xref: /netbsd-src/sys/external/bsd/ipf/netinet/ip_nat.c (revision bdc22b2e01993381dcefeff2bc9b56ca75a4235c)
1 /*	$NetBSD: ip_nat.c,v 1.20 2018/06/03 10:37:23 maxv 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/ipl.h"
91 #include "netinet/ip_fil.h"
92 #include "netinet/ip_nat.h"
93 #include "netinet/ip_frag.h"
94 #include "netinet/ip_state.h"
95 #include "netinet/ip_proxy.h"
96 #include "netinet/ip_lookup.h"
97 #include "netinet/ip_dstlist.h"
98 #include "netinet/ip_sync.h"
99 #if FREEBSD_GE_REV(300000)
100 # include <sys/malloc.h>
101 #endif
102 #ifdef HAS_SYS_MD5_H
103 # include <sys/md5.h>
104 #else
105 # include "md5.h"
106 #endif
107 /* END OF INCLUDES */
108 
109 #undef	SOCKADDR_IN
110 #define	SOCKADDR_IN	struct sockaddr_in
111 
112 #if !defined(lint)
113 #if defined(__NetBSD__)
114 #include <sys/cdefs.h>
115 __KERNEL_RCSID(0, "$NetBSD: ip_nat.c,v 1.20 2018/06/03 10:37:23 maxv Exp $");
116 #else
117 static const char sccsid[] = "@(#)ip_nat.c	1.11 6/5/96 (C) 1995 Darren Reed";
118 static const char rcsid[] = "@(#)Id: ip_nat.c,v 1.1.1.2 2012/07/22 13:45:27 darrenr Exp";
119 #endif
120 #endif
121 
122 
123 #define	NATFSUM(n,v,f)	((v) == 4 ? (n)->f.in4.s_addr : (n)->f.i6[0] + \
124 			 (n)->f.i6[1] + (n)->f.i6[2] + (n)->f.i6[3])
125 #define	NBUMP(x)	softn->(x)++
126 #define	NBUMPD(x, y)	do { \
127 				softn->x.y++; \
128 				DT(y); \
129 			} while (0)
130 #define	NBUMPSIDE(y,x)	softn->ipf_nat_stats.ns_side[y].x++
131 #define	NBUMPSIDED(y,x)	do { softn->ipf_nat_stats.ns_side[y].x++; \
132 			     DT(x); } while (0)
133 #define	NBUMPSIDEX(y,x,z) \
134 			do { softn->ipf_nat_stats.ns_side[y].x++; \
135 			     DT(z); } while (0)
136 #define	NBUMPSIDEDF(y,x)do { softn->ipf_nat_stats.ns_side[y].x++; \
137 			     DT1(x, fr_info_t *, fin); } while (0)
138 
139 frentry_t	ipfnatblock;
140 
141 static const ipftuneable_t ipf_nat_tuneables[] = {
142 	/* nat */
143 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_lock) },
144 		"nat_lock",	0,	1,
145 		stsizeof(ipf_nat_softc_t, ipf_nat_lock),
146 		IPFT_RDONLY,		NULL,	NULL },
147 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_sz) },
148 		"nat_table_size", 1,	0x7fffffff,
149 		stsizeof(ipf_nat_softc_t, ipf_nat_table_sz),
150 		0,			NULL,	ipf_nat_rehash },
151 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_max) },
152 		"nat_table_max", 1,	0x7fffffff,
153 		stsizeof(ipf_nat_softc_t, ipf_nat_table_max),
154 		0,			NULL,	NULL },
155 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_maprules_sz) },
156 		"nat_rules_size", 1,	0x7fffffff,
157 		stsizeof(ipf_nat_softc_t, ipf_nat_maprules_sz),
158 		0,			NULL,	ipf_nat_rehash_rules },
159 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_rdrrules_sz) },
160 		"rdr_rules_size", 1,	0x7fffffff,
161 		stsizeof(ipf_nat_softc_t, ipf_nat_rdrrules_sz),
162 		0,			NULL,	ipf_nat_rehash_rules },
163 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_hostmap_sz) },
164 		"hostmap_size",	1,	0x7fffffff,
165 		stsizeof(ipf_nat_softc_t, ipf_nat_hostmap_sz),
166 		0,			NULL,	ipf_nat_hostmap_rehash },
167 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_maxbucket) },
168 		"nat_maxbucket",1,	0x7fffffff,
169 		stsizeof(ipf_nat_softc_t, ipf_nat_maxbucket),
170 		0,			NULL,	NULL },
171 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_logging) },
172 		"nat_logging",	0,	1,
173 		stsizeof(ipf_nat_softc_t, ipf_nat_logging),
174 		0,			NULL,	NULL },
175 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_doflush) },
176 		"nat_doflush",	0,	1,
177 		stsizeof(ipf_nat_softc_t, ipf_nat_doflush),
178 		0,			NULL,	NULL },
179 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_wm_low) },
180 		"nat_table_wm_low",	1,	99,
181 		stsizeof(ipf_nat_softc_t, ipf_nat_table_wm_low),
182 		0,			NULL,	NULL },
183 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_wm_high) },
184 		"nat_table_wm_high",	2,	100,
185 		stsizeof(ipf_nat_softc_t, ipf_nat_table_wm_high),
186 		0,			NULL,	NULL },
187 	{ { 0 },
188 		NULL,			0,	0,
189 		0,
190 		0,			NULL,	NULL }
191 };
192 
193 /* ======================================================================== */
194 /* How the NAT is organised and works.                                      */
195 /*                                                                          */
196 /* Inside (interface y) NAT       Outside (interface x)                     */
197 /* -------------------- -+- -------------------------------------           */
198 /* Packet going          |   out, processsed by ipf_nat_checkout() for x    */
199 /* ------------>         |   ------------>                                  */
200 /* src=10.1.1.1          |   src=192.1.1.1                                  */
201 /*                       |                                                  */
202 /*                       |   in, processed by ipf_nat_checkin() for x       */
203 /* <------------         |   <------------                                  */
204 /* dst=10.1.1.1          |   dst=192.1.1.1                                  */
205 /* -------------------- -+- -------------------------------------           */
206 /* ipf_nat_checkout() - changes ip_src and if required, sport               */
207 /*             - creates a new mapping, if required.                        */
208 /* ipf_nat_checkin()  - changes ip_dst and if required, dport               */
209 /*                                                                          */
210 /* In the NAT table, internal source is recorded as "in" and externally     */
211 /* seen as "out".                                                           */
212 /* ======================================================================== */
213 
214 
215 #if SOLARIS && !defined(INSTANCES)
216 extern	int		pfil_delayed_copy;
217 #endif
218 
219 static	int	ipf_nat_flush_entry(ipf_main_softc_t *, void *);
220 static	int	ipf_nat_getent(ipf_main_softc_t *, void *, int);
221 static	int	ipf_nat_getsz(ipf_main_softc_t *, void *, int);
222 static	int	ipf_nat_putent(ipf_main_softc_t *, void *, int);
223 static	void	ipf_nat_addmap(ipf_nat_softc_t *, ipnat_t *);
224 static	void	ipf_nat_addrdr(ipf_nat_softc_t *, ipnat_t *);
225 static	int	ipf_nat_builddivertmp(ipf_nat_softc_t *, ipnat_t *);
226 static	int	ipf_nat_clearlist(ipf_main_softc_t *, ipf_nat_softc_t *);
227 static	int	ipf_nat_cmp_rules(ipnat_t *, ipnat_t *);
228 static	int	ipf_nat_decap(fr_info_t *, nat_t *);
229 static	void	ipf_nat_delrule(ipf_main_softc_t *, ipf_nat_softc_t *,
230 				     ipnat_t *, int);
231 static	int	ipf_nat_extraflush(ipf_main_softc_t *, ipf_nat_softc_t *, int);
232 static	int	ipf_nat_finalise(fr_info_t *, nat_t *);
233 static	int	ipf_nat_flushtable(ipf_main_softc_t *, ipf_nat_softc_t *);
234 static	int	ipf_nat_getnext(ipf_main_softc_t *, ipftoken_t *,
235 				ipfgeniter_t *, ipfobj_t *);
236 static	int	ipf_nat_gettable(ipf_main_softc_t *, ipf_nat_softc_t *, char *);
237 static	hostmap_t *ipf_nat_hostmap(ipf_nat_softc_t *, ipnat_t *,
238 				   struct in_addr, struct in_addr,
239 				   struct in_addr, u_32_t);
240 static	int	ipf_nat_icmpquerytype(int);
241 static	int	ipf_nat_iterator(ipf_main_softc_t *, ipftoken_t *,
242 				 ipfgeniter_t *, ipfobj_t *);
243 static	int	ipf_nat_match(fr_info_t *, ipnat_t *);
244 static	int	ipf_nat_matcharray(nat_t *, int *, u_long);
245 static	int	ipf_nat_matchflush(ipf_main_softc_t *, ipf_nat_softc_t *,
246 				   void *);
247 static	void	ipf_nat_mssclamp(tcphdr_t *, u_32_t, fr_info_t *, u_short *);
248 static	int	ipf_nat_newmap(fr_info_t *, nat_t *, natinfo_t *);
249 static	int	ipf_nat_newdivert(fr_info_t *, nat_t *, natinfo_t *);
250 static	int	ipf_nat_newrdr(fr_info_t *, nat_t *, natinfo_t *);
251 static	int	ipf_nat_newrewrite(fr_info_t *, nat_t *, natinfo_t *);
252 static	int	ipf_nat_nextaddr(fr_info_t *, nat_addr_t *, u_32_t *, u_32_t *);
253 static	int	ipf_nat_nextaddrinit(ipf_main_softc_t *, char *,
254 				     nat_addr_t *, int, void *);
255 static	int	ipf_nat_resolverule(ipf_main_softc_t *, ipnat_t *);
256 static	int	ipf_nat_ruleaddrinit(ipf_main_softc_t *,
257 				     ipf_nat_softc_t *, ipnat_t *);
258 static	void	ipf_nat_rule_fini(ipf_main_softc_t *, ipnat_t *);
259 static	int	ipf_nat_rule_init(ipf_main_softc_t *, ipf_nat_softc_t *,
260 				       ipnat_t *);
261 static	int	ipf_nat_siocaddnat(ipf_main_softc_t *, ipf_nat_softc_t *,
262 					ipnat_t *, int);
263 static	void	ipf_nat_siocdelnat(ipf_main_softc_t *, ipf_nat_softc_t *,
264 					ipnat_t *, int);
265 static	void	ipf_nat_tabmove(ipf_nat_softc_t *, nat_t *);
266 
267 /* ------------------------------------------------------------------------ */
268 /* Function:    ipf_nat_main_load                                           */
269 /* Returns:     int - 0 == success, -1 == failure                           */
270 /* Parameters:  Nil                                                         */
271 /*                                                                          */
272 /* The only global NAT structure that needs to be initialised is the filter */
273 /* rule that is used with blocking packets.                                 */
274 /* ------------------------------------------------------------------------ */
275 int
276 ipf_nat_main_load(void)
277 {
278 	bzero((char *)&ipfnatblock, sizeof(ipfnatblock));
279 	ipfnatblock.fr_flags = FR_BLOCK|FR_QUICK;
280 	ipfnatblock.fr_ref = 1;
281 
282 	return 0;
283 }
284 
285 
286 /* ------------------------------------------------------------------------ */
287 /* Function:    ipf_nat_main_unload                                         */
288 /* Returns:     int - 0 == success, -1 == failure                           */
289 /* Parameters:  Nil                                                         */
290 /*                                                                          */
291 /* A null-op function that exists as a placeholder so that the flow in      */
292 /* other functions is obvious.                                              */
293 /* ------------------------------------------------------------------------ */
294 int
295 ipf_nat_main_unload(void)
296 {
297 	return 0;
298 }
299 
300 
301 /* ------------------------------------------------------------------------ */
302 /* Function:    ipf_nat_soft_create                                         */
303 /* Returns:     void * - NULL = failure, else pointer to NAT context        */
304 /* Parameters:  softc(I) - pointer to soft context main structure           */
305 /*                                                                          */
306 /* Allocate the initial soft context structure for NAT and populate it with */
307 /* some default values. Creating the tables is left until we call _init so  */
308 /* that sizes can be changed before we get under way.                       */
309 /* ------------------------------------------------------------------------ */
310 void *
311 ipf_nat_soft_create(ipf_main_softc_t *softc)
312 {
313 	ipf_nat_softc_t *softn;
314 
315 	KMALLOC(softn, ipf_nat_softc_t *);
316 	if (softn == NULL)
317 		return NULL;
318 
319 	bzero((char *)softn, sizeof(*softn));
320 
321 	softn->ipf_nat_tune = ipf_tune_array_copy(softn,
322 						  sizeof(ipf_nat_tuneables),
323 						  ipf_nat_tuneables);
324 	if (softn->ipf_nat_tune == NULL) {
325 		ipf_nat_soft_destroy(softc, softn);
326 		return NULL;
327 	}
328 	if (ipf_tune_array_link(softc, softn->ipf_nat_tune) == -1) {
329 		ipf_nat_soft_destroy(softc, softn);
330 		return NULL;
331 	}
332 
333 	softn->ipf_nat_list_tail = &softn->ipf_nat_list;
334 
335 	softn->ipf_nat_table_max = NAT_TABLE_MAX;
336 	softn->ipf_nat_table_sz = NAT_TABLE_SZ;
337 	softn->ipf_nat_maprules_sz = NAT_SIZE;
338 	softn->ipf_nat_rdrrules_sz = RDR_SIZE;
339 	softn->ipf_nat_hostmap_sz = HOSTMAP_SIZE;
340 	softn->ipf_nat_doflush = 0;
341 #ifdef  IPFILTER_LOG
342 	softn->ipf_nat_logging = 1;
343 #else
344 	softn->ipf_nat_logging = 0;
345 #endif
346 
347 	softn->ipf_nat_defage = DEF_NAT_AGE;
348 	softn->ipf_nat_defipage = IPF_TTLVAL(60);
349 	softn->ipf_nat_deficmpage = IPF_TTLVAL(3);
350 	softn->ipf_nat_table_wm_high = 99;
351 	softn->ipf_nat_table_wm_low = 90;
352 
353 	return softn;
354 }
355 
356 /* ------------------------------------------------------------------------ */
357 /* Function:    ipf_nat_soft_destroy                                        */
358 /* Returns:     Nil                                                         */
359 /* Parameters:  softc(I) - pointer to soft context main structure           */
360 /*                                                                          */
361 /* ------------------------------------------------------------------------ */
362 void
363 ipf_nat_soft_destroy(ipf_main_softc_t *softc, void *arg)
364 {
365 	ipf_nat_softc_t *softn = arg;
366 
367 	if (softn->ipf_nat_tune != NULL) {
368 		ipf_tune_array_unlink(softc, softn->ipf_nat_tune);
369 		KFREES(softn->ipf_nat_tune, sizeof(ipf_nat_tuneables));
370 		softn->ipf_nat_tune = NULL;
371 	}
372 
373 	KFREE(softn);
374 }
375 
376 
377 /* ------------------------------------------------------------------------ */
378 /* Function:    ipf_nat_init                                                */
379 /* Returns:     int - 0 == success, -1 == failure                           */
380 /* Parameters:  softc(I) - pointer to soft context main structure           */
381 /*                                                                          */
382 /* Initialise all of the NAT locks, tables and other structures.            */
383 /* ------------------------------------------------------------------------ */
384 int
385 ipf_nat_soft_init(ipf_main_softc_t *softc, void *arg)
386 {
387 	ipf_nat_softc_t *softn = arg;
388 	ipftq_t *tq;
389 	int i;
390 
391 	KMALLOCS(softn->ipf_nat_table[0], nat_t **, \
392 		 sizeof(nat_t *) * softn->ipf_nat_table_sz);
393 
394 	if (softn->ipf_nat_table[0] != NULL) {
395 		bzero((char *)softn->ipf_nat_table[0],
396 		      softn->ipf_nat_table_sz * sizeof(nat_t *));
397 	} else {
398 		return -1;
399 	}
400 
401 	KMALLOCS(softn->ipf_nat_table[1], nat_t **, \
402 		 sizeof(nat_t *) * softn->ipf_nat_table_sz);
403 
404 	if (softn->ipf_nat_table[1] != NULL) {
405 		bzero((char *)softn->ipf_nat_table[1],
406 		      softn->ipf_nat_table_sz * sizeof(nat_t *));
407 	} else {
408 		return -2;
409 	}
410 
411 	KMALLOCS(softn->ipf_nat_map_rules, ipnat_t **, \
412 		 sizeof(ipnat_t *) * softn->ipf_nat_maprules_sz);
413 
414 	if (softn->ipf_nat_map_rules != NULL) {
415 		bzero((char *)softn->ipf_nat_map_rules,
416 		      softn->ipf_nat_maprules_sz * sizeof(ipnat_t *));
417 	} else {
418 		return -3;
419 	}
420 
421 	KMALLOCS(softn->ipf_nat_rdr_rules, ipnat_t **, \
422 		 sizeof(ipnat_t *) * softn->ipf_nat_rdrrules_sz);
423 
424 	if (softn->ipf_nat_rdr_rules != NULL) {
425 		bzero((char *)softn->ipf_nat_rdr_rules,
426 		      softn->ipf_nat_rdrrules_sz * sizeof(ipnat_t *));
427 	} else {
428 		return -4;
429 	}
430 
431 	KMALLOCS(softn->ipf_hm_maptable, hostmap_t **, \
432 		 sizeof(hostmap_t *) * softn->ipf_nat_hostmap_sz);
433 
434 	if (softn->ipf_hm_maptable != NULL) {
435 		bzero((char *)softn->ipf_hm_maptable,
436 		      sizeof(hostmap_t *) * softn->ipf_nat_hostmap_sz);
437 	} else {
438 		return -5;
439 	}
440 	softn->ipf_hm_maplist = NULL;
441 
442 	KMALLOCS(softn->ipf_nat_stats.ns_side[0].ns_bucketlen, u_int *,
443 		 softn->ipf_nat_table_sz * sizeof(u_int));
444 
445 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen == NULL) {
446 		return -6;
447 	}
448 	bzero((char *)softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
449 	      softn->ipf_nat_table_sz * sizeof(u_int));
450 
451 	KMALLOCS(softn->ipf_nat_stats.ns_side[1].ns_bucketlen, u_int *,
452 		 softn->ipf_nat_table_sz * sizeof(u_int));
453 
454 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen == NULL) {
455 		return -7;
456 	}
457 
458 	bzero((char *)softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
459 	      softn->ipf_nat_table_sz * sizeof(u_int));
460 
461 	if (softn->ipf_nat_maxbucket == 0) {
462 		for (i = softn->ipf_nat_table_sz; i > 0; i >>= 1)
463 			softn->ipf_nat_maxbucket++;
464 		softn->ipf_nat_maxbucket *= 2;
465 	}
466 
467 	ipf_sttab_init(softc, softn->ipf_nat_tcptq);
468 	/*
469 	 * Increase this because we may have "keep state" following this too
470 	 * and packet storms can occur if this is removed too quickly.
471 	 */
472 	softn->ipf_nat_tcptq[IPF_TCPS_CLOSED].ifq_ttl = softc->ipf_tcplastack;
473 	softn->ipf_nat_tcptq[IPF_TCP_NSTATES - 1].ifq_next =
474 							&softn->ipf_nat_udptq;
475 
476 	IPFTQ_INIT(&softn->ipf_nat_udptq, softn->ipf_nat_defage,
477 		   "nat ipftq udp tab");
478 	softn->ipf_nat_udptq.ifq_next = &softn->ipf_nat_udpacktq;
479 
480 	IPFTQ_INIT(&softn->ipf_nat_udpacktq, softn->ipf_nat_defage,
481 		   "nat ipftq udpack tab");
482 	softn->ipf_nat_udpacktq.ifq_next = &softn->ipf_nat_icmptq;
483 
484 	IPFTQ_INIT(&softn->ipf_nat_icmptq, softn->ipf_nat_deficmpage,
485 		   "nat icmp ipftq tab");
486 	softn->ipf_nat_icmptq.ifq_next = &softn->ipf_nat_icmpacktq;
487 
488 	IPFTQ_INIT(&softn->ipf_nat_icmpacktq, softn->ipf_nat_defage,
489 		   "nat icmpack ipftq tab");
490 	softn->ipf_nat_icmpacktq.ifq_next = &softn->ipf_nat_iptq;
491 
492 	IPFTQ_INIT(&softn->ipf_nat_iptq, softn->ipf_nat_defipage,
493 		   "nat ip ipftq tab");
494 	softn->ipf_nat_iptq.ifq_next = &softn->ipf_nat_pending;
495 
496 	IPFTQ_INIT(&softn->ipf_nat_pending, 1, "nat pending ipftq tab");
497 	softn->ipf_nat_pending.ifq_next = NULL;
498 
499 	for (i = 0, tq = softn->ipf_nat_tcptq; i < IPF_TCP_NSTATES; i++, tq++) {
500 		if (tq->ifq_ttl < softn->ipf_nat_deficmpage)
501 			tq->ifq_ttl = softn->ipf_nat_deficmpage;
502 #ifdef LARGE_NAT
503 		else if (tq->ifq_ttl > softn->ipf_nat_defage)
504 			tq->ifq_ttl = softn->ipf_nat_defage;
505 #endif
506 	}
507 
508 	/*
509 	 * Increase this because we may have "keep state" following
510 	 * this too and packet storms can occur if this is removed
511 	 * too quickly.
512 	 */
513 	softn->ipf_nat_tcptq[IPF_TCPS_CLOSED].ifq_ttl = softc->ipf_tcplastack;
514 
515 	MUTEX_INIT(&softn->ipf_nat_new, "ipf nat new mutex");
516 	MUTEX_INIT(&softn->ipf_nat_io, "ipf nat io mutex");
517 
518 	softn->ipf_nat_inited = 1;
519 
520 	return 0;
521 }
522 
523 
524 /* ------------------------------------------------------------------------ */
525 /* Function:    ipf_nat_soft_fini                                           */
526 /* Returns:     Nil                                                         */
527 /* Parameters:  softc(I) - pointer to soft context main structure           */
528 /*                                                                          */
529 /* Free all memory used by NAT structures allocated at runtime.             */
530 /* ------------------------------------------------------------------------ */
531 int
532 ipf_nat_soft_fini(ipf_main_softc_t *softc, void *arg)
533 {
534 	ipf_nat_softc_t *softn = arg;
535 	ipftq_t *ifq, *ifqnext;
536 
537 	(void) ipf_nat_clearlist(softc, softn);
538 	(void) ipf_nat_flushtable(softc, softn);
539 
540 	/*
541 	 * Proxy timeout queues are not cleaned here because although they
542 	 * exist on the NAT list, ipf_proxy_unload is called after unload
543 	 * and the proxies actually are responsible for them being created.
544 	 * Should the proxy timeouts have their own list?  There's no real
545 	 * justification as this is the only complication.
546 	 */
547 	for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifqnext) {
548 		ifqnext = ifq->ifq_next;
549 		if (ipf_deletetimeoutqueue(ifq) == 0)
550 			ipf_freetimeoutqueue(softc, ifq);
551 	}
552 
553 	if (softn->ipf_nat_table[0] != NULL) {
554 		KFREES(softn->ipf_nat_table[0],
555 		       sizeof(nat_t *) * softn->ipf_nat_table_sz);
556 		softn->ipf_nat_table[0] = NULL;
557 	}
558 	if (softn->ipf_nat_table[1] != NULL) {
559 		KFREES(softn->ipf_nat_table[1],
560 		       sizeof(nat_t *) * softn->ipf_nat_table_sz);
561 		softn->ipf_nat_table[1] = NULL;
562 	}
563 	if (softn->ipf_nat_map_rules != NULL) {
564 		KFREES(softn->ipf_nat_map_rules,
565 		       sizeof(ipnat_t *) * softn->ipf_nat_maprules_sz);
566 		softn->ipf_nat_map_rules = NULL;
567 	}
568 	if (softn->ipf_nat_rdr_rules != NULL) {
569 		KFREES(softn->ipf_nat_rdr_rules,
570 		       sizeof(ipnat_t *) * softn->ipf_nat_rdrrules_sz);
571 		softn->ipf_nat_rdr_rules = NULL;
572 	}
573 	if (softn->ipf_hm_maptable != NULL) {
574 		KFREES(softn->ipf_hm_maptable,
575 		       sizeof(hostmap_t *) * softn->ipf_nat_hostmap_sz);
576 		softn->ipf_hm_maptable = NULL;
577 	}
578 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen != NULL) {
579 		KFREES(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
580 		       sizeof(u_int) * softn->ipf_nat_table_sz);
581 		softn->ipf_nat_stats.ns_side[0].ns_bucketlen = NULL;
582 	}
583 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen != NULL) {
584 		KFREES(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
585 		       sizeof(u_int) * softn->ipf_nat_table_sz);
586 		softn->ipf_nat_stats.ns_side[1].ns_bucketlen = NULL;
587 	}
588 
589 	if (softn->ipf_nat_inited == 1) {
590 		softn->ipf_nat_inited = 0;
591 		ipf_sttab_destroy(softn->ipf_nat_tcptq);
592 
593 		MUTEX_DESTROY(&softn->ipf_nat_new);
594 		MUTEX_DESTROY(&softn->ipf_nat_io);
595 
596 		MUTEX_DESTROY(&softn->ipf_nat_udptq.ifq_lock);
597 		MUTEX_DESTROY(&softn->ipf_nat_udpacktq.ifq_lock);
598 		MUTEX_DESTROY(&softn->ipf_nat_icmptq.ifq_lock);
599 		MUTEX_DESTROY(&softn->ipf_nat_icmpacktq.ifq_lock);
600 		MUTEX_DESTROY(&softn->ipf_nat_iptq.ifq_lock);
601 		MUTEX_DESTROY(&softn->ipf_nat_pending.ifq_lock);
602 	}
603 
604 	return 0;
605 }
606 
607 
608 /* ------------------------------------------------------------------------ */
609 /* Function:    ipf_nat_setlock                                             */
610 /* Returns:     Nil                                                         */
611 /* Parameters:  arg(I) - pointer to soft state information                  */
612 /*              tmp(I) - new lock value                                     */
613 /*                                                                          */
614 /* Set the "lock status" of NAT to the value in tmp.                        */
615 /* ------------------------------------------------------------------------ */
616 void
617 ipf_nat_setlock(void *arg, int tmp)
618 {
619 	ipf_nat_softc_t *softn = arg;
620 
621 	softn->ipf_nat_lock = tmp;
622 }
623 
624 
625 /* ------------------------------------------------------------------------ */
626 /* Function:    ipf_nat_addrdr                                              */
627 /* Returns:     Nil                                                         */
628 /* Parameters:  n(I) - pointer to NAT rule to add                           */
629 /*                                                                          */
630 /* Adds a redirect rule to the hash table of redirect rules and the list of */
631 /* loaded NAT rules.  Updates the bitmask indicating which netmasks are in  */
632 /* use by redirect rules.                                                   */
633 /* ------------------------------------------------------------------------ */
634 static void
635 ipf_nat_addrdr(ipf_nat_softc_t *softn, ipnat_t *n)
636 {
637 	ipnat_t **np;
638 	u_32_t j;
639 	u_int hv;
640 	u_int rhv;
641 	int k;
642 
643 	if (n->in_odstatype == FRI_NORMAL) {
644 		k = count4bits(n->in_odstmsk);
645 		ipf_inet_mask_add(k, &softn->ipf_nat_rdr_mask);
646 		j = (n->in_odstaddr & n->in_odstmsk);
647 		rhv = NAT_HASH_FN(j, 0, 0xffffffff);
648 	} else {
649 		ipf_inet_mask_add(0, &softn->ipf_nat_rdr_mask);
650 		j = 0;
651 		rhv = 0;
652 	}
653 	hv = rhv % softn->ipf_nat_rdrrules_sz;
654 	np = softn->ipf_nat_rdr_rules + hv;
655 	while (*np != NULL)
656 		np = &(*np)->in_rnext;
657 	n->in_rnext = NULL;
658 	n->in_prnext = np;
659 	n->in_hv[0] = hv;
660 	n->in_use++;
661 	*np = n;
662 }
663 
664 
665 /* ------------------------------------------------------------------------ */
666 /* Function:    ipf_nat_addmap                                              */
667 /* Returns:     Nil                                                         */
668 /* Parameters:  n(I) - pointer to NAT rule to add                           */
669 /*                                                                          */
670 /* Adds a NAT map rule to the hash table of rules and the list of  loaded   */
671 /* NAT rules.  Updates the bitmask indicating which netmasks are in use by  */
672 /* redirect rules.                                                          */
673 /* ------------------------------------------------------------------------ */
674 static void
675 ipf_nat_addmap(ipf_nat_softc_t *softn, ipnat_t *n)
676 {
677 	ipnat_t **np;
678 	u_32_t j;
679 	u_int hv;
680 	u_int rhv;
681 	int k;
682 
683 	if (n->in_osrcatype == FRI_NORMAL) {
684 		k = count4bits(n->in_osrcmsk);
685 		ipf_inet_mask_add(k, &softn->ipf_nat_map_mask);
686 		j = (n->in_osrcaddr & n->in_osrcmsk);
687 		rhv = NAT_HASH_FN(j, 0, 0xffffffff);
688 	} else {
689 		ipf_inet_mask_add(0, &softn->ipf_nat_map_mask);
690 		j = 0;
691 		rhv = 0;
692 	}
693 	hv = rhv % softn->ipf_nat_maprules_sz;
694 	np = softn->ipf_nat_map_rules + hv;
695 	while (*np != NULL)
696 		np = &(*np)->in_mnext;
697 	n->in_mnext = NULL;
698 	n->in_pmnext = np;
699 	n->in_hv[1] = rhv;
700 	n->in_use++;
701 	*np = n;
702 }
703 
704 
705 /* ------------------------------------------------------------------------ */
706 /* Function:    ipf_nat_delrdr                                              */
707 /* Returns:     Nil                                                         */
708 /* Parameters:  n(I) - pointer to NAT rule to delete                        */
709 /*                                                                          */
710 /* Removes a redirect rule from the hash table of redirect rules.           */
711 /* ------------------------------------------------------------------------ */
712 void
713 ipf_nat_delrdr(ipf_nat_softc_t *softn, ipnat_t *n)
714 {
715 	if (n->in_odstatype == FRI_NORMAL) {
716 		int k = count4bits(n->in_odstmsk);
717 		ipf_inet_mask_del(k, &softn->ipf_nat_rdr_mask);
718 	} else {
719 		ipf_inet_mask_del(0, &softn->ipf_nat_rdr_mask);
720 	}
721 	if (n->in_rnext)
722 		n->in_rnext->in_prnext = n->in_prnext;
723 	*n->in_prnext = n->in_rnext;
724 	n->in_use--;
725 }
726 
727 
728 /* ------------------------------------------------------------------------ */
729 /* Function:    ipf_nat_delmap                                              */
730 /* Returns:     Nil                                                         */
731 /* Parameters:  n(I) - pointer to NAT rule to delete                        */
732 /*                                                                          */
733 /* Removes a NAT map rule from the hash table of NAT map rules.             */
734 /* ------------------------------------------------------------------------ */
735 void
736 ipf_nat_delmap(ipf_nat_softc_t *softn, ipnat_t *n)
737 {
738 	if (n->in_osrcatype == FRI_NORMAL) {
739 		int k = count4bits(n->in_osrcmsk);
740 		ipf_inet_mask_del(k, &softn->ipf_nat_map_mask);
741 	} else {
742 		ipf_inet_mask_del(0, &softn->ipf_nat_map_mask);
743 	}
744 	if (n->in_mnext != NULL)
745 		n->in_mnext->in_pmnext = n->in_pmnext;
746 	*n->in_pmnext = n->in_mnext;
747 	n->in_use--;
748 }
749 
750 
751 /* ------------------------------------------------------------------------ */
752 /* Function:    ipf_nat_hostmap                                             */
753 /* Returns:     struct hostmap* - NULL if no hostmap could be created,      */
754 /*                                else a pointer to the hostmapping to use  */
755 /* Parameters:  np(I)   - pointer to NAT rule                               */
756 /*              real(I) - real IP address                                   */
757 /*              map(I)  - mapped IP address                                 */
758 /*              port(I) - destination port number                           */
759 /* Write Locks: ipf_nat                                                     */
760 /*                                                                          */
761 /* Check if an ip address has already been allocated for a given mapping    */
762 /* that is not doing port based translation.  If is not yet allocated, then */
763 /* create a new entry if a non-NULL NAT rule pointer has been supplied.     */
764 /* ------------------------------------------------------------------------ */
765 static struct hostmap *
766 ipf_nat_hostmap(ipf_nat_softc_t *softn, ipnat_t *np, struct in_addr src,
767     struct in_addr dst, struct in_addr map, u_32_t port)
768 {
769 	hostmap_t *hm;
770 	u_int hv, rhv;
771 
772 	hv = (src.s_addr ^ dst.s_addr);
773 	hv += src.s_addr;
774 	hv += dst.s_addr;
775 	rhv = hv;
776 	hv %= softn->ipf_nat_hostmap_sz;
777 	for (hm = softn->ipf_hm_maptable[hv]; hm; hm = hm->hm_hnext)
778 		if ((hm->hm_osrcip.s_addr == src.s_addr) &&
779 		    (hm->hm_odstip.s_addr == dst.s_addr) &&
780 		    ((np == NULL) || (np == hm->hm_ipnat)) &&
781 		    ((port == 0) || (port == hm->hm_port))) {
782 			softn->ipf_nat_stats.ns_hm_addref++;
783 			hm->hm_ref++;
784 			return hm;
785 		}
786 
787 	if (np == NULL) {
788 		softn->ipf_nat_stats.ns_hm_nullnp++;
789 		return NULL;
790 	}
791 
792 	KMALLOC(hm, hostmap_t *);
793 	if (hm) {
794 		hm->hm_next = softn->ipf_hm_maplist;
795 		hm->hm_pnext = &softn->ipf_hm_maplist;
796 		if (softn->ipf_hm_maplist != NULL)
797 			softn->ipf_hm_maplist->hm_pnext = &hm->hm_next;
798 		softn->ipf_hm_maplist = hm;
799 		hm->hm_hnext = softn->ipf_hm_maptable[hv];
800 		hm->hm_phnext = softn->ipf_hm_maptable + hv;
801 		if (softn->ipf_hm_maptable[hv] != NULL)
802 			softn->ipf_hm_maptable[hv]->hm_phnext = &hm->hm_hnext;
803 		softn->ipf_hm_maptable[hv] = hm;
804 		hm->hm_ipnat = np;
805 		np->in_use++;
806 		hm->hm_osrcip = src;
807 		hm->hm_odstip = dst;
808 		hm->hm_nsrcip = map;
809 		hm->hm_ndstip.s_addr = 0;
810 		hm->hm_ref = 1;
811 		hm->hm_port = port;
812 		hm->hm_hv = rhv;
813 		hm->hm_v = 4;
814 		softn->ipf_nat_stats.ns_hm_new++;
815 	} else {
816 		softn->ipf_nat_stats.ns_hm_newfail++;
817 	}
818 	return hm;
819 }
820 
821 
822 /* ------------------------------------------------------------------------ */
823 /* Function:    ipf_nat_hostmapdel                                          */
824 /* Returns:     Nil                                                         */
825 /* Parameters:  hmp(I) - pointer to hostmap structure pointer               */
826 /* Write Locks: ipf_nat                                                     */
827 /*                                                                          */
828 /* Decrement the references to this hostmap structure by one.  If this      */
829 /* reaches zero then remove it and free it.                                 */
830 /* ------------------------------------------------------------------------ */
831 void
832 ipf_nat_hostmapdel(ipf_main_softc_t *softc, struct hostmap **hmp)
833 {
834 	struct hostmap *hm;
835 
836 	hm = *hmp;
837 	*hmp = NULL;
838 
839 	hm->hm_ref--;
840 	if (hm->hm_ref == 0) {
841 		ipf_nat_rule_deref(softc, &hm->hm_ipnat);
842 		if (hm->hm_hnext)
843 			hm->hm_hnext->hm_phnext = hm->hm_phnext;
844 		*hm->hm_phnext = hm->hm_hnext;
845 		if (hm->hm_next)
846 			hm->hm_next->hm_pnext = hm->hm_pnext;
847 		*hm->hm_pnext = hm->hm_next;
848 		KFREE(hm);
849 	}
850 }
851 
852 
853 /* ------------------------------------------------------------------------ */
854 /* Function:    ipf_fix_outcksum                                            */
855 /* Returns:     Nil                                                         */
856 /* Parameters:  fin(I) - pointer to packet information                      */
857 /*              sp(I)  - location of 16bit checksum to update               */
858 /*              n((I)  - amount to adjust checksum by                       */
859 /*                                                                          */
860 /* Adjusts the 16bit checksum by "n" for packets going out.                 */
861 /* ------------------------------------------------------------------------ */
862 void
863 ipf_fix_outcksum(int cksum, u_short *sp, u_32_t n, u_32_t partial)
864 {
865 	u_short sumshort;
866 	u_32_t sum1;
867 
868 	if (n == 0)
869 		return;
870 
871 	if (cksum == 4) {
872 		*sp = 0;
873 		return;
874 	}
875 	if (cksum == 2) {
876 		sum1 = partial;
877 		sum1 = (sum1 & 0xffff) + (sum1 >> 16);
878 		*sp = htons(sum1);
879 		return;
880 	}
881 	sum1 = (~ntohs(*sp)) & 0xffff;
882 	sum1 += (n);
883 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
884 	/* Again */
885 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
886 	sumshort = ~(u_short)sum1;
887 	*(sp) = htons(sumshort);
888 }
889 
890 
891 /* ------------------------------------------------------------------------ */
892 /* Function:    ipf_fix_incksum                                             */
893 /* Returns:     Nil                                                         */
894 /* Parameters:  fin(I) - pointer to packet information                      */
895 /*              sp(I)  - location of 16bit checksum to update               */
896 /*              n((I)  - amount to adjust checksum by                       */
897 /*                                                                          */
898 /* Adjusts the 16bit checksum by "n" for packets going in.                  */
899 /* ------------------------------------------------------------------------ */
900 void
901 ipf_fix_incksum(int cksum, u_short *sp, u_32_t n, u_32_t partial)
902 {
903 	u_short sumshort;
904 	u_32_t sum1;
905 
906 	if (n == 0)
907 		return;
908 
909 	if (cksum == 4) {
910 		*sp = 0;
911 		return;
912 	}
913 	if (cksum == 2) {
914 		sum1 = partial;
915 		sum1 = (sum1 & 0xffff) + (sum1 >> 16);
916 		*sp = htons(sum1);
917 		return;
918 	}
919 
920 	sum1 = (~ntohs(*sp)) & 0xffff;
921 	sum1 += ~(n) & 0xffff;
922 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
923 	/* Again */
924 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
925 	sumshort = ~(u_short)sum1;
926 	*(sp) = htons(sumshort);
927 }
928 
929 
930 /* ------------------------------------------------------------------------ */
931 /* Function:    ipf_fix_datacksum                                           */
932 /* Returns:     Nil                                                         */
933 /* Parameters:  sp(I)  - location of 16bit checksum to update               */
934 /*              n((I)  - amount to adjust checksum by                       */
935 /*                                                                          */
936 /* Fix_datacksum is used *only* for the adjustments of checksums in the     */
937 /* data section of an IP packet.                                            */
938 /*                                                                          */
939 /* The only situation in which you need to do this is when NAT'ing an       */
940 /* ICMP error message. Such a message, contains in its body the IP header   */
941 /* of the original IP packet, that causes the error.                        */
942 /*                                                                          */
943 /* You can't use fix_incksum or fix_outcksum in that case, because for the  */
944 /* kernel the data section of the ICMP error is just data, and no special   */
945 /* processing like hardware cksum or ntohs processing have been done by the */
946 /* kernel on the data section.                                              */
947 /* ------------------------------------------------------------------------ */
948 void
