xref: /netbsd-src/sys/external/bsd/ipf/netinet/ip_nat.c (revision b7b7574d3bf8eeb51a1fa3977b59142ec6434a55)
1 /*	$NetBSD: ip_nat.c,v 1.12 2014/06/28 08:02:09 darrenr Exp $	*/
2 
3 /*
4  * Copyright (C) 2012 by Darren Reed.
5  *
6  * See the IPFILTER.LICENCE file for details on licencing.
7  */
8 #if defined(KERNEL) || defined(_KERNEL)
9 # undef KERNEL
10 # undef KERNEL
11 # define        KERNEL	1
12 # define        KERNEL	1
13 #endif
14 #include <sys/errno.h>
15 #include <sys/types.h>
16 #include <sys/param.h>
17 #include <sys/time.h>
18 #include <sys/file.h>
19 #if defined(_KERNEL) && \
20     (defined(__NetBSD_Version) && (__NetBSD_Version >= 399002000))
21 # include <sys/kauth.h>
22 #endif
23 #if !defined(_KERNEL)
24 # include <stdio.h>
25 # include <string.h>
26 # include <stdlib.h>
27 # define KERNEL
28 # ifdef _OpenBSD__
29 struct file;
30 # endif
31 # include <sys/uio.h>
32 # undef KERNEL
33 #endif
34 #if defined(_KERNEL) && \
35     defined(__FreeBSD_version) && (__FreeBSD_version >= 220000)
36 # include <sys/filio.h>
37 # include <sys/fcntl.h>
38 #else
39 # include <sys/ioctl.h>
40 #endif
41 #if !defined(AIX)
42 # include <sys/fcntl.h>
43 #endif
44 #if !defined(linux)
45 # include <sys/protosw.h>
46 #endif
47 #include <sys/socket.h>
48 #if defined(_KERNEL)
49 # include <sys/systm.h>
50 # if !defined(__SVR4) && !defined(__svr4__)
51 #  include <sys/mbuf.h>
52 # endif
53 #endif
54 #if defined(__SVR4) || defined(__svr4__)
55 # include <sys/filio.h>
56 # include <sys/byteorder.h>
57 # ifdef KERNEL
58 #  include <sys/dditypes.h>
59 # endif
60 # include <sys/stream.h>
61 # include <sys/kmem.h>
62 #endif
63 #if _FreeBSD_version >= 300000
64 # include <sys/queue.h>
65 #endif
66 #include <net/if.h>
67 #if _FreeBSD_version >= 300000
68 # include <net/if_var.h>
69 #endif
70 #ifdef sun
71 # include <net/af.h>
72 #endif
73 #include <netinet/in.h>
74 #include <netinet/in_systm.h>
75 #include <netinet/ip.h>
76 
77 #ifdef RFC1825
78 # include <vpn/md5.h>
79 # include <vpn/ipsec.h>
80 extern struct ifnet vpnif;
81 #endif
82 
83 #if !defined(linux)
84 # include <netinet/ip_var.h>
85 #endif
86 #include <netinet/tcp.h>
87 #include <netinet/udp.h>
88 #include <netinet/ip_icmp.h>
89 #include "netinet/ip_compat.h"
90 #include <netinet/tcpip.h>
91 #include "netinet/ipl.h"
92 #include "netinet/ip_fil.h"
93 #include "netinet/ip_nat.h"
94 #include "netinet/ip_frag.h"
95 #include "netinet/ip_state.h"
96 #include "netinet/ip_proxy.h"
97 #include "netinet/ip_lookup.h"
98 #include "netinet/ip_dstlist.h"
99 #include "netinet/ip_sync.h"
100 #if FREEBSD_GE_REV(300000)
101 # include <sys/malloc.h>
102 #endif
103 #ifdef HAS_SYS_MD5_H
104 # include <sys/md5.h>
105 #else
106 # include "md5.h"
107 #endif
108 /* END OF INCLUDES */
109 
110 #undef	SOCKADDR_IN
111 #define	SOCKADDR_IN	struct sockaddr_in
112 
113 #if !defined(lint)
114 #if defined(__NetBSD__)
115 #include <sys/cdefs.h>
116 __KERNEL_RCSID(0, "$NetBSD: ip_nat.c,v 1.12 2014/06/28 08:02:09 darrenr Exp $");
117 #else
118 static const char sccsid[] = "@(#)ip_nat.c	1.11 6/5/96 (C) 1995 Darren Reed";
119 static const char rcsid[] = "@(#)Id: ip_nat.c,v 1.1.1.2 2012/07/22 13:45:27 darrenr Exp";
120 #endif
121 #endif
122 
123 
124 #define	NATFSUM(n,v,f)	((v) == 4 ? (n)->f.in4.s_addr : (n)->f.i6[0] + \
125 			 (n)->f.i6[1] + (n)->f.i6[2] + (n)->f.i6[3])
126 #define	NBUMP(x)	softn->(x)++
127 #define	NBUMPD(x, y)	do { \
128 				softn->x.y++; \
129 				DT(y); \
130 			} while (0)
131 #define	NBUMPSIDE(y,x)	softn->ipf_nat_stats.ns_side[y].x++
132 #define	NBUMPSIDED(y,x)	do { softn->ipf_nat_stats.ns_side[y].x++; \
133 			     DT(x); } while (0)
134 #define	NBUMPSIDEX(y,x,z) \
135 			do { softn->ipf_nat_stats.ns_side[y].x++; \
136 			     DT(z); } while (0)
137 #define	NBUMPSIDEDF(y,x)do { softn->ipf_nat_stats.ns_side[y].x++; \
138 			     DT1(x, fr_info_t *, fin); } while (0)
139 
140 frentry_t	ipfnatblock;
141 
142 static ipftuneable_t ipf_nat_tuneables[] = {
143 	/* nat */
144 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_lock) },
145 		"nat_lock",	0,	1,
146 		stsizeof(ipf_nat_softc_t, ipf_nat_lock),
147 		IPFT_RDONLY,		NULL,	NULL },
148 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_sz) },
149 		"nat_table_size", 1,	0x7fffffff,
150 		stsizeof(ipf_nat_softc_t, ipf_nat_table_sz),
151 		0,			NULL,	ipf_nat_rehash },
152 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_max) },
153 		"nat_table_max", 1,	0x7fffffff,
154 		stsizeof(ipf_nat_softc_t, ipf_nat_table_max),
155 		0,			NULL,	NULL },
156 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_maprules_sz) },
157 		"nat_rules_size", 1,	0x7fffffff,
158 		stsizeof(ipf_nat_softc_t, ipf_nat_maprules_sz),
159 		0,			NULL,	ipf_nat_rehash_rules },
160 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_rdrrules_sz) },
161 		"rdr_rules_size", 1,	0x7fffffff,
162 		stsizeof(ipf_nat_softc_t, ipf_nat_rdrrules_sz),
163 		0,			NULL,	ipf_nat_rehash_rules },
164 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_hostmap_sz) },
165 		"hostmap_size",	1,	0x7fffffff,
166 		stsizeof(ipf_nat_softc_t, ipf_nat_hostmap_sz),
167 		0,			NULL,	ipf_nat_hostmap_rehash },
168 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_maxbucket) },
169 		"nat_maxbucket",1,	0x7fffffff,
170 		stsizeof(ipf_nat_softc_t, ipf_nat_maxbucket),
171 		0,			NULL,	NULL },
172 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_logging) },
173 		"nat_logging",	0,	1,
174 		stsizeof(ipf_nat_softc_t, ipf_nat_logging),
175 		0,			NULL,	NULL },
176 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_doflush) },
177 		"nat_doflush",	0,	1,
178 		stsizeof(ipf_nat_softc_t, ipf_nat_doflush),
179 		0,			NULL,	NULL },
180 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_wm_low) },
181 		"nat_table_wm_low",	1,	99,
182 		stsizeof(ipf_nat_softc_t, ipf_nat_table_wm_low),
183 		0,			NULL,	NULL },
184 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_wm_high) },
185 		"nat_table_wm_high",	2,	100,
186 		stsizeof(ipf_nat_softc_t, ipf_nat_table_wm_high),
187 		0,			NULL,	NULL },
188 	{ { 0 },
189 		NULL,			0,	0,
190 		0,
191 		0,			NULL,	NULL }
192 };
193 
194 /* ======================================================================== */
195 /* How the NAT is organised and works.                                      */
196 /*                                                                          */
197 /* Inside (interface y) NAT       Outside (interface x)                     */
198 /* -------------------- -+- -------------------------------------           */
199 /* Packet going          |   out, processsed by ipf_nat_checkout() for x    */
200 /* ------------>         |   ------------>                                  */
201 /* src=10.1.1.1          |   src=192.1.1.1                                  */
202 /*                       |                                                  */
203 /*                       |   in, processed by ipf_nat_checkin() for x       */
204 /* <------------         |   <------------                                  */
205 /* dst=10.1.1.1          |   dst=192.1.1.1                                  */
206 /* -------------------- -+- -------------------------------------           */
207 /* ipf_nat_checkout() - changes ip_src and if required, sport               */
208 /*             - creates a new mapping, if required.                        */
209 /* ipf_nat_checkin()  - changes ip_dst and if required, dport               */
210 /*                                                                          */
211 /* In the NAT table, internal source is recorded as "in" and externally     */
212 /* seen as "out".                                                           */
213 /* ======================================================================== */
214 
215 
216 #if SOLARIS && !defined(INSTANCES)
217 extern	int		pfil_delayed_copy;
218 #endif
219 
220 static	int	ipf_nat_flush_entry(ipf_main_softc_t *, void *);
221 static	int	ipf_nat_getent(ipf_main_softc_t *, void *, int);
222 static	int	ipf_nat_getsz(ipf_main_softc_t *, void *, int);
223 static	int	ipf_nat_putent(ipf_main_softc_t *, void *, int);
224 static	void	ipf_nat_addmap(ipf_nat_softc_t *, ipnat_t *);
225 static	void	ipf_nat_addrdr(ipf_nat_softc_t *, ipnat_t *);
226 static	int	ipf_nat_builddivertmp(ipf_nat_softc_t *, ipnat_t *);
227 static	int	ipf_nat_clearlist(ipf_main_softc_t *, ipf_nat_softc_t *);
228 static	int	ipf_nat_cmp_rules(ipnat_t *, ipnat_t *);
229 static	int	ipf_nat_decap(fr_info_t *, nat_t *);
230 static	void	ipf_nat_delrule(ipf_main_softc_t *, ipf_nat_softc_t *,
231 				     ipnat_t *, int);
232 static	int	ipf_nat_extraflush(ipf_main_softc_t *, ipf_nat_softc_t *, int);
233 static	int	ipf_nat_finalise(fr_info_t *, nat_t *);
234 static	int	ipf_nat_flushtable(ipf_main_softc_t *, ipf_nat_softc_t *);
235 static	int	ipf_nat_getnext(ipf_main_softc_t *, ipftoken_t *,
236 				ipfgeniter_t *, ipfobj_t *);
237 static	int	ipf_nat_gettable(ipf_main_softc_t *, ipf_nat_softc_t *, char *);
238 static	hostmap_t *ipf_nat_hostmap(ipf_nat_softc_t *, ipnat_t *,
239 				   struct in_addr, struct in_addr,
240 				   struct in_addr, u_32_t);
241 static	int	ipf_nat_icmpquerytype(int);
242 static	int	ipf_nat_iterator(ipf_main_softc_t *, ipftoken_t *,
243 				 ipfgeniter_t *, ipfobj_t *);
244 static	int	ipf_nat_match(fr_info_t *, ipnat_t *);
245 static	int	ipf_nat_matcharray(nat_t *, int *, u_long);
246 static	int	ipf_nat_matchflush(ipf_main_softc_t *, ipf_nat_softc_t *,
247 				   void *);
248 static	void	ipf_nat_mssclamp(tcphdr_t *, u_32_t, fr_info_t *, u_short *);
249 static	int	ipf_nat_newmap(fr_info_t *, nat_t *, natinfo_t *);
250 static	int	ipf_nat_newdivert(fr_info_t *, nat_t *, natinfo_t *);
251 static	int	ipf_nat_newrdr(fr_info_t *, nat_t *, natinfo_t *);
252 static	int	ipf_nat_newrewrite(fr_info_t *, nat_t *, natinfo_t *);
253 static	int	ipf_nat_nextaddr(fr_info_t *, nat_addr_t *, u_32_t *, u_32_t *);
254 static	int	ipf_nat_nextaddrinit(ipf_main_softc_t *, char *,
255 				     nat_addr_t *, int, void *);
256 static	int	ipf_nat_resolverule(ipf_main_softc_t *, ipnat_t *);
257 static	int	ipf_nat_ruleaddrinit(ipf_main_softc_t *,
258 				     ipf_nat_softc_t *, ipnat_t *);
259 static	void	ipf_nat_rule_fini(ipf_main_softc_t *, ipnat_t *);
260 static	int	ipf_nat_rule_init(ipf_main_softc_t *, ipf_nat_softc_t *,
261 				       ipnat_t *);
262 static	int	ipf_nat_siocaddnat(ipf_main_softc_t *, ipf_nat_softc_t *,
263 					ipnat_t *, int);
264 static	void	ipf_nat_siocdelnat(ipf_main_softc_t *, ipf_nat_softc_t *,
265 					ipnat_t *, int);
266 static	void	ipf_nat_tabmove(ipf_nat_softc_t *, nat_t *);
267 
268 /* ------------------------------------------------------------------------ */
269 /* Function:    ipf_nat_main_load                                           */
