Copy the NAT structure.
1819 	 */
1820 	bcopy((char *)nat, &ipn->ipn_nat, sizeof(*nat));
1821 
1822 	/*
1823 	 * If we have a pointer to the NAT rule it belongs to, save that too.
1824 	 */
1825 	if (nat->nat_ptr != NULL)
1826 		bcopy((char *)nat->nat_ptr, (char *)&ipn->ipn_ipnat,
1827 		      ipn->ipn_ipnat.in_size);
1828 
1829 	/*
1830 	 * If we also know the NAT entry has an associated filter rule,
1831 	 * save that too.
1832 	 */
1833 	if (nat->nat_fr != NULL)
1834 		bcopy((char *)nat->nat_fr, (char *)&ipn->ipn_fr,
1835 		      sizeof(ipn->ipn_fr));
1836 
1837 	/*
1838 	 * Last but not least, if there is an application proxy session set
1839 	 * up for this NAT entry, then copy that out too, including any
1840 	 * private data saved along side it by the proxy.
1841 	 */
1842 	aps = nat->nat_aps;
1843 	outsize = ipn->ipn_dsize - sizeof(*ipn) + sizeof(ipn->ipn_data);
1844 	if (aps != NULL) {
1845 		char *s;
1846 
1847 		if (outsize < sizeof(*aps)) {
1848 			IPFERROR(60032);
1849 			error = ENOBUFS;
1850 			goto finished;
1851 		}
1852 
1853 		s = ipn->ipn_data;
1854 		bcopy((char *)aps, s, sizeof(*aps));
1855 		s += sizeof(*aps);
1856 		outsize -= sizeof(*aps);
1857 		if ((aps->aps_data != NULL) && (outsize >= aps->aps_psiz))
1858 			bcopy(aps->aps_data, s, aps->aps_psiz);
1859 		else {
1860 			IPFERROR(60033);
1861 			error = ENOBUFS;
1862 		}
1863 	}
1864 	if (error == 0) {
1865 		if (getlock) {
1866 			READ_ENTER(&softc->ipf_nat);
1867 			getlock = 0;
1868 		}
1869 		error = ipf_outobjsz(softc, data, ipn, IPFOBJ_NATSAVE,
1870 				     ipns.ipn_dsize);
1871 	}
1872 
1873 finished:
1874 	if (getlock) {
1875 		READ_ENTER(&softc->ipf_nat);
1876 	}
1877 	if (ipn != NULL) {
1878 		KFREES(ipn, ipns.ipn_dsize);
1879 	}
1880 	return error;
1881 }
1882 
1883 
1884 /* ------------------------------------------------------------------------ */
1885 /* Function:    ipf_nat_putent                                              */
1886 /* Returns:     int - 0 == success, != 0 is the error value.                */
1887 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1888 /*              data(I)    - pointer to natget structure with NAT           */
1889 /*                           structure information to load into the kernel  */
1890 /*              getlock(I) - flag indicating whether or not a write lock    */
1891 /*                           on is already held.                            */
1892 /*                                                                          */
1893 /* Handle SIOCSTPUT.                                                        */
1894 /* Loads a NAT table entry from user space, including a NAT rule, proxy and */
1895 /* firewall rule data structures, if pointers to them indicate so.          */
1896 /* ------------------------------------------------------------------------ */
1897 static int
1898 ipf_nat_putent(ipf_main_softc_t *softc, void *data, int getlock)
1899 {
1900 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1901 	nat_save_t *ipn, *ipnn;
1902 	ap_session_t *aps;
1903 	nat_t *n, *nat;
1904 	frentry_t *fr;
1905 	fr_info_t *fin;
1906 	ipnat_t *in;
1907 	int error;
1908 
1909 	KMALLOC(ipn, nat_save_t *);
1910 	if (ipn == NULL)
1911 		return ENOMEM;
1912 	error = ipf_inobj(softc, data, NULL, ipn, IPFOBJ_NATSAVE);
1913 	if (error != 0)
1914 		return error;
1915 
1916 	/*
1917 	 * Initialise early because of code at junkput label.
1918 	 */
1919 	n = NULL;
1920 	in = NULL;
1921 	aps = NULL;
1922 	nat = NULL;
1923 	ipnn = NULL;
1924 	fin = NULL;
1925 	fr = NULL;
1926 
1927 	/*
1928 	 * New entry, copy in the rest of the NAT entry if it's size is more
1929 	 * than just the nat_t structure.
1930 	 */
1931 	if (ipn->ipn_dsize > sizeof(*ipn)) {
1932 		if (ipn->ipn_dsize > 81920) {
1933 			IPFERROR(60034);
1934 			error = ENOMEM;
1935 			goto junkput;
1936 		}
1937 
1938 		KMALLOCS(ipnn, nat_save_t *, ipn->ipn_dsize);
1939 		if (ipnn == NULL) {
1940 			IPFERROR(60035);
1941 			return ENOMEM;
1942 		}
1943 
1944 		bzero(ipnn, ipn->ipn_dsize);
1945 		error = ipf_inobjsz(softc, data, ipnn, IPFOBJ_NATSAVE,
1946 				    ipn->ipn_dsize);
1947 		if (error != 0) {
1948 			goto junkput;
1949 		}
1950 	} else
1951 		ipnn = ipn;
1952 
1953 	KMALLOC(nat, nat_t *);
1954 	if (nat == NULL) {
1955 		IPFERROR(60037);
1956 		error = ENOMEM;
1957 		goto junkput;
1958 	}
1959 
1960 	bcopy((char *)&ipnn->ipn_nat, (char *)nat, sizeof(*nat));
1961 
1962 	switch (nat->nat_v[0])
1963 	{
1964 	case 4:
1965 #ifdef USE_IENT6
1966 	case 6 :
1967 #endif
1968 		break;
1969 	default :
1970 		IPFERROR(60061);
1971 		error = EPROTONOSUPPORT;
1972 		goto junkput;
1973 		/*NOTREACHED*/
1974 	}
1975 
1976 	/*
1977 	 * Initialize all these so that ipf_nat_delete() doesn't cause a crash.
1978 	 */
1979 	bzero((char *)nat, offsetof(struct nat, nat_tqe));
1980 	nat->nat_tqe.tqe_pnext = NULL;
1981 	nat->nat_tqe.tqe_next = NULL;
1982 	nat->nat_tqe.tqe_ifq = NULL;
1983 	nat->nat_tqe.tqe_parent = nat;
1984 
1985 	/*
1986 	 * Restore the rule associated with this nat session
1987 	 */
1988 	in = ipnn->ipn_nat.nat_ptr;
1989 	if (in != NULL) {
1990 		KMALLOCS(in, ipnat_t *, ipnn->ipn_ipnat.in_size);
1991 		nat->nat_ptr = in;
1992 		if (in == NULL) {
1993 			IPFERROR(60038);
1994 			error = ENOMEM;
1995 			goto junkput;
1996 		}
1997 		bcopy((char *)&ipnn->ipn_ipnat, (char *)in,
1998 		      ipnn->ipn_ipnat.in_size);
1999 		in->in_use = 1;
2000 		in->in_flags |= IPN_DELETE;
2001 
2002 		ATOMIC_INC32(softn->ipf_nat_stats.ns_rules);
2003 
2004 		if (ipf_nat_resolverule(softc, in) != 0) {
2005 			IPFERROR(60039);
2006 			error = ESRCH;
2007 			goto junkput;
2008 		}
2009 	}
2010 
2011 	/*
2012 	 * Check that the NAT entry doesn't already exist in the kernel.
2013 	 *
2014 	 * For NAT_OUTBOUND, we're lookup for a duplicate MAP entry.  To do
2015 	 * this, we check to see if the inbound combination of addresses and
2016 	 * ports is already known.  Similar logic is applied for NAT_INBOUND.
2017 	 *
2018 	 */
2019 	KMALLOC(fin, fr_info_t *);
2020 	if (fin == NULL) {
2021 		error = ENOMEM;
2022 		goto junkput;
2023 	}
2024 	bzero(fin, sizeof(*fin));
2025 	fin->fin_v = nat->nat_v[0];
2026 	fin->fin_p = nat->nat_pr[0];
2027 	fin->fin_rev = nat->nat_rev;
2028 	fin->fin_ifp = nat->nat_ifps[0];
2029 	fin->fin_data[0] = ntohs(nat->nat_ndport);
2030 	fin->fin_data[1] = ntohs(nat->nat_nsport);
2031 
2032 	switch (nat->nat_dir)
2033 	{
2034 	case NAT_OUTBOUND :
2035 	case NAT_DIVERTOUT :
2036 		if (getlock) {
2037 			READ_ENTER(&softc->ipf_nat);
2038 		}
2039 
2040 		fin->fin_v = nat->nat_v[1];
2041 		if (nat->nat_v[1] == 4) {
2042 			n = ipf_nat_inlookup(fin, nat->nat_flags, fin->fin_p,
2043 					     nat->nat_ndstip, nat->nat_nsrcip);
2044 #ifdef USE_INET6
2045 		} else if (nat->nat_v[1] == 6) {
2046 			n = ipf_nat6_inlookup(fin, nat->nat_flags, fin->fin_p,
2047 					      &nat->nat_ndst6.in6,
2048 					      &nat->nat_nsrc6.in6);
2049 #endif
2050 		}
2051 
2052 		if (getlock) {
2053 			RWLOCK_EXIT(&softc->ipf_nat);
2054 		}
2055 		if (n != NULL) {
2056 			IPFERROR(60040);
2057 			error = EEXIST;
2058 			goto junkput;
2059 		}
2060 		break;
2061 
2062 	case NAT_INBOUND :
2063 	case NAT_DIVERTIN :
2064 		if (getlock) {
2065 			READ_ENTER(&softc->ipf_nat);
2066 		}
2067 
2068 		if (fin->fin_v == 4) {
2069 			n = ipf_nat_outlookup(fin, nat->nat_flags, fin->fin_p,
2070 					      nat->nat_ndstip,
2071 					      nat->nat_nsrcip);
2072 #ifdef USE_INET6
2073 		} else if (fin->fin_v == 6) {
2074 			n = ipf_nat6_outlookup(fin, nat->nat_flags, fin->fin_p,
2075 					       &nat->nat_ndst6.in6,
2076 					       &nat->nat_nsrc6.in6);
2077 #endif
2078 		}
2079 
2080 		if (getlock) {
2081 			RWLOCK_EXIT(&softc->ipf_nat);
2082 		}
2083 		if (n != NULL) {
2084 			IPFERROR(60041);
2085 			error = EEXIST;
2086 			goto junkput;
2087 		}
2088 		break;
2089 
2090 	default :
2091 		IPFERROR(60042);
2092 		error = EINVAL;
2093 		goto junkput;
2094 	}
2095 
2096 	/*
2097 	 * Restore ap_session_t structure.  Include the private data allocated
2098 	 * if it was there.
2099 	 */
2100 	aps = nat->nat_aps;
2101 	if (aps != NULL) {
2102 		KMALLOC(aps, ap_session_t *);
2103 		nat->nat_aps = aps;
2104 		if (aps == NULL) {
2105 			IPFERROR(60043);
2106 			error = ENOMEM;
2107 			goto junkput;
2108 		}
2109 		bcopy(ipnn->ipn_data, (char *)aps, sizeof(*aps));
2110 		if (in != NULL)
2111 			aps->aps_apr = in->in_apr;
2112 		else
2113 			aps->aps_apr = NULL;
2114 		if (aps->aps_psiz != 0) {
2115 			if (aps->aps_psiz > 81920) {
2116 				IPFERROR(60044);
2117 				error = ENOMEM;
2118 				goto junkput;
2119 			}
2120 			KMALLOCS(aps->aps_data, void *, aps->aps_psiz);
2121 			if (aps->aps_data == NULL) {
2122 				IPFERROR(60045);
2123 				error = ENOMEM;
2124 				goto junkput;
2125 			}
2126 			bcopy(ipnn->ipn_data + sizeof(*aps), aps->aps_data,
2127 			      aps->aps_psiz);
2128 		} else {
2129 			aps->aps_psiz = 0;
2130 			aps->aps_data = NULL;
2131 		}
2132 	}
2133 
2134 	/*
2135 	 * If there was a filtering rule associated with this entry then
2136 	 * build up a new one.
2137 	 */
2138 	fr = nat->nat_fr;
2139 	if (fr != NULL) {
2140 		if ((nat->nat_flags & SI_NEWFR) != 0) {
2141 			KMALLOC(fr, frentry_t *);
2142 			nat->nat_fr = fr;
2143 			if (fr == NULL) {
2144 				IPFERROR(60046);
2145 				error = ENOMEM;
2146 				goto junkput;
2147 			}
2148 			ipnn->ipn_nat.nat_fr = fr;
2149 			fr->fr_ref = 1;
2150 			(void) ipf_outobj(softc, data, ipnn, IPFOBJ_NATSAVE);
2151 			bcopy((char *)&ipnn->ipn_fr, (char *)fr, sizeof(*fr));
2152 
2153 			fr->fr_ref = 1;
2154 			fr->fr_dsize = 0;
2155 			fr->fr_data = NULL;
2156 			fr->fr_type = FR_T_NONE;
2157 
2158 			MUTEX_NUKE(&fr->fr_lock);
2159 			MUTEX_INIT(&fr->fr_lock, "nat-filter rule lock");
2160 		} else {
2161 			if (getlock) {
2162 				READ_ENTER(&softc->ipf_nat);
2163 			}
2164 			for (n = softn->ipf_nat_instances; n; n = n->nat_next)
2165 				if (n->nat_fr == fr)
2166 					break;
2167 
2168 			if (n != NULL) {
2169 				MUTEX_ENTER(&fr->fr_lock);
2170 				fr->fr_ref++;
2171 				MUTEX_EXIT(&fr->fr_lock);
2172 			}
2173 			if (getlock) {
2174 				RWLOCK_EXIT(&softc->ipf_nat);
2175 			}
2176 
2177 			if (n == NULL) {
2178 				IPFERROR(60047);
2179 				error = ESRCH;
2180 				goto junkput;
2181 			}
2182 		}
2183 	}
2184 
2185 	if (ipnn != ipn) {
2186 		KFREES(ipnn, ipn->ipn_dsize);
2187 		ipnn = NULL;
2188 	}
2189 
2190 	if (getlock) {
2191 		WRITE_ENTER(&softc->ipf_nat);
2192 	}
2193 
2194 	if (fin->fin_v == 4)
2195 		error = ipf_nat_finalise(fin, nat);
2196 #ifdef USE_INET6
2197 	else
2198 		error = ipf_nat6_finalise(fin, nat);
2199 #endif
2200 
2201 	if (getlock) {
2202 		RWLOCK_EXIT(&softc->ipf_nat);
2203 	}
2204 
2205 	if (error == 0)
2206 		return 0;
2207 
2208 	IPFERROR(60048);
2209 	error = ENOMEM;
2210 
2211 junkput:
2212 	if (fin != NULL)
2213 		KFREE(fin);
2214 	if (fr != NULL) {
2215 		(void) ipf_derefrule(softc, &fr);
2216 	}
2217 
2218 	if ((ipnn != NULL) && (ipnn != ipn)) {
2219 		KFREES(ipnn, ipn->ipn_dsize);
2220 	}
2221 	if (ipn != NULL)
2222 		KFREE(ipn);
2223 	if (nat != NULL) {
2224 		if (aps != NULL) {
2225 			if (aps->aps_data != NULL) {
2226 				KFREES(aps->aps_data, aps->aps_psiz);
2227 			}
2228 			KFREE(aps);
2229 		}
2230 		if (in != NULL) {
2231 			if (in->in_apr)
2232 				ipf_proxy_deref(in->in_apr);
2233 			KFREES(in, in->in_size);
2234 		}
2235 		KFREE(nat);
2236 	}
2237 	return error;
2238 }
2239 
2240 
2241 /* ------------------------------------------------------------------------ */
2242 /* Function:    ipf_nat_delete                                              */
2243 /* Returns:     Nil                                                         */
2244 /* Parameters:  softc(I)   - pointer to soft context main structure         */
2245 /*              nat(I)     - pointer to NAT structure to delete             */
2246 /*              logtype(I) - type of LOG record to create before deleting   */
2247 /* Write Lock:  ipf_nat                                                     */
2248 /*                                                                          */
2249 /* Delete a nat entry from the various lists and table.  If NAT logging is  */
2250 /* enabled then generate a NAT log record for this event.                   */
2251 /* ------------------------------------------------------------------------ */
2252 void
2253 ipf_nat_delete(ipf_main_softc_t *softc, struct nat *nat, int logtype)
2254 {
2255 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2256 	int madeorphan = 0, bkt, removed = 0;
2257 	nat_stat_side_t *nss;
2258 	struct ipnat *ipn;
2259 
2260 	if (logtype != 0 && softn->ipf_nat_logging != 0)
2261 		ipf_nat_log(softc, softn, nat, logtype);
2262 
2263 	/*
2264 	 * Take it as a general indication that all the pointers are set if
2265 	 * nat_pnext is set.
2266 	 */
2267 	if (nat->nat_pnext != NULL) {
2268 		removed = 1;
2269 
2270 		bkt = nat->nat_hv[0] % softn->ipf_nat_table_sz;
2271 		nss = &softn->ipf_nat_stats.ns_side[0];
2272 		nss->ns_bucketlen[bkt]--;
2273 		if (nss->ns_bucketlen[bkt] == 0) {
2274 			nss->ns_inuse--;
2275 		}
2276 
2277 		bkt = nat->nat_hv[1] % softn->ipf_nat_table_sz;
2278 		nss = &softn->ipf_nat_stats.ns_side[1];
2279 		nss->ns_bucketlen[bkt]--;
2280 		if (nss->ns_bucketlen[bkt] == 0) {
2281 			nss->ns_inuse--;
2282 		}
2283 
2284 		*nat->nat_pnext = nat->nat_next;
2285 		if (nat->nat_next != NULL) {
2286 			nat->nat_next->nat_pnext = nat->nat_pnext;
2287 			nat->nat_next = NULL;
2288 		}
2289 		nat->nat_pnext = NULL;
2290 
2291 		*nat->nat_phnext[0] = nat->nat_hnext[0];
2292 		if (nat->nat_hnext[0] != NULL) {
2293 			nat->nat_hnext[0]->nat_phnext[0] = nat->nat_phnext[0];
2294 			nat->nat_hnext[0] = NULL;
2295 		}
2296 		nat->nat_phnext[0] = NULL;
2297 
2298 		*nat->nat_phnext[1] = nat->nat_hnext[1];
2299 		if (nat->nat_hnext[1] != NULL) {
2300 			nat->nat_hnext[1]->nat_phnext[1] = nat->nat_phnext[1];
2301 			nat->nat_hnext[1] = NULL;
2302 		}
2303 		nat->nat_phnext[1] = NULL;
2304 
2305 		if ((nat->nat_flags & SI_WILDP) != 0) {
2306 			ATOMIC_DEC32(softn->ipf_nat_stats.ns_wilds);
2307 		}
2308 		madeorphan = 1;
2309 	}
2310 
2311 	if (nat->nat_me != NULL) {
2312 		*nat->nat_me = NULL;
2313 		nat->nat_me = NULL;
2314 		nat->nat_ref--;
2315 		ASSERT(nat->nat_ref >= 0);
2316 	}
2317 
2318 	if (nat->nat_tqe.tqe_ifq != NULL) {
2319 		/*
2320 		 * No call to ipf_freetimeoutqueue() is made here, they are
2321 		 * garbage collected in ipf_nat_expire().
2322 		 */
2323 		(void) ipf_deletequeueentry(&nat->nat_tqe);
2324 	}
2325 
2326 	if (nat->nat_sync) {
2327 		ipf_sync_del_nat(softc->ipf_sync_soft, nat->nat_sync);
2328 		nat->nat_sync = NULL;
2329 	}
2330 
2331 	if (logtype == NL_EXPIRE)
2332 		softn->ipf_nat_stats.ns_expire++;
2333 
2334 	MUTEX_ENTER(&nat->nat_lock);
2335 	/*
2336 	 * NL_DESTROY should only be passed in when we've got nat_ref >= 2.
2337 	 * This happens when a nat'd packet is blocked and we want to throw
2338 	 * away the NAT session.
2339 	 */
2340 	if (logtype == NL_DESTROY) {
2341 		if (nat->nat_ref > 2) {
2342 			nat->nat_ref -= 2;
2343 			MUTEX_EXIT(&nat->nat_lock);
2344 			if (removed)
2345 				softn->ipf_nat_stats.ns_orphans++;
2346 			return;
2347 		}
2348 	} else if (nat->nat_ref > 1) {
2349 		nat->nat_ref--;
2350 		MUTEX_EXIT(&nat->nat_lock);
2351 		if (madeorphan == 1)
2352 			softn->ipf_nat_stats.ns_orphans++;
2353 		return;
2354 	}
2355 	ASSERT(nat->nat_ref >= 0);
2356 	MUTEX_EXIT(&nat->nat_lock);
2357 
2358 	nat->nat_ref = 0;
2359 
2360 	if (madeorphan == 0)
2361 		softn->ipf_nat_stats.ns_orphans--;
2362 
2363 	/*
2364 	 * At this point, nat_ref can be either 0 or -1
2365 	 */
2366 	softn->ipf_nat_stats.ns_proto[nat->nat_pr[0]]--;
2367 
2368 	if (nat->nat_fr != NULL) {
2369 		(void) ipf_derefrule(softc, &nat->nat_fr);
2370 	}
2371 
2372 	if (nat->nat_hm != NULL) {
2373 		ipf_nat_hostmapdel(softc, &nat->nat_hm);
2374 	}
2375 
2376 	/*
2377 	 * If there is an active reference from the nat entry to its parent
2378 	 * rule, decrement the rule's reference count and free it too if no
2379 	 * longer being used.
2380 	 */
2381 	ipn = nat->nat_ptr;
2382 	nat->nat_ptr = NULL;
2383 
2384 	if (ipn != NULL) {
2385 		ipn->in_space++;
2386 		ipf_nat_rule_deref(softc, &ipn);
2387 	}
2388 
2389 	if (nat->nat_aps != NULL) {
2390 		ipf_proxy_free(softc, nat->nat_aps);
2391 		nat->nat_aps = NULL;
2392 	}
2393 
2394 	MUTEX_DESTROY(&nat->nat_lock);
2395 
2396 	softn->ipf_nat_stats.ns_active--;
2397 
2398 	/*
2399 	 * If there's a fragment table entry too for this nat entry, then
2400 	 * dereference that as well.  This is after nat_lock is released
2401 	 * because of Tru64.
2402 	 */
2403 	ipf_frag_natforget(softc, (void *)nat);
2404 
2405 	KFREE(nat);
2406 }
2407 
2408 
2409 /* ------------------------------------------------------------------------ */
2410 /* Function:    ipf_nat_flushtable                                          */
2411 /* Returns:     int - number of NAT rules deleted                           */
2412 /* Parameters:  softc(I) - pointer to soft context main structure           */
2413 /*              softn(I) - pointer to NAT context structure                 */
2414 /* Write Lock:  ipf_nat                                                     */
2415 /*                                                                          */
2416 /* Deletes all currently active NAT sessions.  In deleting each NAT entry a */
2417 /* log record should be emitted in ipf_nat_delete() if NAT logging is       */
2418 /* enabled.                                                                 */
2419 /* ------------------------------------------------------------------------ */
2420 /*
2421  * nat_flushtable - clear the NAT table of all mapping entries.
2422  */
2423 static int
2424 ipf_nat_flushtable(ipf_main_softc_t *softc, ipf_nat_softc_t *softn)
2425 {
2426 	nat_t *nat;
2427 	int j = 0;
2428 
2429 	/*
2430 	 * ALL NAT mappings deleted, so lets just make the deletions
2431 	 * quicker.
2432 	 */
2433 	if (softn->ipf_nat_table[0] != NULL)
2434 		bzero((char *)softn->ipf_nat_table[0],
2435 		      sizeof(softn->ipf_nat_table[0]) *
2436 		      softn->ipf_nat_table_sz);
2437 	if (softn->ipf_nat_table[1] != NULL)
2438 		bzero((char *)softn->ipf_nat_table[1],
2439 		      sizeof(softn->ipf_nat_table[1]) *
2440 		      softn->ipf_nat_table_sz);
2441 
2442 	while ((nat = softn->ipf_nat_instances) != NULL) {
2443 		ipf_nat_delete(softc, nat, NL_FLUSH);
2444 		j++;
2445 	}
2446 
2447 	return j;
2448 }
2449 
2450 
2451 /* ------------------------------------------------------------------------ */
2452 /* Function:    ipf_nat_clearlist                                           */
2453 /* Returns:     int - number of NAT/RDR rules deleted                       */
2454 /* Parameters:  softc(I) - pointer to soft context main structure           */
2455 /*              softn(I) - pointer to NAT context structure                 */
2456 /*                                                                          */
2457 /* Delete all rules in the current list of rules.  There is nothing elegant */
2458 /* about this cleanup: simply free all entries on the list of rules and     */
2459 /* clear out the tables used for hashed NAT rule lookups.                   */
2460 /* ------------------------------------------------------------------------ */
2461 static int
2462 ipf_nat_clearlist(ipf_main_softc_t *softc, ipf_nat_softc_t *softn)
2463 {
2464 	ipnat_t *n;
2465 	int i = 0;
2466 
2467 	if (softn->ipf_nat_map_rules != NULL) {
2468 		bzero((char *)softn->ipf_nat_map_rules,
2469 		      sizeof(*softn->ipf_nat_map_rules) *
2470 		      softn->ipf_nat_maprules_sz);
2471 	}
2472 	if (softn->ipf_nat_rdr_rules != NULL) {
2473 		bzero((char *)softn->ipf_nat_rdr_rules,
2474 		      sizeof(*softn->ipf_nat_rdr_rules) *
2475 		      softn->ipf_nat_rdrrules_sz);
2476 	}
2477 
2478 	while ((n = softn->ipf_nat_list) != NULL) {
2479 		ipf_nat_delrule(softc, softn, n, 0);
2480 		i++;
2481 	}
2482 #if SOLARIS && !defined(INSTANCES)
2483 	pfil_delayed_copy = 1;
2484 #endif
2485 	return i;
2486 }
2487 
2488 
2489 /* ------------------------------------------------------------------------ */
2490 /* Function:    ipf_nat_delrule                                             */
2491 /* Returns:     Nil                                                         */
2492 /* Parameters:  softc(I) - pointer to soft context main structure           */
2493 /*              softn(I) - pointer to NAT context structure                 */
2494 /*              np(I)    - pointer to NAT rule to delete                    */
2495 /*              purge(I) - 1 == allow purge, 0 == prevent purge             */
2496 /* Locks:       WRITE(ipf_nat)                                              */
2497 /*                                                                          */
2498 /* Preventing "purge" from occuring is allowed because when all of the NAT  */
2499 /* rules are being removed, allowing the "purge" to walk through the list   */
2500 /* of NAT sessions, possibly multiple times, would be a large performance   */
2501 /* hit, on the order of O(N^2).                                             */
2502 /* ------------------------------------------------------------------------ */
2503 static void
2504 ipf_nat_delrule(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *np,
2505 	int purge)
2506 {
2507 
2508 	if (np->in_pnext != NULL) {
2509 		*np->in_pnext = np->in_next;
2510 		if (np->in_next != NULL)
2511 			np->in_next->in_pnext = np->in_pnext;
2512 		if (softn->ipf_nat_list_tail == &np->in_next)
2513 			softn->ipf_nat_list_tail = np->in_pnext;
2514 	}
2515 
2516 	if ((purge == 1) && ((np->in_flags & IPN_PURGE) != 0)) {
2517 		nat_t *next;
2518 		nat_t *nat;
2519 
2520 		for (next = softn->ipf_nat_instances; (nat = next) != NULL;) {
2521 			next = nat->nat_next;
2522 			if (nat->nat_ptr == np)
2523 				ipf_nat_delete(softc, nat, NL_PURGE);
2524 		}
2525 	}
2526 
2527 	if ((np->in_flags & IPN_DELETE) == 0) {
2528 		if (np->in_redir & NAT_REDIRECT) {
2529 			switch (np->in_v[0])
2530 			{
2531 			case 4 :
2532 				ipf_nat_delrdr(softn, np);
2533 				break;
2534 #ifdef USE_INET6
2535 			case 6 :
2536 				ipf_nat6_delrdr(softn, np);
2537 				break;
2538 #endif
2539 			}
2540 		}
2541 		if (np->in_redir & (NAT_MAPBLK|NAT_MAP)) {
2542 			switch (np->in_v[0])
2543 			{
2544 			case 4 :
2545 				ipf_nat_delmap(softn, np);
2546 				break;
2547 #ifdef USE_INET6
2548 			case 6 :
2549 				ipf_nat6_delmap(softn, np);
2550 				break;
2551 #endif
2552 			}
2553 		}
2554 	}
2555 
2556 	np->in_flags |= IPN_DELETE;
2557 	ipf_nat_rule_deref(softc, &np);
2558 }
2559 
2560 
2561 /* ------------------------------------------------------------------------ */
2562 /* Function:    ipf_nat_newmap                                              */
2563 /* Returns:     int - -1 == error, 0 == success                             */
2564 /* Parameters:  fin(I) - pointer to packet information                      */
2565 /*              nat(I) - pointer to NAT entry                               */
2566 /*              ni(I)  - pointer to structure with misc. information needed */
2567 /*                       to create new NAT entry.                           */
2568 /*                                                                          */
2569 /* Given an empty NAT structure, populate it with new information about a   */
2570 /* new NAT session, as defined by the matching NAT rule.                    */
2571 /* ni.nai_ip is passed in uninitialised and must be set, in host byte order,*/
2572 /* to the new IP address for the translation.                               */
2573 /* ------------------------------------------------------------------------ */
2574 static int
2575 ipf_nat_newmap(fr_info_t *fin, nat_t *nat, natinfo_t *ni)
2576 {
2577 	ipf_main_softc_t *softc = fin->fin_main_soft;
2578 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2579 	u_short st_port, dport, sport, port, sp, dp;
2580 	struct in_addr in, inb;
2581 	hostmap_t *hm;
2582 	u_32_t flags;
2583 	u_32_t st_ip;
2584 	ipnat_t *np;
2585 	nat_t *natl;
2586 	int l;
2587 
2588 	/*
2589 	 * If it's an outbound packet which doesn't match any existing
2590 	 * record, then create a new port
2591 	 */
2592 	l = 0;
2593 	hm = NULL;
2594 	np = ni->nai_np;
2595 	st_ip = np->in_snip;
2596 	st_port = np->in_spnext;
2597 	flags = nat->nat_flags;
2598 
2599 	if (flags & IPN_ICMPQUERY) {
2600 		sport = fin->fin_data[1];
2601 		dport = 0;
2602 	} else {
2603 		sport = htons(fin->fin_data[0]);
2604 		dport = htons(fin->fin_data[1]);
2605 	}
2606 
2607 	/*
2608 	 * Do a loop until we either run out of entries to try or we find
2609 	 * a NAT mapping that isn't currently being used.  This is done
2610 	 * because the change to the source is not (usually) being fixed.
2611 	 */
2612 	do {
2613 		port = 0;
2614 		in.s_addr = htonl(np->in_snip);
2615 		if (l == 0) {
2616 			/*
2617 			 * Check to see if there is an existing NAT
2618 			 * setup for this IP address pair.
2619 			 */
2620 			hm = ipf_nat_hostmap(softn, np, fin->fin_src,
2621 					     fin->fin_dst, in, 0);
2622 			if (hm != NULL)
2623 				in.s_addr = hm->hm_nsrcip.s_addr;
2624 		} else if ((l == 1) && (hm != NULL)) {
2625 			ipf_nat_hostmapdel(softc, &hm);
2626 		}
2627 		in.s_addr = ntohl(in.s_addr);
2628 
2629 		nat->nat_hm = hm;
2630 
2631 		if ((np->in_nsrcmsk == 0xffffffff) && (np->in_spnext == 0)) {
2632 			if (l > 0) {
2633 				NBUMPSIDEX(1, ns_exhausted, ns_exhausted_1);
2634 				return -1;
2635 			}
2636 		}
2637 
2638 		if (np->in_redir == NAT_BIMAP &&
2639 		    np->in_osrcmsk == np->in_nsrcmsk) {
2640 			/*
2641 			 * map the address block in a 1:1 fashion
2642 			 */
2643 			in.s_addr = np->in_nsrcaddr;
2644 			in.s_addr |= fin->fin_saddr & ~np->in_osrcmsk;
2645 			in.s_addr = ntohl(in.s_addr);
2646 
2647 		} else if (np->in_redir & NAT_MAPBLK) {
2648 			if ((l >= np->in_ppip) || ((l > 0) &&
2649 			     !(flags & IPN_TCPUDP))) {
2650 				NBUMPSIDEX(1, ns_exhausted, ns_exhausted_2);
2651 				return -1;
2652 			}
2653 			/*
2654 			 * map-block - Calculate destination address.
2655 			 */
2656 			in.s_addr = ntohl(fin->fin_saddr);
2657 			in.s_addr &= ntohl(~np->in_osrcmsk);
2658 			inb.s_addr = in.s_addr;
2659 			in.s_addr /= np->in_ippip;
2660 			in.s_addr &= ntohl(~np->in_nsrcmsk);
2661 			in.s_addr += ntohl(np->in_nsrcaddr);
2662 			/*
2663 			 * Calculate destination port.
2664 			 */
2665 			if ((flags & IPN_TCPUDP) &&
2666 			    (np->in_ppip != 0)) {
2667 				port = ntohs(sport) + l;
2668 				port %= np->in_ppip;
2669 				port += np->in_ppip *
2670 					(inb.s_addr % np->in_ippip);
2671 				port += MAPBLK_MINPORT;
2672 				port = htons(port);
2673 			}
2674 
2675 		} else if ((np->in_nsrcaddr == 0) &&
2676 			   (np->in_nsrcmsk == 0xffffffff)) {
2677 			i6addr_t in6;
2678 
2679 			/*
2680 			 * 0/32 - use the interface's IP address.
2681 			 */
2682 			if ((l > 0) ||
2683 			    ipf_ifpaddr(softc, 4, FRI_NORMAL, fin->fin_ifp,
2684 				       &in6, NULL) == -1) {
2685 				NBUMPSIDEX(1, ns_new_ifpaddr, ns_new_ifpaddr_1);
2686 				return -1;
2687 			}
2688 			in.s_addr = ntohl(in6.in4.s_addr);
2689 
2690 		} else if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0)) {
2691 			/*
2692 			 * 0/0 - use the original source address/port.