949 ipf_fix_datacksum(u_short *sp, u_32_t n)
950 {
951 	u_short sumshort;
952 	u_32_t sum1;
953 
954 	if (n == 0)
955 		return;
956 
957 	sum1 = (~ntohs(*sp)) & 0xffff;
958 	sum1 += (n);
959 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
960 	/* Again */
961 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
962 	sumshort = ~(u_short)sum1;
963 	*(sp) = htons(sumshort);
964 }
965 
966 
967 /* ------------------------------------------------------------------------ */
968 /* Function:    ipf_nat_ioctl                                               */
969 /* Returns:     int - 0 == success, != 0 == failure                         */
970 /* Parameters:  softc(I) - pointer to soft context main structure           */
971 /*              data(I)  - pointer to ioctl data                            */
972 /*              cmd(I)   - ioctl command integer                            */
973 /*              mode(I)  - file mode bits used with open                    */
974 /*              uid(I)   - uid of calling process                           */
975 /*              ctx(I)   - pointer used as key for finding context          */
976 /*                                                                          */
977 /* Processes an ioctl call made to operate on the IP Filter NAT device.     */
978 /* ------------------------------------------------------------------------ */
979 int
980 ipf_nat_ioctl(ipf_main_softc_t *softc, void *data, ioctlcmd_t cmd, int mode,
981    int uid, void *ctx)
982 {
983 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
984 	int error = 0, ret, arg, getlock;
985 	ipnat_t *nat, *nt, *n;
986 	ipnat_t natd;
987 	SPL_INT(s);
988 
989 #if BSD_GE_YEAR(199306) && defined(_KERNEL)
990 # if NETBSD_GE_REV(399002000)
991 	if ((mode & FWRITE) &&
992 	     kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_FIREWALL,
993 				     KAUTH_REQ_NETWORK_FIREWALL_FW,
994 				     NULL, NULL, NULL))
995 # else
996 #  if defined(__FreeBSD_version) && (__FreeBSD_version >= 500034)
997 	if (securelevel_ge(curthread->td_ucred, 3) && (mode & FWRITE))
998 #  else
999 	if ((securelevel >= 3) && (mode & FWRITE))
1000 #  endif
1001 # endif
1002 	{
1003 		IPFERROR(60001);
1004 		return EPERM;
1005 	}
1006 #endif
1007 
1008 #if defined(__osf__) && defined(_KERNEL)
1009 	getlock = 0;
1010 #else
1011 	getlock = (mode & NAT_LOCKHELD) ? 0 : 1;
1012 #endif
1013 
1014 	n = NULL;
1015 	nt = NULL;
1016 	nat = NULL;
1017 
1018 	if ((cmd == (ioctlcmd_t)SIOCADNAT) || (cmd == (ioctlcmd_t)SIOCRMNAT) ||
1019 	    (cmd == (ioctlcmd_t)SIOCPURGENAT)) {
1020 		if (mode & NAT_SYSSPACE) {
1021 			bcopy(data, (char *)&natd, sizeof(natd));
1022 			nat = &natd;
1023 			error = 0;
1024 		} else {
1025 			bzero(&natd, sizeof(natd));
1026 			error = ipf_inobj(softc, data, NULL, &natd,
1027 					  IPFOBJ_IPNAT);
1028 			if (error != 0)
1029 				goto done;
1030 
1031 			if (natd.in_size < sizeof(ipnat_t)) {
1032 				error = EINVAL;
1033 				goto done;
1034 			}
1035 			KMALLOCS(nt, ipnat_t *, natd.in_size);
1036 			if (nt == NULL) {
1037 				IPFERROR(60070);
1038 				error = ENOMEM;
1039 				goto done;
1040 			}
1041 			bzero(nt, natd.in_size);
1042 			error = ipf_inobjsz(softc, data, nt, IPFOBJ_IPNAT,
1043 					    natd.in_size);
1044 			if (error)
1045 				goto done;
1046 			nat = nt;
1047 		}
1048 
1049 		/*
1050 		 * For add/delete, look to see if the NAT entry is
1051 		 * already present
1052 		 */
1053 		nat->in_flags &= IPN_USERFLAGS;
1054 		if ((nat->in_redir & NAT_MAPBLK) == 0) {
1055 			if (nat->in_osrcatype == FRI_NORMAL ||
1056 			    nat->in_osrcatype == FRI_NONE)
1057 				nat->in_osrcaddr &= nat->in_osrcmsk;
1058 			if (nat->in_odstatype == FRI_NORMAL ||
1059 			    nat->in_odstatype == FRI_NONE)
1060 				nat->in_odstaddr &= nat->in_odstmsk;
1061 			if ((nat->in_flags & (IPN_SPLIT|IPN_SIPRANGE)) == 0) {
1062 				if (nat->in_nsrcatype == FRI_NORMAL)
1063 					nat->in_nsrcaddr &= nat->in_nsrcmsk;
1064 				if (nat->in_ndstatype == FRI_NORMAL)
1065 					nat->in_ndstaddr &= nat->in_ndstmsk;
1066 			}
1067 		}
1068 
1069 		error = ipf_nat_rule_init(softc, softn, nat);
1070 		if (error != 0)
1071 			goto done;
1072 
1073 		MUTEX_ENTER(&softn->ipf_nat_io);
1074 		for (n = softn->ipf_nat_list; n != NULL; n = n->in_next)
1075 			if (ipf_nat_cmp_rules(nat, n) == 0)
1076 				break;
1077 	}
1078 
1079 	switch (cmd)
1080 	{
1081 #ifdef  IPFILTER_LOG
1082 	case SIOCIPFFB :
1083 	{
1084 		int tmp;
1085 
1086 		if (!(mode & FWRITE)) {
1087 			IPFERROR(60002);
1088 			error = EPERM;
1089 		} else {
1090 			tmp = ipf_log_clear(softc, IPL_LOGNAT);
1091 			error = BCOPYOUT(&tmp, data, sizeof(tmp));
1092 			if (error != 0) {
1093 				IPFERROR(60057);
1094 				error = EFAULT;
1095 			}
1096 		}
1097 		break;
1098 	}
1099 
1100 	case SIOCSETLG :
1101 		if (!(mode & FWRITE)) {
1102 			IPFERROR(60003);
1103 			error = EPERM;
1104 		} else {
1105 			error = BCOPYIN(data, &softn->ipf_nat_logging,
1106 					sizeof(softn->ipf_nat_logging));
1107 			if (error != 0)
1108 				error = EFAULT;
1109 		}
1110 		break;
1111 
1112 	case SIOCGETLG :
1113 		error = BCOPYOUT(&softn->ipf_nat_logging, data,
1114 				 sizeof(softn->ipf_nat_logging));
1115 		if (error != 0) {
1116 			IPFERROR(60004);
1117 			error = EFAULT;
1118 		}
1119 		break;
1120 
1121 	case FIONREAD :
1122 		arg = ipf_log_bytesused(softc, IPL_LOGNAT);
1123 		error = BCOPYOUT(&arg, data, sizeof(arg));
1124 		if (error != 0) {
1125 			IPFERROR(60005);
1126 			error = EFAULT;
1127 		}
1128 		break;
1129 #endif
1130 	case SIOCADNAT :
1131 		if (!(mode & FWRITE)) {
1132 			IPFERROR(60006);
1133 			error = EPERM;
1134 		} else if (n != NULL) {
1135 			natd.in_flineno = n->in_flineno;
1136 			(void) ipf_outobj(softc, data, &natd, IPFOBJ_IPNAT);
1137 			IPFERROR(60007);
1138 			error = EEXIST;
1139 		} else if (nt == NULL) {
1140 			IPFERROR(60008);
1141 			error = ENOMEM;
1142 		}
1143 		if (error != 0) {
1144 			MUTEX_EXIT(&softn->ipf_nat_io);
1145 			break;
1146 		}
1147 		if (nat != nt)
1148 			bcopy((char *)nat, (char *)nt, sizeof(*n));
1149 		error = ipf_nat_siocaddnat(softc, softn, nt, getlock);
1150 		MUTEX_EXIT(&softn->ipf_nat_io);
1151 		if (error == 0) {
1152 			nat = NULL;
1153 			nt = NULL;
1154 		}
1155 		break;
1156 
1157 	case SIOCRMNAT :
1158 	case SIOCPURGENAT :
1159 		if (!(mode & FWRITE)) {
1160 			IPFERROR(60009);
1161 			error = EPERM;
1162 			n = NULL;
1163 		} else if (n == NULL) {
1164 			IPFERROR(60010);
1165 			error = ESRCH;
1166 		}
1167 
1168 		if (error != 0) {
1169 			MUTEX_EXIT(&softn->ipf_nat_io);
1170 			break;
1171 		}
1172 		if (cmd == (ioctlcmd_t)SIOCPURGENAT) {
1173 			error = ipf_outobjsz(softc, data, n, IPFOBJ_IPNAT,
1174 					     n->in_size);
1175 			if (error) {
1176 				MUTEX_EXIT(&softn->ipf_nat_io);
1177 				goto done;
1178 			}
1179 			n->in_flags |= IPN_PURGE;
1180 		}
1181 		ipf_nat_siocdelnat(softc, softn, n, getlock);
1182 
1183 		MUTEX_EXIT(&softn->ipf_nat_io);
1184 		n = NULL;
1185 		break;
1186 
1187 	case SIOCGNATS :
1188 	    {
1189 		natstat_t *nsp = &softn->ipf_nat_stats;
1190 
1191 		nsp->ns_side[0].ns_table = softn->ipf_nat_table[0];
1192 		nsp->ns_side[1].ns_table = softn->ipf_nat_table[1];
1193 		nsp->ns_list = softn->ipf_nat_list;
1194 		nsp->ns_maptable = softn->ipf_hm_maptable;
1195 		nsp->ns_maplist = softn->ipf_hm_maplist;
1196 		nsp->ns_nattab_sz = softn->ipf_nat_table_sz;
1197 		nsp->ns_nattab_max = softn->ipf_nat_table_max;
1198 		nsp->ns_rultab_sz = softn->ipf_nat_maprules_sz;
1199 		nsp->ns_rdrtab_sz = softn->ipf_nat_rdrrules_sz;
1200 		nsp->ns_hostmap_sz = softn->ipf_nat_hostmap_sz;
1201 		nsp->ns_instances = softn->ipf_nat_instances;
1202 		nsp->ns_ticks = softc->ipf_ticks;
1203 #ifdef IPFILTER_LOGGING
1204 		nsp->ns_log_ok = ipf_log_logok(softc, IPF_LOGNAT);
1205 		nsp->ns_log_fail = ipf_log_failures(softc, IPF_LOGNAT);
1206 #else
1207 		nsp->ns_log_ok = 0;
1208 		nsp->ns_log_fail = 0;
1209 #endif
1210 		error = ipf_outobj(softc, data, nsp, IPFOBJ_NATSTAT);
1211 		break;
1212 	    }
1213 
1214 	case SIOCGNATL :
1215 	    {
1216 		natlookup_t nl;
1217 
1218 		error = ipf_inobj(softc, data, NULL, &nl, IPFOBJ_NATLOOKUP);
1219 		if (error == 0) {
1220 			void *ptr;
1221 
1222 			if (getlock) {
1223 				READ_ENTER(&softc->ipf_nat);
1224 			}
1225 
1226 			switch (nl.nl_v)
1227 			{
1228 			case 4 :
1229 				ptr = ipf_nat_lookupredir(softc, &nl);
1230 				break;
1231 #ifdef USE_INET6
1232 			case 6 :
1233 				ptr = ipf_nat6_lookupredir(softc, &nl);
1234 				break;
1235 #endif
1236 			default:
1237 				ptr = NULL;
1238 				break;
1239 			}
1240 
1241 			if (getlock) {
1242 				RWLOCK_EXIT(&softc->ipf_nat);
1243 			}
1244 			if (ptr != NULL) {
1245 				error = ipf_outobj(softc, data, &nl,
1246 						   IPFOBJ_NATLOOKUP);
1247 			} else {
1248 				IPFERROR(60011);
1249 				error = ESRCH;
1250 			}
1251 		}
1252 		break;
1253 	    }
1254 
1255 	case SIOCIPFFL :	/* old SIOCFLNAT & SIOCCNATL */
1256 		if (!(mode & FWRITE)) {
1257 			IPFERROR(60012);
1258 			error = EPERM;
1259 			break;
1260 		}
1261 		if (getlock) {
1262 			WRITE_ENTER(&softc->ipf_nat);
1263 		}
1264 
1265 		error = BCOPYIN(data, &arg, sizeof(arg));
1266 		if (error != 0) {
1267 			IPFERROR(60013);
1268 			error = EFAULT;
1269 		} else {
1270 			if (arg == 0)
1271 				ret = ipf_nat_flushtable(softc, softn);
1272 			else if (arg == 1)
1273 				ret = ipf_nat_clearlist(softc, softn);
1274 			else
1275 				ret = ipf_nat_extraflush(softc, softn, arg);
1276 			ipf_proxy_flush(softc->ipf_proxy_soft, arg);
1277 		}
1278 
1279 		if (getlock) {
1280 			RWLOCK_EXIT(&softc->ipf_nat);
1281 		}
1282 		if (error == 0) {
1283 			error = BCOPYOUT(&ret, data, sizeof(ret));
1284 		}
1285 		break;
1286 
1287 	case SIOCMATCHFLUSH :
1288 		if (!(mode & FWRITE)) {
1289 			IPFERROR(60014);
1290 			error = EPERM;
1291 			break;
1292 		}
1293 		if (getlock) {
1294 			WRITE_ENTER(&softc->ipf_nat);
1295 		}
1296 
1297 		error = ipf_nat_matchflush(softc, softn, data);
1298 
1299 		if (getlock) {
1300 			RWLOCK_EXIT(&softc->ipf_nat);
1301 		}
1302 		break;
1303 
1304 	case SIOCPROXY :
1305 		error = ipf_proxy_ioctl(softc, data, cmd, mode, ctx);
1306 		break;
1307 
1308 	case SIOCSTLCK :
1309 		if (!(mode & FWRITE)) {
1310 			IPFERROR(60015);
1311 			error = EPERM;
1312 		} else {
1313 			error = ipf_lock(data, &softn->ipf_nat_lock);
1314 		}
1315 		break;
1316 
1317 	case SIOCSTPUT :
1318 		if ((mode & FWRITE) != 0) {
1319 			error = ipf_nat_putent(softc, data, getlock);
1320 		} else {
1321 			IPFERROR(60016);
1322 			error = EACCES;
1323 		}
1324 		break;
1325 
1326 	case SIOCSTGSZ :
1327 		if (softn->ipf_nat_lock) {
1328 			error = ipf_nat_getsz(softc, data, getlock);
1329 		} else {
1330 			IPFERROR(60017);
1331 			error = EACCES;
1332 		}
1333 		break;
1334 
1335 	case SIOCSTGET :
1336 		if (softn->ipf_nat_lock) {
1337 			error = ipf_nat_getent(softc, data, getlock);
1338 		} else {
1339 			IPFERROR(60018);
1340 			error = EACCES;
1341 		}
1342 		break;
1343 
1344 	case SIOCGENITER :
1345 	    {
1346 		ipfgeniter_t iter;
1347 		ipftoken_t *token;
1348 		ipfobj_t obj;
1349 
1350 		error = ipf_inobj(softc, data, &obj, &iter, IPFOBJ_GENITER);
1351 		if (error != 0)
1352 			break;
1353 
1354 		SPL_SCHED(s);
1355 		token = ipf_token_find(softc, iter.igi_type, uid, ctx);
1356 		if (token != NULL) {
1357 			error  = ipf_nat_iterator(softc, token, &iter, &obj);
1358 			WRITE_ENTER(&softc->ipf_tokens);
1359 			ipf_token_deref(softc, token);
1360 			RWLOCK_EXIT(&softc->ipf_tokens);
1361 		}
1362 		SPL_X(s);
1363 		break;
1364 	    }
1365 
1366 	case SIOCIPFDELTOK :
1367 		error = BCOPYIN(data, &arg, sizeof(arg));
1368 		if (error == 0) {
1369 			SPL_SCHED(s);
1370 			error = ipf_token_del(softc, arg, uid, ctx);
1371 			SPL_X(s);
1372 		} else {
1373 			IPFERROR(60019);
1374 			error = EFAULT;
1375 		}
1376 		break;
1377 
1378 	case SIOCGTQTAB :
1379 		error = ipf_outobj(softc, data, softn->ipf_nat_tcptq,
1380 				   IPFOBJ_STATETQTAB);
1381 		break;
1382 
1383 	case SIOCGTABL :
1384 		error = ipf_nat_gettable(softc, softn, data);
1385 		break;
1386 
1387 	default :
1388 		IPFERROR(60020);
1389 		error = EINVAL;
1390 		break;
1391 	}
1392 done:
1393 	if (nat != NULL)
1394 		ipf_nat_rule_fini(softc, nat);
1395 	if (nt != NULL)
1396 		KFREES(nt, nt->in_size);
1397 	return error;
1398 }
1399 
1400 
1401 /* ------------------------------------------------------------------------ */
1402 /* Function:    ipf_nat_siocaddnat                                          */
1403 /* Returns:     int - 0 == success, != 0 == failure                         */
1404 /* Parameters:  softc(I) - pointer to soft context main structure           */
1405 /*              softn(I) - pointer to NAT context structure                 */
1406 /*              n(I)       - pointer to new NAT rule                        */
1407 /*              np(I)      - pointer to where to insert new NAT rule        */
1408 /*              getlock(I) - flag indicating if lock on  is held            */
1409 /* Mutex Locks: ipf_nat_io                                                  */
1410 /*                                                                          */
1411 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
1412 /* from information passed to the kernel, then add it  to the appropriate   */
1413 /* NAT rule table(s).                                                       */
1414 /* ------------------------------------------------------------------------ */
1415 static int
1416 ipf_nat_siocaddnat(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *n,
1417 	int getlock)
1418 {
1419 	int error = 0;
1420 
1421 	if (ipf_nat_resolverule(softc, n) != 0) {
1422 		IPFERROR(60022);
1423 		return ENOENT;
1424 	}
1425 
1426 	if ((n->in_age[0] == 0) && (n->in_age[1] != 0)) {
1427 		IPFERROR(60023);
1428 		return EINVAL;
1429 	}
1430 
1431 	if (n->in_redir == (NAT_DIVERTUDP|NAT_MAP)) {
1432 		/*
1433 		 * Prerecord whether or not the destination of the divert
1434 		 * is local or not to the interface the packet is going
1435 		 * to be sent out.
1436 		 */
1437 		n->in_dlocal = ipf_deliverlocal(softc, n->in_v[1],
1438 						n->in_ifps[1], &n->in_ndstip6);
1439 	}
1440 
1441 	if (getlock) {
1442 		WRITE_ENTER(&softc->ipf_nat);
1443 	}
1444 	n->in_next = NULL;
1445 	n->in_pnext = softn->ipf_nat_list_tail;
1446 	*n->in_pnext = n;
1447 	softn->ipf_nat_list_tail = &n->in_next;
1448 	n->in_use++;
1449 
1450 	if (n->in_redir & NAT_REDIRECT) {
1451 		n->in_flags &= ~IPN_NOTDST;
1452 		switch (n->in_v[0])
1453 		{
1454 		case 4 :
1455 			ipf_nat_addrdr(softn, n);
1456 			break;
1457 #ifdef USE_INET6
1458 		case 6 :
1459 			ipf_nat6_addrdr(softn, n);
1460 			break;
1461 #endif
1462 		default :
1463 			break;
1464 		}
1465 		ATOMIC_INC32(softn->ipf_nat_stats.ns_rules_rdr);
1466 	}
1467 
1468 	if (n->in_redir & (NAT_MAP|NAT_MAPBLK)) {
1469 		n->in_flags &= ~IPN_NOTSRC;
1470 		switch (n->in_v[0])
1471 		{
1472 		case 4 :
1473 			ipf_nat_addmap(softn, n);
1474 			break;
1475 #ifdef USE_INET6
1476 		case 6 :
1477 			ipf_nat6_addmap(softn, n);
1478 			break;
1479 #endif
1480 		default :
1481 			break;
1482 		}
1483 		ATOMIC_INC32(softn->ipf_nat_stats.ns_rules_map);
1484 	}
1485 
1486 	if (n->in_age[0] != 0)
1487 		n->in_tqehead[0] = ipf_addtimeoutqueue(softc,
1488 						       &softn->ipf_nat_utqe,
1489 						       n->in_age[0]);
1490 
1491 	if (n->in_age[1] != 0)
1492 		n->in_tqehead[1] = ipf_addtimeoutqueue(softc,
1493 						       &softn->ipf_nat_utqe,
1494 						       n->in_age[1]);
1495 
1496 	MUTEX_INIT(&n->in_lock, "ipnat rule lock");
1497 
1498 	n = NULL;
1499 	ATOMIC_INC32(softn->ipf_nat_stats.ns_rules);
1500 #if SOLARIS && !defined(INSTANCES)
1501 	pfil_delayed_copy = 0;
1502 #endif
1503 	if (getlock) {
1504 		RWLOCK_EXIT(&softc->ipf_nat);			/* WRITE */
1505 	}
1506 
1507 	return error;
1508 }
1509 
1510 
1511 /* ------------------------------------------------------------------------ */
1512 /* Function:    ipf_nat_ruleaddrinit                                        */
1513 /* Parameters:  softc(I) - pointer to soft context main structure           */
1514 /*              softn(I) - pointer to NAT context structure                 */
1515 /*              n(I)     - pointer to NAT rule                              */
1516 /*                                                                          */
1517 /* Initialise all of the NAT address structures in a NAT rule.              */
1518 /* ------------------------------------------------------------------------ */
1519 static int
1520 ipf_nat_ruleaddrinit(ipf_main_softc_t *softc, ipf_nat_softc_t *softn,
1521     ipnat_t *n)
1522 {
1523 	int idx, error;
1524 
1525 	if ((n->in_ndst.na_atype == FRI_LOOKUP) &&
1526 	    (n->in_ndst.na_type != IPLT_DSTLIST)) {
1527 		IPFERROR(60071);
1528 		return EINVAL;
1529 	}
1530 	if ((n->in_nsrc.na_atype == FRI_LOOKUP) &&
1531 	    (n->in_nsrc.na_type != IPLT_DSTLIST)) {
1532 		IPFERROR(60069);
1533 		return EINVAL;
1534 	}
1535 
1536 	if (n->in_redir == NAT_BIMAP) {
1537 		n->in_ndstaddr = n->in_osrcaddr;
1538 		n->in_ndstmsk = n->in_osrcmsk;
1539 		n->in_odstaddr = n->in_nsrcaddr;
1540 		n->in_odstmsk = n->in_nsrcmsk;
1541 
1542 	}
1543 
1544 	if (n->in_redir & NAT_REDIRECT)
1545 		idx = 1;
1546 	else
1547 		idx = 0;
1548 	/*
1549 	 * Initialise all of the address fields.
1550 	 */
1551 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_osrc, 1,
1552 				     n->in_ifps[idx]);
1553 	if (error != 0)
1554 		return error;
1555 
1556 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_odst, 1,
1557 				     n->in_ifps[idx]);
1558 	if (error != 0)
1559 		return error;
1560 
1561 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_nsrc, 1,
1562 				     n->in_ifps[idx]);
1563 	if (error != 0)
1564 		return error;
1565 
1566 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_ndst, 1,
1567 				     n->in_ifps[idx]);
1568 	if (error != 0)
1569 		return error;
1570 
1571 	if (n->in_redir & NAT_DIVERTUDP)
1572 		ipf_nat_builddivertmp(softn, n);
1573 
1574 	return 0;
1575 }
1576 
1577 
1578 /* ------------------------------------------------------------------------ */
1579 /* Function:    ipf_nat_resolvrule                                          */
1580 /* Returns:     Nil                                                         */
1581 /* Parameters:  softc(I) - pointer to soft context main structure           */
1582 /*              n(I)     - pointer to NAT rule                              */
1583 /*                                                                          */
1584 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
1585 /* from information passed to the kernel, then add it  to the appropriate   */
1586 /* NAT rule table(s).                                                       */
1587 /* ------------------------------------------------------------------------ */
1588 static int
1589 ipf_nat_resolverule(ipf_main_softc_t *softc, ipnat_t *n)
1590 {
1591 	char *base;
1592 
1593 	base = n->in_names;
1594 
1595 	n->in_ifps[0] = ipf_resolvenic(softc, base + n->in_ifnames[0],
1596 				       n->in_v[0]);
1597 
1598 	if (n->in_ifnames[1] == -1) {
1599 		n->in_ifnames[1] = n->in_ifnames[0];
1600 		n->in_ifps[1] = n->in_ifps[0];
1601 	} else {
1602 		n->in_ifps[1] = ipf_resolvenic(softc, base + n->in_ifnames[1],
1603 					       n->in_v[1]);
1604 	}
1605 
1606 	if (n->in_plabel != -1) {
1607 		if (n->in_redir & NAT_REDIRECT)
1608 			n->in_apr = ipf_proxy_lookup(softc->ipf_proxy_soft,
1609 						     n->in_pr[0],
1610 						     base + n->in_plabel);
1611 		else
1612 			n->in_apr = ipf_proxy_lookup(softc->ipf_proxy_soft,
1613 						     n->in_pr[1],
1614 						     base + n->in_plabel);
1615 		if (n->in_apr == NULL)
1616 			return -1;
1617 	}
1618 	return 0;
1619 }
1620 
1621 
1622 /* ------------------------------------------------------------------------ */
1623 /* Function:    ipf_nat_siocdelnat                                          */
1624 /* Returns:     int - 0 == success, != 0 == failure                         */
1625 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1626 /*              softn(I)   - pointer to NAT context structure               */
1627 /*              n(I)       - pointer to new NAT rule                        */
1628 /*              getlock(I) - flag indicating if lock on  is held            */
1629 /* Mutex Locks: ipf_nat_io                                                  */
1630 /*                                                                          */
1631 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
1632 /* from information passed to the kernel, then add it  to the appropriate   */
1633 /* NAT rule table(s).                                                       */
1634 /* ------------------------------------------------------------------------ */
1635 static void
1636 ipf_nat_siocdelnat(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *n,
1637 	int getlock)
1638 {
1639 #ifdef IPF_NAT6
1640 	int i;
1641 #endif
1642 
1643 	if (getlock) {
1644 		WRITE_ENTER(&softc->ipf_nat);
1645 	}
1646 
1647 	ipf_nat_delrule(softc, softn, n, 1);
1648 
1649 	if (getlock) {
1650 		RWLOCK_EXIT(&softc->ipf_nat);			/* READ/WRITE */
1651 	}
1652 }
1653 
1654 
1655 /* ------------------------------------------------------------------------ */
1656 /* Function:    ipf_nat_getsz                                               */
1657 /* Returns:     int - 0 == success, != 0 is the error value.                */
1658 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1659 /*              data(I)    - pointer to natget structure with kernel        */
1660 /*                           pointer get the size of.                       */
1661 /*              getlock(I) - flag indicating whether or not the caller      */
1662 /*                           holds a lock on ipf_nat                        */
1663 /*                                                                          */
1664 /* Handle SIOCSTGSZ.                                                        */
1665 /* Return the size of the nat list entry to be copied back to user space.   */
1666 /* The size of the entry is stored in the ng_sz field and the enture natget */
1667 /* structure is copied back to the user.                                    */
1668 /* ------------------------------------------------------------------------ */
1669 static int
1670 ipf_nat_getsz(ipf_main_softc_t *softc, void *data, int getlock)
1671 {
1672 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1673 	ap_session_t *aps;
1674 	nat_t *nat, *n;
1675 	natget_t ng;
1676 	int error;
1677 
1678 	error = BCOPYIN(data, &ng, sizeof(ng));
1679 	if (error != 0) {
1680 		IPFERROR(60024);
1681 		return EFAULT;
1682 	}
1683 
1684 	if (getlock) {
1685 		READ_ENTER(&softc->ipf_nat);
1686 	}
1687 
1688 	nat = ng.ng_ptr;
1689 	if (!nat) {
1690 		nat = softn->ipf_nat_instances;
1691 		ng.ng_sz = 0;
1692 		/*
1693 		 * Empty list so the size returned is 0.  Simple.
1694 		 */
1695 		if (nat == NULL) {
1696 			if (getlock) {
1697 				RWLOCK_EXIT(&softc->ipf_nat);
1698 			}
1699 			error = BCOPYOUT(&ng, data, sizeof(ng));
1700 			if (error != 0) {
1701 				IPFERROR(60025);
1702 				return EFAULT;
1703 			}
1704 			return 0;
1705 		}
1706 	} else {
1707 		/*
1708 		 * Make sure the pointer we're copying from exists in the
1709 		 * current list of entries.  Security precaution to prevent
1710 		 * copying of random kernel data.
1711 		 */
1712 		for (n = softn->ipf_nat_instances; n; n = n->nat_next)
1713 			if (n == nat)
1714 				break;
1715 		if (n == NULL) {
1716 			if (getlock) {
1717 				RWLOCK_EXIT(&softc->ipf_nat);
1718 			}
1719 			IPFERROR(60026);
1720 			return ESRCH;
1721 		}
1722 	}
1723 
1724 	/*
1725 	 * Incluse any space required for proxy data structures.
1726 	 */
1727 	ng.ng_sz = sizeof(nat_save_t);
1728 	aps = nat->nat_aps;
1729 	if (aps != NULL) {
1730 		ng.ng_sz += sizeof(ap_session_t) - 4;
1731 		if (aps->aps_data != 0)
1732 			ng.ng_sz += aps->aps_psiz;
1733 	}
1734 	if (getlock) {
1735 		RWLOCK_EXIT(&softc->ipf_nat);
1736 	}
1737 
1738 	error = BCOPYOUT(&ng, data, sizeof(ng));
1739 	if (error != 0) {
1740 		IPFERROR(60027);
1741 		return EFAULT;
1742 	}
1743 	return 0;
1744 }
1745 
1746 
1747 /* ------------------------------------------------------------------------ */
1748 /* Function:    ipf_nat_getent                                              */
1749 /* Returns:     int - 0 == success, != 0 is the error value.                */
1750 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1751 /*              data(I)    - pointer to natget structure with kernel pointer*/
1752 /*                           to NAT structure to copy out.                  */
1753 /*              getlock(I) - flag indicating whether or not the caller      */
1754 /*                           holds a lock on ipf_nat                        */
1755 /*                                                                          */
1756 /* Handle SIOCSTGET.                                                        */
1757 /* Copies out NAT entry to user space.  Any additional data held for a      */
1758 /* proxy is also copied, as to is the NAT rule which was responsible for it */
1759 /* ------------------------------------------------------------------------ */
1760 static int
1761 ipf_nat_getent(ipf_main_softc_t *softc, void *data, int getlock)
1762 {
1763 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1764 	int error, outsize;
1765 	ap_session_t *aps;
1766 	nat_save_t *ipn, ipns;
1767 	nat_t *n, *nat;
1768 
1769 	error = ipf_inobj(softc, data, NULL, &ipns, IPFOBJ_NATSAVE);
1770 	if (error != 0)
1771 		return error;
1772 
1773 	if ((ipns.ipn_dsize < sizeof(ipns)) || (ipns.ipn_dsize > 81920)) {
1774 		IPFERROR(60028);
1775 		return EINVAL;
1776 	}
1777 
1778 	KMALLOCS(ipn, nat_save_t *, ipns.ipn_dsize);
1779 	if (ipn == NULL) {
1780 		IPFERROR(60029);
1781 		return ENOMEM;
1782 	}
1783 
1784 	if (getlock) {
1785 		READ_ENTER(&softc->ipf_nat);
1786 	}
1787 
1788 	ipn->ipn_dsize = ipns.ipn_dsize;
1789 	nat = ipns.ipn_next;
1790 	if (nat == NULL) {
1791 		nat = softn->ipf_nat_instances;
1792 		if (nat == NULL) {
1793 			if (softn->ipf_nat_instances == NULL) {
1794 				IPFERROR(60030);
1795 				error = ENOENT;
1796 			}
1797 			goto finished;
1798 		}
1799 	} else {
1800 		/*
1801 		 * Make sure the pointer we're copying from exists in the
1802 		 * current list of entries.  Security precaution to prevent
1803 		 * copying of random kernel data.
1804 		 */
1805 		for (n = softn->ipf_nat_instances; n; n = n->nat_next)
1806 			if (n == nat)
1807 				break;
1808 		if (n == NULL) {
1809 			IPFERROR(60031);
1810 			error = ESRCH;
1811 			goto finished;
1812 		}
1813 	}
1814 	ipn->ipn_next = nat->nat_next;
1815 
1816 	/*
1817 	 * Copy the NAT structure.
1818 	 */
1819 	bcopy((char *)nat, &ipn->ipn_nat, sizeof(*nat));
1820 
1821 	/*
1822 	 * If we have a pointer to the NAT rule it belongs to, save that too.
1823 	 */
1824 	if (nat->nat_ptr != NULL)
1825 		bcopy((char *)nat->nat_ptr, (char *)&ipn->ipn_ipnat,
1826 		      ipn->ipn_ipnat.in_size);
1827 
1828 	/*
1829 	 * If we also know the NAT entry has an associated filter rule,
1830 	 * save that too.
1831 	 */
1832 	if (nat->nat_fr != NULL)
1833 		bcopy((char *)nat->nat_fr, (char *)&ipn->ipn_fr,
1834 		      sizeof(ipn->ipn_fr));
1835 
1836 	/*
1837 	 * Last but not least, if there is an application proxy session set
1838 	 * up for this NAT entry, then copy that out too, including any
1839 	 * private data saved along side it by the proxy.
1840 	 */
1841 	aps = nat->nat_aps;
1842 	outsize = ipn->ipn_dsize - sizeof(*ipn) + sizeof(ipn->ipn_data);
1843 	if (aps != NULL) {
1844 		char *s;
1845 
1846 		if (outsize < sizeof(*aps)) {
1847 			IPFERROR(60032);
1848 			error = ENOBUFS;
1849 			goto finished;
1850 		}
1851 
1852 		s = ipn->ipn_data;
1853 		bcopy((char *)aps, s, sizeof(*aps));
1854 		s += sizeof(*aps);
1855 		outsize -= sizeof(*aps);
1856 		if ((aps->aps_data != NULL) && (outsize >= aps->aps_psiz))
1857 			bcopy(aps->aps_data, s, aps->aps_psiz);
1858 		else {
1859 			IPFERROR(60033);
1860 			error = ENOBUFS;
1861 		}
1862 	}
1863 	if (error == 0) {
1864 		if (getlock) {
1865 			READ_ENTER(&softc->ipf_nat);
1866 			getlock = 0;
1867 		}
1868 		error = ipf_outobjsz(softc, data, ipn, IPFOBJ_NATSAVE,
1869 				     ipns.ipn_dsize);
1870 	}
1871 
1872 finished:
1873 	if (getlock) {
1874 		READ_ENTER(&softc->ipf_nat);
1875 	}
1876 	if (ipn != NULL) {
1877 		KFREES(ipn, ipns.ipn_dsize);
1878 	}
1879 	return error;
1880 }
1881 
1882 
1883 /* ------------------------------------------------------------------------ */
1884 /* Function:    ipf_nat_putent                                              */
1885 /* Returns:     int - 0 == success, != 0 is the error value.                */
1886 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1887 /*              data(I)    - pointer to natget structure with NAT           */
1888 /*                           structure information to load into the kernel  */
1889 /*              getlock(I) - flag indicating whether or not a write lock    */
1890 /*                           on is already held.                            */
1891 /*                                                                          */
1892 /* Handle SIOCSTPUT.                                                        */
1893 /* Loads a NAT table entry from user space, including a NAT rule, proxy and */
1894 /* firewall rule data structures, if pointers to them indicate so.          */
1895 /* ------------------------------------------------------------------------ */
1896 static int
1897 ipf_nat_putent(ipf_main_softc_t *softc, void *data, int getlock)
1898 {
1899 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1900 	nat_save_t *ipn, *ipnn;
1901 	ap_session_t *aps;
1902 	nat_t *n, *nat;
1903 	frentry_t *fr;
1904 	fr_info_t *fin;
1905 	ipnat_t *in;
1906 	int error;
1907 
1908 	KMALLOC(ipn, nat_save_t *);
1909 	if (ipn == NULL)
1910 		return ENOMEM;
1911 	error = ipf_inobj(softc, data, NULL, ipn, IPFOBJ_NATSAVE);
1912 	if (error != 0)
1913 		return error;
1914 
1915 	/*
1916 	 * Initialise early because of code at junkput label.
1917 	 */
1918 	n = NULL;
1919 	in = NULL;
1920 	aps = NULL;
1921 	nat = NULL;
1922 	ipnn = NULL;
1923 	fin = NULL;
1924 	fr = NULL;
1925 
1926 	/*
1927 	 * New entry, copy in the rest of the NAT entry if it's size is more
1928 	 * than just the nat_t structure.
1929 	 */
1930 	if (ipn->ipn_dsize > sizeof(*ipn)) {
1931 		if (ipn->ipn_dsize > 81920) {
1932 			IPFERROR(60034);
1933 			error = ENOMEM;
1934 			goto junkput;
1935 		}
1936 
1937 		KMALLOCS(ipnn, nat_save_t *, ipn->ipn_dsize);
1938 		if (ipnn == NULL) {
1939 			IPFERROR(60035);
1940 			return ENOMEM;
1941 		}
1942 
1943 		bzero(ipnn, ipn->ipn_dsize);
1944 		error = ipf_inobjsz(softc, data, ipnn, IPFOBJ_NATSAVE,
1945 				    ipn->ipn_dsize);
1946 		if (error != 0) {
1947 			goto junkput;
1948 		}
1949 	} else
1950 		ipnn = ipn;
1951 
1952 	KMALLOC(nat, nat_t *);
1953 	if (nat == NULL) {
1954 		IPFERROR(60037);
1955 		error = ENOMEM;
1956 		goto junkput;
1957 	}
1958 
1959 	bcopy((char *)&ipnn->ipn_nat, (char *)nat, sizeof(*nat));
1960 
1961 	switch (nat->nat_v[0])
1962 	{
1963 	case 4:
1964 #ifdef USE_IENT6
1965 	case 6 :
1966 #endif
1967 		break;
1968 	default :
1969 		IPFERROR(60061);
1970 		error = EPROTONOSUPPORT;
1971 		goto junkput;
1972 		/*NOTREACHED*/
1973 	}
1974 
1975 	/*
1976 	 * Initialize all these so that ipf_nat_delete() doesn't cause a crash.
1977 	 */
1978 	bzero((char *)nat, offsetof(struct nat, nat_tqe));
1979 	nat->nat_tqe.tqe_pnext = NULL;
1980 	nat->nat_tqe.tqe_next = NULL;
1981 	nat->nat_tqe.tqe_ifq = NULL;
1982 	nat->nat_tqe.tqe_parent = nat;
1983 
1984 	/*
1985 	 * Restore the rule associated with this nat session
1986 	 */
1987 	in = ipnn->ipn_nat.nat_ptr;
1988 	if (in != NULL) {
1989 		KMALLOCS(in, ipnat_t *, ipnn->ipn_ipnat.in_size);
1990 		nat->nat_ptr = in;
1991 		if (in == NULL) {
1992 			IPFERROR(60038);
1993 			error = ENOMEM;
1994 			goto junkput;
1995 		}
1996 		bcopy((char *)&ipnn->ipn_ipnat, (char *)in,
1997 		      ipnn->ipn_ipnat.in_size);
1998 		in->in_use = 1;
1999 		in->in_flags |= IPN_DELETE;
2000 
2001 		ATOMIC_INC32(softn->ipf_nat_stats.ns_rules);
2002 
2003 		if (ipf_nat_resolverule(softc, in) != 0) {
2004 			IPFERROR(60039);
2005 			error = ESRCH;
2006 			goto junkput;
2007 		}
2008 	}
2009 
2010 	/*
2011 	 * Check that the NAT entry doesn't already exist in the kernel.
2012 	 *
2013 	 * For NAT_OUTBOUND, we're lookup for a duplicate MAP entry.  To do
2014 	 * this, we check to see if the inbound combination of addresses and
2015 	 * ports is already known.  Similar logic is applied for NAT_INBOUND.
2016 	 *
2017 	 */
2018 	KMALLOC(fin, fr_info_t *);
2019 	if (fin == NULL) {
2020 		error = ENOMEM;
2021 		goto junkput;
2022 	}
2023 	bzero(fin, sizeof(*fin));
2024 	fin->fin_v = nat->nat_v[0];
2025 	fin->fin_p = nat->nat_pr[0];
2026 	fin->fin_rev = nat->nat_rev;
2027 	fin->fin_ifp = nat->nat_ifps[0];
2028 	fin->fin_data[0] = ntohs(nat->nat_ndport);
2029 	fin->fin_data[1] = ntohs(nat->nat_nsport);
2030 
2031 	switch (nat->nat_dir)
2032 	{
2033 	case NAT_OUTBOUND :
2034 	case NAT_DIVERTOUT :
2035 		if (getlock) {
2036 			READ_ENTER(&softc->ipf_nat);
2037 		}
2038 
2039 		fin->fin_v = nat->nat_v[1];
2040 		if (nat->nat_v[1] == 4) {
2041 			n = ipf_nat_inlookup(fin, nat->nat_flags, fin->fin_p,
2042 					     nat->nat_ndstip, nat->nat_nsrcip);
2043 #ifdef USE_INET6
2044 		} else if (nat->nat_v[1] == 6) {
2045 			n = ipf_nat6_inlookup(fin, nat->nat_flags, fin->fin_p,
2046 					      &nat->nat_ndst6.in6,
2047 					      &nat->nat_nsrc6.in6);
2048 #endif
2049 		}
2050 
2051 		if (getlock) {
2052 			RWLOCK_EXIT(&softc->ipf_nat);
2053 		}
2054 		if (n != NULL) {
2055 			IPFERROR(60040);
2056 			error = EEXIST;
2057 			goto junkput;
2058 		}
2059 		break;
2060 
2061 	case NAT_INBOUND :
2062 	case NAT_DIVERTIN :
2063 		if (getlock) {
2064 			READ_ENTER(&softc->ipf_nat);
2065 		}
2066 
2067 		if (fin->fin_v == 4) {
2068 			n = ipf_nat_outlookup(fin, nat->nat_flags, fin->fin_p,
2069 					      nat->nat_ndstip,
2070 					      nat->nat_nsrcip);
2071 #ifdef USE_INET6
2072 		} else if (fin->fin_v == 6) {
2073 			n = ipf_nat6_outlookup(fin, nat->nat_flags, fin->fin_p,
2074 					       &nat->nat_ndst6.in6,
2075 					       &nat->nat_nsrc6.in6);
2076 #endif
2077 		}
2078 
2079 		if (getlock) {
2080 			RWLOCK_EXIT(&softc->ipf_nat);
2081 		}
2082 		if (n != NULL) {
2083 			IPFERROR(60041);
2084 			error = EEXIST;
2085 			goto junkput;
2086 		}
2087 		break;
2088 
2089 	default :
2090 		IPFERROR(60042);
2091 		error = EINVAL;
2092 		goto junkput;
2093 	}
2094 
2095 	/*
2096 	 * Restore ap_session_t structure.  Include the private data allocated
2097 	 * if it was there.
2098 	 */
2099 	aps = nat->nat_aps;
2100 	if (aps != NULL) {
2101 		KMALLOC(aps, ap_session_t *);
2102 		nat->nat_aps = aps;
2103 		if (aps == NULL) {
2104 			IPFERROR(60043);
2105 			error = ENOMEM;
2106 			goto junkput;
2107 		}
2108 		bcopy(ipnn->ipn_data, (char *)aps, sizeof(*aps));
2109 		if (in != NULL)
2110 			aps->aps_apr = in->in_apr;
2111 		else
2112 			aps->aps_apr = NULL;
2113 		if (aps->aps_psiz != 0) {
2114 			if (aps->aps_psiz > 81920) {
2115 				IPFERROR(60044);
2116 				error = ENOMEM;
2117 				goto junkput;
2118 			}
2119 			KMALLOCS(aps->aps_data, void *, aps->aps_psiz);
2120 			if (aps->aps_data == NULL) {
2121 				IPFERROR(60045);
2122 				error = ENOMEM;
2123 				goto junkput;
2124 			}
2125 			bcopy(ipnn->ipn_data + sizeof(*aps), aps->aps_data,
2126 			      aps->aps_psiz);
2127 		} else {
2128 			aps->aps_psiz = 0;
2129 			aps->aps_data = NULL;
2130 		}
2131 	}
2132 
2133 	/*
2134 	 * If there was a filtering rule associated with this entry then
2135 	 * build up a new one.