270 /* Returns:     int - 0 == success, -1 == failure                           */
271 /* Parameters:  Nil                                                         */
272 /*                                                                          */
273 /* The only global NAT structure that needs to be initialised is the filter */
274 /* rule that is used with blocking packets.                                 */
275 /* ------------------------------------------------------------------------ */
276 int
277 ipf_nat_main_load(void)
278 {
279 	bzero((char *)&ipfnatblock, sizeof(ipfnatblock));
280 	ipfnatblock.fr_flags = FR_BLOCK|FR_QUICK;
281 	ipfnatblock.fr_ref = 1;
282 
283 	return 0;
284 }
285 
286 
287 /* ------------------------------------------------------------------------ */
288 /* Function:    ipf_nat_main_unload                                         */
289 /* Returns:     int - 0 == success, -1 == failure                           */
290 /* Parameters:  Nil                                                         */
291 /*                                                                          */
292 /* A null-op function that exists as a placeholder so that the flow in      */
293 /* other functions is obvious.                                              */
294 /* ------------------------------------------------------------------------ */
295 int
296 ipf_nat_main_unload(void)
297 {
298 	return 0;
299 }
300 
301 
302 /* ------------------------------------------------------------------------ */
303 /* Function:    ipf_nat_soft_create                                         */
304 /* Returns:     void * - NULL = failure, else pointer to NAT context        */
305 /* Parameters:  softc(I) - pointer to soft context main structure           */
306 /*                                                                          */
307 /* Allocate the initial soft context structure for NAT and populate it with */
308 /* some default values. Creating the tables is left until we call _init so  */
309 /* that sizes can be changed before we get under way.                       */
310 /* ------------------------------------------------------------------------ */
311 void *
312 ipf_nat_soft_create(ipf_main_softc_t *softc)
313 {
314 	ipf_nat_softc_t *softn;
315 
316 	KMALLOC(softn, ipf_nat_softc_t *);
317 	if (softn == NULL)
318 		return NULL;
319 
320 	bzero((char *)softn, sizeof(*softn));
321 
322 	softn->ipf_nat_tune = ipf_tune_array_copy(softn,
323 						  sizeof(ipf_nat_tuneables),
324 						  ipf_nat_tuneables);
325 	if (softn->ipf_nat_tune == NULL) {
326 		ipf_nat_soft_destroy(softc, softn);
327 		return NULL;
328 	}
329 	if (ipf_tune_array_link(softc, softn->ipf_nat_tune) == -1) {
330 		ipf_nat_soft_destroy(softc, softn);
331 		return NULL;
332 	}
333 
334 	softn->ipf_nat_list_tail = &softn->ipf_nat_list;
335 
336 	softn->ipf_nat_table_max = NAT_TABLE_MAX;
337 	softn->ipf_nat_table_sz = NAT_TABLE_SZ;
338 	softn->ipf_nat_maprules_sz = NAT_SIZE;
339 	softn->ipf_nat_rdrrules_sz = RDR_SIZE;
340 	softn->ipf_nat_hostmap_sz = HOSTMAP_SIZE;
341 	softn->ipf_nat_doflush = 0;
342 #ifdef  IPFILTER_LOG
343 	softn->ipf_nat_logging = 1;
344 #else
345 	softn->ipf_nat_logging = 0;
346 #endif
347 
348 	softn->ipf_nat_defage = DEF_NAT_AGE;
349 	softn->ipf_nat_defipage = IPF_TTLVAL(60);
350 	softn->ipf_nat_deficmpage = IPF_TTLVAL(3);
351 	softn->ipf_nat_table_wm_high = 99;
352 	softn->ipf_nat_table_wm_low = 90;
353 
354 	return softn;
355 }
356 
357 /* ------------------------------------------------------------------------ */
358 /* Function:    ipf_nat_soft_destroy                                        */
359 /* Returns:     Nil                                                         */
360 /* Parameters:  softc(I) - pointer to soft context main structure           */
361 /*                                                                          */
362 /* ------------------------------------------------------------------------ */
363 void
364 ipf_nat_soft_destroy(ipf_main_softc_t *softc, void *arg)
365 {
366 	ipf_nat_softc_t *softn = arg;
367 
368 	if (softn->ipf_nat_tune != NULL) {
369 		ipf_tune_array_unlink(softc, softn->ipf_nat_tune);
370 		KFREES(softn->ipf_nat_tune, sizeof(ipf_nat_tuneables));
371 		softn->ipf_nat_tune = NULL;
372 	}
373 
374 	KFREE(softn);
375 }
376 
377 
378 /* ------------------------------------------------------------------------ */
379 /* Function:    ipf_nat_init                                                */
380 /* Returns:     int - 0 == success, -1 == failure                           */
381 /* Parameters:  softc(I) - pointer to soft context main structure           */
382 /*                                                                          */
383 /* Initialise all of the NAT locks, tables and other structures.            */
384 /* ------------------------------------------------------------------------ */
385 int
386 ipf_nat_soft_init(ipf_main_softc_t *softc, void *arg)
387 {
388 	ipf_nat_softc_t *softn = arg;
389 	ipftq_t *tq;
390 	int i;
391 
392 	KMALLOCS(softn->ipf_nat_table[0], nat_t **, \
393 		 sizeof(nat_t *) * softn->ipf_nat_table_sz);
394 
395 	if (softn->ipf_nat_table[0] != NULL) {
396 		bzero((char *)softn->ipf_nat_table[0],
397 		      softn->ipf_nat_table_sz * sizeof(nat_t *));
398 	} else {
399 		return -1;
400 	}
401 
402 	KMALLOCS(softn->ipf_nat_table[1], nat_t **, \
403 		 sizeof(nat_t *) * softn->ipf_nat_table_sz);
404 
405 	if (softn->ipf_nat_table[1] != NULL) {
406 		bzero((char *)softn->ipf_nat_table[1],
407 		      softn->ipf_nat_table_sz * sizeof(nat_t *));
408 	} else {
409 		return -2;
410 	}
411 
412 	KMALLOCS(softn->ipf_nat_map_rules, ipnat_t **, \
413 		 sizeof(ipnat_t *) * softn->ipf_nat_maprules_sz);
414 
415 	if (softn->ipf_nat_map_rules != NULL) {
416 		bzero((char *)softn->ipf_nat_map_rules,
417 		      softn->ipf_nat_maprules_sz * sizeof(ipnat_t *));
418 	} else {
419 		return -3;
420 	}
421 
422 	KMALLOCS(softn->ipf_nat_rdr_rules, ipnat_t **, \
423 		 sizeof(ipnat_t *) * softn->ipf_nat_rdrrules_sz);
424 
425 	if (softn->ipf_nat_rdr_rules != NULL) {
426 		bzero((char *)softn->ipf_nat_rdr_rules,
427 		      softn->ipf_nat_rdrrules_sz * sizeof(ipnat_t *));
428 	} else {
429 		return -4;
430 	}
431 
432 	KMALLOCS(softn->ipf_hm_maptable, hostmap_t **, \
433 		 sizeof(hostmap_t *) * softn->ipf_nat_hostmap_sz);
434 
435 	if (softn->ipf_hm_maptable != NULL) {
436 		bzero((char *)softn->ipf_hm_maptable,
437 		      sizeof(hostmap_t *) * softn->ipf_nat_hostmap_sz);
438 	} else {
439 		return -5;
440 	}
441 	softn->ipf_hm_maplist = NULL;
442 
443 	KMALLOCS(softn->ipf_nat_stats.ns_side[0].ns_bucketlen, u_int *,
444 		 softn->ipf_nat_table_sz * sizeof(u_int));
445 
446 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen == NULL) {
447 		return -6;
448 	}
449 	bzero((char *)softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
450 	      softn->ipf_nat_table_sz * sizeof(u_int));
451 
452 	KMALLOCS(softn->ipf_nat_stats.ns_side[1].ns_bucketlen, u_int *,
453 		 softn->ipf_nat_table_sz * sizeof(u_int));
454 
455 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen == NULL) {
456 		return -7;
457 	}
458 
459 	bzero((char *)softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
460 	      softn->ipf_nat_table_sz * sizeof(u_int));
461 
462 	if (softn->ipf_nat_maxbucket == 0) {
463 		for (i = softn->ipf_nat_table_sz; i > 0; i >>= 1)
464 			softn->ipf_nat_maxbucket++;
465 		softn->ipf_nat_maxbucket *= 2;
466 	}
467 
468 	ipf_sttab_init(softc, softn->ipf_nat_tcptq);
469 	/*
470 	 * Increase this because we may have "keep state" following this too
471 	 * and packet storms can occur if this is removed too quickly.
472 	 */
473 	softn->ipf_nat_tcptq[IPF_TCPS_CLOSED].ifq_ttl = softc->ipf_tcplastack;
474 	softn->ipf_nat_tcptq[IPF_TCP_NSTATES - 1].ifq_next =
475 							&softn->ipf_nat_udptq;
476 
477 	IPFTQ_INIT(&softn->ipf_nat_udptq, softn->ipf_nat_defage,
478 		   "nat ipftq udp tab");
479 	softn->ipf_nat_udptq.ifq_next = &softn->ipf_nat_udpacktq;
480 
481 	IPFTQ_INIT(&softn->ipf_nat_udpacktq, softn->ipf_nat_defage,
482 		   "nat ipftq udpack tab");
483 	softn->ipf_nat_udpacktq.ifq_next = &softn->ipf_nat_icmptq;
484 
485 	IPFTQ_INIT(&softn->ipf_nat_icmptq, softn->ipf_nat_deficmpage,
486 		   "nat icmp ipftq tab");
487 	softn->ipf_nat_icmptq.ifq_next = &softn->ipf_nat_icmpacktq;
488 
489 	IPFTQ_INIT(&softn->ipf_nat_icmpacktq, softn->ipf_nat_defage,
490 		   "nat icmpack ipftq tab");
491 	softn->ipf_nat_icmpacktq.ifq_next = &softn->ipf_nat_iptq;
492 
493 	IPFTQ_INIT(&softn->ipf_nat_iptq, softn->ipf_nat_defipage,
494 		   "nat ip ipftq tab");
495 	softn->ipf_nat_iptq.ifq_next = &softn->ipf_nat_pending;
496 
497 	IPFTQ_INIT(&softn->ipf_nat_pending, 1, "nat pending ipftq tab");
498 	softn->ipf_nat_pending.ifq_next = NULL;
499 
500 	for (i = 0, tq = softn->ipf_nat_tcptq; i < IPF_TCP_NSTATES; i++, tq++) {
501 		if (tq->ifq_ttl < softn->ipf_nat_deficmpage)
502 			tq->ifq_ttl = softn->ipf_nat_deficmpage;
503 #ifdef LARGE_NAT
504 		else if (tq->ifq_ttl > softn->ipf_nat_defage)
505 			tq->ifq_ttl = softn->ipf_nat_defage;
506 #endif
507 	}
508 
509 	/*
510 	 * Increase this because we may have "keep state" following
511 	 * this too and packet storms can occur if this is removed
512 	 * too quickly.
513 	 */
514 	softn->ipf_nat_tcptq[IPF_TCPS_CLOSED].ifq_ttl = softc->ipf_tcplastack;
515 
516 	MUTEX_INIT(&softn->ipf_nat_new, "ipf nat new mutex");
517 	MUTEX_INIT(&softn->ipf_nat_io, "ipf nat io mutex");
518 
519 	softn->ipf_nat_inited = 1;
520 
521 	return 0;
522 }
523 
524 
525 /* ------------------------------------------------------------------------ */
526 /* Function:    ipf_nat_soft_fini                                           */
527 /* Returns:     Nil                                                         */
528 /* Parameters:  softc(I) - pointer to soft context main structure           */
529 /*                                                                          */
530 /* Free all memory used by NAT structures allocated at runtime.             */
531 /* ------------------------------------------------------------------------ */
532 int
533 ipf_nat_soft_fini(ipf_main_softc_t *softc, void *arg)
534 {
535 	ipf_nat_softc_t *softn = arg;
536 	ipftq_t *ifq, *ifqnext;
537 
538 	(void) ipf_nat_clearlist(softc, softn);
539 	(void) ipf_nat_flushtable(softc, softn);
540 
541 	/*
542 	 * Proxy timeout queues are not cleaned here because although they
543 	 * exist on the NAT list, ipf_proxy_unload is called after unload
544 	 * and the proxies actually are responsible for them being created.
545 	 * Should the proxy timeouts have their own list?  There's no real
546 	 * justification as this is the only complication.