2693 			 */
2694 			if (l > 0) {
2695 				NBUMPSIDEX(1, ns_exhausted, ns_exhausted_3);
2696 				return -1;
2697 			}
2698 			in.s_addr = ntohl(fin->fin_saddr);
2699 
2700 		} else if ((np->in_nsrcmsk != 0xffffffff) &&
2701 			   (np->in_spnext == 0) && ((l > 0) || (hm == NULL)))
2702 			np->in_snip++;
2703 
2704 		natl = NULL;
2705 
2706 		if ((flags & IPN_TCPUDP) &&
2707 		    ((np->in_redir & NAT_MAPBLK) == 0) &&
2708 		    (np->in_flags & IPN_AUTOPORTMAP)) {
2709 			/*
2710 			 * "ports auto" (without map-block)
2711 			 */
2712 			if ((l > 0) && np->in_ppip && (l % np->in_ppip == 0)) {
2713 				if (l > np->in_space) {
2714 					return -1;
2715 				} else if ((l > np->in_ppip) &&
2716 					   np->in_nsrcmsk != 0xffffffff)
2717 					np->in_snip++;
2718 			}
2719 			if (np->in_ppip != 0) {
2720 				port = ntohs(sport);
2721 				port += (l % np->in_ppip);
2722 				port %= np->in_ppip;
2723 				port += np->in_ppip *
2724 					(ntohl(fin->fin_saddr) %
2725 					 np->in_ippip);
2726 				port += MAPBLK_MINPORT;
2727 				port = htons(port);
2728 			}
2729 
2730 		} else if (((np->in_redir & NAT_MAPBLK) == 0) &&
2731 			   (flags & IPN_TCPUDPICMP) && (np->in_spnext != 0)) {
2732 			/*
2733 			 * Standard port translation.  Select next port.
2734 			 */
2735 			if (np->in_flags & IPN_SEQUENTIAL) {
2736 				port = np->in_spnext;
2737 			} else {
2738 				port = ipf_random() % (np->in_spmax -
2739 						       np->in_spmin + 1);
2740 				port += np->in_spmin;
2741 			}
2742 			port = htons(port);
2743 			np->in_spnext++;
2744 
2745 			if (np->in_spnext > np->in_spmax) {
2746 				np->in_spnext = np->in_spmin;
2747 				if (np->in_nsrcmsk != 0xffffffff)
2748 					np->in_snip++;
2749 			}
2750 		}
2751 
2752 		if (np->in_flags & IPN_SIPRANGE) {
2753 			if (np->in_snip > ntohl(np->in_nsrcmsk))
2754 				np->in_snip = ntohl(np->in_nsrcaddr);
2755 		} else {
2756 			if ((np->in_nsrcmsk != 0xffffffff) &&
2757 			    ((np->in_snip + 1) & ntohl(np->in_nsrcmsk)) >
2758 			    ntohl(np->in_nsrcaddr))
2759 				np->in_snip = ntohl(np->in_nsrcaddr) + 1;
2760 		}
2761 
2762 		if ((port == 0) && (flags & (IPN_TCPUDPICMP|IPN_ICMPQUERY)))
2763 			port = sport;
2764 
2765 		/*
2766 		 * Here we do a lookup of the connection as seen from
2767 		 * the outside.  If an IP# pair already exists, try
2768 		 * again.  So if you have A->B becomes C->B, you can
2769 		 * also have D->E become C->E but not D->B causing
2770 		 * another C->B.  Also take protocol and ports into
2771 		 * account when determining whether a pre-existing
2772 		 * NAT setup will cause an external conflict where
2773 		 * this is appropriate.
2774 		 */
2775 		inb.s_addr = htonl(in.s_addr);
2776 		sp = fin->fin_data[0];
2777 		dp = fin->fin_data[1];
2778 		fin->fin_data[0] = fin->fin_data[1];
2779 		fin->fin_data[1] = ntohs(port);
2780 		natl = ipf_nat_inlookup(fin, flags & ~(SI_WILDP|NAT_SEARCH),
2781 					(u_int)fin->fin_p, fin->fin_dst, inb);
2782 		fin->fin_data[0] = sp;
2783 		fin->fin_data[1] = dp;
2784 
2785 		/*
2786 		 * Has the search wrapped around and come back to the
2787 		 * start ?
2788 		 */
2789 		if ((natl != NULL) &&
2790 		    (np->in_spnext != 0) && (st_port == np->in_spnext) &&
2791 		    (np->in_snip != 0) && (st_ip == np->in_snip)) {
2792 			NBUMPSIDED(1, ns_wrap);
2793 			return -1;
2794 		}
2795 		l++;
2796 	} while (natl != NULL);
2797 
2798 	/* Setup the NAT table */
2799 	nat->nat_osrcip = fin->fin_src;
2800 	nat->nat_nsrcaddr = htonl(in.s_addr);
2801 	nat->nat_odstip = fin->fin_dst;
2802 	nat->nat_ndstip = fin->fin_dst;
2803 	if (nat->nat_hm == NULL)
2804 		nat->nat_hm = ipf_nat_hostmap(softn, np, fin->fin_src,
2805 					      fin->fin_dst, nat->nat_nsrcip,
2806 					      0);
2807 
2808 	if (flags & IPN_TCPUDP) {
2809 		nat->nat_osport = sport;
2810 		nat->nat_nsport = port;	/* sport */
2811 		nat->nat_odport = dport;
2812 		nat->nat_ndport = dport;
2813 		((tcphdr_t *)fin->fin_dp)->th_sport = port;
2814 	} else if (flags & IPN_ICMPQUERY) {
2815 		nat->nat_oicmpid = fin->fin_data[1];
2816 		((icmphdr_t *)fin->fin_dp)->icmp_id = port;
2817 		nat->nat_nicmpid = port;
2818 	}
2819 	return 0;
2820 }
2821 
2822 
2823 /* ------------------------------------------------------------------------ */
2824 /* Function:    ipf_nat_newrdr                                              */
2825 /* Returns:     int - -1 == error, 0 == success (no move), 1 == success and */
2826 /*                    allow rule to be moved if IPN_ROUNDR is set.          */
2827 /* Parameters:  fin(I) - pointer to packet information                      */
2828 /*              nat(I) - pointer to NAT entry                               */
2829 /*              ni(I)  - pointer to structure with misc. information needed */
2830 /*                       to create new NAT entry.                           */
2831 /*                                                                          */
2832 /* ni.nai_ip is passed in uninitialised and must be set, in host byte order,*/
2833 /* to the new IP address for the translation.                               */
2834 /* ------------------------------------------------------------------------ */
2835 static int
2836 ipf_nat_newrdr(fr_info_t *fin, nat_t *nat, natinfo_t *ni)
2837 {
2838 	ipf_main_softc_t *softc = fin->fin_main_soft;
2839 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2840 	u_short nport, dport, sport;
2841 	struct in_addr in, inb;
2842 	u_short sp, dp;
2843 	hostmap_t *hm;
2844 	u_32_t flags;
2845 	ipnat_t *np;
2846 	nat_t *natl;
2847 	int move;
2848 
2849 	move = 1;
2850 	hm = NULL;
2851 	in.s_addr = 0;
2852 	np = ni->nai_np;
2853 	flags = nat->nat_flags;
2854 
2855 	if (flags & IPN_ICMPQUERY) {
2856 		dport = fin->fin_data[1];
2857 		sport = 0;
2858 	} else {
2859 		sport = htons(fin->fin_data[0]);
2860 		dport = htons(fin->fin_data[1]);
2861 	}
2862 
2863 	/* TRACE sport, dport */
2864 
2865 
2866 	/*
2867 	 * If the matching rule has IPN_STICKY set, then we want to have the
2868 	 * same rule kick in as before.  Why would this happen?  If you have
2869 	 * a collection of rdr rules with "round-robin sticky", the current
2870 	 * packet might match a different one to the previous connection but
2871 	 * we want the same destination to be used.
2872 	 */
2873 	if (((np->in_flags & (IPN_ROUNDR|IPN_SPLIT)) != 0) &&
2874 	    ((np->in_flags & IPN_STICKY) != 0)) {
2875 		hm = ipf_nat_hostmap(softn, NULL, fin->fin_src, fin->fin_dst,
2876 				     in, (u_32_t)dport);
2877 		if (hm != NULL) {
2878 			in.s_addr = ntohl(hm->hm_ndstip.s_addr);
2879 			np = hm->hm_ipnat;
2880 			ni->nai_np = np;
2881 			move = 0;
2882 			ipf_nat_hostmapdel(softc, &hm);
2883 		}
2884 	}
2885 
2886 	/*
2887 	 * Otherwise, it's an inbound packet. Most likely, we don't
2888 	 * want to rewrite source ports and source addresses. Instead,
2889 	 * we want to rewrite to a fixed internal address and fixed
2890 	 * internal port.
2891 	 */
2892 	if (np->in_flags & IPN_SPLIT) {
2893 		in.s_addr = np->in_dnip;
2894 
2895 		if ((np->in_flags & (IPN_ROUNDR|IPN_STICKY)) == IPN_STICKY) {
2896 			hm = ipf_nat_hostmap(softn, NULL, fin->fin_src,
2897 					     fin->fin_dst, in, (u_32_t)dport);
2898 			if (hm != NULL) {
2899 				in.s_addr = hm->hm_ndstip.s_addr;
2900 				move = 0;
2901 			}
2902 		}
2903 
2904 		if (hm == NULL || hm->hm_ref == 1) {
2905 			if (np->in_ndstaddr == htonl(in.s_addr)) {
2906 				np->in_dnip = ntohl(np->in_ndstmsk);
2907 				move = 0;
2908 			} else {
2909 				np->in_dnip = ntohl(np->in_ndstaddr);
2910 			}
2911 		}
2912 		if (hm != NULL)
2913 			ipf_nat_hostmapdel(softc, &hm);
2914 
2915 	} else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0xffffffff)) {
2916 		i6addr_t in6;
2917 
2918 		/*
2919 		 * 0/32 - use the interface's IP address.
2920 		 */
2921 		if (ipf_ifpaddr(softc, 4, FRI_NORMAL, fin->fin_ifp,
2922 			       &in6, NULL) == -1) {
2923 			NBUMPSIDEX(0, ns_new_ifpaddr, ns_new_ifpaddr_2);
2924 			return -1;
2925 		}
2926 		in.s_addr = ntohl(in6.in4.s_addr);
2927 
2928 	} else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk== 0)) {
2929 		/*
2930 		 * 0/0 - use the original destination address/port.
2931 		 */
2932 		in.s_addr = ntohl(fin->fin_daddr);
2933 
2934 	} else if (np->in_redir == NAT_BIMAP &&
2935 		   np->in_ndstmsk == np->in_odstmsk) {
2936 		/*
2937 		 * map the address block in a 1:1 fashion
2938 		 */
2939 		in.s_addr = np->in_ndstaddr;
2940 		in.s_addr |= fin->fin_daddr & ~np->in_ndstmsk;
2941 		in.s_addr = ntohl(in.s_addr);
2942 	} else {
2943 		in.s_addr = ntohl(np->in_ndstaddr);
2944 	}
2945 
2946 	if ((np->in_dpnext == 0) || ((flags & NAT_NOTRULEPORT) != 0))
2947 		nport = dport;
2948 	else {
2949 		/*
2950 		 * Whilst not optimized for the case where
2951 		 * pmin == pmax, the gain is not significant.
2952 		 */
2953 		if (((np->in_flags & IPN_FIXEDDPORT) == 0) &&
2954 		    (np->in_odport != np->in_dtop)) {
2955 			nport = ntohs(dport) - np->in_odport + np->in_dpmax;
2956 			nport = htons(nport);
2957 		} else {
2958 			nport = htons(np->in_dpnext);
2959 			np->in_dpnext++;
2960 			if (np->in_dpnext > np->in_dpmax)
2961 				np->in_dpnext = np->in_dpmin;
2962 		}
2963 	}
2964 
2965 	/*
2966 	 * When the redirect-to address is set to 0.0.0.0, just
2967 	 * assume a blank `forwarding' of the packet.  We don't
2968 	 * setup any translation for this either.
2969 	 */
2970 	if (in.s_addr == 0) {
2971 		if (nport == dport) {
2972 			NBUMPSIDED(0, ns_xlate_null);
2973 			return -1;
2974 		}
2975 		in.s_addr = ntohl(fin->fin_daddr);
2976 	}
2977 
2978 	/*
2979 	 * Check to see if this redirect mapping already exists and if
2980 	 * it does, return "failure" (allowing it to be created will just
2981 	 * cause one or both of these "connections" to stop working.)
2982 	 */
2983 	inb.s_addr = htonl(in.s_addr);
2984 	sp = fin->fin_data[0];
2985 	dp = fin->fin_data[1];
2986 	fin->fin_data[1] = fin->fin_data[0];
2987 	fin->fin_data[0] = ntohs(nport);
2988 	natl = ipf_nat_outlookup(fin, flags & ~(SI_WILDP|NAT_SEARCH),
2989 			     (u_int)fin->fin_p, inb, fin->fin_src);
2990 	fin->fin_data[0] = sp;
2991 	fin->fin_data[1] = dp;
2992 	if (natl != NULL) {
2993 		DT2(ns_new_xlate_exists, fr_info_t *, fin, nat_t *, natl);
2994 		NBUMPSIDE(0, ns_xlate_exists);
2995 		return -1;
2996 	}
2997 
2998 	nat->nat_ndstaddr = htonl(in.s_addr);
2999 	nat->nat_odstip = fin->fin_dst;
3000 	nat->nat_nsrcip = fin->fin_src;
3001 	nat->nat_osrcip = fin->fin_src;
3002 	if ((nat->nat_hm == NULL) && ((np->in_flags & IPN_STICKY) != 0))
3003 		nat->nat_hm = ipf_nat_hostmap(softn, np, fin->fin_src,
3004 					      fin->fin_dst, in, (u_32_t)dport);
3005 
3006 	if (flags & IPN_TCPUDP) {
3007 		nat->nat_odport = dport;
3008 		nat->nat_ndport = nport;
3009 		nat->nat_osport = sport;
3010 		nat->nat_nsport = sport;
3011 		((tcphdr_t *)fin->fin_dp)->th_dport = nport;
3012 	} else if (flags & IPN_ICMPQUERY) {
3013 		nat->nat_oicmpid = fin->fin_data[1];
3014 		((icmphdr_t *)fin->fin_dp)->icmp_id = nport;
3015 		nat->nat_nicmpid = nport;
3016 	}
3017 
3018 	return move;
3019 }
3020 
3021 /* ------------------------------------------------------------------------ */
3022 /* Function:    ipf_nat_add                                                 */
3023 /* Returns:     nat_t* - NULL == failure to create new NAT structure,       */
3024 /*                       else pointer to new NAT structure                  */
3025 /* Parameters:  fin(I)       - pointer to packet information                */
3026 /*              np(I)        - pointer to NAT rule                          */
3027 /*              natsave(I)   - pointer to where to store NAT struct pointer */
3028 /*              flags(I)     - flags describing the current packet          */
3029 /*              direction(I) - direction of packet (in/out)                 */
3030 /* Write Lock:  ipf_nat                                                     */
3031 /*                                                                          */
3032 /* Attempts to create a new NAT entry.  Does not actually change the packet */
3033 /* in any way.                                                              */
3034 /*                                                                          */
3035 /* This fucntion is in three main parts: (1) deal with creating a new NAT   */
3036 /* structure for a "MAP" rule (outgoing NAT translation); (2) deal with     */
3037 /* creating a new NAT structure for a "RDR" rule (incoming NAT translation) */
3038 /* and (3) building that structure and putting it into the NAT table(s).    */
3039 /*                                                                          */
3040 /* NOTE: natsave should NOT be used top point back to an ipstate_t struct   */
3041 /*       as it can result in memory being corrupted.                        */
3042 /* ------------------------------------------------------------------------ */
3043 nat_t *
3044 ipf_nat_add(fr_info_t *fin, ipnat_t *np, nat_t **natsave, u_int flags,
3045     int direction)
3046 {
3047 	ipf_main_softc_t *softc = fin->fin_main_soft;
3048 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3049 	hostmap_t *hm = NULL;
3050 	nat_t *nat, *natl;
3051 	natstat_t *nsp;
3052 	u_int nflags;
3053 	natinfo_t ni;
3054 	int move;
3055 
3056 	memset(&ni, 0, sizeof ni);	/* XXX gcc */
3057 	nsp = &softn->ipf_nat_stats;
3058 
3059 	if ((nsp->ns_active * 100 / softn->ipf_nat_table_max) >
3060 	    softn->ipf_nat_table_wm_high) {
3061 		softn->ipf_nat_doflush = 1;
3062 	}
3063 
3064 	if (nsp->ns_active >= softn->ipf_nat_table_max) {
3065 		NBUMPSIDED(fin->fin_out, ns_table_max);
3066 		return NULL;
3067 	}
3068 
3069 	move = 1;
3070 	nflags = np->in_flags & flags;
3071 	nflags &= NAT_FROMRULE;
3072 
3073 	ni.nai_np = np;
3074 	ni.nai_dport = 0;
3075 	ni.nai_sport = 0;
3076 
3077 	/* Give me a new nat */
3078 	KMALLOC(nat, nat_t *);
3079 	if (nat == NULL) {
3080 		NBUMPSIDED(fin->fin_out, ns_memfail);
3081 		/*
3082 		 * Try to automatically tune the max # of entries in the
3083 		 * table allowed to be less than what will cause kmem_alloc()
3084 		 * to fail and try to eliminate panics due to out of memory
3085 		 * conditions arising.
3086 		 */
3087 		if ((softn->ipf_nat_table_max > softn->ipf_nat_table_sz) &&
3088 		    (nsp->ns_active > 100)) {
3089 			softn->ipf_nat_table_max = nsp->ns_active - 100;
3090 			printf("table_max reduced to %d\n",
3091 				softn->ipf_nat_table_max);
3092 		}
3093 		return NULL;
3094 	}
3095 
3096 	if (flags & IPN_ICMPQUERY) {
3097 		/*
3098 		 * In the ICMP query NAT code, we translate the ICMP id fields
3099 		 * to make them unique. This is indepedent of the ICMP type
3100 		 * (e.g. in the unlikely event that a host sends an echo and
3101 		 * an tstamp request with the same id, both packets will have
3102 		 * their ip address/id field changed in the same way).
3103 		 */
3104 		/* The icmp_id field is used by the sender to identify the
3105 		 * process making the icmp request. (the receiver justs
3106 		 * copies it back in its response). So, it closely matches
3107 		 * the concept of source port. We overlay sport, so we can
3108 		 * maximally reuse the existing code.
3109 		 */
3110 		ni.nai_sport = fin->fin_data[1];
3111 		ni.nai_dport = 0;
3112 	}
3113 
3114 	bzero((char *)nat, sizeof(*nat));
3115 	nat->nat_flags = flags;
3116 	nat->nat_redir = np->in_redir;
3117 	nat->nat_dir = direction;
3118 	nat->nat_pr[0] = fin->fin_p;
3119 	nat->nat_pr[1] = fin->fin_p;
3120 
3121 	/*
3122 	 * Search the current table for a match and create a new mapping
3123 	 * if there is none found.
3124 	 */
3125 	if (np->in_redir & NAT_DIVERTUDP) {
3126 		move = ipf_nat_newdivert(fin, nat, &ni);
3127 
3128 	} else if (np->in_redir & NAT_REWRITE) {
3129 		move = ipf_nat_newrewrite(fin, nat, &ni);
3130 
3131 	} else if (direction == NAT_OUTBOUND) {
3132 		/*
3133 		 * We can now arrange to call this for the same connection
3134 		 * because ipf_nat_new doesn't protect the code path into
3135 		 * this function.
3136 		 */
3137 		natl = ipf_nat_outlookup(fin, nflags, (u_int)fin->fin_p,
3138 				     fin->fin_src, fin->fin_dst);
3139 		if (natl != NULL) {
3140 			KFREE(nat);
3141 			nat = natl;
3142 			goto done;
3143 		}
3144 
3145 		move = ipf_nat_newmap(fin, nat, &ni);
3146 	} else {
3147 		/*
3148 		 * NAT_INBOUND is used for redirects rules
3149 		 */
3150 		natl = ipf_nat_inlookup(fin, nflags, (u_int)fin->fin_p,
3151 					fin->fin_src, fin->fin_dst);
3152 		if (natl != NULL) {
3153 			KFREE(nat);
3154 			nat = natl;
3155 			goto done;
3156 		}
3157 
3158 		move = ipf_nat_newrdr(fin, nat, &ni);
3159 	}
3160 	if (move == -1)
3161 		goto badnat;
3162 
3163 	np = ni.nai_np;
3164 
3165 	nat->nat_mssclamp = np->in_mssclamp;
3166 	nat->nat_me = natsave;
3167 	nat->nat_fr = fin->fin_fr;
3168 	nat->nat_rev = fin->fin_rev;
3169 	nat->nat_ptr = np;
3170 	nat->nat_dlocal = np->in_dlocal;
3171 
3172 	if ((np->in_apr != NULL) && ((nat->nat_flags & NAT_SLAVE) == 0)) {
3173 		if (ipf_proxy_new(fin, nat) == -1) {
3174 			NBUMPSIDED(fin->fin_out, ns_appr_fail);
3175 			goto badnat;
3176 		}
3177 	}
3178 
3179 	nat->nat_ifps[0] = np->in_ifps[0];
3180 	if (np->in_ifps[0] != NULL) {
3181 		COPYIFNAME(np->in_v[0], np->in_ifps[0], nat->nat_ifnames[0]);
3182 	}
3183 
3184 	nat->nat_ifps[1] = np->in_ifps[1];
3185 	if (np->in_ifps[1] != NULL) {
3186 		COPYIFNAME(np->in_v[1], np->in_ifps[1], nat->nat_ifnames[1]);
3187 	}
3188 
3189 	if (ipf_nat_finalise(fin, nat) == -1) {
3190 		goto badnat;
3191 	}
3192 
3193 	np->in_use++;
3194 
3195 	if ((move == 1) && (np->in_flags & IPN_ROUNDR)) {
3196 		if ((np->in_redir & (NAT_REDIRECT|NAT_MAP)) == NAT_REDIRECT) {
3197 			ipf_nat_delrdr(softn, np);
3198 			ipf_nat_addrdr(softn, np);
3199 		} else if ((np->in_redir & (NAT_REDIRECT|NAT_MAP)) == NAT_MAP) {
3200 			ipf_nat_delmap(softn, np);
3201 			ipf_nat_addmap(softn, np);
3202 		}
3203 	}
3204 
3205 	if (flags & SI_WILDP)
3206 		nsp->ns_wilds++;
3207 	nsp->ns_proto[nat->nat_pr[0]]++;
3208 
3209 	goto done;
3210 badnat:
3211 	DT2(ns_badnatnew, fr_info_t *, fin, nat_t *, nat);
3212 	NBUMPSIDE(fin->fin_out, ns_badnatnew);
3213 	if ((hm = nat->nat_hm) != NULL)
3214 		ipf_nat_hostmapdel(softc, &hm);
3215 	KFREE(nat);
3216 	nat = NULL;
3217 done:
3218 	if (nat != NULL && np != NULL)
3219 		np->in_hits++;
3220 	if (natsave != NULL)
3221 		*natsave = nat;
3222 	return nat;
3223 }
3224 
3225 
3226 /* ------------------------------------------------------------------------ */
3227 /* Function:    ipf_nat_finalise                                            */
3228 /* Returns:     int - 0 == sucess, -1 == failure                            */
3229 /* Parameters:  fin(I) - pointer to packet information                      */
3230 /*              nat(I) - pointer to NAT entry                               */
3231 /* Write Lock:  ipf_nat                                                     */
3232 /*                                                                          */
3233 /* This is the tail end of constructing a new NAT entry and is the same     */
3234 /* for both IPv4 and IPv6.                                                  */
3235 /* ------------------------------------------------------------------------ */
3236 /*ARGSUSED*/
3237 static int
3238 ipf_nat_finalise(fr_info_t *fin, nat_t *nat)
3239 {
3240 	ipf_main_softc_t *softc = fin->fin_main_soft;
3241 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3242 	u_32_t sum1, sum2, sumd;
3243 	frentry_t *fr;
3244 	u_32_t flags;
3245 #if SOLARIS && defined(_KERNEL) && (SOLARIS2 >= 6) && defined(ICK_M_CTL_MAGIC)
3246 	qpktinfo_t *qpi = fin->fin_qpi;
3247 #endif
3248 
3249 	flags = nat->nat_flags;
3250 
3251 	switch (nat->nat_pr[0])
3252 	{
3253 	case IPPROTO_ICMP :
3254 		sum1 = LONG_SUM(ntohs(nat->nat_oicmpid));
3255 		sum2 = LONG_SUM(ntohs(nat->nat_nicmpid));
3256 		CALC_SUMD(sum1, sum2, sumd);
3257 		nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
3258 
3259 		break;
3260 
3261 	default :
3262 		sum1 = LONG_SUM(ntohl(nat->nat_osrcaddr) + \
3263 				ntohs(nat->nat_osport));
3264 		sum2 = LONG_SUM(ntohl(nat->nat_nsrcaddr) + \
3265 				ntohs(nat->nat_nsport));
3266 		CALC_SUMD(sum1, sum2, sumd);
3267 		nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
3268 
3269 		sum1 = LONG_SUM(ntohl(nat->nat_odstaddr) + \
3270 				ntohs(nat->nat_odport));
3271 		sum2 = LONG_SUM(ntohl(nat->nat_ndstaddr) + \
3272 				ntohs(nat->nat_ndport));
3273 		CALC_SUMD(sum1, sum2, sumd);
3274 		nat->nat_sumd[0] += (sumd & 0xffff) + (sumd >> 16);
3275 		break;
3276 	}
3277 
3278 	/*
3279 	 * Compute the partial checksum, just in case.
3280 	 * This is only ever placed into outbound packets so care needs
3281 	 * to be taken over which pair of addresses are used.
3282 	 */
3283 	if (nat->nat_dir == NAT_OUTBOUND) {
3284 		sum1 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
3285 		sum1 += LONG_SUM(ntohl(nat->nat_ndstaddr));
3286 	} else {
3287 		sum1 = LONG_SUM(ntohl(nat->nat_osrcaddr));
3288 		sum1 += LONG_SUM(ntohl(nat->nat_odstaddr));
3289 	}
3290 	sum1 += nat->nat_pr[1];
3291 	nat->nat_sumd[1] = (sum1 & 0xffff) + (sum1 >> 16);
3292 
3293 	sum1 = LONG_SUM(ntohl(nat->nat_osrcaddr));
3294 	sum2 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
3295 	CALC_SUMD(sum1, sum2, sumd);
3296 	nat->nat_ipsumd = (sumd & 0xffff) + (sumd >> 16);
3297 
3298 	sum1 = LONG_SUM(ntohl(nat->nat_odstaddr));
3299 	sum2 = LONG_SUM(ntohl(nat->nat_ndstaddr));
3300 	CALC_SUMD(sum1, sum2, sumd);
3301 	nat->nat_ipsumd += (sumd & 0xffff) + (sumd >> 16);
3302 
3303 	nat->nat_v[0] = 4;
3304 	nat->nat_v[1] = 4;
3305 
3306 	if ((nat->nat_ifps[0] != NULL) && (nat->nat_ifps[0] != (void *)-1)) {
3307 		nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
3308 	}
3309 
3310 	if ((nat->nat_ifps[1] != NULL) && (nat->nat_ifps[1] != (void *)-1)) {
3311 		nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
3312 	}
3313 
3314 	if ((nat->nat_flags & SI_CLONE) == 0)
3315 		nat->nat_sync = ipf_sync_new(softc, SMC_NAT, fin, nat);
3316 
3317 	if (ipf_nat_insert(softc, softn, nat) == 0) {
3318 		if (softn->ipf_nat_logging)
3319 			ipf_nat_log(softc, softn, nat, NL_NEW);
3320 		fr = nat->nat_fr;
3321 		if (fr != NULL) {
3322 			MUTEX_ENTER(&fr->fr_lock);
3323 			fr->fr_ref++;
3324 			MUTEX_EXIT(&fr->fr_lock);
3325 		}
3326 		return 0;
3327 	}
3328 
3329 	NBUMPSIDED(fin->fin_out, ns_unfinalised);
3330 	/*
3331 	 * nat_insert failed, so cleanup time...
3332 	 */
3333 	if (nat->nat_sync != NULL)
3334 		ipf_sync_del_nat(softc->ipf_sync_soft, nat->nat_sync);
3335 	return -1;
3336 }
3337 
3338 
3339 /* ------------------------------------------------------------------------ */
3340 /* Function:    ipf_nat_insert                                              */
3341 /* Returns:     int - 0 == sucess, -1 == failure                            */
3342 /* Parameters:  softc(I) - pointer to soft context main structure           */
3343 /*              softn(I) - pointer to NAT context structure                 */
3344 /*              nat(I) - pointer to NAT structure                           */
3345 /* Write Lock:  ipf_nat                                                     */
3346 /*                                                                          */
3347 /* Insert a NAT entry into the hash tables for searching and add it to the  */
3348 /* list of active NAT entries.  Adjust global counters when complete.       */
3349 /* ------------------------------------------------------------------------ */
3350 int
3351 ipf_nat_insert(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, nat_t *nat)
3352 {
3353 	u_int hv0, hv1;
3354 	u_int sp, dp;
3355 	ipnat_t *in;
3356 
3357 	/*
3358 	 * Try and return an error as early as possible, so calculate the hash
3359 	 * entry numbers first and then proceed.
3360 	 */
3361 	if ((nat->nat_flags & (SI_W_SPORT|SI_W_DPORT)) == 0) {
3362 		if ((nat->nat_flags & IPN_TCPUDP) != 0) {
3363 			sp = nat->nat_osport;
3364 			dp = nat->nat_odport;
3365 		} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
3366 			sp = 0;
3367 			dp = nat->nat_oicmpid;
3368 		} else {
3369 			sp = 0;
3370 			dp = 0;
3371 		}
3372 		hv0 = NAT_HASH_FN(nat->nat_osrcaddr, sp, 0xffffffff);
3373 		hv0 = NAT_HASH_FN(nat->nat_odstaddr, hv0 + dp, 0xffffffff);
3374 		/*
3375 		 * TRACE nat_osrcaddr, nat_osport, nat_odstaddr,
3376 		 * nat_odport, hv0
3377 		 */
3378 
3379 		if ((nat->nat_flags & IPN_TCPUDP) != 0) {
3380 			sp = nat->nat_nsport;
3381 			dp = nat->nat_ndport;
3382 		} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
3383 			sp = 0;
3384 			dp = nat->nat_nicmpid;
3385 		} else {
3386 			sp = 0;
3387 			dp = 0;
3388 		}
3389 		hv1 = NAT_HASH_FN(nat->nat_nsrcaddr, sp, 0xffffffff);
3390 		hv1 = NAT_HASH_FN(nat->nat_ndstaddr, hv1 + dp, 0xffffffff);
3391 		/*
3392 		 * TRACE nat_nsrcaddr, nat_nsport, nat_ndstaddr,
3393 		 * nat_ndport, hv1
3394 		 */
3395 	} else {
3396 		hv0 = NAT_HASH_FN(nat->nat_osrcaddr, 0, 0xffffffff);
3397 		hv0 = NAT_HASH_FN(nat->nat_odstaddr, hv0, 0xffffffff);
3398 		/* TRACE nat_osrcaddr, nat_odstaddr, hv0 */
3399 
3400 		hv1 = NAT_HASH_FN(nat->nat_nsrcaddr, 0, 0xffffffff);
3401 		hv1 = NAT_HASH_FN(nat->nat_ndstaddr, hv1, 0xffffffff);
3402 		/* TRACE nat_nsrcaddr, nat_ndstaddr, hv1 */
3403 	}
3404 
3405 	nat->nat_hv[0] = hv0;
3406 	nat->nat_hv[1] = hv1;
3407 
3408 	MUTEX_INIT(&nat->nat_lock, "nat entry lock");
3409 
3410 	in = nat->nat_ptr;
3411 	nat->nat_ref = nat->nat_me ? 2 : 1;
3412 
3413 	nat->nat_ifnames[0][LIFNAMSIZ - 1] = '\0';
3414 	nat->nat_ifps[0] = ipf_resolvenic(softc, nat->nat_ifnames[0], 4);
3415 
3416 	if (nat->nat_ifnames[1][0] != '\0') {
3417 		nat->nat_ifnames[1][LIFNAMSIZ - 1] = '\0';
3418 		nat->nat_ifps[1] = ipf_resolvenic(softc,
3419 						  nat->nat_ifnames[1], 4);
3420 	} else if (in->in_ifnames[1] != -1) {
3421 		char *name;
3422 
3423 		name = in->in_names + in->in_ifnames[1];
3424 		if (name[1] != '\0' && name[0] != '-' && name[0] != '*') {
3425 			(void) strncpy(nat->nat_ifnames[1],
3426 				       nat->nat_ifnames[0], LIFNAMSIZ);
3427 			nat->nat_ifnames[1][LIFNAMSIZ - 1] = '\0';
3428 			nat->nat_ifps[1] = nat->nat_ifps[0];
3429 		}
3430 	}
3431 	if ((nat->nat_ifps[0] != NULL) && (nat->nat_ifps[0] != (void *)-1)) {
3432 		nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
3433 	}
3434 	if ((nat->nat_ifps[1] != NULL) && (nat->nat_ifps[1] != (void *)-1)) {
3435 		nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
3436 	}
3437 
3438 	return ipf_nat_hashtab_add(softc, softn, nat);
3439 }
3440 
3441 
3442 /* ------------------------------------------------------------------------ */
3443 /* Function:    ipf_nat_hashtab_add                                         */
3444 /* Parameters:  softc(I) - pointer to soft context main structure           */
3445 /*              softn(I) - pointer to NAT context structure                 */
3446 /*              nat(I) - pointer to NAT structure                           */
3447 /*                                                                          */
3448 /* Handle the insertion of a NAT entry into the table/list.                 */
3449 /* ------------------------------------------------------------------------ */
3450 int
3451 ipf_nat_hashtab_add(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, nat_t *nat)
3452 {
3453 	nat_t **natp;
3454 	u_int hv0;
3455 	u_int hv1;
3456 
3457 	hv0 = nat->nat_hv[0] % softn->ipf_nat_table_sz;
3458 	hv1 = nat->nat_hv[1] % softn->ipf_nat_table_sz;
3459 
3460 	if (nat->nat_dir == NAT_INBOUND || nat->nat_dir == NAT_DIVERTIN) {
3461 		u_int swap;
3462 
3463 		swap = hv0;
3464 		hv0 = hv1;
3465 		hv1 = swap;
3466 	}
3467 
3468 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen[hv0] >=
3469 	    softn->ipf_nat_maxbucket) {
3470 		DT1(ns_bucket_max_0, int,
3471 		    softn->ipf_nat_stats.ns_side[0].ns_bucketlen[hv0]);
3472 		NBUMPSIDE(0, ns_bucket_max);
3473 		return -1;
3474 	}
3475 
3476 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen[hv1] >=
3477 	    softn->ipf_nat_maxbucket) {
3478 		DT1(ns_bucket_max_1, int,
3479 		    softn->ipf_nat_stats.ns_side[1].ns_bucketlen[hv1]);
3480 		NBUMPSIDE(1, ns_bucket_max);
3481 		return -1;
3482 	}
3483 
3484 	/*
3485 	 * The ordering of operations in the list and hash table insertion
3486 	 * is very important.  The last operation for each task should be
3487 	 * to update the top of the list, after all the "nexts" have been
3488 	 * done so that walking the list while it is being done does not
3489 	 * find strange pointers.