2136 	 */
2137 	fr = nat->nat_fr;
2138 	if (fr != NULL) {
2139 		if ((nat->nat_flags & SI_NEWFR) != 0) {
2140 			KMALLOC(fr, frentry_t *);
2141 			nat->nat_fr = fr;
2142 			if (fr == NULL) {
2143 				IPFERROR(60046);
2144 				error = ENOMEM;
2145 				goto junkput;
2146 			}
2147 			ipnn->ipn_nat.nat_fr = fr;
2148 			fr->fr_ref = 1;
2149 			(void) ipf_outobj(softc, data, ipnn, IPFOBJ_NATSAVE);
2150 			bcopy((char *)&ipnn->ipn_fr, (char *)fr, sizeof(*fr));
2151 
2152 			fr->fr_ref = 1;
2153 			fr->fr_dsize = 0;
2154 			fr->fr_data = NULL;
2155 			fr->fr_type = FR_T_NONE;
2156 
2157 			MUTEX_NUKE(&fr->fr_lock);
2158 			MUTEX_INIT(&fr->fr_lock, "nat-filter rule lock");
2159 		} else {
2160 			if (getlock) {
2161 				READ_ENTER(&softc->ipf_nat);
2162 			}
2163 			for (n = softn->ipf_nat_instances; n; n = n->nat_next)
2164 				if (n->nat_fr == fr)
2165 					break;
2166 
2167 			if (n != NULL) {
2168 				MUTEX_ENTER(&fr->fr_lock);
2169 				fr->fr_ref++;
2170 				MUTEX_EXIT(&fr->fr_lock);
2171 			}
2172 			if (getlock) {
2173 				RWLOCK_EXIT(&softc->ipf_nat);
2174 			}
2175 
2176 			if (n == NULL) {
2177 				IPFERROR(60047);
2178 				error = ESRCH;
2179 				goto junkput;
2180 			}
2181 		}
2182 	}
2183 
2184 	if (ipnn != ipn) {
2185 		KFREES(ipnn, ipn->ipn_dsize);
2186 		ipnn = NULL;
2187 	}
2188 
2189 	if (getlock) {
2190 		WRITE_ENTER(&softc->ipf_nat);
2191 	}
2192 
2193 	if (fin->fin_v == 4)
2194 		error = ipf_nat_finalise(fin, nat);
2195 #ifdef USE_INET6
2196 	else
2197 		error = ipf_nat6_finalise(fin, nat);
2198 #endif
2199 
2200 	if (getlock) {
2201 		RWLOCK_EXIT(&softc->ipf_nat);
2202 	}
2203 
2204 	if (error == 0)
2205 		return 0;
2206 
2207 	IPFERROR(60048);
2208 	error = ENOMEM;
2209 
2210 junkput:
2211 	if (fin != NULL)
2212 		KFREE(fin);
2213 	if (fr != NULL) {
2214 		(void) ipf_derefrule(softc, &fr);
2215 	}
2216 
2217 	if ((ipnn != NULL) && (ipnn != ipn)) {
2218 		KFREES(ipnn, ipn->ipn_dsize);
2219 	}
2220 	if (ipn != NULL)
2221 		KFREE(ipn);
2222 	if (nat != NULL) {
2223 		if (aps != NULL) {
2224 			if (aps->aps_data != NULL) {
2225 				KFREES(aps->aps_data, aps->aps_psiz);
2226 			}
2227 			KFREE(aps);
2228 		}
2229 		if (in != NULL) {
2230 			if (in->in_apr)
2231 				ipf_proxy_deref(in->in_apr);
2232 			KFREES(in, in->in_size);
2233 		}
2234 		KFREE(nat);
2235 	}
2236 	return error;
2237 }
2238 
2239 
2240 /* ------------------------------------------------------------------------ */
2241 /* Function:    ipf_nat_delete                                              */
2242 /* Returns:     Nil                                                         */
2243 /* Parameters:  softc(I)   - pointer to soft context main structure         */
2244 /*              nat(I)     - pointer to NAT structure to delete             */
2245 /*              logtype(I) - type of LOG record to create before deleting   */
2246 /* Write Lock:  ipf_nat                                                     */
2247 /*                                                                          */
2248 /* Delete a nat entry from the various lists and table.  If NAT logging is  */
2249 /* enabled then generate a NAT log record for this event.                   */
2250 /* ------------------------------------------------------------------------ */
2251 void
2252 ipf_nat_delete(ipf_main_softc_t *softc, struct nat *nat, int logtype)
2253 {
2254 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2255 	int madeorphan = 0, removed = 0;
2256 	u_int bkt;
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 		ASSERT(nss->ns_bucketlen[bkt] > 0);
2273 		nss->ns_bucketlen[bkt]--;
2274 		if (nss->ns_bucketlen[bkt] == 0) {
2275 			nss->ns_inuse--;
2276 		}
2277 
2278 		bkt = nat->nat_hv[1] % softn->ipf_nat_table_sz;
2279 		nss = &softn->ipf_nat_stats.ns_side[1];
2280 		ASSERT(nss->ns_bucketlen[bkt] > 0);
2281 		nss->ns_bucketlen[bkt]--;
2282 		if (nss->ns_bucketlen[bkt] == 0) {
2283 			nss->ns_inuse--;
2284 		}
2285 
2286 		*nat->nat_pnext = nat->nat_next;
2287 		if (nat->nat_next != NULL) {
2288 			nat->nat_next->nat_pnext = nat->nat_pnext;
2289 			nat->nat_next = NULL;
2290 		}
2291 		nat->nat_pnext = NULL;
2292 
2293 		*nat->nat_phnext[0] = nat->nat_hnext[0];
2294 		if (nat->nat_hnext[0] != NULL) {
2295 			nat->nat_hnext[0]->nat_phnext[0] = nat->nat_phnext[0];
2296 			nat->nat_hnext[0] = NULL;
2297 		}
2298 		nat->nat_phnext[0] = NULL;
2299 
2300 		*nat->nat_phnext[1] = nat->nat_hnext[1];
2301 		if (nat->nat_hnext[1] != NULL) {
2302 			nat->nat_hnext[1]->nat_phnext[1] = nat->nat_phnext[1];
2303 			nat->nat_hnext[1] = NULL;
2304 		}
2305 		nat->nat_phnext[1] = NULL;
2306 
2307 		if ((nat->nat_flags & SI_WILDP) != 0) {
2308 			ATOMIC_DEC32(softn->ipf_nat_stats.ns_wilds);
2309 		}
2310 		madeorphan = 1;
2311 	}
2312 
2313 	if (nat->nat_me != NULL) {
2314 		*nat->nat_me = NULL;
2315 		nat->nat_me = NULL;
2316 		nat->nat_ref--;
2317 		ASSERT(nat->nat_ref >= 0);
2318 	}
2319 
2320 	if (nat->nat_tqe.tqe_ifq != NULL) {
2321 		/*
2322 		 * No call to ipf_freetimeoutqueue() is made here, they are
2323 		 * garbage collected in ipf_nat_expire().
2324 		 */
2325 		(void) ipf_deletequeueentry(&nat->nat_tqe);
2326 	}
2327 
2328 	if (nat->nat_sync) {
2329 		ipf_sync_del_nat(softc->ipf_sync_soft, nat->nat_sync);
2330 		nat->nat_sync = NULL;
2331 	}
2332 
2333 	if (logtype == NL_EXPIRE)
2334 		softn->ipf_nat_stats.ns_expire++;
2335 
2336 	MUTEX_ENTER(&nat->nat_lock);
2337 	/*
2338 	 * NL_DESTROY should only be passed in when we've got nat_ref >= 2.
2339 	 * This happens when a nat'd packet is blocked and we want to throw
2340 	 * away the NAT session.
2341 	 */
2342 	if (logtype == NL_DESTROY) {
2343 		if (nat->nat_ref > 2) {
2344 			nat->nat_ref -= 2;
2345 			MUTEX_EXIT(&nat->nat_lock);
2346 			if (removed)
2347 				softn->ipf_nat_stats.ns_orphans++;
2348 			return;
2349 		}
2350 	} else if (nat->nat_ref > 1) {
2351 		nat->nat_ref--;
2352 		MUTEX_EXIT(&nat->nat_lock);
2353 		if (madeorphan == 1)
2354 			softn->ipf_nat_stats.ns_orphans++;
2355 		return;
2356 	}
2357 	ASSERT(nat->nat_ref >= 0);
2358 	MUTEX_EXIT(&nat->nat_lock);
2359 
2360 	nat->nat_ref = 0;
2361 
2362 	if (madeorphan == 0)
2363 		softn->ipf_nat_stats.ns_orphans--;
2364 
2365 	/*
2366 	 * At this point, nat_ref can be either 0 or -1
2367 	 */
2368 	softn->ipf_nat_stats.ns_proto[nat->nat_pr[0]]--;
2369 
2370 	if (nat->nat_fr != NULL) {
2371 		(void) ipf_derefrule(softc, &nat->nat_fr);
2372 	}
2373 
2374 	if (nat->nat_hm != NULL) {
2375 		ipf_nat_hostmapdel(softc, &nat->nat_hm);
2376 	}
2377 
2378 	/*
2379 	 * If there is an active reference from the nat entry to its parent
2380 	 * rule, decrement the rule's reference count and free it too if no
2381 	 * longer being used.
2382 	 */
2383 	ipn = nat->nat_ptr;
2384 	nat->nat_ptr = NULL;
2385 
2386 	if (ipn != NULL) {
2387 		ipn->in_space++;
2388 		ipf_nat_rule_deref(softc, &ipn);
2389 	}
2390 
2391 	if (nat->nat_aps != NULL) {
2392 		ipf_proxy_free(softc, nat->nat_aps);
2393 		nat->nat_aps = NULL;
2394 	}
2395 
2396 	MUTEX_DESTROY(&nat->nat_lock);
2397 
2398 	softn->ipf_nat_stats.ns_active--;
2399 
2400 	/*
2401 	 * If there's a fragment table entry too for this nat entry, then
2402 	 * dereference that as well.  This is after nat_lock is released
2403 	 * because of Tru64.
2404 	 */
2405 	ipf_frag_natforget(softc, (void *)nat);
2406 
2407 	KFREE(nat);
2408 }
2409 
2410 
2411 /* ------------------------------------------------------------------------ */
2412 /* Function:    ipf_nat_flushtable                                          */
2413 /* Returns:     int - number of NAT rules deleted                           */
2414 /* Parameters:  softc(I) - pointer to soft context main structure           */
2415 /*              softn(I) - pointer to NAT context structure                 */
2416 /* Write Lock:  ipf_nat                                                     */
2417 /*                                                                          */
2418 /* Deletes all currently active NAT sessions.  In deleting each NAT entry a */
2419 /* log record should be emitted in ipf_nat_delete() if NAT logging is       */
2420 /* enabled.                                                                 */
2421 /* ------------------------------------------------------------------------ */
2422 /*
2423  * nat_flushtable - clear the NAT table of all mapping entries.
2424  */
2425 static int
2426 ipf_nat_flushtable(ipf_main_softc_t *softc, ipf_nat_softc_t *softn)
2427 {
2428 	nat_t *nat;
2429 	int j = 0;
2430 
2431 	/*
2432 	 * ALL NAT mappings deleted, so lets just make the deletions
2433 	 * quicker.
2434 	 */
2435 	if (softn->ipf_nat_table[0] != NULL)
2436 		bzero((char *)softn->ipf_nat_table[0],
2437 		      sizeof(softn->ipf_nat_table[0]) *
2438 		      softn->ipf_nat_table_sz);
2439 	if (softn->ipf_nat_table[1] != NULL)
2440 		bzero((char *)softn->ipf_nat_table[1],
2441 		      sizeof(softn->ipf_nat_table[1]) *
2442 		      softn->ipf_nat_table_sz);
2443 
2444 	while ((nat = softn->ipf_nat_instances) != NULL) {
2445 		ipf_nat_delete(softc, nat, NL_FLUSH);
2446 		j++;
2447 	}
2448 
2449 	return j;
2450 }
2451 
2452 
2453 /* ------------------------------------------------------------------------ */
2454 /* Function:    ipf_nat_clearlist                                           */
2455 /* Returns:     int - number of NAT/RDR rules deleted                       */
2456 /* Parameters:  softc(I) - pointer to soft context main structure           */
2457 /*              softn(I) - pointer to NAT context structure                 */
2458 /*                                                                          */
2459 /* Delete all rules in the current list of rules.  There is nothing elegant */
2460 /* about this cleanup: simply free all entries on the list of rules and     */
2461 /* clear out the tables used for hashed NAT rule lookups.                   */
2462 /* ------------------------------------------------------------------------ */
2463 static int
2464 ipf_nat_clearlist(ipf_main_softc_t *softc, ipf_nat_softc_t *softn)
2465 {
2466 	ipnat_t *n;
2467 	int i = 0;
2468 
2469 	if (softn->ipf_nat_map_rules != NULL) {
2470 		bzero((char *)softn->ipf_nat_map_rules,
2471 		      sizeof(*softn->ipf_nat_map_rules) *
2472 		      softn->ipf_nat_maprules_sz);
2473 	}
2474 	if (softn->ipf_nat_rdr_rules != NULL) {
2475 		bzero((char *)softn->ipf_nat_rdr_rules,
2476 		      sizeof(*softn->ipf_nat_rdr_rules) *
2477 		      softn->ipf_nat_rdrrules_sz);
2478 	}
2479 
2480 	while ((n = softn->ipf_nat_list) != NULL) {
2481 		ipf_nat_delrule(softc, softn, n, 0);
2482 		i++;
2483 	}
2484 #if SOLARIS && !defined(INSTANCES)
2485 	pfil_delayed_copy = 1;
2486 #endif
2487 	return i;
2488 }
2489 
2490 
2491 /* ------------------------------------------------------------------------ */
2492 /* Function:    ipf_nat_delrule                                             */
2493 /* Returns:     Nil                                                         */
2494 /* Parameters:  softc(I) - pointer to soft context main structure           */
2495 /*              softn(I) - pointer to NAT context structure                 */
2496 /*              np(I)    - pointer to NAT rule to delete                    */
2497 /*              purge(I) - 1 == allow purge, 0 == prevent purge             */
2498 /* Locks:       WRITE(ipf_nat)                                              */
2499 /*                                                                          */
2500 /* Preventing "purge" from occuring is allowed because when all of the NAT  */
2501 /* rules are being removed, allowing the "purge" to walk through the list   */
2502 /* of NAT sessions, possibly multiple times, would be a large performance   */
2503 /* hit, on the order of O(N^2).                                             */
2504 /* ------------------------------------------------------------------------ */
2505 static void
2506 ipf_nat_delrule(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *np,
2507 	int purge)
2508 {
2509 
2510 	if (np->in_pnext != NULL) {
2511 		*np->in_pnext = np->in_next;
2512 		if (np->in_next != NULL)
2513 			np->in_next->in_pnext = np->in_pnext;
2514 		if (softn->ipf_nat_list_tail == &np->in_next)
2515 			softn->ipf_nat_list_tail = np->in_pnext;
2516 	}
2517 
2518 	if ((purge == 1) && ((np->in_flags & IPN_PURGE) != 0)) {
2519 		nat_t *next;
2520 		nat_t *nat;
2521 
2522 		for (next = softn->ipf_nat_instances; (nat = next) != NULL;) {
2523 			next = nat->nat_next;
2524 			if (nat->nat_ptr == np)
2525 				ipf_nat_delete(softc, nat, NL_PURGE);
2526 		}
2527 	}
2528 
2529 	if ((np->in_flags & IPN_DELETE) == 0) {
2530 		if (np->in_redir & NAT_REDIRECT) {
2531 			switch (np->in_v[0])
2532 			{
2533 			case 4 :
2534 				ipf_nat_delrdr(softn, np);
2535 				break;
2536 #ifdef USE_INET6
2537 			case 6 :
2538 				ipf_nat6_delrdr(softn, np);
2539 				break;
2540 #endif
2541 			}
2542 		}
2543 		if (np->in_redir & (NAT_MAPBLK|NAT_MAP)) {
2544 			switch (np->in_v[0])
2545 			{
2546 			case 4 :
2547 				ipf_nat_delmap(softn, np);
2548 				break;
2549 #ifdef USE_INET6
2550 			case 6 :
2551 				ipf_nat6_delmap(softn, np);
2552 				break;
2553 #endif
2554 			}
2555 		}
2556 	}
2557 
2558 	np->in_flags |= IPN_DELETE;
2559 	ipf_nat_rule_deref(softc, &np);
2560 }
2561 
2562 
2563 /* ------------------------------------------------------------------------ */
2564 /* Function:    ipf_nat_newmap                                              */
2565 /* Returns:     int - -1 == error, 0 == success                             */
2566 /* Parameters:  fin(I) - pointer to packet information                      */
2567 /*              nat(I) - pointer to NAT entry                               */
2568 /*              ni(I)  - pointer to structure with misc. information needed */
2569 /*                       to create new NAT entry.                           */
2570 /*                                                                          */
2571 /* Given an empty NAT structure, populate it with new information about a   */
2572 /* new NAT session, as defined by the matching NAT rule.                    */
2573 /* ni.nai_ip is passed in uninitialised and must be set, in host byte order,*/
2574 /* to the new IP address for the translation.                               */
2575 /* ------------------------------------------------------------------------ */
2576 static int
2577 ipf_nat_newmap(fr_info_t *fin, nat_t *nat, natinfo_t *ni)
2578 {
2579 	ipf_main_softc_t *softc = fin->fin_main_soft;
2580 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2581 	u_short st_port, dport, sport, port, sp, dp;
2582 	struct in_addr in, inb;
2583 	hostmap_t *hm;
2584 	u_32_t flags;
2585 	u_32_t st_ip;
2586 	ipnat_t *np;
2587 	nat_t *natl;
2588 	int l;
2589 
2590 	/*
2591 	 * If it's an outbound packet which doesn't match any existing
2592 	 * record, then create a new port
2593 	 */
2594 	l = 0;
2595 	hm = NULL;
2596 	np = ni->nai_np;
2597 	st_ip = np->in_snip;
2598 	st_port = np->in_spnext;
2599 	flags = nat->nat_flags;
2600 
2601 	if (flags & IPN_ICMPQUERY) {
2602 		sport = fin->fin_data[1];
2603 		dport = 0;
2604 	} else {
2605 		sport = htons(fin->fin_data[0]);
2606 		dport = htons(fin->fin_data[1]);
2607 	}
2608 
2609 	/*
2610 	 * Do a loop until we either run out of entries to try or we find
2611 	 * a NAT mapping that isn't currently being used.  This is done
2612 	 * because the change to the source is not (usually) being fixed.
2613 	 */
2614 	do {
2615 		port = 0;
2616 		in.s_addr = htonl(np->in_snip);
2617 		if (l == 0) {
2618 			/*
2619 			 * Check to see if there is an existing NAT
2620 			 * setup for this IP address pair.
2621 			 */
2622 			hm = ipf_nat_hostmap(softn, np, fin->fin_src,
2623 					     fin->fin_dst, in, 0);
2624 			if (hm != NULL)
2625 				in.s_addr = hm->hm_nsrcip.s_addr;
2626 		} else if ((l == 1) && (hm != NULL)) {
2627 			ipf_nat_hostmapdel(softc, &hm);
2628 		}
2629 		in.s_addr = ntohl(in.s_addr);
2630 
2631 		nat->nat_hm = hm;
2632 
2633 		if ((np->in_nsrcmsk == 0xffffffff) && (np->in_spnext == 0)) {
2634 			if (l > 0) {
2635 				NBUMPSIDEX(1, ns_exhausted, ns_exhausted_1);
2636 				return -1;
2637 			}
2638 		}
2639 
2640 		if (np->in_redir == NAT_BIMAP &&
2641 		    np->in_osrcmsk == np->in_nsrcmsk) {
2642 			/*
2643 			 * map the address block in a 1:1 fashion
2644 			 */
2645 			in.s_addr = np->in_nsrcaddr;
2646 			in.s_addr |= fin->fin_saddr & ~np->in_osrcmsk;
2647 			in.s_addr = ntohl(in.s_addr);
2648 
2649 		} else if (np->in_redir & NAT_MAPBLK) {
2650 			if ((l >= np->in_ppip) || ((l > 0) &&
2651 			     !(flags & IPN_TCPUDP))) {
2652 				NBUMPSIDEX(1, ns_exhausted, ns_exhausted_2);
2653 				return -1;
2654 			}
2655 			/*
2656 			 * map-block - Calculate destination address.
2657 			 */
2658 			in.s_addr = ntohl(fin->fin_saddr);
2659 			in.s_addr &= ntohl(~np->in_osrcmsk);
2660 			inb.s_addr = in.s_addr;
2661 			in.s_addr /= np->in_ippip;
2662 			in.s_addr &= ntohl(~np->in_nsrcmsk);
2663 			in.s_addr += ntohl(np->in_nsrcaddr);
2664 			/*
2665 			 * Calculate destination port.
2666 			 */
2667 			if ((flags & IPN_TCPUDP) &&
2668 			    (np->in_ppip != 0)) {
2669 				port = ntohs(sport) + l;
2670 				port %= np->in_ppip;
2671 				port += np->in_ppip *
2672 					(inb.s_addr % np->in_ippip);
2673 				port += MAPBLK_MINPORT;
2674 				port = htons(port);
2675 			}
2676 
2677 		} else if ((np->in_nsrcaddr == 0) &&
2678 			   (np->in_nsrcmsk == 0xffffffff)) {
2679 			i6addr_t in6;
2680 
2681 			/*
2682 			 * 0/32 - use the interface's IP address.
2683 			 */
2684 			if ((l > 0) ||
2685 			    ipf_ifpaddr(softc, 4, FRI_NORMAL, fin->fin_ifp,
2686 				       &in6, NULL) == -1) {
2687 				NBUMPSIDEX(1, ns_new_ifpaddr, ns_new_ifpaddr_1);
2688 				return -1;
2689 			}
2690 			in.s_addr = ntohl(in6.in4.s_addr);
2691 
2692 		} else if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0)) {
2693 			/*
2694 			 * 0/0 - use the original source address/port.
2695 			 */
2696 			if (l > 0) {
2697 				NBUMPSIDEX(1, ns_exhausted, ns_exhausted_3);
2698 				return -1;
2699 			}
2700 			in.s_addr = ntohl(fin->fin_saddr);
2701 
2702 		} else if ((np->in_nsrcmsk != 0xffffffff) &&
2703 			   (np->in_spnext == 0) && ((l > 0) || (hm == NULL)))
2704 			np->in_snip++;
2705 
2706 		natl = NULL;
2707 
2708 		if ((flags & IPN_TCPUDP) &&
2709 		    ((np->in_redir & NAT_MAPBLK) == 0) &&
2710 		    (np->in_flags & IPN_AUTOPORTMAP)) {
2711 			/*
2712 			 * "ports auto" (without map-block)
2713 			 */
2714 			if ((l > 0) && np->in_ppip && (l % np->in_ppip == 0)) {
2715 				if (l > np->in_space) {
2716 					return -1;
2717 				} else if ((l > np->in_ppip) &&
2718 					   np->in_nsrcmsk != 0xffffffff)
2719 					np->in_snip++;
2720 			}
2721 			if (np->in_ppip != 0) {
2722 				port = ntohs(sport);
2723 				port += (l % np->in_ppip);
2724 				port %= np->in_ppip;
2725 				port += np->in_ppip *
2726 					(ntohl(fin->fin_saddr) %
2727 					 np->in_ippip);
2728 				port += MAPBLK_MINPORT;
2729 				port = htons(port);
2730 			}
2731 
2732 		} else if (((np->in_redir & NAT_MAPBLK) == 0) &&
2733 			   (flags & IPN_TCPUDPICMP) && (np->in_spnext != 0)) {
2734 			/*
2735 			 * Standard port translation.  Select next port.
2736 			 */
2737 			if (np->in_flags & IPN_SEQUENTIAL) {
2738 				port = np->in_spnext;
2739 			} else {
2740 				port = ipf_random() % (np->in_spmax -
2741 						       np->in_spmin + 1);
2742 				port += np->in_spmin;
2743 			}
2744 			port = htons(port);
2745 			np->in_spnext++;
2746 
2747 			if (np->in_spnext > np->in_spmax) {
2748 				np->in_spnext = np->in_spmin;
2749 				if (np->in_nsrcmsk != 0xffffffff)
2750 					np->in_snip++;
2751 			}
2752 		}
2753 
2754 		if (np->in_flags & IPN_SIPRANGE) {
2755 			if (np->in_snip > ntohl(np->in_nsrcmsk))
2756 				np->in_snip = ntohl(np->in_nsrcaddr);
2757 		} else {
2758 			if ((np->in_nsrcmsk != 0xffffffff) &&
2759 			    ((np->in_snip + 1) & ntohl(np->in_nsrcmsk)) >
2760 			    ntohl(np->in_nsrcaddr))
2761 				np->in_snip = ntohl(np->in_nsrcaddr) + 1;
2762 		}
2763 
2764 		if ((port == 0) && (flags & (IPN_TCPUDPICMP|IPN_ICMPQUERY)))
2765 			port = sport;
2766 
2767 		/*
2768 		 * Here we do a lookup of the connection as seen from
2769 		 * the outside.  If an IP# pair already exists, try
2770 		 * again.  So if you have A->B becomes C->B, you can
2771 		 * also have D->E become C->E but not D->B causing
2772 		 * another C->B.  Also take protocol and ports into
2773 		 * account when determining whether a pre-existing
2774 		 * NAT setup will cause an external conflict where
2775 		 * this is appropriate.
2776 		 */
2777 		inb.s_addr = htonl(in.s_addr);
2778 		sp = fin->fin_data[0];
2779 		dp = fin->fin_data[1];
2780 		fin->fin_data[0] = fin->fin_data[1];
2781 		fin->fin_data[1] = ntohs(port);
2782 		natl = ipf_nat_inlookup(fin, flags & ~(SI_WILDP|NAT_SEARCH),
2783 					(u_int)fin->fin_p, fin->fin_dst, inb);
2784 		fin->fin_data[0] = sp;
2785 		fin->fin_data[1] = dp;
2786 
2787 		/*
2788 		 * Has the search wrapped around and come back to the
2789 		 * start ?
2790 		 */
2791 		if ((natl != NULL) &&
2792 		    (np->in_spnext != 0) && (st_port == np->in_spnext) &&
2793 		    (np->in_snip != 0) && (st_ip == np->in_snip)) {
2794 			NBUMPSIDED(1, ns_wrap);
2795 			return -1;
2796 		}
2797 		l++;
2798 	} while (natl != NULL);
2799 
2800 	/* Setup the NAT table */
2801 	nat->nat_osrcip = fin->fin_src;
2802 	nat->nat_nsrcaddr = htonl(in.s_addr);
2803 	nat->nat_odstip = fin->fin_dst;
2804 	nat->nat_ndstip = fin->fin_dst;
2805 	if (nat->nat_hm == NULL)
2806 		nat->nat_hm = ipf_nat_hostmap(softn, np, fin->fin_src,
2807 					      fin->fin_dst, nat->nat_nsrcip,
2808 					      0);
2809 
2810 	if (flags & IPN_TCPUDP) {
2811 		nat->nat_osport = sport;
2812 		nat->nat_nsport = port;	/* sport */
2813 		nat->nat_odport = dport;
2814 		nat->nat_ndport = dport;
2815 		((tcphdr_t *)fin->fin_dp)->th_sport = port;
2816 	} else if (flags & IPN_ICMPQUERY) {
2817 		nat->nat_oicmpid = fin->fin_data[1];
2818 		((icmphdr_t *)fin->fin_dp)->icmp_id = port;
2819 		nat->nat_nicmpid = port;
2820 	}
2821 	return 0;
2822 }
2823 
2824 
2825 /* ------------------------------------------------------------------------ */
2826 /* Function:    ipf_nat_newrdr                                              */
2827 /* Returns:     int - -1 == error, 0 == success (no move), 1 == success and */
2828 /*                    allow rule to be moved if IPN_ROUNDR is set.          */
2829 /* Parameters:  fin(I) - pointer to packet information                      */
2830 /*              nat(I) - pointer to NAT entry                               */
2831 /*              ni(I)  - pointer to structure with misc. information needed */
2832 /*                       to create new NAT entry.                           */
2833 /*                                                                          */
2834 /* ni.nai_ip is passed in uninitialised and must be set, in host byte order,*/
2835 /* to the new IP address for the translation.                               */
2836 /* ------------------------------------------------------------------------ */
2837 static int
2838 ipf_nat_newrdr(fr_info_t *fin, nat_t *nat, natinfo_t *ni)
2839 {
2840 	ipf_main_softc_t *softc = fin->fin_main_soft;
2841 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2842 	u_short nport, dport, sport;
2843 	struct in_addr in, inb;
2844 	u_short sp, dp;
2845 	hostmap_t *hm;
2846 	u_32_t flags;
2847 	ipnat_t *np;
2848 	nat_t *natl;
2849 	int move;
2850 
2851 	move = 1;
2852 	hm = NULL;
2853 	in.s_addr = 0;
2854 	np = ni->nai_np;
2855 	flags = nat->nat_flags;
2856 
2857 	if (flags & IPN_ICMPQUERY) {
2858 		dport = fin->fin_data[1];
2859 		sport = 0;
2860 	} else {
2861 		sport = htons(fin->fin_data[0]);
2862 		dport = htons(fin->fin_data[1]);
2863 	}
2864 
2865 	/* TRACE sport, dport */
2866 
2867 
2868 	/*
2869 	 * If the matching rule has IPN_STICKY set, then we want to have the
2870 	 * same rule kick in as before.  Why would this happen?  If you have
2871 	 * a collection of rdr rules with "round-robin sticky", the current
2872 	 * packet might match a different one to the previous connection but
2873 	 * we want the same destination to be used.
2874 	 */
2875 	if (((np->in_flags & (IPN_ROUNDR|IPN_SPLIT)) != 0) &&
2876 	    ((np->in_flags & IPN_STICKY) != 0)) {
2877 		hm = ipf_nat_hostmap(softn, NULL, fin->fin_src, fin->fin_dst,
2878 				     in, (u_32_t)dport);
2879 		if (hm != NULL) {
2880 			in.s_addr = ntohl(hm->hm_ndstip.s_addr);
2881 			np = hm->hm_ipnat;
2882 			ni->nai_np = np;
2883 			move = 0;
2884 			ipf_nat_hostmapdel(softc, &hm);
2885 		}
2886 	}
2887 
2888 	/*
2889 	 * Otherwise, it's an inbound packet. Most likely, we don't
2890 	 * want to rewrite source ports and source addresses. Instead,
2891 	 * we want to rewrite to a fixed internal address and fixed
2892 	 * internal port.
2893 	 */
2894 	if (np->in_flags & IPN_SPLIT) {
2895 		in.s_addr = np->in_dnip;
2896 		inb.s_addr = htonl(in.s_addr);
2897 
2898 		if ((np->in_flags & (IPN_ROUNDR|IPN_STICKY)) == IPN_STICKY) {
2899 			hm = ipf_nat_hostmap(softn, NULL, fin->fin_src,
2900 					     fin->fin_dst, inb, (u_32_t)dport);
2901 			if (hm != NULL) {
2902 				in.s_addr = hm->hm_ndstip.s_addr;
2903 				move = 0;
2904 			}
2905 		}
2906 
2907 		if (hm == NULL || hm->hm_ref == 1) {
2908 			if (np->in_ndstaddr == htonl(in.s_addr)) {
2909 				np->in_dnip = ntohl(np->in_ndstmsk);
2910 				move = 0;
2911 			} else {
2912 				np->in_dnip = ntohl(np->in_ndstaddr);
2913 			}
2914 		}
2915 		if (hm != NULL)
2916 			ipf_nat_hostmapdel(softc, &hm);
2917 
2918 	} else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0xffffffff)) {
2919 		i6addr_t in6;
2920 
2921 		/*
2922 		 * 0/32 - use the interface's IP address.
2923 		 */
2924 		if (ipf_ifpaddr(softc, 4, FRI_NORMAL, fin->fin_ifp,
2925 			       &in6, NULL) == -1) {
2926 			NBUMPSIDEX(0, ns_new_ifpaddr, ns_new_ifpaddr_2);
2927 			return -1;
2928 		}
2929 		in.s_addr = ntohl(in6.in4.s_addr);
2930 
2931 	} else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk== 0)) {
2932 		/*
2933 		 * 0/0 - use the original destination address/port.
2934 		 */
2935 		in.s_addr = ntohl(fin->fin_daddr);
2936 
2937 	} else if (np->in_redir == NAT_BIMAP &&
2938 		   np->in_ndstmsk == np->in_odstmsk) {
2939 		/*
2940 		 * map the address block in a 1:1 fashion
2941 		 */
2942 		in.s_addr = np->in_ndstaddr;
2943 		in.s_addr |= fin->fin_daddr & ~np->in_ndstmsk;
2944 		in.s_addr = ntohl(in.s_addr);
2945 	} else {
2946 		in.s_addr = ntohl(np->in_ndstaddr);
2947 	}
2948 
2949 	if ((np->in_dpnext == 0) || ((flags & NAT_NOTRULEPORT) != 0))
2950 		nport = dport;
2951 	else {
2952 		/*
2953 		 * Whilst not optimized for the case where
2954 		 * pmin == pmax, the gain is not significant.
2955 		 */
2956 		if (((np->in_flags & IPN_FIXEDDPORT) == 0) &&
2957 		    (np->in_odport != np->in_dtop)) {
2958 			nport = ntohs(dport) - np->in_odport + np->in_dpmax;
2959 			nport = htons(nport);
2960 		} else {
2961 			nport = htons(np->in_dpnext);
2962 			np->in_dpnext++;
2963 			if (np->in_dpnext > np->in_dpmax)
2964 				np->in_dpnext = np->in_dpmin;
2965 		}
2966 	}
2967 
2968 	/*
2969 	 * When the redirect-to address is set to 0.0.0.0, just
2970 	 * assume a blank `forwarding' of the packet.  We don't
2971 	 * setup any translation for this either.
2972 	 */
2973 	if (in.s_addr == 0) {
2974 		if (nport == dport) {
2975 			NBUMPSIDED(0, ns_xlate_null);
2976 			return -1;
2977 		}
2978 		in.s_addr = ntohl(fin->fin_daddr);
2979 	}
2980 
2981 	/*
2982 	 * Check to see if this redirect mapping already exists and if
2983 	 * it does, return "failure" (allowing it to be created will just
2984 	 * cause one or both of these "connections" to stop working.)
2985 	 */
2986 	inb.s_addr = htonl(in.s_addr);
2987 	sp = fin->fin_data[0];
2988 	dp = fin->fin_data[1];
2989 	fin->fin_data[1] = fin->fin_data[0];
2990 	fin->fin_data[0] = ntohs(nport);
2991 	natl = ipf_nat_outlookup(fin, flags & ~(SI_WILDP|NAT_SEARCH),
2992 			     (u_int)fin->fin_p, inb, fin->fin_src);
2993 	fin->fin_data[0] = sp;
2994 	fin->fin_data[1] = dp;
2995 	if (natl != NULL) {
2996 		DT2(ns_new_xlate_exists, fr_info_t *, fin, nat_t *, natl);
2997 		NBUMPSIDE(0, ns_xlate_exists);
2998 		return -1;
2999 	}
3000 
3001 	nat->nat_ndstaddr = htonl(in.s_addr);
3002 	nat->nat_odstip = fin->fin_dst;
3003 	nat->nat_nsrcip = fin->fin_src;
3004 	nat->nat_osrcip = fin->fin_src;
3005 	if ((nat->nat_hm == NULL) && ((np->in_flags & IPN_STICKY) != 0))
3006 		nat->nat_hm = ipf_nat_hostmap(softn, np, fin->fin_src,
3007 					      fin->fin_dst, inb, (u_32_t)dport);
3008 
3009 	if (flags & IPN_TCPUDP) {
3010 		nat->nat_odport = dport;
3011 		nat->nat_ndport = nport;
3012 		nat->nat_osport = sport;
3013 		nat->nat_nsport = sport;
3014 		((tcphdr_t *)fin->fin_dp)->th_dport = nport;
3015 	} else if (flags & IPN_ICMPQUERY) {
3016 		nat->nat_oicmpid = fin->fin_data[1];
3017 		((icmphdr_t *)fin->fin_dp)->icmp_id = nport;
3018 		nat->nat_nicmpid = nport;
3019 	}
3020 
3021 	return move;
3022 }
3023 
3024 /* ------------------------------------------------------------------------ */
3025 /* Function:    ipf_nat_add                                                 */
3026 /* Returns:     nat_t* - NULL == failure to create new NAT structure,       */
3027 /*                       else pointer to new NAT structure                  */
3028 /* Parameters:  fin(I)       - pointer to packet information                */
3029 /*              np(I)        - pointer to NAT rule                          */
3030 /*              natsave(I)   - pointer to where to store NAT struct pointer */
3031 /*              flags(I)     - flags describing the current packet          */
3032 /*              direction(I) - direction of packet (in/out)                 */
3033 /* Write Lock:  ipf_nat                                                     */
3034 /*                                                                          */
3035 /* Attempts to create a new NAT entry.  Does not actually change the packet */
3036 /* in any way.                                                              */
3037 /*                                                                          */
3038 /* This function is in three main parts: (1) deal with creating a new NAT   */
3039 /* structure for a "MAP" rule (outgoing NAT translation); (2) deal with     */
3040 /* creating a new NAT structure for a "RDR" rule (incoming NAT translation) */
3041 /* and (3) building that structure and putting it into the NAT table(s).    */
3042 /*                                                                          */
3043 /* NOTE: natsave should NOT be used to point back to an ipstate_t struct    */
3044 /*       as it can result in memory being corrupted.                        */
3045 /* ------------------------------------------------------------------------ */
3046 nat_t *
3047 ipf_nat_add(fr_info_t *fin, ipnat_t *np, nat_t **natsave, u_int flags,
3048     int direction)
3049 {
3050 	ipf_main_softc_t *softc = fin->fin_main_soft;
3051 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3052 	hostmap_t *hm = NULL;
3053 	nat_t *nat, *natl;
3054 	natstat_t *nsp;
3055 	u_int nflags;
3056 	natinfo_t ni;
3057 	int move;
3058 
3059 	memset(&ni, 0, sizeof ni);	/* XXX gcc */
3060 	nsp = &softn->ipf_nat_stats;
3061 
3062 	if ((nsp->ns_active * 100 / softn->ipf_nat_table_max) >
3063 	    softn->ipf_nat_table_wm_high) {
3064 		softn->ipf_nat_doflush = 1;
3065 	}
3066 
3067 	if (nsp->ns_active >= softn->ipf_nat_table_max) {
3068 		NBUMPSIDED(fin->fin_out, ns_table_max);
3069 		return NULL;
3070 	}
3071 
3072 	move = 1;
3073 	nflags = np->in_flags & flags;
3074 	nflags &= NAT_FROMRULE;
3075 
3076 	ni.nai_np = np;
3077 	ni.nai_dport = 0;
3078 	ni.nai_sport = 0;
3079 
3080 	/* Give me a new nat */
3081 	KMALLOC(nat, nat_t *);
3082 	if (nat == NULL) {
3083 		NBUMPSIDED(fin->fin_out, ns_memfail);
3084 		/*
3085 		 * Try to automatically tune the max # of entries in the
3086 		 * table allowed to be less than what will cause kmem_alloc()
3087 		 * to fail and try to eliminate panics due to out of memory
3088 		 * conditions arising.
3089 		 */
3090 		if ((softn->ipf_nat_table_max > softn->ipf_nat_table_sz) &&
3091 		    (nsp->ns_active > 100)) {
3092 			softn->ipf_nat_table_max = nsp->ns_active - 100;
3093 			printf("table_max reduced to %d\n",
3094 				softn->ipf_nat_table_max);
3095 		}
3096 		return NULL;
3097 	}
3098 
3099 	if (flags & IPN_ICMPQUERY) {
3100 		/*
3101 		 * In the ICMP query NAT code, we translate the ICMP id fields
3102 		 * to make them unique. This is indepedent of the ICMP type
3103 		 * (e.g. in the unlikely event that a host sends an echo and
3104 		 * an tstamp request with the same id, both packets will have
3105 		 * their ip address/id field changed in the same way).
3106 		 */
3107 		/* The icmp_id field is used by the sender to identify the
3108 		 * process making the icmp request. (the receiver justs
3109 		 * copies it back in its response). So, it closely matches
3110 		 * the concept of source port. We overlay sport, so we can
3111 		 * maximally reuse the existing code.
3112 		 */
3113 		ni.nai_sport = fin->fin_data[1];
3114 		ni.nai_dport = 0;
3115 	}
3116 
3117 	bzero((char *)nat, sizeof(*nat));
3118 	nat->nat_flags = flags;
3119 	nat->nat_redir = np->in_redir;
3120 	nat->nat_dir = direction;
3121 	nat->nat_pr[0] = fin->fin_p;
3122 	nat->nat_pr[1] = fin->fin_p;
3123 
3124 	/*
3125 	 * Search the current table for a match and create a new mapping
3126 	 * if there is none found.
3127 	 */
3128 	if (np->in_redir & NAT_DIVERTUDP) {
3129 		move = ipf_nat_newdivert(fin, nat, &ni);
3130 
3131 	} else if (np->in_redir & NAT_REWRITE) {
3132 		move = ipf_nat_newrewrite(fin, nat, &ni);
3133 
3134 	} else if (direction == NAT_OUTBOUND) {
3135 		/*
3136 		 * We can now arrange to call this for the same connection
3137 		 * because ipf_nat_new doesn't protect the code path into
3138 		 * this function.