547 	 */
548 	for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifqnext) {
549 		ifqnext = ifq->ifq_next;
550 		if (ipf_deletetimeoutqueue(ifq) == 0)
551 			ipf_freetimeoutqueue(softc, ifq);
552 	}
553 
554 	if (softn->ipf_nat_table[0] != NULL) {
555 		KFREES(softn->ipf_nat_table[0],
556 		       sizeof(nat_t *) * softn->ipf_nat_table_sz);
557 		softn->ipf_nat_table[0] = NULL;
558 	}
559 	if (softn->ipf_nat_table[1] != NULL) {
560 		KFREES(softn->ipf_nat_table[1],
561 		       sizeof(nat_t *) * softn->ipf_nat_table_sz);
562 		softn->ipf_nat_table[1] = NULL;
563 	}
564 	if (softn->ipf_nat_map_rules != NULL) {
565 		KFREES(softn->ipf_nat_map_rules,
566 		       sizeof(ipnat_t *) * softn->ipf_nat_maprules_sz);
567 		softn->ipf_nat_map_rules = NULL;
568 	}
569 	if (softn->ipf_nat_rdr_rules != NULL) {
570 		KFREES(softn->ipf_nat_rdr_rules,
571 		       sizeof(ipnat_t *) * softn->ipf_nat_rdrrules_sz);
572 		softn->ipf_nat_rdr_rules = NULL;
573 	}
574 	if (softn->ipf_hm_maptable != NULL) {
575 		KFREES(softn->ipf_hm_maptable,
576 		       sizeof(hostmap_t *) * softn->ipf_nat_hostmap_sz);
577 		softn->ipf_hm_maptable = NULL;
578 	}
579 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen != NULL) {
580 		KFREES(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
581 		       sizeof(u_int) * softn->ipf_nat_table_sz);
582 		softn->ipf_nat_stats.ns_side[0].ns_bucketlen = NULL;
583 	}
584 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen != NULL) {
585 		KFREES(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
586 		       sizeof(u_int) * softn->ipf_nat_table_sz);
587 		softn->ipf_nat_stats.ns_side[1].ns_bucketlen = NULL;
588 	}
589 
590 	if (softn->ipf_nat_inited == 1) {
591 		softn->ipf_nat_inited = 0;
592 		ipf_sttab_destroy(softn->ipf_nat_tcptq);
593 
594 		MUTEX_DESTROY(&softn->ipf_nat_new);
595 		MUTEX_DESTROY(&softn->ipf_nat_io);
596 
597 		MUTEX_DESTROY(&softn->ipf_nat_udptq.ifq_lock);
598 		MUTEX_DESTROY(&softn->ipf_nat_udpacktq.ifq_lock);
599 		MUTEX_DESTROY(&softn->ipf_nat_icmptq.ifq_lock);
600 		MUTEX_DESTROY(&softn->ipf_nat_icmpacktq.ifq_lock);
601 		MUTEX_DESTROY(&softn->ipf_nat_iptq.ifq_lock);
602 		MUTEX_DESTROY(&softn->ipf_nat_pending.ifq_lock);
603 	}
604 
605 	return 0;
606 }
607 
608 
609 /* ------------------------------------------------------------------------ */
610 /* Function:    ipf_nat_setlock                                             */
611 /* Returns:     Nil                                                         */
612 /* Parameters:  arg(I) - pointer to soft state information                  */
613 /*              tmp(I) - new lock value                                     */
614 /*                                                                          */
615 /* Set the "lock status" of NAT to the value in tmp.                        */
616 /* ------------------------------------------------------------------------ */
617 void
618 ipf_nat_setlock(void *arg, int tmp)
619 {
620 	ipf_nat_softc_t *softn = arg;
621 
622 	softn->ipf_nat_lock = tmp;
623 }
624 
625 
626 /* ------------------------------------------------------------------------ */
627 /* Function:    ipf_nat_addrdr                                              */
628 /* Returns:     Nil                                                         */
629 /* Parameters:  n(I) - pointer to NAT rule to add                           */
630 /*                                                                          */
631 /* Adds a redirect rule to the hash table of redirect rules and the list of */
632 /* loaded NAT rules.  Updates the bitmask indicating which netmasks are in  */
633 /* use by redirect rules.                                                   */
634 /* ------------------------------------------------------------------------ */
635 static void
636 ipf_nat_addrdr(ipf_nat_softc_t *softn, ipnat_t *n)
637 {
638 	ipnat_t **np;
639 	u_32_t j;
640 	u_int hv;
641 	u_int rhv;
642 	int k;
643 
644 	if (n->in_odstatype == FRI_NORMAL) {
645 		k = count4bits(n->in_odstmsk);
646 		ipf_inet_mask_add(k, &softn->ipf_nat_rdr_mask);
647 		j = (n->in_odstaddr & n->in_odstmsk);
648 		rhv = NAT_HASH_FN(j, 0, 0xffffffff);
649 	} else {
650 		ipf_inet_mask_add(0, &softn->ipf_nat_rdr_mask);
651 		j = 0;
652 		rhv = 0;
653 	}
654 	hv = rhv % softn->ipf_nat_rdrrules_sz;
655 	np = softn->ipf_nat_rdr_rules + hv;
656 	while (*np != NULL)
657 		np = &(*np)->in_rnext;
658 	n->in_rnext = NULL;
659 	n->in_prnext = np;
660 	n->in_hv[0] = hv;
661 	n->in_use++;
662 	*np = n;
663 }
664 
665 
666 /* ------------------------------------------------------------------------ */
667 /* Function:    ipf_nat_addmap                                              */
668 /* Returns:     Nil                                                         */
669 /* Parameters:  n(I) - pointer to NAT rule to add                           */
670 /*                                                                          */
671 /* Adds a NAT map rule to the hash table of rules and the list of  loaded   */
672 /* NAT rules.  Updates the bitmask indicating which netmasks are in use by  */
673 /* redirect rules.                                                          */
674 /* ------------------------------------------------------------------------ */
675 static void
676 ipf_nat_addmap(ipf_nat_softc_t *softn, ipnat_t *n)
677 {
678 	ipnat_t **np;
679 	u_32_t j;
680 	u_int hv;
681 	u_int rhv;
682 	int k;
683 
684 	if (n->in_osrcatype == FRI_NORMAL) {
685 		k = count4bits(n->in_osrcmsk);
686 		ipf_inet_mask_add(k, &softn->ipf_nat_map_mask);
687 		j = (n->in_osrcaddr & n->in_osrcmsk);
688 		rhv = NAT_HASH_FN(j, 0, 0xffffffff);
689 	} else {
690 		ipf_inet_mask_add(0, &softn->ipf_nat_map_mask);
691 		j = 0;
692 		rhv = 0;
693 	}
694 	hv = rhv % softn->ipf_nat_maprules_sz;
695 	np = softn->ipf_nat_map_rules + hv;
696 	while (*np != NULL)
697 		np = &(*np)->in_mnext;
698 	n->in_mnext = NULL;
699 	n->in_pmnext = np;
700 	n->in_hv[1] = rhv;
701 	n->in_use++;
702 	*np = n;
703 }
704 
705 
706 /* ------------------------------------------------------------------------ */
707 /* Function:    ipf_nat_delrdr                                              */
708 /* Returns:     Nil                                                         */
709 /* Parameters:  n(I) - pointer to NAT rule to delete                        */
710 /*                                                                          */
711 /* Removes a redirect rule from the hash table of redirect rules.           */
712 /* ------------------------------------------------------------------------ */
713 void
714 ipf_nat_delrdr(ipf_nat_softc_t *softn, ipnat_t *n)
715 {
716 	if (n->in_odstatype == FRI_NORMAL) {
717 		int k = count4bits(n->in_odstmsk);
718 		ipf_inet_mask_del(k, &softn->ipf_nat_rdr_mask);
719 	} else {
720 		ipf_inet_mask_del(0, &softn->ipf_nat_rdr_mask);
721 	}
722 	if (n->in_rnext)
723 		n->in_rnext->in_prnext = n->in_prnext;
724 	*n->in_prnext = n->in_rnext;
725 	n->in_use--;
726 }
727 
728 
729 /* ------------------------------------------------------------------------ */
730 /* Function:    ipf_nat_delmap                                              */
731 /* Returns:     Nil                                                         */
732 /* Parameters:  n(I) - pointer to NAT rule to delete                        */
733 /*                                                                          */
734 /* Removes a NAT map rule from the hash table of NAT map rules.             */
735 /* ------------------------------------------------------------------------ */
736 void
737 ipf_nat_delmap(ipf_nat_softc_t *softn, ipnat_t *n)
738 {
739 	if (n->in_osrcatype == FRI_NORMAL) {
740 		int k = count4bits(n->in_osrcmsk);
741 		ipf_inet_mask_del(k, &softn->ipf_nat_map_mask);
742 	} else {
743 		ipf_inet_mask_del(0, &softn->ipf_nat_map_mask);
744 	}
745 	if (n->in_mnext != NULL)
746 		n->in_mnext->in_pmnext = n->in_pmnext;
747 	*n->in_pmnext = n->in_mnext;
748 	n->in_use--;
749 }
750 
751 
752 /* ------------------------------------------------------------------------ */
753 /* Function:    ipf_nat_hostmap                                             */
754 /* Returns:     struct hostmap* - NULL if no hostmap could be created,      */
755 /*                                else a pointer to the hostmapping to use  */
756 /* Parameters:  np(I)   - pointer to NAT rule                               */
757 /*              real(I) - real IP address                                   */
758 /*              map(I)  - mapped IP address                                 */
759 /*              port(I) - destination port number                           */
760 /* Write Locks: ipf_nat                                                     */
761 /*                                                                          */
762 /* Check if an ip address has already been allocated for a given mapping    */
763 /* that is not doing port based translation.  If is not yet allocated, then */
764 /* create a new entry if a non-NULL NAT rule pointer has been supplied.     */
765 /* ------------------------------------------------------------------------ */
766 static struct hostmap *
767 ipf_nat_hostmap(ipf_nat_softc_t *softn, ipnat_t *np, struct in_addr src,
768     struct in_addr dst, struct in_addr map, u_32_t port)
769 {
770 	hostmap_t *hm;
771 	u_int hv, rhv;
772 
773 	hv = (src.s_addr ^ dst.s_addr);
774 	hv += src.s_addr;
775 	hv += dst.s_addr;
776 	rhv = hv;
777 	hv %= softn->ipf_nat_hostmap_sz;
778 	for (hm = softn->ipf_hm_maptable[hv]; hm; hm = hm->hm_hnext)
779 		if ((hm->hm_osrcip.s_addr == src.s_addr) &&
780 		    (hm->hm_odstip.s_addr == dst.s_addr) &&
781 		    ((np == NULL) || (np == hm->hm_ipnat)) &&
782 		    ((port == 0) || (port == hm->hm_port))) {
783 			softn->ipf_nat_stats.ns_hm_addref++;
784 			hm->hm_ref++;
785 			return hm;
786 		}
787 
788 	if (np == NULL) {
789 		softn->ipf_nat_stats.ns_hm_nullnp++;
790 		return NULL;
791 	}
792 
793 	KMALLOC(hm, hostmap_t *);
794 	if (hm) {
795 		hm->hm_next = softn->ipf_hm_maplist;
796 		hm->hm_pnext = &softn->ipf_hm_maplist;
797 		if (softn->ipf_hm_maplist != NULL)
798 			softn->ipf_hm_maplist->hm_pnext = &hm->hm_next;
799 		softn->ipf_hm_maplist = hm;
800 		hm->hm_hnext = softn->ipf_hm_maptable[hv];
801 		hm->hm_phnext = softn->ipf_hm_maptable + hv;
802 		if (softn->ipf_hm_maptable[hv] != NULL)
803 			softn->ipf_hm_maptable[hv]->hm_phnext = &hm->hm_hnext;
804 		softn->ipf_hm_maptable[hv] = hm;
805 		hm->hm_ipnat = np;
806 		np->in_use++;
807 		hm->hm_osrcip = src;
808 		hm->hm_odstip = dst;
809 		hm->hm_nsrcip = map;
810 		hm->hm_ndstip.s_addr = 0;
811 		hm->hm_ref = 1;
812 		hm->hm_port = port;
813 		hm->hm_hv = rhv;
814 		hm->hm_v = 4;
815 		softn->ipf_nat_stats.ns_hm_new++;
816 	} else {
817 		softn->ipf_nat_stats.ns_hm_newfail++;
818 	}
819 	return hm;
820 }
821 
822 
823 /* ------------------------------------------------------------------------ */
824 /* Function:    ipf_nat_hostmapdel                                          */
825 /* Returns:     Nil                                                         */
826 /* Parameters:  hmp(I) - pointer to hostmap structure pointer               */
827 /* Write Locks: ipf_nat                                                     */
828 /*                                                                          */
829 /* Decrement the references to this hostmap structure by one.  If this      */
830 /* reaches zero then remove it and free it.                                 */
831 /* ------------------------------------------------------------------------ */
832 void
833 ipf_nat_hostmapdel(ipf_main_softc_t *softc, struct hostmap **hmp)
834 {
835 	struct hostmap *hm;
836 
837 	hm = *hmp;
838 	*hmp = NULL;
839 
840 	hm->hm_ref--;
841 	if (hm->hm_ref == 0) {
842 		ipf_nat_rule_deref(softc, &hm->hm_ipnat);
843 		if (hm->hm_hnext)
844 			hm->hm_hnext->hm_phnext = hm->hm_phnext;
845 		*hm->hm_phnext = hm->hm_hnext;
846 		if (hm->hm_next)
847 			hm->hm_next->hm_pnext = hm->hm_pnext;
848 		*hm->hm_pnext = hm->hm_next;
849 		KFREE(hm);
850 	}
851 }
852 
853 
854 /* ------------------------------------------------------------------------ */
855 /* Function:    ipf_fix_outcksum                                            */
856 /* Returns:     Nil                                                         */
857 /* Parameters:  fin(I) - pointer to packet information                      */
858 /*              sp(I)  - location of 16bit checksum to update               */
859 /*              n((I)  - amount to adjust checksum by                       */
860 /*                                                                          */
861 /* Adjusts the 16bit checksum by "n" for packets going out.                 */
862 /* ------------------------------------------------------------------------ */
863 void
864 ipf_fix_outcksum(int cksum, u_short *sp, u_32_t n, u_32_t partial)
865 {
866 	u_short sumshort;
867 	u_32_t sum1;
868 
869 	if (n == 0)
870 		return;
871 
872 	if (cksum == 4) {
873 		*sp = 0;
874 		return;
875 	}