3490 	 *
3491 	 * Global list of NAT instances
3492 	 */
3493 	nat->nat_next = softn->ipf_nat_instances;
3494 	nat->nat_pnext = &softn->ipf_nat_instances;
3495 	if (softn->ipf_nat_instances)
3496 		softn->ipf_nat_instances->nat_pnext = &nat->nat_next;
3497 	softn->ipf_nat_instances = nat;
3498 
3499 	/*
3500 	 * Inbound hash table.
3501 	 */
3502 	natp = &softn->ipf_nat_table[0][hv0];
3503 	nat->nat_phnext[0] = natp;
3504 	nat->nat_hnext[0] = *natp;
3505 	if (*natp) {
3506 		(*natp)->nat_phnext[0] = &nat->nat_hnext[0];
3507 	} else {
3508 		NBUMPSIDE(0, ns_inuse);
3509 	}
3510 	*natp = nat;
3511 	NBUMPSIDE(0, ns_bucketlen[hv0]);
3512 
3513 	/*
3514 	 * Outbound hash table.
3515 	 */
3516 	natp = &softn->ipf_nat_table[1][hv1];
3517 	nat->nat_phnext[1] = natp;
3518 	nat->nat_hnext[1] = *natp;
3519 	if (*natp)
3520 		(*natp)->nat_phnext[1] = &nat->nat_hnext[1];
3521 	else {
3522 		NBUMPSIDE(1, ns_inuse);
3523 	}
3524 	*natp = nat;
3525 	NBUMPSIDE(1, ns_bucketlen[hv1]);
3526 
3527 	ipf_nat_setqueue(softc, softn, nat);
3528 
3529 	if (nat->nat_dir & NAT_OUTBOUND) {
3530 		NBUMPSIDE(1, ns_added);
3531 	} else {
3532 		NBUMPSIDE(0, ns_added);
3533 	}
3534 	softn->ipf_nat_stats.ns_active++;
3535 	return 0;
3536 }
3537 
3538 
3539 /* ------------------------------------------------------------------------ */
3540 /* Function:    ipf_nat_icmperrorlookup                                     */
3541 /* Returns:     nat_t* - point to matching NAT structure                    */
3542 /* Parameters:  fin(I) - pointer to packet information                      */
3543 /*              dir(I) - direction of packet (in/out)                       */
3544 /*                                                                          */
3545 /* Check if the ICMP error message is related to an existing TCP, UDP or    */
3546 /* ICMP query nat entry.  It is assumed that the packet is already of the   */
3547 /* the required length.                                                     */
3548 /* ------------------------------------------------------------------------ */
3549 nat_t *
3550 ipf_nat_icmperrorlookup(fr_info_t *fin, int dir)
3551 {
3552 	ipf_main_softc_t *softc = fin->fin_main_soft;
3553 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3554 	int flags = 0, type, minlen;
3555 	icmphdr_t *icmp, *orgicmp;
3556 	nat_stat_side_t *nside;
3557 	tcphdr_t *tcp = NULL;
3558 	u_short data[2];
3559 	nat_t *nat;
3560 	ip_t *oip;
3561 	u_int p;
3562 
3563 	icmp = fin->fin_dp;
3564 	type = icmp->icmp_type;
3565 	nside = &softn->ipf_nat_stats.ns_side[fin->fin_out];
3566 	/*
3567 	 * Does it at least have the return (basic) IP header ?
3568 	 * Only a basic IP header (no options) should be with an ICMP error
3569 	 * header.  Also, if it's not an error type, then return.
3570 	 */
3571 	if ((fin->fin_hlen != sizeof(ip_t)) || !(fin->fin_flx & FI_ICMPERR)) {
3572 		ATOMIC_INCL(nside->ns_icmp_basic);
3573 		return NULL;
3574 	}
3575 
3576 	/*
3577 	 * Check packet size
3578 	 */
3579 	oip = (ip_t *)((char *)fin->fin_dp + 8);
3580 	minlen = IP_HL(oip) << 2;
3581 	if ((minlen < sizeof(ip_t)) ||
3582 	    (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen)) {
3583 		ATOMIC_INCL(nside->ns_icmp_size);
3584 		return NULL;
3585 	}
3586 
3587 	/*
3588 	 * Is the buffer big enough for all of it ?  It's the size of the IP
3589 	 * header claimed in the encapsulated part which is of concern.  It
3590 	 * may be too big to be in this buffer but not so big that it's
3591 	 * outside the ICMP packet, leading to TCP deref's causing problems.
3592 	 * This is possible because we don't know how big oip_hl is when we
3593 	 * do the pullup early in ipf_check() and thus can't gaurantee it is
3594 	 * all here now.
3595 	 */
3596 #ifdef  ipf_nat_KERNEL
3597 	{
3598 	mb_t *m;
3599 
3600 	m = fin->fin_m;
3601 # if defined(MENTAT)
3602 	if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN >
3603 	    (char *)m->b_wptr) {
3604 		ATOMIC_INCL(nside->ns_icmp_mbuf);
3605 		return NULL;
3606 	}
3607 # else
3608 	if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN >
3609 	    (char *)fin->fin_ip + M_LEN(m)) {
3610 		ATOMIC_INCL(nside->ns_icmp_mbuf);
3611 		return NULL;
3612 	}
3613 # endif
3614 	}
3615 #endif
3616 
3617 	if (fin->fin_daddr != oip->ip_src.s_addr) {
3618 		ATOMIC_INCL(nside->ns_icmp_address);
3619 		return NULL;
3620 	}
3621 
3622 	p = oip->ip_p;
3623 	if (p == IPPROTO_TCP)
3624 		flags = IPN_TCP;
3625 	else if (p == IPPROTO_UDP)
3626 		flags = IPN_UDP;
3627 	else if (p == IPPROTO_ICMP) {
3628 		orgicmp = (icmphdr_t *)((char *)oip + (IP_HL(oip) << 2));
3629 
3630 		/* see if this is related to an ICMP query */
3631 		if (ipf_nat_icmpquerytype(orgicmp->icmp_type)) {
3632 			data[0] = fin->fin_data[0];
3633 			data[1] = fin->fin_data[1];
3634 			fin->fin_data[0] = 0;
3635 			fin->fin_data[1] = orgicmp->icmp_id;
3636 
3637 			flags = IPN_ICMPERR|IPN_ICMPQUERY;
3638 			/*
3639 			 * NOTE : dir refers to the direction of the original
3640 			 *        ip packet. By definition the icmp error
3641 			 *        message flows in the opposite direction.
3642 			 */
3643 			if (dir == NAT_INBOUND)
3644 				nat = ipf_nat_inlookup(fin, flags, p,
3645 						       oip->ip_dst,
3646 						       oip->ip_src);
3647 			else
3648 				nat = ipf_nat_outlookup(fin, flags, p,
3649 							oip->ip_dst,
3650 							oip->ip_src);
3651 			fin->fin_data[0] = data[0];
3652 			fin->fin_data[1] = data[1];
3653 			return nat;
3654 		}
3655 	}
3656 
3657 	if (flags & IPN_TCPUDP) {
3658 		minlen += 8;		/* + 64bits of data to get ports */
3659 		/* TRACE (fin,minlen) */
3660 		if (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen) {
3661 			ATOMIC_INCL(nside->ns_icmp_short);
3662 			return NULL;
3663 		}
3664 
3665 		data[0] = fin->fin_data[0];
3666 		data[1] = fin->fin_data[1];
3667 		tcp = (tcphdr_t *)((char *)oip + (IP_HL(oip) << 2));
3668 		fin->fin_data[0] = ntohs(tcp->th_dport);
3669 		fin->fin_data[1] = ntohs(tcp->th_sport);
3670 
3671 		if (dir == NAT_INBOUND) {
3672 			nat = ipf_nat_inlookup(fin, flags, p, oip->ip_dst,
3673 					       oip->ip_src);
3674 		} else {
3675 			nat = ipf_nat_outlookup(fin, flags, p, oip->ip_dst,
3676 					    oip->ip_src);
3677 		}
3678 		fin->fin_data[0] = data[0];
3679 		fin->fin_data[1] = data[1];
3680 		return nat;
3681 	}
3682 	if (dir == NAT_INBOUND)
3683 		nat = ipf_nat_inlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
3684 	else
3685 		nat = ipf_nat_outlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
3686 
3687 	return nat;
3688 }
3689 
3690 
3691 /* ------------------------------------------------------------------------ */
3692 /* Function:    ipf_nat_icmperror                                           */
3693 /* Returns:     nat_t* - point to matching NAT structure                    */
3694 /* Parameters:  fin(I)    - pointer to packet information                   */
3695 /*              nflags(I) - NAT flags for this packet                       */
3696 /*              dir(I)    - direction of packet (in/out)                    */
3697 /*                                                                          */
3698 /* Fix up an ICMP packet which is an error message for an existing NAT      */
3699 /* session.  This will correct both packet header data and checksums.       */
3700 /*                                                                          */
3701 /* This should *ONLY* be used for incoming ICMP error packets to make sure  */
3702 /* a NAT'd ICMP packet gets correctly recognised.                           */
3703 /* ------------------------------------------------------------------------ */
3704 nat_t *
3705 ipf_nat_icmperror(fr_info_t *fin, u_int *nflags, int dir)
3706 {
3707 	ipf_main_softc_t *softc = fin->fin_main_soft;
3708 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3709 	u_32_t sum1, sum2, sumd, sumd2;
3710 	struct in_addr a1, a2, a3, a4;
3711 	int flags, dlen, odst;
3712 	icmphdr_t *icmp;
3713 	u_short *csump;
3714 	tcphdr_t *tcp;
3715 	nat_t *nat;
3716 	ip_t *oip;
3717 	void *dp;
3718 
3719 	if ((fin->fin_flx & (FI_SHORT|FI_FRAGBODY))) {
3720 		NBUMPSIDED(fin->fin_out, ns_icmp_short);
3721 		return NULL;
3722 	}
3723 
3724 	/*
3725 	 * ipf_nat_icmperrorlookup() will return NULL for `defective' packets.
3726 	 */
3727 	if ((fin->fin_v != 4) || !(nat = ipf_nat_icmperrorlookup(fin, dir))) {
3728 		NBUMPSIDED(fin->fin_out, ns_icmp_notfound);
3729 		return NULL;
3730 	}
3731 
3732 	tcp = NULL;
3733 	csump = NULL;
3734 	flags = 0;
3735 	sumd2 = 0;
3736 	*nflags = IPN_ICMPERR;
3737 	icmp = fin->fin_dp;
3738 	oip = (ip_t *)&icmp->icmp_ip;
3739 	dp = (((char *)oip) + (IP_HL(oip) << 2));
3740 	if (oip->ip_p == IPPROTO_TCP) {
3741 		tcp = (tcphdr_t *)dp;
3742 		csump = (u_short *)&tcp->th_sum;
3743 		flags = IPN_TCP;
3744 	} else if (oip->ip_p == IPPROTO_UDP) {
3745 		udphdr_t *udp;
3746 
3747 		udp = (udphdr_t *)dp;
3748 		tcp = (tcphdr_t *)dp;
3749 		csump = (u_short *)&udp->uh_sum;
3750 		flags = IPN_UDP;
3751 	} else if (oip->ip_p == IPPROTO_ICMP)
3752 		flags = IPN_ICMPQUERY;
3753 	dlen = fin->fin_plen - ((char *)dp - (char *)fin->fin_ip);
3754 
3755 	/*
3756 	 * Need to adjust ICMP header to include the real IP#'s and
3757 	 * port #'s.  Only apply a checksum change relative to the
3758 	 * IP address change as it will be modified again in ipf_nat_checkout
3759 	 * for both address and port.  Two checksum changes are
3760 	 * necessary for the two header address changes.  Be careful
3761 	 * to only modify the checksum once for the port # and twice
3762 	 * for the IP#.
3763 	 */
3764 
3765 	/*
3766 	 * Step 1
3767 	 * Fix the IP addresses in the offending IP packet. You also need
3768 	 * to adjust the IP header checksum of that offending IP packet.
3769 	 *
3770 	 * Normally, you would expect that the ICMP checksum of the
3771 	 * ICMP error message needs to be adjusted as well for the
3772 	 * IP address change in oip.
3773 	 * However, this is a NOP, because the ICMP checksum is
3774 	 * calculated over the complete ICMP packet, which includes the
3775 	 * changed oip IP addresses and oip->ip_sum. However, these
3776 	 * two changes cancel each other out (if the delta for
3777 	 * the IP address is x, then the delta for ip_sum is minus x),
3778 	 * so no change in the icmp_cksum is necessary.
3779 	 *
3780 	 * Inbound ICMP
3781 	 * ------------
3782 	 * MAP rule, SRC=a,DST=b -> SRC=c,DST=b
3783 	 * - response to outgoing packet (a,b)=>(c,b) (OIP_SRC=c,OIP_DST=b)
3784 	 * - OIP_SRC(c)=nat_newsrcip,          OIP_DST(b)=nat_newdstip
3785 	 *=> OIP_SRC(c)=nat_oldsrcip,          OIP_DST(b)=nat_olddstip
3786 	 *
3787 	 * RDR rule, SRC=a,DST=b -> SRC=a,DST=c
3788 	 * - response to outgoing packet (c,a)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3789 	 * - OIP_SRC(b)=nat_olddstip,          OIP_DST(a)=nat_oldsrcip
3790 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3791 	 *
3792 	 * REWRITE out rule, SRC=a,DST=b -> SRC=c,DST=d
3793 	 * - response to outgoing packet (a,b)=>(c,d) (OIP_SRC=c,OIP_DST=d)
3794 	 * - OIP_SRC(c)=nat_newsrcip,          OIP_DST(d)=nat_newdstip
3795 	 *=> OIP_SRC(c)=nat_oldsrcip,          OIP_DST(d)=nat_olddstip
3796 	 *
3797 	 * REWRITE in rule, SRC=a,DST=b -> SRC=c,DST=d
3798 	 * - response to outgoing packet (d,c)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3799 	 * - OIP_SRC(b)=nat_olddstip,          OIP_DST(a)=nat_oldsrcip
3800 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3801 	 *
3802 	 * Outbound ICMP
3803 	 * -------------
3804 	 * MAP rule, SRC=a,DST=b -> SRC=c,DST=b
3805 	 * - response to incoming packet (b,c)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3806 	 * - OIP_SRC(b)=nat_olddstip,          OIP_DST(a)=nat_oldsrcip
3807 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3808 	 *
3809 	 * RDR rule, SRC=a,DST=b -> SRC=a,DST=c
3810 	 * - response to incoming packet (a,b)=>(a,c) (OIP_SRC=a,OIP_DST=c)
3811 	 * - OIP_SRC(a)=nat_newsrcip,          OIP_DST(c)=nat_newdstip
3812 	 *=> OIP_SRC(a)=nat_oldsrcip,          OIP_DST(c)=nat_olddstip
3813 	 *
3814 	 * REWRITE out rule, SRC=a,DST=b -> SRC=c,DST=d
3815 	 * - response to incoming packet (d,c)=>(b,a) (OIP_SRC=c,OIP_DST=d)
3816 	 * - OIP_SRC(c)=nat_olddstip,          OIP_DST(d)=nat_oldsrcip
3817 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3818 	 *
3819 	 * REWRITE in rule, SRC=a,DST=b -> SRC=c,DST=d
3820 	 * - response to incoming packet (a,b)=>(c,d) (OIP_SRC=b,OIP_DST=a)
3821 	 * - OIP_SRC(b)=nat_newsrcip,          OIP_DST(a)=nat_newdstip
3822 	 *=> OIP_SRC(a)=nat_oldsrcip,          OIP_DST(c)=nat_olddstip
3823 	 */
3824 
3825 	if (((fin->fin_out == 0) && ((nat->nat_redir & NAT_MAP) != 0)) ||
3826 	    ((fin->fin_out == 1) && ((nat->nat_redir & NAT_REDIRECT) != 0))) {
3827 		a1.s_addr = ntohl(nat->nat_osrcaddr);
3828 		a4.s_addr = ntohl(oip->ip_src.s_addr);
3829 		a3.s_addr = ntohl(nat->nat_odstaddr);
3830 		a2.s_addr = ntohl(oip->ip_dst.s_addr);
3831 		oip->ip_src.s_addr = htonl(a1.s_addr);
3832 		oip->ip_dst.s_addr = htonl(a3.s_addr);
3833 		odst = 1;
3834 	} else {
3835 		a1.s_addr = ntohl(nat->nat_ndstaddr);
3836 		a2.s_addr = ntohl(oip->ip_dst.s_addr);
3837 		a3.s_addr = ntohl(nat->nat_nsrcaddr);
3838 		a4.s_addr = ntohl(oip->ip_src.s_addr);
3839 		oip->ip_dst.s_addr = htonl(a3.s_addr);
3840 		oip->ip_src.s_addr = htonl(a1.s_addr);
3841 		odst = 0;
3842 	}
3843 	sum1 = 0;
3844 	sum2 = 0;
3845 	sumd = 0;
3846 	CALC_SUMD(a2.s_addr, a3.s_addr, sum1);
3847 	CALC_SUMD(a4.s_addr, a1.s_addr, sum2);
3848 	sumd = sum2 + sum1;
3849 	if (sumd != 0)
3850 		ipf_fix_datacksum(&oip->ip_sum, sumd);
3851 
3852 	sumd2 = sumd;
3853 	sum1 = 0;
3854 	sum2 = 0;
3855 
3856 	/*
3857 	 * Fix UDP pseudo header checksum to compensate for the
3858 	 * IP address change.
3859 	 */
3860 	if (((flags & IPN_TCPUDP) != 0) && (dlen >= 4)) {
3861 		u_32_t sum3, sum4, sumt;
3862 
3863 		/*
3864 		 * Step 2 :
3865 		 * For offending TCP/UDP IP packets, translate the ports as
3866 		 * well, based on the NAT specification. Of course such
3867 		 * a change may be reflected in the ICMP checksum as well.
3868 		 *
3869 		 * Since the port fields are part of the TCP/UDP checksum
3870 		 * of the offending IP packet, you need to adjust that checksum
3871 		 * as well... except that the change in the port numbers should
3872 		 * be offset by the checksum change.  However, the TCP/UDP
3873 		 * checksum will also need to change if there has been an
3874 		 * IP address change.
3875 		 */
3876 		if (odst == 1) {
3877 			sum1 = ntohs(nat->nat_osport);
3878 			sum4 = ntohs(tcp->th_sport);
3879 			sum3 = ntohs(nat->nat_odport);
3880 			sum2 = ntohs(tcp->th_dport);
3881 
3882 			tcp->th_sport = htons(sum1);
3883 			tcp->th_dport = htons(sum3);
3884 		} else {
3885 			sum1 = ntohs(nat->nat_ndport);
3886 			sum2 = ntohs(tcp->th_dport);
3887 			sum3 = ntohs(nat->nat_nsport);
3888 			sum4 = ntohs(tcp->th_sport);
3889 
3890 			tcp->th_dport = htons(sum3);
3891 			tcp->th_sport = htons(sum1);
3892 		}
3893 		CALC_SUMD(sum4, sum1, sumt);
3894 		sumd += sumt;
3895 		CALC_SUMD(sum2, sum3, sumt);
3896 		sumd += sumt;
3897 
3898 		if (sumd != 0 || sumd2 != 0) {
3899 			/*
3900 			 * At this point, sumd is the delta to apply to the
3901 			 * TCP/UDP header, given the changes in both the IP
3902 			 * address and the ports and sumd2 is the delta to
3903 			 * apply to the ICMP header, given the IP address
3904 			 * change delta that may need to be applied to the
3905 			 * TCP/UDP checksum instead.
3906 			 *
3907 			 * If we will both the IP and TCP/UDP checksums
3908 			 * then the ICMP checksum changes by the address
3909 			 * delta applied to the TCP/UDP checksum.  If we
3910 			 * do not change the TCP/UDP checksum them we
3911 			 * apply the delta in ports to the ICMP checksum.
3912 			 */
3913 			if (oip->ip_p == IPPROTO_UDP) {
3914 				if ((dlen >= 8) && (*csump != 0)) {
3915 					ipf_fix_datacksum(csump, sumd);
3916 				} else {
3917 					CALC_SUMD(sum1, sum4, sumd2);
3918 					CALC_SUMD(sum3, sum2, sumt);
3919 					sumd2 += sumt;
3920 				}
3921 			} else if (oip->ip_p == IPPROTO_TCP) {
3922 				if (dlen >= 18) {
3923 					ipf_fix_datacksum(csump, sumd);
3924 				} else {
3925 					CALC_SUMD(sum1, sum4, sumd2);
3926 					CALC_SUMD(sum3, sum2, sumt);
3927 					sumd2 += sumt;
3928 				}
3929 			}
3930 			if (sumd2 != 0) {
3931 				sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3932 				sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3933 				sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3934 				ipf_fix_incksum(0, &icmp->icmp_cksum, sumd2, 0);
3935 			}
3936 		}
3937 	} else if (((flags & IPN_ICMPQUERY) != 0) && (dlen >= 8)) {
3938 		icmphdr_t *orgicmp;
3939 
3940 		/*
3941 		 * XXX - what if this is bogus hl and we go off the end ?
3942 		 * In this case, ipf_nat_icmperrorlookup() will have
3943 		 * returned NULL.
3944 		 */
3945 		orgicmp = (icmphdr_t *)dp;
3946 
3947 		if (odst == 1) {
3948 			if (orgicmp->icmp_id != nat->nat_osport) {
3949 
3950 				/*
3951 				 * Fix ICMP checksum (of the offening ICMP
3952 				 * query packet) to compensate the change
3953 				 * in the ICMP id of the offending ICMP
3954 				 * packet.
3955 				 *
3956 				 * Since you modify orgicmp->icmp_id with
3957 				 * a delta (say x) and you compensate that
3958 				 * in origicmp->icmp_cksum with a delta
3959 				 * minus x, you don't have to adjust the
3960 				 * overall icmp->icmp_cksum
3961 				 */
3962 				sum1 = ntohs(orgicmp->icmp_id);
3963 				sum2 = ntohs(nat->nat_oicmpid);
3964 				CALC_SUMD(sum1, sum2, sumd);
3965 				orgicmp->icmp_id = nat->nat_oicmpid;
3966 				ipf_fix_datacksum(&orgicmp->icmp_cksum, sumd);
3967 			}
3968 		} /* nat_dir == NAT_INBOUND is impossible for icmp queries */
3969 	}
3970 	return nat;
3971 }
3972 
3973 
3974 /*
3975  *       MAP-IN    MAP-OUT   RDR-IN   RDR-OUT
3976  * osrc    X       == src    == src      X
3977  * odst    X       == dst    == dst      X
3978  * nsrc  == dst      X         X      == dst
3979  * ndst  == src      X         X      == src
3980  * MAP = NAT_OUTBOUND, RDR = NAT_INBOUND
3981  */
3982 /*
3983  * NB: these lookups don't lock access to the list, it assumed that it has
3984  * already been done!
3985  */
3986 /* ------------------------------------------------------------------------ */
3987 /* Function:    ipf_nat_inlookup                                            */
3988 /* Returns:     nat_t* - NULL == no match,                                  */
3989 /*                       else pointer to matching NAT entry                 */
3990 /* Parameters:  fin(I)    - pointer to packet information                   */
3991 /*              flags(I)  - NAT flags for this packet                       */
3992 /*              p(I)      - protocol for this packet                        */
3993 /*              src(I)    - source IP address                               */
3994 /*              mapdst(I) - destination IP address                          */
3995 /*                                                                          */
3996 /* Lookup a nat entry based on the mapped destination ip address/port and   */
3997 /* real source address/port.  We use this lookup when receiving a packet,   */
3998 /* we're looking for a table entry, based on the destination address.       */
3999 /*                                                                          */
4000 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY.         */
4001 /*                                                                          */
4002 /* NOTE: IT IS ASSUMED THAT  IS ONLY HELD WITH A READ LOCK WHEN             */
4003 /*       THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags.             */
4004 /*                                                                          */
4005 /* flags   -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if   */
4006 /*            the packet is of said protocol                                */
4007 /* ------------------------------------------------------------------------ */
4008 nat_t *
4009 ipf_nat_inlookup(fr_info_t *fin, u_int flags, u_int p, struct in_addr src,
4010     struct in_addr mapdst)
4011 {
4012 	ipf_main_softc_t *softc = fin->fin_main_soft;
4013 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4014 	u_short sport, dport;
4015 	grehdr_t *gre;
4016 	ipnat_t *ipn;
4017 	u_int sflags;
4018 	nat_t *nat;
4019 	int nflags;
4020 	u_32_t dst;
4021 	void *ifp;
4022 	u_int hv, rhv;
4023 
4024 	ifp = fin->fin_ifp;
4025 	gre = NULL;
4026 	dst = mapdst.s_addr;
4027 	sflags = flags & NAT_TCPUDPICMP;
4028 
4029 	switch (p)
4030 	{
4031 	case IPPROTO_TCP :
4032 	case IPPROTO_UDP :
4033 		sport = htons(fin->fin_data[0]);
4034 		dport = htons(fin->fin_data[1]);
4035 		break;
4036 	case IPPROTO_ICMP :
4037 		if (flags & IPN_ICMPERR) {
4038 			sport = fin->fin_data[1];
4039 			dport = 0;
4040 		} else {
4041 			dport = fin->fin_data[1];
4042 			sport = 0;
4043 		}
4044 		break;
4045 	default :
4046 		sport = 0;
4047 		dport = 0;
4048 		break;
4049 	}
4050 
4051 
4052 	if ((flags & SI_WILDP) != 0)
4053 		goto find_in_wild_ports;
4054 
4055 	rhv = NAT_HASH_FN(dst, dport, 0xffffffff);
4056 	rhv = NAT_HASH_FN(src.s_addr, rhv + sport, 0xffffffff);
4057 	hv = rhv % softn->ipf_nat_table_sz;
4058 	nat = softn->ipf_nat_table[1][hv];
4059 	/* TRACE dst, dport, src, sport, hv, nat */
4060 
4061 	for (; nat; nat = nat->nat_hnext[1]) {
4062 		if (nat->nat_ifps[0] != NULL) {
4063 			if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
4064 				continue;
4065 		}
4066 
4067 		if (nat->nat_pr[0] != p)
4068 			continue;
4069 
4070 		switch (nat->nat_dir)
4071 		{
4072 		case NAT_INBOUND :
4073 		case NAT_DIVERTIN :
4074 			if (nat->nat_v[0] != 4)
4075 				continue;
4076 			if (nat->nat_osrcaddr != src.s_addr ||
4077 			    nat->nat_odstaddr != dst)
4078 				continue;
4079 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4080 				if (nat->nat_osport != sport)
4081 					continue;
4082 				if (nat->nat_odport != dport)
4083 					continue;
4084 
4085 			} else if (p == IPPROTO_ICMP) {
4086 				if (nat->nat_osport != dport) {
4087 					continue;
4088 				}
4089 			}
4090 			break;
4091 		case NAT_DIVERTOUT :
4092 			if (nat->nat_dlocal)
4093 				continue;
4094 		case NAT_OUTBOUND :
4095 			if (nat->nat_v[1] != 4)
4096 				continue;
4097 			if (nat->nat_dlocal)
4098 				continue;
4099 			if (nat->nat_dlocal)
4100 				continue;
4101 			if (nat->nat_ndstaddr != src.s_addr ||
4102 			    nat->nat_nsrcaddr != dst)
4103 				continue;
4104 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4105 				if (nat->nat_ndport != sport)
4106 					continue;
4107 				if (nat->nat_nsport != dport)
4108 					continue;
4109 
4110 			} else if (p == IPPROTO_ICMP) {
4111 				if (nat->nat_osport != dport) {
4112 					continue;
4113 				}
4114 			}
4115 			break;
4116 		}
4117 
4118 
4119 		if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4120 			ipn = nat->nat_ptr;
4121 			if ((ipn != NULL) && (nat->nat_aps != NULL))
4122 				if (ipf_proxy_match(fin, nat) != 0)
4123 					continue;
4124 		}
4125 		if ((nat->nat_ifps[0] == NULL) && (ifp != NULL)) {
4126 			nat->nat_ifps[0] = ifp;
4127 			nat->nat_mtu[0] = GETIFMTU_4(ifp);
4128 		}
4129 		return nat;
4130 	}
4131 
4132 	/*
4133 	 * So if we didn't find it but there are wildcard members in the hash
4134 	 * table, go back and look for them.  We do this search and update here
4135 	 * because it is modifying the NAT table and we want to do this only
4136 	 * for the first packet that matches.  The exception, of course, is
4137 	 * for "dummy" (FI_IGNORE) lookups.