3139 		 */
3140 		natl = ipf_nat_outlookup(fin, nflags, (u_int)fin->fin_p,
3141 				     fin->fin_src, fin->fin_dst);
3142 		if (natl != NULL) {
3143 			KFREE(nat);
3144 			nat = natl;
3145 			goto done;
3146 		}
3147 
3148 		move = ipf_nat_newmap(fin, nat, &ni);
3149 	} else {
3150 		/*
3151 		 * NAT_INBOUND is used for redirects rules
3152 		 */
3153 		natl = ipf_nat_inlookup(fin, nflags, (u_int)fin->fin_p,
3154 					fin->fin_src, fin->fin_dst);
3155 		if (natl != NULL) {
3156 			KFREE(nat);
3157 			nat = natl;
3158 			goto done;
3159 		}
3160 
3161 		move = ipf_nat_newrdr(fin, nat, &ni);
3162 	}
3163 	if (move == -1)
3164 		goto badnat;
3165 
3166 	np = ni.nai_np;
3167 
3168 	nat->nat_mssclamp = np->in_mssclamp;
3169 	nat->nat_me = natsave;
3170 	nat->nat_fr = fin->fin_fr;
3171 	nat->nat_rev = fin->fin_rev;
3172 	nat->nat_ptr = np;
3173 	nat->nat_dlocal = np->in_dlocal;
3174 
3175 	if ((np->in_apr != NULL) && ((nat->nat_flags & NAT_SLAVE) == 0)) {
3176 		if (ipf_proxy_new(fin, nat) == -1) {
3177 			NBUMPSIDED(fin->fin_out, ns_appr_fail);
3178 			goto badnat;
3179 		}
3180 	}
3181 
3182 	nat->nat_ifps[0] = np->in_ifps[0];
3183 	if (np->in_ifps[0] != NULL) {
3184 		COPYIFNAME(np->in_v[0], np->in_ifps[0], nat->nat_ifnames[0]);
3185 	}
3186 
3187 	nat->nat_ifps[1] = np->in_ifps[1];
3188 	if (np->in_ifps[1] != NULL) {
3189 		COPYIFNAME(np->in_v[1], np->in_ifps[1], nat->nat_ifnames[1]);
3190 	}
3191 
3192 	if (ipf_nat_finalise(fin, nat) == -1) {
3193 		goto badnat;
3194 	}
3195 
3196 	np->in_use++;
3197 
3198 	if ((move == 1) && (np->in_flags & IPN_ROUNDR)) {
3199 		if ((np->in_redir & (NAT_REDIRECT|NAT_MAP)) == NAT_REDIRECT) {
3200 			ipf_nat_delrdr(softn, np);
3201 			ipf_nat_addrdr(softn, np);
3202 		} else if ((np->in_redir & (NAT_REDIRECT|NAT_MAP)) == NAT_MAP) {
3203 			ipf_nat_delmap(softn, np);
3204 			ipf_nat_addmap(softn, np);
3205 		}
3206 	}
3207 
3208 	if (flags & SI_WILDP)
3209 		nsp->ns_wilds++;
3210 	nsp->ns_proto[nat->nat_pr[0]]++;
3211 
3212 	goto done;
3213 badnat:
3214 	DT2(ns_badnatnew, fr_info_t *, fin, nat_t *, nat);
3215 	NBUMPSIDE(fin->fin_out, ns_badnatnew);
3216 	if ((hm = nat->nat_hm) != NULL)
3217 		ipf_nat_hostmapdel(softc, &hm);
3218 	KFREE(nat);
3219 	nat = NULL;
3220 done:
3221 	if (nat != NULL && np != NULL)
3222 		np->in_hits++;
3223 	if (natsave != NULL)
3224 		*natsave = nat;
3225 	return nat;
3226 }
3227 
3228 
3229 /* ------------------------------------------------------------------------ */
3230 /* Function:    ipf_nat_finalise                                            */
3231 /* Returns:     int - 0 == sucess, -1 == failure                            */
3232 /* Parameters:  fin(I) - pointer to packet information                      */
3233 /*              nat(I) - pointer to NAT entry                               */
3234 /* Write Lock:  ipf_nat                                                     */
3235 /*                                                                          */
3236 /* This is the tail end of constructing a new NAT entry and is the same     */
3237 /* for both IPv4 and IPv6.                                                  */
3238 /* ------------------------------------------------------------------------ */
3239 /*ARGSUSED*/
3240 static int
3241 ipf_nat_finalise(fr_info_t *fin, nat_t *nat)
3242 {
3243 	ipf_main_softc_t *softc = fin->fin_main_soft;
3244 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3245 	u_32_t sum1, sum2, sumd;
3246 	frentry_t *fr;
3247 #if SOLARIS && defined(_KERNEL) && (SOLARIS2 >= 6) && defined(ICK_M_CTL_MAGIC)
3248 	qpktinfo_t *qpi = fin->fin_qpi;
3249 #endif
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 	int ret;
3357 
3358 	/*
3359 	 * Try and return an error as early as possible, so calculate the hash
3360 	 * entry numbers first and then proceed.
3361 	 */
3362 	if ((nat->nat_flags & (SI_W_SPORT|SI_W_DPORT)) == 0) {
3363 		if ((nat->nat_flags & IPN_TCPUDP) != 0) {
3364 			sp = nat->nat_osport;
3365 			dp = nat->nat_odport;
3366 		} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
3367 			sp = 0;
3368 			dp = nat->nat_oicmpid;
3369 		} else {
3370 			sp = 0;
3371 			dp = 0;
3372 		}
3373 		hv0 = NAT_HASH_FN(nat->nat_osrcaddr, sp, 0xffffffff);
3374 		hv0 = NAT_HASH_FN(nat->nat_odstaddr, hv0 + dp, 0xffffffff);
3375 		/*
3376 		 * TRACE nat_osrcaddr, nat_osport, nat_odstaddr,
3377 		 * nat_odport, hv0
3378 		 */
3379 
3380 		if ((nat->nat_flags & IPN_TCPUDP) != 0) {
3381 			sp = nat->nat_nsport;
3382 			dp = nat->nat_ndport;
3383 		} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
3384 			sp = 0;
3385 			dp = nat->nat_nicmpid;
3386 		} else {
3387 			sp = 0;
3388 			dp = 0;
3389 		}
3390 		hv1 = NAT_HASH_FN(nat->nat_nsrcaddr, sp, 0xffffffff);
3391 		hv1 = NAT_HASH_FN(nat->nat_ndstaddr, hv1 + dp, 0xffffffff);
3392 		/*
3393 		 * TRACE nat_nsrcaddr, nat_nsport, nat_ndstaddr,
3394 		 * nat_ndport, hv1
3395 		 */
3396 	} else {
3397 		hv0 = NAT_HASH_FN(nat->nat_osrcaddr, 0, 0xffffffff);
3398 		hv0 = NAT_HASH_FN(nat->nat_odstaddr, hv0, 0xffffffff);
3399 		/* TRACE nat_osrcaddr, nat_odstaddr, hv0 */
3400 
3401 		hv1 = NAT_HASH_FN(nat->nat_nsrcaddr, 0, 0xffffffff);
3402 		hv1 = NAT_HASH_FN(nat->nat_ndstaddr, hv1, 0xffffffff);
3403 		/* TRACE nat_nsrcaddr, nat_ndstaddr, hv1 */
3404 	}
3405 
3406 	if ((nat->nat_dir & NAT_OUTBOUND) == NAT_OUTBOUND) {
3407 		nat->nat_hv[0] = hv0;
3408 		nat->nat_hv[1] = hv1;
3409 	} else {
3410 		nat->nat_hv[0] = hv1;
3411 		nat->nat_hv[1] = hv0;
3412 	}
3413 
3414 	MUTEX_INIT(&nat->nat_lock, "nat entry lock");
3415 
3416 	in = nat->nat_ptr;
3417 	nat->nat_ref = nat->nat_me ? 2 : 1;
3418 
3419 	nat->nat_ifnames[0][LIFNAMSIZ - 1] = '\0';
3420 	nat->nat_ifps[0] = ipf_resolvenic(softc, nat->nat_ifnames[0], 4);
3421 
3422 	if (nat->nat_ifnames[1][0] != '\0') {
3423 		nat->nat_ifnames[1][LIFNAMSIZ - 1] = '\0';
3424 		nat->nat_ifps[1] = ipf_resolvenic(softc,
3425 						  nat->nat_ifnames[1], 4);
3426 	} else if (in->in_ifnames[1] != -1) {
3427 		char *name;
3428 
3429 		name = in->in_names + in->in_ifnames[1];
3430 		if (name[1] != '\0' && name[0] != '-' && name[0] != '*') {
3431 			(void) strncpy(nat->nat_ifnames[1],
3432 				       nat->nat_ifnames[0], LIFNAMSIZ);
3433 			nat->nat_ifnames[1][LIFNAMSIZ - 1] = '\0';
3434 			nat->nat_ifps[1] = nat->nat_ifps[0];
3435 		}
3436 	}
3437 	if ((nat->nat_ifps[0] != NULL) && (nat->nat_ifps[0] != (void *)-1)) {
3438 		nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
3439 	}
3440 	if ((nat->nat_ifps[1] != NULL) && (nat->nat_ifps[1] != (void *)-1)) {
3441 		nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
3442 	}
3443 
3444 	ret = ipf_nat_hashtab_add(softc, softn, nat);
3445 	if (ret != 0)
3446 		MUTEX_DESTROY(&nat->nat_lock);
3447 	return ret;
3448 }
3449 
3450 
3451 /* ------------------------------------------------------------------------ */
3452 /* Function:    ipf_nat_hashtab_add                                         */
3453 /* Returns:     int - 0 == sucess, -1 == failure                            */
3454 /* Parameters:  softc(I) - pointer to soft context main structure           */
3455 /*              softn(I) - pointer to NAT context structure                 */
3456 /*              nat(I) - pointer to NAT structure                           */
3457 /* Write Lock:  ipf_nat                                                     */
3458 /*                                                                          */
3459 /* Handle the insertion of a NAT entry into the table/list.                 */
3460 /* ------------------------------------------------------------------------ */
3461 int
3462 ipf_nat_hashtab_add(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, nat_t *nat)
3463 {
3464 	nat_t **natp;
3465 	u_int hv0;
3466 	u_int hv1;
3467 
3468 	hv0 = nat->nat_hv[0] % softn->ipf_nat_table_sz;
3469 	hv1 = nat->nat_hv[1] % softn->ipf_nat_table_sz;
3470 
3471 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen[hv0] >=
3472 	    softn->ipf_nat_maxbucket) {
3473 		DT1(ns_bucket_max_0, int,
3474 		    softn->ipf_nat_stats.ns_side[0].ns_bucketlen[hv0]);
3475 		NBUMPSIDE(0, ns_bucket_max);
3476 		return -1;
3477 	}
3478 
3479 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen[hv1] >=
3480 	    softn->ipf_nat_maxbucket) {
3481 		DT1(ns_bucket_max_1, int,
3482 		    softn->ipf_nat_stats.ns_side[1].ns_bucketlen[hv1]);
3483 		NBUMPSIDE(1, ns_bucket_max);
3484 		return -1;
3485 	}
3486 
3487 	/*
3488 	 * The ordering of operations in the list and hash table insertion
3489 	 * is very important.  The last operation for each task should be
3490 	 * to update the top of the list, after all the "nexts" have been
3491 	 * done so that walking the list while it is being done does not
3492 	 * find strange pointers.
3493 	 *
3494 	 * Global list of NAT instances
3495 	 */
3496 	nat->nat_next = softn->ipf_nat_instances;
3497 	nat->nat_pnext = &softn->ipf_nat_instances;
3498 	if (softn->ipf_nat_instances)
3499 		softn->ipf_nat_instances->nat_pnext = &nat->nat_next;
3500 	softn->ipf_nat_instances = nat;
3501 
3502 	/*
3503 	 * Inbound hash table.
3504 	 */
3505 	natp = &softn->ipf_nat_table[0][hv0];
3506 	nat->nat_phnext[0] = natp;
3507 	nat->nat_hnext[0] = *natp;
3508 	if (*natp) {
3509 		(*natp)->nat_phnext[0] = &nat->nat_hnext[0];
3510 	} else {
3511 		NBUMPSIDE(0, ns_inuse);
3512 	}
3513 	*natp = nat;
3514 	NBUMPSIDE(0, ns_bucketlen[hv0]);
3515 
3516 	/*
3517 	 * Outbound hash table.
3518 	 */
3519 	natp = &softn->ipf_nat_table[1][hv1];
3520 	nat->nat_phnext[1] = natp;
3521 	nat->nat_hnext[1] = *natp;
3522 	if (*natp)
3523 		(*natp)->nat_phnext[1] = &nat->nat_hnext[1];
3524 	else {
3525 		NBUMPSIDE(1, ns_inuse);
3526 	}
3527 	*natp = nat;
3528 	NBUMPSIDE(1, ns_bucketlen[hv1]);
3529 
3530 	ipf_nat_setqueue(softc, softn, nat);
3531 
3532 	if (nat->nat_dir & NAT_OUTBOUND) {
3533 		NBUMPSIDE(1, ns_added);
3534 	} else {
3535 		NBUMPSIDE(0, ns_added);
3536 	}
3537 	softn->ipf_nat_stats.ns_active++;
3538 	return 0;
3539 }
3540 
3541 
3542 /* ------------------------------------------------------------------------ */
3543 /* Function:    ipf_nat_icmperrorlookup                                     */
3544 /* Returns:     nat_t* - point to matching NAT structure                    */
3545 /* Parameters:  fin(I) - pointer to packet information                      */
3546 /*              dir(I) - direction of packet (in/out)                       */
3547 /*                                                                          */
3548 /* Check if the ICMP error message is related to an existing TCP, UDP or    */
3549 /* ICMP query nat entry.  It is assumed that the packet is already of the   */
3550 /* the required length.                                                     */
3551 /* ------------------------------------------------------------------------ */
3552 nat_t *
3553 ipf_nat_icmperrorlookup(fr_info_t *fin, int dir)
3554 {
3555 	ipf_main_softc_t *softc = fin->fin_main_soft;
3556 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3557 	int flags = 0, minlen;
3558 	icmphdr_t *orgicmp;
3559 	nat_stat_side_t *nside;
3560 	tcphdr_t *tcp = NULL;
3561 	u_short data[2];
3562 	nat_t *nat;
3563 	ip_t *oip;
3564 	u_int p;
3565 
3566 	nside = &softn->ipf_nat_stats.ns_side[fin->fin_out];
3567 	/*
3568 	 * Does it at least have the return (basic) IP header ?
3569 	 * Only a basic IP header (no options) should be with an ICMP error
3570 	 * header.  Also, if it's not an error type, then return.
3571 	 */
3572 	if ((fin->fin_hlen != sizeof(ip_t)) || !(fin->fin_flx & FI_ICMPERR)) {
3573 		ATOMIC_INCL(nside->ns_icmp_basic);
3574 		return NULL;
3575 	}
3576 
3577 	/*
3578 	 * Check packet size
3579 	 */
3580 	oip = (ip_t *)((char *)fin->fin_dp + 8);
3581 	minlen = IP_HL(oip) << 2;
3582 	if ((minlen < sizeof(ip_t)) ||
3583 	    (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen)) {
3584 		ATOMIC_INCL(nside->ns_icmp_size);
3585 		return NULL;
3586 	}
3587 
3588 	/*
3589 	 * Is the buffer big enough for all of it ?  It's the size of the IP
3590 	 * header claimed in the encapsulated part which is of concern.  It
3591 	 * may be too big to be in this buffer but not so big that it's
3592 	 * outside the ICMP packet, leading to TCP deref's causing problems.
3593 	 * This is possible because we don't know how big oip_hl is when we
3594 	 * do the pullup early in ipf_check() and thus can't gaurantee it is
3595 	 * all here now.
3596 	 */
3597 #ifdef  ipf_nat_KERNEL
3598 	{
3599 	mb_t *m;
3600 
3601 	m = fin->fin_m;
3602 # if defined(MENTAT)
3603 	if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN >
3604 	    (char *)m->b_wptr) {
3605 		ATOMIC_INCL(nside->ns_icmp_mbuf);
3606 		return NULL;
3607 	}
3608 # else
3609 	if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN >
3610 	    (char *)fin->fin_ip + M_LEN(m)) {
3611 		ATOMIC_INCL(nside->ns_icmp_mbuf);
3612 		return NULL;
3613 	}
3614 # endif
3615 	}
3616 #endif
3617 
3618 	if (fin->fin_daddr != oip->ip_src.s_addr) {
3619 		ATOMIC_INCL(nside->ns_icmp_address);
3620 		return NULL;
3621 	}
3622 
3623 	p = oip->ip_p;
3624 	if (p == IPPROTO_TCP)
3625 		flags = IPN_TCP;
3626 	else if (p == IPPROTO_UDP)
3627 		flags = IPN_UDP;
3628 	else if (p == IPPROTO_ICMP) {
3629 		orgicmp = (icmphdr_t *)((char *)oip + (IP_HL(oip) << 2));
3630 
3631 		/* see if this is related to an ICMP query */
3632 		if (ipf_nat_icmpquerytype(orgicmp->icmp_type)) {
3633 			data[0] = fin->fin_data[0];
3634 			data[1] = fin->fin_data[1];
3635 			fin->fin_data[0] = 0;
3636 			fin->fin_data[1] = orgicmp->icmp_id;
3637 
3638 			flags = IPN_ICMPERR|IPN_ICMPQUERY;
3639 			/*
3640 			 * NOTE : dir refers to the direction of the original
3641 			 *        ip packet. By definition the icmp error
3642 			 *        message flows in the opposite direction.
3643 			 */
3644 			if (dir == NAT_INBOUND)
3645 				nat = ipf_nat_inlookup(fin, flags, p,
3646 						       oip->ip_dst,
3647 						       oip->ip_src);
3648 			else
3649 				nat = ipf_nat_outlookup(fin, flags, p,
3650 							oip->ip_dst,
3651 							oip->ip_src);
3652 			fin->fin_data[0] = data[0];
3653 			fin->fin_data[1] = data[1];
3654 			return nat;
3655 		}
3656 	}
3657 
3658 	if (flags & IPN_TCPUDP) {
3659 		minlen += 8;		/* + 64bits of data to get ports */
3660 		/* TRACE (fin,minlen) */
3661 		if (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen) {
3662 			ATOMIC_INCL(nside->ns_icmp_short);
3663 			return NULL;
3664 		}
3665 
3666 		data[0] = fin->fin_data[0];
3667 		data[1] = fin->fin_data[1];
3668 		tcp = (tcphdr_t *)((char *)oip + (IP_HL(oip) << 2));
3669 		fin->fin_data[0] = ntohs(tcp->th_dport);
3670 		fin->fin_data[1] = ntohs(tcp->th_sport);
3671 
3672 		if (dir == NAT_INBOUND) {
3673 			nat = ipf_nat_inlookup(fin, flags, p, oip->ip_dst,
3674 					       oip->ip_src);
3675 		} else {
3676 			nat = ipf_nat_outlookup(fin, flags, p, oip->ip_dst,
3677 					    oip->ip_src);
3678 		}
3679 		fin->fin_data[0] = data[0];
3680 		fin->fin_data[1] = data[1];
3681 		return nat;
3682 	}
3683 	if (dir == NAT_INBOUND)
3684 		nat = ipf_nat_inlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
3685 	else
3686 		nat = ipf_nat_outlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
3687 
3688 	return nat;
3689 }
3690 
3691 
3692 /* ------------------------------------------------------------------------ */
3693 /* Function:    ipf_nat_icmperror                                           */
3694 /* Returns:     nat_t* - point to matching NAT structure                    */
3695 /* Parameters:  fin(I)    - pointer to packet information                   */
3696 /*              nflags(I) - NAT flags for this packet                       */
3697 /*              dir(I)    - direction of packet (in/out)                    */
3698 /*                                                                          */
3699 /* Fix up an ICMP packet which is an error message for an existing NAT      */
3700 /* session.  This will correct both packet header data and checksums.       */
3701 /*                                                                          */
3702 /* This should *ONLY* be used for incoming ICMP error packets to make sure  */
3703 /* a NAT'd ICMP packet gets correctly recognised.                           */
3704 /* ------------------------------------------------------------------------ */
3705 nat_t *
3706 ipf_nat_icmperror(fr_info_t *fin, u_int *nflags, int dir)
3707 {
3708 	ipf_main_softc_t *softc = fin->fin_main_soft;
3709 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3710 	u_32_t sum1, sum2, sumd, sumd2;
3711 	struct in_addr a1, a2, a3, a4;
3712 	int flags, dlen, odst;
3713 	icmphdr_t *icmp;
3714 	u_short *csump;
3715 	tcphdr_t *tcp;
3716 	nat_t *nat;
3717 	ip_t *oip;
3718 	void *dp;
3719 
3720 	if ((fin->fin_flx & (FI_SHORT|FI_FRAGBODY))) {
3721 		NBUMPSIDED(fin->fin_out, ns_icmp_short);
3722 		return NULL;
3723 	}
3724 
3725 	/*
3726 	 * ipf_nat_icmperrorlookup() will return NULL for `defective' packets.
3727 	 */
3728 	if ((fin->fin_v != 4) || !(nat = ipf_nat_icmperrorlookup(fin, dir))) {
3729 		NBUMPSIDED(fin->fin_out, ns_icmp_notfound);
3730 		return NULL;
3731 	}
3732 
3733 	tcp = NULL;
3734 	csump = NULL;
3735 	flags = 0;
3736 	sumd2 = 0;
3737 	*nflags = IPN_ICMPERR;
3738 	icmp = fin->fin_dp;
3739 	oip = (ip_t *)&icmp->icmp_ip;
3740 	dp = (((char *)oip) + (IP_HL(oip) << 2));
3741 	if (oip->ip_p == IPPROTO_TCP) {
3742 		tcp = (tcphdr_t *)dp;
3743 		csump = (u_short *)&tcp->th_sum;
3744 		flags = IPN_TCP;
3745 	} else if (oip->ip_p == IPPROTO_UDP) {
3746 		udphdr_t *udp;
3747 
3748 		udp = (udphdr_t *)dp;
3749 		tcp = (tcphdr_t *)dp;
3750 		csump = (u_short *)&udp->uh_sum;
3751 		flags = IPN_UDP;
3752 	} else if (oip->ip_p == IPPROTO_ICMP)
3753 		flags = IPN_ICMPQUERY;
3754 	dlen = fin->fin_plen - ((char *)dp - (char *)fin->fin_ip);
3755 
3756 	/*
3757 	 * Need to adjust ICMP header to include the real IP#'s and
3758 	 * port #'s.  Only apply a checksum change relative to the
3759 	 * IP address change as it will be modified again in ipf_nat_checkout
3760 	 * for both address and port.  Two checksum changes are
3761 	 * necessary for the two header address changes.  Be careful
3762 	 * to only modify the checksum once for the port # and twice
3763 	 * for the IP#.
3764 	 */
3765 
3766 	/*
3767 	 * Step 1
3768 	 * Fix the IP addresses in the offending IP packet. You also need
3769 	 * to adjust the IP header checksum of that offending IP packet.
3770 	 *
3771 	 * Normally, you would expect that the ICMP checksum of the
3772 	 * ICMP error message needs to be adjusted as well for the
3773 	 * IP address change in oip.
3774 	 * However, this is a NOP, because the ICMP checksum is
3775 	 * calculated over the complete ICMP packet, which includes the
3776 	 * changed oip IP addresses and oip->ip_sum. However, these
3777 	 * two changes cancel each other out (if the delta for
3778 	 * the IP address is x, then the delta for ip_sum is minus x),
3779 	 * so no change in the icmp_cksum is necessary.
3780 	 *
3781 	 * Inbound ICMP
3782 	 * ------------
3783 	 * MAP rule, SRC=a,DST=b -> SRC=c,DST=b
3784 	 * - response to outgoing packet (a,b)=>(c,b) (OIP_SRC=c,OIP_DST=b)
3785 	 * - OIP_SRC(c)=nat_newsrcip,          OIP_DST(b)=nat_newdstip
3786 	 *=> OIP_SRC(c)=nat_oldsrcip,          OIP_DST(b)=nat_olddstip
3787 	 *
3788 	 * RDR rule, SRC=a,DST=b -> SRC=a,DST=c
3789 	 * - response to outgoing packet (c,a)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3790 	 * - OIP_SRC(b)=nat_olddstip,          OIP_DST(a)=nat_oldsrcip
3791 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3792 	 *
3793 	 * REWRITE out rule, SRC=a,DST=b -> SRC=c,DST=d
3794 	 * - response to outgoing packet (a,b)=>(c,d) (OIP_SRC=c,OIP_DST=d)
3795 	 * - OIP_SRC(c)=nat_newsrcip,          OIP_DST(d)=nat_newdstip
3796 	 *=> OIP_SRC(c)=nat_oldsrcip,          OIP_DST(d)=nat_olddstip
3797 	 *
3798 	 * REWRITE in rule, SRC=a,DST=b -> SRC=c,DST=d
3799 	 * - response to outgoing packet (d,c)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3800 	 * - OIP_SRC(b)=nat_olddstip,          OIP_DST(a)=nat_oldsrcip
3801 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3802 	 *
3803 	 * Outbound ICMP
3804 	 * -------------
3805 	 * MAP rule, SRC=a,DST=b -> SRC=c,DST=b
3806 	 * - response to incoming packet (b,c)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3807 	 * - OIP_SRC(b)=nat_olddstip,          OIP_DST(a)=nat_oldsrcip
3808 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3809 	 *
3810 	 * RDR rule, SRC=a,DST=b -> SRC=a,DST=c
3811 	 * - response to incoming packet (a,b)=>(a,c) (OIP_SRC=a,OIP_DST=c)
3812 	 * - OIP_SRC(a)=nat_newsrcip,          OIP_DST(c)=nat_newdstip
3813 	 *=> OIP_SRC(a)=nat_oldsrcip,          OIP_DST(c)=nat_olddstip
3814 	 *
3815 	 * REWRITE out rule, SRC=a,DST=b -> SRC=c,DST=d
3816 	 * - response to incoming packet (d,c)=>(b,a) (OIP_SRC=c,OIP_DST=d)
3817 	 * - OIP_SRC(c)=nat_olddstip,          OIP_DST(d)=nat_oldsrcip
3818 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3819 	 *
3820 	 * REWRITE in rule, SRC=a,DST=b -> SRC=c,DST=d
3821 	 * - response to incoming packet (a,b)=>(c,d) (OIP_SRC=b,OIP_DST=a)
3822 	 * - OIP_SRC(b)=nat_newsrcip,          OIP_DST(a)=nat_newdstip
3823 	 *=> OIP_SRC(a)=nat_oldsrcip,          OIP_DST(c)=nat_olddstip
3824 	 */
3825 
3826 	if (((fin->fin_out == 0) && ((nat->nat_redir & NAT_MAP) != 0)) ||
3827 	    ((fin->fin_out == 1) && ((nat->nat_redir & NAT_REDIRECT) != 0))) {
3828 		a1.s_addr = ntohl(nat->nat_osrcaddr);
3829 		a4.s_addr = ntohl(oip->ip_src.s_addr);
3830 		a3.s_addr = ntohl(nat->nat_odstaddr);
3831 		a2.s_addr = ntohl(oip->ip_dst.s_addr);
3832 		oip->ip_src.s_addr = htonl(a1.s_addr);
3833 		oip->ip_dst.s_addr = htonl(a3.s_addr);
3834 		odst = 1;
3835 	} else {
3836 		a1.s_addr = ntohl(nat->nat_ndstaddr);
3837 		a2.s_addr = ntohl(oip->ip_dst.s_addr);
3838 		a3.s_addr = ntohl(nat->nat_nsrcaddr);
3839 		a4.s_addr = ntohl(oip->ip_src.s_addr);
3840 		oip->ip_dst.s_addr = htonl(a3.s_addr);
3841 		oip->ip_src.s_addr = htonl(a1.s_addr);
3842 		odst = 0;
3843 	}
3844 	sum1 = 0;
3845 	sum2 = 0;
3846 	sumd = 0;
3847 	CALC_SUMD(a2.s_addr, a3.s_addr, sum1);
3848 	CALC_SUMD(a4.s_addr, a1.s_addr, sum2);
3849 	sumd = sum2 + sum1;
3850 	if (sumd != 0)
3851 		ipf_fix_datacksum(&oip->ip_sum, sumd);
3852 
3853 	sumd2 = sumd;
3854 	sum1 = 0;
3855 	sum2 = 0;
3856 
3857 	/*
3858 	 * Fix UDP pseudo header checksum to compensate for the
3859 	 * IP address change.
3860 	 */
3861 	if (((flags & IPN_TCPUDP) != 0) && (dlen >= 4)) {
3862 		u_32_t sum3, sum4, sumt;
3863 
3864 		/*
3865 		 * Step 2 :
3866 		 * For offending TCP/UDP IP packets, translate the ports as
3867 		 * well, based on the NAT specification. Of course such
3868 		 * a change may be reflected in the ICMP checksum as well.
3869 		 *
3870 		 * Since the port fields are part of the TCP/UDP checksum
3871 		 * of the offending IP packet, you need to adjust that checksum
3872 		 * as well... except that the change in the port numbers should
3873 		 * be offset by the checksum change.  However, the TCP/UDP
3874 		 * checksum will also need to change if there has been an
3875 		 * IP address change.
3876 		 */
3877 		if (odst == 1) {
3878 			sum1 = ntohs(nat->nat_osport);
3879 			sum4 = ntohs(tcp->th_sport);
3880 			sum3 = ntohs(nat->nat_odport);
3881 			sum2 = ntohs(tcp->th_dport);
3882 
3883 			tcp->th_sport = htons(sum1);
3884 			tcp->th_dport = htons(sum3);
3885 		} else {
3886 			sum1 = ntohs(nat->nat_ndport);
3887 			sum2 = ntohs(tcp->th_dport);
3888 			sum3 = ntohs(nat->nat_nsport);
3889 			sum4 = ntohs(tcp->th_sport);
3890 
3891 			tcp->th_dport = htons(sum3);
3892 			tcp->th_sport = htons(sum1);
3893 		}
3894 		CALC_SUMD(sum4, sum1, sumt);
3895 		sumd += sumt;
3896 		CALC_SUMD(sum2, sum3, sumt);
3897 		sumd += sumt;
3898 
3899 		if (sumd != 0 || sumd2 != 0) {
3900 			/*
3901 			 * At this point, sumd is the delta to apply to the
3902 			 * TCP/UDP header, given the changes in both the IP
3903 			 * address and the ports and sumd2 is the delta to
3904 			 * apply to the ICMP header, given the IP address
3905 			 * change delta that may need to be applied to the
3906 			 * TCP/UDP checksum instead.
3907 			 *
3908 			 * If we will both the IP and TCP/UDP checksums
3909 			 * then the ICMP checksum changes by the address
3910 			 * delta applied to the TCP/UDP checksum.  If we
3911 			 * do not change the TCP/UDP checksum them we
3912 			 * apply the delta in ports to the ICMP checksum.
3913 			 */
3914 			if (oip->ip_p == IPPROTO_UDP) {
3915 				if ((dlen >= 8) && (*csump != 0)) {
3916 					ipf_fix_datacksum(csump, sumd);
3917 				} else {
3918 					CALC_SUMD(sum1, sum4, sumd2);
3919 					CALC_SUMD(sum3, sum2, sumt);
3920 					sumd2 += sumt;
3921 				}
3922 			} else if (oip->ip_p == IPPROTO_TCP) {
3923 				if (dlen >= 18) {
3924 					ipf_fix_datacksum(csump, sumd);
3925 				} else {
3926 					CALC_SUMD(sum1, sum4, sumd2);
3927 					CALC_SUMD(sum3, sum2, sumt);
3928 					sumd2 += sumt;
3929 				}
3930 			}
3931 			if (sumd2 != 0) {
3932 				sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3933 				sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3934 				sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3935 				ipf_fix_incksum(0, &icmp->icmp_cksum, sumd2, 0);
3936 			}
3937 		}
3938 	} else if (((flags & IPN_ICMPQUERY) != 0) && (dlen >= 8)) {
3939 		icmphdr_t *orgicmp;
3940 
3941 		/*
3942 		 * XXX - what if this is bogus hl and we go off the end ?
3943 		 * In this case, ipf_nat_icmperrorlookup() will have
3944 		 * returned NULL.
3945 		 */
3946 		orgicmp = (icmphdr_t *)dp;
3947 
3948 		if (odst == 1) {
3949 			if (orgicmp->icmp_id != nat->nat_osport) {
3950 
3951 				/*
3952 				 * Fix ICMP checksum (of the offening ICMP
3953 				 * query packet) to compensate the change
3954 				 * in the ICMP id of the offending ICMP
3955 				 * packet.
3956 				 *
3957 				 * Since you modify orgicmp->icmp_id with
3958 				 * a delta (say x) and you compensate that
3959 				 * in origicmp->icmp_cksum with a delta
3960 				 * minus x, you don't have to adjust the
3961 				 * overall icmp->icmp_cksum
3962 				 */
3963 				sum1 = ntohs(orgicmp->icmp_id);
3964 				sum2 = ntohs(nat->nat_oicmpid);
3965 				CALC_SUMD(sum1, sum2, sumd);
3966 				orgicmp->icmp_id = nat->nat_oicmpid;
3967 				ipf_fix_datacksum(&orgicmp->icmp_cksum, sumd);
3968 			}
3969 		} /* nat_dir == NAT_INBOUND is impossible for icmp queries */
3970 	}
3971 	return nat;
3972 }
3973 
3974 
3975 /*
3976  *       MAP-IN    MAP-OUT   RDR-IN   RDR-OUT
3977  * osrc    X       == src    == src      X
3978  * odst    X       == dst    == dst      X
3979  * nsrc  == dst      X         X      == dst
3980  * ndst  == src      X         X      == src
3981  * MAP = NAT_OUTBOUND, RDR = NAT_INBOUND
3982  */
3983 /*
3984  * NB: these lookups don't lock access to the list, it assumed that it has
3985  * already been done!
3986  */
3987 /* ------------------------------------------------------------------------ */
3988 /* Function:    ipf_nat_inlookup                                            */
3989 /* Returns:     nat_t* - NULL == no match,                                  */
3990 /*                       else pointer to matching NAT entry                 */
3991 /* Parameters:  fin(I)    - pointer to packet information                   */
3992 /*              flags(I)  - NAT flags for this packet                       */
3993 /*              p(I)      - protocol for this packet                        */
3994 /*              src(I)    - source IP address                               */
3995 /*              mapdst(I) - destination IP address                          */
3996 /*                                                                          */
3997 /* Lookup a nat entry based on the mapped destination ip address/port and   */
3998 /* real source address/port.  We use this lookup when receiving a packet,   */
3999 /* we're looking for a table entry, based on the destination address.       */
4000 /*                                                                          */
4001 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY.         */
4002 /*                                                                          */
4003 /* NOTE: IT IS ASSUMED THAT  IS ONLY HELD WITH A READ LOCK WHEN             */
4004 /*       THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags.             */
4005 /*                                                                          */
4006 /* flags   -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if   */
4007 /*            the packet is of said protocol                                */
4008 /* ------------------------------------------------------------------------ */
4009 nat_t *
4010 ipf_nat_inlookup(fr_info_t *fin, u_int flags, u_int p, struct in_addr src,
4011     struct in_addr mapdst)
4012 {
4013 	ipf_main_softc_t *softc = fin->fin_main_soft;
4014 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4015 	u_short sport, dport;
4016 	ipnat_t *ipn;
4017 	nat_t *nat;
4018 	int nflags;
4019 	u_32_t dst;
4020 	void *ifp;
4021 	u_int hv, rhv;
4022 
4023 	ifp = fin->fin_ifp;
4024 	dst = mapdst.s_addr;
4025 
4026 	switch (p)
4027 	{
4028 	case IPPROTO_TCP :
4029 	case IPPROTO_UDP :
4030 		sport = htons(fin->fin_data[0]);
4031 		dport = htons(fin->fin_data[1]);
4032 		break;
4033 	case IPPROTO_ICMP :
4034 		sport = 0;
4035 		dport = fin->fin_data[1];
4036 		break;
4037 	default :
4038 		sport = 0;
4039 		dport = 0;
4040 		break;
4041 	}
4042 
4043 
4044 	if ((flags & SI_WILDP) != 0)
4045 		goto find_in_wild_ports;
4046 
4047 	rhv = NAT_HASH_FN(dst, dport, 0xffffffff);
4048 	rhv = NAT_HASH_FN(src.s_addr, rhv + sport, 0xffffffff);
4049 	hv = rhv % softn->ipf_nat_table_sz;
4050 	nat = softn->ipf_nat_table[1][hv];
4051 	/* TRACE dst, dport, src, sport, hv, nat */
4052 
4053 	for (; nat; nat = nat->nat_hnext[1]) {
4054 		if (nat->nat_ifps[0] != NULL) {
4055 			if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
4056 				continue;
4057 		}
4058 
4059 		if (nat->nat_pr[0] != p)
4060 			continue;
4061 
4062 		switch (nat->nat_dir)
4063 		{
4064 		case NAT_INBOUND :
4065 		case NAT_DIVERTIN :
4066 			if (nat->nat_v[0] != 4)
4067 				continue;
4068 			if (nat->nat_osrcaddr != src.s_addr ||
4069 			    nat->nat_odstaddr != dst)
4070 				continue;
4071 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4072 				if (nat->nat_osport != sport)
4073 					continue;
4074 				if (nat->nat_odport != dport)
4075 					continue;
4076 
4077 			} else if (p == IPPROTO_ICMP) {
4078 				if (nat->nat_oicmpid != dport) {
4079 					continue;
4080 				}
4081 			}
4082 			break;
4083 		case NAT_DIVERTOUT :
4084 			if (nat->nat_dlocal)
4085 				continue;
4086 		case NAT_OUTBOUND :
4087 			if (nat->nat_v[1] != 4)
4088 				continue;
4089 			if (nat->nat_dlocal)
4090 				continue;
4091 			if (nat->nat_dlocal)
4092 				continue;
4093 			if (nat->nat_ndstaddr != src.s_addr ||
4094 			    nat->nat_nsrcaddr != dst)
4095 				continue;
4096 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4097 				if (nat->nat_ndport != sport)
4098 					continue;
4099 				if (nat->nat_nsport != dport)
4100 					continue;
4101 
4102 			} else if (p == IPPROTO_ICMP) {
4103 				if (nat->nat_nicmpid != dport) {
4104 					continue;
4105 				}
4106 			}
4107 			break;
4108 		}
4109 
4110 
4111 		if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4112 			ipn = nat->nat_ptr;
4113 			if ((ipn != NULL) && (nat->nat_aps != NULL))
4114 				if (ipf_proxy_match(fin, nat) != 0)
4115 					continue;
4116 		}
4117 		if ((nat->nat_ifps[0] == NULL) && (ifp != NULL)) {
4118 			nat->nat_ifps[0] = ifp;
4119 			nat->nat_mtu[0] = GETIFMTU_4(ifp);
4120 		}
4121 		return nat;
4122 	}
4123 
4124 	/*
4125 	 * So if we didn't find it but there are wildcard members in the hash
4126 	 * table, go back and look for them.  We do this search and update here
4127 	 * because it is modifying the NAT table and we want to do this only
4128 	 * for the first packet that matches.  The exception, of course, is
4129 	 * for "dummy" (FI_IGNORE) lookups.