876 	if (cksum == 2) {
877 		sum1 = partial;
878 		sum1 = (sum1 & 0xffff) + (sum1 >> 16);
879 		*sp = htons(sum1);
880 		return;
881 	}
882 	sum1 = (~ntohs(*sp)) & 0xffff;
883 	sum1 += (n);
884 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
885 	/* Again */
886 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
887 	sumshort = ~(u_short)sum1;
888 	*(sp) = htons(sumshort);
889 }
890 
891 
892 /* ------------------------------------------------------------------------ */
893 /* Function:    ipf_fix_incksum                                             */
894 /* Returns:     Nil                                                         */
895 /* Parameters:  fin(I) - pointer to packet information                      */
896 /*              sp(I)  - location of 16bit checksum to update               */
897 /*              n((I)  - amount to adjust checksum by                       */
898 /*                                                                          */
899 /* Adjusts the 16bit checksum by "n" for packets going in.                  */
900 /* ------------------------------------------------------------------------ */
901 void
902 ipf_fix_incksum(int cksum, u_short *sp, u_32_t n, u_32_t partial)
903 {
904 	u_short sumshort;
905 	u_32_t sum1;
906 
907 	if (n == 0)
908 		return;
909 
910 	if (cksum == 4) {
911 		*sp = 0;
912 		return;
913 	}
914 	if (cksum == 2) {
915 		sum1 = partial;
916 		sum1 = (sum1 & 0xffff) + (sum1 >> 16);
917 		*sp = htons(sum1);
918 		return;
919 	}
920 
921 	sum1 = (~ntohs(*sp)) & 0xffff;
922 	sum1 += ~(n) & 0xffff;
923 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
924 	/* Again */
925 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
926 	sumshort = ~(u_short)sum1;
927 	*(sp) = htons(sumshort);
928 }
929 
930 
931 /* ------------------------------------------------------------------------ */
932 /* Function:    ipf_fix_datacksum                                           */
933 /* Returns:     Nil                                                         */
934 /* Parameters:  sp(I)  - location of 16bit checksum to update               */
935 /*              n((I)  - amount to adjust checksum by                       */
936 /*                                                                          */
937 /* Fix_datacksum is used *only* for the adjustments of checksums in the     */
938 /* data section of an IP packet.                                            */
939 /*                                                                          */
940 /* The only situation in which you need to do this is when NAT'ing an       */
941 /* ICMP error message. Such a message, contains in its body the IP header   */
942 /* of the original IP packet, that causes the error.                        */
943 /*                                                                          */
944 /* You can't use fix_incksum or fix_outcksum in that case, because for the  */
945 /* kernel the data section of the ICMP error is just data, and no special   */
946 /* processing like hardware cksum or ntohs processing have been done by the */
947 /* kernel on the data section.                                              */
948 /* ------------------------------------------------------------------------ */
949 void
950 ipf_fix_datacksum(u_short *sp, u_32_t n)
951 {
952 	u_short sumshort;
953 	u_32_t sum1;
954 
955 	if (n == 0)
956 		return;
957 
958 	sum1 = (~ntohs(*sp)) & 0xffff;
959 	sum1 += (n);
960 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
961 	/* Again */
962 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
963 	sumshort = ~(u_short)sum1;
964 	*(sp) = htons(sumshort);
965 }
966 
967 
968 /* ------------------------------------------------------------------------ */
969 /* Function:    ipf_nat_ioctl                                               */
970 /* Returns:     int - 0 == success, != 0 == failure                         */
971 /* Parameters:  softc(I) - pointer to soft context main structure           */
972 /*              data(I)  - pointer to ioctl data                            */
973 /*              cmd(I)   - ioctl command integer                            */
974 /*              mode(I)  - file mode bits used with open                    */
975 /*              uid(I)   - uid of calling process                           */
976 /*              ctx(I)   - pointer used as key for finding context          */
977 /*                                                                          */
978 /* Processes an ioctl call made to operate on the IP Filter NAT device.     */
979 /* ------------------------------------------------------------------------ */
980 int
981 ipf_nat_ioctl(ipf_main_softc_t *softc, void *data, ioctlcmd_t cmd, int mode,
982    int uid, void *ctx)
983 {
984 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
985 	int error = 0, ret, arg, getlock;
986 	ipnat_t *nat, *nt, *n;
987 	ipnat_t natd;
988 	SPL_INT(s);
989 
990 #if BSD_GE_YEAR(199306) && defined(_KERNEL)
991 # if NETBSD_GE_REV(399002000)
992 	if ((mode & FWRITE) &&
993 	     kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_FIREWALL,
994 				     KAUTH_REQ_NETWORK_FIREWALL_FW,
995 				     NULL, NULL, NULL))
996 # else
997 #  if defined(__FreeBSD_version) && (__FreeBSD_version >= 500034)
998 	if (securelevel_ge(curthread->td_ucred, 3) && (mode & FWRITE))
999 #  else
1000 	if ((securelevel >= 3) && (mode & FWRITE))
1001 #  endif
1002 # endif
1003 	{
1004 		IPFERROR(60001);
1005 		return EPERM;
1006 	}
1007 #endif
1008 
1009 #if defined(__osf__) && defined(_KERNEL)
1010 	getlock = 0;
1011 #else
1012 	getlock = (mode & NAT_LOCKHELD) ? 0 : 1;
1013 #endif
1014 
1015 	n = NULL;
1016 	nt = NULL;
1017 	nat = NULL;
1018 
1019 	if ((cmd == (ioctlcmd_t)SIOCADNAT) || (cmd == (ioctlcmd_t)SIOCRMNAT) ||
1020 	    (cmd == (ioctlcmd_t)SIOCPURGENAT)) {
1021 		if (mode & NAT_SYSSPACE) {
1022 			bcopy(data, (char *)&natd, sizeof(natd));
1023 			nat = &natd;
1024 			error = 0;
1025 		} else {
1026 			bzero(&natd, sizeof(natd));
1027 			error = ipf_inobj(softc, data, NULL, &natd,
1028 					  IPFOBJ_IPNAT);
1029 			if (error != 0)
1030 				goto done;
1031 
1032 			if (natd.in_size < sizeof(ipnat_t)) {
1033 				error = EINVAL;
1034 				goto done;
1035 			}
1036 			KMALLOCS(nt, ipnat_t *, natd.in_size);
1037 			if (nt == NULL) {
1038 				IPFERROR(60070);
1039 				error = ENOMEM;
1040 				goto done;
1041 			}
1042 			bzero(nt, natd.in_size);
1043 			error = ipf_inobjsz(softc, data, nt, IPFOBJ_IPNAT,
1044 					    natd.in_size);
1045 			if (error)
1046 				goto done;
1047 			nat = nt;
1048 		}
1049 
1050 		/*
1051 		 * For add/delete, look to see if the NAT entry is
1052 		 * already present
1053 		 */
1054 		nat->in_flags &= IPN_USERFLAGS;
1055 		if ((nat->in_redir & NAT_MAPBLK) == 0) {
1056 			if (nat->in_osrcatype == FRI_NORMAL ||
1057 			    nat->in_osrcatype == FRI_NONE)
1058 				nat->in_osrcaddr &= nat->in_osrcmsk;
1059 			if (nat->in_odstatype == FRI_NORMAL ||
1060 			    nat->in_odstatype == FRI_NONE)
1061 				nat->in_odstaddr &= nat->in_odstmsk;
1062 			if ((nat->in_flags & (IPN_SPLIT|IPN_SIPRANGE)) == 0) {
1063 				if (nat->in_nsrcatype == FRI_NORMAL)
1064 					nat->in_nsrcaddr &= nat->in_nsrcmsk;
1065 				if (nat->in_ndstatype == FRI_NORMAL)
1066 					nat->in_ndstaddr &= nat->in_ndstmsk;
1067 			}
1068 		}
1069 
1070 		error = ipf_nat_rule_init(softc, softn, nat);
1071 		if (error != 0)
1072 			goto done;
1073 
1074 		MUTEX_ENTER(&softn->ipf_nat_io);
1075 		for (n = softn->ipf_nat_list; n != NULL; n = n->in_next)
1076 			if (ipf_nat_cmp_rules(nat, n) == 0)
1077 				break;
1078 	}
1079 
1080 	switch (cmd)
1081 	{
1082 #ifdef  IPFILTER_LOG
1083 	case SIOCIPFFB :
1084 	{
1085 		int tmp;
1086 
1087 		if (!(mode & FWRITE)) {
1088 			IPFERROR(60002);
1089 			error = EPERM;
1090 		} else {
1091 			tmp = ipf_log_clear(softc, IPL_LOGNAT);
1092 			error = BCOPYOUT(&tmp, data, sizeof(tmp));
1093 			if (error != 0) {
1094 				IPFERROR(60057);
1095 				error = EFAULT;
1096 			}
1097 		}
1098 		break;
1099 	}
1100 
1101 	case SIOCSETLG :
1102 		if (!(mode & FWRITE)) {
1103 			IPFERROR(60003);
1104 			error = EPERM;
1105 		} else {
1106 			error = BCOPYIN(data, &softn->ipf_nat_logging,
1107 					sizeof(softn->ipf_nat_logging));
1108 			if (error != 0)
1109 				error = EFAULT;
1110 		}
1111 		break;
1112 
1113 	case SIOCGETLG :
1114 		error = BCOPYOUT(&softn->ipf_nat_logging, data,
1115 				 sizeof(softn->ipf_nat_logging));
1116 		if (error != 0) {
1117 			IPFERROR(60004);
1118 			error = EFAULT;
1119 		}
1120 		break;
1121 
1122 	case FIONREAD :
1123 		arg = ipf_log_bytesused(softc, IPL_LOGNAT);
1124 		error = BCOPYOUT(&arg, data, sizeof(arg));
1125 		if (error != 0) {
1126 			IPFERROR(60005);
1127 			error = EFAULT;
1128 		}
1129 		break;
1130 #endif
1131 	case SIOCADNAT :
1132 		if (!(mode & FWRITE)) {
1133 			IPFERROR(60006);
1134 			error = EPERM;
1135 		} else if (n != NULL) {
1136 			natd.in_flineno = n->in_flineno;
1137 			(void) ipf_outobj(softc, data, &natd, IPFOBJ_IPNAT);
1138 			IPFERROR(60007);
1139 			error = EEXIST;
1140 		} else if (nt == NULL) {
1141 			IPFERROR(60008);
1142 			error = ENOMEM;
1143 		}
1144 		if (error != 0) {
1145 			MUTEX_EXIT(&softn->ipf_nat_io);
1146 			break;
1147 		}
1148 		if (nat != nt)
1149 			bcopy((char *)nat, (char *)nt, sizeof(*n));
1150 		error = ipf_nat_siocaddnat(softc, softn, nt, getlock);
1151 		MUTEX_EXIT(&softn->ipf_nat_io);
1152 		if (error == 0) {
1153 			nat = NULL;
1154 			nt = NULL;
1155 		}
1156 		break;
1157 
1158 	case SIOCRMNAT :
1159 	case SIOCPURGENAT :
1160 		if (!(mode & FWRITE)) {
1161 			IPFERROR(60009);
1162 			error = EPERM;
1163 			n = NULL;
1164 		} else if (n == NULL) {
1165 			IPFERROR(60010);
1166 			error = ESRCH;
1167 		}
1168 
1169 		if (error != 0) {
1170 			MUTEX_EXIT(&softn->ipf_nat_io);
1171 			break;
1172 		}
1173 		if (cmd == (ioctlcmd_t)SIOCPURGENAT) {
1174 			error = ipf_outobjsz(softc, data, n, IPFOBJ_IPNAT,
1175 					     n->in_size);
1176 			if (error) {
1177 				MUTEX_EXIT(&softn->ipf_nat_io);
1178 				goto done;
1179 			}
1180 			n->in_flags |= IPN_PURGE;
1181 		}
1182 		ipf_nat_siocdelnat(softc, softn, n, getlock);
1183 
1184 		MUTEX_EXIT(&softn->ipf_nat_io);
1185 		n = NULL;
1186 		break;
1187 
1188 	case SIOCGNATS :
1189 	    {
1190 		natstat_t *nsp = &softn->ipf_nat_stats;
1191 
1192 		nsp->ns_side[0].ns_table = softn->ipf_nat_table[0];
1193 		nsp->ns_side[1].ns_table = softn->ipf_nat_table[1];
1194 		nsp->ns_list = softn->ipf_nat_list;
1195 		nsp->ns_maptable = softn->ipf_hm_maptable;
1196 		nsp->ns_maplist = softn->ipf_hm_maplist;
1197 		nsp->ns_nattab_sz = softn->ipf_nat_table_sz;
1198 		nsp->ns_nattab_max = softn->ipf_nat_table_max;
1199 		nsp->ns_rultab_sz = softn->ipf_nat_maprules_sz;
1200 		nsp->ns_rdrtab_sz = softn->ipf_nat_rdrrules_sz;
1201 		nsp->ns_hostmap_sz = softn->ipf_nat_hostmap_sz;
1202 		nsp->ns_instances = softn->ipf_nat_instances;
1203 		nsp->ns_ticks = softc->ipf_ticks;
1204 #ifdef IPFILTER_LOGGING
1205 		nsp->ns_log_ok = ipf_log_logok(softc, IPF_LOGNAT);
1206 		nsp->ns_log_fail = ipf_log_failures(softc, IPF_LOGNAT);
1207 #else
1208 		nsp->ns_log_ok = 0;
1209 		nsp->ns_log_fail = 0;
1210 #endif
1211 		error = ipf_outobj(softc, data, nsp, IPFOBJ_NATSTAT);
1212 		break;
1213 	    }
1214 
1215 	case SIOCGNATL :
1216 	    {
1217 		natlookup_t nl;
1218 
1219 		error = ipf_inobj(softc, data, NULL, &nl, IPFOBJ_NATLOOKUP);
1220 		if (error == 0) {
1221 			void *ptr;
1222 
1223 			if (getlock) {
1224 				READ_ENTER(&softc->ipf_nat);
1225 			}
1226 
1227 			switch (nl.nl_v)
1228 			{
1229 			case 4 :
1230 				ptr = ipf_nat_lookupredir(&nl);
1231 				break;
1232 #ifdef USE_INET6
1233 			case 6 :
1234 				ptr = ipf_nat6_lookupredir(&nl);
1235 				break;
1236 #endif
1237 			default:
1238 				ptr = NULL;
1239 				break;
1240 			}
1241 
1242 			if (getlock) {
1243 				RWLOCK_EXIT(&softc->ipf_nat);
1244 			}
1245 			if (ptr != NULL) {
1246 				error = ipf_outobj(softc, data, &nl,
1247 						   IPFOBJ_NATLOOKUP);
1248 			} else {
1249 				IPFERROR(60011);
1250 				error = ESRCH;
1251 			}
1252 		}
1253 		break;
1254 	    }
1255 
1256 	case SIOCIPFFL :	/* old SIOCFLNAT & SIOCCNATL */
1257 		if (!(mode & FWRITE)) {
1258 			IPFERROR(60012);
1259 			error = EPERM;
1260 			break;
1261 		}
1262 		if (getlock) {
1263 			WRITE_ENTER(&softc->ipf_nat);
1264 		}
1265 
1266 		error = BCOPYIN(data, &arg, sizeof(arg));
1267 		if (error != 0) {
1268 			IPFERROR(60013);
1269 			error = EFAULT;
1270 		} else {
1271 			if (arg == 0)
1272 				ret = ipf_nat_flushtable(softc, softn);
1273 			else if (arg == 1)
1274 				ret = ipf_nat_clearlist(softc, softn);
1275 			else
1276 				ret = ipf_nat_extraflush(softc, softn, arg);
1277 			ipf_proxy_flush(softc->ipf_proxy_soft, arg);
1278 		}
1279 
1280 		if (getlock) {
1281 			RWLOCK_EXIT(&softc->ipf_nat);
1282 		}
1283 		if (error == 0) {
1284 			error = BCOPYOUT(&ret, data, sizeof(ret));
1285 		}
1286 		break;
1287 
1288 	case SIOCMATCHFLUSH :
1289 		if (!(mode & FWRITE)) {
1290 			IPFERROR(60014);
1291 			error = EPERM;
1292 			break;
1293 		}
1294 		if (getlock) {