4138 	 */
4139 find_in_wild_ports:
4140 	if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH)) {
4141 		NBUMPSIDEX(0, ns_lookup_miss, ns_lookup_miss_0);
4142 		return NULL;
4143 	}
4144 	if (softn->ipf_nat_stats.ns_wilds == 0 || (fin->fin_flx & FI_NOWILD)) {
4145 		NBUMPSIDEX(0, ns_lookup_nowild, ns_lookup_nowild_0);
4146 		return NULL;
4147 	}
4148 
4149 	RWLOCK_EXIT(&softc->ipf_nat);
4150 
4151 	hv = NAT_HASH_FN(dst, 0, 0xffffffff);
4152 	hv = NAT_HASH_FN(src.s_addr, hv, softn->ipf_nat_table_sz);
4153 	WRITE_ENTER(&softc->ipf_nat);
4154 
4155 	nat = softn->ipf_nat_table[1][hv];
4156 	/* TRACE dst, src, hv, nat */
4157 	for (; nat; nat = nat->nat_hnext[1]) {
4158 		if (nat->nat_ifps[0] != NULL) {
4159 			if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
4160 				continue;
4161 		}
4162 
4163 		if (nat->nat_pr[0] != fin->fin_p)
4164 			continue;
4165 
4166 		switch (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND))
4167 		{
4168 		case NAT_INBOUND :
4169 			if (nat->nat_v[0] != 4)
4170 				continue;
4171 			if (nat->nat_osrcaddr != src.s_addr ||
4172 			    nat->nat_odstaddr != dst)
4173 				continue;
4174 			break;
4175 		case NAT_OUTBOUND :
4176 			if (nat->nat_v[1] != 4)
4177 				continue;
4178 			if (nat->nat_ndstaddr != src.s_addr ||
4179 			    nat->nat_nsrcaddr != dst)
4180 				continue;
4181 			break;
4182 		}
4183 
4184 		nflags = nat->nat_flags;
4185 		if (!(nflags & (NAT_TCPUDP|SI_WILDP)))
4186 			continue;
4187 
4188 		if (ipf_nat_wildok(nat, (int)sport, (int)dport, nflags,
4189 				   NAT_INBOUND) == 1) {
4190 			if ((fin->fin_flx & FI_IGNORE) != 0)
4191 				break;
4192 			if ((nflags & SI_CLONE) != 0) {
4193 				nat = ipf_nat_clone(fin, nat);
4194 				if (nat == NULL)
4195 					break;
4196 			} else {
4197 				MUTEX_ENTER(&softn->ipf_nat_new);
4198 				softn->ipf_nat_stats.ns_wilds--;
4199 				MUTEX_EXIT(&softn->ipf_nat_new);
4200 			}
4201 
4202 			if (nat->nat_dir == NAT_INBOUND) {
4203 				if (nat->nat_osport == 0) {
4204 					nat->nat_osport = sport;
4205 					nat->nat_nsport = sport;
4206 				}
4207 				if (nat->nat_odport == 0) {
4208 					nat->nat_odport = dport;
4209 					nat->nat_ndport = dport;
4210 				}
4211 			} else if (nat->nat_dir == NAT_OUTBOUND) {
4212 				if (nat->nat_osport == 0) {
4213 					nat->nat_osport = dport;
4214 					nat->nat_nsport = dport;
4215 				}
4216 				if (nat->nat_odport == 0) {
4217 					nat->nat_odport = sport;
4218 					nat->nat_ndport = sport;
4219 				}
4220 			}
4221 			if ((nat->nat_ifps[0] == NULL) && (ifp != NULL)) {
4222 				nat->nat_ifps[0] = ifp;
4223 				nat->nat_mtu[0] = GETIFMTU_4(ifp);
4224 			}
4225 			nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
4226 			ipf_nat_tabmove(softn, nat);
4227 			break;
4228 		}
4229 	}
4230 
4231 	MUTEX_DOWNGRADE(&softc->ipf_nat);
4232 
4233 	if (nat == NULL) {
4234 		NBUMPSIDE(0, ns_lookup_miss);
4235 	}
4236 	return nat;
4237 }
4238 
4239 
4240 /* ------------------------------------------------------------------------ */
4241 /* Function:    ipf_nat_tabmove                                             */
4242 /* Returns:     Nil                                                         */
4243 /* Parameters:  softn(I) - pointer to NAT context structure                 */
4244 /*              nat(I)   - pointer to NAT structure                         */
4245 /* Write Lock:  ipf_nat                                                     */
4246 /*                                                                          */
4247 /* This function is only called for TCP/UDP NAT table entries where the     */
4248 /* original was placed in the table without hashing on the ports and we now */
4249 /* want to include hashing on port numbers.                                 */
4250 /* ------------------------------------------------------------------------ */
4251 static void
4252 ipf_nat_tabmove(ipf_nat_softc_t *softn, nat_t *nat)
4253 {
4254 	u_int hv0, hv1, rhv0, rhv1;
4255 	natstat_t *nsp;
4256 	nat_t **natp;
4257 
4258 	if (nat->nat_flags & SI_CLONE)
4259 		return;
4260 
4261 	nsp = &softn->ipf_nat_stats;
4262 	/*
4263 	 * Remove the NAT entry from the old location
4264 	 */
4265 	if (nat->nat_hnext[0])
4266 		nat->nat_hnext[0]->nat_phnext[0] = nat->nat_phnext[0];
4267 	*nat->nat_phnext[0] = nat->nat_hnext[0];
4268 	nsp->ns_side[0].ns_bucketlen[nat->nat_hv[0] %
4269 				     softn->ipf_nat_table_sz]--;
4270 
4271 	if (nat->nat_hnext[1])
4272 		nat->nat_hnext[1]->nat_phnext[1] = nat->nat_phnext[1];
4273 	*nat->nat_phnext[1] = nat->nat_hnext[1];
4274 	nsp->ns_side[1].ns_bucketlen[nat->nat_hv[1] %
4275 				     softn->ipf_nat_table_sz]--;
4276 
4277 	/*
4278 	 * Add into the NAT table in the new position
4279 	 */
4280 	rhv0 = NAT_HASH_FN(nat->nat_osrcaddr, nat->nat_osport, 0xffffffff);
4281 	rhv0 = NAT_HASH_FN(nat->nat_odstaddr, rhv0 + nat->nat_odport,
4282 			   0xffffffff);
4283 	rhv1 = NAT_HASH_FN(nat->nat_nsrcaddr, nat->nat_nsport, 0xffffffff);
4284 	rhv1 = NAT_HASH_FN(nat->nat_ndstaddr, rhv1 + nat->nat_ndport,
4285 			   0xffffffff);
4286 
4287 	hv0 = rhv0 % softn->ipf_nat_table_sz;
4288 	hv1 = rhv1 % softn->ipf_nat_table_sz;
4289 
4290 	if (nat->nat_dir == NAT_INBOUND || nat->nat_dir == NAT_DIVERTIN) {
4291 		u_int swap;
4292 
4293 		swap = hv0;
4294 		hv0 = hv1;
4295 		hv1 = swap;
4296 	}
4297 
4298 	/* TRACE nat_osrcaddr, nat_osport, nat_odstaddr, nat_odport, hv0 */
4299 	/* TRACE nat_nsrcaddr, nat_nsport, nat_ndstaddr, nat_ndport, hv1 */
4300 
4301 	nat->nat_hv[0] = rhv0;
4302 	natp = &softn->ipf_nat_table[0][hv0];
4303 	if (*natp)
4304 		(*natp)->nat_phnext[0] = &nat->nat_hnext[0];
4305 	nat->nat_phnext[0] = natp;
4306 	nat->nat_hnext[0] = *natp;
4307 	*natp = nat;
4308 	nsp->ns_side[0].ns_bucketlen[hv0]++;
4309 
4310 	nat->nat_hv[1] = rhv1;
4311 	natp = &softn->ipf_nat_table[1][hv1];
4312 	if (*natp)
4313 		(*natp)->nat_phnext[1] = &nat->nat_hnext[1];
4314 	nat->nat_phnext[1] = natp;
4315 	nat->nat_hnext[1] = *natp;
4316 	*natp = nat;
4317 	nsp->ns_side[1].ns_bucketlen[hv1]++;
4318 }
4319 
4320 
4321 /* ------------------------------------------------------------------------ */
4322 /* Function:    ipf_nat_outlookup                                           */
4323 /* Returns:     nat_t* - NULL == no match,                                  */
4324 /*                       else pointer to matching NAT entry                 */
4325 /* Parameters:  fin(I)   - pointer to packet information                    */
4326 /*              flags(I) - NAT flags for this packet                        */
4327 /*              p(I)     - protocol for this packet                         */
4328 /*              src(I)   - source IP address                                */
4329 /*              dst(I)   - destination IP address                           */
4330 /*              rw(I)    - 1 == write lock on  held, 0 == read lock.        */
4331 /*                                                                          */
4332 /* Lookup a nat entry based on the source 'real' ip address/port and        */
4333 /* destination address/port.  We use this lookup when sending a packet out, */
4334 /* we're looking for a table entry, based on the source address.            */
4335 /*                                                                          */
4336 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY.         */
4337 /*                                                                          */
4338 /* NOTE: IT IS ASSUMED THAT  IS ONLY HELD WITH A READ LOCK WHEN             */
4339 /*       THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags.             */
4340 /*                                                                          */
4341 /* flags   -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if   */
4342 /*            the packet is of said protocol                                */
4343 /* ------------------------------------------------------------------------ */
4344 nat_t *
4345 ipf_nat_outlookup(fr_info_t *fin, u_int flags, u_int p, struct in_addr src,
4346    struct in_addr dst)
4347 {
4348 	ipf_main_softc_t *softc = fin->fin_main_soft;
4349 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4350 	u_short sport, dport;
4351 	u_int sflags;
4352 	ipnat_t *ipn;
4353 	nat_t *nat;
4354 	void *ifp;
4355 	u_int hv;
4356 
4357 	ifp = fin->fin_ifp;
4358 	sflags = flags & IPN_TCPUDPICMP;
4359 	sport = 0;
4360 	dport = 0;
4361 
4362 	switch (p)
4363 	{
4364 	case IPPROTO_TCP :
4365 	case IPPROTO_UDP :
4366 		sport = htons(fin->fin_data[0]);
4367 		dport = htons(fin->fin_data[1]);
4368 		break;
4369 	case IPPROTO_ICMP :
4370 		if (flags & IPN_ICMPERR)
4371 			sport = fin->fin_data[1];
4372 		else
4373 			dport = fin->fin_data[1];
4374 		break;
4375 	default :
4376 		break;
4377 	}
4378 
4379 	if ((flags & SI_WILDP) != 0)
4380 		goto find_out_wild_ports;
4381 
4382 	hv = NAT_HASH_FN(src.s_addr, sport, 0xffffffff);
4383 	hv = NAT_HASH_FN(dst.s_addr, hv + dport, softn->ipf_nat_table_sz);
4384 	nat = softn->ipf_nat_table[0][hv];
4385 
4386 	/* TRACE src, sport, dst, dport, hv, nat */
4387 
4388 	for (; nat; nat = nat->nat_hnext[0]) {
4389 		if (nat->nat_ifps[1] != NULL) {
4390 			if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
4391 				continue;
4392 		}
4393 
4394 		if (nat->nat_pr[1] != p)
4395 			continue;
4396 
4397 		switch (nat->nat_dir)
4398 		{
4399 		case NAT_INBOUND :
4400 		case NAT_DIVERTIN :
4401 			if (nat->nat_v[1] != 4)
4402 				continue;
4403 			if (nat->nat_ndstaddr != src.s_addr ||
4404 			    nat->nat_nsrcaddr != dst.s_addr)
4405 				continue;
4406 
4407 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4408 				if (nat->nat_ndport != sport)
4409 					continue;
4410 				if (nat->nat_nsport != dport)
4411 					continue;
4412 
4413 			} else if (p == IPPROTO_ICMP) {
4414 				if (nat->nat_osport != dport) {
4415 					continue;
4416 				}
4417 			}
4418 			break;
4419 		case NAT_OUTBOUND :
4420 		case NAT_DIVERTOUT :
4421 			if (nat->nat_v[0] != 4)
4422 				continue;
4423 			if (nat->nat_osrcaddr != src.s_addr ||
4424 			    nat->nat_odstaddr != dst.s_addr)
4425 				continue;
4426 
4427 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4428 				if (nat->nat_odport != dport)
4429 					continue;
4430 				if (nat->nat_osport != sport)
4431 					continue;
4432 
4433 			} else if (p == IPPROTO_ICMP) {
4434 				if (nat->nat_osport != dport) {
4435 					continue;
4436 				}
4437 			}
4438 			break;
4439 		}
4440 
4441 		ipn = nat->nat_ptr;
4442 		if ((ipn != NULL) && (nat->nat_aps != NULL))
4443 			if (ipf_proxy_match(fin, nat) != 0)
4444 				continue;
4445 
4446 		if ((nat->nat_ifps[1] == NULL) && (ifp != NULL)) {
4447 			nat->nat_ifps[1] = ifp;
4448 			nat->nat_mtu[1] = GETIFMTU_4(ifp);
4449 		}
4450 		return nat;
4451 	}
4452 
4453 	/*
4454 	 * So if we didn't find it but there are wildcard members in the hash
4455 	 * table, go back and look for them.  We do this search and update here
4456 	 * because it is modifying the NAT table and we want to do this only
4457 	 * for the first packet that matches.  The exception, of course, is
4458 	 * for "dummy" (FI_IGNORE) lookups.
4459 	 */
4460 find_out_wild_ports:
4461 	if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH)) {
4462 		NBUMPSIDEX(1, ns_lookup_miss, ns_lookup_miss_1);
4463 		return NULL;
4464 	}
4465 	if (softn->ipf_nat_stats.ns_wilds == 0 || (fin->fin_flx & FI_NOWILD)) {
4466 		NBUMPSIDEX(1, ns_lookup_nowild, ns_lookup_nowild_1);
4467 		return NULL;
4468 	}
4469 
4470 	RWLOCK_EXIT(&softc->ipf_nat);
4471 
4472 	hv = NAT_HASH_FN(src.s_addr, 0, 0xffffffff);
4473 	hv = NAT_HASH_FN(dst.s_addr, hv, softn->ipf_nat_table_sz);
4474 
4475 	WRITE_ENTER(&softc->ipf_nat);
4476 
4477 	nat = softn->ipf_nat_table[0][hv];
4478 	for (; nat; nat = nat->nat_hnext[0]) {
4479 		if (nat->nat_ifps[1] != NULL) {
4480 			if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
4481 				continue;
4482 		}
4483 
4484 		if (nat->nat_pr[1] != fin->fin_p)
4485 			continue;
4486 
4487 		switch (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND))
4488 		{
4489 		case NAT_INBOUND :
4490 			if (nat->nat_v[1] != 4)
4491 				continue;
4492 			if (nat->nat_ndstaddr != src.s_addr ||
4493 			    nat->nat_nsrcaddr != dst.s_addr)
4494 				continue;
4495 			break;
4496 		case NAT_OUTBOUND :
4497 			if (nat->nat_v[0] != 4)
4498 				continue;
4499 			if (nat->nat_osrcaddr != src.s_addr ||
4500 			    nat->nat_odstaddr != dst.s_addr)
4501 				continue;
4502 			break;
4503 		}
4504 
4505 		if (!(nat->nat_flags & (NAT_TCPUDP|SI_WILDP)))
4506 			continue;
4507 
4508 		if (ipf_nat_wildok(nat, (int)sport, (int)dport, nat->nat_flags,
4509 				   NAT_OUTBOUND) == 1) {
4510 			if ((fin->fin_flx & FI_IGNORE) != 0)
4511 				break;
4512 			if ((nat->nat_flags & SI_CLONE) != 0) {
4513 				nat = ipf_nat_clone(fin, nat);
4514 				if (nat == NULL)
4515 					break;
4516 			} else {
4517 				MUTEX_ENTER(&softn->ipf_nat_new);
4518 				softn->ipf_nat_stats.ns_wilds--;
4519 				MUTEX_EXIT(&softn->ipf_nat_new);
4520 			}
4521 
4522 			if (nat->nat_dir == NAT_OUTBOUND) {
4523 				if (nat->nat_osport == 0) {
4524 					nat->nat_osport = sport;
4525 					nat->nat_nsport = sport;
4526 				}
4527 				if (nat->nat_odport == 0) {
4528 					nat->nat_odport = dport;
4529 					nat->nat_ndport = dport;
4530 				}
4531 			} else if (nat->nat_dir == NAT_INBOUND) {
4532 				if (nat->nat_osport == 0) {
4533 					nat->nat_osport = dport;
4534 					nat->nat_nsport = dport;
4535 				}
4536 				if (nat->nat_odport == 0) {
4537 					nat->nat_odport = sport;
4538 					nat->nat_ndport = sport;
4539 				}
4540 			}
4541 			if ((nat->nat_ifps[1] == NULL) && (ifp != NULL)) {
4542 				nat->nat_ifps[1] = ifp;
4543 				nat->nat_mtu[1] = GETIFMTU_4(ifp);
4544 			}
4545 			nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
4546 			ipf_nat_tabmove(softn, nat);
4547 			break;
4548 		}
4549 	}
4550 
4551 	MUTEX_DOWNGRADE(&softc->ipf_nat);
4552 
4553 	if (nat == NULL) {
4554 		NBUMPSIDE(1, ns_lookup_miss);
4555 	}
4556 	return nat;
4557 }
4558 
4559 
4560 /* ------------------------------------------------------------------------ */
4561 /* Function:    ipf_nat_lookupredir                                         */
4562 /* Returns:     nat_t* - NULL == no match,                                  */
4563 /*                       else pointer to matching NAT entry                 */
4564 /* Parameters:  np(I) - pointer to description of packet to find NAT table  */
4565 /*                      entry for.                                          */
4566 /*                                                                          */
4567 /* Lookup the NAT tables to search for a matching redirect                  */
4568 /* The contents of natlookup_t should imitate those found in a packet that  */
4569 /* would be translated - ie a packet coming in for RDR or going out for MAP.*/
4570 /* We can do the lookup in one of two ways, imitating an inbound or         */
4571 /* outbound  packet.  By default we assume outbound, unless IPN_IN is set.  */
4572 /* For IN, the fields are set as follows:                                   */
4573 /*     nl_real* = source information                                        */
4574 /*     nl_out* = destination information (translated)                       */
4575 /* For an out packet, the fields are set like this:                         */
4576 /*     nl_in* = source information (untranslated)                           */
4577 /*     nl_out* = destination information (translated)                       */
4578 /* ------------------------------------------------------------------------ */
4579 nat_t *
4580 ipf_nat_lookupredir(natlookup_t *np)
4581 {
4582 	fr_info_t fi;
4583 	nat_t *nat;
4584 
4585 	bzero((char *)&fi, sizeof(fi));
4586 	if (np->nl_flags & IPN_IN) {
4587 		fi.fin_data[0] = ntohs(np->nl_realport);
4588 		fi.fin_data[1] = ntohs(np->nl_outport);
4589 	} else {
4590 		fi.fin_data[0] = ntohs(np->nl_inport);
4591 		fi.fin_data[1] = ntohs(np->nl_outport);
4592 	}
4593 	if (np->nl_flags & IPN_TCP)
4594 		fi.fin_p = IPPROTO_TCP;
4595 	else if (np->nl_flags & IPN_UDP)
4596 		fi.fin_p = IPPROTO_UDP;
4597 	else if (np->nl_flags & (IPN_ICMPERR|IPN_ICMPQUERY))
4598 		fi.fin_p = IPPROTO_ICMP;
4599 
4600 	/*
4601 	 * We can do two sorts of lookups:
4602 	 * - IPN_IN: we have the `real' and `out' address, look for `in'.
4603 	 * - default: we have the `in' and `out' address, look for `real'.
4604 	 */
4605 	if (np->nl_flags & IPN_IN) {
4606 		if ((nat = ipf_nat_inlookup(&fi, np->nl_flags, fi.fin_p,
4607 					    np->nl_realip, np->nl_outip))) {
4608 			np->nl_inip = nat->nat_odstip;
4609 			np->nl_inport = nat->nat_odport;
4610 		}
4611 	} else {
4612 		/*
4613 		 * If nl_inip is non null, this is a lookup based on the real
4614 		 * ip address. Else, we use the fake.
4615 		 */
4616 		if ((nat = ipf_nat_outlookup(&fi, np->nl_flags, fi.fin_p,
4617 					 np->nl_inip, np->nl_outip))) {
4618 
4619 			if ((np->nl_flags & IPN_FINDFORWARD) != 0) {
4620 				fr_info_t fin;
4621 				bzero((char *)&fin, sizeof(fin));
4622 				fin.fin_p = nat->nat_pr[0];
4623 				fin.fin_data[0] = ntohs(nat->nat_ndport);
4624 				fin.fin_data[1] = ntohs(nat->nat_nsport);
4625 				if (ipf_nat_inlookup(&fin, np->nl_flags,
4626 						     fin.fin_p, nat->nat_ndstip,
4627 						     nat->nat_nsrcip) != NULL) {
4628 					np->nl_flags &= ~IPN_FINDFORWARD;
4629 				}
4630 			}
4631 
4632 			np->nl_realip = nat->nat_ndstip;
4633 			np->nl_realport = nat->nat_ndport;
4634 		}
4635  	}
4636 
4637 	return nat;
4638 }
4639 
4640 
4641 /* ------------------------------------------------------------------------ */
4642 /* Function:    ipf_nat_match                                               */
4643 /* Returns:     int - 0 == no match, 1 == match                             */
4644 /* Parameters:  fin(I)   - pointer to packet information                    */
4645 /*              np(I)    - pointer to NAT rule                              */
4646 /*                                                                          */
4647 /* Pull the matching of a packet against a NAT rule out of that complex     */
4648 /* loop inside ipf_nat_checkin() and lay it out properly in its own function. */
4649 /* ------------------------------------------------------------------------ */
4650 static int
4651 ipf_nat_match(fr_info_t *fin, ipnat_t *np)
4652 {
4653 	ipf_main_softc_t *softc = fin->fin_main_soft;
4654 	frtuc_t *ft;
4655 	int match;
4656 
4657 	match = 0;
4658 	switch (np->in_osrcatype)
4659 	{
4660 	case FRI_NORMAL :
4661 		match = ((fin->fin_saddr & np->in_osrcmsk) != np->in_osrcaddr);
4662 		break;
4663 	case FRI_LOOKUP :
4664 		match = (*np->in_osrcfunc)(softc, np->in_osrcptr,
4665 					   4, &fin->fin_saddr, fin->fin_plen);
4666 		break;
4667 	}
4668 	match ^= ((np->in_flags & IPN_NOTSRC) != 0);
4669 	if (match)
4670 		return 0;
4671 
4672 	match = 0;
4673 	switch (np->in_odstatype)
4674 	{
4675 	case FRI_NORMAL :
4676 		match = ((fin->fin_daddr & np->in_odstmsk) != np->in_odstaddr);
4677 		break;
4678 	case FRI_LOOKUP :
4679 		match = (*np->in_odstfunc)(softc, np->in_odstptr,
4680 					   4, &fin->fin_daddr, fin->fin_plen);
4681 		break;
4682 	}
4683 
4684 	match ^= ((np->in_flags & IPN_NOTDST) != 0);
4685 	if (match)
4686 		return 0;
4687 
4688 	ft = &np->in_tuc;
4689 	if (!(fin->fin_flx & FI_TCPUDP) ||
4690 	    (fin->fin_flx & (FI_SHORT|FI_FRAGBODY))) {
4691 		if (ft->ftu_scmp || ft->ftu_dcmp)
4692 			return 0;
4693 		return 1;
4694 	}
4695 
4696 	return ipf_tcpudpchk(&fin->fin_fi, ft);
4697 }
4698 
4699 
4700 /* ------------------------------------------------------------------------ */
4701 /* Function:    ipf_nat_update                                              */
4702 /* Returns:     Nil                                                         */
4703 /* Parameters:  fin(I) - pointer to packet information                      */
4704 /*              nat(I) - pointer to NAT structure                           */
4705 /*                                                                          */
4706 /* Updates the lifetime of a NAT table entry for non-TCP packets.  Must be  */
4707 /* called with fin_rev updated - i.e. after calling ipf_nat_proto().        */
4708 /*                                                                          */
4709 /* This *MUST* be called after ipf_nat_proto() as it expects fin_rev to     */
4710 /* already be set.                                                          */
4711 /* ------------------------------------------------------------------------ */
4712 void
4713 ipf_nat_update(fr_info_t *fin, nat_t *nat)
4714 {
4715 	ipf_main_softc_t *softc = fin->fin_main_soft;
4716 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4717 	ipftq_t *ifq, *ifq2;
4718 	ipftqent_t *tqe;
4719 	ipnat_t *np = nat->nat_ptr;
4720 
4721 	tqe = &nat->nat_tqe;
4722 	ifq = tqe->tqe_ifq;
4723 
4724 	/*
4725 	 * We allow over-riding of NAT timeouts from NAT rules, even for
4726 	 * TCP, however, if it is TCP and there is no rule timeout set,
4727 	 * then do not update the timeout here.
4728 	 */
4729 	if (np != NULL) {
4730 		np->in_bytes[fin->fin_rev] += fin->fin_plen;
4731 		ifq2 = np->in_tqehead[fin->fin_rev];
4732 	} else {
4733 		ifq2 = NULL;
4734 	}
4735 
4736 	if (nat->nat_pr[0] == IPPROTO_TCP && ifq2 == NULL) {
4737 		(void) ipf_tcp_age(&nat->nat_tqe, fin, softn->ipf_nat_tcptq,
4738 				   0, 2);
4739 	} else {
4740 		if (ifq2 == NULL) {
4741 			if (nat->nat_pr[0] == IPPROTO_UDP)
4742 				ifq2 = fin->fin_rev ? &softn->ipf_nat_udpacktq :
4743 						      &softn->ipf_nat_udptq;
4744 			else if (nat->nat_pr[0] == IPPROTO_ICMP ||
4745 				 nat->nat_pr[0] == IPPROTO_ICMPV6)
4746 				ifq2 = fin->fin_rev ? &softn->ipf_nat_icmpacktq:
4747 						      &softn->ipf_nat_icmptq;
4748 			else
4749 				ifq2 = &softn->ipf_nat_iptq;
4750 		}
4751 
4752 		ipf_movequeue(softc->ipf_ticks, tqe, ifq, ifq2);
4753 	}
4754 }
4755 
4756 
4757 /* ------------------------------------------------------------------------ */
4758 /* Function:    ipf_nat_checkout                                            */
4759 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
4760 /*                     0 == no packet translation occurred,                 */
4761 /*                     1 == packet was successfully translated.             */
4762 /* Parameters:  fin(I)   - pointer to packet information                    */
4763 /*              passp(I) - pointer to filtering result flags                */
4764 /*                                                                          */
4765 /* Check to see if an outcoming packet should be changed.  ICMP packets are */
4766 /* first checked to see if they match an existing entry (if an error),      */
4767 /* otherwise a search of the current NAT table is made.  If neither results */
4768 /* in a match then a search for a matching NAT rule is made.  Create a new  */
4769 /* NAT entry if a we matched a NAT rule.  Lastly, actually change the       */
4770 /* packet header(s) as required.                                            */
4771 /* ------------------------------------------------------------------------ */
4772 int
4773 ipf_nat_checkout(fr_info_t *fin, u_32_t *passp)
4774 {
4775 	ipnat_t *np = NULL, *npnext;
4776 	struct ifnet *ifp, *sifp;
4777 	ipf_main_softc_t *softc;
4778 	ipf_nat_softc_t *softn;
4779 	icmphdr_t *icmp = NULL;
4780 	tcphdr_t *tcp = NULL;
4781 	int rval, natfailed;
4782 	u_int nflags = 0;
4783 	u_32_t ipa, iph;
4784 	int natadd = 1;
4785 	frentry_t *fr;
4786 	nat_t *nat;
4787 
4788 	if (fin->fin_v == 6) {
4789 #ifdef USE_INET6
4790 		return ipf_nat6_checkout(fin, passp);
4791 #else
4792 		return 0;
4793 #endif
4794 	}
4795 
4796 	softc = fin->fin_main_soft;
4797 	softn = softc->ipf_nat_soft;
4798 
4799 	if (softn->ipf_nat_lock != 0)
4800 		return 0;
4801 	if (softn->ipf_nat_stats.ns_rules == 0 &&
4802 	    softn->ipf_nat_instances == NULL)
4803 		return 0;
4804 
4805 	natfailed = 0;
4806 	fr = fin->fin_fr;
4807 	sifp = fin->fin_ifp;
4808 	if (fr != NULL) {
4809 		ifp = fr->fr_tifs[fin->fin_rev].fd_ptr;
4810 		if ((ifp != NULL) && (ifp != (void *)-1))
4811 			fin->fin_ifp = ifp;
4812 	}
4813 	ifp = fin->fin_ifp;
4814 
4815 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
4816 		switch (fin->fin_p)
4817 		{
4818 		case IPPROTO_TCP :
4819 			nflags = IPN_TCP;
4820 			break;
4821 		case IPPROTO_UDP :
4822 			nflags = IPN_UDP;
4823 			break;
4824 		case IPPROTO_ICMP :
4825 			icmp = fin->fin_dp;
4826 
4827 			/*
4828 			 * This is an incoming packet, so the destination is
4829 			 * the icmp_id and the source port equals 0
4830 			 */
4831 			if ((fin->fin_flx & FI_ICMPQUERY) != 0)
4832 				nflags = IPN_ICMPQUERY;
4833 			break;
4834 		default :
4835 			break;
4836 		}
4837 
4838 		if ((nflags & IPN_TCPUDP))
4839 			tcp = fin->fin_dp;
4840 	}
4841 
4842 	ipa = fin->fin_saddr;
4843 
4844 	READ_ENTER(&softc->ipf_nat);
4845 
4846 	if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
4847 	    (nat = ipf_nat_icmperror(fin, &nflags, NAT_OUTBOUND)))
4848 		/*EMPTY*/;
4849 	else if ((fin->fin_flx & FI_FRAG) && (nat = ipf_frag_natknown(fin)))
4850 		natadd = 0;
4851 	else if ((nat = ipf_nat_outlookup(fin, nflags|NAT_SEARCH,
4852 				      (u_int)fin->fin_p, fin->fin_src,
4853 				      fin->fin_dst))) {
4854 		nflags = nat->nat_flags;
4855 	} else if (fin->fin_off == 0) {
4856 		u_32_t hv, msk, nmsk = 0;
4857 
4858 		/*
4859 		 * If there is no current entry in the nat table for this IP#,
4860 		 * create one for it (if there is a matching rule).
4861 		 */
4862 maskloop:
4863 		msk = softn->ipf_nat_map_active_masks[nmsk];
4864 		iph = ipa & msk;
4865 		hv = NAT_HASH_FN(iph, 0, softn->ipf_nat_maprules_sz);
4866 retry_roundrobin:
4867 		for (np = softn->ipf_nat_map_rules[hv]; np; np = npnext) {
4868 			npnext = np->in_mnext;
4869 			if ((np->in_ifps[1] && (np->in_ifps[1] != ifp)))
4870 				continue;
4871 			if (np->in_v[0] != 4)
4872 				continue;
4873 			if (np->in_pr[1] && (np->in_pr[1] != fin->fin_p))
4874 				continue;
4875 			if ((np->in_flags & IPN_RF) &&
4876 			    !(np->in_flags & nflags))
4877 				continue;
4878 			if (np->in_flags & IPN_FILTER) {
4879 				switch (ipf_nat_match(fin, np))
4880 				{
4881 				case 0 :
4882 					continue;
4883 				case -1 :
4884 					rval = -1;
4885 					goto outmatchfail;
4886 				case 1 :
4887 				default :
4888 					break;
4889 				}
4890 			} else if ((ipa & np->in_osrcmsk) != np->in_osrcaddr)
4891 				continue;
4892 
4893 			if ((fr != NULL) &&
4894 			    !ipf_matchtag(&np->in_tag, &fr->fr_nattag))
4895 				continue;
4896 
4897 			if (np->in_plabel != -1) {
4898 				if (((np->in_flags & IPN_FILTER) == 0) &&
4899 				    (np->in_odport != fin->fin_data[1]))
4900 					continue;
4901 				if (ipf_proxy_ok(fin, tcp, np) == 0)
4902 					continue;
4903 			}
4904 
4905 			if (np->in_flags & IPN_NO) {
4906 				np->in_hits++;
4907 				break;
4908 			}
4909 			MUTEX_ENTER(&softn->ipf_nat_new);
4910 			/*
4911 			 * If we've matched a round-robin rule but it has
4912 			 * moved in the list since we got it, start over as
4913 			 * this is now no longer correct.
4914 			 */
4915 			if (npnext != np->in_mnext) {
4916 				if ((np->in_flags & IPN_ROUNDR) != 0) {
4917 					MUTEX_EXIT(&softn->ipf_nat_new);
4918 					goto retry_roundrobin;
4919 				}
4920 				npnext = np->in_mnext;
4921 			}
4922 
4923 			nat = ipf_nat_add(fin, np, NULL, nflags, NAT_OUTBOUND);
4924 			MUTEX_EXIT(&softn->ipf_nat_new);
4925 			if (nat != NULL) {
4926 				natfailed = 0;
4927 				break;
4928 			}
4929 			natfailed = -1;
4930 		}
4931 		if ((np == NULL) && (nmsk < softn->ipf_nat_map_max)) {
4932 			nmsk++;
4933 			goto maskloop;
4934 		}
4935 	}
4936 
4937 	if (nat != NULL) {
4938 		rval = ipf_nat_out(fin, nat, natadd, nflags);
4939 		if (rval == 1) {
4940 			MUTEX_ENTER(&nat->nat_lock);
4941 			ipf_nat_update(fin, nat);
4942 			nat->nat_bytes[1] += fin->fin_plen;
4943 			nat->nat_pkts[1]++;
4944 			fin->fin_pktnum = nat->nat_pkts[1];
4945 			MUTEX_EXIT(&nat->nat_lock);
4946 		}
4947 	} else
4948 		rval = natfailed;
4949 outmatchfail:
4950 	RWLOCK_EXIT(&softc->ipf_nat);
4951 
4952 	switch (rval)
4953 	{
4954 	case -1 :
4955 		if (passp != NULL) {
4956 			DT1(frb_natv4out, fr_info_t *, fin);
4957 			NBUMPSIDED(1, ns_drop);
4958 			*passp = FR_BLOCK;
4959 			fin->fin_reason = FRB_NATV4;
4960 		}
4961 		fin->fin_flx |= FI_BADNAT;
4962 		NBUMPSIDED(1, ns_badnat);
4963 		break;
4964 	case 0 :
4965 		NBUMPSIDE(1, ns_ignored);
4966 		break;
4967 	case 1 :
4968 		NBUMPSIDE(1, ns_translated);
4969 		break;
4970 	}
4971 	fin->fin_ifp = sifp;
4972 	return rval;
4973 }
4974 
4975 /* ------------------------------------------------------------------------ */
4976 /* Function:    ipf_nat_out                                                 */
4977 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
4978 /*                     1 == packet was successfully translated.             */
4979 /* Parameters:  fin(I)    - pointer to packet information                   */
4980 /*              nat(I)    - pointer to NAT structure                        */
4981 /*              natadd(I) - flag indicating if it is safe to add frag cache */
4982 /*              nflags(I) - NAT flags set for this packet                   */
4983 /*                                                                          */
4984 /* Translate a packet coming "out" on an interface.                         */
4985 /* ------------------------------------------------------------------------ */
4986 int
4987 ipf_nat_out(fr_info_t *fin, nat_t *nat, int natadd, u_32_t nflags)
4988 {
4989 	ipf_main_softc_t *softc = fin->fin_main_soft;
4990 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4991 	icmphdr_t *icmp;
4992 	tcphdr_t *tcp;
4993 	ipnat_t *np;
4994 	int skip;
4995 	int i;
4996 
4997 	tcp = NULL;
4998 	icmp = NULL;
4999 	np = nat->nat_ptr;
5000 
5001 	if ((natadd != 0) && (fin->fin_flx & FI_FRAG) && (np != NULL))
5002 		(void) ipf_frag_natnew(softc, fin, 0, nat);
5003 
5004 	/*
5005 	 * Fix up checksums, not by recalculating them, but
5006 	 * simply computing adjustments.
5007 	 * This is only done for STREAMS based IP implementations where the
5008 	 * checksum has already been calculated by IP.  In all other cases,
5009 	 * IPFilter is called before the checksum needs calculating so there
5010 	 * is no call to modify whatever is in the header now.
5011 	 */
5012 	if (nflags == IPN_ICMPERR) {
5013 		u_32_t s1, s2, sumd, msumd;
5014 
5015 		s1 = LONG_SUM(ntohl(fin->fin_saddr));
5016 		if (nat->nat_dir == NAT_OUTBOUND) {
5017 			s2 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
5018 		} else {
5019 			s2 = LONG_SUM(ntohl(nat->nat_odstaddr));
5020 		}
5021 		CALC_SUMD(s1, s2, sumd);
5022 		msumd = sumd;
5023 
5024 		s1 = LONG_SUM(ntohl(fin->fin_daddr));
5025 		if (nat->nat_dir == NAT_OUTBOUND) {
5026 			s2 = LONG_SUM(ntohl(nat->nat_ndstaddr));
5027 		} else {
5028 			s2 = LONG_SUM(ntohl(nat->nat_osrcaddr));
5029 		}
5030 		CALC_SUMD(s1, s2, sumd);
5031 		msumd += sumd;
5032 
5033 		ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, msumd, 0);
5034 	}
5035 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5036     defined(linux) || defined(BRIDGE_IPF)
5037 	else {
5038 		/*
5039 		 * Strictly speaking, this isn't necessary on BSD
5040 		 * kernels because they do checksum calculation after
5041 		 * this code has run BUT if ipfilter is being used
5042 		 * to do NAT as a bridge, that code doesn't exist.
5043 		 */
5044 		switch (nat->nat_dir)
5045 		{
5046 		case NAT_OUTBOUND :
5047 			ipf_fix_outcksum(fin->fin_cksum & FI_CK_L4PART,
5048 					 &fin->fin_ip->ip_sum,
5049 					 nat->nat_ipsumd, 0);
5050 			break;
5051 
5052 		case NAT_INBOUND :
5053 			ipf_fix_incksum(fin->fin_cksum & FI_CK_L4PART,
5054 					&fin->fin_ip->ip_sum,
5055 					nat->nat_ipsumd, 0);
5056 			break;
5057 
5058 		default :
5059 			break;
5060 		}
5061 	}
5062 #endif
5063 
5064 	/*
5065 	 * Address assignment is after the checksum modification because
5066 	 * we are using the address in the packet for determining the
5067 	 * correct checksum offset (the ICMP error could be coming from
5068 	 * anyone...)