4130 	 */
4131 find_in_wild_ports:
4132 	if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH)) {
4133 		NBUMPSIDEX(0, ns_lookup_miss, ns_lookup_miss_0);
4134 		return NULL;
4135 	}
4136 	if (softn->ipf_nat_stats.ns_wilds == 0 || (fin->fin_flx & FI_NOWILD)) {
4137 		NBUMPSIDEX(0, ns_lookup_nowild, ns_lookup_nowild_0);
4138 		return NULL;
4139 	}
4140 
4141 	RWLOCK_EXIT(&softc->ipf_nat);
4142 
4143 	hv = NAT_HASH_FN(dst, 0, 0xffffffff);
4144 	hv = NAT_HASH_FN(src.s_addr, hv, softn->ipf_nat_table_sz);
4145 	WRITE_ENTER(&softc->ipf_nat);
4146 
4147 	nat = softn->ipf_nat_table[1][hv];
4148 	/* TRACE dst, src, hv, nat */
4149 	for (; nat; nat = nat->nat_hnext[1]) {
4150 		if (nat->nat_ifps[0] != NULL) {
4151 			if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
4152 				continue;
4153 		}
4154 
4155 		if (nat->nat_pr[0] != fin->fin_p)
4156 			continue;
4157 
4158 		switch (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND))
4159 		{
4160 		case NAT_INBOUND :
4161 			if (nat->nat_v[0] != 4)
4162 				continue;
4163 			if (nat->nat_osrcaddr != src.s_addr ||
4164 			    nat->nat_odstaddr != dst)
4165 				continue;
4166 			break;
4167 		case NAT_OUTBOUND :
4168 			if (nat->nat_v[1] != 4)
4169 				continue;
4170 			if (nat->nat_ndstaddr != src.s_addr ||
4171 			    nat->nat_nsrcaddr != dst)
4172 				continue;
4173 			break;
4174 		}
4175 
4176 		nflags = nat->nat_flags;
4177 		if (!(nflags & (NAT_TCPUDP|SI_WILDP)))
4178 			continue;
4179 
4180 		if (ipf_nat_wildok(nat, (int)sport, (int)dport, nflags,
4181 				   NAT_INBOUND) == 1) {
4182 			if ((fin->fin_flx & FI_IGNORE) != 0)
4183 				break;
4184 			if ((nflags & SI_CLONE) != 0) {
4185 				nat = ipf_nat_clone(fin, nat);
4186 				if (nat == NULL)
4187 					break;
4188 			} else {
4189 				MUTEX_ENTER(&softn->ipf_nat_new);
4190 				softn->ipf_nat_stats.ns_wilds--;
4191 				MUTEX_EXIT(&softn->ipf_nat_new);
4192 			}
4193 
4194 			if (nat->nat_dir == NAT_INBOUND) {
4195 				if (nat->nat_osport == 0) {
4196 					nat->nat_osport = sport;
4197 					nat->nat_nsport = sport;
4198 				}
4199 				if (nat->nat_odport == 0) {
4200 					nat->nat_odport = dport;
4201 					nat->nat_ndport = dport;
4202 				}
4203 			} else if (nat->nat_dir == NAT_OUTBOUND) {
4204 				if (nat->nat_osport == 0) {
4205 					nat->nat_osport = dport;
4206 					nat->nat_nsport = dport;
4207 				}
4208 				if (nat->nat_odport == 0) {
4209 					nat->nat_odport = sport;
4210 					nat->nat_ndport = sport;
4211 				}
4212 			}
4213 			if ((nat->nat_ifps[0] == NULL) && (ifp != NULL)) {
4214 				nat->nat_ifps[0] = ifp;
4215 				nat->nat_mtu[0] = GETIFMTU_4(ifp);
4216 			}
4217 			nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
4218 			ipf_nat_tabmove(softn, nat);
4219 			break;
4220 		}
4221 	}
4222 
4223 	MUTEX_DOWNGRADE(&softc->ipf_nat);
4224 
4225 	if (nat == NULL) {
4226 		NBUMPSIDE(0, ns_lookup_miss);
4227 	}
4228 	return nat;
4229 }
4230 
4231 
4232 /* ------------------------------------------------------------------------ */
4233 /* Function:    ipf_nat_tabmove                                             */
4234 /* Returns:     Nil                                                         */
4235 /* Parameters:  softn(I) - pointer to NAT context structure                 */
4236 /*              nat(I)   - pointer to NAT structure                         */
4237 /* Write Lock:  ipf_nat                                                     */
4238 /*                                                                          */
4239 /* This function is only called for TCP/UDP NAT table entries where the     */
4240 /* original was placed in the table without hashing on the ports and we now */
4241 /* want to include hashing on port numbers.                                 */
4242 /* ------------------------------------------------------------------------ */
4243 static void
4244 ipf_nat_tabmove(ipf_nat_softc_t *softn, nat_t *nat)
4245 {
4246 	u_int hv0, hv1, rhv0, rhv1;
4247 	natstat_t *nsp;
4248 	nat_t **natp;
4249 
4250 	if (nat->nat_flags & SI_CLONE)
4251 		return;
4252 
4253 	nsp = &softn->ipf_nat_stats;
4254 	/*
4255 	 * Remove the NAT entry from the old location
4256 	 */
4257 	if (nat->nat_hnext[0])
4258 		nat->nat_hnext[0]->nat_phnext[0] = nat->nat_phnext[0];
4259 	*nat->nat_phnext[0] = nat->nat_hnext[0];
4260 	hv0 = nat->nat_hv[0] % softn->ipf_nat_table_sz;
4261 	hv1 = nat->nat_hv[1] % softn->ipf_nat_table_sz;
4262 
4263 	ASSERT(nsp->ns_side[0].ns_bucketlen[hv0] > 0);
4264 	nsp->ns_side[0].ns_bucketlen[hv0]--;
4265 
4266 	if (nat->nat_hnext[1])
4267 		nat->nat_hnext[1]->nat_phnext[1] = nat->nat_phnext[1];
4268 	*nat->nat_phnext[1] = nat->nat_hnext[1];
4269 	ASSERT(nsp->ns_side[1].ns_bucketlen[hv1] > 0);
4270 	nsp->ns_side[1].ns_bucketlen[hv1]--;
4271 
4272 	/*
4273 	 * Add into the NAT table in the new position
4274 	 */
4275 	rhv0 = NAT_HASH_FN(nat->nat_osrcaddr, nat->nat_osport, 0xffffffff);
4276 	rhv0 = NAT_HASH_FN(nat->nat_odstaddr, rhv0 + nat->nat_odport,
4277 			   0xffffffff);
4278 	rhv1 = NAT_HASH_FN(nat->nat_nsrcaddr, nat->nat_nsport, 0xffffffff);
4279 	rhv1 = NAT_HASH_FN(nat->nat_ndstaddr, rhv1 + nat->nat_ndport,
4280 			   0xffffffff);
4281 
4282 	if ((nat->nat_dir & NAT_OUTBOUND) == NAT_OUTBOUND) {
4283 		nat->nat_hv[0] = rhv0;
4284 		nat->nat_hv[1] = rhv1;
4285 	} else {
4286 		nat->nat_hv[0] = rhv1;
4287 		nat->nat_hv[1] = rhv0;
4288 	}
4289 
4290 	hv0 = nat->nat_hv[0] % softn->ipf_nat_table_sz;
4291 	hv1 = nat->nat_hv[1] % softn->ipf_nat_table_sz;
4292 
4293 	/* TRACE nat_osrcaddr, nat_osport, nat_odstaddr, nat_odport, hv0 */
4294 	/* TRACE nat_nsrcaddr, nat_nsport, nat_ndstaddr, nat_ndport, hv1 */
4295 
4296 	natp = &softn->ipf_nat_table[0][hv0];
4297 	if (*natp)
4298 		(*natp)->nat_phnext[0] = &nat->nat_hnext[0];
4299 	nat->nat_phnext[0] = natp;
4300 	nat->nat_hnext[0] = *natp;
4301 	*natp = nat;
4302 	nsp->ns_side[0].ns_bucketlen[hv0]++;
4303 
4304 	natp = &softn->ipf_nat_table[1][hv1];
4305 	if (*natp)
4306 		(*natp)->nat_phnext[1] = &nat->nat_hnext[1];
4307 	nat->nat_phnext[1] = natp;
4308 	nat->nat_hnext[1] = *natp;
4309 	*natp = nat;
4310 	nsp->ns_side[1].ns_bucketlen[hv1]++;
4311 }
4312 
4313 
4314 /* ------------------------------------------------------------------------ */
4315 /* Function:    ipf_nat_outlookup                                           */
4316 /* Returns:     nat_t* - NULL == no match,                                  */
4317 /*                       else pointer to matching NAT entry                 */
4318 /* Parameters:  fin(I)   - pointer to packet information                    */
4319 /*              flags(I) - NAT flags for this packet                        */
4320 /*              p(I)     - protocol for this packet                         */
4321 /*              src(I)   - source IP address                                */
4322 /*              dst(I)   - destination IP address                           */
4323 /*              rw(I)    - 1 == write lock on  held, 0 == read lock.        */
4324 /*                                                                          */
4325 /* Lookup a nat entry based on the source 'real' ip address/port and        */
4326 /* destination address/port.  We use this lookup when sending a packet out, */
4327 /* we're looking for a table entry, based on the source address.            */
4328 /*                                                                          */
4329 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY.         */
4330 /*                                                                          */
4331 /* NOTE: IT IS ASSUMED THAT  IS ONLY HELD WITH A READ LOCK WHEN             */
4332 /*       THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags.             */
4333 /*                                                                          */
4334 /* flags   -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if   */
4335 /*            the packet is of said protocol                                */
4336 /* ------------------------------------------------------------------------ */
4337 nat_t *
4338 ipf_nat_outlookup(fr_info_t *fin, u_int flags, u_int p, struct in_addr src,
4339    struct in_addr dst)
4340 {
4341 	ipf_main_softc_t *softc = fin->fin_main_soft;
4342 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4343 	u_short sport, dport;
4344 	ipnat_t *ipn;
4345 	nat_t *nat;
4346 	void *ifp;
4347 	u_int hv;
4348 
4349 	ifp = fin->fin_ifp;
4350 
4351 	switch (p)
4352 	{
4353 	case IPPROTO_TCP :
4354 	case IPPROTO_UDP :
4355 		sport = htons(fin->fin_data[0]);
4356 		dport = htons(fin->fin_data[1]);
4357 		break;
4358 	case IPPROTO_ICMP :
4359 		sport = 0;
4360 		dport = fin->fin_data[1];
4361 		break;
4362 	default :
4363 		sport = 0;
4364 		dport = 0;
4365 		break;
4366 	}
4367 
4368 	if ((flags & SI_WILDP) != 0)
4369 		goto find_out_wild_ports;
4370 
4371 	hv = NAT_HASH_FN(src.s_addr, sport, 0xffffffff);
4372 	hv = NAT_HASH_FN(dst.s_addr, hv + dport, softn->ipf_nat_table_sz);
4373 	nat = softn->ipf_nat_table[0][hv];
4374 
4375 	/* TRACE src, sport, dst, dport, hv, nat */
4376 
4377 	for (; nat; nat = nat->nat_hnext[0]) {
4378 		if (nat->nat_ifps[1] != NULL) {
4379 			if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
4380 				continue;
4381 		}
4382 
4383 		if (nat->nat_pr[1] != p)
4384 			continue;
4385 
4386 		switch (nat->nat_dir)
4387 		{
4388 		case NAT_INBOUND :
4389 		case NAT_DIVERTIN :
4390 			if (nat->nat_v[1] != 4)
4391 				continue;
4392 			if (nat->nat_ndstaddr != src.s_addr ||
4393 			    nat->nat_nsrcaddr != dst.s_addr)
4394 				continue;
4395 
4396 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4397 				if (nat->nat_ndport != sport)
4398 					continue;
4399 				if (nat->nat_nsport != dport)
4400 					continue;
4401 
4402 			} else if (p == IPPROTO_ICMP) {
4403 				if (nat->nat_nicmpid != dport) {
4404 					continue;
4405 				}
4406 			}
4407 			break;
4408 		case NAT_OUTBOUND :
4409 		case NAT_DIVERTOUT :
4410 			if (nat->nat_v[0] != 4)
4411 				continue;
4412 			if (nat->nat_osrcaddr != src.s_addr ||
4413 			    nat->nat_odstaddr != dst.s_addr)
4414 				continue;
4415 
4416 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4417 				if (nat->nat_odport != dport)
4418 					continue;
4419 				if (nat->nat_osport != sport)
4420 					continue;
4421 
4422 			} else if (p == IPPROTO_ICMP) {
4423 				if (nat->nat_oicmpid != dport) {
4424 					continue;
4425 				}
4426 			}
4427 			break;
4428 		}
4429 
4430 		ipn = nat->nat_ptr;
4431 		if ((ipn != NULL) && (nat->nat_aps != NULL))
4432 			if (ipf_proxy_match(fin, nat) != 0)
4433 				continue;
4434 
4435 		if ((nat->nat_ifps[1] == NULL) && (ifp != NULL)) {
4436 			nat->nat_ifps[1] = ifp;
4437 			nat->nat_mtu[1] = GETIFMTU_4(ifp);
4438 		}
4439 		return nat;
4440 	}
4441 
4442 	/*
4443 	 * So if we didn't find it but there are wildcard members in the hash
4444 	 * table, go back and look for them.  We do this search and update here
4445 	 * because it is modifying the NAT table and we want to do this only
4446 	 * for the first packet that matches.  The exception, of course, is
4447 	 * for "dummy" (FI_IGNORE) lookups.
4448 	 */
4449 find_out_wild_ports:
4450 	if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH)) {
4451 		NBUMPSIDEX(1, ns_lookup_miss, ns_lookup_miss_1);
4452 		return NULL;
4453 	}
4454 	if (softn->ipf_nat_stats.ns_wilds == 0 || (fin->fin_flx & FI_NOWILD)) {
4455 		NBUMPSIDEX(1, ns_lookup_nowild, ns_lookup_nowild_1);
4456 		return NULL;
4457 	}
4458 
4459 	RWLOCK_EXIT(&softc->ipf_nat);
4460 
4461 	hv = NAT_HASH_FN(src.s_addr, 0, 0xffffffff);
4462 	hv = NAT_HASH_FN(dst.s_addr, hv, softn->ipf_nat_table_sz);
4463 
4464 	WRITE_ENTER(&softc->ipf_nat);
4465 
4466 	nat = softn->ipf_nat_table[0][hv];
4467 	for (; nat; nat = nat->nat_hnext[0]) {
4468 		if (nat->nat_ifps[1] != NULL) {
4469 			if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
4470 				continue;
4471 		}
4472 
4473 		if (nat->nat_pr[1] != fin->fin_p)
4474 			continue;
4475 
4476 		switch (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND))
4477 		{
4478 		case NAT_INBOUND :
4479 			if (nat->nat_v[1] != 4)
4480 				continue;
4481 			if (nat->nat_ndstaddr != src.s_addr ||
4482 			    nat->nat_nsrcaddr != dst.s_addr)
4483 				continue;
4484 			break;
4485 		case NAT_OUTBOUND :
4486 			if (nat->nat_v[0] != 4)
4487 				continue;
4488 			if (nat->nat_osrcaddr != src.s_addr ||
4489 			    nat->nat_odstaddr != dst.s_addr)
4490 				continue;
4491 			break;
4492 		}
4493 
4494 		if (!(nat->nat_flags & (NAT_TCPUDP|SI_WILDP)))
4495 			continue;
4496 
4497 		if (ipf_nat_wildok(nat, (int)sport, (int)dport, nat->nat_flags,
4498 				   NAT_OUTBOUND) == 1) {
4499 			if ((fin->fin_flx & FI_IGNORE) != 0)
4500 				break;
4501 			if ((nat->nat_flags & SI_CLONE) != 0) {
4502 				nat = ipf_nat_clone(fin, nat);
4503 				if (nat == NULL)
4504 					break;
4505 			} else {
4506 				MUTEX_ENTER(&softn->ipf_nat_new);
4507 				softn->ipf_nat_stats.ns_wilds--;
4508 				MUTEX_EXIT(&softn->ipf_nat_new);
4509 			}
4510 
4511 			if (nat->nat_dir == NAT_OUTBOUND) {
4512 				if (nat->nat_osport == 0) {
4513 					nat->nat_osport = sport;
4514 					nat->nat_nsport = sport;
4515 				}
4516 				if (nat->nat_odport == 0) {
4517 					nat->nat_odport = dport;
4518 					nat->nat_ndport = dport;
4519 				}
4520 			} else if (nat->nat_dir == NAT_INBOUND) {
4521 				if (nat->nat_osport == 0) {
4522 					nat->nat_osport = dport;
4523 					nat->nat_nsport = dport;
4524 				}
4525 				if (nat->nat_odport == 0) {
4526 					nat->nat_odport = sport;
4527 					nat->nat_ndport = sport;
4528 				}
4529 			}
4530 			if ((nat->nat_ifps[1] == NULL) && (ifp != NULL)) {
4531 				nat->nat_ifps[1] = ifp;
4532 				nat->nat_mtu[1] = GETIFMTU_4(ifp);
4533 			}
4534 			nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
4535 			ipf_nat_tabmove(softn, nat);
4536 			break;
4537 		}
4538 	}
4539 
4540 	MUTEX_DOWNGRADE(&softc->ipf_nat);
4541 
4542 	if (nat == NULL) {
4543 		NBUMPSIDE(1, ns_lookup_miss);
4544 	}
4545 	return nat;
4546 }
4547 
4548 
4549 /* ------------------------------------------------------------------------ */
4550 /* Function:    ipf_nat_lookupredir                                         */
4551 /* Returns:     nat_t* - NULL == no match,                                  */
4552 /*                       else pointer to matching NAT entry                 */
4553 /* Parameters:  softc(I) - pointer to soft context main structure           */
4554 /*              np(I)    - pointer to description of packet to find NAT     */
4555 /*                         table entry for.                                 */
4556 /*                                                                          */
4557 /* Lookup the NAT tables to search for a matching redirect                  */
4558 /* The contents of natlookup_t should imitate those found in a packet that  */
4559 /* would be translated - ie a packet coming in for RDR or going out for MAP.*/
4560 /* We can do the lookup in one of two ways, imitating an inbound or         */
4561 /* outbound  packet.  By default we assume outbound, unless IPN_IN is set.  */
4562 /* For IN, the fields are set as follows:                                   */
4563 /*     nl_real* = source information                                        */
4564 /*     nl_out* = destination information (translated)                       */
4565 /* For an out packet, the fields are set like this:                         */
4566 /*     nl_in* = source information (untranslated)                           */
4567 /*     nl_out* = destination information (translated)                       */
4568 /* ------------------------------------------------------------------------ */
4569 nat_t *
4570 ipf_nat_lookupredir(ipf_main_softc_t *softc, natlookup_t *np)
4571 {
4572 	fr_info_t fi;
4573 	nat_t *nat;
4574 
4575 	bzero((char *)&fi, sizeof(fi));
4576 	fi.fin_main_soft = softc;
4577 	if (np->nl_flags & IPN_IN) {
4578 		fi.fin_data[0] = ntohs(np->nl_realport);
4579 		fi.fin_data[1] = ntohs(np->nl_outport);
4580 	} else {
4581 		fi.fin_data[0] = ntohs(np->nl_inport);
4582 		fi.fin_data[1] = ntohs(np->nl_outport);
4583 	}
4584 	if (np->nl_flags & IPN_TCP)
4585 		fi.fin_p = IPPROTO_TCP;
4586 	else if (np->nl_flags & IPN_UDP)
4587 		fi.fin_p = IPPROTO_UDP;
4588 	else if (np->nl_flags & (IPN_ICMPERR|IPN_ICMPQUERY))
4589 		fi.fin_p = IPPROTO_ICMP;
4590 
4591 	/*
4592 	 * We can do two sorts of lookups:
4593 	 * - IPN_IN: we have the `real' and `out' address, look for `in'.
4594 	 * - default: we have the `in' and `out' address, look for `real'.
4595 	 */
4596 	if (np->nl_flags & IPN_IN) {
4597 		if ((nat = ipf_nat_inlookup(&fi, np->nl_flags, fi.fin_p,
4598 					    np->nl_realip, np->nl_outip))) {
4599 			np->nl_inip = nat->nat_odstip;
4600 			np->nl_inport = nat->nat_odport;
4601 		}
4602 	} else {
4603 		/*
4604 		 * If nl_inip is non null, this is a lookup based on the real
4605 		 * ip address. Else, we use the fake.
4606 		 */
4607 		if ((nat = ipf_nat_outlookup(&fi, np->nl_flags, fi.fin_p,
4608 					 np->nl_inip, np->nl_outip))) {
4609 
4610 			if ((np->nl_flags & IPN_FINDFORWARD) != 0) {
4611 				fr_info_t fin;
4612 				bzero((char *)&fin, sizeof(fin));
4613 				fin.fin_p = nat->nat_pr[0];
4614 				fin.fin_data[0] = ntohs(nat->nat_ndport);
4615 				fin.fin_data[1] = ntohs(nat->nat_nsport);
4616 				if (ipf_nat_inlookup(&fin, np->nl_flags,
4617 						     fin.fin_p, nat->nat_ndstip,
4618 						     nat->nat_nsrcip) != NULL) {
4619 					np->nl_flags &= ~IPN_FINDFORWARD;
4620 				}
4621 			}
4622 
4623 			np->nl_realip = nat->nat_odstip;
4624 			np->nl_realport = nat->nat_odport;
4625 		}
4626  	}
4627 
4628 	return nat;
4629 }
4630 
4631 
4632 /* ------------------------------------------------------------------------ */
4633 /* Function:    ipf_nat_match                                               */
4634 /* Returns:     int - 0 == no match, 1 == match                             */
4635 /* Parameters:  fin(I)   - pointer to packet information                    */
4636 /*              np(I)    - pointer to NAT rule                              */
4637 /*                                                                          */
4638 /* Pull the matching of a packet against a NAT rule out of that complex     */
4639 /* loop inside ipf_nat_checkin() and lay it out properly in its own function. */
4640 /* ------------------------------------------------------------------------ */
4641 static int
4642 ipf_nat_match(fr_info_t *fin, ipnat_t *np)
4643 {
4644 	ipf_main_softc_t *softc = fin->fin_main_soft;
4645 	frtuc_t *ft;
4646 	int match;
4647 
4648 	match = 0;
4649 	switch (np->in_osrcatype)
4650 	{
4651 	case FRI_NORMAL :
4652 		match = ((fin->fin_saddr & np->in_osrcmsk) != np->in_osrcaddr);
4653 		break;
4654 	case FRI_LOOKUP :
4655 		match = (*np->in_osrcfunc)(softc, np->in_osrcptr,
4656 					   4, &fin->fin_saddr, fin->fin_plen);
4657 		break;
4658 	}
4659 	match ^= ((np->in_flags & IPN_NOTSRC) != 0);
4660 	if (match)
4661 		return 0;
4662 
4663 	match = 0;
4664 	switch (np->in_odstatype)
4665 	{
4666 	case FRI_NORMAL :
4667 		match = ((fin->fin_daddr & np->in_odstmsk) != np->in_odstaddr);
4668 		break;
4669 	case FRI_LOOKUP :
4670 		match = (*np->in_odstfunc)(softc, np->in_odstptr,
4671 					   4, &fin->fin_daddr, fin->fin_plen);
4672 		break;
4673 	}
4674 
4675 	match ^= ((np->in_flags & IPN_NOTDST) != 0);
4676 	if (match)
4677 		return 0;
4678 
4679 	ft = &np->in_tuc;
4680 	if (!(fin->fin_flx & FI_TCPUDP) ||
4681 	    (fin->fin_flx & (FI_SHORT|FI_FRAGBODY))) {
4682 		if (ft->ftu_scmp || ft->ftu_dcmp)
4683 			return 0;
4684 		return 1;
4685 	}
4686 
4687 	return ipf_tcpudpchk(&fin->fin_fi, ft);
4688 }
4689 
4690 
4691 /* ------------------------------------------------------------------------ */
4692 /* Function:    ipf_nat_update                                              */
4693 /* Returns:     Nil                                                         */
4694 /* Parameters:  fin(I) - pointer to packet information                      */
4695 /*              nat(I) - pointer to NAT structure                           */
4696 /*                                                                          */
4697 /* Updates the lifetime of a NAT table entry for non-TCP packets.  Must be  */
4698 /* called with fin_rev updated - i.e. after calling ipf_nat_proto().        */
4699 /*                                                                          */
4700 /* This *MUST* be called after ipf_nat_proto() as it expects fin_rev to     */
4701 /* already be set.                                                          */
4702 /* ------------------------------------------------------------------------ */
4703 void
4704 ipf_nat_update(fr_info_t *fin, nat_t *nat)
4705 {
4706 	ipf_main_softc_t *softc = fin->fin_main_soft;
4707 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4708 	ipftq_t *ifq, *ifq2;
4709 	ipftqent_t *tqe;
4710 	ipnat_t *np = nat->nat_ptr;
4711 
4712 	tqe = &nat->nat_tqe;
4713 	ifq = tqe->tqe_ifq;
4714 
4715 	/*
4716 	 * We allow over-riding of NAT timeouts from NAT rules, even for
4717 	 * TCP, however, if it is TCP and there is no rule timeout set,
4718 	 * then do not update the timeout here.
4719 	 */
4720 	if (np != NULL) {
4721 		np->in_bytes[fin->fin_rev] += fin->fin_plen;
4722 		ifq2 = np->in_tqehead[fin->fin_rev];
4723 	} else {
4724 		ifq2 = NULL;
4725 	}
4726 
4727 	if (nat->nat_pr[0] == IPPROTO_TCP && ifq2 == NULL) {
4728 		(void) ipf_tcp_age(&nat->nat_tqe, fin, softn->ipf_nat_tcptq,
4729 				   0, 2);
4730 	} else {
4731 		if (ifq2 == NULL) {
4732 			if (nat->nat_pr[0] == IPPROTO_UDP)
4733 				ifq2 = fin->fin_rev ? &softn->ipf_nat_udpacktq :
4734 						      &softn->ipf_nat_udptq;
4735 			else if (nat->nat_pr[0] == IPPROTO_ICMP ||
4736 				 nat->nat_pr[0] == IPPROTO_ICMPV6)
4737 				ifq2 = fin->fin_rev ? &softn->ipf_nat_icmpacktq:
4738 						      &softn->ipf_nat_icmptq;
4739 			else
4740 				ifq2 = &softn->ipf_nat_iptq;
4741 		}
4742 
4743 		ipf_movequeue(softc->ipf_ticks, tqe, ifq, ifq2);
4744 	}
4745 }
4746 
4747 
4748 /* ------------------------------------------------------------------------ */
4749 /* Function:    ipf_nat_checkout                                            */
4750 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
4751 /*                     0 == no packet translation occurred,                 */
4752 /*                     1 == packet was successfully translated.             */
4753 /* Parameters:  fin(I)   - pointer to packet information                    */
4754 /*              passp(I) - pointer to filtering result flags                */
4755 /*                                                                          */
4756 /* Check to see if an outcoming packet should be changed.  ICMP packets are */
4757 /* first checked to see if they match an existing entry (if an error),      */
4758 /* otherwise a search of the current NAT table is made.  If neither results */
4759 /* in a match then a search for a matching NAT rule is made.  Create a new  */
4760 /* NAT entry if a we matched a NAT rule.  Lastly, actually change the       */
4761 /* packet header(s) as required.                                            */
4762 /* ------------------------------------------------------------------------ */
4763 int
4764 ipf_nat_checkout(fr_info_t *fin, u_32_t *passp)
4765 {
4766 	ipnat_t *np = NULL, *npnext;
4767 	struct ifnet *ifp, *sifp;
4768 	ipf_main_softc_t *softc;
4769 	ipf_nat_softc_t *softn;
4770 	tcphdr_t *tcp = NULL;
4771 	int rval, natfailed;
4772 	u_int nflags = 0;
4773 	u_32_t ipa, iph;
4774 	int natadd = 1;
4775 	frentry_t *fr;
4776 	nat_t *nat;
4777 
4778 	if (fin->fin_v == 6) {
4779 #ifdef USE_INET6
4780 		return ipf_nat6_checkout(fin, passp);
4781 #else
4782 		return 0;
4783 #endif
4784 	}
4785 
4786 	softc = fin->fin_main_soft;
4787 	softn = softc->ipf_nat_soft;
4788 
4789 	if (softn->ipf_nat_lock != 0)
4790 		return 0;
4791 	if (softn->ipf_nat_stats.ns_rules == 0 &&
4792 	    softn->ipf_nat_instances == NULL)
4793 		return 0;
4794 
4795 	natfailed = 0;
4796 	fr = fin->fin_fr;
4797 	sifp = fin->fin_ifp;
4798 	if (fr != NULL) {
4799 		ifp = fr->fr_tifs[fin->fin_rev].fd_ptr;
4800 		if ((ifp != NULL) && (ifp != (void *)-1))
4801 			fin->fin_ifp = ifp;
4802 	}
4803 	ifp = fin->fin_ifp;
4804 
4805 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
4806 		switch (fin->fin_p)
4807 		{
4808 		case IPPROTO_TCP :
4809 			nflags = IPN_TCP;
4810 			break;
4811 		case IPPROTO_UDP :
4812 			nflags = IPN_UDP;
4813 			break;
4814 		case IPPROTO_ICMP :
4815 			/*
4816 			 * This is an incoming packet, so the destination is
4817 			 * the icmp_id and the source port equals 0
4818 			 */
4819 			if ((fin->fin_flx & FI_ICMPQUERY) != 0)
4820 				nflags = IPN_ICMPQUERY;
4821 			break;
4822 		default :
4823 			break;
4824 		}
4825 
4826 		if ((nflags & IPN_TCPUDP))
4827 			tcp = fin->fin_dp;
4828 	}
4829 
4830 	ipa = fin->fin_saddr;
4831 
4832 	READ_ENTER(&softc->ipf_nat);
4833 
4834 	if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
4835 	    (nat = ipf_nat_icmperror(fin, &nflags, NAT_OUTBOUND)))
4836 		/*EMPTY*/;
4837 	else if ((fin->fin_flx & FI_FRAG) && (nat = ipf_frag_natknown(fin)))
4838 		natadd = 0;
4839 	else if ((nat = ipf_nat_outlookup(fin, nflags|NAT_SEARCH,
4840 				      (u_int)fin->fin_p, fin->fin_src,
4841 				      fin->fin_dst))) {
4842 		nflags = nat->nat_flags;
4843 	} else if (fin->fin_off == 0) {
4844 		u_32_t hv, msk, nmsk = 0;
4845 
4846 		/*
4847 		 * If there is no current entry in the nat table for this IP#,
4848 		 * create one for it (if there is a matching rule).
4849 		 */
4850 maskloop:
4851 		msk = softn->ipf_nat_map_active_masks[nmsk];
4852 		iph = ipa & msk;
4853 		hv = NAT_HASH_FN(iph, 0, softn->ipf_nat_maprules_sz);
4854 retry_roundrobin:
4855 		for (np = softn->ipf_nat_map_rules[hv]; np; np = npnext) {
4856 			npnext = np->in_mnext;
4857 			if ((np->in_ifps[1] && (np->in_ifps[1] != ifp)))
4858 				continue;
4859 			if (np->in_v[0] != 4)
4860 				continue;
4861 			if (np->in_pr[1] && (np->in_pr[1] != fin->fin_p))
4862 				continue;
4863 			if ((np->in_flags & IPN_RF) &&
4864 			    !(np->in_flags & nflags))
4865 				continue;
4866 			if (np->in_flags & IPN_FILTER) {
4867 				switch (ipf_nat_match(fin, np))
4868 				{
4869 				case 0 :
4870 					continue;
4871 				case -1 :
4872 					rval = -1;
4873 					goto outmatchfail;
4874 				case 1 :
4875 				default :
4876 					break;
4877 				}
4878 			} else if ((ipa & np->in_osrcmsk) != np->in_osrcaddr)
4879 				continue;
4880 
4881 			if ((fr != NULL) &&
4882 			    !ipf_matchtag(&np->in_tag, &fr->fr_nattag))
4883 				continue;
4884 
4885 			if (np->in_plabel != -1) {
4886 				if (((np->in_flags & IPN_FILTER) == 0) &&
4887 				    (np->in_odport != fin->fin_data[1]))
4888 					continue;
4889 				if (ipf_proxy_ok(fin, tcp, np) == 0)
4890 					continue;
4891 			}
4892 
4893 			if (np->in_flags & IPN_NO) {
4894 				np->in_hits++;
4895 				break;
4896 			}
4897 			MUTEX_ENTER(&softn->ipf_nat_new);
4898 			/*
4899 			 * If we've matched a round-robin rule but it has
4900 			 * moved in the list since we got it, start over as
4901 			 * this is now no longer correct.
4902 			 */
4903 			if (npnext != np->in_mnext) {
4904 				if ((np->in_flags & IPN_ROUNDR) != 0) {
4905 					MUTEX_EXIT(&softn->ipf_nat_new);
4906 					goto retry_roundrobin;
4907 				}
4908 				npnext = np->in_mnext;
4909 			}
4910 
4911 			nat = ipf_nat_add(fin, np, NULL, nflags, NAT_OUTBOUND);
4912 			MUTEX_EXIT(&softn->ipf_nat_new);
4913 			if (nat != NULL) {
4914 				natfailed = 0;
4915 				break;
4916 			}
4917 			natfailed = -1;
4918 		}
4919 		if ((np == NULL) && (nmsk < softn->ipf_nat_map_max)) {
4920 			nmsk++;
4921 			goto maskloop;
4922 		}
4923 	}
4924 
4925 	if (nat != NULL) {
4926 		rval = ipf_nat_out(fin, nat, natadd, nflags);
4927 		if (rval == 1) {
4928 			MUTEX_ENTER(&nat->nat_lock);
4929 			ipf_nat_update(fin, nat);
4930 			nat->nat_bytes[1] += fin->fin_plen;
4931 			nat->nat_pkts[1]++;
4932 			fin->fin_pktnum = nat->nat_pkts[1];
4933 			MUTEX_EXIT(&nat->nat_lock);
4934 		}
4935 	} else
4936 		rval = natfailed;
4937 outmatchfail:
4938 	RWLOCK_EXIT(&softc->ipf_nat);
4939 
4940 	switch (rval)
4941 	{
4942 	case -1 :
4943 		if (passp != NULL) {
4944 			DT1(frb_natv4out, fr_info_t *, fin);
4945 			NBUMPSIDED(1, ns_drop);
4946 			*passp = FR_BLOCK;
4947 			fin->fin_reason = FRB_NATV4;
4948 		}
4949 		fin->fin_flx |= FI_BADNAT;
4950 		NBUMPSIDED(1, ns_badnat);
4951 		break;
4952 	case 0 :
4953 		NBUMPSIDE(1, ns_ignored);
4954 		break;
4955 	case 1 :
4956 		NBUMPSIDE(1, ns_translated);
4957 		break;
4958 	}
4959 	fin->fin_ifp = sifp;
4960 	return rval;
4961 }
4962 
4963 /* ------------------------------------------------------------------------ */
4964 /* Function:    ipf_nat_out                                                 */
4965 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
4966 /*                     1 == packet was successfully translated.             */
4967 /* Parameters:  fin(I)    - pointer to packet information                   */
4968 /*              nat(I)    - pointer to NAT structure                        */
4969 /*              natadd(I) - flag indicating if it is safe to add frag cache */
4970 /*              nflags(I) - NAT flags set for this packet                   */
4971 /*                                                                          */
4972 /* Translate a packet coming "out" on an interface.                         */
4973 /* ------------------------------------------------------------------------ */
4974 int
4975 ipf_nat_out(fr_info_t *fin, nat_t *nat, int natadd, u_32_t nflags)
4976 {
4977 	ipf_main_softc_t *softc = fin->fin_main_soft;
4978 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4979 	icmphdr_t *icmp;
4980 	tcphdr_t *tcp;
4981 	ipnat_t *np;
4982 	int skip;
4983 	int i;
4984 
4985 	tcp = NULL;
4986 	icmp = NULL;
4987 	np = nat->nat_ptr;
4988 
4989 	if ((natadd != 0) && (fin->fin_flx & FI_FRAG) && (np != NULL))
4990 		(void) ipf_frag_natnew(softc, fin, 0, nat);
4991 
4992 	/*
4993 	 * Fix up checksums, not by recalculating them, but
4994 	 * simply computing adjustments.
4995 	 * This is only done for STREAMS based IP implementations where the
4996 	 * checksum has already been calculated by IP.  In all other cases,
4997 	 * IPFilter is called before the checksum needs calculating so there
4998 	 * is no call to modify whatever is in the header now.
4999 	 */
5000 	if (nflags == IPN_ICMPERR) {
5001 		u_32_t s1, s2, sumd, msumd;
5002 
5003 		s1 = LONG_SUM(ntohl(fin->fin_saddr));
5004 		if (nat->nat_dir == NAT_OUTBOUND) {
5005 			s2 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
5006 		} else {
5007 			s2 = LONG_SUM(ntohl(nat->nat_odstaddr));
5008 		}
5009 		CALC_SUMD(s1, s2, sumd);
5010 		msumd = sumd;
5011 
5012 		s1 = LONG_SUM(ntohl(fin->fin_daddr));
5013 		if (nat->nat_dir == NAT_OUTBOUND) {
5014 			s2 = LONG_SUM(ntohl(nat->nat_ndstaddr));
5015 		} else {
5016 			s2 = LONG_SUM(ntohl(nat->nat_osrcaddr));
5017 		}
5018 		CALC_SUMD(s1, s2, sumd);
5019 		msumd += sumd;
5020 
5021 		ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, msumd, 0);
5022 	}
5023 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5024     defined(linux) || defined(BRIDGE_IPF)
5025 	else {
5026 		/*
5027 		 * Strictly speaking, this isn't necessary on BSD
5028 		 * kernels because they do checksum calculation after
5029 		 * this code has run BUT if ipfilter is being used
5030 		 * to do NAT as a bridge, that code doesn't exist.
5031 		 */
5032 		switch (nat->nat_dir)
5033 		{
5034 		case NAT_OUTBOUND :
5035 			ipf_fix_outcksum(fin->fin_cksum & FI_CK_L4PART,
5036 					 &fin->fin_ip->ip_sum,
5037 					 nat->nat_ipsumd, 0);
5038 			break;
5039 
5040 		case NAT_INBOUND :
5041 			ipf_fix_incksum(fin->fin_cksum & FI_CK_L4PART,
5042 					&fin->fin_ip->ip_sum,
5043 					nat->nat_ipsumd, 0);
5044 			break;
5045 
5046 		default :
5047 			break;
5048 		}
5049 	}
5050 #endif
5051 
5052 	/*
5053 	 * Address assignment is after the checksum modification because
5054 	 * we are using the address in the packet for determining the
5055 	 * correct checksum offset (the ICMP error could be coming from
5056 	 * anyone...)
5057 	 */
5058 	switch (nat->nat_dir)
5059 	{
5060 	case NAT_OUTBOUND :
5061 		fin->fin_ip->ip_src = nat->nat_nsrcip;
5062 		fin->fin_saddr = nat->nat_nsrcaddr;
5063 		fin->fin_ip->ip_dst = nat->nat_ndstip;
5064 		fin->fin_daddr = nat->nat_ndstaddr;
5065 		break;
5066 
5067 	case NAT_INBOUND :
5068 		fin->fin_ip->ip_src = nat->nat_odstip;
5069 		fin->fin_saddr = nat->nat_ndstaddr;
5070 		fin->fin_ip->ip_dst = nat->nat_osrcip;
5071 		fin->fin_daddr = nat->nat_nsrcaddr;
5072 		break;
5073 
5074 	case NAT_DIVERTIN :
5075 	    {
5076 		mb_t *m;
5077 
5078 		skip = ipf_nat_decap(fin, nat);
5079 		if (skip <= 0) {
5080 			NBUMPSIDED(1, ns_decap_fail);
5081 			return -1;
5082 		}
5083 
5084 		m = fin->fin_m;
5085 
5086 #if defined(MENTAT) && defined(_KERNEL)
5087 		m->b_rptr += skip;
5088 #else
5089 		m->m_data += skip;
5090 		m->m_len -= skip;
5091 
5092 # ifdef M_PKTHDR
5093 		if (m->m_flags & M_PKTHDR)
5094 			m->m_pkthdr.len -= skip;
5095 # endif
5096 #endif
5097 
5098 		MUTEX_ENTER(&nat->nat_lock);
5099 		ipf_nat_update(fin, nat);
5100 		MUTEX_EXIT(&nat->nat_lock);
5101 		fin->fin_flx |= FI_NATED;
5102 		if (np != NULL && np->in_tag.ipt_num[0] != 0)
5103 			fin->fin_nattag = &np->in_tag;
5104 		return 1;
5105 		/* NOTREACHED */
5106 	    }
5107 
5108 	case NAT_DIVERTOUT :
5109 	    {
5110 		u_32_t s1, s2, sumd;
5111 		udphdr_t *uh;
5112 		ip_t *ip;
5113 		mb_t *m;
5114 
5115 		m = M_DUP(np->in_divmp);
5116 		if (m == NULL) {
5117 			NBUMPSIDED(1, ns_divert_dup);
5118 			return -1;
5119 		}
5120 
5121 		ip = MTOD(m, ip_t *);
5122 		ip->ip_id = htons(ipf_nextipid(fin));
5123 		s2 = ntohs(ip->ip_id);
5124 
5125 		s1 = ip->ip_len;
5126 		ip->ip_len = ntohs(ip->ip_len);
5127 		ip->ip_len += fin->fin_plen;
5128 		ip->ip_len = htons(ip->ip_len);
5129 		s2 += ntohs(ip->ip_len);
5130 		CALC_SUMD(s1, s2, sumd);
5131 
5132 		uh = (udphdr_t *)(ip + 1);
5133 		uh->uh_ulen += fin->fin_plen;
5134 		uh->uh_ulen = htons(uh->uh_ulen);
5135 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5136     defined(linux) || defined(BRIDGE_IPF)
5137 		ipf_fix_outcksum(0, &ip->ip_sum, sumd, 0);
5138 #endif
5139 
5140 		PREP_MB_T(fin, m);
5141 
5142 		fin->fin_src = ip->ip_src;
5143 		fin->fin_dst = ip->ip_dst;
5144 		fin->fin_ip = ip;
5145 		fin->fin_plen += sizeof(ip_t) + 8;	/* UDP + IPv4 hdr */
5146 		fin->fin_dlen += sizeof(ip_t) + 8;	/* UDP + IPv4 hdr */
5147 
5148 		nflags &= ~IPN_TCPUDPICMP;
5149 
5150 		break;
5151 	    }
5152 
5153 	default :
5154 		break;
5155 	}
5156 
5157 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5158 		u_short *csump;
5159 
5160 		if ((nat->nat_nsport != 0) && (nflags & IPN_TCPUDP)) {
5161 			tcp = fin->fin_dp;
5162 
5163 			switch (nat->nat_dir)
5164 			{
5165 			case NAT_OUTBOUND :
5166 				tcp->th_sport = nat->nat_nsport;
5167 				fin->fin_data[0] = ntohs(nat->nat_nsport);
5168 				tcp->th_dport = nat->nat_ndport;
5169 				fin->fin_data[1] = ntohs(nat->nat_ndport);
5170 				break;
5171 
5172 			case NAT_INBOUND :
5173 				tcp->th_sport = nat->nat_odport;
5174 				fin->fin_data[0] = ntohs(nat->nat_odport);
5175 				tcp->th_dport = nat->nat_osport;
5176 				fin->fin_data[1] = ntohs(nat->nat_osport);
5177 				break;
5178 			}
5179 		}
5180 
5181 		if ((nat->nat_oicmpid != 0) && (nflags & IPN_ICMPQUERY)) {
5182 			icmp = fin->fin_dp;
5183 
5184 			switch (nat->nat_dir)
5185 			{
5186 			case NAT_OUTBOUND :
5187 				icmp->icmp_id = nat->nat_nicmpid;
5188 				break;
5189 			case NAT_INBOUND :
5190 				icmp->icmp_id = nat->nat_oicmpid;
5191 				break;
5192 			}
5193 		}
5194 
5195 		csump = ipf_nat_proto(fin, nat, nflags);
5196 
5197 		/*
5198 		 * The above comments do not hold for layer 4 (or higher)
5199 		 * checksums...
5200 		 */
5201 		if (csump != NULL) {
5202 			if (nat->nat_dir == NAT_OUTBOUND)
5203 				ipf_fix_outcksum(fin->fin_cksum, csump,
5204 						 nat->nat_sumd[0],
5205 						 nat->nat_sumd[1] +
5206 						 fin->fin_dlen);
5207 			else
5208 				ipf_fix_incksum(fin->fin_cksum, csump,
5209 						nat->nat_sumd[0],
5210 						nat->nat_sumd[1] +
5211 						fin->fin_dlen);
5212 		}
5213 	}
5214 
5215 	ipf_sync_update(softc, SMC_NAT, fin, nat->nat_sync);
5216 	/* ------------------------------------------------------------- */
5217 	/* A few quick notes:                                            */
5218 	/*      Following are test conditions prior to calling the       */
5219 	/*      ipf_proxy_check routine.                                 */
5220 	/*                                                               */
5221 	/*      A NULL tcp indicates a non TCP/UDP packet.  When dealing */
5222 	/*      with a redirect rule, we attempt to match the packet's   */
5223 	/*      source port against in_dport, otherwise we'd compare the */
5224 	/*      packet's destination.                                    */
5225 	/* ------------------------------------------------------------- */
5226 	if ((np != NULL) && (np->in_apr != NULL)) {
5227 		i = ipf_proxy_check(fin, nat);
5228 		if (i == 0) {
5229 			i = 1;
5230 		} else if (i == -1) {
5231 			NBUMPSIDED(1, ns_ipf_proxy_fail);
5232 		}
5233 	} else {
5234 		i = 1;
5235 	}
5236 	fin->fin_flx |= FI_NATED;
5237 	return i;
5238 }
5239 
5240 
5241 /* ------------------------------------------------------------------------ */
5242 /* Function:    ipf_nat_checkin                                             */
5243 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
5244 /*                     0 == no packet translation occurred,                 */
5245 /*                     1 == packet was successfully translated.             */
5246 /* Parameters:  fin(I)   - pointer to packet information                    */
5247 /*              passp(I) - pointer to filtering result flags                */
5248 /*                                                                          */
5249 /* Check to see if an incoming packet should be changed.  ICMP packets are  */
5250 /* first checked to see if they match an existing entry (if an error),      */
5251 /* otherwise a search of the current NAT table is made.  If neither results */
5252 /* in a match then a search for a matching NAT rule is made.  Create a new  */
5253 /* NAT entry if a we matched a NAT rule.  Lastly, actually change the       */
5254 /* packet header(s) as required.                                            */
5255 /* ------------------------------------------------------------------------ */
5256 int
5257 ipf_nat_checkin(fr_info_t *fin, u_32_t *passp)
5258 {
5259 	ipf_main_softc_t *softc;
5260 	ipf_nat_softc_t *softn;
5261 	u_int nflags, natadd;
5262 	ipnat_t *np, *npnext;
5263 	int rval, natfailed;
5264 	struct ifnet *ifp;
5265 	struct in_addr in;
5266 	icmphdr_t *icmp;
5267 	tcphdr_t *tcp;
5268 	u_short dport;
5269 	nat_t *nat;
5270 	u_32_t iph;
5271 
5272 	softc = fin->fin_main_soft;
5273 	softn = softc->ipf_nat_soft;
5274 
5275 	if (softn->ipf_nat_lock != 0)
5276 		return 0;
5277 	if (softn->ipf_nat_stats.ns_rules == 0 &&
5278 	    softn->ipf_nat_instances == NULL)
5279 		return 0;
5280 
5281 	tcp = NULL;
5282 	icmp = NULL;
5283 	dport = 0;
5284 	natadd = 1;
5285 	nflags = 0;
5286 	natfailed = 0;
5287 	ifp = fin->fin_ifp;
5288 
5289 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5290 		switch (fin->fin_p)
5291 		{
5292 		case IPPROTO_TCP :
5293 			nflags = IPN_TCP;
5294 			break;
5295 		case IPPROTO_UDP :
5296 			nflags = IPN_UDP;
5297 			break;
5298 		case IPPROTO_ICMP :
5299 			icmp = fin->fin_dp;
5300 
5301 			/*
5302 			 * This is an incoming packet, so the destination is
5303 			 * the icmp_id and the source port equals 0
5304 			 */
5305 			if ((fin->fin_flx & FI_ICMPQUERY) != 0) {
5306 				nflags = IPN_ICMPQUERY;
5307 				dport = icmp->icmp_id;
5308 			} break;
5309 		default :
5310 			break;
5311 		}
5312 
5313 		if ((nflags & IPN_TCPUDP)) {
5314 			tcp = fin->fin_dp;
5315 			dport = fin->fin_data[1];
5316 		}
5317 	}
5318 
5319 	in = fin->fin_dst;
5320 
5321 	READ_ENTER(&softc->ipf_nat);
5322 
5323 	if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
5324 	    (nat = ipf_nat_icmperror(fin, &nflags, NAT_INBOUND)))
5325 		/*EMPTY*/;
5326 	else if ((fin->fin_flx & FI_FRAG) && (nat = ipf_frag_natknown(fin)))
5327 		natadd = 0;
5328 	else if ((nat = ipf_nat_inlookup(fin, nflags|NAT_SEARCH,
5329 					 (u_int)fin->fin_p,
5330 					 fin->fin_src, in))) {
5331 		nflags = nat->nat_flags;
5332 	} else if (fin->fin_off == 0) {
5333 		u_32_t hv, msk, rmsk = 0;
5334 
5335 		/*
5336 		 * If there is no current entry in the nat table for this IP#,
5337 		 * create one for it (if there is a matching rule).