1295 			WRITE_ENTER(&softc->ipf_nat);
1296 		}
1297 
1298 		error = ipf_nat_matchflush(softc, softn, data);
1299 
1300 		if (getlock) {
1301 			RWLOCK_EXIT(&softc->ipf_nat);
1302 		}
1303 		break;
1304 
1305 	case SIOCPROXY :
1306 		error = ipf_proxy_ioctl(softc, data, cmd, mode, ctx);
1307 		break;
1308 
1309 	case SIOCSTLCK :
1310 		if (!(mode & FWRITE)) {
1311 			IPFERROR(60015);
1312 			error = EPERM;
1313 		} else {
1314 			error = ipf_lock(data, &softn->ipf_nat_lock);
1315 		}
1316 		break;
1317 
1318 	case SIOCSTPUT :
1319 		if ((mode & FWRITE) != 0) {
1320 			error = ipf_nat_putent(softc, data, getlock);
1321 		} else {
1322 			IPFERROR(60016);
1323 			error = EACCES;
1324 		}
1325 		break;
1326 
1327 	case SIOCSTGSZ :
1328 		if (softn->ipf_nat_lock) {
1329 			error = ipf_nat_getsz(softc, data, getlock);
1330 		} else {
1331 			IPFERROR(60017);
1332 			error = EACCES;
1333 		}
1334 		break;
1335 
1336 	case SIOCSTGET :
1337 		if (softn->ipf_nat_lock) {
1338 			error = ipf_nat_getent(softc, data, getlock);
1339 		} else {
1340 			IPFERROR(60018);
1341 			error = EACCES;
1342 		}
1343 		break;
1344 
1345 	case SIOCGENITER :
1346 	    {
1347 		ipfgeniter_t iter;
1348 		ipftoken_t *token;
1349 		ipfobj_t obj;
1350 
1351 		error = ipf_inobj(softc, data, &obj, &iter, IPFOBJ_GENITER);
1352 		if (error != 0)
1353 			break;
1354 
1355 		SPL_SCHED(s);
1356 		token = ipf_token_find(softc, iter.igi_type, uid, ctx);
1357 		if (token != NULL) {
1358 			error  = ipf_nat_iterator(softc, token, &iter, &obj);
1359 			WRITE_ENTER(&softc->ipf_tokens);
1360 			ipf_token_deref(softc, token);
1361 			RWLOCK_EXIT(&softc->ipf_tokens);
1362 		}
1363 		SPL_X(s);
1364 		break;
1365 	    }
1366 
1367 	case SIOCIPFDELTOK :
1368 		error = BCOPYIN(data, &arg, sizeof(arg));
1369 		if (error == 0) {
1370 			SPL_SCHED(s);
1371 			error = ipf_token_del(softc, arg, uid, ctx);
1372 			SPL_X(s);
1373 		} else {
1374 			IPFERROR(60019);
1375 			error = EFAULT;
1376 		}
1377 		break;
1378 
1379 	case SIOCGTQTAB :
1380 		error = ipf_outobj(softc, data, softn->ipf_nat_tcptq,
1381 				   IPFOBJ_STATETQTAB);
1382 		break;
1383 
1384 	case SIOCGTABL :
1385 		error = ipf_nat_gettable(softc, softn, data);
1386 		break;
1387 
1388 	default :
1389 		IPFERROR(60020);
1390 		error = EINVAL;
1391 		break;
1392 	}
1393 done:
1394 	if (nat != NULL)
1395 		ipf_nat_rule_fini(softc, nat);
1396 	if (nt != NULL)
1397 		KFREES(nt, nt->in_size);
1398 	return error;
1399 }
1400 
1401 
1402 /* ------------------------------------------------------------------------ */
1403 /* Function:    ipf_nat_siocaddnat                                          */
1404 /* Returns:     int - 0 == success, != 0 == failure                         */
1405 /* Parameters:  softc(I) - pointer to soft context main structure           */
1406 /*              softn(I) - pointer to NAT context structure                 */
1407 /*              n(I)       - pointer to new NAT rule                        */
1408 /*              np(I)      - pointer to where to insert new NAT rule        */
1409 /*              getlock(I) - flag indicating if lock on  is held            */
1410 /* Mutex Locks: ipf_nat_io                                                  */
1411 /*                                                                          */
1412 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
1413 /* from information passed to the kernel, then add it  to the appropriate   */
1414 /* NAT rule table(s).                                                       */
1415 /* ------------------------------------------------------------------------ */
1416 static int
1417 ipf_nat_siocaddnat(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *n,
1418 	int getlock)
1419 {
1420 	int error = 0;
1421 
1422 	if (ipf_nat_resolverule(softc, n) != 0) {
1423 		IPFERROR(60022);
1424 		return ENOENT;
1425 	}
1426 
1427 	if ((n->in_age[0] == 0) && (n->in_age[1] != 0)) {
1428 		IPFERROR(60023);
1429 		return EINVAL;
1430 	}
1431 
1432 	if (n->in_redir == (NAT_DIVERTUDP|NAT_MAP)) {
1433 		/*
1434 		 * Prerecord whether or not the destination of the divert
1435 		 * is local or not to the interface the packet is going
1436 		 * to be sent out.
1437 		 */
1438 		n->in_dlocal = ipf_deliverlocal(softc, n->in_v[1],
1439 						n->in_ifps[1], &n->in_ndstip6);
1440 	}
1441 
1442 	if (getlock) {
1443 		WRITE_ENTER(&softc->ipf_nat);
1444 	}
1445 	n->in_next = NULL;
1446 	n->in_pnext = softn->ipf_nat_list_tail;
1447 	*n->in_pnext = n;
1448 	softn->ipf_nat_list_tail = &n->in_next;
1449 	n->in_use++;
1450 
1451 	if (n->in_redir & NAT_REDIRECT) {
1452 		n->in_flags &= ~IPN_NOTDST;
1453 		switch (n->in_v[0])
1454 		{
1455 		case 4 :
1456 			ipf_nat_addrdr(softn, n);
1457 			break;
1458 #ifdef USE_INET6
1459 		case 6 :
1460 			ipf_nat6_addrdr(softn, n);
1461 			break;
1462 #endif
1463 		default :
1464 			break;
1465 		}
1466 		ATOMIC_INC32(softn->ipf_nat_stats.ns_rules_rdr);
1467 	}
1468 
1469 	if (n->in_redir & (NAT_MAP|NAT_MAPBLK)) {
1470 		n->in_flags &= ~IPN_NOTSRC;
1471 		switch (n->in_v[0])
1472 		{
1473 		case 4 :
1474 			ipf_nat_addmap(softn, n);
1475 			break;
1476 #ifdef USE_INET6
1477 		case 6 :
1478 			ipf_nat6_addmap(softn, n);
1479 			break;
1480 #endif
1481 		default :
1482 			break;
1483 		}
1484 		ATOMIC_INC32(softn->ipf_nat_stats.ns_rules_map);
1485 	}
1486 
1487 	if (n->in_age[0] != 0)
1488 		n->in_tqehead[0] = ipf_addtimeoutqueue(softc,
1489 						       &softn->ipf_nat_utqe,
1490 						       n->in_age[0]);
1491 
1492 	if (n->in_age[1] != 0)
1493 		n->in_tqehead[1] = ipf_addtimeoutqueue(softc,
1494 						       &softn->ipf_nat_utqe,
1495 						       n->in_age[1]);
1496 
1497 	MUTEX_INIT(&n->in_lock, "ipnat rule lock");
1498 
1499 	n = NULL;
1500 	ATOMIC_INC32(softn->ipf_nat_stats.ns_rules);
1501 #if SOLARIS && !defined(INSTANCES)
1502 	pfil_delayed_copy = 0;
1503 #endif
1504 	if (getlock) {
1505 		RWLOCK_EXIT(&softc->ipf_nat);			/* WRITE */
1506 	}
1507 
1508 	return error;
1509 }
1510 
1511 
1512 /* ------------------------------------------------------------------------ */
1513 /* Function:    ipf_nat_ruleaddrinit                                        */
1514 /* Parameters:  softc(I) - pointer to soft context main structure           */
1515 /*              softn(I) - pointer to NAT context structure                 */
1516 /*              n(I)     - pointer to NAT rule                              */
1517 /*                                                                          */
1518 /* Initialise all of the NAT address structures in a NAT rule.              */
1519 /* ------------------------------------------------------------------------ */
1520 static int
1521 ipf_nat_ruleaddrinit(ipf_main_softc_t *softc, ipf_nat_softc_t *softn,
1522     ipnat_t *n)
1523 {
1524 	int idx, error;
1525 
1526 	if ((n->in_ndst.na_atype == FRI_LOOKUP) &&
1527 	    (n->in_ndst.na_type != IPLT_DSTLIST)) {
1528 		IPFERROR(60071);
1529 		return EINVAL;
1530 	}
1531 	if ((n->in_nsrc.na_atype == FRI_LOOKUP) &&
1532 	    (n->in_nsrc.na_type != IPLT_DSTLIST)) {
1533 		IPFERROR(60069);
1534 		return EINVAL;
1535 	}
1536 
1537 	if (n->in_redir == NAT_BIMAP) {
1538 		n->in_ndstaddr = n->in_osrcaddr;
1539 		n->in_ndstmsk = n->in_osrcmsk;
1540 		n->in_odstaddr = n->in_nsrcaddr;
1541 		n->in_odstmsk = n->in_nsrcmsk;
1542 
1543 	}
1544 
1545 	if (n->in_redir & NAT_REDIRECT)
1546 		idx = 1;
1547 	else
1548 		idx = 0;
1549 	/*
1550 	 * Initialise all of the address fields.
1551 	 */
1552 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_osrc, 1,
1553 				     n->in_ifps[idx]);
1554 	if (error != 0)
1555 		return error;
1556 
1557 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_odst, 1,
1558 				     n->in_ifps[idx]);
1559 	if (error != 0)
1560 		return error;
1561 
1562 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_nsrc, 1,
1563 				     n->in_ifps[idx]);
1564 	if (error != 0)
1565 		return error;
1566 
1567 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_ndst, 1,
1568 				     n->in_ifps[idx]);
1569 	if (error != 0)
1570 		return error;
1571 
1572 	if (n->in_redir & NAT_DIVERTUDP)
1573 		ipf_nat_builddivertmp(softn, n);
1574 
1575 	return 0;
1576 }
1577 
1578 
1579 /* ------------------------------------------------------------------------ */
1580 /* Function:    ipf_nat_resolvrule                                          */
1581 /* Returns:     Nil                                                         */
1582 /* Parameters:  softc(I) - pointer to soft context main structure           */
1583 /*              n(I)     - pointer to NAT rule                              */
1584 /*                                                                          */
1585 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
1586 /* from information passed to the kernel, then add it  to the appropriate   */
1587 /* NAT rule table(s).                                                       */
1588 /* ------------------------------------------------------------------------ */
1589 static int
1590 ipf_nat_resolverule(ipf_main_softc_t *softc, ipnat_t *n)
1591 {
1592 	char *base;
1593 
1594 	base = n->in_names;
1595 
1596 	n->in_ifps[0] = ipf_resolvenic(softc, base + n->in_ifnames[0],
1597 				       n->in_v[0]);
1598 
1599 	if (n->in_ifnames[1] == -1) {
1600 		n->in_ifnames[1] = n->in_ifnames[0];
1601 		n->in_ifps[1] = n->in_ifps[0];
1602 	} else {
1603 		n->in_ifps[1] = ipf_resolvenic(softc, base + n->in_ifnames[1],
1604 					       n->in_v[1]);
1605 	}
1606 
1607 	if (n->in_plabel != -1) {
1608 		if (n->in_redir & NAT_REDIRECT)
1609 			n->in_apr = ipf_proxy_lookup(softc->ipf_proxy_soft,
1610 						     n->in_pr[0],
1611 						     base + n->in_plabel);
1612 		else
1613 			n->in_apr = ipf_proxy_lookup(softc->ipf_proxy_soft,
1614 						     n->in_pr[1],
1615 						     base + n->in_plabel);
1616 		if (n->in_apr == NULL)
1617 			return -1;
1618 	}
1619 	return 0;
1620 }
1621 
1622 
1623 /* ------------------------------------------------------------------------ */
1624 /* Function:    ipf_nat_siocdelnat                                          */
1625 /* Returns:     int - 0 == success, != 0 == failure                         */
1626 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1627 /*              softn(I)   - pointer to NAT context structure               */
1628 /*              n(I)       - pointer to new NAT rule                        */
1629 /*              getlock(I) - flag indicating if lock on  is held            */
1630 /* Mutex Locks: ipf_nat_io                                                  */
1631 /*                                                                          */
1632 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
1633 /* from information passed to the kernel, then add it  to the appropriate   */
1634 /* NAT rule table(s).                                                       */
1635 /* ------------------------------------------------------------------------ */
1636 static void
1637 ipf_nat_siocdelnat(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *n,
1638 	int getlock)
1639 {
1640 #ifdef IPF_NAT6
1641 	int i;
1642 #endif
1643 
1644 	if (getlock) {
1645 		WRITE_ENTER(&softc->ipf_nat);
1646 	}
1647 
1648 	ipf_nat_delrule(softc, softn, n, 1);
1649 
1650 	if (getlock) {
1651 		RWLOCK_EXIT(&softc->ipf_nat);			/* READ/WRITE */
1652 	}
1653 }
1654 
1655 
1656 /* ------------------------------------------------------------------------ */
1657 /* Function:    ipf_nat_getsz                                               */
1658 /* Returns:     int - 0 == success, != 0 is the error value.                */
1659 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1660 /*              data(I)    - pointer to natget structure with kernel        */
1661 /*                           pointer get the size of.                       */
1662 /*              getlock(I) - flag indicating whether or not the caller      */
1663 /*                           holds a lock on ipf_nat                        */
1664 /*                                                                          */
1665 /* Handle SIOCSTGSZ.                                                        */
1666 /* Return the size of the nat list entry to be copied back to user space.   */
1667 /* The size of the entry is stored in the ng_sz field and the enture natget */
1668 /* structure is copied back to the user.                                    */
1669 /* ------------------------------------------------------------------------ */
1670 static int
1671 ipf_nat_getsz(ipf_main_softc_t *softc, void *data, int getlock)
1672 {
1673 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1674 	ap_session_t *aps;
1675 	nat_t *nat, *n;
1676 	natget_t ng;
1677 	int error;
1678 
1679 	error = BCOPYIN(data, &ng, sizeof(ng));
1680 	if (error != 0) {
1681 		IPFERROR(60024);
1682 		return EFAULT;
1683 	}
1684 
1685 	if (getlock) {
1686 		READ_ENTER(&softc->ipf_nat);
1687 	}
1688 
1689 	nat = ng.ng_ptr;
1690 	if (!nat) {
1691 		nat = softn->ipf_nat_instances;
1692 		ng.ng_sz = 0;
1693 		/*
1694 		 * Empty list so the size returned is 0.  Simple.