5069 	 */
5070 	switch (nat->nat_dir)
5071 	{
5072 	case NAT_OUTBOUND :
5073 		fin->fin_ip->ip_src = nat->nat_nsrcip;
5074 		fin->fin_saddr = nat->nat_nsrcaddr;
5075 		fin->fin_ip->ip_dst = nat->nat_ndstip;
5076 		fin->fin_daddr = nat->nat_ndstaddr;
5077 		break;
5078 
5079 	case NAT_INBOUND :
5080 		fin->fin_ip->ip_src = nat->nat_odstip;
5081 		fin->fin_saddr = nat->nat_ndstaddr;
5082 		fin->fin_ip->ip_dst = nat->nat_osrcip;
5083 		fin->fin_daddr = nat->nat_nsrcaddr;
5084 		break;
5085 
5086 	case NAT_DIVERTIN :
5087 	    {
5088 		mb_t *m;
5089 
5090 		skip = ipf_nat_decap(fin, nat);
5091 		if (skip <= 0) {
5092 			NBUMPSIDED(1, ns_decap_fail);
5093 			return -1;
5094 		}
5095 
5096 		m = fin->fin_m;
5097 
5098 #if defined(MENTAT) && defined(_KERNEL)
5099 		m->b_rptr += skip;
5100 #else
5101 		m->m_data += skip;
5102 		m->m_len -= skip;
5103 
5104 # ifdef M_PKTHDR
5105 		if (m->m_flags & M_PKTHDR)
5106 			m->m_pkthdr.len -= skip;
5107 # endif
5108 #endif
5109 
5110 		MUTEX_ENTER(&nat->nat_lock);
5111 		ipf_nat_update(fin, nat);
5112 		MUTEX_EXIT(&nat->nat_lock);
5113 		fin->fin_flx |= FI_NATED;
5114 		if (np != NULL && np->in_tag.ipt_num[0] != 0)
5115 			fin->fin_nattag = &np->in_tag;
5116 		return 1;
5117 		/* NOTREACHED */
5118 	    }
5119 
5120 	case NAT_DIVERTOUT :
5121 	    {
5122 		u_32_t s1, s2, sumd;
5123 		udphdr_t *uh;
5124 		ip_t *ip;
5125 		mb_t *m;
5126 
5127 		m = M_DUP(np->in_divmp);
5128 		if (m == NULL) {
5129 			NBUMPSIDED(1, ns_divert_dup);
5130 			return -1;
5131 		}
5132 
5133 		ip = MTOD(m, ip_t *);
5134 		ip->ip_id = htons(ipf_nextipid(fin));
5135 		s2 = ntohs(ip->ip_id);
5136 
5137 		s1 = ip->ip_len;
5138 		ip->ip_len = ntohs(ip->ip_len);
5139 		ip->ip_len += fin->fin_plen;
5140 		ip->ip_len = htons(ip->ip_len);
5141 		s2 += ntohs(ip->ip_len);
5142 		CALC_SUMD(s1, s2, sumd);
5143 
5144 		uh = (udphdr_t *)(ip + 1);
5145 		uh->uh_ulen += fin->fin_plen;
5146 		uh->uh_ulen = htons(uh->uh_ulen);
5147 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5148     defined(linux) || defined(BRIDGE_IPF)
5149 		ipf_fix_outcksum(0, &ip->ip_sum, sumd, 0);
5150 #endif
5151 
5152 		PREP_MB_T(fin, m);
5153 
5154 		fin->fin_src = ip->ip_src;
5155 		fin->fin_dst = ip->ip_dst;
5156 		fin->fin_ip = ip;
5157 		fin->fin_plen += sizeof(ip_t) + 8;	/* UDP + IPv4 hdr */
5158 		fin->fin_dlen += sizeof(ip_t) + 8;	/* UDP + IPv4 hdr */
5159 
5160 		nflags &= ~IPN_TCPUDPICMP;
5161 
5162 		break;
5163 	    }
5164 
5165 	default :
5166 		break;
5167 	}
5168 
5169 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5170 		u_short *csump;
5171 
5172 		if ((nat->nat_nsport != 0) && (nflags & IPN_TCPUDP)) {
5173 			tcp = fin->fin_dp;
5174 
5175 			switch (nat->nat_dir)
5176 			{
5177 			case NAT_OUTBOUND :
5178 				tcp->th_sport = nat->nat_nsport;
5179 				fin->fin_data[0] = ntohs(nat->nat_nsport);
5180 				tcp->th_dport = nat->nat_ndport;
5181 				fin->fin_data[1] = ntohs(nat->nat_ndport);
5182 				break;
5183 
5184 			case NAT_INBOUND :
5185 				tcp->th_sport = nat->nat_odport;
5186 				fin->fin_data[0] = ntohs(nat->nat_odport);
5187 				tcp->th_dport = nat->nat_osport;
5188 				fin->fin_data[1] = ntohs(nat->nat_osport);
5189 				break;
5190 			}
5191 		}
5192 
5193 		if ((nat->nat_nsport != 0) && (nflags & IPN_ICMPQUERY)) {
5194 			icmp = fin->fin_dp;
5195 			icmp->icmp_id = nat->nat_nicmpid;
5196 		}
5197 
5198 		csump = ipf_nat_proto(fin, nat, nflags);
5199 
5200 		/*
5201 		 * The above comments do not hold for layer 4 (or higher)
5202 		 * checksums...
5203 		 */
5204 		if (csump != NULL) {
5205 			if (nat->nat_dir == NAT_OUTBOUND)
5206 				ipf_fix_outcksum(fin->fin_cksum, csump,
5207 						 nat->nat_sumd[0],
5208 						 nat->nat_sumd[1] +
5209 						 fin->fin_dlen);
5210 			else
5211 				ipf_fix_incksum(fin->fin_cksum, csump,
5212 						nat->nat_sumd[0],
5213 						nat->nat_sumd[1] +
5214 						fin->fin_dlen);
5215 		}
5216 	}
5217 
5218 	ipf_sync_update(softc, SMC_NAT, fin, nat->nat_sync);
5219 	/* ------------------------------------------------------------- */
5220 	/* A few quick notes:                                            */
5221 	/*      Following are test conditions prior to calling the       */
5222 	/*      ipf_proxy_check routine.                                 */
5223 	/*                                                               */
5224 	/*      A NULL tcp indicates a non TCP/UDP packet.  When dealing */
5225 	/*      with a redirect rule, we attempt to match the packet's   */
5226 	/*      source port against in_dport, otherwise we'd compare the */
5227 	/*      packet's destination.                                    */
5228 	/* ------------------------------------------------------------- */
5229 	if ((np != NULL) && (np->in_apr != NULL)) {
5230 		i = ipf_proxy_check(fin, nat);
5231 		if (i == 0) {
5232 			i = 1;
5233 		} else if (i == -1) {
5234 			NBUMPSIDED(1, ns_ipf_proxy_fail);
5235 		}
5236 	} else {
5237 		i = 1;
5238 	}
5239 	fin->fin_flx |= FI_NATED;
5240 	return i;
5241 }
5242 
5243 
5244 /* ------------------------------------------------------------------------ */
5245 /* Function:    ipf_nat_checkin                                             */
5246 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
5247 /*                     0 == no packet translation occurred,                 */
5248 /*                     1 == packet was successfully translated.             */
5249 /* Parameters:  fin(I)   - pointer to packet information                    */
5250 /*              passp(I) - pointer to filtering result flags                */
5251 /*                                                                          */
5252 /* Check to see if an incoming packet should be changed.  ICMP packets are  */
5253 /* first checked to see if they match an existing entry (if an error),      */
5254 /* otherwise a search of the current NAT table is made.  If neither results */
5255 /* in a match then a search for a matching NAT rule is made.  Create a new  */
5256 /* NAT entry if a we matched a NAT rule.  Lastly, actually change the       */
5257 /* packet header(s) as required.                                            */
5258 /* ------------------------------------------------------------------------ */
5259 int
5260 ipf_nat_checkin(fr_info_t *fin, u_32_t *passp)
5261 {
5262 	ipf_main_softc_t *softc;
5263 	ipf_nat_softc_t *softn;
5264 	u_int nflags, natadd;
5265 	ipnat_t *np, *npnext;
5266 	int rval, natfailed;
5267 	struct ifnet *ifp;
5268 	struct in_addr in;
5269 	icmphdr_t *icmp;
5270 	tcphdr_t *tcp;
5271 	u_short dport;
5272 	nat_t *nat;
5273 	u_32_t iph;
5274 
5275 	softc = fin->fin_main_soft;
5276 	softn = softc->ipf_nat_soft;
5277 
5278 	if (softn->ipf_nat_lock != 0)
5279 		return 0;
5280 	if (softn->ipf_nat_stats.ns_rules == 0 &&
5281 	    softn->ipf_nat_instances == NULL)
5282 		return 0;
5283 
5284 	tcp = NULL;
5285 	icmp = NULL;
5286 	dport = 0;
5287 	natadd = 1;
5288 	nflags = 0;
5289 	natfailed = 0;
5290 	ifp = fin->fin_ifp;
5291 
5292 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5293 		switch (fin->fin_p)
5294 		{
5295 		case IPPROTO_TCP :
5296 			nflags = IPN_TCP;
5297 			break;
5298 		case IPPROTO_UDP :
5299 			nflags = IPN_UDP;
5300 			break;
5301 		case IPPROTO_ICMP :
5302 			icmp = fin->fin_dp;
5303 
5304 			/*
5305 			 * This is an incoming packet, so the destination is
5306 			 * the icmp_id and the source port equals 0
5307 			 */
5308 			if ((fin->fin_flx & FI_ICMPQUERY) != 0) {
5309 				nflags = IPN_ICMPQUERY;
5310 				dport = icmp->icmp_id;
5311 			} break;
5312 		default :
5313 			break;
5314 		}
5315 
5316 		if ((nflags & IPN_TCPUDP)) {
5317 			tcp = fin->fin_dp;
5318 			dport = fin->fin_data[1];
5319 		}
5320 	}
5321 
5322 	in = fin->fin_dst;
5323 
5324 	READ_ENTER(&softc->ipf_nat);
5325 
5326 	if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
5327 	    (nat = ipf_nat_icmperror(fin, &nflags, NAT_INBOUND)))
5328 		/*EMPTY*/;
5329 	else if ((fin->fin_flx & FI_FRAG) && (nat = ipf_frag_natknown(fin)))
5330 		natadd = 0;
5331 	else if ((nat = ipf_nat_inlookup(fin, nflags|NAT_SEARCH,
5332 					 (u_int)fin->fin_p,
5333 					 fin->fin_src, in))) {
5334 		nflags = nat->nat_flags;
5335 	} else if (fin->fin_off == 0) {
5336 		u_32_t hv, msk, rmsk = 0;
5337 
5338 		/*
5339 		 * If there is no current entry in the nat table for this IP#,
5340 		 * create one for it (if there is a matching rule).
5341 		 */
5342 maskloop:
5343 		msk = softn->ipf_nat_rdr_active_masks[rmsk];
5344 		iph = in.s_addr & msk;
5345 		hv = NAT_HASH_FN(iph, 0, softn->ipf_nat_rdrrules_sz);
5346 retry_roundrobin:
5347 		/* TRACE (iph,msk,rmsk,hv,softn->ipf_nat_rdrrules_sz) */
5348 		for (np = softn->ipf_nat_rdr_rules[hv]; np; np = npnext) {
5349 			npnext = np->in_rnext;
5350 			if (np->in_ifps[0] && (np->in_ifps[0] != ifp))
5351 				continue;
5352 			if (np->in_v[0] != 4)
5353 				continue;
5354 			if (np->in_pr[0] && (np->in_pr[0] != fin->fin_p))
5355 				continue;
5356 			if ((np->in_flags & IPN_RF) && !(np->in_flags & nflags))
5357 				continue;
5358 			if (np->in_flags & IPN_FILTER) {
5359 				switch (ipf_nat_match(fin, np))
5360 				{
5361 				case 0 :
5362 					continue;
5363 				case -1 :
5364 					rval = -1;
5365 					goto inmatchfail;
5366 				case 1 :
5367 				default :
5368 					break;
5369 				}
5370 			} else {
5371 				if ((in.s_addr & np->in_odstmsk) !=
5372 				    np->in_odstaddr)
5373 					continue;
5374 				if (np->in_odport &&
5375 				    ((np->in_dtop < dport) ||
5376 				     (dport < np->in_odport)))
5377 					continue;
5378 			}
5379 
5380 			if (np->in_plabel != -1) {
5381 				if (!ipf_proxy_ok(fin, tcp, np)) {
5382 					continue;
5383 				}
5384 			}
5385 
5386 			if (np->in_flags & IPN_NO) {
5387 				np->in_hits++;
5388 				break;
5389 			}
5390 
5391 			MUTEX_ENTER(&softn->ipf_nat_new);
5392 			/*
5393 			 * If we've matched a round-robin rule but it has
5394 			 * moved in the list since we got it, start over as
5395 			 * this is now no longer correct.
5396 			 */
5397 			if (npnext != np->in_rnext) {
5398 				if ((np->in_flags & IPN_ROUNDR) != 0) {
5399 					MUTEX_EXIT(&softn->ipf_nat_new);
5400 					goto retry_roundrobin;
5401 				}
5402 				npnext = np->in_rnext;
5403 			}
5404 
5405 			nat = ipf_nat_add(fin, np, NULL, nflags, NAT_INBOUND);
5406 			MUTEX_EXIT(&softn->ipf_nat_new);
5407 			if (nat != NULL) {
5408 				natfailed = 0;
5409 				break;
5410 			}
5411 			natfailed = -1;
5412 		}
5413 		if ((np == NULL) && (rmsk < softn->ipf_nat_rdr_max)) {
5414 			rmsk++;
5415 			goto maskloop;
5416 		}
5417 	}
5418 
5419 	if (nat != NULL) {
5420 		rval = ipf_nat_in(fin, nat, natadd, nflags);
5421 		if (rval == 1) {
5422 			MUTEX_ENTER(&nat->nat_lock);
5423 			ipf_nat_update(fin, nat);
5424 			nat->nat_bytes[0] += fin->fin_plen;
5425 			nat->nat_pkts[0]++;
5426 			fin->fin_pktnum = nat->nat_pkts[0];
5427 			MUTEX_EXIT(&nat->nat_lock);
5428 		}
5429 	} else
5430 		rval = natfailed;
5431 inmatchfail:
5432 	RWLOCK_EXIT(&softc->ipf_nat);
5433 
5434 	switch (rval)
5435 	{
5436 	case -1 :
5437 		if (passp != NULL) {
5438 			DT1(frb_natv4in, fr_info_t *, fin);
5439 			NBUMPSIDED(0, ns_drop);
5440 			*passp = FR_BLOCK;
5441 			fin->fin_reason = FRB_NATV4;
5442 		}
5443 		fin->fin_flx |= FI_BADNAT;
5444 		NBUMPSIDED(0, ns_badnat);
5445 		break;
5446 	case 0 :
5447 		NBUMPSIDE(0, ns_ignored);
5448 		break;
5449 	case 1 :
5450 		NBUMPSIDE(0, ns_translated);
5451 		break;
5452 	}
5453 	return rval;
5454 }
5455 
5456 
5457 /* ------------------------------------------------------------------------ */
5458 /* Function:    ipf_nat_in                                                  */
5459 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
5460 /*                     1 == packet was successfully translated.             */
5461 /* Parameters:  fin(I)    - pointer to packet information                   */
5462 /*              nat(I)    - pointer to NAT structure                        */
5463 /*              natadd(I) - flag indicating if it is safe to add frag cache */
5464 /*              nflags(I) - NAT flags set for this packet                   */
5465 /* Locks Held:  ipf_nat(READ)                                               */
5466 /*                                                                          */
5467 /* Translate a packet coming "in" on an interface.                          */
5468 /* ------------------------------------------------------------------------ */
5469 int
5470 ipf_nat_in(fr_info_t *fin, nat_t *nat, int natadd, u_32_t nflags)
5471 {
5472 	ipf_main_softc_t *softc = fin->fin_main_soft;
5473 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5474 	u_32_t sumd, ipsumd, sum1, sum2;
5475 	icmphdr_t *icmp;
5476 	tcphdr_t *tcp;
5477 	ipnat_t *np;
5478 	int skip;
5479 	int i;
5480 
5481 	tcp = NULL;
5482 	np = nat->nat_ptr;
5483 	fin->fin_fr = nat->nat_fr;
5484 
5485 	if (np != NULL) {
5486 		if ((natadd != 0) && (fin->fin_flx & FI_FRAG))
5487 			(void) ipf_frag_natnew(softc, fin, 0, nat);
5488 
5489 	/* ------------------------------------------------------------- */
5490 	/* A few quick notes:                                            */
5491 	/*      Following are test conditions prior to calling the       */
5492 	/*      ipf_proxy_check routine.                                 */
5493 	/*                                                               */
5494 	/*      A NULL tcp indicates a non TCP/UDP packet.  When dealing */
5495 	/*      with a map rule, we attempt to match the packet's        */
5496 	/*      source port against in_dport, otherwise we'd compare the */
5497 	/*      packet's destination.                                    */
5498 	/* ------------------------------------------------------------- */
5499 		if (np->in_apr != NULL) {
5500 			i = ipf_proxy_check(fin, nat);
5501 			if (i == -1) {
5502 				NBUMPSIDED(0, ns_ipf_proxy_fail);
5503 				return -1;
5504 			}
5505 		}
5506 	}
5507 
5508 	ipf_sync_update(softc, SMC_NAT, fin, nat->nat_sync);
5509 
5510 	ipsumd = nat->nat_ipsumd;
5511 	/*
5512 	 * Fix up checksums, not by recalculating them, but
5513 	 * simply computing adjustments.
5514 	 * Why only do this for some platforms on inbound packets ?
5515 	 * Because for those that it is done, IP processing is yet to happen
5516 	 * and so the IPv4 header checksum has not yet been evaluated.
5517 	 * Perhaps it should always be done for the benefit of things like
5518 	 * fast forwarding (so that it doesn't need to be recomputed) but with
5519 	 * header checksum offloading, perhaps it is a moot point.
5520 	 */
5521 
5522 	switch (nat->nat_dir)
5523 	{
5524 	case NAT_INBOUND :
5525 		if ((fin->fin_flx & FI_ICMPERR) == 0) {
5526 			fin->fin_ip->ip_src = nat->nat_nsrcip;
5527 			fin->fin_saddr = nat->nat_nsrcaddr;
5528 		} else {
5529 			sum1 = nat->nat_osrcaddr;
5530 			sum2 = nat->nat_nsrcaddr;
5531 			CALC_SUMD(sum1, sum2, sumd);
5532 			ipsumd -= sumd;
5533 		}
5534 		fin->fin_ip->ip_dst = nat->nat_ndstip;
5535 		fin->fin_daddr = nat->nat_ndstaddr;
5536 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5537      defined(__osf__) || defined(linux)
5538 		ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, ipsumd, 0);
5539 #endif
5540 		break;
5541 
5542 	case NAT_OUTBOUND :
5543 		if ((fin->fin_flx & FI_ICMPERR) == 0) {
5544 			fin->fin_ip->ip_src = nat->nat_odstip;
5545 			fin->fin_saddr = nat->nat_odstaddr;
5546 		} else {
5547 			sum1 = nat->nat_odstaddr;
5548 			sum2 = nat->nat_ndstaddr;
5549 			CALC_SUMD(sum1, sum2, sumd);
5550 			ipsumd -= sumd;
5551 		}
5552 		fin->fin_ip->ip_dst = nat->nat_osrcip;
5553 		fin->fin_daddr = nat->nat_osrcaddr;
5554 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5555      defined(__osf__) || defined(linux)
5556 		ipf_fix_incksum(0, &fin->fin_ip->ip_sum, ipsumd, 0);
5557 #endif
5558 		break;
5559 
5560 	case NAT_DIVERTIN :
5561 	    {
5562 		udphdr_t *uh;
5563 		ip_t *ip;
5564 		mb_t *m;
5565 
5566 		m = M_DUP(np->in_divmp);
5567 		if (m == NULL) {
5568 			NBUMPSIDED(0, ns_divert_dup);
5569 			return -1;
5570 		}
5571 
5572 		ip = MTOD(m, ip_t *);
5573 		ip->ip_id = htons(ipf_nextipid(fin));
5574 		sum1 = ntohs(ip->ip_len);
5575 		ip->ip_len = ntohs(ip->ip_len);
5576 		ip->ip_len += fin->fin_plen;
5577 		ip->ip_len = htons(ip->ip_len);
5578 
5579 		uh = (udphdr_t *)(ip + 1);
5580 		uh->uh_ulen += fin->fin_plen;
5581 		uh->uh_ulen = htons(uh->uh_ulen);
5582 
5583 		sum2 = ntohs(ip->ip_id) + ntohs(ip->ip_len);
5584 		sum2 += ntohs(ip->ip_off) & IP_DF;
5585 		CALC_SUMD(sum1, sum2, sumd);
5586 
5587 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5588      defined(__osf__) || defined(linux)
5589 		ipf_fix_outcksum(0, &ip->ip_sum, sumd, 0);
5590 #endif
5591 		PREP_MB_T(fin, m);
5592 
5593 		fin->fin_ip = ip;
5594 		fin->fin_plen += sizeof(ip_t) + 8;	/* UDP + new IPv4 hdr */
5595 		fin->fin_dlen += sizeof(ip_t) + 8;	/* UDP + old IPv4 hdr */
5596 
5597 		nflags &= ~IPN_TCPUDPICMP;
5598 
5599 		break;
5600 	    }
5601 
5602 	case NAT_DIVERTOUT :
5603 	    {
5604 		mb_t *m;
5605 
5606 		skip = ipf_nat_decap(fin, nat);
5607 		if (skip <= 0) {
5608 			NBUMPSIDED(0, ns_decap_fail);
5609 			return -1;
5610 		}
5611 
5612 		m = fin->fin_m;
5613 
5614 #if defined(MENTAT) && defined(_KERNEL)
5615 		m->b_rptr += skip;
5616 #else
5617 		m->m_data += skip;
5618 		m->m_len -= skip;
5619 
5620 # ifdef M_PKTHDR
5621 		if (m->m_flags & M_PKTHDR)
5622 			m->m_pkthdr.len -= skip;
5623 # endif
5624 #endif
5625 
5626 		ipf_nat_update(fin, nat);
5627 		nflags &= ~IPN_TCPUDPICMP;
5628 		fin->fin_flx |= FI_NATED;
5629 		if (np != NULL && np->in_tag.ipt_num[0] != 0)
5630 			fin->fin_nattag = &np->in_tag;
5631 		return 1;
5632 		/* NOTREACHED */
5633 	    }
5634 	}
5635 	if (nflags & IPN_TCPUDP)
5636 		tcp = fin->fin_dp;
5637 
5638 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5639 		u_short *csump;
5640 
5641 		if ((nat->nat_odport != 0) && (nflags & IPN_TCPUDP)) {
5642 			switch (nat->nat_dir)
5643 			{
5644 			case NAT_INBOUND :
5645 				tcp->th_sport = nat->nat_nsport;
5646 				fin->fin_data[0] = ntohs(nat->nat_nsport);
5647 				tcp->th_dport = nat->nat_ndport;
5648 				fin->fin_data[1] = ntohs(nat->nat_ndport);
5649 				break;
5650 
5651 			case NAT_OUTBOUND :
5652 				tcp->th_sport = nat->nat_odport;
5653 				fin->fin_data[0] = ntohs(nat->nat_odport);
5654 				tcp->th_dport = nat->nat_osport;
5655 				fin->fin_data[1] = ntohs(nat->nat_osport);
5656 				break;
5657 			}
5658 		}
5659 
5660 
5661 		if ((nat->nat_odport != 0) && (nflags & IPN_ICMPQUERY)) {
5662 			icmp = fin->fin_dp;
5663 
5664 			icmp->icmp_id = nat->nat_nicmpid;
5665 		}
5666 
5667 		csump = ipf_nat_proto(fin, nat, nflags);
5668 
5669 		/*
5670 		 * The above comments do not hold for layer 4 (or higher)
5671 		 * checksums...
5672 		 */
5673 		if (csump != NULL) {
5674 			if (nat->nat_dir == NAT_OUTBOUND)
5675 				ipf_fix_incksum(0, csump, nat->nat_sumd[0], 0);
5676 			else
5677 				ipf_fix_outcksum(0, csump, nat->nat_sumd[0], 0);
5678 		}
5679 	}
5680 
5681 	fin->fin_flx |= FI_NATED;
5682 	if (np != NULL && np->in_tag.ipt_num[0] != 0)
5683 		fin->fin_nattag = &np->in_tag;
5684 	return 1;
5685 }
5686 
5687 
5688 /* ------------------------------------------------------------------------ */
5689 /* Function:    ipf_nat_proto                                               */
5690 /* Returns:     u_short* - pointer to transport header checksum to update,  */
5691 /*                         NULL if the transport protocol is not recognised */
5692 /*                         as needing a checksum update.                    */
5693 /* Parameters:  fin(I)    - pointer to packet information                   */
5694 /*              nat(I)    - pointer to NAT structure                        */
5695 /*              nflags(I) - NAT flags set for this packet                   */
5696 /*                                                                          */
5697 /* Return the pointer to the checksum field for each protocol so understood.*/
5698 /* If support for making other changes to a protocol header is required,    */
5699 /* that is not strictly 'address' translation, such as clamping the MSS in  */
5700 /* TCP down to a specific value, then do it from here.                      */
5701 /* ------------------------------------------------------------------------ */
5702 u_short *
5703 ipf_nat_proto(fr_info_t *fin, nat_t *nat, u_int nflags)
5704 {
5705 	icmphdr_t *icmp;
5706 	u_short *csump;
5707 	tcphdr_t *tcp;
5708 	udphdr_t *udp;
5709 
5710 	csump = NULL;
5711 	if (fin->fin_out == 0) {
5712 		fin->fin_rev = (nat->nat_dir & NAT_OUTBOUND);
5713 	} else {
5714 		fin->fin_rev = ((nat->nat_dir & NAT_OUTBOUND) == 0);
5715 	}
5716 
5717 	switch (fin->fin_p)
5718 	{
5719 	case IPPROTO_TCP :
5720 		tcp = fin->fin_dp;
5721 
5722 		if ((nflags & IPN_TCP) != 0)
5723 			csump = &tcp->th_sum;
5724 
5725 		/*
5726 		 * Do a MSS CLAMPING on a SYN packet,
5727 		 * only deal IPv4 for now.
5728 		 */
5729 		if ((nat->nat_mssclamp != 0) && (tcp->th_flags & TH_SYN) != 0)
5730 			ipf_nat_mssclamp(tcp, nat->nat_mssclamp, fin, csump);
5731 
5732 		break;
5733 
5734 	case IPPROTO_UDP :
5735 		udp = fin->fin_dp;
5736 
5737 		if ((nflags & IPN_UDP) != 0) {
5738 			if (udp->uh_sum != 0)
5739 				csump = &udp->uh_sum;
5740 		}
5741 		break;
5742 
5743 	case IPPROTO_ICMP :
5744 		icmp = fin->fin_dp;
5745 
5746 		if ((nflags & IPN_ICMPQUERY) != 0) {
5747 			if (icmp->icmp_cksum != 0)
5748 				csump = &icmp->icmp_cksum;
5749 		}
5750 		break;
5751 
5752 #ifdef USE_INET6
5753 	case IPPROTO_ICMPV6 :
5754 	    {
5755 		struct icmp6_hdr *icmp6 = (struct icmp6_hdr *)fin->fin_dp;
5756 
5757 		icmp6 = fin->fin_dp;
5758 
5759 		if ((nflags & IPN_ICMPQUERY) != 0) {
5760 			if (icmp6->icmp6_cksum != 0)
5761 				csump = &icmp6->icmp6_cksum;
5762 		}
5763 		break;
5764 	    }
5765 #endif
5766 	}
5767 	return csump;
5768 }
5769 
5770 
5771 /* ------------------------------------------------------------------------ */
5772 /* Function:    ipf_nat_expire                                              */
5773 /* Returns:     Nil                                                         */
5774 /* Parameters:  softc(I) - pointer to soft context main structure           */
5775 /*                                                                          */
5776 /* Check all of the timeout queues for entries at the top which need to be  */
5777 /* expired.                                                                 */
5778 /* ------------------------------------------------------------------------ */
5779 void
5780 ipf_nat_expire(ipf_main_softc_t *softc)
5781 {
5782 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5783 	ipftq_t *ifq, *ifqnext;
5784 	ipftqent_t *tqe, *tqn;
5785 	int i;
5786 	SPL_INT(s);
5787 
5788 	SPL_NET(s);
5789 	WRITE_ENTER(&softc->ipf_nat);
5790 	for (ifq = softn->ipf_nat_tcptq, i = 0; ifq != NULL;
5791 	     ifq = ifq->ifq_next) {
5792 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
5793 			if (tqe->tqe_die > softc->ipf_ticks)
5794 				break;
5795 			tqn = tqe->tqe_next;
5796 			ipf_nat_delete(softc, tqe->tqe_parent, NL_EXPIRE);
5797 		}
5798 	}
5799 
5800 	for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifq->ifq_next) {
5801 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
5802 			if (tqe->tqe_die > softc->ipf_ticks)
5803 				break;
5804 			tqn = tqe->tqe_next;
5805 			ipf_nat_delete(softc, tqe->tqe_parent, NL_EXPIRE);
5806 		}
5807 	}
5808 
5809 	for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifqnext) {
5810 		ifqnext = ifq->ifq_next;
5811 
5812 		if (((ifq->ifq_flags & IFQF_DELETE) != 0) &&
5813 		    (ifq->ifq_ref == 0)) {
5814 			ipf_freetimeoutqueue(softc, ifq);
5815 		}
5816 	}
5817 
5818 	if (softn->ipf_nat_doflush != 0) {
5819 		ipf_nat_extraflush(softc, softn, 2);
5820 		softn->ipf_nat_doflush = 0;
5821 	}
5822 
5823 	RWLOCK_EXIT(&softc->ipf_nat);
5824 	SPL_X(s);
5825 }
5826 
5827 
5828 /* ------------------------------------------------------------------------ */
5829 /* Function:    ipf_nat_sync                                                */
5830 /* Returns:     Nil                                                         */
5831 /* Parameters:  softc(I) - pointer to soft context main structure           */
5832 /*              ifp(I) - pointer to network interface                       */
5833 /*                                                                          */
5834 /* Walk through all of the currently active NAT sessions, looking for those */
5835 /* which need to have their translated address updated.                     */
5836 /* ------------------------------------------------------------------------ */
5837 void
5838 ipf_nat_sync(ipf_main_softc_t *softc, void *ifp)
5839 {
5840 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5841 	u_32_t sum1, sum2, sumd;
5842 	i6addr_t in;
5843 	ipnat_t *n;
5844 	nat_t *nat;
5845 	void *ifp2;
5846 	int idx;
5847 	SPL_INT(s);
5848 
5849 	if (softc->ipf_running <= 0)
5850 		return;
5851 
5852 	/*
5853 	 * Change IP addresses for NAT sessions for any protocol except TCP
5854 	 * since it will break the TCP connection anyway.  The only rules
5855 	 * which will get changed are those which are "map ... -> 0/32",
5856 	 * where the rule specifies the address is taken from the interface.
5857 	 */
5858 	SPL_NET(s);
5859 	WRITE_ENTER(&softc->ipf_nat);
5860 
5861 	if (softc->ipf_running <= 0) {
5862 		RWLOCK_EXIT(&softc->ipf_nat);
5863 		return;
5864 	}
5865 
5866 	for (nat = softn->ipf_nat_instances; nat; nat = nat->nat_next) {
5867 		if ((nat->nat_flags & IPN_TCP) != 0)
5868 			continue;
5869 
5870 		n = nat->nat_ptr;
5871 		if (n != NULL) {
5872 			if (n->in_v[1] == 4) {
5873 				if (n->in_redir & NAT_MAP) {
5874 					if ((n->in_nsrcaddr != 0) ||
5875 					    (n->in_nsrcmsk != 0xffffffff))
5876 						continue;
5877 				} else if (n->in_redir & NAT_REDIRECT) {
5878 					if ((n->in_ndstaddr != 0) ||
5879 					    (n->in_ndstmsk != 0xffffffff))
5880 						continue;
5881 				}
5882 			}
5883 #ifdef USE_INET6
5884 			if (n->in_v[1] == 4) {
5885 				if (n->in_redir & NAT_MAP) {
5886 					if (!IP6_ISZERO(&n->in_nsrcaddr) ||
5887 					    !IP6_ISONES(&n->in_nsrcmsk))
5888 						continue;
5889 				} else if (n->in_redir & NAT_REDIRECT) {
5890 					if (!IP6_ISZERO(&n->in_ndstaddr) ||
5891 					    !IP6_ISONES(&n->in_ndstmsk))
5892 						continue;
5893 				}
5894 			}
5895 #endif
5896 		}
5897 
5898 		if (((ifp == NULL) || (ifp == nat->nat_ifps[0]) ||
5899 		     (ifp == nat->nat_ifps[1]))) {
5900 			nat->nat_ifps[0] = GETIFP(nat->nat_ifnames[0],
5901 						  nat->nat_v[0]);
5902 			if ((nat->nat_ifps[0] != NULL) &&
5903 			    (nat->nat_ifps[0] != (void *)-1)) {
5904 				nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
5905 			}
5906 			if (nat->nat_ifnames[1][0] != '\0') {
5907 				nat->nat_ifps[1] = GETIFP(nat->nat_ifnames[1],
5908 							  nat->nat_v[1]);
5909 			} else {
5910 				nat->nat_ifps[1] = nat->nat_ifps[0];
5911 			}
5912 			if ((nat->nat_ifps[1] != NULL) &&
5913 			    (nat->nat_ifps[1] != (void *)-1)) {
5914 				nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
5915 			}
5916 			ifp2 = nat->nat_ifps[0];
5917 			if (ifp2 == NULL)
5918 				continue;
5919 
5920 			/*
5921 			 * Change the map-to address to be the same as the
5922 			 * new one.