5338 		 */
5339 maskloop:
5340 		msk = softn->ipf_nat_rdr_active_masks[rmsk];
5341 		iph = in.s_addr & msk;
5342 		hv = NAT_HASH_FN(iph, 0, softn->ipf_nat_rdrrules_sz);
5343 retry_roundrobin:
5344 		/* TRACE (iph,msk,rmsk,hv,softn->ipf_nat_rdrrules_sz) */
5345 		for (np = softn->ipf_nat_rdr_rules[hv]; np; np = npnext) {
5346 			npnext = np->in_rnext;
5347 			if (np->in_ifps[0] && (np->in_ifps[0] != ifp))
5348 				continue;
5349 			if (np->in_v[0] != 4)
5350 				continue;
5351 			if (np->in_pr[0] && (np->in_pr[0] != fin->fin_p))
5352 				continue;
5353 			if ((np->in_flags & IPN_RF) && !(np->in_flags & nflags))
5354 				continue;
5355 			if (np->in_flags & IPN_FILTER) {
5356 				switch (ipf_nat_match(fin, np))
5357 				{
5358 				case 0 :
5359 					continue;
5360 				case -1 :
5361 					rval = -1;
5362 					goto inmatchfail;
5363 				case 1 :
5364 				default :
5365 					break;
5366 				}
5367 			} else {
5368 				if ((in.s_addr & np->in_odstmsk) !=
5369 				    np->in_odstaddr)
5370 					continue;
5371 				if (np->in_odport &&
5372 				    ((np->in_dtop < dport) ||
5373 				     (dport < np->in_odport)))
5374 					continue;
5375 			}
5376 
5377 			if (np->in_plabel != -1) {
5378 				if (!ipf_proxy_ok(fin, tcp, np)) {
5379 					continue;
5380 				}
5381 			}
5382 
5383 			if (np->in_flags & IPN_NO) {
5384 				np->in_hits++;
5385 				break;
5386 			}
5387 
5388 			MUTEX_ENTER(&softn->ipf_nat_new);
5389 			/*
5390 			 * If we've matched a round-robin rule but it has
5391 			 * moved in the list since we got it, start over as
5392 			 * this is now no longer correct.
5393 			 */
5394 			if (npnext != np->in_rnext) {
5395 				if ((np->in_flags & IPN_ROUNDR) != 0) {
5396 					MUTEX_EXIT(&softn->ipf_nat_new);
5397 					goto retry_roundrobin;
5398 				}
5399 				npnext = np->in_rnext;
5400 			}
5401 
5402 			nat = ipf_nat_add(fin, np, NULL, nflags, NAT_INBOUND);
5403 			MUTEX_EXIT(&softn->ipf_nat_new);
5404 			if (nat != NULL) {
5405 				natfailed = 0;
5406 				break;
5407 			}
5408 			natfailed = -1;
5409 		}
5410 		if ((np == NULL) && (rmsk < softn->ipf_nat_rdr_max)) {
5411 			rmsk++;
5412 			goto maskloop;
5413 		}
5414 	}
5415 
5416 	if (nat != NULL) {
5417 		rval = ipf_nat_in(fin, nat, natadd, nflags);
5418 		if (rval == 1) {
5419 			MUTEX_ENTER(&nat->nat_lock);
5420 			ipf_nat_update(fin, nat);
5421 			nat->nat_bytes[0] += fin->fin_plen;
5422 			nat->nat_pkts[0]++;
5423 			fin->fin_pktnum = nat->nat_pkts[0];
5424 			MUTEX_EXIT(&nat->nat_lock);
5425 		}
5426 	} else
5427 		rval = natfailed;
5428 inmatchfail:
5429 	RWLOCK_EXIT(&softc->ipf_nat);
5430 
5431 	switch (rval)
5432 	{
5433 	case -1 :
5434 		if (passp != NULL) {
5435 			DT1(frb_natv4in, fr_info_t *, fin);
5436 			NBUMPSIDED(0, ns_drop);
5437 			*passp = FR_BLOCK;
5438 			fin->fin_reason = FRB_NATV4;
5439 		}
5440 		fin->fin_flx |= FI_BADNAT;
5441 		NBUMPSIDED(0, ns_badnat);
5442 		break;
5443 	case 0 :
5444 		NBUMPSIDE(0, ns_ignored);
5445 		break;
5446 	case 1 :
5447 		NBUMPSIDE(0, ns_translated);
5448 		break;
5449 	}
5450 	return rval;
5451 }
5452 
5453 
5454 /* ------------------------------------------------------------------------ */
5455 /* Function:    ipf_nat_in                                                  */
5456 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
5457 /*                     1 == packet was successfully translated.             */
5458 /* Parameters:  fin(I)    - pointer to packet information                   */
5459 /*              nat(I)    - pointer to NAT structure                        */
5460 /*              natadd(I) - flag indicating if it is safe to add frag cache */
5461 /*              nflags(I) - NAT flags set for this packet                   */
5462 /* Locks Held:  ipf_nat(READ)                                               */
5463 /*                                                                          */
5464 /* Translate a packet coming "in" on an interface.                          */
5465 /* ------------------------------------------------------------------------ */
5466 int
5467 ipf_nat_in(fr_info_t *fin, nat_t *nat, int natadd, u_32_t nflags)
5468 {
5469 	ipf_main_softc_t *softc = fin->fin_main_soft;
5470 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5471 	u_32_t sumd, ipsumd, sum1, sum2;
5472 	icmphdr_t *icmp;
5473 	tcphdr_t *tcp;
5474 	ipnat_t *np;
5475 	int skip;
5476 	int i;
5477 
5478 	tcp = NULL;
5479 	np = nat->nat_ptr;
5480 	fin->fin_fr = nat->nat_fr;
5481 
5482 	if (np != NULL) {
5483 		if ((natadd != 0) && (fin->fin_flx & FI_FRAG))
5484 			(void) ipf_frag_natnew(softc, fin, 0, nat);
5485 
5486 	/* ------------------------------------------------------------- */
5487 	/* A few quick notes:                                            */
5488 	/*      Following are test conditions prior to calling the       */
5489 	/*      ipf_proxy_check routine.                                 */
5490 	/*                                                               */
5491 	/*      A NULL tcp indicates a non TCP/UDP packet.  When dealing */
5492 	/*      with a map rule, we attempt to match the packet's        */
5493 	/*      source port against in_dport, otherwise we'd compare the */
5494 	/*      packet's destination.                                    */
5495 	/* ------------------------------------------------------------- */
5496 		if (np->in_apr != NULL) {
5497 			i = ipf_proxy_check(fin, nat);
5498 			if (i == -1) {
5499 				NBUMPSIDED(0, ns_ipf_proxy_fail);
5500 				return -1;
5501 			}
5502 		}
5503 	}
5504 
5505 	ipf_sync_update(softc, SMC_NAT, fin, nat->nat_sync);
5506 
5507 	ipsumd = nat->nat_ipsumd;
5508 	/*
5509 	 * Fix up checksums, not by recalculating them, but
5510 	 * simply computing adjustments.
5511 	 * Why only do this for some platforms on inbound packets ?
5512 	 * Because for those that it is done, IP processing is yet to happen
5513 	 * and so the IPv4 header checksum has not yet been evaluated.
5514 	 * Perhaps it should always be done for the benefit of things like
5515 	 * fast forwarding (so that it doesn't need to be recomputed) but with
5516 	 * header checksum offloading, perhaps it is a moot point.
5517 	 */
5518 
5519 	switch (nat->nat_dir)
5520 	{
5521 	case NAT_INBOUND :
5522 		if ((fin->fin_flx & FI_ICMPERR) == 0) {
5523 			fin->fin_ip->ip_src = nat->nat_nsrcip;
5524 			fin->fin_saddr = nat->nat_nsrcaddr;
5525 		} else {
5526 			sum1 = nat->nat_osrcaddr;
5527 			sum2 = nat->nat_nsrcaddr;
5528 			CALC_SUMD(sum1, sum2, sumd);
5529 			ipsumd -= sumd;
5530 		}
5531 		fin->fin_ip->ip_dst = nat->nat_ndstip;
5532 		fin->fin_daddr = nat->nat_ndstaddr;
5533 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5534      defined(__osf__) || defined(linux)
5535 		ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, ipsumd, 0);
5536 #endif
5537 		break;
5538 
5539 	case NAT_OUTBOUND :
5540 		if ((fin->fin_flx & FI_ICMPERR) == 0) {
5541 			fin->fin_ip->ip_src = nat->nat_odstip;
5542 			fin->fin_saddr = nat->nat_odstaddr;
5543 		} else {
5544 			sum1 = nat->nat_odstaddr;
5545 			sum2 = nat->nat_ndstaddr;
5546 			CALC_SUMD(sum1, sum2, sumd);
5547 			ipsumd -= sumd;
5548 		}
5549 		fin->fin_ip->ip_dst = nat->nat_osrcip;
5550 		fin->fin_daddr = nat->nat_osrcaddr;
5551 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5552      defined(__osf__) || defined(linux)
5553 		ipf_fix_incksum(0, &fin->fin_ip->ip_sum, ipsumd, 0);
5554 #endif
5555 		break;
5556 
5557 	case NAT_DIVERTIN :
5558 	    {
5559 		udphdr_t *uh;
5560 		ip_t *ip;
5561 		mb_t *m;
5562 
5563 		m = M_DUP(np->in_divmp);
5564 		if (m == NULL) {
5565 			NBUMPSIDED(0, ns_divert_dup);
5566 			return -1;
5567 		}
5568 
5569 		ip = MTOD(m, ip_t *);
5570 		ip->ip_id = htons(ipf_nextipid(fin));
5571 		sum1 = ntohs(ip->ip_len);
5572 		ip->ip_len = ntohs(ip->ip_len);
5573 		ip->ip_len += fin->fin_plen;
5574 		ip->ip_len = htons(ip->ip_len);
5575 
5576 		uh = (udphdr_t *)(ip + 1);
5577 		uh->uh_ulen += fin->fin_plen;
5578 		uh->uh_ulen = htons(uh->uh_ulen);
5579 
5580 		sum2 = ntohs(ip->ip_id) + ntohs(ip->ip_len);
5581 		sum2 += ntohs(ip->ip_off) & IP_DF;
5582 		CALC_SUMD(sum1, sum2, sumd);
5583 
5584 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5585      defined(__osf__) || defined(linux)
5586 		ipf_fix_outcksum(0, &ip->ip_sum, sumd, 0);
5587 #endif
5588 		PREP_MB_T(fin, m);
5589 
5590 		fin->fin_ip = ip;
5591 		fin->fin_plen += sizeof(ip_t) + 8;	/* UDP + new IPv4 hdr */
5592 		fin->fin_dlen += sizeof(ip_t) + 8;	/* UDP + old IPv4 hdr */
5593 
5594 		nflags &= ~IPN_TCPUDPICMP;
5595 
5596 		break;
5597 	    }
5598 
5599 	case NAT_DIVERTOUT :
5600 	    {
5601 		mb_t *m;
5602 
5603 		skip = ipf_nat_decap(fin, nat);
5604 		if (skip <= 0) {
5605 			NBUMPSIDED(0, ns_decap_fail);
5606 			return -1;
5607 		}
5608 
5609 		m = fin->fin_m;
5610 
5611 #if defined(MENTAT) && defined(_KERNEL)
5612 		m->b_rptr += skip;
5613 #else
5614 		m->m_data += skip;
5615 		m->m_len -= skip;
5616 
5617 # ifdef M_PKTHDR
5618 		if (m->m_flags & M_PKTHDR)
5619 			m->m_pkthdr.len -= skip;
5620 # endif
5621 #endif
5622 
5623 		ipf_nat_update(fin, nat);
5624 		nflags &= ~IPN_TCPUDPICMP;
5625 		fin->fin_flx |= FI_NATED;
5626 		if (np != NULL && np->in_tag.ipt_num[0] != 0)
5627 			fin->fin_nattag = &np->in_tag;
5628 		return 1;
5629 		/* NOTREACHED */
5630 	    }
5631 	}
5632 	if (nflags & IPN_TCPUDP)
5633 		tcp = fin->fin_dp;
5634 
5635 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5636 		u_short *csump;
5637 
5638 		if ((nat->nat_odport != 0) && (nflags & IPN_TCPUDP)) {
5639 			switch (nat->nat_dir)
5640 			{
5641 			case NAT_INBOUND :
5642 				tcp->th_sport = nat->nat_nsport;
5643 				fin->fin_data[0] = ntohs(nat->nat_nsport);
5644 				tcp->th_dport = nat->nat_ndport;
5645 				fin->fin_data[1] = ntohs(nat->nat_ndport);
5646 				break;
5647 
5648 			case NAT_OUTBOUND :
5649 				tcp->th_sport = nat->nat_odport;
5650 				fin->fin_data[0] = ntohs(nat->nat_odport);
5651 				tcp->th_dport = nat->nat_osport;
5652 				fin->fin_data[1] = ntohs(nat->nat_osport);
5653 				break;
5654 			}
5655 		}
5656 
5657 
5658 		if ((nat->nat_oicmpid != 0) && (nflags & IPN_ICMPQUERY)) {
5659 			icmp = fin->fin_dp;
5660 
5661 			switch (nat->nat_dir)
5662 			{
5663 			case NAT_INBOUND :
5664 				icmp->icmp_id = nat->nat_nicmpid;
5665 				break;
5666 			case NAT_OUTBOUND :
5667 				icmp->icmp_id = nat->nat_oicmpid;
5668 				break;
5669 			}
5670 		}
5671 
5672 		csump = ipf_nat_proto(fin, nat, nflags);
5673 
5674 		/*
5675 		 * The above comments do not hold for layer 4 (or higher)
5676 		 * checksums...
5677 		 */
5678 		if (csump != NULL) {
5679 			if (nat->nat_dir == NAT_OUTBOUND)
5680 				ipf_fix_incksum(0, csump, nat->nat_sumd[0], 0);
5681 			else
5682 				ipf_fix_outcksum(0, csump, nat->nat_sumd[0], 0);
5683 		}
5684 	}
5685 
5686 	fin->fin_flx |= FI_NATED;
5687 	if (np != NULL && np->in_tag.ipt_num[0] != 0)
5688 		fin->fin_nattag = &np->in_tag;
5689 	return 1;
5690 }
5691 
5692 
5693 /* ------------------------------------------------------------------------ */
5694 /* Function:    ipf_nat_proto                                               */
5695 /* Returns:     u_short* - pointer to transport header checksum to update,  */
5696 /*                         NULL if the transport protocol is not recognised */
5697 /*                         as needing a checksum update.                    */
5698 /* Parameters:  fin(I)    - pointer to packet information                   */
5699 /*              nat(I)    - pointer to NAT structure                        */
5700 /*              nflags(I) - NAT flags set for this packet                   */
5701 /*                                                                          */
5702 /* Return the pointer to the checksum field for each protocol so understood.*/
5703 /* If support for making other changes to a protocol header is required,    */
5704 /* that is not strictly 'address' translation, such as clamping the MSS in  */
5705 /* TCP down to a specific value, then do it from here.                      */
5706 /* ------------------------------------------------------------------------ */
5707 u_short *
5708 ipf_nat_proto(fr_info_t *fin, nat_t *nat, u_int nflags)
5709 {
5710 	icmphdr_t *icmp;
5711 	u_short *csump;
5712 	tcphdr_t *tcp;
5713 	udphdr_t *udp;
5714 
5715 	csump = NULL;
5716 	if (fin->fin_out == 0) {
5717 		fin->fin_rev = (nat->nat_dir & NAT_OUTBOUND);
5718 	} else {
5719 		fin->fin_rev = ((nat->nat_dir & NAT_OUTBOUND) == 0);
5720 	}
5721 
5722 	switch (fin->fin_p)
5723 	{
5724 	case IPPROTO_TCP :
5725 		tcp = fin->fin_dp;
5726 
5727 		if ((nflags & IPN_TCP) != 0)
5728 			csump = &tcp->th_sum;
5729 
5730 		/*
5731 		 * Do a MSS CLAMPING on a SYN packet,
5732 		 * only deal IPv4 for now.
5733 		 */
5734 		if ((nat->nat_mssclamp != 0) && (tcp->th_flags & TH_SYN) != 0)
5735 			ipf_nat_mssclamp(tcp, nat->nat_mssclamp, fin, csump);
5736 
5737 		break;
5738 
5739 	case IPPROTO_UDP :
5740 		udp = fin->fin_dp;
5741 
5742 		if ((nflags & IPN_UDP) != 0) {
5743 			if (udp->uh_sum != 0)
5744 				csump = &udp->uh_sum;
5745 		}
5746 		break;
5747 
5748 	case IPPROTO_ICMP :
5749 		icmp = fin->fin_dp;
5750 
5751 		if ((nflags & IPN_ICMPQUERY) != 0) {
5752 			if (icmp->icmp_cksum != 0)
5753 				csump = &icmp->icmp_cksum;
5754 		}
5755 		break;
5756 
5757 #ifdef USE_INET6
5758 	case IPPROTO_ICMPV6 :
5759 	    {
5760 		struct icmp6_hdr *icmp6 = (struct icmp6_hdr *)fin->fin_dp;
5761 
5762 		icmp6 = fin->fin_dp;
5763 
5764 		if ((nflags & IPN_ICMPQUERY) != 0) {
5765 			if (icmp6->icmp6_cksum != 0)
5766 				csump = &icmp6->icmp6_cksum;
5767 		}
5768 		break;
5769 	    }
5770 #endif
5771 	}
5772 	return csump;
5773 }
5774 
5775 
5776 /* ------------------------------------------------------------------------ */
5777 /* Function:    ipf_nat_expire                                              */
5778 /* Returns:     Nil                                                         */
5779 /* Parameters:  softc(I) - pointer to soft context main structure           */
5780 /*                                                                          */
5781 /* Check all of the timeout queues for entries at the top which need to be  */
5782 /* expired.                                                                 */
5783 /* ------------------------------------------------------------------------ */
5784 void
5785 ipf_nat_expire(ipf_main_softc_t *softc)
5786 {
5787 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5788 	ipftq_t *ifq, *ifqnext;
5789 	ipftqent_t *tqe, *tqn;
5790 	int i;
5791 	SPL_INT(s);
5792 
5793 	SPL_NET(s);
5794 	WRITE_ENTER(&softc->ipf_nat);
5795 	for (ifq = softn->ipf_nat_tcptq, i = 0; ifq != NULL;
5796 	     ifq = ifq->ifq_next) {
5797 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
5798 			if (tqe->tqe_die > softc->ipf_ticks)
5799 				break;
5800 			tqn = tqe->tqe_next;
5801 			ipf_nat_delete(softc, tqe->tqe_parent, NL_EXPIRE);
5802 		}
5803 	}
5804 
5805 	for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifq->ifq_next) {
5806 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
5807 			if (tqe->tqe_die > softc->ipf_ticks)
5808 				break;
5809 			tqn = tqe->tqe_next;
5810 			ipf_nat_delete(softc, tqe->tqe_parent, NL_EXPIRE);
5811 		}
5812 	}
5813 
5814 	for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifqnext) {
5815 		ifqnext = ifq->ifq_next;
5816 
5817 		if (((ifq->ifq_flags & IFQF_DELETE) != 0) &&
5818 		    (ifq->ifq_ref == 0)) {
5819 			ipf_freetimeoutqueue(softc, ifq);
5820 		}
5821 	}
5822 
5823 	if (softn->ipf_nat_doflush != 0) {
5824 		ipf_nat_extraflush(softc, softn, 2);
5825 		softn->ipf_nat_doflush = 0;
5826 	}
5827 
5828 	RWLOCK_EXIT(&softc->ipf_nat);
5829 	SPL_X(s);
5830 }
5831 
5832 
5833 /* ------------------------------------------------------------------------ */
5834 /* Function:    ipf_nat_sync                                                */
5835 /* Returns:     Nil                                                         */
5836 /* Parameters:  softc(I) - pointer to soft context main structure           */
5837 /*              ifp(I) - pointer to network interface                       */
5838 /*                                                                          */
5839 /* Walk through all of the currently active NAT sessions, looking for those */
5840 /* which need to have their translated address updated.                     */
5841 /* ------------------------------------------------------------------------ */
5842 void
5843 ipf_nat_sync(ipf_main_softc_t *softc, void *ifp)
5844 {
5845 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5846 	u_32_t sum1, sum2, sumd;
5847 	i6addr_t in;
5848 	ipnat_t *n;
5849 	nat_t *nat;
5850 	void *ifp2;
5851 	int idx;
5852 	SPL_INT(s);
5853 
5854 	if (softc->ipf_running <= 0)
5855 		return;
5856 
5857 	/*
5858 	 * Change IP addresses for NAT sessions for any protocol except TCP
5859 	 * since it will break the TCP connection anyway.  The only rules
5860 	 * which will get changed are those which are "map ... -> 0/32",
5861 	 * where the rule specifies the address is taken from the interface.
5862 	 */
5863 	SPL_NET(s);
5864 	WRITE_ENTER(&softc->ipf_nat);
5865 
5866 	if (softc->ipf_running <= 0) {
5867 		RWLOCK_EXIT(&softc->ipf_nat);
5868 		return;
5869 	}
5870 
5871 	for (nat = softn->ipf_nat_instances; nat; nat = nat->nat_next) {
5872 		if ((nat->nat_flags & IPN_TCP) != 0)
5873 			continue;
5874 
5875 		n = nat->nat_ptr;
5876 		if (n != NULL) {
5877 			if (n->in_v[1] == 4) {
5878 				if (n->in_redir & NAT_MAP) {
5879 					if ((n->in_nsrcaddr != 0) ||
5880 					    (n->in_nsrcmsk != 0xffffffff))
5881 						continue;
5882 				} else if (n->in_redir & NAT_REDIRECT) {
5883 					if ((n->in_ndstaddr != 0) ||
5884 					    (n->in_ndstmsk != 0xffffffff))
5885 						continue;
5886 				}
5887 			}
5888 #ifdef USE_INET6
5889 			if (n->in_v[1] == 4) {
5890 				if (n->in_redir & NAT_MAP) {
5891 					if (!IP6_ISZERO(&n->in_nsrcaddr) ||
5892 					    !IP6_ISONES(&n->in_nsrcmsk))
5893 						continue;
5894 				} else if (n->in_redir & NAT_REDIRECT) {
5895 					if (!IP6_ISZERO(&n->in_ndstaddr) ||
5896 					    !IP6_ISONES(&n->in_ndstmsk))
5897 						continue;
5898 				}
5899 			}
5900 #endif
5901 		}
5902 
5903 		if (((ifp == NULL) || (ifp == nat->nat_ifps[0]) ||
5904 		     (ifp == nat->nat_ifps[1]))) {
5905 			nat->nat_ifps[0] = GETIFP(nat->nat_ifnames[0],
5906 						  nat->nat_v[0]);
5907 			if ((nat->nat_ifps[0] != NULL) &&
5908 			    (nat->nat_ifps[0] != (void *)-1)) {
5909 				nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
5910 			}
5911 			if (nat->nat_ifnames[1][0] != '\0') {
5912 				nat->nat_ifps[1] = GETIFP(nat->nat_ifnames[1],
5913 							  nat->nat_v[1]);
5914 			} else {
5915 				nat->nat_ifps[1] = nat->nat_ifps[0];
5916 			}
5917 			if ((nat->nat_ifps[1] != NULL) &&
5918 			    (nat->nat_ifps[1] != (void *)-1)) {
5919 				nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
5920 			}
5921 			ifp2 = nat->nat_ifps[0];
5922 			if (ifp2 == NULL)
5923 				continue;
5924 
5925 			/*
5926 			 * Change the map-to address to be the same as the
5927 			 * new one.
5928 			 */
5929 			sum1 = NATFSUM(nat, nat->nat_v[1], nat_nsrc6);
5930 			if (ipf_ifpaddr(softc, nat->nat_v[0], FRI_NORMAL, ifp2,
5931 				       &in, NULL) != -1) {
5932 				if (nat->nat_v[0] == 4)
5933 					nat->nat_nsrcip = in.in4;
5934 			}
5935 			sum2 = NATFSUM(nat, nat->nat_v[1], nat_nsrc6);
5936 
5937 			if (sum1 == sum2)
5938 				continue;
5939 			/*
5940 			 * Readjust the checksum adjustment to take into
5941 			 * account the new IP#.
5942 			 */
5943 			CALC_SUMD(sum1, sum2, sumd);
5944 			/* XXX - dont change for TCP when solaris does
5945 			 * hardware checksumming.
5946 			 */
5947 			sumd += nat->nat_sumd[0];
5948 			nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
5949 			nat->nat_sumd[1] = nat->nat_sumd[0];
5950 		}
5951 	}
5952 
5953 	for (n = softn->ipf_nat_list; (n != NULL); n = n->in_next) {
5954 		char *base = n->in_names;
5955 
5956 		if ((ifp == NULL) || (n->in_ifps[0] == ifp))
5957 			n->in_ifps[0] = ipf_resolvenic(softc,
5958 						       base + n->in_ifnames[0],
5959 						       n->in_v[0]);
5960 		if ((ifp == NULL) || (n->in_ifps[1] == ifp))
5961 			n->in_ifps[1] = ipf_resolvenic(softc,
5962 						       base + n->in_ifnames[1],
5963 						       n->in_v[1]);
5964 
5965 		if (n->in_redir & NAT_REDIRECT)
5966 			idx = 1;
5967 		else
5968 			idx = 0;
5969 
5970 		if (((ifp == NULL) || (n->in_ifps[idx] == ifp)) &&
5971 		    (n->in_ifps[idx] != NULL &&
5972 		     n->in_ifps[idx] != (void *)-1)) {
5973 
5974 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_osrc,
5975 					     0, n->in_ifps[idx]);
5976 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_odst,
5977 					     0, n->in_ifps[idx]);
5978 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_nsrc,
5979 					     0, n->in_ifps[idx]);
5980 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_ndst,
5981 					     0, n->in_ifps[idx]);
5982 		}
5983 	}
5984 	RWLOCK_EXIT(&softc->ipf_nat);
5985 	SPL_X(s);
5986 }
5987 
5988 
5989 /* ------------------------------------------------------------------------ */
5990 /* Function:    ipf_nat_icmpquerytype                                       */
5991 /* Returns:     int - 1 == success, 0 == failure                            */
5992 /* Parameters:  icmptype(I) - ICMP type number                              */
5993 /*                                                                          */
5994 /* Tests to see if the ICMP type number passed is a query/response type or  */
5995 /* not.                                                                     */
5996 /* ------------------------------------------------------------------------ */
5997 static int
5998 ipf_nat_icmpquerytype(int icmptype)
5999 {
6000 
6001 	/*
6002 	 * For the ICMP query NAT code, it is essential that both the query
6003 	 * and the reply match on the NAT rule. Because the NAT structure
6004 	 * does not keep track of the icmptype, and a single NAT structure
6005 	 * is used for all icmp types with the same src, dest and id, we
6006 	 * simply define the replies as queries as well. The funny thing is,
6007 	 * although it seems silly to call a reply a query, this is exactly
6008 	 * as it is defined in the IPv4 specification
6009 	 */
6010 	switch (icmptype)
6011 	{
6012 	case ICMP_ECHOREPLY:
6013 	case ICMP_ECHO:
6014 	/* route advertisement/sollicitation is currently unsupported: */
6015 	/* it would require rewriting the ICMP data section            */
6016 	case ICMP_TSTAMP:
6017 	case ICMP_TSTAMPREPLY:
6018 	case ICMP_IREQ:
6019 	case ICMP_IREQREPLY:
6020 	case ICMP_MASKREQ:
6021 	case ICMP_MASKREPLY:
6022 		return 1;
6023 	default:
6024 		return 0;
6025 	}
6026 }
6027 
6028 
6029 /* ------------------------------------------------------------------------ */
6030 /* Function:    nat_log                                                     */
6031 /* Returns:     Nil                                                         */
6032 /* Parameters:  softc(I) - pointer to soft context main structure           */
6033 /*              softn(I) - pointer to NAT context structure                 */
6034 /*              nat(I)    - pointer to NAT structure                        */
6035 /*              action(I) - action related to NAT structure being performed */
6036 /*                                                                          */
6037 /* Creates a NAT log entry.                                                 */
6038 /* ------------------------------------------------------------------------ */
6039 void
6040 ipf_nat_log(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, struct nat *nat,
6041     u_int action)
6042 {
6043 #ifdef	IPFILTER_LOG
6044 # ifndef LARGE_NAT
6045 	struct ipnat *np;
6046 	int rulen;
6047 # endif
6048 	struct natlog natl;
6049 	void *items[1];
6050 	size_t sizes[1];
6051 	int types[1];
6052 
6053 	bcopy((char *)&nat->nat_osrc6, (char *)&natl.nl_osrcip,
6054 	      sizeof(natl.nl_osrcip));
6055 	bcopy((char *)&nat->nat_nsrc6, (char *)&natl.nl_nsrcip,
6056 	      sizeof(natl.nl_nsrcip));
6057 	bcopy((char *)&nat->nat_odst6, (char *)&natl.nl_odstip,
6058 	      sizeof(natl.nl_odstip));
6059 	bcopy((char *)&nat->nat_ndst6, (char *)&natl.nl_ndstip,
6060 	      sizeof(natl.nl_ndstip));
6061 
6062 	natl.nl_bytes[0] = nat->nat_bytes[0];
6063 	natl.nl_bytes[1] = nat->nat_bytes[1];
6064 	natl.nl_pkts[0] = nat->nat_pkts[0];
6065 	natl.nl_pkts[1] = nat->nat_pkts[1];
6066 	natl.nl_odstport = nat->nat_odport;
6067 	natl.nl_osrcport = nat->nat_osport;
6068 	natl.nl_nsrcport = nat->nat_nsport;
6069 	natl.nl_ndstport = nat->nat_ndport;
6070 	natl.nl_p[0] = nat->nat_pr[0];
6071 	natl.nl_p[1] = nat->nat_pr[1];
6072 	natl.nl_v[0] = nat->nat_v[0];
6073 	natl.nl_v[1] = nat->nat_v[1];
6074 	natl.nl_type = nat->nat_redir;
6075 	natl.nl_action = action;
6076 	natl.nl_rule = -1;
6077 
6078 	bcopy(nat->nat_ifnames[0], natl.nl_ifnames[0],
6079 	      sizeof(nat->nat_ifnames[0]));
6080 	bcopy(nat->nat_ifnames[1], natl.nl_ifnames[1],
6081 	      sizeof(nat->nat_ifnames[1]));
6082 
6083 # ifndef LARGE_NAT
6084 	if (nat->nat_ptr != NULL) {
6085 		for (rulen = 0, np = softn->ipf_nat_list; np != NULL;
6086 		     np = np->in_next, rulen++)
6087 			if (np == nat->nat_ptr) {
6088 				natl.nl_rule = rulen;
6089 				break;
6090 			}
6091 	}
6092 # endif
6093 	items[0] = &natl;
6094 	sizes[0] = sizeof(natl);
6095 	types[0] = 0;
6096 
6097 	(void) ipf_log_items(softc, IPL_LOGNAT, NULL, items, sizes, types, 1);
6098 #endif
6099 }
6100 
6101 
6102 #if defined(__OpenBSD__)
6103 /* ------------------------------------------------------------------------ */
6104 /* Function:    ipf_nat_ifdetach                                            */
6105 /* Returns:     Nil                                                         */
6106 /* Parameters:  ifp(I) - pointer to network interface                       */
6107 /*                                                                          */
6108 /* Compatibility interface for OpenBSD to trigger the correct updating of   */
6109 /* interface references within IPFilter.                                    */
6110 /* ------------------------------------------------------------------------ */
6111 void
6112 ipf_nat_ifdetach(ifp)
6113 	void *ifp;
6114 {
6115 	ipf_main_softc_t *softc;
6116 
6117 	softc = ipf_get_softc(0);
6118 
6119 	ipf_sync(ifp);
6120 	return;
6121 }
6122 #endif
6123 
6124 
6125 /* ------------------------------------------------------------------------ */
6126 /* Function:    ipf_nat_rule_deref                                          */
6127 /* Returns:     Nil                                                         */
6128 /* Parameters:  softc(I) - pointer to soft context main structure           */
6129 /*              inp(I)   - pointer to pointer to NAT rule                   */
6130 /* Write Locks: ipf_nat                                                     */
6131 /*                                                                          */
6132 /* Dropping the refernce count for a rule means that whatever held the      */
6133 /* pointer to this rule (*inp) is no longer interested in it and when the   */
6134 /* reference count drops to zero, any resources allocated for the rule can  */
6135 /* be released and the rule itself free'd.                                  */
6136 /* ------------------------------------------------------------------------ */
6137 void
6138 ipf_nat_rule_deref(ipf_main_softc_t *softc, ipnat_t **inp)
6139 {
6140 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6141 	ipnat_t *n;
6142 
6143 	n = *inp;
6144 	*inp = NULL;
6145 	n->in_use--;
6146 	if (n->in_use > 0)
6147 		return;
6148 
6149 	if (n->in_apr != NULL)
6150 		ipf_proxy_deref(n->in_apr);
6151 
6152 	ipf_nat_rule_fini(softc, n);
6153 
6154 	if (n->in_redir & NAT_REDIRECT) {
6155 		if ((n->in_flags & IPN_PROXYRULE) == 0) {
6156 			ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules_rdr);
6157 		}
6158 	}
6159 	if (n->in_redir & (NAT_MAP|NAT_MAPBLK)) {
6160 		if ((n->in_flags & IPN_PROXYRULE) == 0) {
6161 			ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules_map);
6162 		}
6163 	}
6164 
6165 	if (n->in_tqehead[0] != NULL) {
6166 		if (ipf_deletetimeoutqueue(n->in_tqehead[0]) == 0) {
6167 			ipf_freetimeoutqueue(softc, n->in_tqehead[1]);
6168 		}
6169 	}
6170 
6171 	if (n->in_tqehead[1] != NULL) {
6172 		if (ipf_deletetimeoutqueue(n->in_tqehead[1]) == 0) {
6173 			ipf_freetimeoutqueue(softc, n->in_tqehead[1]);
6174 		}
6175 	}
6176 
6177 	if ((n->in_flags & IPN_PROXYRULE) == 0) {
6178 		ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules);
6179 	}
6180 
6181 	MUTEX_DESTROY(&n->in_lock);
6182 
6183 	KFREES(n, n->in_size);
6184 
6185 #if SOLARIS && !defined(INSTANCES)
6186 	if (softn->ipf_nat_stats.ns_rules == 0)
6187 		pfil_delayed_copy = 1;
6188 #endif
6189 }
6190 
6191 
6192 /* ------------------------------------------------------------------------ */
6193 /* Function:    ipf_nat_deref                                               */
6194 /* Returns:     Nil                                                         */
6195 /* Parameters:  softc(I) - pointer to soft context main structure           */
6196 /*              natp(I)  - pointer to pointer to NAT table entry            */
6197 /*                                                                          */
6198 /* Decrement the reference counter for this NAT table entry and free it if  */
6199 /* there are no more things using it.                                       */
6200 /*                                                                          */
6201 /* IF nat_ref == 1 when this function is called, then we have an orphan nat */
6202 /* structure *because* it only gets called on paths _after_ nat_ref has been*/
6203 /* incremented.  If nat_ref == 1 then we shouldn't decrement it here        */
6204 /* because nat_delete() will do that and send nat_ref to -1.                */
6205 /*                                                                          */
6206 /* Holding the lock on nat_lock is required to serialise nat_delete() being */
6207 /* called from a NAT flush ioctl with a deref happening because of a packet.*/
6208 /* ------------------------------------------------------------------------ */
6209 void
6210 ipf_nat_deref(ipf_main_softc_t *softc, nat_t **natp)
6211 {
6212 	nat_t *nat;
6213 
6214 	nat = *natp;
6215 	*natp = NULL;
6216 
6217 	MUTEX_ENTER(&nat->nat_lock);
6218 	if (nat->nat_ref > 1) {
6219 		nat->nat_ref--;
6220 		ASSERT(nat->nat_ref >= 0);
6221 		MUTEX_EXIT(&nat->nat_lock);
6222 		return;
6223 	}
6224 	MUTEX_EXIT(&nat->nat_lock);
6225 
6226 	WRITE_ENTER(&softc->ipf_nat);
6227 	ipf_nat_delete(softc, nat, NL_EXPIRE);
6228 	RWLOCK_EXIT(&softc->ipf_nat);
6229 }
6230 
6231 
6232 /* ------------------------------------------------------------------------ */
6233 /* Function:    ipf_nat_clone                                               */
6234 /* Returns:     ipstate_t* - NULL == cloning failed,                        */
6235 /*                           else pointer to new state structure            */
6236 /* Parameters:  fin(I) - pointer to packet information                      */
6237 /*              is(I)  - pointer to master state structure                  */
6238 /* Write Lock:  ipf_nat                                                     */
6239 /*                                                                          */
6240 /* Create a "duplcate" state table entry from the master.                   */
6241 /* ------------------------------------------------------------------------ */
6242 nat_t *
6243 ipf_nat_clone(fr_info_t *fin, nat_t *nat)
6244 {
6245 	ipf_main_softc_t *softc = fin->fin_main_soft;
6246 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6247 	frentry_t *fr;
6248 	nat_t *clone;
6249 	ipnat_t *np;
6250 
6251 	KMALLOC(clone, nat_t *);
6252 	if (clone == NULL) {
6253 		NBUMPSIDED(fin->fin_out, ns_clone_nomem);
6254 		return NULL;
6255 	}
6256 	bcopy((char *)nat, (char *)clone, sizeof(*clone));
6257 
6258 	MUTEX_NUKE(&clone->nat_lock);
6259 
6260 	clone->nat_rev = fin->fin_rev;
6261 	clone->nat_aps = NULL;
6262 	/*
6263 	 * Initialize all these so that ipf_nat_delete() doesn't cause a crash.
6264 	 */
6265 	clone->nat_tqe.tqe_pnext = NULL;
6266 	clone->nat_tqe.tqe_next = NULL;
6267 	clone->nat_tqe.tqe_ifq = NULL;
6268 	clone->nat_tqe.tqe_parent = clone;
6269 
6270 	clone->nat_flags &= ~SI_CLONE;
6271 	clone->nat_flags |= SI_CLONED;
6272 
6273 	if (clone->nat_hm)
6274 		clone->nat_hm->hm_ref++;
6275 
6276 	if (ipf_nat_insert(softc, softn, clone) == -1) {
6277 		KFREE(clone);
6278 		NBUMPSIDED(fin->fin_out, ns_insert_fail);
6279 		return NULL;
6280 	}
6281 
6282 	np = clone->nat_ptr;
6283 	if (np != NULL) {
6284 		if (softn->ipf_nat_logging)
6285 			ipf_nat_log(softc, softn, clone, NL_CLONE);
6286 		np->in_use++;
6287 	}
6288 	fr = clone->nat_fr;
6289 	if (fr != NULL) {
6290 		MUTEX_ENTER(&fr->fr_lock);
6291 		fr->fr_ref++;
6292 		MUTEX_EXIT(&fr->fr_lock);
6293 	}
6294 
6295 
6296 	/*
6297 	 * Because the clone is created outside the normal loop of things and
6298 	 * TCP has special needs in terms of state, initialise the timeout
6299 	 * state of the new NAT from here.
6300 	 */
6301 	if (clone->nat_pr[0] == IPPROTO_TCP) {
6302 		(void) ipf_tcp_age(&clone->nat_tqe, fin, softn->ipf_nat_tcptq,
6303 				   clone->nat_flags, 2);
6304 	}
6305 	clone->nat_sync = ipf_sync_new(softc, SMC_NAT, fin, clone);
6306 	if (softn->ipf_nat_logging)
6307 		ipf_nat_log(softc, softn, clone, NL_CLONE);
6308 	return clone;
6309 }
6310 
6311 
6312 /* ------------------------------------------------------------------------ */
6313 /* Function:   ipf_nat_wildok                                               */
6314 /* Returns:    int - 1 == packet's ports match wildcards                    */
6315 /*                   0 == packet's ports don't match wildcards              */
6316 /* Parameters: nat(I)   - NAT entry                                         */
6317 /*             sport(I) - source port                                       */
6318 /*             dport(I) - destination port                                  */
6319 /*             flags(I) - wildcard flags                                    */
6320 /*             dir(I)   - packet direction                                  */
6321 /*                                                                          */
6322 /* Use NAT entry and packet direction to determine which combination of     */
6323 /* wildcard flags should be used.                                           */
6324 /* ------------------------------------------------------------------------ */
6325 int
6326 ipf_nat_wildok(nat_t *nat, int sport, int dport, int flags, int dir)
6327 {
6328 	/*
6329 	 * When called by       dir is set to
6330 	 * nat_inlookup         NAT_INBOUND (0)
6331 	 * nat_outlookup        NAT_OUTBOUND (1)
6332 	 *
6333 	 * We simply combine the packet's direction in dir with the original
6334 	 * "intended" direction of that NAT entry in nat->nat_dir to decide
6335 	 * which combination of wildcard flags to allow.