1695 		 */
1696 		if (nat == NULL) {
1697 			if (getlock) {
1698 				RWLOCK_EXIT(&softc->ipf_nat);
1699 			}
1700 			error = BCOPYOUT(&ng, data, sizeof(ng));
1701 			if (error != 0) {
1702 				IPFERROR(60025);
1703 				return EFAULT;
1704 			}
1705 			return 0;
1706 		}
1707 	} else {
1708 		/*
1709 		 * Make sure the pointer we're copying from exists in the
1710 		 * current list of entries.  Security precaution to prevent
1711 		 * copying of random kernel data.
1712 		 */
1713 		for (n = softn->ipf_nat_instances; n; n = n->nat_next)
1714 			if (n == nat)
1715 				break;
1716 		if (n == NULL) {
1717 			if (getlock) {
1718 				RWLOCK_EXIT(&softc->ipf_nat);
1719 			}
1720 			IPFERROR(60026);
1721 			return ESRCH;
1722 		}
1723 	}
1724 
1725 	/*
1726 	 * Incluse any space required for proxy data structures.
1727 	 */
1728 	ng.ng_sz = sizeof(nat_save_t);
1729 	aps = nat->nat_aps;
1730 	if (aps != NULL) {
1731 		ng.ng_sz += sizeof(ap_session_t) - 4;
1732 		if (aps->aps_data != 0)
1733 			ng.ng_sz += aps->aps_psiz;
1734 	}
1735 	if (getlock) {
1736 		RWLOCK_EXIT(&softc->ipf_nat);
1737 	}
1738 
1739 	error = BCOPYOUT(&ng, data, sizeof(ng));
1740 	if (error != 0) {
1741 		IPFERROR(60027);
1742 		return EFAULT;
1743 	}
1744 	return 0;
1745 }
1746 
1747 
1748 /* ------------------------------------------------------------------------ */
1749 /* Function:    ipf_nat_getent                                              */
1750 /* Returns:     int - 0 == success, != 0 is the error value.                */
1751 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1752 /*              data(I)    - pointer to natget structure with kernel pointer*/
1753 /*                           to NAT structure to copy out.                  */
1754 /*              getlock(I) - flag indicating whether or not the caller      */
1755 /*                           holds a lock on ipf_nat                        */
1756 /*                                                                          */
1757 /* Handle SIOCSTGET.                                                        */
1758 /* Copies out NAT entry to user space.  Any additional data held for a      */
1759 /* proxy is also copied, as to is the NAT rule which was responsible for it */
1760 /* ------------------------------------------------------------------------ */
1761 static int
1762 ipf_nat_getent(ipf_main_softc_t *softc, void *data, int getlock)
1763 {
1764 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1765 	int error, outsize;
1766 	ap_session_t *aps;
1767 	nat_save_t *ipn, ipns;
1768 	nat_t *n, *nat;
1769 
1770 	error = ipf_inobj(softc, data, NULL, &ipns, IPFOBJ_NATSAVE);
1771 	if (error != 0)
1772 		return error;
1773 
1774 	if ((ipns.ipn_dsize < sizeof(ipns)) || (ipns.ipn_dsize > 81920)) {
1775 		IPFERROR(60028);
1776 		return EINVAL;
1777 	}
1778 
1779 	KMALLOCS(ipn, nat_save_t *, ipns.ipn_dsize);
1780 	if (ipn == NULL) {
1781 		IPFERROR(60029);
1782 		return ENOMEM;
1783 	}
1784 
1785 	if (getlock) {
1786 		READ_ENTER(&softc->ipf_nat);
1787 	}
1788 
1789 	ipn->ipn_dsize = ipns.ipn_dsize;
1790 	nat = ipns.ipn_next;
1791 	if (nat == NULL) {
1792 		nat = softn->ipf_nat_instances;
1793 		if (nat == NULL) {
1794 			if (softn->ipf_nat_instances == NULL) {
1795 				IPFERROR(60030);
1796 				error = ENOENT;
1797 			}
1798 			goto finished;
1799 		}
1800 	} else {
1801 		/*
1802 		 * Make sure the pointer we're copying from exists in the
1803 		 * current list of entries.  Security precaution to prevent
1804 		 * copying of random kernel data.
1805 		 */
1806 		for (n = softn->ipf_nat_instances; n; n = n->nat_next)
1807 			if (n == nat)
1808 				break;
1809 		if (n == NULL) {
1810 			IPFERROR(60031);
1811 			error = ESRCH;
1812 			goto finished;
1813 		}
1814 	}
1815 	ipn->ipn_next = nat->nat_next;
1816 
1817 	/*
1818 	 * Copy the NAT structure.
1819 	 */
1820 	bcopy((char *)nat, &ipn->ipn_nat, sizeof(*nat));
1821 
1822 	/*
1823 	 * If we have a pointer to the NAT rule it belongs to, save that too.
1824 	 */
1825 	if (nat->nat_ptr != NULL)
1826 		bcopy((char *)nat->nat_ptr, (char *)&ipn->ipn_ipnat,
1827 		      ipn->ipn_ipnat.in_size);
1828 
1829 	/*
1830 	 * If we also know the NAT entry has an associated filter rule,
1831 	 * save that too.
1832 	 */
1833 	if (nat->nat_fr != NULL)
1834 		bcopy((char *)nat->nat_fr, (char *)&ipn->ipn_fr,
1835 		      sizeof(ipn->ipn_fr));
1836 
1837 	/*
1838 	 * Last but not least, if there is an application proxy session set
1839 	 * up for this NAT entry, then copy that out too, including any
1840 	 * private data saved along side it by the proxy.
1841 	 */
1842 	aps = nat->nat_aps;
1843 	outsize = ipn->ipn_dsize - sizeof(*ipn) + sizeof(ipn->ipn_data);
1844 	if (aps != NULL) {
1845 		char *s;
1846 
1847 		if (outsize < sizeof(*aps)) {
1848 			IPFERROR(60032);
1849 			error = ENOBUFS;
1850 			goto finished;
1851 		}
1852 
1853 		s = ipn->ipn_data;
1854 		bcopy((char *)aps, s, sizeof(*aps));
1855 		s += sizeof(*aps);
1856 		outsize -= sizeof(*aps);
1857 		if ((aps->aps_data != NULL) && (outsize >= aps->aps_psiz))
1858 			bcopy(aps->aps_data, s, aps->aps_psiz);
1859 		else {
1860 			IPFERROR(60033);
1861 			error = ENOBUFS;
1862 		}
1863 	}
1864 	if (error == 0) {
1865 		if (getlock) {
1866 			READ_ENTER(&softc->ipf_nat);
1867 			getlock = 0;
1868 		}
1869 		error = ipf_outobjsz(softc, data, ipn, IPFOBJ_NATSAVE,
1870 				     ipns.ipn_dsize);
1871 	}
1872 
1873 finished:
1874 	if (getlock) {
1875 		READ_ENTER(&softc->ipf_nat);
1876 	}
1877 	if (ipn != NULL) {
1878 		KFREES(ipn, ipns.ipn_dsize);
1879 	}
1880 	return error;
1881 }
1882 
1883 
1884 /* ------------------------------------------------------------------------ */
1885 /* Function:    ipf_nat_putent                                              */
1886 /* Returns:     int - 0 == success, != 0 is the error value.                */
1887 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1888 /*              data(I)    - pointer to natget structure with NAT           */
1889 /*                           structure information to load into the kernel  */
1890 /*              getlock(I) - flag indicating whether or not a write lock    */
1891 /*                           on is already held.                            */
1892 /*                                                                          */
1893 /* Handle SIOCSTPUT.                                                        */
1894 /* Loads a NAT table entry from user space, including a NAT rule, proxy and */
1895 /* firewall rule data structures, if pointers to them indicate so.          */
1896 /* ------------------------------------------------------------------------ */
1897 static int
1898 ipf_nat_putent(ipf_main_softc_t *softc, void *data, int getlock)
1899 {
1900 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1901 	nat_save_t *ipn, *ipnn;
1902 	ap_session_t *aps;
1903 	nat_t *n, *nat;
1904 	frentry_t *fr;
1905 	fr_info_t *fin;
1906 	ipnat_t *in;
1907 	int error;
1908 
1909 	KMALLOC(ipn, nat_save_t *);
1910 	if (ipn == NULL)
1911 		return ENOMEM;
1912 	error = ipf_inobj(softc, data, NULL, ipn, IPFOBJ_NATSAVE);
1913 	if (error != 0)
1914 		return error;
1915 
1916 	/*
1917 	 * Initialise early because of code at junkput label.
1918 	 */
1919 	n = NULL;
1920 	in = NULL;
1921 	aps = NULL;
1922 	nat = NULL;
1923 	ipnn = NULL;
1924 	fin = NULL;
1925 	fr = NULL;
1926 
1927 	/*
1928 	 * New entry, copy in the rest of the NAT entry if it's size is more
1929 	 * than just the nat_t structure.
1930 	 */
1931 	if (ipn->ipn_dsize > sizeof(*ipn)) {
1932 		if (ipn->ipn_dsize > 81920) {
1933 			IPFERROR(60034);
1934 			error = ENOMEM;
1935 			goto junkput;
1936 		}
1937 
1938 		KMALLOCS(ipnn, nat_save_t *, ipn->ipn_dsize);
1939 		if (ipnn == NULL) {
1940 			IPFERROR(60035);
1941 			return ENOMEM;
1942 		}
1943 
1944 		bzero(ipnn, ipn->ipn_dsize);
1945 		error = ipf_inobjsz(softc, data, ipnn, IPFOBJ_NATSAVE,
1946 				    ipn->ipn_dsize);
1947 		if (error != 0) {
1948 			goto junkput;
1949 		}
1950 	} else
1951 		ipnn = ipn;
1952 
1953 	KMALLOC(nat, nat_t *);
1954 	if (nat == NULL) {
1955 		IPFERROR(60037);
1956 		error = ENOMEM;
1957 		goto junkput;
1958 	}
1959 
1960 	bcopy((char *)&ipnn->ipn_nat, (char *)nat, sizeof(*nat));
1961 
1962 	switch (nat->nat_v[0])
1963 	{
1964 	case 4:
1965 #ifdef USE_IENT6
1966 	case 6 :
1967 #endif
1968 		break;
1969 	default :
1970 		IPFERROR(60061);
1971 		error = EPROTONOSUPPORT;
1972 		goto junkput;
1973 		/*NOTREACHED*/
1974 	}
1975 
1976 	/*
1977 	 * Initialize all these so that ipf_nat_delete() doesn't cause a crash.
1978 	 */
1979 	bzero((char *)nat, offsetof(struct nat, nat_tqe));
1980 	nat->nat_tqe.tqe_pnext = NULL;
1981 	nat->nat_tqe.tqe_next = NULL;
1982 	nat->nat_tqe.tqe_ifq = NULL;
1983 	nat->nat_tqe.tqe_parent = nat;
1984 
1985 	/*
1986 	 * Restore the rule associated with this nat session
1987 	 */
1988 	in = ipnn->ipn_nat.nat_ptr;
1989 	if (in != NULL) {
1990 		KMALLOCS(in, ipnat_t *, ipnn->ipn_ipnat.in_size);
1991 		nat->nat_ptr = in;
1992 		if (in == NULL) {
1993 			IPFERROR(60038);
1994 			error = ENOMEM;
1995 			goto junkput;
1996 		}
1997 		bcopy((char *)&ipnn->ipn_ipnat, (char *)in,
1998 		      ipnn->ipn_ipnat.in_size);
1999 		in->in_use = 1;
2000 		in->in_flags |= IPN_DELETE;
2001 
2002 		ATOMIC_INC32(softn->ipf_nat_stats.ns_rules);
2003 
2004 		if (ipf_nat_resolverule(softc, in) != 0) {
2005 			IPFERROR(60039);
2006 			error = ESRCH;
2007 			goto junkput;
2008 		}
2009 	}
2010 
2011 	/*
2012 	 * Check that the NAT entry doesn't already exist in the kernel.
2013 	 *
2014 	 * For NAT_OUTBOUND, we're lookup for a duplicate MAP entry.  To do
2015 	 * this, we check to see if the inbound combination of addresses and
2016 	 * ports is already known.  Similar logic is applied for NAT_INBOUND.