5923 			 */
5924 			sum1 = NATFSUM(nat, nat->nat_v[1], nat_nsrc6);
5925 			if (ipf_ifpaddr(softc, nat->nat_v[0], FRI_NORMAL, ifp2,
5926 				       &in, NULL) != -1) {
5927 				if (nat->nat_v[0] == 4)
5928 					nat->nat_nsrcip = in.in4;
5929 			}
5930 			sum2 = NATFSUM(nat, nat->nat_v[1], nat_nsrc6);
5931 
5932 			if (sum1 == sum2)
5933 				continue;
5934 			/*
5935 			 * Readjust the checksum adjustment to take into
5936 			 * account the new IP#.
5937 			 */
5938 			CALC_SUMD(sum1, sum2, sumd);
5939 			/* XXX - dont change for TCP when solaris does
5940 			 * hardware checksumming.
5941 			 */
5942 			sumd += nat->nat_sumd[0];
5943 			nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
5944 			nat->nat_sumd[1] = nat->nat_sumd[0];
5945 		}
5946 	}
5947 
5948 	for (n = softn->ipf_nat_list; (n != NULL); n = n->in_next) {
5949 		char *base = n->in_names;
5950 
5951 		if ((ifp == NULL) || (n->in_ifps[0] == ifp))
5952 			n->in_ifps[0] = ipf_resolvenic(softc,
5953 						       base + n->in_ifnames[0],
5954 						       n->in_v[0]);
5955 		if ((ifp == NULL) || (n->in_ifps[1] == ifp))
5956 			n->in_ifps[1] = ipf_resolvenic(softc,
5957 						       base + n->in_ifnames[1],
5958 						       n->in_v[1]);
5959 
5960 		if (n->in_redir & NAT_REDIRECT)
5961 			idx = 1;
5962 		else
5963 			idx = 0;
5964 
5965 		if (((ifp == NULL) || (n->in_ifps[idx] == ifp)) &&
5966 		    (n->in_ifps[idx] != NULL &&
5967 		     n->in_ifps[idx] != (void *)-1)) {
5968 
5969 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_osrc,
5970 					     0, n->in_ifps[idx]);
5971 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_odst,
5972 					     0, n->in_ifps[idx]);
5973 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_nsrc,
5974 					     0, n->in_ifps[idx]);
5975 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_ndst,
5976 					     0, n->in_ifps[idx]);
5977 		}
5978 	}
5979 	RWLOCK_EXIT(&softc->ipf_nat);
5980 	SPL_X(s);
5981 }
5982 
5983 
5984 /* ------------------------------------------------------------------------ */
5985 /* Function:    ipf_nat_icmpquerytype                                       */
5986 /* Returns:     int - 1 == success, 0 == failure                            */
5987 /* Parameters:  icmptype(I) - ICMP type number                              */
5988 /*                                                                          */
5989 /* Tests to see if the ICMP type number passed is a query/response type or  */
5990 /* not.                                                                     */
5991 /* ------------------------------------------------------------------------ */
5992 static int
5993 ipf_nat_icmpquerytype(int icmptype)
5994 {
5995 
5996 	/*
5997 	 * For the ICMP query NAT code, it is essential that both the query
5998 	 * and the reply match on the NAT rule. Because the NAT structure
5999 	 * does not keep track of the icmptype, and a single NAT structure
6000 	 * is used for all icmp types with the same src, dest and id, we
6001 	 * simply define the replies as queries as well. The funny thing is,
6002 	 * although it seems silly to call a reply a query, this is exactly
6003 	 * as it is defined in the IPv4 specification
6004 	 */
6005 	switch (icmptype)
6006 	{
6007 	case ICMP_ECHOREPLY:
6008 	case ICMP_ECHO:
6009 	/* route aedvertisement/solliciation is currently unsupported: */
6010 	/* it would require rewriting the ICMP data section            */
6011 	case ICMP_TSTAMP:
6012 	case ICMP_TSTAMPREPLY:
6013 	case ICMP_IREQ:
6014 	case ICMP_IREQREPLY:
6015 	case ICMP_MASKREQ:
6016 	case ICMP_MASKREPLY:
6017 		return 1;
6018 	default:
6019 		return 0;
6020 	}
6021 }
6022 
6023 
6024 /* ------------------------------------------------------------------------ */
6025 /* Function:    nat_log                                                     */
6026 /* Returns:     Nil                                                         */
6027 /* Parameters:  softc(I) - pointer to soft context main structure           */
6028 /*              softn(I) - pointer to NAT context structure                 */
6029 /*              nat(I)    - pointer to NAT structure                        */
6030 /*              action(I) - action related to NAT structure being performed */
6031 /*                                                                          */
6032 /* Creates a NAT log entry.                                                 */
6033 /* ------------------------------------------------------------------------ */
6034 void
6035 ipf_nat_log(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, struct nat *nat,
6036     u_int action)
6037 {
6038 #ifdef	IPFILTER_LOG
6039 # ifndef LARGE_NAT
6040 	struct ipnat *np;
6041 	int rulen;
6042 # endif
6043 	struct natlog natl;
6044 	void *items[1];
6045 	size_t sizes[1];
6046 	int types[1];
6047 
6048 	bcopy((char *)&nat->nat_osrc6, (char *)&natl.nl_osrcip,
6049 	      sizeof(natl.nl_osrcip));
6050 	bcopy((char *)&nat->nat_nsrc6, (char *)&natl.nl_nsrcip,
6051 	      sizeof(natl.nl_nsrcip));
6052 	bcopy((char *)&nat->nat_odst6, (char *)&natl.nl_odstip,
6053 	      sizeof(natl.nl_odstip));
6054 	bcopy((char *)&nat->nat_ndst6, (char *)&natl.nl_ndstip,
6055 	      sizeof(natl.nl_ndstip));
6056 
6057 	natl.nl_bytes[0] = nat->nat_bytes[0];
6058 	natl.nl_bytes[1] = nat->nat_bytes[1];
6059 	natl.nl_pkts[0] = nat->nat_pkts[0];
6060 	natl.nl_pkts[1] = nat->nat_pkts[1];
6061 	natl.nl_odstport = nat->nat_odport;
6062 	natl.nl_osrcport = nat->nat_osport;
6063 	natl.nl_nsrcport = nat->nat_nsport;
6064 	natl.nl_ndstport = nat->nat_ndport;
6065 	natl.nl_p[0] = nat->nat_pr[0];
6066 	natl.nl_p[1] = nat->nat_pr[1];
6067 	natl.nl_v[0] = nat->nat_v[0];
6068 	natl.nl_v[1] = nat->nat_v[1];
6069 	natl.nl_type = nat->nat_redir;
6070 	natl.nl_action = action;
6071 	natl.nl_rule = -1;
6072 
6073 	bcopy(nat->nat_ifnames[0], natl.nl_ifnames[0],
6074 	      sizeof(nat->nat_ifnames[0]));
6075 	bcopy(nat->nat_ifnames[1], natl.nl_ifnames[1],
6076 	      sizeof(nat->nat_ifnames[1]));
6077 
6078 # ifndef LARGE_NAT
6079 	if (nat->nat_ptr != NULL) {
6080 		for (rulen = 0, np = softn->ipf_nat_list; np != NULL;
6081 		     np = np->in_next, rulen++)
6082 			if (np == nat->nat_ptr) {
6083 				natl.nl_rule = rulen;
6084 				break;
6085 			}
6086 	}
6087 # endif
6088 	items[0] = &natl;
6089 	sizes[0] = sizeof(natl);
6090 	types[0] = 0;
6091 
6092 	(void) ipf_log_items(softc, IPL_LOGNAT, NULL, items, sizes, types, 1);
6093 #endif
6094 }
6095 
6096 
6097 #if defined(__OpenBSD__)
6098 /* ------------------------------------------------------------------------ */
6099 /* Function:    ipf_nat_ifdetach                                            */
6100 /* Returns:     Nil                                                         */
6101 /* Parameters:  ifp(I) - pointer to network interface                       */
6102 /*                                                                          */
6103 /* Compatibility interface for OpenBSD to trigger the correct updating of   */
6104 /* interface references within IPFilter.                                    */
6105 /* ------------------------------------------------------------------------ */
6106 void
6107 ipf_nat_ifdetach(ifp)
6108 	void *ifp;
6109 {
6110 	ipf_main_softc_t *softc;
6111 
6112 	softc = ipf_get_softc(0);
6113 
6114 	ipf_sync(ifp);
6115 	return;
6116 }
6117 #endif
6118 
6119 
6120 /* ------------------------------------------------------------------------ */
6121 /* Function:    ipf_nat_rule_deref                                          */
6122 /* Returns:     Nil                                                         */
6123 /* Parameters:  softc(I) - pointer to soft context main structure           */
6124 /*              inp(I)   - pointer to pointer to NAT rule                   */
6125 /* Write Locks: ipf_nat                                                     */
6126 /*                                                                          */
6127 /* Dropping the refernce count for a rule means that whatever held the      */
6128 /* pointer to this rule (*inp) is no longer interested in it and when the   */
6129 /* reference count drops to zero, any resources allocated for the rule can  */
6130 /* be released and the rule itself free'd.                                  */
6131 /* ------------------------------------------------------------------------ */
6132 void
6133 ipf_nat_rule_deref(ipf_main_softc_t *softc, ipnat_t **inp)
6134 {
6135 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6136 	ipnat_t *n;
6137 
6138 	n = *inp;
6139 	*inp = NULL;
6140 	n->in_use--;
6141 	if (n->in_use > 0)
6142 		return;
6143 
6144 	if (n->in_apr != NULL)
6145 		ipf_proxy_deref(n->in_apr);
6146 
6147 	ipf_nat_rule_fini(softc, n);
6148 
6149 	if (n->in_redir & NAT_REDIRECT) {
6150 		if ((n->in_flags & IPN_PROXYRULE) == 0) {
6151 			ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules_rdr);
6152 		}
6153 	}
6154 	if (n->in_redir & (NAT_MAP|NAT_MAPBLK)) {
6155 		if ((n->in_flags & IPN_PROXYRULE) == 0) {
6156 			ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules_map);
6157 		}
6158 	}
6159 
6160 	if (n->in_tqehead[0] != NULL) {
6161 		if (ipf_deletetimeoutqueue(n->in_tqehead[0]) == 0) {
6162 			ipf_freetimeoutqueue(softc, n->in_tqehead[1]);
6163 		}
6164 	}
6165 
6166 	if (n->in_tqehead[1] != NULL) {
6167 		if (ipf_deletetimeoutqueue(n->in_tqehead[1]) == 0) {
6168 			ipf_freetimeoutqueue(softc, n->in_tqehead[1]);
6169 		}
6170 	}
6171 
6172 	if ((n->in_flags & IPN_PROXYRULE) == 0) {
6173 		ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules);
6174 	}
6175 
6176 	MUTEX_DESTROY(&n->in_lock);
6177 
6178 	KFREES(n, n->in_size);
6179 
6180 #if SOLARIS && !defined(INSTANCES)
6181 	if (softn->ipf_nat_stats.ns_rules == 0)
6182 		pfil_delayed_copy = 1;
6183 #endif
6184 }
6185 
6186 
6187 /* ------------------------------------------------------------------------ */
6188 /* Function:    ipf_nat_deref                                               */
6189 /* Returns:     Nil                                                         */
6190 /* Parameters:  softc(I) - pointer to soft context main structure           */
6191 /*              natp(I)  - pointer to pointer to NAT table entry            */
6192 /*                                                                          */
6193 /* Decrement the reference counter for this NAT table entry and free it if  */
6194 /* there are no more things using it.                                       */
6195 /*                                                                          */
6196 /* IF nat_ref == 1 when this function is called, then we have an orphan nat */
6197 /* structure *because* it only gets called on paths _after_ nat_ref has been*/
6198 /* incremented.  If nat_ref == 1 then we shouldn't decrement it here        */
6199 /* because nat_delete() will do that and send nat_ref to -1.                */
6200 /*                                                                          */
6201 /* Holding the lock on nat_lock is required to serialise nat_delete() being */
6202 /* called from a NAT flush ioctl with a deref happening because of a packet.*/
6203 /* ------------------------------------------------------------------------ */
6204 void
6205 ipf_nat_deref(ipf_main_softc_t *softc, nat_t **natp)
6206 {
6207 	nat_t *nat;
6208 
6209 	nat = *natp;
6210 	*natp = NULL;
6211 
6212 	MUTEX_ENTER(&nat->nat_lock);
6213 	if (nat->nat_ref > 1) {
6214 		nat->nat_ref--;
6215 		ASSERT(nat->nat_ref >= 0);
6216 		MUTEX_EXIT(&nat->nat_lock);
6217 		return;
6218 	}
6219 	MUTEX_EXIT(&nat->nat_lock);
6220 
6221 	WRITE_ENTER(&softc->ipf_nat);
6222 	ipf_nat_delete(softc, nat, NL_EXPIRE);
6223 	RWLOCK_EXIT(&softc->ipf_nat);
6224 }
6225 
6226 
6227 /* ------------------------------------------------------------------------ */
6228 /* Function:    ipf_nat_clone                                               */
6229 /* Returns:     ipstate_t* - NULL == cloning failed,                        */
6230 /*                           else pointer to new state structure            */
6231 /* Parameters:  fin(I) - pointer to packet information                      */
6232 /*              is(I)  - pointer to master state structure                  */
6233 /* Write Lock:  ipf_nat                                                     */
6234 /*                                                                          */
6235 /* Create a "duplcate" state table entry from the master.                   */
6236 /* ------------------------------------------------------------------------ */
6237 nat_t *
6238 ipf_nat_clone(fr_info_t *fin, nat_t *nat)
6239 {
6240 	ipf_main_softc_t *softc = fin->fin_main_soft;
6241 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6242 	frentry_t *fr;
6243 	nat_t *clone;
6244 	ipnat_t *np;
6245 
6246 	KMALLOC(clone, nat_t *);
6247 	if (clone == NULL) {
6248 		NBUMPSIDED(fin->fin_out, ns_clone_nomem);
6249 		return NULL;
6250 	}
6251 	bcopy((char *)nat, (char *)clone, sizeof(*clone));
6252 
6253 	MUTEX_NUKE(&clone->nat_lock);
6254 
6255 	clone->nat_rev = fin->fin_rev;
6256 	clone->nat_aps = NULL;
6257 	/*
6258 	 * Initialize all these so that ipf_nat_delete() doesn't cause a crash.
6259 	 */
6260 	clone->nat_tqe.tqe_pnext = NULL;
6261 	clone->nat_tqe.tqe_next = NULL;
6262 	clone->nat_tqe.tqe_ifq = NULL;
6263 	clone->nat_tqe.tqe_parent = clone;
6264 
6265 	clone->nat_flags &= ~SI_CLONE;
6266 	clone->nat_flags |= SI_CLONED;
6267 
6268 	if (clone->nat_hm)
6269 		clone->nat_hm->hm_ref++;
6270 
6271 	if (ipf_nat_insert(softc, softn, clone) == -1) {
6272 		KFREE(clone);
6273 		NBUMPSIDED(fin->fin_out, ns_insert_fail);
6274 		return NULL;
6275 	}
6276 
6277 	np = clone->nat_ptr;
6278 	if (np != NULL) {
6279 		if (softn->ipf_nat_logging)
6280 			ipf_nat_log(softc, softn, clone, NL_CLONE);
6281 		np->in_use++;
6282 	}
6283 	fr = clone->nat_fr;
6284 	if (fr != NULL) {
6285 		MUTEX_ENTER(&fr->fr_lock);
6286 		fr->fr_ref++;
6287 		MUTEX_EXIT(&fr->fr_lock);
6288 	}
6289 
6290 
6291 	/*
6292 	 * Because the clone is created outside the normal loop of things and
6293 	 * TCP has special needs in terms of state, initialise the timeout
6294 	 * state of the new NAT from here.
6295 	 */
6296 	if (clone->nat_pr[0] == IPPROTO_TCP) {
6297 		(void) ipf_tcp_age(&clone->nat_tqe, fin, softn->ipf_nat_tcptq,
6298 				   clone->nat_flags, 2);
6299 	}
6300 	clone->nat_sync = ipf_sync_new(softc, SMC_NAT, fin, clone);
6301 	if (softn->ipf_nat_logging)
6302 		ipf_nat_log(softc, softn, clone, NL_CLONE);
6303 	return clone;
6304 }
6305 
6306 
6307 /* ------------------------------------------------------------------------ */
6308 /* Function:   ipf_nat_wildok                                               */
6309 /* Returns:    int - 1 == packet's ports match wildcards                    */
6310 /*                   0 == packet's ports don't match wildcards              */
6311 /* Parameters: nat(I)   - NAT entry                                         */
6312 /*             sport(I) - source port                                       */
6313 /*             dport(I) - destination port                                  */
6314 /*             flags(I) - wildcard flags                                    */
6315 /*             dir(I)   - packet direction                                  */
6316 /*                                                                          */
6317 /* Use NAT entry and packet direction to determine which combination of     */
6318 /* wildcard flags should be used.                                           */
6319 /* ------------------------------------------------------------------------ */
6320 int
6321 ipf_nat_wildok(nat_t *nat, int sport, int dport, int flags, int dir)
6322 {
6323 	/*
6324 	 * When called by       dir is set to
6325 	 * nat_inlookup         NAT_INBOUND (0)
6326 	 * nat_outlookup        NAT_OUTBOUND (1)
6327 	 *
6328 	 * We simply combine the packet's direction in dir with the original
6329 	 * "intended" direction of that NAT entry in nat->nat_dir to decide
6330 	 * which combination of wildcard flags to allow.
6331 	 */
6332 	switch ((dir << 1) | (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND)))
6333 	{
6334 	case 3: /* outbound packet / outbound entry */
6335 		if (((nat->nat_osport == sport) ||
6336 		    (flags & SI_W_SPORT)) &&
6337 		    ((nat->nat_odport == dport) ||
6338 		    (flags & SI_W_DPORT)))
6339 			return 1;
6340 		break;
6341 	case 2: /* outbound packet / inbound entry */
6342 		if (((nat->nat_osport == dport) ||
6343 		    (flags & SI_W_SPORT)) &&
6344 		    ((nat->nat_odport == sport) ||
6345 		    (flags & SI_W_DPORT)))
6346 			return 1;
6347 		break;
6348 	case 1: /* inbound packet / outbound entry */
6349 		if (((nat->nat_osport == dport) ||
6350 		    (flags & SI_W_SPORT)) &&
6351 		    ((nat->nat_odport == sport) ||
6352 		    (flags & SI_W_DPORT)))
6353 			return 1;
6354 		break;
6355 	case 0: /* inbound packet / inbound entry */
6356 		if (((nat->nat_osport == sport) ||
6357 		    (flags & SI_W_SPORT)) &&
6358 		    ((nat->nat_odport == dport) ||
6359 		    (flags & SI_W_DPORT)))
6360 			return 1;
6361 		break;
6362 	default:
6363 		break;
6364 	}
6365 
6366 	return(0);
6367 }
6368 
6369 
6370 /* ------------------------------------------------------------------------ */
6371 /* Function:    nat_mssclamp                                                */
6372 /* Returns:     Nil                                                         */
6373 /* Parameters:  tcp(I)    - pointer to TCP header                           */
6374 /*              maxmss(I) - value to clamp the TCP MSS to                   */
6375 /*              fin(I)    - pointer to packet information                   */
6376 /*              csump(I)  - pointer to TCP checksum                         */
6377 /*                                                                          */
6378 /* Check for MSS option and clamp it if necessary.  If found and changed,   */
6379 /* then the TCP header checksum will be updated to reflect the change in    */
6380 /* the MSS.                                                                 */
6381 /* ------------------------------------------------------------------------ */
6382 static void
6383 ipf_nat_mssclamp(tcphdr_t *tcp, u_32_t maxmss, fr_info_t *fin, u_short *csump)
6384 {
6385 	u_char *cp, *ep, opt;
6386 	int hlen, advance;
6387 	u_32_t mss, sumd;
6388 
6389 	hlen = TCP_OFF(tcp) << 2;
6390 	if (hlen > sizeof(*tcp)) {
6391 		cp = (u_char *)tcp + sizeof(*tcp);
6392 		ep = (u_char *)tcp + hlen;
6393 
6394 		while (cp < ep) {
6395 			opt = cp[0];
6396 			if (opt == TCPOPT_EOL)
6397 				break;
6398 			else if (opt == TCPOPT_NOP) {
6399 				cp++;
6400 				continue;
6401 			}
6402 
6403 			if (cp + 1 >= ep)
6404 				break;
6405 			advance = cp[1];
6406 			if ((cp + advance > ep) || (advance <= 0))
6407 				break;
6408 			switch (opt)
6409 			{
6410 			case TCPOPT_MAXSEG:
6411 				if (advance != 4)
6412 					break;
6413 				mss = cp[2] * 256 + cp[3];
6414 				if (mss > maxmss) {
6415 					cp[2] = maxmss / 256;
6416 					cp[3] = maxmss & 0xff;
6417 					CALC_SUMD(mss, maxmss, sumd);
6418 					ipf_fix_outcksum(0, csump, sumd, 0);
6419 				}
6420 				break;
6421 			default:
6422 				/* ignore unknown options */
6423 				break;
6424 			}
6425 
6426 			cp += advance;
6427 		}
6428 	}
6429 }
6430 
6431 
6432 /* ------------------------------------------------------------------------ */
6433 /* Function:    ipf_nat_setqueue                                            */
6434 /* Returns:     Nil                                                         */
6435 /* Parameters:  softc(I) - pointer to soft context main structure           */
6436 /*              softn(I) - pointer to NAT context structure                 */
6437 /*              nat(I)- pointer to NAT structure                            */
6438 /* Locks:       ipf_nat (read or write)                                     */
6439 /*                                                                          */
6440 /* Put the NAT entry on its default queue entry, using rev as a helped in   */
6441 /* determining which queue it should be placed on.                          */
6442 /* ------------------------------------------------------------------------ */
6443 void
6444 ipf_nat_setqueue(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, nat_t *nat)
6445 {
6446 	ipftq_t *oifq, *nifq;
6447 	int rev = nat->nat_rev;
6448 
6449 	if (nat->nat_ptr != NULL)
6450 		nifq = nat->nat_ptr->in_tqehead[rev];
6451 	else
6452 		nifq = NULL;
6453 
6454 	if (nifq == NULL) {
6455 		switch (nat->nat_pr[0])
6456 		{
6457 		case IPPROTO_UDP :
6458 			nifq = &softn->ipf_nat_udptq;
6459 			break;
6460 		case IPPROTO_ICMP :
6461 			nifq = &softn->ipf_nat_icmptq;
6462 			break;
6463 		case IPPROTO_TCP :
6464 			nifq = softn->ipf_nat_tcptq +
6465 			       nat->nat_tqe.tqe_state[rev];
6466 			break;
6467 		default :
6468 			nifq = &softn->ipf_nat_iptq;
6469 			break;
6470 		}
6471 	}
6472 
6473 	oifq = nat->nat_tqe.tqe_ifq;
6474 	/*
6475 	 * If it's currently on a timeout queue, move it from one queue to
6476 	 * another, else put it on the end of the newly determined queue.
6477 	 */
6478 	if (oifq != NULL)
6479 		ipf_movequeue(softc->ipf_ticks, &nat->nat_tqe, oifq, nifq);
6480 	else
6481 		ipf_queueappend(softc->ipf_ticks, &nat->nat_tqe, nifq, nat);
6482 	return;
6483 }
6484 
6485 
6486 /* ------------------------------------------------------------------------ */
6487 /* Function:    nat_getnext                                                 */
6488 /* Returns:     int - 0 == ok, else error                                   */
6489 /* Parameters:  softc(I) - pointer to soft context main structure           */
6490 /*              t(I)   - pointer to ipftoken structure                      */
6491 /*              itp(I) - pointer to ipfgeniter_t structure                  */
6492 /*                                                                          */
6493 /* Fetch the next nat/ipnat structure pointer from the linked list and      */
6494 /* copy it out to the storage space pointed to by itp_data.  The next item  */
6495 /* in the list to look at is put back in the ipftoken struture.             */
6496 /* ------------------------------------------------------------------------ */
6497 static int
6498 ipf_nat_getnext(ipf_main_softc_t *softc, ipftoken_t *t, ipfgeniter_t *itp,
6499     ipfobj_t *objp)
6500 {
6501 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6502 	hostmap_t *hm, *nexthm = NULL, zerohm;
6503 	ipnat_t *ipn, *nextipnat = NULL, zeroipn;
6504 	nat_t *nat, *nextnat = NULL, zeronat;
6505 	int error = 0;
6506 	void *nnext;
6507 
6508 	if (itp->igi_nitems != 1) {
6509 		IPFERROR(60075);
6510 		return ENOSPC;
6511 	}
6512 
6513 	READ_ENTER(&softc->ipf_nat);
6514 
6515 	switch (itp->igi_type)
6516 	{
6517 	case IPFGENITER_HOSTMAP :
6518 		hm = t->ipt_data;
6519 		if (hm == NULL) {
6520 			nexthm = softn->ipf_hm_maplist;
6521 		} else {
6522 			nexthm = hm->hm_next;
6523 		}
6524 		if (nexthm != NULL) {
6525 			ATOMIC_INC32(nexthm->hm_ref);
6526 			t->ipt_data = nexthm;
6527 		} else {
6528 			bzero(&zerohm, sizeof(zerohm));
6529 			nexthm = &zerohm;
6530 			t->ipt_data = NULL;
6531 		}
6532 		nnext = nexthm->hm_next;
6533 		break;
6534 
6535 	case IPFGENITER_IPNAT :
6536 		ipn = t->ipt_data;
6537 		if (ipn == NULL) {
6538 			nextipnat = softn->ipf_nat_list;
6539 		} else {
6540 			nextipnat = ipn->in_next;
6541 		}
6542 		if (nextipnat != NULL) {
6543 			ATOMIC_INC32(nextipnat->in_use);
6544 			t->ipt_data = nextipnat;
6545 		} else {
6546 			bzero(&zeroipn, sizeof(zeroipn));
6547 			nextipnat = &zeroipn;
6548 			t->ipt_data = NULL;
6549 		}
6550 		nnext = nextipnat->in_next;
6551 		break;
6552 
6553 	case IPFGENITER_NAT :
6554 		nat = t->ipt_data;
6555 		if (nat == NULL) {
6556 			nextnat = softn->ipf_nat_instances;
6557 		} else {
6558 			nextnat = nat->nat_next;
6559 		}
6560 		if (nextnat != NULL) {
6561 			MUTEX_ENTER(&nextnat->nat_lock);
6562 			nextnat->nat_ref++;
6563 			MUTEX_EXIT(&nextnat->nat_lock);
6564 			t->ipt_data = nextnat;
6565 		} else {
6566 			bzero(&zeronat, sizeof(zeronat));
6567 			nextnat = &zeronat;
6568 			t->ipt_data = NULL;
6569 		}
6570 		nnext = nextnat->nat_next;
6571 		break;
6572 
6573 	default :
6574 		RWLOCK_EXIT(&softc->ipf_nat);
6575 		IPFERROR(60055);
6576 		return EINVAL;
6577 	}
6578 
6579 	RWLOCK_EXIT(&softc->ipf_nat);
6580 
6581 	objp->ipfo_ptr = itp->igi_data;
6582 
6583 	switch (itp->igi_type)
6584 	{
6585 	case IPFGENITER_HOSTMAP :
6586 		error = COPYOUT(nexthm, objp->ipfo_ptr, sizeof(*nexthm));
6587 		if (error != 0) {
6588 			IPFERROR(60049);
6589 			error = EFAULT;
6590 		}
6591 		if (hm != NULL) {
6592 			WRITE_ENTER(&softc->ipf_nat);
6593 			ipf_nat_hostmapdel(softc, &hm);
6594 			RWLOCK_EXIT(&softc->ipf_nat);
6595 		}
6596 		break;
6597 
6598 	case IPFGENITER_IPNAT :
6599 		objp->ipfo_size = nextipnat->in_size;
6600 		objp->ipfo_type = IPFOBJ_IPNAT;
6601 		error = ipf_outobjk(softc, objp, nextipnat);
6602 		if (ipn != NULL) {
6603 			WRITE_ENTER(&softc->ipf_nat);
6604 			ipf_nat_rule_deref(softc, &ipn);
6605 			RWLOCK_EXIT(&softc->ipf_nat);
6606 		}
6607 		break;
6608 
6609 	case IPFGENITER_NAT :
6610 		objp->ipfo_size = sizeof(nat_t);
6611 		objp->ipfo_type = IPFOBJ_NAT;
6612 		error = ipf_outobjk(softc, objp, nextnat);
6613 		if (nat != NULL)
6614 			ipf_nat_deref(softc, &nat);
6615 
6616 		break;
6617 	}
6618 
6619 	if (nnext == NULL)
6620 		ipf_token_mark_complete(t);
6621 
6622 	return error;
6623 }
6624 
6625 
6626 /* ------------------------------------------------------------------------ */
6627 /* Function:    nat_extraflush                                              */
6628 /* Returns:     int - 0 == success, -1 == failure                           */
6629 /* Parameters:  softc(I) - pointer to soft context main structure           */
6630 /*              softn(I) - pointer to NAT context structure                 */
6631 /*              which(I) - how to flush the active NAT table                */
6632 /* Write Locks: ipf_nat                                                     */
6633 /*                                                                          */
6634 /* Flush nat tables.  Three actions currently defined:                      */
6635 /* which == 0 : flush all nat table entries                                 */
6636 /* which == 1 : flush TCP connections which have started to close but are   */
6637 /*	      stuck for some reason.                                        */
6638 /* which == 2 : flush TCP connections which have been idle for a long time, */
6639 /*	      starting at > 4 days idle and working back in successive half-*/
6640 /*	      days to at most 12 hours old.  If this fails to free enough   */
6641 /*            slots then work backwards in half hour slots to 30 minutes.   */
6642 /*            If that too fails, then work backwards in 30 second intervals */
6643 /*            for the last 30 minutes to at worst 30 seconds idle.          */
6644 /* ------------------------------------------------------------------------ */
6645 static int
6646 ipf_nat_extraflush(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, int which)
6647 {
6648 	nat_t *nat, **natp;
6649 	ipftqent_t *tqn;
6650 	ipftq_t *ifq;
6651 	int removed;
6652 	SPL_INT(s);
6653 
6654 	removed = 0;
6655 
6656 	SPL_NET(s);
6657 	switch (which)
6658 	{
6659 	case 0 :
6660 		softn->ipf_nat_stats.ns_flush_all++;
6661 		/*
6662 		 * Style 0 flush removes everything...
6663 		 */
6664 		for (natp = &softn->ipf_nat_instances;
6665 		     ((nat = *natp) != NULL); ) {
6666 			ipf_nat_delete(softc, nat, NL_FLUSH);
6667 			removed++;
6668 		}
6669 		break;
6670 
6671 	case 1 :
6672 		softn->ipf_nat_stats.ns_flush_closing++;
6673 		/*
6674 		 * Since we're only interested in things that are closing,
6675 		 * we can start with the appropriate timeout queue.
6676 		 */
6677 		for (ifq = softn->ipf_nat_tcptq + IPF_TCPS_CLOSE_WAIT;
6678 		     ifq != NULL; ifq = ifq->ifq_next) {
6679 
6680 			for (tqn = ifq->ifq_head; tqn != NULL; ) {
6681 				nat = tqn->tqe_parent;
6682 				tqn = tqn->tqe_next;
6683 				if (nat->nat_pr[0] != IPPROTO_TCP ||
6684 				    nat->nat_pr[1] != IPPROTO_TCP)
6685 					break;
6686 				ipf_nat_delete(softc, nat, NL_EXPIRE);
6687 				removed++;
6688 			}
6689 		}
6690 
6691 		/*
6692 		 * Also need to look through the user defined queues.
6693 		 */
6694 		for (ifq = softn->ipf_nat_utqe; ifq != NULL;
6695 		     ifq = ifq->ifq_next) {
6696 			for (tqn = ifq->ifq_head; tqn != NULL; ) {
6697 				nat = tqn->tqe_parent;
6698 				tqn = tqn->tqe_next;
6699 				if (nat->nat_pr[0] != IPPROTO_TCP ||
6700 				    nat->nat_pr[1] != IPPROTO_TCP)
6701 					continue;
6702 
6703 				if ((nat->nat_tcpstate[0] >
6704 				     IPF_TCPS_ESTABLISHED) &&
6705 				    (nat->nat_tcpstate[1] >
6706 				     IPF_TCPS_ESTABLISHED)) {
6707 					ipf_nat_delete(softc, nat, NL_EXPIRE);
6708 					removed++;
6709 				}
6710 			}
6711 		}
6712 		break;
6713 
6714 		/*
6715 		 * Args 5-11 correspond to flushing those particular states
6716 		 * for TCP connections.
6717 		 */
6718 	case IPF_TCPS_CLOSE_WAIT :
6719 	case IPF_TCPS_FIN_WAIT_1 :
6720 	case IPF_TCPS_CLOSING :
6721 	case IPF_TCPS_LAST_ACK :
6722 	case IPF_TCPS_FIN_WAIT_2 :
6723 	case IPF_TCPS_TIME_WAIT :
6724 	case IPF_TCPS_CLOSED :
6725 		softn->ipf_nat_stats.ns_flush_state++;
6726 		tqn = softn->ipf_nat_tcptq[which].ifq_head;
6727 		while (tqn != NULL) {
6728 			nat = tqn->tqe_parent;
6729 			tqn = tqn->tqe_next;
6730 			ipf_nat_delete(softc, nat, NL_FLUSH);
6731 			removed++;
6732 		}
6733 		break;
6734 
6735 	default :
6736 		if (which < 30)
6737 			break;
6738 
6739 		softn->ipf_nat_stats.ns_flush_timeout++;
6740 		/*
6741 		 * Take a large arbitrary number to mean the number of seconds
6742 		 * for which which consider to be the maximum value we'll allow
6743 		 * the expiration to be.
6744 		 */
6745 		which = IPF_TTLVAL(which);
6746 		for (natp = &softn->ipf_nat_instances;
6747 		     ((nat = *natp) != NULL); ) {
6748 			if (softc->ipf_ticks - nat->nat_touched > which) {
6749 				ipf_nat_delete(softc, nat, NL_FLUSH);
6750 				removed++;
6751 			} else
6752 				natp = &nat->nat_next;
6753 		}
6754 		break;
6755 	}
6756 
6757 	if (which != 2) {
6758 		SPL_X(s);
6759 		return removed;
6760 	}
6761 
6762 	softn->ipf_nat_stats.ns_flush_queue++;
6763 
6764 	/*
6765 	 * Asked to remove inactive entries because the table is full, try
6766 	 * again, 3 times, if first attempt failed with a different criteria
6767 	 * each time.  The order tried in must be in decreasing age.