6336 	 */
6337 	switch ((dir << 1) | (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND)))
6338 	{
6339 	case 3: /* outbound packet / outbound entry */
6340 		if (((nat->nat_osport == sport) ||
6341 		    (flags & SI_W_SPORT)) &&
6342 		    ((nat->nat_odport == dport) ||
6343 		    (flags & SI_W_DPORT)))
6344 			return 1;
6345 		break;
6346 	case 2: /* outbound packet / inbound entry */
6347 		if (((nat->nat_osport == dport) ||
6348 		    (flags & SI_W_SPORT)) &&
6349 		    ((nat->nat_odport == sport) ||
6350 		    (flags & SI_W_DPORT)))
6351 			return 1;
6352 		break;
6353 	case 1: /* inbound packet / outbound entry */
6354 		if (((nat->nat_osport == dport) ||
6355 		    (flags & SI_W_SPORT)) &&
6356 		    ((nat->nat_odport == sport) ||
6357 		    (flags & SI_W_DPORT)))
6358 			return 1;
6359 		break;
6360 	case 0: /* inbound packet / inbound entry */
6361 		if (((nat->nat_osport == sport) ||
6362 		    (flags & SI_W_SPORT)) &&
6363 		    ((nat->nat_odport == dport) ||
6364 		    (flags & SI_W_DPORT)))
6365 			return 1;
6366 		break;
6367 	default:
6368 		break;
6369 	}
6370 
6371 	return(0);
6372 }
6373 
6374 
6375 /* ------------------------------------------------------------------------ */
6376 /* Function:    nat_mssclamp                                                */
6377 /* Returns:     Nil                                                         */
6378 /* Parameters:  tcp(I)    - pointer to TCP header                           */
6379 /*              maxmss(I) - value to clamp the TCP MSS to                   */
6380 /*              fin(I)    - pointer to packet information                   */
6381 /*              csump(I)  - pointer to TCP checksum                         */
6382 /*                                                                          */
6383 /* Check for MSS option and clamp it if necessary.  If found and changed,   */
6384 /* then the TCP header checksum will be updated to reflect the change in    */
6385 /* the MSS.                                                                 */
6386 /* ------------------------------------------------------------------------ */
6387 static void
6388 ipf_nat_mssclamp(tcphdr_t *tcp, u_32_t maxmss, fr_info_t *fin, u_short *csump)
6389 {
6390 	u_char *cp, *ep, opt;
6391 	int hlen, advance;
6392 	u_32_t mss, sumd;
6393 
6394 	hlen = TCP_OFF(tcp) << 2;
6395 	if (hlen > sizeof(*tcp)) {
6396 		cp = (u_char *)tcp + sizeof(*tcp);
6397 		ep = (u_char *)tcp + hlen;
6398 
6399 		while (cp < ep) {
6400 			opt = cp[0];
6401 			if (opt == TCPOPT_EOL)
6402 				break;
6403 			else if (opt == TCPOPT_NOP) {
6404 				cp++;
6405 				continue;
6406 			}
6407 
6408 			if (cp + 1 >= ep)
6409 				break;
6410 			advance = cp[1];
6411 			if ((cp + advance > ep) || (advance <= 0))
6412 				break;
6413 			switch (opt)
6414 			{
6415 			case TCPOPT_MAXSEG:
6416 				if (advance != 4)
6417 					break;
6418 				mss = cp[2] * 256 + cp[3];
6419 				if (mss > maxmss) {
6420 					cp[2] = maxmss / 256;
6421 					cp[3] = maxmss & 0xff;
6422 					CALC_SUMD(mss, maxmss, sumd);
6423 					ipf_fix_outcksum(0, csump, sumd, 0);
6424 				}
6425 				break;
6426 			default:
6427 				/* ignore unknown options */
6428 				break;
6429 			}
6430 
6431 			cp += advance;
6432 		}
6433 	}
6434 }
6435 
6436 
6437 /* ------------------------------------------------------------------------ */
6438 /* Function:    ipf_nat_setqueue                                            */
6439 /* Returns:     Nil                                                         */
6440 /* Parameters:  softc(I) - pointer to soft context main structure           */
6441 /*              softn(I) - pointer to NAT context structure                 */
6442 /*              nat(I)- pointer to NAT structure                            */
6443 /* Locks:       ipf_nat (read or write)                                     */
6444 /*                                                                          */
6445 /* Put the NAT entry on its default queue entry, using rev as a helped in   */
6446 /* determining which queue it should be placed on.                          */
6447 /* ------------------------------------------------------------------------ */
6448 void
6449 ipf_nat_setqueue(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, nat_t *nat)
6450 {
6451 	ipftq_t *oifq, *nifq;
6452 	int rev = nat->nat_rev;
6453 
6454 	if (nat->nat_ptr != NULL)
6455 		nifq = nat->nat_ptr->in_tqehead[rev];
6456 	else
6457 		nifq = NULL;
6458 
6459 	if (nifq == NULL) {
6460 		switch (nat->nat_pr[0])
6461 		{
6462 		case IPPROTO_UDP :
6463 			nifq = &softn->ipf_nat_udptq;
6464 			break;
6465 		case IPPROTO_ICMP :
6466 			nifq = &softn->ipf_nat_icmptq;
6467 			break;
6468 		case IPPROTO_TCP :
6469 			nifq = softn->ipf_nat_tcptq +
6470 			       nat->nat_tqe.tqe_state[rev];
6471 			break;
6472 		default :
6473 			nifq = &softn->ipf_nat_iptq;
6474 			break;
6475 		}
6476 	}
6477 
6478 	oifq = nat->nat_tqe.tqe_ifq;
6479 	/*
6480 	 * If it's currently on a timeout queue, move it from one queue to
6481 	 * another, else put it on the end of the newly determined queue.
6482 	 */
6483 	if (oifq != NULL)
6484 		ipf_movequeue(softc->ipf_ticks, &nat->nat_tqe, oifq, nifq);
6485 	else
6486 		ipf_queueappend(softc->ipf_ticks, &nat->nat_tqe, nifq, nat);
6487 	return;
6488 }
6489 
6490 
6491 /* ------------------------------------------------------------------------ */
6492 /* Function:    nat_getnext                                                 */
6493 /* Returns:     int - 0 == ok, else error                                   */
6494 /* Parameters:  softc(I) - pointer to soft context main structure           */
6495 /*              t(I)   - pointer to ipftoken structure                      */
6496 /*              itp(I) - pointer to ipfgeniter_t structure                  */
6497 /*                                                                          */
6498 /* Fetch the next nat/ipnat structure pointer from the linked list and      */
6499 /* copy it out to the storage space pointed to by itp_data.  The next item  */
6500 /* in the list to look at is put back in the ipftoken struture.             */
6501 /* ------------------------------------------------------------------------ */
6502 static int
6503 ipf_nat_getnext(ipf_main_softc_t *softc, ipftoken_t *t, ipfgeniter_t *itp,
6504     ipfobj_t *objp)
6505 {
6506 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6507 	hostmap_t *hm, *nexthm = NULL, zerohm;
6508 	ipnat_t *ipn, *nextipnat = NULL, zeroipn;
6509 	nat_t *nat, *nextnat = NULL, zeronat;
6510 	int error = 0;
6511 	void *nnext;
6512 
6513 	if (itp->igi_nitems != 1) {
6514 		IPFERROR(60075);
6515 		return ENOSPC;
6516 	}
6517 
6518 	READ_ENTER(&softc->ipf_nat);
6519 
6520 	switch (itp->igi_type)
6521 	{
6522 	case IPFGENITER_HOSTMAP :
6523 		hm = t->ipt_data;
6524 		if (hm == NULL) {
6525 			nexthm = softn->ipf_hm_maplist;
6526 		} else {
6527 			nexthm = hm->hm_next;
6528 		}
6529 		if (nexthm != NULL) {
6530 			ATOMIC_INC32(nexthm->hm_ref);
6531 			t->ipt_data = nexthm;
6532 		} else {
6533 			bzero(&zerohm, sizeof(zerohm));
6534 			nexthm = &zerohm;
6535 			t->ipt_data = NULL;
6536 		}
6537 		nnext = nexthm->hm_next;
6538 		break;
6539 
6540 	case IPFGENITER_IPNAT :
6541 		ipn = t->ipt_data;
6542 		if (ipn == NULL) {
6543 			nextipnat = softn->ipf_nat_list;
6544 		} else {
6545 			nextipnat = ipn->in_next;
6546 		}
6547 		if (nextipnat != NULL) {
6548 			ATOMIC_INC32(nextipnat->in_use);
6549 			t->ipt_data = nextipnat;
6550 		} else {
6551 			bzero(&zeroipn, sizeof(zeroipn));
6552 			nextipnat = &zeroipn;
6553 			t->ipt_data = NULL;
6554 		}
6555 		nnext = nextipnat->in_next;
6556 		break;
6557 
6558 	case IPFGENITER_NAT :
6559 		nat = t->ipt_data;
6560 		if (nat == NULL) {
6561 			nextnat = softn->ipf_nat_instances;
6562 		} else {
6563 			nextnat = nat->nat_next;
6564 		}
6565 		if (nextnat != NULL) {
6566 			MUTEX_ENTER(&nextnat->nat_lock);
6567 			nextnat->nat_ref++;
6568 			MUTEX_EXIT(&nextnat->nat_lock);
6569 			t->ipt_data = nextnat;
6570 		} else {
6571 			bzero(&zeronat, sizeof(zeronat));
6572 			nextnat = &zeronat;
6573 			t->ipt_data = NULL;
6574 		}
6575 		nnext = nextnat->nat_next;
6576 		break;
6577 
6578 	default :
6579 		RWLOCK_EXIT(&softc->ipf_nat);
6580 		IPFERROR(60055);
6581 		return EINVAL;
6582 	}
6583 
6584 	RWLOCK_EXIT(&softc->ipf_nat);
6585 
6586 	objp->ipfo_ptr = itp->igi_data;
6587 
6588 	switch (itp->igi_type)
6589 	{
6590 	case IPFGENITER_HOSTMAP :
6591 		error = COPYOUT(nexthm, objp->ipfo_ptr, sizeof(*nexthm));
6592 		if (error != 0) {
6593 			IPFERROR(60049);
6594 			error = EFAULT;
6595 		}
6596 		if (hm != NULL) {
6597 			WRITE_ENTER(&softc->ipf_nat);
6598 			ipf_nat_hostmapdel(softc, &hm);
6599 			RWLOCK_EXIT(&softc->ipf_nat);
6600 		}
6601 		break;
6602 
6603 	case IPFGENITER_IPNAT :
6604 		objp->ipfo_size = nextipnat->in_size;
6605 		objp->ipfo_type = IPFOBJ_IPNAT;
6606 		error = ipf_outobjk(softc, objp, nextipnat);
6607 		if (ipn != NULL) {
6608 			WRITE_ENTER(&softc->ipf_nat);
6609 			ipf_nat_rule_deref(softc, &ipn);
6610 			RWLOCK_EXIT(&softc->ipf_nat);
6611 		}
6612 		break;
6613 
6614 	case IPFGENITER_NAT :
6615 		objp->ipfo_size = sizeof(nat_t);
6616 		objp->ipfo_type = IPFOBJ_NAT;
6617 		error = ipf_outobjk(softc, objp, nextnat);
6618 		if (nat != NULL)
6619 			ipf_nat_deref(softc, &nat);
6620 
6621 		break;
6622 	}
6623 
6624 	if (nnext == NULL)
6625 		ipf_token_mark_complete(t);
6626 
6627 	return error;
6628 }
6629 
6630 
6631 /* ------------------------------------------------------------------------ */
6632 /* Function:    nat_extraflush                                              */
6633 /* Returns:     int - 0 == success, -1 == failure                           */
6634 /* Parameters:  softc(I) - pointer to soft context main structure           */
6635 /*              softn(I) - pointer to NAT context structure                 */
6636 /*              which(I) - how to flush the active NAT table                */
6637 /* Write Locks: ipf_nat                                                     */
6638 /*                                                                          */
6639 /* Flush nat tables.  Three actions currently defined:                      */
6640 /* which == 0 : flush all nat table entries                                 */
6641 /* which == 1 : flush TCP connections which have started to close but are   */
6642 /*	      stuck for some reason.                                        */
6643 /* which == 2 : flush TCP connections which have been idle for a long time, */
6644 /*	      starting at > 4 days idle and working back in successive half-*/
6645 /*	      days to at most 12 hours old.  If this fails to free enough   */
6646 /*            slots then work backwards in half hour slots to 30 minutes.   */
6647 /*            If that too fails, then work backwards in 30 second intervals */
6648 /*            for the last 30 minutes to at worst 30 seconds idle.          */
6649 /* ------------------------------------------------------------------------ */
6650 static int
6651 ipf_nat_extraflush(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, int which)
6652 {
6653 	nat_t *nat, **natp;
6654 	ipftqent_t *tqn;
6655 	ipftq_t *ifq;
6656 	int removed;
6657 	SPL_INT(s);
6658 
6659 	removed = 0;
6660 
6661 	SPL_NET(s);
6662 	switch (which)
6663 	{
6664 	case 0 :
6665 		softn->ipf_nat_stats.ns_flush_all++;
6666 		/*
6667 		 * Style 0 flush removes everything...
6668 		 */
6669 		for (natp = &softn->ipf_nat_instances;
6670 		     ((nat = *natp) != NULL); ) {
6671 			ipf_nat_delete(softc, nat, NL_FLUSH);
6672 			removed++;
6673 		}
6674 		break;
6675 
6676 	case 1 :
6677 		softn->ipf_nat_stats.ns_flush_closing++;
6678 		/*
6679 		 * Since we're only interested in things that are closing,
6680 		 * we can start with the appropriate timeout queue.
6681 		 */
6682 		for (ifq = softn->ipf_nat_tcptq + IPF_TCPS_CLOSE_WAIT;
6683 		     ifq != NULL; ifq = ifq->ifq_next) {
6684 
6685 			for (tqn = ifq->ifq_head; tqn != NULL; ) {
6686 				nat = tqn->tqe_parent;
6687 				tqn = tqn->tqe_next;
6688 				if (nat->nat_pr[0] != IPPROTO_TCP ||
6689 				    nat->nat_pr[1] != IPPROTO_TCP)
6690 					break;
6691 				ipf_nat_delete(softc, nat, NL_EXPIRE);
6692 				removed++;
6693 			}
6694 		}
6695 
6696 		/*
6697 		 * Also need to look through the user defined queues.
6698 		 */
6699 		for (ifq = softn->ipf_nat_utqe; ifq != NULL;
6700 		     ifq = ifq->ifq_next) {
6701 			for (tqn = ifq->ifq_head; tqn != NULL; ) {
6702 				nat = tqn->tqe_parent;
6703 				tqn = tqn->tqe_next;
6704 				if (nat->nat_pr[0] != IPPROTO_TCP ||
6705 				    nat->nat_pr[1] != IPPROTO_TCP)
6706 					continue;
6707 
6708 				if ((nat->nat_tcpstate[0] >
6709 				     IPF_TCPS_ESTABLISHED) &&
6710 				    (nat->nat_tcpstate[1] >
6711 				     IPF_TCPS_ESTABLISHED)) {
6712 					ipf_nat_delete(softc, nat, NL_EXPIRE);
6713 					removed++;
6714 				}
6715 			}
6716 		}
6717 		break;
6718 
6719 		/*
6720 		 * Args 5-11 correspond to flushing those particular states
6721 		 * for TCP connections.
6722 		 */
6723 	case IPF_TCPS_CLOSE_WAIT :
6724 	case IPF_TCPS_FIN_WAIT_1 :
6725 	case IPF_TCPS_CLOSING :
6726 	case IPF_TCPS_LAST_ACK :
6727 	case IPF_TCPS_FIN_WAIT_2 :
6728 	case IPF_TCPS_TIME_WAIT :
6729 	case IPF_TCPS_CLOSED :
6730 		softn->ipf_nat_stats.ns_flush_state++;
6731 		tqn = softn->ipf_nat_tcptq[which].ifq_head;
6732 		while (tqn != NULL) {
6733 			nat = tqn->tqe_parent;
6734 			tqn = tqn->tqe_next;
6735 			ipf_nat_delete(softc, nat, NL_FLUSH);
6736 			removed++;
6737 		}
6738 		break;
6739 
6740 	default :
6741 		if (which < 30)
6742 			break;
6743 
6744 		softn->ipf_nat_stats.ns_flush_timeout++;
6745 		/*
6746 		 * Take a large arbitrary number to mean the number of seconds
6747 		 * for which which consider to be the maximum value we'll allow
6748 		 * the expiration to be.
6749 		 */
6750 		which = IPF_TTLVAL(which);
6751 		for (natp = &softn->ipf_nat_instances;
6752 		     ((nat = *natp) != NULL); ) {
6753 			if (softc->ipf_ticks - nat->nat_touched > which) {
6754 				ipf_nat_delete(softc, nat, NL_FLUSH);
6755 				removed++;
6756 			} else
6757 				natp = &nat->nat_next;
6758 		}
6759 		break;
6760 	}
6761 
6762 	if (which != 2) {
6763 		SPL_X(s);
6764 		return removed;
6765 	}
6766 
6767 	softn->ipf_nat_stats.ns_flush_queue++;
6768 
6769 	/*
6770 	 * Asked to remove inactive entries because the table is full, try
6771 	 * again, 3 times, if first attempt failed with a different criteria
6772 	 * each time.  The order tried in must be in decreasing age.
6773 	 * Another alternative is to implement random drop and drop N entries
6774 	 * at random until N have been freed up.
6775 	 */
6776 	if (softc->ipf_ticks - softn->ipf_nat_last_force_flush >
6777 	    IPF_TTLVAL(5)) {
6778 		softn->ipf_nat_last_force_flush = softc->ipf_ticks;
6779 
6780 		removed = ipf_queueflush(softc, ipf_nat_flush_entry,
6781 					 softn->ipf_nat_tcptq,
6782 					 softn->ipf_nat_utqe,
6783 					 &softn->ipf_nat_stats.ns_active,
6784 					 softn->ipf_nat_table_sz,
6785 					 softn->ipf_nat_table_wm_low);
6786 	}
6787 
6788 	SPL_X(s);
6789 	return removed;
6790 }
6791 
6792 
6793 /* ------------------------------------------------------------------------ */
6794 /* Function:    ipf_nat_flush_entry                                         */
6795 /* Returns:     0 - always succeeds                                         */
6796 /* Parameters:  softc(I) - pointer to soft context main structure           */
6797 /*              entry(I) - pointer to NAT entry                             */
6798 /* Write Locks: ipf_nat                                                     */
6799 /*                                                                          */
6800 /* This function is a stepping stone between ipf_queueflush() and           */
6801 /* nat_dlete().  It is used so we can provide a uniform interface via the   */
6802 /* ipf_queueflush() function.  Since the nat_delete() function returns void */
6803 /* we translate that to mean it always succeeds in deleting something.      */
6804 /* ------------------------------------------------------------------------ */
6805 static int
6806 ipf_nat_flush_entry(ipf_main_softc_t *softc, void *entry)
6807 {
6808 	ipf_nat_delete(softc, entry, NL_FLUSH);
6809 	return 0;
6810 }
6811 
6812 
6813 /* ------------------------------------------------------------------------ */
6814 /* Function:    ipf_nat_iterator                                            */
6815 /* Returns:     int - 0 == ok, else error                                   */
6816 /* Parameters:  softc(I) - pointer to soft context main structure           */
6817 /*              token(I) - pointer to ipftoken structure                    */
6818 /*              itp(I)   - pointer to ipfgeniter_t structure                */
6819 /*              obj(I)   - pointer to data description structure            */
6820 /*                                                                          */
6821 /* This function acts as a handler for the SIOCGENITER ioctls that use a    */
6822 /* generic structure to iterate through a list.  There are three different  */
6823 /* linked lists of NAT related information to go through: NAT rules, active */
6824 /* NAT mappings and the NAT fragment cache.                                 */
6825 /* ------------------------------------------------------------------------ */
6826 static int
6827 ipf_nat_iterator(ipf_main_softc_t *softc, ipftoken_t *token, ipfgeniter_t *itp,
6828     ipfobj_t *obj)
6829 {
6830 	int error;
6831 
6832 	if (itp->igi_data == NULL) {
6833 		IPFERROR(60052);
6834 		return EFAULT;
6835 	}
6836 
6837 	switch (itp->igi_type)
6838 	{
6839 	case IPFGENITER_HOSTMAP :
6840 	case IPFGENITER_IPNAT :
6841 	case IPFGENITER_NAT :
6842 		error = ipf_nat_getnext(softc, token, itp, obj);
6843 		break;
6844 
6845 	case IPFGENITER_NATFRAG :
6846 		error = ipf_frag_nat_next(softc, token, itp);
6847 		break;
6848 	default :
6849 		IPFERROR(60053);
6850 		error = EINVAL;
6851 		break;
6852 	}
6853 
6854 	return error;
6855 }
6856 
6857 
6858 /* ------------------------------------------------------------------------ */
6859 /* Function:    ipf_nat_setpending                                          */
6860 /* Returns:     Nil                                                         */
6861 /* Parameters:  softc(I) - pointer to soft context main structure           */
6862 /*              nat(I)   - pointer to NAT structure                         */
6863 /* Locks:       ipf_nat (read or write)                                     */
6864 /*                                                                          */
6865 /* Put the NAT entry on to the pending queue - this queue has a very short  */
6866 /* lifetime where items are put that can't be deleted straight away because */
6867 /* of locking issues but we want to delete them ASAP, anyway.  In calling   */
6868 /* this function, it is assumed that the owner (if there is one, as shown   */
6869 /* by nat_me) is no longer interested in it.                                */
6870 /* ------------------------------------------------------------------------ */
6871 void
6872 ipf_nat_setpending(ipf_main_softc_t *softc, nat_t *nat)
6873 {
6874 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6875 	ipftq_t *oifq;
6876 
6877 	oifq = nat->nat_tqe.tqe_ifq;
6878 	if (oifq != NULL)
6879 		ipf_movequeue(softc->ipf_ticks, &nat->nat_tqe, oifq,
6880 			      &softn->ipf_nat_pending);
6881 	else
6882 		ipf_queueappend(softc->ipf_ticks, &nat->nat_tqe,
6883 				&softn->ipf_nat_pending, nat);
6884 
6885 	if (nat->nat_me != NULL) {
6886 		*nat->nat_me = NULL;
6887 		nat->nat_me = NULL;
6888 		nat->nat_ref--;
6889 		ASSERT(nat->nat_ref >= 0);
6890 	}
6891 }
6892 
6893 
6894 /* ------------------------------------------------------------------------ */
6895 /* Function:    nat_newrewrite                                              */
6896 /* Returns:     int - -1 == error, 0 == success (no move), 1 == success and */
6897 /*                    allow rule to be moved if IPN_ROUNDR is set.          */
6898 /* Parameters:  fin(I) - pointer to packet information                      */
6899 /*              nat(I) - pointer to NAT entry                               */
6900 /*              ni(I)  - pointer to structure with misc. information needed */
6901 /*                       to create new NAT entry.                           */
6902 /* Write Lock:  ipf_nat                                                     */
6903 /*                                                                          */
6904 /* This function is responsible for setting up an active NAT session where  */
6905 /* we are changing both the source and destination parameters at the same   */
6906 /* time.  The loop in here works differently to elsewhere - each iteration  */
6907 /* is responsible for changing a single parameter that can be incremented.  */
6908 /* So one pass may increase the source IP#, next source port, next dest. IP#*/
6909 /* and the last destination port for a total of 4 iterations to try each.   */
6910 /* This is done to try and exhaustively use the translation space available.*/
6911 /* ------------------------------------------------------------------------ */
6912 static int
6913 ipf_nat_newrewrite(fr_info_t *fin, nat_t *nat, natinfo_t *nai)
6914 {
6915 	int src_search = 1;
6916 	int dst_search = 1;
6917 	fr_info_t frnat;
6918 	u_32_t flags;
6919 	u_short swap;
6920 	ipnat_t *np;
6921 	nat_t *natl;
6922 	int l = 0;
6923 	int changed;
6924 
6925 	natl = NULL;
6926 	changed = -1;
6927 	np = nai->nai_np;
6928 	flags = nat->nat_flags;
6929 	bcopy((char *)fin, (char *)&frnat, sizeof(*fin));
6930 
6931 	nat->nat_hm = NULL;
6932 
6933 	do {
6934 		changed = -1;
6935 		/* TRACE (l, src_search, dst_search, np) */
6936 
6937 		if ((src_search == 0) && (np->in_spnext == 0) &&
6938 		    (dst_search == 0) && (np->in_dpnext == 0)) {
6939 			if (l > 0)
6940 				return -1;
6941 		}
6942 
6943 		/*
6944 		 * Find a new source address
6945 		 */
6946 		if (ipf_nat_nextaddr(fin, &np->in_nsrc, &frnat.fin_saddr,
6947 				     &frnat.fin_saddr) == -1) {
6948 			return -1;
6949 		}
6950 
6951 		if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0xffffffff)) {
6952 			src_search = 0;
6953 			if (np->in_stepnext == 0)
6954 				np->in_stepnext = 1;
6955 
6956 		} else if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0)) {
6957 			src_search = 0;
6958 			if (np->in_stepnext == 0)
6959 				np->in_stepnext = 1;
6960 
6961 		} else if (np->in_nsrcmsk == 0xffffffff) {
6962 			src_search = 0;
6963 			if (np->in_stepnext == 0)
6964 				np->in_stepnext = 1;
6965 
6966 		} else if (np->in_nsrcmsk != 0xffffffff) {
6967 			if (np->in_stepnext == 0 && changed == -1) {
6968 				np->in_snip++;
6969 				np->in_stepnext++;
6970 				changed = 0;
6971 			}
6972 		}
6973 
6974 		if ((flags & IPN_TCPUDPICMP) != 0) {
6975 			if (np->in_spnext != 0)
6976 				frnat.fin_data[0] = np->in_spnext;
6977 
6978 			/*
6979 			 * Standard port translation.  Select next port.
6980 			 */
6981 			if ((flags & IPN_FIXEDSPORT) != 0) {
6982 				np->in_stepnext = 2;
6983 			} else if ((np->in_stepnext == 1) &&
6984 				   (changed == -1) && (natl != NULL)) {
6985 				np->in_spnext++;
6986 				np->in_stepnext++;
6987 				changed = 1;
6988 				if (np->in_spnext > np->in_spmax)
6989 					np->in_spnext = np->in_spmin;
6990 			}
6991 		} else {
6992 			np->in_stepnext = 2;
6993 		}
6994 		np->in_stepnext &= 0x3;
6995 
6996 		/*
6997 		 * Find a new destination address
6998 		 */
6999 		/* TRACE (fin, np, l, frnat) */
7000 
7001 		if (ipf_nat_nextaddr(fin, &np->in_ndst, &frnat.fin_daddr,
7002 				     &frnat.fin_daddr) == -1)
7003 			return -1;
7004 		if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0xffffffff)) {
7005 			dst_search = 0;
7006 			if (np->in_stepnext == 2)
7007 				np->in_stepnext = 3;
7008 
7009 		} else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0)) {
7010 			dst_search = 0;
7011 			if (np->in_stepnext == 2)
7012 				np->in_stepnext = 3;
7013 
7014 		} else if (np->in_ndstmsk == 0xffffffff) {
7015 			dst_search = 0;
7016 			if (np->in_stepnext == 2)
7017 				np->in_stepnext = 3;
7018 
7019 		} else if (np->in_ndstmsk != 0xffffffff) {
7020 			if ((np->in_stepnext == 2) && (changed == -1) &&
7021 			    (natl != NULL)) {
7022 				changed = 2;
7023 				np->in_stepnext++;
7024 				np->in_dnip++;
7025 			}
7026 		}
7027 
7028 		if ((flags & IPN_TCPUDPICMP) != 0) {
7029 			if (np->in_dpnext != 0)
7030 				frnat.fin_data[1] = np->in_dpnext;
7031 
7032 			/*
7033 			 * Standard port translation.  Select next port.
7034 			 */
7035 			if ((flags & IPN_FIXEDDPORT) != 0) {
7036 				np->in_stepnext = 0;
7037 			} else if (np->in_stepnext == 3 && changed == -1) {
7038 				np->in_dpnext++;
7039 				np->in_stepnext++;
7040 				changed = 3;
7041 				if (np->in_dpnext > np->in_dpmax)
7042 					np->in_dpnext = np->in_dpmin;
7043 			}
7044 		} else {
7045 			if (np->in_stepnext == 3)
7046 				np->in_stepnext = 0;
7047 		}
7048 
7049 		/* TRACE (frnat) */
7050 
7051 		/*
7052 		 * Here we do a lookup of the connection as seen from
7053 		 * the outside.  If an IP# pair already exists, try
7054 		 * again.  So if you have A->B becomes C->B, you can
7055 		 * also have D->E become C->E but not D->B causing
7056 		 * another C->B.  Also take protocol and ports into
7057 		 * account when determining whether a pre-existing
7058 		 * NAT setup will cause an external conflict where
7059 		 * this is appropriate.
7060 		 *
7061 		 * fin_data[] is swapped around because we are doing a
7062 		 * lookup of the packet is if it were moving in the opposite
7063 		 * direction of the one we are working with now.
7064 		 */
7065 		if (flags & IPN_TCPUDP) {
7066 			swap = frnat.fin_data[0];
7067 			frnat.fin_data[0] = frnat.fin_data[1];
7068 			frnat.fin_data[1] = swap;
7069 		}
7070 		if (fin->fin_out == 1) {
7071 			natl = ipf_nat_inlookup(&frnat,
7072 						flags & ~(SI_WILDP|NAT_SEARCH),
7073 						(u_int)frnat.fin_p,
7074 						frnat.fin_dst, frnat.fin_src);
7075 
7076 		} else {
7077 			natl = ipf_nat_outlookup(&frnat,
7078 						 flags & ~(SI_WILDP|NAT_SEARCH),
7079 						 (u_int)frnat.fin_p,
7080 						 frnat.fin_dst, frnat.fin_src);
7081 		}
7082 		if (flags & IPN_TCPUDP) {
7083 			swap = frnat.fin_data[0];
7084 			frnat.fin_data[0] = frnat.fin_data[1];
7085 			frnat.fin_data[1] = swap;
7086 		}
7087 
7088 		/* TRACE natl, in_stepnext, l */
7089 
7090 		if ((natl != NULL) && (l > 8))	/* XXX 8 is arbitrary */
7091 			return -1;
7092 
7093 		np->in_stepnext &= 0x3;
7094 
7095 		l++;
7096 		changed = -1;
7097 	} while (natl != NULL);
7098 
7099 	nat->nat_osrcip = fin->fin_src;
7100 	nat->nat_odstip = fin->fin_dst;
7101 	nat->nat_nsrcip = frnat.fin_src;
7102 	nat->nat_ndstip = frnat.fin_dst;
7103 
7104 	if ((flags & IPN_TCPUDP) != 0) {
7105 		nat->nat_osport = htons(fin->fin_data[0]);
7106 		nat->nat_odport = htons(fin->fin_data[1]);
7107 		nat->nat_nsport = htons(frnat.fin_data[0]);
7108 		nat->nat_ndport = htons(frnat.fin_data[1]);
7109 	} else if ((flags & IPN_ICMPQUERY) != 0) {
7110 		nat->nat_oicmpid = fin->fin_data[1];
7111 		nat->nat_nicmpid = frnat.fin_data[1];
7112 	}
7113 
7114 	return 0;
7115 }
7116 
7117 
7118 /* ------------------------------------------------------------------------ */
7119 /* Function:    nat_newdivert                                               */
7120 /* Returns:     int - -1 == error, 0 == success                             */
7121 /* Parameters:  fin(I) - pointer to packet information                      */
7122 /*              nat(I) - pointer to NAT entry                               */
7123 /*              ni(I)  - pointer to structure with misc. information needed */
7124 /*                       to create new NAT entry.                           */
7125 /* Write Lock:  ipf_nat                                                     */
7126 /*                                                                          */
7127 /* Create a new NAT  divert session as defined by the NAT rule.  This is    */
7128 /* somewhat different to other NAT session creation routines because we     */
7129 /* do not iterate through either port numbers or IP addresses, searching    */
7130 /* for a unique mapping, however, a complimentary duplicate check is made.  */
7131 /* ------------------------------------------------------------------------ */
7132 static int
7133 ipf_nat_newdivert(fr_info_t *fin, nat_t *nat, natinfo_t *nai)
7134 {
7135 	ipf_main_softc_t *softc = fin->fin_main_soft;
7136 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7137 	fr_info_t frnat;
7138 	ipnat_t *np;
7139 	nat_t *natl;
7140 	int p;
7141 
7142 	np = nai->nai_np;
7143 	bcopy((char *)fin, (char *)&frnat, sizeof(*fin));
7144 
7145 	nat->nat_pr[0] = 0;
7146 	nat->nat_osrcaddr = fin->fin_saddr;
7147 	nat->nat_odstaddr = fin->fin_daddr;
7148 	frnat.fin_saddr = htonl(np->in_snip);
7149 	frnat.fin_daddr = htonl(np->in_dnip);
7150 	if ((nat->nat_flags & IPN_TCPUDP) != 0) {
7151 		nat->nat_osport = htons(fin->fin_data[0]);
7152 		nat->nat_odport = htons(fin->fin_data[1]);
7153 	} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
7154 		nat->nat_oicmpid = fin->fin_data[1];
7155 	}
7156 
7157 	if (np->in_redir & NAT_DIVERTUDP) {
7158 		frnat.fin_data[0] = np->in_spnext;
7159 		frnat.fin_data[1] = np->in_dpnext;
7160 		frnat.fin_flx |= FI_TCPUDP;
7161 		p = IPPROTO_UDP;
7162 	} else {
7163 		frnat.fin_flx &= ~FI_TCPUDP;
7164 		p = IPPROTO_IPIP;
7165 	}
7166 
7167 	if (fin->fin_out == 1) {
7168 		natl = ipf_nat_inlookup(&frnat, 0, p,
7169 					frnat.fin_dst, frnat.fin_src);
7170 
7171 	} else {
7172 		natl = ipf_nat_outlookup(&frnat, 0, p,
7173 					 frnat.fin_dst, frnat.fin_src);
7174 	}
7175 
7176 	if (natl != NULL) {
7177 		NBUMPSIDED(fin->fin_out, ns_divert_exist);
7178 		return -1;
7179 	}
7180 
7181 	nat->nat_nsrcaddr = frnat.fin_saddr;
7182 	nat->nat_ndstaddr = frnat.fin_daddr;
7183 	if ((nat->nat_flags & IPN_TCPUDP) != 0) {
7184 		nat->nat_nsport = htons(frnat.fin_data[0]);
7185 		nat->nat_ndport = htons(frnat.fin_data[1]);
7186 	} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
7187 		nat->nat_nicmpid = frnat.fin_data[1];
7188 	}
7189 
7190 	nat->nat_pr[fin->fin_out] = fin->fin_p;
7191 	nat->nat_pr[1 - fin->fin_out] = p;
7192 
7193 	if (np->in_redir & NAT_REDIRECT)
7194 		nat->nat_dir = NAT_DIVERTIN;
7195 	else
7196 		nat->nat_dir = NAT_DIVERTOUT;
7197 
7198 	return 0;
7199 }
7200 
7201 
7202 /* ------------------------------------------------------------------------ */
7203 /* Function:    nat_builddivertmp                                           */
7204 /* Returns:     int - -1 == error, 0 == success                             */
7205 /* Parameters:  softn(I) - pointer to NAT context structure                 */
7206 /*              np(I)    - pointer to a NAT rule                            */
7207 /*                                                                          */
7208 /* For divert rules, a skeleton packet representing what will be prepended  */
7209 /* to the real packet is created.  Even though we don't have the full       */
7210 /* packet here, a checksum is calculated that we update later when we       */
7211 /* fill in the final details.  At present a 0 checksum for UDP is being set */
7212 /* here because it is expected that divert will be used for localhost.      */
7213 /* ------------------------------------------------------------------------ */
7214 static int
7215 ipf_nat_builddivertmp(ipf_nat_softc_t *softn, ipnat_t *np)
7216 {
7217 	udphdr_t *uh;
7218 	size_t len;
7219 	ip_t *ip;
7220 
7221 	if ((np->in_redir & NAT_DIVERTUDP) != 0)
7222 		len = sizeof(ip_t) + sizeof(udphdr_t);
7223 	else
7224 		len = sizeof(ip_t);
7225 
7226 	ALLOC_MB_T(np->in_divmp, len);
7227 	if (np->in_divmp == NULL) {
7228 		NBUMPD(ipf_nat_stats, ns_divert_build);
7229 		return -1;
7230 	}
7231 
7232 	/*
7233 	 * First, the header to get the packet diverted to the new destination
7234 	 */
7235 	ip = MTOD(np->in_divmp, ip_t *);
7236 	IP_V_A(ip, 4);
7237 	IP_HL_A(ip, 5);
7238 	ip->ip_tos = 0;
7239 	if ((np->in_redir & NAT_DIVERTUDP) != 0)
7240 		ip->ip_p = IPPROTO_UDP;
7241 	else
7242 		ip->ip_p = IPPROTO_IPIP;
7243 	ip->ip_ttl = 255;
7244 	ip->ip_off = 0;
7245 	ip->ip_sum = 0;
7246 	ip->ip_len = htons(len);
7247 	ip->ip_id = 0;
7248 	ip->ip_src.s_addr = htonl(np->in_snip);
7249 	ip->ip_dst.s_addr = htonl(np->in_dnip);
7250 	ip->ip_sum = ipf_cksum((u_short *)ip, sizeof(*ip));
7251 
7252 	if (np->in_redir & NAT_DIVERTUDP) {
7253 		uh = (udphdr_t *)(ip + 1);
7254 		uh->uh_sum = 0;
7255 		uh->uh_ulen = 8;
7256 		uh->uh_sport = htons(np->in_spnext);
7257 		uh->uh_dport = htons(np->in_dpnext);
7258 	}
7259 
7260 	return 0;
7261 }
7262 
7263 
7264 #define	MINDECAP	(sizeof(ip_t) + sizeof(udphdr_t) + sizeof(ip_t))
7265 
7266 /* ------------------------------------------------------------------------ */
7267 /* Function:    nat_decap                                                   */
7268 /* Returns:     int - -1 == error, 0 == success                             */
7269 /* Parameters:  fin(I) - pointer to packet information                      */
7270 /*              nat(I) - pointer to current NAT session                     */
7271 /*                                                                          */
7272 /* This function is responsible for undoing a packet's encapsulation in the */
7273 /* reverse of an encap/divert rule.  After removing the outer encapsulation */
7274 /* it is necessary to call ipf_makefrip() again so that the contents of 'fin'*/
7275 /* match the "new" packet as it may still be used by IPFilter elsewhere.    */
7276 /* We use "dir" here as the basis for some of the expectations about the    */
7277 /* outer header.  If we return an error, the goal is to leave the original  */
7278 /* packet information undisturbed - this falls short at the end where we'd  */
7279 /* need to back a backup copy of "fin" - expensive.                         */
7280 /* ------------------------------------------------------------------------ */
7281 static int
7282 ipf_nat_decap(fr_info_t *fin, nat_t *nat)
7283 {
7284 	ipf_main_softc_t *softc = fin->fin_main_soft;
7285 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7286 	char *hdr;
7287 	int hlen;
7288 	int skip;
7289 	mb_t *m;
7290 
7291 	if ((fin->fin_flx & FI_ICMPERR) != 0) {
7292 		/*
7293 		 * ICMP packets don't get decapsulated, instead what we need
7294 		 * to do is change the ICMP reply from including (in the data
7295 		 * portion for errors) the encapsulated packet that we sent
7296 		 * out to something that resembles the original packet prior
7297 		 * to encapsulation.  This isn't done here - all we're doing
7298 		 * here is changing the outer address to ensure that it gets
7299 		 * targetted back to the correct system.
7300 		 */
7301 
7302 		if (nat->nat_dir & NAT_OUTBOUND) {
7303 			u_32_t sum1, sum2, sumd;
7304 
7305 			sum1 = ntohl(fin->fin_daddr);
7306 			sum2 = ntohl(nat->nat_osrcaddr);
7307 			CALC_SUMD(sum1, sum2, sumd);
7308 			fin->fin_ip->ip_dst = nat->nat_osrcip;
7309 			fin->fin_daddr = nat->nat_osrcaddr;
7310 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
7311      defined(__osf__) || defined(linux)
7312 			ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, sumd, 0);
7313 #endif
7314 		}
7315 		return 0;
7316 	}
7317 
7318 	m = fin->fin_m;
7319 	skip = fin->fin_hlen;
7320 
7321 	switch (nat->nat_dir)
7322 	{
7323 	case NAT_DIVERTIN :
7324 	case NAT_DIVERTOUT :
7325 		if (fin->fin_plen < MINDECAP)
7326 			return -1;
7327 		skip += sizeof(udphdr_t);
7328 		break;
7329 
7330 	case NAT_ENCAPIN :
7331 	case NAT_ENCAPOUT :
7332 		if (fin->fin_plen < (skip + sizeof(ip_t)))
7333 			return -1;
7334 		break;
7335 	default :
7336 		return -1;
7337 		/* NOTREACHED */
7338 	}
7339 
7340 	/*
7341 	 * The aim here is to keep the original packet details in "fin" for
7342 	 * as long as possible so that returning with an error is for the
7343 	 * original packet and there is little undoing work to do.
7344 	 */
7345 	if (M_LEN(m) < skip + sizeof(ip_t)) {
7346 		if (ipf_pr_pullup(fin, skip + sizeof(ip_t)) == -1)
7347 			return -1;
7348 	}
7349 
7350 	hdr = MTOD(fin->fin_m, char *);
7351 	fin->fin_ip = (ip_t *)(hdr + skip);
7352 	hlen = IP_HL(fin->fin_ip) << 2;
7353 
7354 	if (ipf_pr_pullup(fin, skip + hlen) == -1) {
7355 		NBUMPSIDED(fin->fin_out, ns_decap_pullup);
7356 		return -1;
7357 	}
7358 
7359 	fin->fin_hlen = hlen;
7360 	fin->fin_dlen -= skip;
7361 	fin->fin_plen -= skip;
7362 	fin->fin_ipoff += skip;
7363 
7364 	if (ipf_makefrip(hlen, (ip_t *)hdr, fin) == -1) {
7365 		NBUMPSIDED(fin->fin_out, ns_decap_bad);
7366 		return -1;
7367 	}
7368 
7369 	return skip;
7370 }
7371 
7372 
7373 /* ------------------------------------------------------------------------ */
7374 /* Function:    nat_nextaddr                                                */
7375 /* Returns:     int - -1 == bad input (no new address),                     */
7376 /*                     0 == success and dst has new address                 */
7377 /* Parameters:  fin(I) - pointer to packet information                      */
7378 /*              na(I)  - how to generate new address                        */
7379 /*              old(I) - original address being replaced                    */
7380 /*              dst(O) - where to put the new address                       */
7381 /* Write Lock:  ipf_nat                                                     */
7382 /*                                                                          */
7383 /* This function uses the contents of the "na" structure, in combination    */
7384 /* with "old" to produce a new address to store in "dst".  Not all of the   */
7385 /* possible uses of "na" will result in a new address.                      */
7386 /* ------------------------------------------------------------------------ */
7387 static int
7388 ipf_nat_nextaddr(fr_info_t *fin, nat_addr_t *na, u_32_t *old, u_32_t *dst)
7389 {
7390 	ipf_main_softc_t *softc = fin->fin_main_soft;
7391 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7392 	u_32_t amin, amax, new;
7393 	i6addr_t newip;
7394 	int error;
7395 
7396 	new = 0;
7397 	amin = na->na_addr[0].in4.s_addr;
7398 
7399 	switch (na->na_atype)
7400 	{
7401 	case FRI_RANGE :
7402 		amax = na->na_addr[1].in4.s_addr;
7403 		break;
7404 
7405 	case FRI_NETMASKED :
7406 	case FRI_DYNAMIC :
7407 	case FRI_NORMAL :
7408 		/*
7409 		 * Compute the maximum address by adding the inverse of the
7410 		 * netmask to the minimum address.