2017 	 *
2018 	 */
2019 	KMALLOC(fin, fr_info_t *);
2020 	if (fin == NULL) {
2021 		error = ENOMEM;
2022 		goto junkput;
2023 	}
2024 	bzero(fin, sizeof(*fin));
2025 	fin->fin_v = nat->nat_v[0];
2026 	fin->fin_p = nat->nat_pr[0];
2027 	fin->fin_rev = nat->nat_rev;
2028 	fin->fin_ifp = nat->nat_ifps[0];
2029 	fin->fin_data[0] = ntohs(nat->nat_ndport);
2030 	fin->fin_data[1] = ntohs(nat->nat_nsport);
2031 
2032 	switch (nat->nat_dir)
2033 	{
2034 	case NAT_OUTBOUND :
2035 	case NAT_DIVERTOUT :
2036 		if (getlock) {
2037 			READ_ENTER(&softc->ipf_nat);
2038 		}
2039 
2040 		fin->fin_v = nat->nat_v[1];
2041 		if (nat->nat_v[1] == 4) {
2042 			n = ipf_nat_inlookup(fin, nat->nat_flags, fin->fin_p,
2043 					     nat->nat_ndstip, nat->nat_nsrcip);
2044 #ifdef USE_INET6
2045 		} else if (nat->nat_v[1] == 6) {
2046 			n = ipf_nat6_inlookup(fin, nat->nat_flags, fin->fin_p,
2047 					      &nat->nat_ndst6.in6,
2048 					      &nat->nat_nsrc6.in6);
2049 #endif
2050 		}
2051 
2052 		if (getlock) {
2053 			RWLOCK_EXIT(&softc->ipf_nat);
2054 		}
2055 		if (n != NULL) {
2056 			IPFERROR(60040);
2057 			error = EEXIST;
2058 			goto junkput;
2059 		}
2060 		break;
2061 
2062 	case NAT_INBOUND :
2063 	case NAT_DIVERTIN :
2064 		if (getlock) {
2065 			READ_ENTER(&softc->ipf_nat);
2066 		}
2067 
2068 		if (fin->fin_v == 4) {
2069 			n = ipf_nat_outlookup(fin, nat->nat_flags, fin->fin_p,
2070 					      nat->nat_ndstip,
2071 					      nat->nat_nsrcip);
2072 #ifdef USE_INET6
2073 		} else if (fin->fin_v == 6) {
2074 			n = ipf_nat6_outlookup(fin, nat->nat_flags, fin->fin_p,
2075 					       &nat->nat_ndst6.in6,
2076 					       &nat->nat_nsrc6.in6);
2077 #endif
2078 		}
2079 
2080 		if (getlock) {
2081 			RWLOCK_EXIT(&softc->ipf_nat);
2082 		}
2083 		if (n != NULL) {
2084 			IPFERROR(60041);
2085 			error = EEXIST;
2086 			goto junkput;
2087 		}
2088 		break;
2089 
2090 	default :
2091 		IPFERROR(60042);
2092 		error = EINVAL;
2093 		goto junkput;
2094 	}
2095 
2096 	/*
2097 	 * Restore ap_session_t structure.  Include the private data allocated
2098 	 * if it was there.
2099 	 */
2100 	aps = nat->nat_aps;
2101 	if (aps != NULL) {
2102 		KMALLOC(aps, ap_session_t *);
2103 		nat->nat_aps = aps;
2104 		if (aps == NULL) {
2105 			IPFERROR(60043);
2106 			error = ENOMEM;
2107 			goto junkput;
2108 		}
2109 		bcopy(ipnn->ipn_data, (char *)aps, sizeof(*aps));
2110 		if (in != NULL)
2111 			aps->aps_apr = in->in_apr;
2112 		else
2113 			aps->aps_apr = NULL;
2114 		if (aps->aps_psiz != 0) {
2115 			if (aps->aps_psiz > 81920) {
2116 				IPFERROR(60044);
2117 				error = ENOMEM;
2118 				goto junkput;
2119 			}
2120 			KMALLOCS(aps->aps_data, void *, aps->aps_psiz);
2121 			if (aps->aps_data == NULL) {
2122 				IPFERROR(60045);
2123 				error = ENOMEM;
2124 				goto junkput;
2125 			}
2126 			bcopy(ipnn->ipn_data + sizeof(*aps), aps->aps_data,
2127 			      aps->aps_psiz);
2128 		} else {
2129 			aps->aps_psiz = 0;
2130 			aps->aps_data = NULL;
2131 		}
2132 	}
2133 
2134 	/*
2135 	 * If there was a filtering rule associated with this entry then
2136 	 * build up a new one.
2137 	 */
2138 	fr = nat->nat_fr;
2139 	if (fr != NULL) {
2140 		if ((nat->nat_flags & SI_NEWFR) != 0) {
2141 			KMALLOC(fr, frentry_t *);
2142 			nat->nat_fr = fr;
2143 			if (fr == NULL) {
2144 				IPFERROR(60046);
2145 				error = ENOMEM;
2146 				goto junkput;
2147 			}
2148 			ipnn->ipn_nat.nat_fr = fr;
2149 			fr->fr_ref = 1;
2150 			(void) ipf_outobj(softc, data, ipnn, IPFOBJ_NATSAVE);
2151 			bcopy((char *)&ipnn->ipn_fr, (char *)fr, sizeof(*fr));
2152 
2153 			fr->fr_ref = 1;
2154 			fr->fr_dsize = 0;
2155 			fr->fr_data = NULL;
2156 			fr->fr_type = FR_T_NONE;
2157 
2158 			MUTEX_NUKE(&fr->fr_lock);
2159 			MUTEX_INIT(&fr->fr_lock, "nat-filter rule lock");
2160 		} else {
2161 			if (getlock) {
2162 				READ_ENTER(&softc->ipf_nat);
2163 			}
2164 			for (n = softn->ipf_nat_instances; n; n = n->nat_next)
2165 				if (n->nat_fr == fr)
2166 					break;
2167 
2168 			if (n != NULL) {
2169 				MUTEX_ENTER(&fr->fr_lock);
2170 				fr->fr_ref++;
2171 				MUTEX_EXIT(&fr->fr_lock);
2172 			}
2173 			if (getlock) {
2174 				RWLOCK_EXIT(&softc->ipf_nat);
2175 			}
2176 
2177 			if (n == NULL) {
2178 				IPFERROR(60047);
2179 				error = ESRCH;
2180 				goto junkput;
2181 			}
2182 		}
2183 	}
2184 
2185 	if (ipnn != ipn) {
2186 		KFREES(ipnn, ipn->ipn_dsize);
2187 		ipnn = NULL;
2188 	}
2189 
2190 	if (getlock) {
2191 		WRITE_ENTER(&softc->ipf_nat);
2192 	}
2193 
2194 	if (fin->fin_v == 4)
2195 		error = ipf_nat_finalise(fin, nat);
2196 #ifdef USE_INET6
2197 	else
2198 		error = ipf_nat6_finalise(fin, nat);
2199 #endif
2200 
2201 	if (getlock) {
2202 		RWLOCK_EXIT(&softc->ipf_nat);
2203 	}
2204 
2205 	if (error == 0)
2206 		return 0;
2207 
2208 	IPFERROR(60048);
2209 	error = ENOMEM;
2210 
2211 junkput:
2212 	if (fin != NULL)
2213 		KFREE(fin);
2214 	if (fr != NULL) {
2215 		(void) ipf_derefrule(softc, &fr);
2216 	}
2217 
2218 	if ((ipnn != NULL) && (ipnn != ipn)) {
2219 		KFREES(ipnn, ipn->ipn_dsize);
2220 	}
2221 	if (ipn != NULL)
2222 		KFREE(ipn);
2223 	if (nat != NULL) {
2224 		if (aps != NULL) {
2225 			if (aps->aps_data != NULL) {
2226 				KFREES(aps->aps_data, aps->aps_psiz);
2227 			}
2228 			KFREE(aps);
2229 		}
2230 		if (in != NULL) {
2231 			if (in->in_apr)
2232 				ipf_proxy_deref(in->in_apr);
2233 			KFREES(in, in->in_size);
2234 		}
2235 		KFREE(nat);
2236 	}
2237 	return error;
2238 }
2239 
2240 
2241 /* ------------------------------------------------------------------------ */
2242 /* Function:    ipf_nat_delete                                              */
2243 /* Returns:     Nil                                                         */
2244 /* Parameters:  softc(I)   - pointer to soft context main structure         */
2245 /*              nat(I)     - pointer to NAT structure to delete             */
2246 /*              logtype(I) - type of LOG record to create before deleting   */
2247 /* Write Lock:  ipf_nat                                                     */
2248 /*                                                                          */
2249 /* Delete a nat entry from the various lists and table.  If NAT logging is  */
2250 /* enabled then generate a NAT log record for this event.                   */
2251 /* ------------------------------------------------------------------------ */
2252 void
2253 ipf_nat_delete(ipf_main_softc_t *softc, struct nat *nat, int logtype)
2254 {
2255 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2256 	int madeorphan = 0, removed = 0;
2257 	u_int bkt;
2258 	nat_stat_side_t *nss;
2259 	struct ipnat *ipn;
2260 
2261 	if (logtype != 0 && softn->ipf_nat_logging != 0)
2262 		ipf_nat_log(softc, softn, nat, logtype);
2263 
2264 	/*
2265 	 * Take it as a general indication that all the pointers are set if
2266 	 * nat_pnext is set.
2267 	 */
2268 	if (nat->nat_pnext != NULL) {
2269 		removed = 1;
2270 
2271 		bkt = nat->nat_hv[0] % softn->ipf_nat_table_sz;
2272 		nss = &softn->ipf_nat_stats.ns_side[0];
2273 		ASSERT(nss->ns_bucketlen[bkt] > 0);
2274 		nss->ns_bucketlen[bkt]--;
2275 		if (nss->ns_bucketlen[bkt] == 0) {
2276 			nss->ns_inuse--;
2277 		}
2278 
2279 		bkt = nat->nat_hv[1] % softn->ipf_nat_table_sz;
2280 		nss = &softn->ipf_nat_stats.ns_side[1];
2281 		ASSERT(nss->ns_bucketlen[bkt] > 0);
2282 		nss->ns_bucketlen[bkt]--;
2283 		if (nss->ns_bucketlen[bkt] == 0) {
2284 			nss->ns_inuse--;
2285 		}
2286 
2287 		*nat->nat_pnext = nat->nat_next;
2288 		if (nat->nat_next != NULL) {
2289 			nat->nat_next->nat_pnext = nat->nat_pnext;
2290 			nat->nat_next = NULL;
2291 		}
2292 		nat->nat_pnext = NULL;
2293 
2294 		*nat->nat_phnext[0] = nat->nat_hnext[0];
2295 		if (nat->nat_hnext[0] != NULL) {
2296 			nat->nat_hnext[0]->nat_phnext[0] = nat->nat_phnext[0];
2297 			nat->nat_hnext[0] = NULL;
2298 		}
2299 		nat->nat_phnext[0] = NULL;
2300 
2301 		*nat->nat_phnext[1] = nat->nat_hnext[1];
2302 		if (nat->nat_hnext[1] != NULL) {
2303 			nat->nat_hnext[1]->nat_phnext[1] = nat->nat_phnext[1];
2304 			nat->nat_hnext[1] = NULL;
2305 		}
2306 		nat->nat_phnext[1] = NULL;
2307 
2308 		if ((nat->nat_flags & SI_WILDP) != 0) {
2309 			ATOMIC_DEC32(softn->ipf_nat_stats.ns_wilds);
2310 		}
2311 		madeorphan = 1;
2312 	}
2313 
2314 	if (nat->nat_me != NULL) {
2315 		*nat->nat_me = NULL;
2316 		nat->nat_me = NULL;
2317 		nat->nat_ref--;
2318 		ASSERT(nat->nat_ref >= 0);
2319 	}
2320 
2321 	if (nat->nat_tqe.tqe_ifq != NULL) {
2322 		/*
2323 		 * No call to ipf_freetimeoutqueue() is made here, they are
2324 		 * garbage collected in ipf_nat_expire().
2325 		 */
2326 		(void) ipf_deletequeueentry(&nat->nat_tqe);
2327 	}
2328 
2329 	if (nat->nat_sync) {
2330 		ipf_sync_del_nat(softc->ipf_sync_soft, nat->nat_sync);
2331 		nat->nat_sync = NULL;
2332 	}
2333 
2334 	if (logtype == NL_EXPIRE)
2335 		softn->ipf_nat_stats.ns_expire++;
2336 
2337 	MUTEX_ENTER(&nat->nat_lock);
2338 	/*
2339 	 * NL_DESTROY should only be passed in when we've got nat_ref >= 2.
2340 	 * This happens when a nat'd packet is blocked and we want to throw
2341 	 * away the NAT session.
2342 	 */
2343 	if (logtype == NL_DESTROY) {
2344 		if (nat->nat_ref > 2) {
2345 			nat->nat_ref -= 2;
2346 			MUTEX_EXIT(&nat->nat_lock);
2347 			if (removed)
2348 				softn->ipf_nat_stats.ns_orphans++;
2349 			return;
2350 		}
2351 	} else if (nat->nat_ref > 1) {
2352 		nat->nat_ref--;
2353 		MUTEX_EXIT(&nat->nat_lock);
2354 		if (madeorphan == 1)
2355 			softn->ipf_nat_stats.ns_orphans++;
2356 		return;
2357 	}
2358 	ASSERT(nat->nat_ref >= 0);
2359 	MUTEX_EXIT(&nat->nat_lock);
2360 
2361 	nat->nat_ref = 0;
2362 
2363 	if (madeorphan == 0)
2364 		softn->ipf_nat_stats.ns_orphans--;
2365 
2366 	/*
2367 	 * At this point, nat_ref can be either 0 or -1
2368 	 */
2369 	softn->ipf_nat_stats.ns_proto[nat->nat_pr[0]]--;
2370 
2371 	if (nat->nat_fr != NULL) {
2372 		(void) ipf_derefrule(softc, &nat->nat_fr);
2373 	}
2374 
2375 	if (nat->nat_hm != NULL) {
2376 		ipf_nat_hostmapdel(softc, &nat->nat_hm);
2377 	}
2378 
2379 	/*
2380 	 * If there is an active reference from the nat entry to its parent
2381 	 * rule, decrement the rule's reference count and free it too if no
2382 	 * longer being used.
2383 	 */
2384 	ipn = nat->nat_ptr;
2385 	nat->nat_ptr = NULL;
2386 
2387 	if (ipn != NULL) {
2388 		ipn->in_space++;
2389 		ipf_nat_rule_deref(softc, &ipn);
2390 	}
2391 
2392 	if (nat->nat_aps != NULL) {
2393 		ipf_proxy_free(softc, nat->nat_aps);
2394 		nat->nat_aps = NULL;
2395 	}
2396 
2397 	MUTEX_DESTROY(&nat->nat_lock);
2398 
2399 	softn->ipf_nat_stats.ns_active--;
2400 
2401 	/*
2402 	 * If there's a fragment table entry too for this nat entry, then
2403 	 * dereference that as well.  This is after nat_lock is released
2404 	 * because of Tru64.