6768 	 * Another alternative is to implement random drop and drop N entries
6769 	 * at random until N have been freed up.
6770 	 */
6771 	if (softc->ipf_ticks - softn->ipf_nat_last_force_flush >
6772 	    IPF_TTLVAL(5)) {
6773 		softn->ipf_nat_last_force_flush = softc->ipf_ticks;
6774 
6775 		removed = ipf_queueflush(softc, ipf_nat_flush_entry,
6776 					 softn->ipf_nat_tcptq,
6777 					 softn->ipf_nat_utqe,
6778 					 &softn->ipf_nat_stats.ns_active,
6779 					 softn->ipf_nat_table_sz,
6780 					 softn->ipf_nat_table_wm_low);
6781 	}
6782 
6783 	SPL_X(s);
6784 	return removed;
6785 }
6786 
6787 
6788 /* ------------------------------------------------------------------------ */
6789 /* Function:    ipf_nat_flush_entry                                         */
6790 /* Returns:     0 - always succeeds                                         */
6791 /* Parameters:  softc(I) - pointer to soft context main structure           */
6792 /*              entry(I) - pointer to NAT entry                             */
6793 /* Write Locks: ipf_nat                                                     */
6794 /*                                                                          */
6795 /* This function is a stepping stone between ipf_queueflush() and           */
6796 /* nat_dlete().  It is used so we can provide a uniform interface via the   */
6797 /* ipf_queueflush() function.  Since the nat_delete() function returns void */
6798 /* we translate that to mean it always succeeds in deleting something.      */
6799 /* ------------------------------------------------------------------------ */
6800 static int
6801 ipf_nat_flush_entry(ipf_main_softc_t *softc, void *entry)
6802 {
6803 	ipf_nat_delete(softc, entry, NL_FLUSH);
6804 	return 0;
6805 }
6806 
6807 
6808 /* ------------------------------------------------------------------------ */
6809 /* Function:    ipf_nat_iterator                                            */
6810 /* Returns:     int - 0 == ok, else error                                   */
6811 /* Parameters:  softc(I) - pointer to soft context main structure           */
6812 /*              token(I) - pointer to ipftoken structure                    */
6813 /*              itp(I)   - pointer to ipfgeniter_t structure                */
6814 /*              obj(I)   - pointer to data description structure            */
6815 /*                                                                          */
6816 /* This function acts as a handler for the SIOCGENITER ioctls that use a    */
6817 /* generic structure to iterate through a list.  There are three different  */
6818 /* linked lists of NAT related information to go through: NAT rules, active */
6819 /* NAT mappings and the NAT fragment cache.                                 */
6820 /* ------------------------------------------------------------------------ */
6821 static int
6822 ipf_nat_iterator(ipf_main_softc_t *softc, ipftoken_t *token, ipfgeniter_t *itp,
6823     ipfobj_t *obj)
6824 {
6825 	int error;
6826 
6827 	if (itp->igi_data == NULL) {
6828 		IPFERROR(60052);
6829 		return EFAULT;
6830 	}
6831 
6832 	switch (itp->igi_type)
6833 	{
6834 	case IPFGENITER_HOSTMAP :
6835 	case IPFGENITER_IPNAT :
6836 	case IPFGENITER_NAT :
6837 		error = ipf_nat_getnext(softc, token, itp, obj);
6838 		break;
6839 
6840 	case IPFGENITER_NATFRAG :
6841 		error = ipf_frag_nat_next(softc, token, itp);
6842 		break;
6843 	default :
6844 		IPFERROR(60053);
6845 		error = EINVAL;
6846 		break;
6847 	}
6848 
6849 	return error;
6850 }
6851 
6852 
6853 /* ------------------------------------------------------------------------ */
6854 /* Function:    ipf_nat_setpending                                          */
6855 /* Returns:     Nil                                                         */
6856 /* Parameters:  softc(I) - pointer to soft context main structure           */
6857 /*              nat(I)   - pointer to NAT structure                         */
6858 /* Locks:       ipf_nat (read or write)                                     */
6859 /*                                                                          */
6860 /* Put the NAT entry on to the pending queue - this queue has a very short  */
6861 /* lifetime where items are put that can't be deleted straight away because */
6862 /* of locking issues but we want to delete them ASAP, anyway.  In calling   */
6863 /* this function, it is assumed that the owner (if there is one, as shown   */
6864 /* by nat_me) is no longer interested in it.                                */
6865 /* ------------------------------------------------------------------------ */
6866 void
6867 ipf_nat_setpending(ipf_main_softc_t *softc, nat_t *nat)
6868 {
6869 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6870 	ipftq_t *oifq;
6871 
6872 	oifq = nat->nat_tqe.tqe_ifq;
6873 	if (oifq != NULL)
6874 		ipf_movequeue(softc->ipf_ticks, &nat->nat_tqe, oifq,
6875 			      &softn->ipf_nat_pending);
6876 	else
6877 		ipf_queueappend(softc->ipf_ticks, &nat->nat_tqe,
6878 				&softn->ipf_nat_pending, nat);
6879 
6880 	if (nat->nat_me != NULL) {
6881 		*nat->nat_me = NULL;
6882 		nat->nat_me = NULL;
6883 		nat->nat_ref--;
6884 		ASSERT(nat->nat_ref >= 0);
6885 	}
6886 }
6887 
6888 
6889 /* ------------------------------------------------------------------------ */
6890 /* Function:    nat_newrewrite                                              */
6891 /* Returns:     int - -1 == error, 0 == success (no move), 1 == success and */
6892 /*                    allow rule to be moved if IPN_ROUNDR is set.          */
6893 /* Parameters:  fin(I) - pointer to packet information                      */
6894 /*              nat(I) - pointer to NAT entry                               */
6895 /*              ni(I)  - pointer to structure with misc. information needed */
6896 /*                       to create new NAT entry.                           */
6897 /* Write Lock:  ipf_nat                                                     */
6898 /*                                                                          */
6899 /* This function is responsible for setting up an active NAT session where  */
6900 /* we are changing both the source and destination parameters at the same   */
6901 /* time.  The loop in here works differently to elsewhere - each iteration  */
6902 /* is responsible for changing a single parameter that can be incremented.  */
6903 /* So one pass may increase the source IP#, next source port, next dest. IP#*/
6904 /* and the last destination port for a total of 4 iterations to try each.   */
6905 /* This is done to try and exhaustively use the translation space available.*/
6906 /* ------------------------------------------------------------------------ */
6907 static int
6908 ipf_nat_newrewrite(fr_info_t *fin, nat_t *nat, natinfo_t *nai)
6909 {
6910 	int src_search = 1;
6911 	int dst_search = 1;
6912 	fr_info_t frnat;
6913 	u_32_t flags;
6914 	u_short swap;
6915 	ipnat_t *np;
6916 	nat_t *natl;
6917 	int l = 0;
6918 	int changed;
6919 
6920 	natl = NULL;
6921 	changed = -1;
6922 	np = nai->nai_np;
6923 	flags = nat->nat_flags;
6924 	bcopy((char *)fin, (char *)&frnat, sizeof(*fin));
6925 
6926 	nat->nat_hm = NULL;
6927 
6928 	do {
6929 		changed = -1;
6930 		/* TRACE (l, src_search, dst_search, np) */
6931 
6932 		if ((src_search == 0) && (np->in_spnext == 0) &&
6933 		    (dst_search == 0) && (np->in_dpnext == 0)) {
6934 			if (l > 0)
6935 				return -1;
6936 		}
6937 
6938 		/*
6939 		 * Find a new source address
6940 		 */
6941 		if (ipf_nat_nextaddr(fin, &np->in_nsrc, &frnat.fin_saddr,
6942 				     &frnat.fin_saddr) == -1) {
6943 			return -1;
6944 		}
6945 
6946 		if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0xffffffff)) {
6947 			src_search = 0;
6948 			if (np->in_stepnext == 0)
6949 				np->in_stepnext = 1;
6950 
6951 		} else if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0)) {
6952 			src_search = 0;
6953 			if (np->in_stepnext == 0)
6954 				np->in_stepnext = 1;
6955 
6956 		} else if (np->in_nsrcmsk == 0xffffffff) {
6957 			src_search = 0;
6958 			if (np->in_stepnext == 0)
6959 				np->in_stepnext = 1;
6960 
6961 		} else if (np->in_nsrcmsk != 0xffffffff) {
6962 			if (np->in_stepnext == 0 && changed == -1) {
6963 				np->in_snip++;
6964 				np->in_stepnext++;
6965 				changed = 0;
6966 			}
6967 		}
6968 
6969 		if ((flags & IPN_TCPUDPICMP) != 0) {
6970 			if (np->in_spnext != 0)
6971 				frnat.fin_data[0] = np->in_spnext;
6972 
6973 			/*
6974 			 * Standard port translation.  Select next port.
6975 			 */
6976 			if ((flags & IPN_FIXEDSPORT) != 0) {
6977 				np->in_stepnext = 2;
6978 			} else if ((np->in_stepnext == 1) &&
6979 				   (changed == -1) && (natl != NULL)) {
6980 				np->in_spnext++;
6981 				np->in_stepnext++;
6982 				changed = 1;
6983 				if (np->in_spnext > np->in_spmax)
6984 					np->in_spnext = np->in_spmin;
6985 			}
6986 		} else {
6987 			np->in_stepnext = 2;
6988 		}
6989 		np->in_stepnext &= 0x3;
6990 
6991 		/*
6992 		 * Find a new destination address
6993 		 */
6994 		/* TRACE (fin, np, l, frnat) */
6995 
6996 		if (ipf_nat_nextaddr(fin, &np->in_ndst, &frnat.fin_daddr,
6997 				     &frnat.fin_daddr) == -1)
6998 			return -1;
6999 		if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0xffffffff)) {
7000 			dst_search = 0;
7001 			if (np->in_stepnext == 2)
7002 				np->in_stepnext = 3;
7003 
7004 		} else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0)) {
7005 			dst_search = 0;
7006 			if (np->in_stepnext == 2)
7007 				np->in_stepnext = 3;
7008 
7009 		} else if (np->in_ndstmsk == 0xffffffff) {
7010 			dst_search = 0;
7011 			if (np->in_stepnext == 2)
7012 				np->in_stepnext = 3;
7013 
7014 		} else if (np->in_ndstmsk != 0xffffffff) {
7015 			if ((np->in_stepnext == 2) && (changed == -1) &&
7016 			    (natl != NULL)) {
7017 				changed = 2;
7018 				np->in_stepnext++;
7019 				np->in_dnip++;
7020 			}
7021 		}
7022 
7023 		if ((flags & IPN_TCPUDPICMP) != 0) {
7024 			if (np->in_dpnext != 0)
7025 				frnat.fin_data[1] = np->in_dpnext;
7026 
7027 			/*
7028 			 * Standard port translation.  Select next port.
7029 			 */
7030 			if ((flags & IPN_FIXEDDPORT) != 0) {
7031 				np->in_stepnext = 0;
7032 			} else if (np->in_stepnext == 3 && changed == -1) {
7033 				np->in_dpnext++;
7034 				np->in_stepnext++;
7035 				changed = 3;
7036 				if (np->in_dpnext > np->in_dpmax)
7037 					np->in_dpnext = np->in_dpmin;
7038 			}
7039 		} else {
7040 			if (np->in_stepnext == 3)
7041 				np->in_stepnext = 0;
7042 		}
7043 
7044 		/* TRACE (frnat) */
7045 
7046 		/*
7047 		 * Here we do a lookup of the connection as seen from
7048 		 * the outside.  If an IP# pair already exists, try
7049 		 * again.  So if you have A->B becomes C->B, you can
7050 		 * also have D->E become C->E but not D->B causing
7051 		 * another C->B.  Also take protocol and ports into
7052 		 * account when determining whether a pre-existing
7053 		 * NAT setup will cause an external conflict where
7054 		 * this is appropriate.
7055 		 *
7056 		 * fin_data[] is swapped around because we are doing a
7057 		 * lookup of the packet is if it were moving in the opposite
7058 		 * direction of the one we are working with now.
7059 		 */
7060 		if (flags & IPN_TCPUDP) {
7061 			swap = frnat.fin_data[0];
7062 			frnat.fin_data[0] = frnat.fin_data[1];
7063 			frnat.fin_data[1] = swap;
7064 		}
7065 		if (fin->fin_out == 1) {
7066 			natl = ipf_nat_inlookup(&frnat,
7067 						flags & ~(SI_WILDP|NAT_SEARCH),
7068 						(u_int)frnat.fin_p,
7069 						frnat.fin_dst, frnat.fin_src);
7070 
7071 		} else {
7072 			natl = ipf_nat_outlookup(&frnat,
7073 						 flags & ~(SI_WILDP|NAT_SEARCH),
7074 						 (u_int)frnat.fin_p,
7075 						 frnat.fin_dst, frnat.fin_src);
7076 		}
7077 		if (flags & IPN_TCPUDP) {
7078 			swap = frnat.fin_data[0];
7079 			frnat.fin_data[0] = frnat.fin_data[1];
7080 			frnat.fin_data[1] = swap;
7081 		}
7082 
7083 		/* TRACE natl, in_stepnext, l */
7084 
7085 		if ((natl != NULL) && (l > 8))	/* XXX 8 is arbitrary */
7086 			return -1;
7087 
7088 		np->in_stepnext &= 0x3;
7089 
7090 		l++;
7091 		changed = -1;
7092 	} while (natl != NULL);
7093 
7094 	nat->nat_osrcip = fin->fin_src;
7095 	nat->nat_odstip = fin->fin_dst;
7096 	nat->nat_nsrcip = frnat.fin_src;
7097 	nat->nat_ndstip = frnat.fin_dst;
7098 
7099 	if ((flags & IPN_TCPUDP) != 0) {
7100 		nat->nat_osport = htons(fin->fin_data[0]);
7101 		nat->nat_odport = htons(fin->fin_data[1]);
7102 		nat->nat_nsport = htons(frnat.fin_data[0]);
7103 		nat->nat_ndport = htons(frnat.fin_data[1]);
7104 	} else if ((flags & IPN_ICMPQUERY) != 0) {
7105 		nat->nat_oicmpid = fin->fin_data[1];
7106 		nat->nat_nicmpid = frnat.fin_data[1];
7107 	}
7108 
7109 	return 0;
7110 }
7111 
7112 
7113 /* ------------------------------------------------------------------------ */
7114 /* Function:    nat_newdivert                                               */
7115 /* Returns:     int - -1 == error, 0 == success                             */
7116 /* Parameters:  fin(I) - pointer to packet information                      */
7117 /*              nat(I) - pointer to NAT entry                               */
7118 /*              ni(I)  - pointer to structure with misc. information needed */
7119 /*                       to create new NAT entry.                           */
7120 /* Write Lock:  ipf_nat                                                     */
7121 /*                                                                          */
7122 /* Create a new NAT  divert session as defined by the NAT rule.  This is    */
7123 /* somewhat different to other NAT session creation routines because we     */
7124 /* do not iterate through either port numbers or IP addresses, searching    */
7125 /* for a unique mapping, however, a complimentary duplicate check is made.  */
7126 /* ------------------------------------------------------------------------ */
7127 static int
7128 ipf_nat_newdivert(fr_info_t *fin, nat_t *nat, natinfo_t *nai)
7129 {
7130 	ipf_main_softc_t *softc = fin->fin_main_soft;
7131 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7132 	fr_info_t frnat;
7133 	ipnat_t *np;
7134 	nat_t *natl;
7135 	int p;
7136 
7137 	np = nai->nai_np;
7138 	bcopy((char *)fin, (char *)&frnat, sizeof(*fin));
7139 
7140 	nat->nat_pr[0] = 0;
7141 	nat->nat_osrcaddr = fin->fin_saddr;
7142 	nat->nat_odstaddr = fin->fin_daddr;
7143 	frnat.fin_saddr = htonl(np->in_snip);
7144 	frnat.fin_daddr = htonl(np->in_dnip);
7145 	if ((nat->nat_flags & IPN_TCPUDP) != 0) {
7146 		nat->nat_osport = htons(fin->fin_data[0]);
7147 		nat->nat_odport = htons(fin->fin_data[1]);
7148 	} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
7149 		nat->nat_oicmpid = fin->fin_data[1];
7150 	}
7151 
7152 	if (np->in_redir & NAT_DIVERTUDP) {
7153 		frnat.fin_data[0] = np->in_spnext;
7154 		frnat.fin_data[1] = np->in_dpnext;
7155 		frnat.fin_flx |= FI_TCPUDP;
7156 		p = IPPROTO_UDP;
7157 	} else {
7158 		frnat.fin_flx &= ~FI_TCPUDP;
7159 		p = IPPROTO_IPIP;
7160 	}
7161 
7162 	if (fin->fin_out == 1) {
7163 		natl = ipf_nat_inlookup(&frnat, 0, p,
7164 					frnat.fin_dst, frnat.fin_src);
7165 
7166 	} else {
7167 		natl = ipf_nat_outlookup(&frnat, 0, p,
7168 					 frnat.fin_dst, frnat.fin_src);
7169 	}
7170 
7171 	if (natl != NULL) {
7172 		NBUMPSIDED(fin->fin_out, ns_divert_exist);
7173 		return -1;
7174 	}
7175 
7176 	nat->nat_nsrcaddr = frnat.fin_saddr;
7177 	nat->nat_ndstaddr = frnat.fin_daddr;
7178 	if ((nat->nat_flags & IPN_TCPUDP) != 0) {
7179 		nat->nat_nsport = htons(frnat.fin_data[0]);
7180 		nat->nat_ndport = htons(frnat.fin_data[1]);
7181 	} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
7182 		nat->nat_nicmpid = frnat.fin_data[1];
7183 	}
7184 
7185 	nat->nat_pr[fin->fin_out] = fin->fin_p;
7186 	nat->nat_pr[1 - fin->fin_out] = p;
7187 
7188 	if (np->in_redir & NAT_REDIRECT)
7189 		nat->nat_dir = NAT_DIVERTIN;
7190 	else
7191 		nat->nat_dir = NAT_DIVERTOUT;
7192 
7193 	return 0;
7194 }
7195 
7196 
7197 /* ------------------------------------------------------------------------ */
7198 /* Function:    nat_builddivertmp                                           */
7199 /* Returns:     int - -1 == error, 0 == success                             */
7200 /* Parameters:  softn(I) - pointer to NAT context structure                 */
7201 /*              np(I)    - pointer to a NAT rule                            */
7202 /*                                                                          */
7203 /* For divert rules, a skeleton packet representing what will be prepended  */
7204 /* to the real packet is created.  Even though we don't have the full       */
7205 /* packet here, a checksum is calculated that we update later when we       */
7206 /* fill in the final details.  At present a 0 checksum for UDP is being set */
7207 /* here because it is expected that divert will be used for localhost.      */
7208 /* ------------------------------------------------------------------------ */
7209 static int
7210 ipf_nat_builddivertmp(ipf_nat_softc_t *softn, ipnat_t *np)
7211 {
7212 	udphdr_t *uh;
7213 	size_t len;
7214 	ip_t *ip;
7215 
7216 	if ((np->in_redir & NAT_DIVERTUDP) != 0)
7217 		len = sizeof(ip_t) + sizeof(udphdr_t);
7218 	else
7219 		len = sizeof(ip_t);
7220 
7221 	ALLOC_MB_T(np->in_divmp, len);
7222 	if (np->in_divmp == NULL) {
7223 		NBUMPD(ipf_nat_stats, ns_divert_build);
7224 		return -1;
7225 	}
7226 
7227 	/*
7228 	 * First, the header to get the packet diverted to the new destination
7229 	 */
7230 	ip = MTOD(np->in_divmp, ip_t *);
7231 	IP_V_A(ip, 4);
7232 	IP_HL_A(ip, 5);
7233 	ip->ip_tos = 0;
7234 	if ((np->in_redir & NAT_DIVERTUDP) != 0)
7235 		ip->ip_p = IPPROTO_UDP;
7236 	else
7237 		ip->ip_p = IPPROTO_IPIP;
7238 	ip->ip_ttl = 255;
7239 	ip->ip_off = 0;
7240 	ip->ip_sum = 0;
7241 	ip->ip_len = htons(len);
7242 	ip->ip_id = 0;
7243 	ip->ip_src.s_addr = htonl(np->in_snip);
7244 	ip->ip_dst.s_addr = htonl(np->in_dnip);
7245 	ip->ip_sum = ipf_cksum((u_short *)ip, sizeof(*ip));
7246 
7247 	if (np->in_redir & NAT_DIVERTUDP) {
7248 		uh = (udphdr_t *)(ip + 1);
7249 		uh->uh_sum = 0;
7250 		uh->uh_ulen = 8;
7251 		uh->uh_sport = htons(np->in_spnext);
7252 		uh->uh_dport = htons(np->in_dpnext);
7253 	}
7254 
7255 	return 0;
7256 }
7257 
7258 
7259 #define	MINDECAP	(sizeof(ip_t) + sizeof(udphdr_t) + sizeof(ip_t))
7260 
7261 /* ------------------------------------------------------------------------ */
7262 /* Function:    nat_decap                                                   */
7263 /* Returns:     int - -1 == error, 0 == success                             */
7264 /* Parameters:  fin(I) - pointer to packet information                      */
7265 /*              nat(I) - pointer to current NAT session                     */
7266 /*                                                                          */
7267 /* This function is responsible for undoing a packet's encapsulation in the */
7268 /* reverse of an encap/divert rule.  After removing the outer encapsulation */
7269 /* it is necessary to call ipf_makefrip() again so that the contents of 'fin'*/
7270 /* match the "new" packet as it may still be used by IPFilter elsewhere.    */
7271 /* We use "dir" here as the basis for some of the expectations about the    */
7272 /* outer header.  If we return an error, the goal is to leave the original  */
7273 /* packet information undisturbed - this falls short at the end where we'd  */
7274 /* need to back a backup copy of "fin" - expensive.                         */
7275 /* ------------------------------------------------------------------------ */
7276 static int
7277 ipf_nat_decap(fr_info_t *fin, nat_t *nat)
7278 {
7279 	ipf_main_softc_t *softc = fin->fin_main_soft;
7280 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7281 	char *hdr;
7282 	int hlen;
7283 	int skip;
7284 	mb_t *m;
7285 
7286 	if ((fin->fin_flx & FI_ICMPERR) != 0) {
7287 		/*
7288 		 * ICMP packets don't get decapsulated, instead what we need
7289 		 * to do is change the ICMP reply from including (in the data
7290 		 * portion for errors) the encapsulated packet that we sent
7291 		 * out to something that resembles the original packet prior
7292 		 * to encapsulation.  This isn't done here - all we're doing
7293 		 * here is changing the outer address to ensure that it gets
7294 		 * targetted back to the correct system.
7295 		 */
7296 
7297 		if (nat->nat_dir & NAT_OUTBOUND) {
7298 			u_32_t sum1, sum2, sumd;
7299 
7300 			sum1 = ntohl(fin->fin_daddr);
7301 			sum2 = ntohl(nat->nat_osrcaddr);
7302 			CALC_SUMD(sum1, sum2, sumd);
7303 			fin->fin_ip->ip_dst = nat->nat_osrcip;
7304 			fin->fin_daddr = nat->nat_osrcaddr;
7305 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
7306      defined(__osf__) || defined(linux)
7307 			ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, sumd, 0);
7308 #endif
7309 		}
7310 		return 0;
7311 	}
7312 
7313 	m = fin->fin_m;
7314 	skip = fin->fin_hlen;
7315 
7316 	switch (nat->nat_dir)
7317 	{
7318 	case NAT_DIVERTIN :
7319 	case NAT_DIVERTOUT :
7320 		if (fin->fin_plen < MINDECAP)
7321 			return -1;
7322 		skip += sizeof(udphdr_t);
7323 		break;
7324 
7325 	case NAT_ENCAPIN :
7326 	case NAT_ENCAPOUT :
7327 		if (fin->fin_plen < (skip + sizeof(ip_t)))
7328 			return -1;
7329 		break;
7330 	default :
7331 		return -1;
7332 		/* NOTREACHED */
7333 	}
7334 
7335 	/*
7336 	 * The aim here is to keep the original packet details in "fin" for
7337 	 * as long as possible so that returning with an error is for the
7338 	 * original packet and there is little undoing work to do.
7339 	 */
7340 	if (M_LEN(m) < skip + sizeof(ip_t)) {
7341 		if (ipf_pr_pullup(fin, skip + sizeof(ip_t)) == -1)
7342 			return -1;
7343 	}
7344 
7345 	hdr = MTOD(fin->fin_m, char *);
7346 	fin->fin_ip = (ip_t *)(hdr + skip);
7347 	hlen = IP_HL(fin->fin_ip) << 2;
7348 
7349 	if (ipf_pr_pullup(fin, skip + hlen) == -1) {
7350 		NBUMPSIDED(fin->fin_out, ns_decap_pullup);
7351 		return -1;
7352 	}
7353 
7354 	fin->fin_hlen = hlen;
7355 	fin->fin_dlen -= skip;
7356 	fin->fin_plen -= skip;
7357 	fin->fin_ipoff += skip;
7358 
7359 	if (ipf_makefrip(hlen, (ip_t *)hdr, fin) == -1) {
7360 		NBUMPSIDED(fin->fin_out, ns_decap_bad);
7361 		return -1;
7362 	}
7363 
7364 	return skip;
7365 }
7366 
7367 
7368 /* ------------------------------------------------------------------------ */
7369 /* Function:    nat_nextaddr                                                */
7370 /* Returns:     int - -1 == bad input (no new address),                     */
7371 /*                     0 == success and dst has new address                 */
7372 /* Parameters:  fin(I) - pointer to packet information                      */
7373 /*              na(I)  - how to generate new address                        */
7374 /*              old(I) - original address being replaced                    */
7375 /*              dst(O) - where to put the new address                       */
7376 /* Write Lock:  ipf_nat                                                     */
7377 /*                                                                          */
7378 /* This function uses the contents of the "na" structure, in combination    */
7379 /* with "old" to produce a new address to store in "dst".  Not all of the   */
7380 /* possible uses of "na" will result in a new address.                      */
7381 /* ------------------------------------------------------------------------ */
7382 static int
7383 ipf_nat_nextaddr(fr_info_t *fin, nat_addr_t *na, u_32_t *old, u_32_t *dst)
7384 {
7385 	ipf_main_softc_t *softc = fin->fin_main_soft;
7386 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7387 	u_32_t amin, amax, new;
7388 	i6addr_t newip;
7389 	int error;
7390 
7391 	new = 0;
7392 	amin = na->na_addr[0].in4.s_addr;
7393 
7394 	switch (na->na_atype)
7395 	{
7396 	case FRI_RANGE :
7397 		amax = na->na_addr[1].in4.s_addr;
7398 		break;
7399 
7400 	case FRI_NETMASKED :
7401 	case FRI_DYNAMIC :
7402 	case FRI_NORMAL :
7403 		/*
7404 		 * Compute the maximum address by adding the inverse of the
7405 		 * netmask to the minimum address.
7406 		 */
7407 		amax = ~na->na_addr[1].in4.s_addr;
7408 		amax |= amin;
7409 		break;
7410 
7411 	case FRI_LOOKUP :
7412 		break;
7413 
7414 	case FRI_BROADCAST :
7415 	case FRI_PEERADDR :
7416 	case FRI_NETWORK :
7417 	default :
7418 		return -1;
7419 	}
7420 
7421 	error = -1;
7422 
7423 	if (na->na_atype == FRI_LOOKUP) {
7424 		if (na->na_type == IPLT_DSTLIST) {
7425 			error = ipf_dstlist_select_node(fin, na->na_ptr, dst,
7426 							NULL);
7427 		} else {
7428 			NBUMPSIDE(fin->fin_out, ns_badnextaddr);
7429 		}
7430 
7431 	} else if (na->na_atype == IPLT_NONE) {
7432 		/*
7433 		 * 0/0 as the new address means leave it alone.