7411 		 */
7412 		amax = ~na->na_addr[1].in4.s_addr;
7413 		amax |= amin;
7414 		break;
7415 
7416 	case FRI_LOOKUP :
7417 		break;
7418 
7419 	case FRI_BROADCAST :
7420 	case FRI_PEERADDR :
7421 	case FRI_NETWORK :
7422 	default :
7423 		return -1;
7424 	}
7425 
7426 	error = -1;
7427 
7428 	if (na->na_atype == FRI_LOOKUP) {
7429 		if (na->na_type == IPLT_DSTLIST) {
7430 			error = ipf_dstlist_select_node(fin, na->na_ptr, dst,
7431 							NULL);
7432 		} else {
7433 			NBUMPSIDE(fin->fin_out, ns_badnextaddr);
7434 		}
7435 
7436 	} else if (na->na_atype == IPLT_NONE) {
7437 		/*
7438 		 * 0/0 as the new address means leave it alone.
7439 		 */
7440 		if (na->na_addr[0].in4.s_addr == 0 &&
7441 		    na->na_addr[1].in4.s_addr == 0) {
7442 			new = *old;
7443 
7444 		/*
7445 		 * 0/32 means get the interface's address
7446 		 */
7447 		} else if (na->na_addr[0].in4.s_addr == 0 &&
7448 			   na->na_addr[1].in4.s_addr == 0xffffffff) {
7449 			if (ipf_ifpaddr(softc, 4, na->na_atype,
7450 					fin->fin_ifp, &newip, NULL) == -1) {
7451 				NBUMPSIDED(fin->fin_out, ns_ifpaddrfail);
7452 				return -1;
7453 			}
7454 			new = newip.in4.s_addr;
7455 		} else {
7456 			new = htonl(na->na_nextip);
7457 		}
7458 		*dst = new;
7459 		error = 0;
7460 
7461 	} else {
7462 		NBUMPSIDE(fin->fin_out, ns_badnextaddr);
7463 	}
7464 
7465 	return error;
7466 }
7467 
7468 
7469 /* ------------------------------------------------------------------------ */
7470 /* Function:    nat_nextaddrinit                                            */
7471 /* Returns:     int - 0 == success, else error number                       */
7472 /* Parameters:  softc(I) - pointer to soft context main structure           */
7473 /*              na(I)      - NAT address information for generating new addr*/
7474 /*              initial(I) - flag indicating if it is the first call for    */
7475 /*                           this "na" structure.                           */
7476 /*              ifp(I)     - network interface to derive address            */
7477 /*                           information from.                              */
7478 /*                                                                          */
7479 /* This function is expected to be called in two scenarious: when a new NAT */
7480 /* rule is loaded into the kernel and when the list of NAT rules is sync'd  */
7481 /* up with the valid network interfaces (possibly due to them changing.)    */
7482 /* To distinguish between these, the "initial" parameter is used.  If it is */
7483 /* 1 then this indicates the rule has just been reloaded and 0 for when we  */
7484 /* are updating information.  This difference is important because in       */
7485 /* instances where we are not updating address information associated with  */
7486 /* a network interface, we don't want to disturb what the "next" address to */
7487 /* come out of ipf_nat_nextaddr() will be.                                  */
7488 /* ------------------------------------------------------------------------ */
7489 static int
7490 ipf_nat_nextaddrinit(ipf_main_softc_t *softc, char *base, nat_addr_t *na,
7491     int initial, void *ifp)
7492 {
7493 
7494 	switch (na->na_atype)
7495 	{
7496 	case FRI_LOOKUP :
7497 		if (na->na_subtype == 0) {
7498 			na->na_ptr = ipf_lookup_res_num(softc, IPL_LOGNAT,
7499 							na->na_type,
7500 							na->na_num,
7501 							&na->na_func);
7502 		} else if (na->na_subtype == 1) {
7503 			na->na_ptr = ipf_lookup_res_name(softc, IPL_LOGNAT,
7504 							 na->na_type,
7505 							 base + na->na_num,
7506 							 &na->na_func);
7507 		}
7508 		if (na->na_func == NULL) {
7509 			IPFERROR(60060);
7510 			return ESRCH;
7511 		}
7512 		if (na->na_ptr == NULL) {
7513 			IPFERROR(60056);
7514 			return ESRCH;
7515 		}
7516 		break;
7517 
7518 	case FRI_DYNAMIC :
7519 	case FRI_BROADCAST :
7520 	case FRI_NETWORK :
7521 	case FRI_NETMASKED :
7522 	case FRI_PEERADDR :
7523 		if (ifp != NULL)
7524 			(void )ipf_ifpaddr(softc, 4, na->na_atype, ifp,
7525 					   &na->na_addr[0], &na->na_addr[1]);
7526 		break;
7527 
7528 	case FRI_SPLIT :
7529 	case FRI_RANGE :
7530 		if (initial)
7531 			na->na_nextip = ntohl(na->na_addr[0].in4.s_addr);
7532 		break;
7533 
7534 	case FRI_NONE :
7535 		na->na_addr[0].in4.s_addr &= na->na_addr[1].in4.s_addr;
7536 		return 0;
7537 
7538 	case FRI_NORMAL :
7539 		na->na_addr[0].in4.s_addr &= na->na_addr[1].in4.s_addr;
7540 		break;
7541 
7542 	default :
7543 		IPFERROR(60054);
7544 		return EINVAL;
7545 	}
7546 
7547 	if (initial && (na->na_atype == FRI_NORMAL)) {
7548 		if (na->na_addr[0].in4.s_addr == 0) {
7549 			if ((na->na_addr[1].in4.s_addr == 0xffffffff) ||
7550 			    (na->na_addr[1].in4.s_addr == 0)) {
7551 				return 0;
7552 			}
7553 		}
7554 
7555 		if (na->na_addr[1].in4.s_addr == 0xffffffff) {
7556 			na->na_nextip = ntohl(na->na_addr[0].in4.s_addr);
7557 		} else {
7558 			na->na_nextip = ntohl(na->na_addr[0].in4.s_addr) + 1;
7559 		}
7560 	}
7561 
7562 	return 0;
7563 }
7564 
7565 
7566 /* ------------------------------------------------------------------------ */
7567 /* Function:    ipf_nat_matchflush                                          */
7568 /* Returns:     int - -1 == error, 0 == success                             */
7569 /* Parameters:  softc(I) - pointer to soft context main structure           */
7570 /*              softn(I) - pointer to NAT context structure                 */
7571 /*              nat(I)   - pointer to current NAT session                   */
7572 /*                                                                          */
7573 /* ------------------------------------------------------------------------ */
7574 static int
7575 ipf_nat_matchflush(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, void *data)
7576 {
7577 	int *array, flushed, error;
7578 	nat_t *nat, *natnext;
7579 	ipfobj_t obj;
7580 
7581 	error = ipf_matcharray_load(softc, data, &obj, &array);
7582 	if (error != 0)
7583 		return error;
7584 
7585 	flushed = 0;
7586 
7587 	for (nat = softn->ipf_nat_instances; nat != NULL; nat = natnext) {
7588 		natnext = nat->nat_next;
7589 		if (ipf_nat_matcharray(nat, array, softc->ipf_ticks) == 0) {
7590 			ipf_nat_delete(softc, nat, NL_FLUSH);
7591 			flushed++;
7592 		}
7593 	}
7594 
7595 	obj.ipfo_retval = flushed;
7596 	error = BCOPYOUT(&obj, data, sizeof(obj));
7597 
7598 	KFREES(array, array[0] * sizeof(*array));
7599 
7600 	return error;
7601 }
7602 
7603 
7604 /* ------------------------------------------------------------------------ */
7605 /* Function:    ipf_nat_matcharray                                          */
7606 /* Returns:     int - -1 == error, 0 == success                             */
7607 /* Parameters:  fin(I) - pointer to packet information                      */
7608 /*              nat(I) - pointer to current NAT session                     */
7609 /*                                                                          */
7610 /* ------------------------------------------------------------------------ */
7611 static int
7612 ipf_nat_matcharray(nat_t *nat, int *array, u_long ticks)
7613 {
7614 	int i, n, *x, e, p;
7615 
7616 	e = 0;
7617 	n = array[0];
7618 	x = array + 1;
7619 
7620 	for (; n > 0; x += 3 + x[2]) {
7621 		if (x[0] == IPF_EXP_END)
7622 			break;
7623 		e = 0;
7624 
7625 		n -= x[2] + 3;
7626 		if (n < 0)
7627 			break;
7628 
7629 		p = x[0] >> 16;
7630 		if (p != 0 && p != nat->nat_pr[1])
7631 			break;
7632 
7633 		switch (x[0])
7634 		{
7635 		case IPF_EXP_IP_PR :
7636 			for (i = 0; !e && i < x[2]; i++) {
7637 				e |= (nat->nat_pr[1] == x[i + 3]);
7638 			}
7639 			break;
7640 
7641 		case IPF_EXP_IP_SRCADDR :
7642 			if (nat->nat_v[0] == 4) {
7643 				for (i = 0; !e && i < x[2]; i++) {
7644 					e |= ((nat->nat_osrcaddr & x[i + 4]) ==
7645 					      x[i + 3]);
7646 				}
7647 			}
7648 			if (nat->nat_v[1] == 4) {
7649 				for (i = 0; !e && i < x[2]; i++) {
7650 					e |= ((nat->nat_nsrcaddr & x[i + 4]) ==
7651 					      x[i + 3]);
7652 				}
7653 			}
7654 			break;
7655 
7656 		case IPF_EXP_IP_DSTADDR :
7657 			if (nat->nat_v[0] == 4) {
7658 				for (i = 0; !e && i < x[2]; i++) {
7659 					e |= ((nat->nat_odstaddr & x[i + 4]) ==
7660 					      x[i + 3]);
7661 				}
7662 			}
7663 			if (nat->nat_v[1] == 4) {
7664 				for (i = 0; !e && i < x[2]; i++) {
7665 					e |= ((nat->nat_ndstaddr & x[i + 4]) ==
7666 					      x[i + 3]);
7667 				}
7668 			}
7669 			break;
7670 
7671 		case IPF_EXP_IP_ADDR :
7672 			for (i = 0; !e && i < x[2]; i++) {
7673 				if (nat->nat_v[0] == 4) {
7674 					e |= ((nat->nat_osrcaddr & x[i + 4]) ==
7675 					      x[i + 3]);
7676 				}
7677 				if (nat->nat_v[1] == 4) {
7678 					e |= ((nat->nat_nsrcaddr & x[i + 4]) ==
7679 					      x[i + 3]);
7680 				}
7681 				if (nat->nat_v[0] == 4) {
7682 					e |= ((nat->nat_odstaddr & x[i + 4]) ==
7683 					      x[i + 3]);
7684 				}
7685 				if (nat->nat_v[1] == 4) {
7686 					e |= ((nat->nat_ndstaddr & x[i + 4]) ==
7687 					      x[i + 3]);
7688 				}
7689 			}
7690 			break;
7691 
7692 #ifdef USE_INET6
7693 		case IPF_EXP_IP6_SRCADDR :
7694 			if (nat->nat_v[0] == 6) {
7695 				for (i = 0; !e && i < x[3]; i++) {
7696 					e |= IP6_MASKEQ(&nat->nat_osrc6,
7697 							x + i + 7, x + i + 3);
7698 				}
7699 			}
7700 			if (nat->nat_v[1] == 6) {
7701 				for (i = 0; !e && i < x[3]; i++) {
7702 					e |= IP6_MASKEQ(&nat->nat_nsrc6,
7703 							x + i + 7, x + i + 3);
7704 				}
7705 			}
7706 			break;
7707 
7708 		case IPF_EXP_IP6_DSTADDR :
7709 			if (nat->nat_v[0] == 6) {
7710 				for (i = 0; !e && i < x[3]; i++) {
7711 					e |= IP6_MASKEQ(&nat->nat_odst6,
7712 							x + i + 7,
7713 							x + i + 3);
7714 				}
7715 			}
7716 			if (nat->nat_v[1] == 6) {
7717 				for (i = 0; !e && i < x[3]; i++) {
7718 					e |= IP6_MASKEQ(&nat->nat_ndst6,
7719 							x + i + 7,
7720 							x + i + 3);
7721 				}
7722 			}
7723 			break;
7724 
7725 		case IPF_EXP_IP6_ADDR :
7726 			for (i = 0; !e && i < x[3]; i++) {
7727 				if (nat->nat_v[0] == 6) {
7728 					e |= IP6_MASKEQ(&nat->nat_osrc6,
7729 							x + i + 7,
7730 							x + i + 3);
7731 				}
7732 				if (nat->nat_v[0] == 6) {
7733 					e |= IP6_MASKEQ(&nat->nat_odst6,
7734 							x + i + 7,
7735 							x + i + 3);
7736 				}
7737 				if (nat->nat_v[1] == 6) {
7738 					e |= IP6_MASKEQ(&nat->nat_nsrc6,
7739 							x + i + 7,
7740 							x + i + 3);
7741 				}
7742 				if (nat->nat_v[1] == 6) {
7743 					e |= IP6_MASKEQ(&nat->nat_ndst6,
7744 							x + i + 7,
7745 							x + i + 3);
7746 				}
7747 			}
7748 			break;
7749 #endif
7750 
7751 		case IPF_EXP_UDP_PORT :
7752 		case IPF_EXP_TCP_PORT :
7753 			for (i = 0; !e && i < x[2]; i++) {
7754 				e |= (nat->nat_nsport == x[i + 3]) ||
7755 				     (nat->nat_ndport == x[i + 3]);
7756 			}
7757 			break;
7758 
7759 		case IPF_EXP_UDP_SPORT :
7760 		case IPF_EXP_TCP_SPORT :
7761 			for (i = 0; !e && i < x[2]; i++) {
7762 				e |= (nat->nat_nsport == x[i + 3]);
7763 			}
7764 			break;
7765 
7766 		case IPF_EXP_UDP_DPORT :
7767 		case IPF_EXP_TCP_DPORT :
7768 			for (i = 0; !e && i < x[2]; i++) {
7769 				e |= (nat->nat_ndport == x[i + 3]);
7770 			}
7771 			break;
7772 
7773 		case IPF_EXP_TCP_STATE :
7774 			for (i = 0; !e && i < x[2]; i++) {
7775 				e |= (nat->nat_tcpstate[0] == x[i + 3]) ||
7776 				     (nat->nat_tcpstate[1] == x[i + 3]);
7777 			}
7778 			break;
7779 
7780 		case IPF_EXP_IDLE_GT :
7781 			e |= (ticks - nat->nat_touched > x[3]);
7782 			break;
7783 		}
7784 		e ^= x[1];
7785 
7786 		if (!e)
7787 			break;
7788 	}
7789 
7790 	return e;
7791 }
7792 
7793 
7794 /* ------------------------------------------------------------------------ */
7795 /* Function:    ipf_nat_gettable                                            */
7796 /* Returns:     int     - 0 = success, else error                           */
7797 /* Parameters:  softc(I) - pointer to soft context main structure           */
7798 /*              softn(I) - pointer to NAT context structure                 */
7799 /*              data(I)  - pointer to ioctl data                            */
7800 /*                                                                          */
7801 /* This function handles ioctl requests for tables of nat information.      */
7802 /* At present the only table it deals with is the hash bucket statistics.   */
7803 /* ------------------------------------------------------------------------ */
7804 static int
7805 ipf_nat_gettable(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, char *data)
7806 {
7807 	ipftable_t table;
7808 	int error;
7809 
7810 	error = ipf_inobj(softc, data, NULL, &table, IPFOBJ_GTABLE);
7811 	if (error != 0)
7812 		return error;
7813 
7814 	switch (table.ita_type)
7815 	{
7816 	case IPFTABLE_BUCKETS_NATIN :
7817 		error = COPYOUT(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
7818 				table.ita_table,
7819 				softn->ipf_nat_table_sz * sizeof(u_int));
7820 		break;
7821 
7822 	case IPFTABLE_BUCKETS_NATOUT :
7823 		error = COPYOUT(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
7824 				table.ita_table,
7825 				softn->ipf_nat_table_sz * sizeof(u_int));
7826 		break;
7827 
7828 	default :
7829 		IPFERROR(60058);
7830 		return EINVAL;
7831 	}
7832 
7833 	if (error != 0) {
7834 		IPFERROR(60059);
7835 		error = EFAULT;
7836 	}
7837 	return error;
7838 }
7839 
7840 
7841 /* ------------------------------------------------------------------------ */
7842 /* Function:    ipf_nat_settimeout                                          */
7843 /* Returns:     int  - 0 = success, else failure			    */
7844 /* Parameters:  softc(I) - pointer to soft context main structure           */
7845 /*              t(I) - pointer to tunable                                   */
7846 /*              p(I) - pointer to new tuning data                           */
7847 /*                                                                          */
7848 /* Apply the timeout change to the NAT timeout queues.                      */
7849 /* ------------------------------------------------------------------------ */
7850 int
7851 ipf_nat_settimeout(struct ipf_main_softc_s *softc, ipftuneable_t *t,
7852     ipftuneval_t *p)
7853 {
7854 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7855 
7856 	if (!strncmp(t->ipft_name, "tcp_", 4))
7857 		return ipf_settimeout_tcp(t, p, softn->ipf_nat_tcptq);
7858 
7859 	if (!strcmp(t->ipft_name, "udp_timeout")) {
7860 		ipf_apply_timeout(&softn->ipf_nat_udptq, p->ipftu_int);
7861 	} else if (!strcmp(t->ipft_name, "udp_ack_timeout")) {
7862 		ipf_apply_timeout(&softn->ipf_nat_udpacktq, p->ipftu_int);
7863 	} else if (!strcmp(t->ipft_name, "icmp_timeout")) {
7864 		ipf_apply_timeout(&softn->ipf_nat_icmptq, p->ipftu_int);
7865 	} else if (!strcmp(t->ipft_name, "icmp_ack_timeout")) {
7866 		ipf_apply_timeout(&softn->ipf_nat_icmpacktq, p->ipftu_int);
7867 	} else if (!strcmp(t->ipft_name, "ip_timeout")) {
7868 		ipf_apply_timeout(&softn->ipf_nat_iptq, p->ipftu_int);
7869 	} else {
7870 		IPFERROR(60062);
7871 		return ESRCH;
7872 	}
7873 	return 0;
7874 }
7875 
7876 
7877 /* ------------------------------------------------------------------------ */
7878 /* Function:    ipf_nat_rehash                                              */
7879 /* Returns:     int  - 0 = success, else failure			    */
7880 /* Parameters:  softc(I) - pointer to soft context main structure           */
7881 /*              t(I) - pointer to tunable                                   */
7882 /*              p(I) - pointer to new tuning data                           */
7883 /*                                                                          */
7884 /* To change the size of the basic NAT table, we need to first allocate the */
7885 /* new tables (lest it fails and we've got nowhere to store all of the NAT  */
7886 /* sessions currently active) and then walk through the entire list and     */
7887 /* insert them into the table.  There are two tables here: an inbound one   */
7888 /* and an outbound one.  Each NAT entry goes into each table once.          */
7889 /* ------------------------------------------------------------------------ */
7890 int
7891 ipf_nat_rehash(ipf_main_softc_t *softc, ipftuneable_t *t, ipftuneval_t *p)
7892 {
7893 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7894 	nat_t **newtab[2], *nat, **natp;
7895 	u_int *bucketlens[2];
7896 	u_int maxbucket;
7897 	u_int newsize;
7898 	int error;
7899 	u_int hv;
7900 	int i;
7901 
7902 	newsize = p->ipftu_int;
7903 	/*
7904 	 * In case there is nothing to do...
7905 	 */
7906 	if (newsize == softn->ipf_nat_table_sz)
7907 		return 0;
7908 
7909 	newtab[0] = NULL;
7910 	newtab[1] = NULL;
7911 	bucketlens[0] = NULL;
7912 	bucketlens[1] = NULL;
7913 	/*
7914 	 * 4 tables depend on the NAT table size: the inbound looking table,
7915 	 * the outbound lookup table and the hash chain length for each.
7916 	 */
7917 	KMALLOCS(newtab[0], nat_t **, newsize * sizeof(nat_t *));
7918 	if (newtab[0] == NULL) {
7919 		error = 60063;
7920 		goto badrehash;
7921 	}
7922 
7923 	KMALLOCS(newtab[1], nat_t **, newsize * sizeof(nat_t *));
7924 	if (newtab[1] == NULL) {
7925 		error = 60064;
7926 		goto badrehash;
7927 	}
7928 
7929 	KMALLOCS(bucketlens[0], u_int *, newsize * sizeof(u_int));
7930 	if (bucketlens[0] == NULL) {
7931 		error = 60065;
7932 		goto badrehash;
7933 	}
7934 
7935 	KMALLOCS(bucketlens[1], u_int *, newsize * sizeof(u_int));
7936 	if (bucketlens[1] == NULL) {
7937 		error = 60066;
7938 		goto badrehash;
7939 	}
7940 
7941 	/*
7942 	 * Recalculate the maximum length based on the new size.
7943 	 */
7944 	for (maxbucket = 0, i = newsize; i > 0; i >>= 1)
7945 		maxbucket++;
7946 	maxbucket *= 2;
7947 
7948 	bzero((char *)newtab[0], newsize * sizeof(nat_t *));
7949 	bzero((char *)newtab[1], newsize * sizeof(nat_t *));
7950 	bzero((char *)bucketlens[0], newsize * sizeof(u_int));
7951 	bzero((char *)bucketlens[1], newsize * sizeof(u_int));
7952 
7953 	WRITE_ENTER(&softc->ipf_nat);
7954 
7955 	if (softn->ipf_nat_table[0] != NULL) {
7956 		KFREES(softn->ipf_nat_table[0],
7957 		       softn->ipf_nat_table_sz *
7958 		       sizeof(*softn->ipf_nat_table[0]));
7959 	}
7960 	softn->ipf_nat_table[0] = newtab[0];
7961 
7962 	if (softn->ipf_nat_table[1] != NULL) {
7963 		KFREES(softn->ipf_nat_table[1],
7964 		       softn->ipf_nat_table_sz *
7965 		       sizeof(*softn->ipf_nat_table[1]));
7966 	}
7967 	softn->ipf_nat_table[1] = newtab[1];
7968 
7969 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen != NULL) {
7970 		KFREES(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
7971 		       softn->ipf_nat_table_sz * sizeof(u_int));
7972 	}
7973 	softn->ipf_nat_stats.ns_side[0].ns_bucketlen = bucketlens[0];
7974 
7975 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen != NULL) {
7976 		KFREES(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
7977 		       softn->ipf_nat_table_sz * sizeof(u_int));
7978 	}
7979 	softn->ipf_nat_stats.ns_side[1].ns_bucketlen = bucketlens[1];
7980 
7981 	softn->ipf_nat_maxbucket = maxbucket;
7982 	softn->ipf_nat_table_sz = newsize;
7983 	/*
7984 	 * Walk through the entire list of NAT table entries and put them
7985 	 * in the new NAT table, somewhere.  Because we have a new table,
7986 	 * we need to restart the counter of how many chains are in use.
7987 	 */
7988 	softn->ipf_nat_stats.ns_side[0].ns_inuse = 0;
7989 	softn->ipf_nat_stats.ns_side[1].ns_inuse = 0;
7990 
7991 	for (nat = softn->ipf_nat_instances; nat != NULL; nat = nat->nat_next) {
7992 		nat->nat_hnext[0] = NULL;
7993 		nat->nat_phnext[0] = NULL;
7994 		hv = nat->nat_hv[0] % softn->ipf_nat_table_sz;
7995 
7996 		natp = &softn->ipf_nat_table[0][hv];
7997 		if (*natp) {
7998 			(*natp)->nat_phnext[0] = &nat->nat_hnext[0];
7999 		} else {
8000 			NBUMPSIDE(0, ns_inuse);
8001 		}
8002 		nat->nat_phnext[0] = natp;
8003 		nat->nat_hnext[0] = *natp;
8004 		*natp = nat;
8005 		NBUMPSIDE(0, ns_bucketlen[hv]);
8006 
8007 		nat->nat_hnext[1] = NULL;
8008 		nat->nat_phnext[1] = NULL;
8009 		hv = nat->nat_hv[1] % softn->ipf_nat_table_sz;
8010 
8011 		natp = &softn->ipf_nat_table[1][hv];
8012 		if (*natp) {
8013 			(*natp)->nat_phnext[1] = &nat->nat_hnext[1];
8014 		} else {
8015 			NBUMPSIDE(1, ns_inuse);
8016 		}
8017 		nat->nat_phnext[1] = natp;
8018 		nat->nat_hnext[1] = *natp;
8019 		*natp = nat;
8020 		NBUMPSIDE(1, ns_bucketlen[hv]);
8021 	}
8022 	RWLOCK_EXIT(&softc->ipf_nat);
8023 
8024 	return 0;
8025 
8026 badrehash:
8027 	if (bucketlens[1] != NULL) {
8028 		KFREES(bucketlens[0], newsize * sizeof(u_int));
8029 	}
8030 	if (bucketlens[0] != NULL) {
8031 		KFREES(bucketlens[0], newsize * sizeof(u_int));
8032 	}
8033 	if (newtab[0] != NULL) {
8034 		KFREES(newtab[0], newsize * sizeof(nat_t *));
8035 	}
8036 	if (newtab[1] != NULL) {
8037 		KFREES(newtab[1], newsize * sizeof(nat_t *));
8038 	}
8039 	IPFERROR(error);
8040 	return ENOMEM;
8041 }
8042 
8043 
8044 /* ------------------------------------------------------------------------ */
8045 /* Function:    ipf_nat_rehash_rules                                        */
8046 /* Returns:     int  - 0 = success, else failure			    */
8047 /* Parameters:  softc(I) - pointer to soft context main structure           */
8048 /*              t(I) - pointer to tunable                                   */
8049 /*              p(I) - pointer to new tuning data                           */
8050 /*                                                                          */
8051 /* All of the NAT rules hang off of a hash table that is searched with a    */
8052 /* hash on address after the netmask is applied.  There is a different table*/
8053 /* for both inbound rules (rdr) and outbound (map.)  The resizing will only */
8054 /* affect one of these two tables.                                          */
8055 /* ------------------------------------------------------------------------ */
8056 int
8057 ipf_nat_rehash_rules(ipf_main_softc_t *softc, ipftuneable_t *t, ipftuneval_t *p)
8058 {
8059 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8060 	ipnat_t **newtab, *np, ***old, **npp;
8061 	u_int newsize;
8062 	u_int mask;
8063 	u_int hv;
8064 
8065 	newsize = p->ipftu_int;
8066 	/*
8067 	 * In case there is nothing to do...
8068 	 */
8069 	if (newsize == *t->ipft_pint)
8070 		return 0;
8071 
8072 	/*
8073 	 * All inbound rules have the NAT_REDIRECT bit set in in_redir and
8074 	 * all outbound rules have either NAT_MAP or MAT_MAPBLK set.
8075 	 * This if statement allows for some more generic code to be below,
8076 	 * rather than two huge gobs of code that almost do the same thing.
8077 	 */
8078 	if (t->ipft_pint == &softn->ipf_nat_rdrrules_sz) {
8079 		old = &softn->ipf_nat_rdr_rules;
8080 		mask = NAT_REDIRECT;
8081 	} else {
8082 		old = &softn->ipf_nat_map_rules;
8083 		mask = NAT_MAP|NAT_MAPBLK;
8084 	}
8085 
8086 	KMALLOCS(newtab, ipnat_t **, newsize * sizeof(ipnat_t *));
8087 	if (newtab == NULL) {
8088 		IPFERROR(60067);
8089 		return ENOMEM;
8090 	}
8091 
8092 	bzero((char *)newtab, newsize * sizeof(ipnat_t *));
8093 
8094 	WRITE_ENTER(&softc->ipf_nat);
8095 
8096 	if (*old != NULL) {
8097 		KFREES(*old, *t->ipft_pint * sizeof(ipnat_t **));
8098 	}
8099 	*old = newtab;
8100 	*t->ipft_pint = newsize;
8101 
8102 	for (np = softn->ipf_nat_list; np != NULL; np = np->in_next) {
8103 		if ((np->in_redir & mask) == 0)
8104 			continue;
8105 
8106 		if (np->in_redir & NAT_REDIRECT) {
8107 			np->in_rnext = NULL;
8108 			hv = np->in_hv[0] % newsize;
8109 			for (npp = newtab + hv; *npp != NULL; )
8110 				npp = &(*npp)->in_rnext;
8111 			np->in_prnext = npp;
8112 			*npp = np;
8113 		}
8114 		if (np->in_redir & NAT_MAP) {
8115 			np->in_mnext = NULL;
8116 			hv = np->in_hv[1] % newsize;
8117 			for (npp = newtab + hv; *npp != NULL; )
8118 				npp = &(*npp)->in_mnext;
8119 			np->in_pmnext = npp;
8120 			*npp = np;
8121 		}
8122 
8123 	}
8124 	RWLOCK_EXIT(&softc->ipf_nat);
8125 
8126 	return 0;
8127 }
8128 
8129 
8130 /* ------------------------------------------------------------------------ */
8131 /* Function:    ipf_nat_hostmap_rehash                                      */
8132 /* Returns:     int  - 0 = success, else failure			    */
8133 /* Parameters:  softc(I) - pointer to soft context main structure           */
8134 /*              t(I) - pointer to tunable                                   */
8135 /*              p(I) - pointer to new tuning data                           */
8136 /*                                                                          */
8137 /* Allocate and populate a new hash table that will contain a reference to  */
8138 /* all of the active IP# translations currently in place.                   */
8139 /* ------------------------------------------------------------------------ */
8140 int
8141 ipf_nat_hostmap_rehash(ipf_main_softc_t *softc, ipftuneable_t *t,
8142     ipftuneval_t *p)
8143 {
8144 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8145 	hostmap_t *hm, **newtab;
8146 	u_int newsize;
8147 	u_int hv;
8148 
8149 	newsize = p->ipftu_int;
8150 	/*
8151 	 * In case there is nothing to do...
8152 	 */
8153 	if (newsize == *t->ipft_pint)
8154 		return 0;
8155 
8156 	KMALLOCS(newtab, hostmap_t **, newsize * sizeof(hostmap_t *));
8157 	if (newtab == NULL) {
8158 		IPFERROR(60068);
8159 		return ENOMEM;
8160 	}
8161 
8162 	bzero((char *)newtab, newsize * sizeof(hostmap_t *));
8163 
8164 	WRITE_ENTER(&softc->ipf_nat);
8165 	if (softn->ipf_hm_maptable != NULL) {
8166 		KFREES(softn->ipf_hm_maptable,
8167 		       softn->ipf_nat_hostmap_sz * sizeof(hostmap_t *));
8168 	}
8169 	softn->ipf_hm_maptable = newtab;
8170 	softn->ipf_nat_hostmap_sz = newsize;
8171 
8172 	for (hm = softn->ipf_hm_maplist; hm != NULL; hm = hm->hm_next) {
8173 		hv = hm->hm_hv % softn->ipf_nat_hostmap_sz;
8174 		hm->hm_hnext = softn->ipf_hm_maptable[hv];
8175 		hm->hm_phnext = softn->ipf_hm_maptable + hv;
8176 		if (softn->ipf_hm_maptable[hv] != NULL)
8177 			softn->ipf_hm_maptable[hv]->hm_phnext = &hm->hm_hnext;
8178 		softn->ipf_hm_maptable[hv] = hm;
8179 	}
8180 	RWLOCK_EXIT(&softc->ipf_nat);
8181 
8182 	return 0;
8183 }
8184 
8185 
8186 /* ------------------------------------------------------------------------ */
8187 /* Function:    ipf_nat_add_tq                                              */
8188 /* Parameters:  softc(I) - pointer to soft context main structure           */
8189 /*                                                                          */
8190 /* ------------------------------------------------------------------------ */
8191 ipftq_t *
8192 ipf_nat_add_tq(ipf_main_softc_t *softc, int ttl)
8193 {
8194 	ipf_nat_softc_t *softs = softc->ipf_nat_soft;
8195 
8196 	return ipf_addtimeoutqueue(softc, &softs->ipf_nat_utqe, ttl);
8197 }
8198 
8199 /* ------------------------------------------------------------------------ */
8200 /* Function:    ipf_nat_uncreate                                            */
8201 /* Returns:     Nil                                                         */
8202 /* Parameters:  fin(I) - pointer to packet information                      */
8203 /*                                                                          */
8204 /* This function is used to remove a NAT entry from the NAT table when we   */
8205 /* decide that the create was actually in error. It is thus assumed that    */
8206 /* fin_flx will have both FI_NATED and FI_NATNEW set. Because we're dealing */
8207 /* with the translated packet (not the original), we have to reverse the    */
8208 /* lookup. Although doing the lookup is expensive (relatively speaking), it */
8209 /* is not anticipated that this will be a frequent occurance for normal     */
8210 /* traffic patterns.                                                        */
8211 /* ------------------------------------------------------------------------ */
8212 void
8213 ipf_nat_uncreate(fr_info_t *fin)
8214 {
8215 	ipf_main_softc_t *softc = fin->fin_main_soft;
8216 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8217 	int nflags;
8218 	nat_t *nat;
8219 
8220 	switch (fin->fin_p)
8221 	{
8222 	case IPPROTO_TCP :
8223 		nflags = IPN_TCP;
8224 		break;
8225 	case IPPROTO_UDP :
8226 		nflags = IPN_UDP;
8227 		break;
8228 	default :
8229 		nflags = 0;
8230 		break;
8231 	}
8232 
8233 	WRITE_ENTER(&softc->ipf_nat);
8234 
8235 	if (fin->fin_out == 0) {
8236 		nat = ipf_nat_outlookup(fin, nflags, (u_int)fin->fin_p,
8237 					fin->fin_dst, fin->fin_src);
8238 	} else {
8239 		nat = ipf_nat_inlookup(fin, nflags, (u_int)fin->fin_p,
8240 				       fin->fin_src, fin->fin_dst);
8241 	}
8242 
8243 	if (nat != NULL) {
8244 		NBUMPSIDE(fin->fin_out, ns_uncreate[0]);
8245 		ipf_nat_delete(softc, nat, NL_DESTROY);
8246 	} else {
8247 		NBUMPSIDE(fin->fin_out, ns_uncreate[1]);
8248 	}
8249 
8250 	RWLOCK_EXIT(&softc->ipf_nat);
8251 }
8252 
8253 
8254 /* ------------------------------------------------------------------------ */
8255 /* Function:    ipf_nat_cmp_rules                                           */
8256 /* Returns:     int   - 0 == success, else rules do not match.              */
8257 /* Parameters:  n1(I) - first rule to compare                               */
8258 /*              n2(I) - first rule to compare                               */
8259 /*                                                                          */
8260 /* Compare two rules using pointers to each rule. A straight bcmp will not  */
8261 /* work as some fields (such as in_dst, in_pkts) actually do change once    */
8262 /* the rule has been loaded into the kernel. Whilst this function returns   */
8263 /* various non-zero returns, they're strictly to aid in debugging. Use of   */
8264 /* this function should simply care if the result is zero or not.           */
8265 /* ------------------------------------------------------------------------ */
8266 static int
8267 ipf_nat_cmp_rules(ipnat_t *n1, ipnat_t *n2)
8268 {
8269 	if (n1->in_size != n2->in_size)
8270 		return 1;
8271 
8272 	if (bcmp((char *)&n1->in_v, (char *)&n2->in_v,
8273 		 offsetof(ipnat_t, in_ndst) - offsetof(ipnat_t, in_v)) != 0)
8274 		return 2;
8275 
8276 	if (bcmp((char *)&n1->in_tuc, (char *)&n2->in_tuc,
8277 		 n1->in_size - offsetof(ipnat_t, in_tuc)) != 0)
8278 		return 3;
8279 	if (n1->in_ndst.na_atype != n2->in_ndst.na_atype)
8280 		return 5;
8281 	if (n1->in_ndst.na_function != n2->in_ndst.na_function)
8282 		return 6;
8283 	if (bcmp((char *)&n1->in_ndst.na_addr, (char *)&n2->in_ndst.na_addr,
8284 		 sizeof(n1->in_ndst.na_addr)))
8285 		return 7;
8286 	if (n1->in_nsrc.na_atype != n2->in_nsrc.na_atype)
8287 		return 8;
8288 	if (n1->in_nsrc.na_function != n2->in_nsrc.na_function)
8289 		return 9;
8290 	if (bcmp((char *)&n1->in_nsrc.na_addr, (char *)&n2->in_nsrc.na_addr,
8291 		 sizeof(n1->in_nsrc.na_addr)))
8292 		return 10;
8293 	if (n1->in_odst.na_atype != n2->in_odst.na_atype)
8294 		return 11;
8295 	if (n1->in_odst.na_function != n2->in_odst.na_function)
8296 		return 12;
8297 	if (bcmp((char *)&n1->in_odst.na_addr, (char *)&n2->in_odst.na_addr,
8298 		 sizeof(n1->in_odst.na_addr)))
8299 		return 13;
8300 	if (n1->in_osrc.na_atype != n2->in_osrc.na_atype)
8301 		return 14;
8302 	if (n1->in_osrc.na_function != n2->in_osrc.na_function)
8303 		return 15;
8304 	if (bcmp((char *)&n1->in_osrc.na_addr, (char *)&n2->in_osrc.na_addr,
8305 		 sizeof(n1->in_osrc.na_addr)))
8306 		return 16;
8307 	return 0;
8308 }
8309 
8310 
8311 /* ------------------------------------------------------------------------ */
8312 /* Function:    ipf_nat_rule_init                                           */
8313 /* Returns:     int   - 0 == success, else rules do not match.              */
8314 /* Parameters:  softc(I) - pointer to soft context main structure           */
8315 /*              softn(I) - pointer to NAT context structure                 */
8316 /*              n(I)     - first rule to compare                            */
8317 /*                                                                          */
8318 /* ------------------------------------------------------------------------ */
8319 static int
8320 ipf_nat_rule_init(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *n)
8321 {
8322 	int error = 0;
8323 
8324 	if ((n->in_flags & IPN_SIPRANGE) != 0)
8325 		n->in_nsrcatype = FRI_RANGE;
8326 
8327 	if ((n->in_flags & IPN_DIPRANGE) != 0)
8328 		n->in_ndstatype = FRI_RANGE;
8329 
8330 	if ((n->in_flags & IPN_SPLIT) != 0)
8331 		n->in_ndstatype = FRI_SPLIT;
8332 
8333 	if ((n->in_redir & (NAT_MAP|NAT_REWRITE|NAT_DIVERTUDP)) != 0)
8334 		n->in_spnext = n->in_spmin;
8335 
8336 	if ((n->in_redir & (NAT_REWRITE|NAT_DIVERTUDP)) != 0) {
8337 		n->in_dpnext = n->in_dpmin;
8338 	} else if (n->in_redir == NAT_REDIRECT) {
8339 		n->in_dpnext = n->in_dpmin;
8340 	}
8341 
8342 	n->in_stepnext = 0;
8343 
8344 	switch (n->in_v[0])
8345 	{
8346 	case 4 :
8347 		error = ipf_nat_ruleaddrinit(softc, softn, n);
8348 		if (error != 0)
8349 			return error;
8350 		break;
8351 #ifdef USE_INET6
8352 	case 6 :
8353 		error = ipf_nat6_ruleaddrinit(softc, softn, n);
8354 		if (error != 0)
8355 			return error;
8356 		break;
8357 #endif
8358 	default :
8359 		break;
8360 	}
8361 
8362 	if (n->in_redir == (NAT_DIVERTUDP|NAT_MAP)) {
8363 		/*
8364 		 * Prerecord whether or not the destination of the divert
8365 		 * is local or not to the interface the packet is going
8366 		 * to be sent out.
8367 		 */
8368 		n->in_dlocal = ipf_deliverlocal(softc, n->in_v[1],
8369 						n->in_ifps[1], &n->in_ndstip6);
8370 	}
8371 
8372 	return error;
8373 }
8374 
8375 
8376 /* ------------------------------------------------------------------------ */
8377 /* Function:    ipf_nat_rule_fini                                           */
8378 /* Returns:     int   - 0 == success, else rules do not match.              */
8379 /* Parameters:  softc(I) - pointer to soft context main structure           */
8380 /*              n(I)     - rule to work on                                  */
8381 /*                                                                          */
8382 /* This function is used to release any objects that were referenced during */
8383 /* the rule initialisation. This is useful both when free'ing the rule and  */
8384 /* when handling ioctls that need to initialise these fields but not        */
8385 /* actually use them after the ioctl processing has finished.               */
8386 /* ------------------------------------------------------------------------ */
8387 static void
8388 ipf_nat_rule_fini(ipf_main_softc_t *softc, ipnat_t *n)
8389 {
8390 	if (n->in_odst.na_atype == FRI_LOOKUP && n->in_odst.na_ptr != NULL)
8391 		ipf_lookup_deref(softc, n->in_odst.na_type, n->in_odst.na_ptr);
8392 
8393 	if (n->in_osrc.na_atype == FRI_LOOKUP && n->in_osrc.na_ptr != NULL)
8394 		ipf_lookup_deref(softc, n->in_osrc.na_type, n->in_osrc.na_ptr);
8395 
8396 	if (n->in_ndst.na_atype == FRI_LOOKUP && n->in_ndst.na_ptr != NULL)
8397 		ipf_lookup_deref(softc, n->in_ndst.na_type, n->in_ndst.na_ptr);
8398 
8399 	if (n->in_nsrc.na_atype == FRI_LOOKUP && n->in_nsrc.na_ptr != NULL)
8400 		ipf_lookup_deref(softc, n->in_nsrc.na_type, n->in_nsrc.na_ptr);
8401 
8402 	if (n->in_divmp != NULL)
8403 		FREE_MB_T(n->in_divmp);
8404 }
8405