7434 		 */
7435 		if (na->na_addr[0].in4.s_addr == 0 &&
7436 		    na->na_addr[1].in4.s_addr == 0) {
7437 			new = *old;
7438 
7439 		/*
7440 		 * 0/32 means get the interface's address
7441 		 */
7442 		} else if (na->na_addr[0].in4.s_addr == 0 &&
7443 			   na->na_addr[1].in4.s_addr == 0xffffffff) {
7444 			if (ipf_ifpaddr(softc, 4, na->na_atype,
7445 					fin->fin_ifp, &newip, NULL) == -1) {
7446 				NBUMPSIDED(fin->fin_out, ns_ifpaddrfail);
7447 				return -1;
7448 			}
7449 			new = newip.in4.s_addr;
7450 		} else {
7451 			new = htonl(na->na_nextip);
7452 		}
7453 		*dst = new;
7454 		error = 0;
7455 
7456 	} else {
7457 		NBUMPSIDE(fin->fin_out, ns_badnextaddr);
7458 	}
7459 
7460 	return error;
7461 }
7462 
7463 
7464 /* ------------------------------------------------------------------------ */
7465 /* Function:    nat_nextaddrinit                                            */
7466 /* Returns:     int - 0 == success, else error number                       */
7467 /* Parameters:  softc(I) - pointer to soft context main structure           */
7468 /*              na(I)      - NAT address information for generating new addr*/
7469 /*              initial(I) - flag indicating if it is the first call for    */
7470 /*                           this "na" structure.                           */
7471 /*              ifp(I)     - network interface to derive address            */
7472 /*                           information from.                              */
7473 /*                                                                          */
7474 /* This function is expected to be called in two scenarious: when a new NAT */
7475 /* rule is loaded into the kernel and when the list of NAT rules is sync'd  */
7476 /* up with the valid network interfaces (possibly due to them changing.)    */
7477 /* To distinguish between these, the "initial" parameter is used.  If it is */
7478 /* 1 then this indicates the rule has just been reloaded and 0 for when we  */
7479 /* are updating information.  This difference is important because in       */
7480 /* instances where we are not updating address information associated with  */
7481 /* a network interface, we don't want to disturb what the "next" address to */
7482 /* come out of ipf_nat_nextaddr() will be.                                  */
7483 /* ------------------------------------------------------------------------ */
7484 static int
7485 ipf_nat_nextaddrinit(ipf_main_softc_t *softc, char *base, nat_addr_t *na,
7486     int initial, void *ifp)
7487 {
7488 
7489 	switch (na->na_atype)
7490 	{
7491 	case FRI_LOOKUP :
7492 		if (na->na_subtype == 0) {
7493 			na->na_ptr = ipf_lookup_res_num(softc, IPL_LOGNAT,
7494 							na->na_type,
7495 							na->na_num,
7496 							&na->na_func);
7497 		} else if (na->na_subtype == 1) {
7498 			na->na_ptr = ipf_lookup_res_name(softc, IPL_LOGNAT,
7499 							 na->na_type,
7500 							 base + na->na_num,
7501 							 &na->na_func);
7502 		}
7503 		if (na->na_func == NULL) {
7504 			IPFERROR(60060);
7505 			return ESRCH;
7506 		}
7507 		if (na->na_ptr == NULL) {
7508 			IPFERROR(60056);
7509 			return ESRCH;
7510 		}
7511 		break;
7512 
7513 	case FRI_DYNAMIC :
7514 	case FRI_BROADCAST :
7515 	case FRI_NETWORK :
7516 	case FRI_NETMASKED :
7517 	case FRI_PEERADDR :
7518 		if (ifp != NULL)
7519 			(void )ipf_ifpaddr(softc, 4, na->na_atype, ifp,
7520 					   &na->na_addr[0], &na->na_addr[1]);
7521 		break;
7522 
7523 	case FRI_SPLIT :
7524 	case FRI_RANGE :
7525 		if (initial)
7526 			na->na_nextip = ntohl(na->na_addr[0].in4.s_addr);
7527 		break;
7528 
7529 	case FRI_NONE :
7530 		na->na_addr[0].in4.s_addr &= na->na_addr[1].in4.s_addr;
7531 		return 0;
7532 
7533 	case FRI_NORMAL :
7534 		na->na_addr[0].in4.s_addr &= na->na_addr[1].in4.s_addr;
7535 		break;
7536 
7537 	default :
7538 		IPFERROR(60054);
7539 		return EINVAL;
7540 	}
7541 
7542 	if (initial && (na->na_atype == FRI_NORMAL)) {
7543 		if (na->na_addr[0].in4.s_addr == 0) {
7544 			if ((na->na_addr[1].in4.s_addr == 0xffffffff) ||
7545 			    (na->na_addr[1].in4.s_addr == 0)) {
7546 				return 0;
7547 			}
7548 		}
7549 
7550 		if (na->na_addr[1].in4.s_addr == 0xffffffff) {
7551 			na->na_nextip = ntohl(na->na_addr[0].in4.s_addr);
7552 		} else {
7553 			na->na_nextip = ntohl(na->na_addr[0].in4.s_addr) + 1;
7554 		}
7555 	}
7556 
7557 	return 0;
7558 }
7559 
7560 
7561 /* ------------------------------------------------------------------------ */
7562 /* Function:    ipf_nat_matchflush                                          */
7563 /* Returns:     int - -1 == error, 0 == success                             */
7564 /* Parameters:  softc(I) - pointer to soft context main structure           */
7565 /*              softn(I) - pointer to NAT context structure                 */
7566 /*              nat(I)   - pointer to current NAT session                   */
7567 /*                                                                          */
7568 /* ------------------------------------------------------------------------ */
7569 static int
7570 ipf_nat_matchflush(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, void *data)
7571 {
7572 	int *array, flushed, error;
7573 	nat_t *nat, *natnext;
7574 	ipfobj_t obj;
7575 
7576 	error = ipf_matcharray_load(softc, data, &obj, &array);
7577 	if (error != 0)
7578 		return error;
7579 
7580 	flushed = 0;
7581 
7582 	for (nat = softn->ipf_nat_instances; nat != NULL; nat = natnext) {
7583 		natnext = nat->nat_next;
7584 		if (ipf_nat_matcharray(nat, array, softc->ipf_ticks) == 0) {
7585 			ipf_nat_delete(softc, nat, NL_FLUSH);
7586 			flushed++;
7587 		}
7588 	}
7589 
7590 	obj.ipfo_retval = flushed;
7591 	error = BCOPYOUT(&obj, data, sizeof(obj));
7592 
7593 	KFREES(array, array[0] * sizeof(*array));
7594 
7595 	return error;
7596 }
7597 
7598 
7599 /* ------------------------------------------------------------------------ */
7600 /* Function:    ipf_nat_matcharray                                          */
7601 /* Returns:     int - -1 == error, 0 == success                             */
7602 /* Parameters:  fin(I) - pointer to packet information                      */
7603 /*              nat(I) - pointer to current NAT session                     */
7604 /*                                                                          */
7605 /* ------------------------------------------------------------------------ */
7606 static int
7607 ipf_nat_matcharray(nat_t *nat, int *array, u_long ticks)
7608 {
7609 	int i, n, *x, e, p;
7610 
7611 	e = 0;
7612 	n = array[0];
7613 	x = array + 1;
7614 
7615 	for (; n > 0; x += 3 + x[2]) {
7616 		if (x[0] == IPF_EXP_END)
7617 			break;
7618 		e = 0;
7619 
7620 		n -= x[2] + 3;
7621 		if (n < 0)
7622 			break;
7623 
7624 		p = x[0] >> 16;
7625 		if (p != 0 && p != nat->nat_pr[1])
7626 			break;
7627 
7628 		switch (x[0])
7629 		{
7630 		case IPF_EXP_IP_PR :
7631 			for (i = 0; !e && i < x[2]; i++) {
7632 				e |= (nat->nat_pr[1] == x[i + 3]);
7633 			}
7634 			break;
7635 
7636 		case IPF_EXP_IP_SRCADDR :
7637 			if (nat->nat_v[0] == 4) {
7638 				for (i = 0; !e && i < x[2]; i++) {
7639 					e |= ((nat->nat_osrcaddr & x[i + 4]) ==
7640 					      x[i + 3]);
7641 				}
7642 			}
7643 			if (nat->nat_v[1] == 4) {
7644 				for (i = 0; !e && i < x[2]; i++) {
7645 					e |= ((nat->nat_nsrcaddr & x[i + 4]) ==
7646 					      x[i + 3]);
7647 				}
7648 			}
7649 			break;
7650 
7651 		case IPF_EXP_IP_DSTADDR :
7652 			if (nat->nat_v[0] == 4) {
7653 				for (i = 0; !e && i < x[2]; i++) {
7654 					e |= ((nat->nat_odstaddr & x[i + 4]) ==
7655 					      x[i + 3]);
7656 				}
7657 			}
7658 			if (nat->nat_v[1] == 4) {
7659 				for (i = 0; !e && i < x[2]; i++) {
7660 					e |= ((nat->nat_ndstaddr & x[i + 4]) ==
7661 					      x[i + 3]);
7662 				}
7663 			}
7664 			break;
7665 
7666 		case IPF_EXP_IP_ADDR :
7667 			for (i = 0; !e && i < x[2]; i++) {
7668 				if (nat->nat_v[0] == 4) {
7669 					e |= ((nat->nat_osrcaddr & x[i + 4]) ==
7670 					      x[i + 3]);
7671 				}
7672 				if (nat->nat_v[1] == 4) {
7673 					e |= ((nat->nat_nsrcaddr & x[i + 4]) ==
7674 					      x[i + 3]);
7675 				}
7676 				if (nat->nat_v[0] == 4) {
7677 					e |= ((nat->nat_odstaddr & x[i + 4]) ==
7678 					      x[i + 3]);
7679 				}
7680 				if (nat->nat_v[1] == 4) {
7681 					e |= ((nat->nat_ndstaddr & x[i + 4]) ==
7682 					      x[i + 3]);
7683 				}
7684 			}
7685 			break;
7686 
7687 #ifdef USE_INET6
7688 		case IPF_EXP_IP6_SRCADDR :
7689 			if (nat->nat_v[0] == 6) {
7690 				for (i = 0; !e && i < x[3]; i++) {
7691 					e |= IP6_MASKEQ(&nat->nat_osrc6,
7692 							x + i + 7, x + i + 3);
7693 				}
7694 			}
7695 			if (nat->nat_v[1] == 6) {
7696 				for (i = 0; !e && i < x[3]; i++) {
7697 					e |= IP6_MASKEQ(&nat->nat_nsrc6,
7698 							x + i + 7, x + i + 3);
7699 				}
7700 			}
7701 			break;
7702 
7703 		case IPF_EXP_IP6_DSTADDR :
7704 			if (nat->nat_v[0] == 6) {
7705 				for (i = 0; !e && i < x[3]; i++) {
7706 					e |= IP6_MASKEQ(&nat->nat_odst6,
7707 							x + i + 7,
7708 							x + i + 3);
7709 				}
7710 			}
7711 			if (nat->nat_v[1] == 6) {
7712 				for (i = 0; !e && i < x[3]; i++) {
7713 					e |= IP6_MASKEQ(&nat->nat_ndst6,
7714 							x + i + 7,
7715 							x + i + 3);
7716 				}
7717 			}
7718 			break;
7719 
7720 		case IPF_EXP_IP6_ADDR :
7721 			for (i = 0; !e && i < x[3]; i++) {
7722 				if (nat->nat_v[0] == 6) {
7723 					e |= IP6_MASKEQ(&nat->nat_osrc6,
7724 							x + i + 7,
7725 							x + i + 3);
7726 				}
7727 				if (nat->nat_v[0] == 6) {
7728 					e |= IP6_MASKEQ(&nat->nat_odst6,
7729 							x + i + 7,
7730 							x + i + 3);
7731 				}
7732 				if (nat->nat_v[1] == 6) {
7733 					e |= IP6_MASKEQ(&nat->nat_nsrc6,
7734 							x + i + 7,
7735 							x + i + 3);
7736 				}
7737 				if (nat->nat_v[1] == 6) {
7738 					e |= IP6_MASKEQ(&nat->nat_ndst6,
7739 							x + i + 7,
7740 							x + i + 3);
7741 				}
7742 			}
7743 			break;
7744 #endif
7745 
7746 		case IPF_EXP_UDP_PORT :
7747 		case IPF_EXP_TCP_PORT :
7748 			for (i = 0; !e && i < x[2]; i++) {
7749 				e |= (nat->nat_nsport == x[i + 3]) ||
7750 				     (nat->nat_ndport == x[i + 3]);
7751 			}
7752 			break;
7753 
7754 		case IPF_EXP_UDP_SPORT :
7755 		case IPF_EXP_TCP_SPORT :
7756 			for (i = 0; !e && i < x[2]; i++) {
7757 				e |= (nat->nat_nsport == x[i + 3]);
7758 			}
7759 			break;
7760 
7761 		case IPF_EXP_UDP_DPORT :
7762 		case IPF_EXP_TCP_DPORT :
7763 			for (i = 0; !e && i < x[2]; i++) {
7764 				e |= (nat->nat_ndport == x[i + 3]);
7765 			}
7766 			break;
7767 
7768 		case IPF_EXP_TCP_STATE :
7769 			for (i = 0; !e && i < x[2]; i++) {
7770 				e |= (nat->nat_tcpstate[0] == x[i + 3]) ||
7771 				     (nat->nat_tcpstate[1] == x[i + 3]);
7772 			}
7773 			break;
7774 
7775 		case IPF_EXP_IDLE_GT :
7776 			e |= (ticks - nat->nat_touched > x[3]);
7777 			break;
7778 		}
7779 		e ^= x[1];
7780 
7781 		if (!e)
7782 			break;
7783 	}
7784 
7785 	return e;
7786 }
7787 
7788 
7789 /* ------------------------------------------------------------------------ */
7790 /* Function:    ipf_nat_gettable                                            */
7791 /* Returns:     int     - 0 = success, else error                           */
7792 /* Parameters:  softc(I) - pointer to soft context main structure           */
7793 /*              softn(I) - pointer to NAT context structure                 */
7794 /*              data(I)  - pointer to ioctl data                            */
7795 /*                                                                          */
7796 /* This function handles ioctl requests for tables of nat information.      */
7797 /* At present the only table it deals with is the hash bucket statistics.   */
7798 /* ------------------------------------------------------------------------ */
7799 static int
7800 ipf_nat_gettable(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, char *data)
7801 {
7802 	ipftable_t table;
7803 	int error;
7804 
7805 	error = ipf_inobj(softc, data, NULL, &table, IPFOBJ_GTABLE);
7806 	if (error != 0)
7807 		return error;
7808 
7809 	switch (table.ita_type)
7810 	{
7811 	case IPFTABLE_BUCKETS_NATIN :
7812 		error = COPYOUT(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
7813 				table.ita_table,
7814 				softn->ipf_nat_table_sz * sizeof(u_int));
7815 		break;
7816 
7817 	case IPFTABLE_BUCKETS_NATOUT :
7818 		error = COPYOUT(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
7819 				table.ita_table,
7820 				softn->ipf_nat_table_sz * sizeof(u_int));
7821 		break;
7822 
7823 	default :
7824 		IPFERROR(60058);
7825 		return EINVAL;
7826 	}
7827 
7828 	if (error != 0) {
7829 		IPFERROR(60059);
7830 		error = EFAULT;
7831 	}
7832 	return error;
7833 }
7834 
7835 
7836 /* ------------------------------------------------------------------------ */
7837 /* Function:    ipf_nat_settimeout                                          */
7838 /* Returns:     int  - 0 = success, else failure			    */
7839 /* Parameters:  softc(I) - pointer to soft context main structure           */
7840 /*              t(I) - pointer to tunable                                   */
7841 /*              p(I) - pointer to new tuning data                           */
7842 /*                                                                          */
7843 /* Apply the timeout change to the NAT timeout queues.                      */
7844 /* ------------------------------------------------------------------------ */
7845 int
7846 ipf_nat_settimeout(struct ipf_main_softc_s *softc, ipftuneable_t *t,
7847     ipftuneval_t *p)
7848 {
7849 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7850 
7851 	if (!strncmp(t->ipft_name, "tcp_", 4))
7852 		return ipf_settimeout_tcp(t, p, softn->ipf_nat_tcptq);
7853 
7854 	if (!strcmp(t->ipft_name, "udp_timeout")) {
7855 		ipf_apply_timeout(&softn->ipf_nat_udptq, p->ipftu_int);
7856 	} else if (!strcmp(t->ipft_name, "udp_ack_timeout")) {
7857 		ipf_apply_timeout(&softn->ipf_nat_udpacktq, p->ipftu_int);
7858 	} else if (!strcmp(t->ipft_name, "icmp_timeout")) {
7859 		ipf_apply_timeout(&softn->ipf_nat_icmptq, p->ipftu_int);
7860 	} else if (!strcmp(t->ipft_name, "icmp_ack_timeout")) {
7861 		ipf_apply_timeout(&softn->ipf_nat_icmpacktq, p->ipftu_int);
7862 	} else if (!strcmp(t->ipft_name, "ip_timeout")) {
7863 		ipf_apply_timeout(&softn->ipf_nat_iptq, p->ipftu_int);
7864 	} else {
7865 		IPFERROR(60062);
7866 		return ESRCH;
7867 	}
7868 	return 0;
7869 }
7870 
7871 
7872 /* ------------------------------------------------------------------------ */
7873 /* Function:    ipf_nat_rehash                                              */
7874 /* Returns:     int  - 0 = success, else failure			    */
7875 /* Parameters:  softc(I) - pointer to soft context main structure           */
7876 /*              t(I) - pointer to tunable                                   */
7877 /*              p(I) - pointer to new tuning data                           */
7878 /*                                                                          */
7879 /* To change the size of the basic NAT table, we need to first allocate the */
7880 /* new tables (lest it fails and we've got nowhere to store all of the NAT  */
7881 /* sessions currently active) and then walk through the entire list and     */
7882 /* insert them into the table.  There are two tables here: an inbound one   */
7883 /* and an outbound one.  Each NAT entry goes into each table once.          */
7884 /* ------------------------------------------------------------------------ */
7885 int
7886 ipf_nat_rehash(ipf_main_softc_t *softc, ipftuneable_t *t, ipftuneval_t *p)
7887 {
7888 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7889 	nat_t **newtab[2], *nat, **natp;
7890 	u_int *bucketlens[2];
7891 	u_int maxbucket;
7892 	u_int newsize;
7893 	int error;
7894 	u_int hv;
7895 	int i;
7896 
7897 	newsize = p->ipftu_int;
7898 	/*
7899 	 * In case there is nothing to do...
7900 	 */
7901 	if (newsize == softn->ipf_nat_table_sz)
7902 		return 0;
7903 
7904 	newtab[0] = NULL;
7905 	newtab[1] = NULL;
7906 	bucketlens[0] = NULL;
7907 	bucketlens[1] = NULL;
7908 	/*
7909 	 * 4 tables depend on the NAT table size: the inbound looking table,
7910 	 * the outbound lookup table and the hash chain length for each.
7911 	 */
7912 	KMALLOCS(newtab[0], nat_t **, newsize * sizeof(nat_t *));
7913 	if (newtab == NULL) {
7914 		error = 60063;
7915 		goto badrehash;
7916 	}
7917 
7918 	KMALLOCS(newtab[1], nat_t **, newsize * sizeof(nat_t *));
7919 	if (newtab == NULL) {
7920 		error = 60064;
7921 		goto badrehash;
7922 	}
7923 
7924 	KMALLOCS(bucketlens[0], u_int *, newsize * sizeof(u_int));
7925 	if (bucketlens[0] == NULL) {
7926 		error = 60065;
7927 		goto badrehash;
7928 	}
7929 
7930 	KMALLOCS(bucketlens[1], u_int *, newsize * sizeof(u_int));
7931 	if (bucketlens[1] == NULL) {
7932 		error = 60066;
7933 		goto badrehash;
7934 	}
7935 
7936 	/*
7937 	 * Recalculate the maximum length based on the new size.
7938 	 */
7939 	for (maxbucket = 0, i = newsize; i > 0; i >>= 1)
7940 		maxbucket++;
7941 	maxbucket *= 2;
7942 
7943 	bzero((char *)newtab[0], newsize * sizeof(nat_t *));
7944 	bzero((char *)newtab[1], newsize * sizeof(nat_t *));
7945 	bzero((char *)bucketlens[0], newsize * sizeof(u_int));
7946 	bzero((char *)bucketlens[1], newsize * sizeof(u_int));
7947 
7948 	WRITE_ENTER(&softc->ipf_nat);
7949 
7950 	if (softn->ipf_nat_table[0] != NULL) {
7951 		KFREES(softn->ipf_nat_table[0],
7952 		       softn->ipf_nat_table_sz *
7953 		       sizeof(*softn->ipf_nat_table[0]));
7954 	}
7955 	softn->ipf_nat_table[0] = newtab[0];
7956 
7957 	if (softn->ipf_nat_table[1] != NULL) {
7958 		KFREES(softn->ipf_nat_table[1],
7959 		       softn->ipf_nat_table_sz *
7960 		       sizeof(*softn->ipf_nat_table[1]));
7961 	}
7962 	softn->ipf_nat_table[1] = newtab[1];
7963 
7964 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen != NULL) {
7965 		KFREES(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
7966 		       softn->ipf_nat_table_sz * sizeof(u_int));
7967 	}
7968 	softn->ipf_nat_stats.ns_side[0].ns_bucketlen = bucketlens[0];
7969 
7970 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen != NULL) {
7971 		KFREES(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
7972 		       softn->ipf_nat_table_sz * sizeof(u_int));
7973 	}
7974 	softn->ipf_nat_stats.ns_side[1].ns_bucketlen = bucketlens[1];
7975 
7976 	softn->ipf_nat_maxbucket = maxbucket;
7977 	softn->ipf_nat_table_sz = newsize;
7978 	/*
7979 	 * Walk through the entire list of NAT table entries and put them
7980 	 * in the new NAT table, somewhere.  Because we have a new table,
7981 	 * we need to restart the counter of how many chains are in use.
7982 	 */
7983 	softn->ipf_nat_stats.ns_side[0].ns_inuse = 0;
7984 	softn->ipf_nat_stats.ns_side[1].ns_inuse = 0;
7985 
7986 	for (nat = softn->ipf_nat_instances; nat != NULL; nat = nat->nat_next) {
7987 		nat->nat_hnext[0] = NULL;
7988 		nat->nat_phnext[0] = NULL;
7989 		hv = nat->nat_hv[0] % softn->ipf_nat_table_sz;
7990 
7991 		natp = &softn->ipf_nat_table[0][hv];
7992 		if (*natp) {
7993 			(*natp)->nat_phnext[0] = &nat->nat_hnext[0];
7994 		} else {
7995 			NBUMPSIDE(0, ns_inuse);
7996 		}
7997 		nat->nat_phnext[0] = natp;
7998 		nat->nat_hnext[0] = *natp;
7999 		*natp = nat;
8000 		NBUMPSIDE(0, ns_bucketlen[hv]);
8001 
8002 		nat->nat_hnext[1] = NULL;
8003 		nat->nat_phnext[1] = NULL;
8004 		hv = nat->nat_hv[1] % softn->ipf_nat_table_sz;
8005 
8006 		natp = &softn->ipf_nat_table[1][hv];
8007 		if (*natp) {
8008 			(*natp)->nat_phnext[1] = &nat->nat_hnext[1];
8009 		} else {
8010 			NBUMPSIDE(1, ns_inuse);
8011 		}
8012 		nat->nat_phnext[1] = natp;
8013 		nat->nat_hnext[1] = *natp;
8014 		*natp = nat;
8015 		NBUMPSIDE(1, ns_bucketlen[hv]);
8016 	}
8017 	RWLOCK_EXIT(&softc->ipf_nat);
8018 
8019 	return 0;
8020 
8021 badrehash:
8022 	if (bucketlens[1] != NULL) {
8023 		KFREES(bucketlens[0], newsize * sizeof(u_int));
8024 	}
8025 	if (bucketlens[0] != NULL) {
8026 		KFREES(bucketlens[0], newsize * sizeof(u_int));
8027 	}
8028 	if (newtab[0] != NULL) {
8029 		KFREES(newtab[0], newsize * sizeof(nat_t *));
8030 	}
8031 	if (newtab[1] != NULL) {
8032 		KFREES(newtab[1], newsize * sizeof(nat_t *));
8033 	}
8034 	IPFERROR(error);
8035 	return ENOMEM;
8036 }
8037 
8038 
8039 /* ------------------------------------------------------------------------ */
8040 /* Function:    ipf_nat_rehash_rules                                        */
8041 /* Returns:     int  - 0 = success, else failure			    */
8042 /* Parameters:  softc(I) - pointer to soft context main structure           */
8043 /*              t(I) - pointer to tunable                                   */
8044 /*              p(I) - pointer to new tuning data                           */
8045 /*                                                                          */
8046 /* All of the NAT rules hang off of a hash table that is searched with a    */
8047 /* hash on address after the netmask is applied.  There is a different table*/
8048 /* for both inbound rules (rdr) and outbound (map.)  The resizing will only */
8049 /* affect one of these two tables.                                          */
8050 /* ------------------------------------------------------------------------ */
8051 int
8052 ipf_nat_rehash_rules(ipf_main_softc_t *softc, ipftuneable_t *t, ipftuneval_t *p)
8053 {
8054 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8055 	ipnat_t **newtab, *np, ***old, **npp;
8056 	u_int newsize;
8057 	u_int mask;
8058 	u_int hv;
8059 
8060 	newsize = p->ipftu_int;
8061 	/*
8062 	 * In case there is nothing to do...
8063 	 */
8064 	if (newsize == *t->ipft_pint)
8065 		return 0;
8066 
8067 	/*
8068 	 * All inbound rules have the NAT_REDIRECT bit set in in_redir and
8069 	 * all outbound rules have either NAT_MAP or MAT_MAPBLK set.
8070 	 * This if statement allows for some more generic code to be below,
8071 	 * rather than two huge gobs of code that almost do the same thing.
8072 	 */
8073 	if (t->ipft_pint == &softn->ipf_nat_rdrrules_sz) {
8074 		old = &softn->ipf_nat_rdr_rules;
8075 		mask = NAT_REDIRECT;
8076 	} else {
8077 		old = &softn->ipf_nat_map_rules;
8078 		mask = NAT_MAP|NAT_MAPBLK;
8079 	}
8080 
8081 	KMALLOCS(newtab, ipnat_t **, newsize * sizeof(ipnat_t *));
8082 	if (newtab == NULL) {
8083 		IPFERROR(60067);
8084 		return ENOMEM;
8085 	}
8086 
8087 	bzero((char *)newtab, newsize * sizeof(ipnat_t *));
8088 
8089 	WRITE_ENTER(&softc->ipf_nat);
8090 
8091 	if (*old != NULL) {
8092 		KFREES(*old, *t->ipft_pint * sizeof(ipnat_t **));
8093 	}
8094 	*old = newtab;
8095 	*t->ipft_pint = newsize;
8096 
8097 	for (np = softn->ipf_nat_list; np != NULL; np = np->in_next) {
8098 		if ((np->in_redir & mask) == 0)
8099 			continue;
8100 
8101 		if (np->in_redir & NAT_REDIRECT) {
8102 			np->in_rnext = NULL;
8103 			hv = np->in_hv[0] % newsize;
8104 			for (npp = newtab + hv; *npp != NULL; )
8105 				npp = &(*npp)->in_rnext;
8106 			np->in_prnext = npp;
8107 			*npp = np;
8108 		}
8109 		if (np->in_redir & NAT_MAP) {
8110 			np->in_mnext = NULL;
8111 			hv = np->in_hv[1] % newsize;
8112 			for (npp = newtab + hv; *npp != NULL; )
8113 				npp = &(*npp)->in_mnext;
8114 			np->in_pmnext = npp;
8115 			*npp = np;
8116 		}
8117 
8118 	}
8119 	RWLOCK_EXIT(&softc->ipf_nat);
8120 
8121 	return 0;
8122 }
8123 
8124 
8125 /* ------------------------------------------------------------------------ */
8126 /* Function:    ipf_nat_hostmap_rehash                                      */
8127 /* Returns:     int  - 0 = success, else failure			    */
8128 /* Parameters:  softc(I) - pointer to soft context main structure           */
8129 /*              t(I) - pointer to tunable                                   */
8130 /*              p(I) - pointer to new tuning data                           */
8131 /*                                                                          */
8132 /* Allocate and populate a new hash table that will contain a reference to  */
8133 /* all of the active IP# translations currently in place.                   */
8134 /* ------------------------------------------------------------------------ */
8135 int
8136 ipf_nat_hostmap_rehash(ipf_main_softc_t *softc, ipftuneable_t *t,
8137     ipftuneval_t *p)
8138 {
8139 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8140 	hostmap_t *hm, **newtab;
8141 	u_int newsize;
8142 	u_int hv;
8143 
8144 	newsize = p->ipftu_int;
8145 	/*
8146 	 * In case there is nothing to do...
8147 	 */
8148 	if (newsize == *t->ipft_pint)
8149 		return 0;
8150 
8151 	KMALLOCS(newtab, hostmap_t **, newsize * sizeof(hostmap_t *));
8152 	if (newtab == NULL) {
8153 		IPFERROR(60068);
8154 		return ENOMEM;
8155 	}
8156 
8157 	bzero((char *)newtab, newsize * sizeof(hostmap_t *));
8158 
8159 	WRITE_ENTER(&softc->ipf_nat);
8160 	if (softn->ipf_hm_maptable != NULL) {
8161 		KFREES(softn->ipf_hm_maptable,
8162 		       softn->ipf_nat_hostmap_sz * sizeof(hostmap_t *));
8163 	}
8164 	softn->ipf_hm_maptable = newtab;
8165 	softn->ipf_nat_hostmap_sz = newsize;
8166 
8167 	for (hm = softn->ipf_hm_maplist; hm != NULL; hm = hm->hm_next) {
8168 		hv = hm->hm_hv % softn->ipf_nat_hostmap_sz;
8169 		hm->hm_hnext = softn->ipf_hm_maptable[hv];
8170 		hm->hm_phnext = softn->ipf_hm_maptable + hv;
8171 		if (softn->ipf_hm_maptable[hv] != NULL)
8172 			softn->ipf_hm_maptable[hv]->hm_phnext = &hm->hm_hnext;
8173 		softn->ipf_hm_maptable[hv] = hm;
8174 	}
8175 	RWLOCK_EXIT(&softc->ipf_nat);
8176 
8177 	return 0;
8178 }
8179 
8180 
8181 /* ------------------------------------------------------------------------ */
8182 /* Function:    ipf_nat_add_tq                                              */
8183 /* Parameters:  softc(I) - pointer to soft context main structure           */
8184 /*                                                                          */
8185 /* ------------------------------------------------------------------------ */
8186 ipftq_t *
8187 ipf_nat_add_tq(ipf_main_softc_t *softc, int ttl)
8188 {
8189 	ipf_nat_softc_t *softs = softc->ipf_nat_soft;
8190 
8191 	return ipf_addtimeoutqueue(softc, &softs->ipf_nat_utqe, ttl);
8192 }
8193 
8194 /* ------------------------------------------------------------------------ */
8195 /* Function:    ipf_nat_uncreate                                            */
8196 /* Returns:     Nil                                                         */
8197 /* Parameters:  fin(I) - pointer to packet information                      */
8198 /*                                                                          */
8199 /* This function is used to remove a NAT entry from the NAT table when we   */
8200 /* decide that the create was actually in error. It is thus assumed that    */
8201 /* fin_flx will have both FI_NATED and FI_NATNEW set. Because we're dealing */
8202 /* with the translated packet (not the original), we have to reverse the    */
8203 /* lookup. Although doing the lookup is expensive (relatively speaking), it */
8204 /* is not anticipated that this will be a frequent occurance for normal     */
8205 /* traffic patterns.                                                        */
8206 /* ------------------------------------------------------------------------ */
8207 void
8208 ipf_nat_uncreate(fr_info_t *fin)
8209 {
8210 	ipf_main_softc_t *softc = fin->fin_main_soft;
8211 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8212 	int nflags;
8213 	nat_t *nat;
8214 
8215 	switch (fin->fin_p)
8216 	{
8217 	case IPPROTO_TCP :
8218 		nflags = IPN_TCP;
8219 		break;
8220 	case IPPROTO_UDP :
8221 		nflags = IPN_UDP;
8222 		break;
8223 	default :
8224 		nflags = 0;
8225 		break;
8226 	}
8227 
8228 	WRITE_ENTER(&softc->ipf_nat);
8229 
8230 	if (fin->fin_out == 0) {
8231 		nat = ipf_nat_outlookup(fin, nflags, (u_int)fin->fin_p,
8232 					fin->fin_dst, fin->fin_src);
8233 	} else {
8234 		nat = ipf_nat_inlookup(fin, nflags, (u_int)fin->fin_p,
8235 				       fin->fin_src, fin->fin_dst);
8236 	}
8237 
8238 	if (nat != NULL) {
8239 		NBUMPSIDE(fin->fin_out, ns_uncreate[0]);
8240 		ipf_nat_delete(softc, nat, NL_DESTROY);
8241 	} else {
8242 		NBUMPSIDE(fin->fin_out, ns_uncreate[1]);
8243 	}
8244 
8245 	RWLOCK_EXIT(&softc->ipf_nat);
8246 }
8247 
8248 
8249 /* ------------------------------------------------------------------------ */
8250 /* Function:    ipf_nat_cmp_rules                                           */
8251 /* Returns:     int   - 0 == success, else rules do not match.              */
8252 /* Parameters:  n1(I) - first rule to compare                               */
8253 /*              n2(I) - first rule to compare                               */
8254 /*                                                                          */
8255 /* Compare two rules using pointers to each rule. A straight bcmp will not  */
8256 /* work as some fields (such as in_dst, in_pkts) actually do change once    */
8257 /* the rule has been loaded into the kernel. Whilst this function returns   */
8258 /* various non-zero returns, they're strictly to aid in debugging. Use of   */
8259 /* this function should simply care if the result is zero or not.           */
8260 /* ------------------------------------------------------------------------ */
8261 static int
8262 ipf_nat_cmp_rules(ipnat_t *n1, ipnat_t *n2)
8263 {
8264 	if (n1->in_size != n2->in_size)
8265 		return 1;
8266 
8267 	if (bcmp((char *)&n1->in_v, (char *)&n2->in_v,
8268 		 offsetof(ipnat_t, in_ndst) - offsetof(ipnat_t, in_v)) != 0)
8269 		return 2;
8270 
8271 	if (bcmp((char *)&n1->in_tuc, (char *)&n2->in_tuc,
8272 		 n1->in_size - offsetof(ipnat_t, in_tuc)) != 0)
8273 		return 3;
8274 	if (n1->in_ndst.na_atype != n2->in_ndst.na_atype)
8275 		return 5;
8276 	if (n1->in_ndst.na_function != n2->in_ndst.na_function)
8277 		return 6;
8278 	if (bcmp((char *)&n1->in_ndst.na_addr, (char *)&n2->in_ndst.na_addr,
8279 		 sizeof(n1->in_ndst.na_addr)))
8280 		return 7;
8281 	if (n1->in_nsrc.na_atype != n2->in_nsrc.na_atype)
8282 		return 8;
8283 	if (n1->in_nsrc.na_function != n2->in_nsrc.na_function)
8284 		return 9;
8285 	if (bcmp((char *)&n1->in_nsrc.na_addr, (char *)&n2->in_nsrc.na_addr,
8286 		 sizeof(n1->in_nsrc.na_addr)))
8287 		return 10;
8288 	if (n1->in_odst.na_atype != n2->in_odst.na_atype)
8289 		return 11;
8290 	if (n1->in_odst.na_function != n2->in_odst.na_function)
8291 		return 12;
8292 	if (bcmp((char *)&n1->in_odst.na_addr, (char *)&n2->in_odst.na_addr,
8293 		 sizeof(n1->in_odst.na_addr)))
8294 		return 13;
8295 	if (n1->in_osrc.na_atype != n2->in_osrc.na_atype)
8296 		return 14;
8297 	if (n1->in_osrc.na_function != n2->in_osrc.na_function)
8298 		return 15;
8299 	if (bcmp((char *)&n1->in_osrc.na_addr, (char *)&n2->in_osrc.na_addr,
8300 		 sizeof(n1->in_osrc.na_addr)))
8301 		return 16;
8302 	return 0;
8303 }
8304 
8305 
8306 /* ------------------------------------------------------------------------ */
8307 /* Function:    ipf_nat_rule_init                                           */
8308 /* Returns:     int   - 0 == success, else rules do not match.              */
8309 /* Parameters:  softc(I) - pointer to soft context main structure           */
8310 /*              softn(I) - pointer to NAT context structure                 */
8311 /*              n(I)     - first rule to compare                            */
8312 /*                                                                          */
8313 /* ------------------------------------------------------------------------ */
8314 static int
8315 ipf_nat_rule_init(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *n)
8316 {
8317 	int error = 0;
8318 
8319 	if ((n->in_flags & IPN_SIPRANGE) != 0)
8320 		n->in_nsrcatype = FRI_RANGE;
8321 
8322 	if ((n->in_flags & IPN_DIPRANGE) != 0)
8323 		n->in_ndstatype = FRI_RANGE;
8324 
8325 	if ((n->in_flags & IPN_SPLIT) != 0)
8326 		n->in_ndstatype = FRI_SPLIT;
8327 
8328 	if ((n->in_redir & (NAT_MAP|NAT_REWRITE|NAT_DIVERTUDP)) != 0)
8329 		n->in_spnext = n->in_spmin;
8330 
8331 	if ((n->in_redir & (NAT_REWRITE|NAT_DIVERTUDP)) != 0) {
8332 		n->in_dpnext = n->in_dpmin;
8333 	} else if (n->in_redir == NAT_REDIRECT) {
8334 		n->in_dpnext = n->in_dpmin;
8335 	}
8336 
8337 	n->in_stepnext = 0;
8338 
8339 	switch (n->in_v[0])
8340 	{
8341 	case 4 :
8342 		error = ipf_nat_ruleaddrinit(softc, softn, n);
8343 		if (error != 0)
8344 			return error;
8345 		break;
8346 #ifdef USE_INET6
8347 	case 6 :
8348 		error = ipf_nat6_ruleaddrinit(softc, softn, n);
8349 		if (error != 0)
8350 			return error;
8351 		break;
8352 #endif
8353 	default :
8354 		break;
8355 	}
8356 
8357 	if (n->in_redir == (NAT_DIVERTUDP|NAT_MAP)) {
8358 		/*
8359 		 * Prerecord whether or not the destination of the divert
8360 		 * is local or not to the interface the packet is going
8361 		 * to be sent out.
8362 		 */
8363 		n->in_dlocal = ipf_deliverlocal(softc, n->in_v[1],
8364 						n->in_ifps[1], &n->in_ndstip6);
8365 	}
8366 
8367 	return error;
8368 }
8369 
8370 
8371 /* ------------------------------------------------------------------------ */
8372 /* Function:    ipf_nat_rule_fini                                           */
8373 /* Returns:     int   - 0 == success, else rules do not match.              */
8374 /* Parameters:  softc(I) - pointer to soft context main structure           */
8375 /*              n(I)     - rule to work on                                  */
8376 /*                                                                          */
8377 /* This function is used to release any objects that were referenced during */
8378 /* the rule initialisation. This is useful both when free'ing the rule and  */
8379 /* when handling ioctls that need to initialise these fields but not        */
8380 /* actually use them after the ioctl processing has finished.               */
8381 /* ------------------------------------------------------------------------ */
8382 static void
8383 ipf_nat_rule_fini(ipf_main_softc_t *softc, ipnat_t *n)
8384 {
8385 	if (n->in_odst.na_atype == FRI_LOOKUP && n->in_odst.na_ptr != NULL)
8386 		ipf_lookup_deref(softc, n->in_odst.na_type, n->in_odst.na_ptr);
8387 
8388 	if (n->in_osrc.na_atype == FRI_LOOKUP && n->in_osrc.na_ptr != NULL)
8389 		ipf_lookup_deref(softc, n->in_osrc.na_type, n->in_osrc.na_ptr);
8390 
8391 	if (n->in_ndst.na_atype == FRI_LOOKUP && n->in_ndst.na_ptr != NULL)
8392 		ipf_lookup_deref(softc, n->in_ndst.na_type, n->in_ndst.na_ptr);
8393 
8394 	if (n->in_nsrc.na_atype == FRI_LOOKUP && n->in_nsrc.na_ptr != NULL)
8395 		ipf_lookup_deref(softc, n->in_nsrc.na_type, n->in_nsrc.na_ptr);
8396 
8397 	if (n->in_divmp != NULL)
8398 		FREE_MB_T(n->in_divmp);
8399 }
8400