2405 	 */
2406 	ipf_frag_natforget(softc, (void *)nat);
2407 
2408 	KFREE(nat);
2409 }
2410 
2411 
2412 /* ------------------------------------------------------------------------ */
2413 /* Function:    ipf_nat_flushtable                                          */
2414 /* Returns:     int - number of NAT rules deleted                           */
2415 /* Parameters:  softc(I) - pointer to soft context main structure           */
2416 /*              softn(I) - pointer to NAT context structure                 */
2417 /* Write Lock:  ipf_nat                                                     */
2418 /*                                                                          */
2419 /* Deletes all currently active NAT sessions.  In deleting each NAT entry a */
2420 /* log record should be emitted in ipf_nat_delete() if NAT logging is       */
2421 /* enabled.                                                                 */
2422 /* ------------------------------------------------------------------------ */
2423 /*
2424  * nat_flushtable - clear the NAT table of all mapping entries.
2425  */
2426 static int
2427 ipf_nat_flushtable(ipf_main_softc_t *softc, ipf_nat_softc_t *softn)
2428 {
2429 	nat_t *nat;
2430 	int j = 0;
2431 
2432 	/*
2433 	 * ALL NAT mappings deleted, so lets just make the deletions
2434 	 * quicker.
2435 	 */
2436 	if (softn->ipf_nat_table[0] != NULL)
2437 		bzero((char *)softn->ipf_nat_table[0],
2438 		      sizeof(softn->ipf_nat_table[0]) *
2439 		      softn->ipf_nat_table_sz);
2440 	if (softn->ipf_nat_table[1] != NULL)
2441 		bzero((char *)softn->ipf_nat_table[1],
2442 		      sizeof(softn->ipf_nat_table[1]) *
2443 		      softn->ipf_nat_table_sz);
2444 
2445 	while ((nat = softn->ipf_nat_instances) != NULL) {
2446 		ipf_nat_delete(softc, nat, NL_FLUSH);
2447 		j++;
2448 	}
2449 
2450 	return j;
2451 }
2452 
2453 
2454 /* ------------------------------------------------------------------------ */
2455 /* Function:    ipf_nat_clearlist                                           */
2456 /* Returns:     int - number of NAT/RDR rules deleted                       */
2457 /* Parameters:  softc(I) - pointer to soft context main structure           */
2458 /*              softn(I) - pointer to NAT context structure                 */
2459 /*                                                                          */
2460 /* Delete all rules in the current list of rules.  There is nothing elegant */
2461 /* about this cleanup: simply free all entries on the list of rules and     */
2462 /* clear out the tables used for hashed NAT rule lookups.                   */
2463 /* ------------------------------------------------------------------------ */
2464 static int
2465 ipf_nat_clearlist(ipf_main_softc_t *softc, ipf_nat_softc_t *softn)
2466 {
2467 	ipnat_t *n;
2468 	int i = 0;
2469 
2470 	if (softn->ipf_nat_map_rules != NULL) {
2471 		bzero((char *)softn->ipf_nat_map_rules,
2472 		      sizeof(*softn->ipf_nat_map_rules) *
2473 		      softn->ipf_nat_maprules_sz);
2474 	}
2475 	if (softn->ipf_nat_rdr_rules != NULL) {
2476 		bzero((char *)softn->ipf_nat_rdr_rules,
2477 		      sizeof(*softn->ipf_nat_rdr_rules) *
2478 		      softn->ipf_nat_rdrrules_sz);
2479 	}
2480 
2481 	while ((n = softn->ipf_nat_list) != NULL) {
2482 		ipf_nat_delrule(softc, softn, n, 0);
2483 		i++;
2484 	}
2485 #if SOLARIS && !defined(INSTANCES)
2486 	pfil_delayed_copy = 1;
2487 #endif
2488 	return i;
2489 }
2490 
2491 
2492 /* ------------------------------------------------------------------------ */
2493 /* Function:    ipf_nat_delrule                                             */
2494 /* Returns:     Nil                                                         */
2495 /* Parameters:  softc(I) - pointer to soft context main structure           */
2496 /*              softn(I) - pointer to NAT context structure                 */
2497 /*              np(I)    - pointer to NAT rule to delete                    */
2498 /*              purge(I) - 1 == allow purge, 0 == prevent purge             */
2499 /* Locks:       WRITE(ipf_nat)                                              */
2500 /*                                                                          */
2501 /* Preventing "purge" from occuring is allowed because when all of the NAT  */
2502 /* rules are being removed, allowing the "purge" to walk through the list   */
2503 /* of NAT sessions, possibly multiple times, would be a large performance   */
2504 /* hit, on the order of O(N^2).                                             */
2505 /* ------------------------------------------------------------------------ */
2506 static void
2507 ipf_nat_delrule(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *np,
2508 	int purge)
2509 {
2510 
2511 	if (np->in_pnext != NULL) {
2512 		*np->in_pnext = np->in_next;
2513 		if (np->in_next != NULL)
2514 			np->in_next->in_pnext = np->in_pnext;
2515 		if (softn->ipf_nat_list_tail == &np->in_next)
2516 			softn->ipf_nat_list_tail = np->in_pnext;
2517 	}
2518 
2519 	if ((purge == 1) && ((np->in_flags & IPN_PURGE) != 0)) {
2520 		nat_t *next;
2521 		nat_t *nat;
2522 
2523 		for (next = softn->ipf_nat_instances; (nat = next) != NULL;) {
2524 			next = nat->nat_next;
2525 			if (nat->nat_ptr == np)
2526 				ipf_nat_delete(softc, nat, NL_PURGE);
2527 		}
2528 	}
2529 
2530 	if ((np->in_flags & IPN_DELETE) == 0) {
2531 		if (np->in_redir & NAT_REDIRECT) {
2532 			switch (np->in_v[0])
2533 			{
2534 			case 4 :
2535 				ipf_nat_delrdr(softn, np);
2536 				break;
2537 #ifdef USE_INET6
2538 			case 6 :
2539 				ipf_nat6_delrdr(softn, np);
2540 				break;
2541 #endif
2542 			}
2543 		}
2544 		if (np->in_redir & (NAT_MAPBLK|NAT_MAP)) {
2545 			switch (np->in_v[0])
2546 			{
2547 			case 4 :
2548 				ipf_nat_delmap(softn, np);
2549 				break;
2550 #ifdef USE_INET6
2551 			case 6 :
2552 				ipf_nat6_delmap(softn, np);
2553 				break;
2554 #endif
2555 			}
2556 		}
2557 	}
2558 
2559 	np->in_flags |= IPN_DELETE;
2560 	ipf_nat_rule_deref(softc, &np);
2561 }
2562 
2563 
2564 /* ------------------------------------------------------------------------ */
2565 /* Function:    ipf_nat_newmap                                              */
2566 /* Returns:     int - -1 == error, 0 == success                             */
2567 /* Parameters:  fin(I) - pointer to packet information                      */
2568 /*              nat(I) - pointer to NAT entry                               */
2569 /*              ni(I)  - pointer to structure with misc. information needed */
2570 /*                       to create new NAT entry.                           */
2571 /*                                                                          */
2572 /* Given an empty NAT structure, populate it with new information about a   */
2573 /* new NAT session, as defined by the matching NAT rule.                    */
2574 /* ni.nai_ip is passed in uninitialised and must be set, in host byte order,*/
2575 /* to the new IP address for the translation.                               */
2576 /* ------------------------------------------------------------------------ */
2577 static int
2578 ipf_nat_newmap(fr_info_t *fin, nat_t *nat, natinfo_t *ni)
2579 {
2580 	ipf_main_softc_t *softc = fin->fin_main_soft;
2581 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2582 	u_short st_port, dport, sport, port, sp, dp;
2583 	struct in_addr in, inb;
2584 	hostmap_t *hm;
2585 	u_32_t flags;
2586 	u_32_t st_ip;
2587 	ipnat_t *np;
2588 	nat_t *natl;
2589 	int l;
2590 
2591 	/*
2592 	 * If it's an outbound packet which doesn't match any existing
2593 	 * record, then create a new port
2594 	 */
2595 	l = 0;
2596 	hm = NULL;
2597 	np = ni->nai_np;
2598 	st_ip = np->in_snip;
2599 	st_port = np->in_spnext;
2600 	flags = nat->nat_flags;
2601 
2602 	if (flags & IPN_ICMPQUERY) {
2603 		sport = fin->fin_data[1];
2604 		dport = 0;
2605 	} else {
2606 		sport = htons(fin->fin_data[0]);
2607 		dport = htons(fin->fin_data[1]);
2608 	}
2609 
2610 	/*
2611 	 * Do a loop until we either run out of entries to try or we find
2612 	 * a NAT mapping that isn't currently being used.  This is done
2613 	 * because the change to the source is not (usually) being fixed.
2614 	 */
2615 	do {
2616 		port = 0;
2617 		in.s_addr = htonl(np->in_snip);
2618 		if (l == 0) {
2619 			/*
2620 			 * Check to see if there is an existing NAT
2621 			 * setup for this IP address pair.
2622 			 */
2623 			hm = ipf_nat_hostmap(softn, np, fin->fin_src,
2624 					     fin->fin_dst, in, 0);
2625 			if (hm != NULL)
2626 				in.s_addr = hm->hm_nsrcip.s_addr;
2627 		} else if ((l == 1) && (hm != NULL)) {
2628 			ipf_nat_hostmapdel(softc, &hm);
2629 		}
2630 		in.s_addr = ntohl(in.s_addr);
2631 
2632 		nat->nat_hm = hm;
2633 
2634 		if ((np->in_nsrcmsk == 0xffffffff) && (np->in_spnext == 0)) {
2635 			if (l > 0) {
2636 				NBUMPSIDEX(1, ns_exhausted, ns_exhausted_1);
2637 				return -1;
2638 			}
2639 		}
2640 
2641 		if (np->in_redir == NAT_BIMAP &&
2642 		    np->in_osrcmsk == np->in_nsrcmsk) {
2643 			/*
2644 			 * map the address block in a 1:1 fashion
2645 			 */
2646 			in.s_addr = np->in_nsrcaddr;
2647 			in.s_addr |= fin->fin_saddr & ~np->in_osrcmsk;
2648 			in.s_addr = ntohl(in.s_addr);
2649 
2650 		} else if (np->in_redir & NAT_MAPBLK) {
2651 			if ((l >= np->in_ppip) || ((l > 0) &&
2652 			     !(flags & IPN_TCPUDP))) {
2653 				NBUMPSIDEX(1, ns_exhausted, ns_exhausted_2);
2654 				return -1;
2655 			}
2656 			/*
2657 			 * map-block - Calculate destination address.
2658 			 */
2659 			in.s_addr = ntohl(fin->fin_saddr);
2660 			in.s_addr &= ntohl(~np->in_osrcmsk);
2661 			inb.s_addr = in.s_addr;
2662 			in.s_addr /= np->in_ippip;
2663 			in.s_addr &= ntohl(~np->in_nsrcmsk);
2664 			in.s_addr += ntohl(np->in_nsrcaddr);
2665 			/*
2666 			 * Calculate destination port.
2667 			 */
2668 			if ((flags & IPN_TCPUDP) &&
2669 			    (np->in_ppip != 0)) {
2670 				port = ntohs(sport) + l;
2671 				port %= np->in_ppip;
2672 				port += np->in_ppip *
2673 					(inb.s_addr % np->in_ippip);
2674 				port += MAPBLK_MINPORT;
2675 				port = htons(port);
2676 			}
2677 
2678 		} else if ((np->in_nsrcaddr == 0) &&
2679 			   (np->in_nsrcmsk == 0xffffffff)) {
2680 			i6addr_t in6;
2681 
2682 			/*
2683 			 * 0/32 - use the interface's IP address.
2684 			 */
2685 			if ((l > 0) ||
2686 			    ipf_ifpaddr(softc, 4, FRI_NORMAL, fin->fin_ifp,
2687 				       &in6, NULL) == -1) {
2688 				NBUMPSIDEX(1, ns_new_ifpaddr, ns_new_ifpaddr_1);
2689 				return -1;
2690 			}
2691 			in.s_addr = ntohl(in6.in4.s_addr);
2692 
2693 		} else if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0)) {
2694 			/*
2695 			 * 0/0 - use the original source address/port.
2696 			 */
2697 			if (l > 0) {
2698 				NBUMPSIDEX(1, ns_exhausted, ns_exhausted_3);
2699 				return -1;
2700 			}
2701 			in.s_addr = ntohl(fin->fin_saddr);
2702 
2703 		} else if ((np->in_nsrcmsk != 0xffffffff) &&
2704 			   (np->in_spnext == 0) && ((l > 0) || (hm == NULL)))
2705 			np->in_snip++;
2706 
2707 		natl = NULL;
2708 
2709 		if ((flags & IPN_TCPUDP) &&
2710 		    ((np->in_redir & NAT_MAPBLK) == 0) &&
2711 		    (np->in_flags & IPN_AUTOPORTMAP)) {
2712 			/*
2713 			 * "ports auto" (without map-block)
2714 			 */
2715 			if ((l > 0) && np->in_ppip && (l % np->in_ppip == 0)) {
2716 				if (l > np->in_space) {
2717 					return -1;
2718 				} else if ((l > np->in_ppip) &&
2719 					   np->in_nsrcmsk != 0xffffffff)
2720 					np->in_snip++;
2721 			}
2722 			if (np->in_ppip != 0) {
2723 				port = ntohs(sport);
2724 				port += (l % np->in_ppip);
2725 				port %= np->in_ppip;
2726 				port += np->in_ppip *
2727 					(ntohl(fin->fin_saddr) %
2728 					 np->in_ippip);
2729 				port += MAPBLK_MINPORT;
2730 				port = htons(port);
2731 			}
2732 
2733 		} else if (((np->in_redir & NAT_MAPBLK) == 0) &&
2734 			   (flags & IPN_TCPUDPICMP) && (np->in_spnext != 0)) {
2735 			/*
2736 			 * Standard port translation.  Select next port.
2737 			 */
2738 			if (np->in_flags & IPN_SEQUENTIAL) {
2739 				port = np->in_spnext;
2740 			} else {
2741 				port = ipf_random() % (np->in_spmax -
2742 						       np->in_spmin + 1);
2743 				port += np->in_spmin;
2744 			}
2745 			port = htons(port);
2746 			np->in_spnext++;
2747 
2748 			if (np->in_spnext > np->in_spmax) {
2749 				np->in_spnext = np->in_spmin;
2750 				if (np->in_nsrcmsk != 0xffffffff)
2751 					np->in_snip++;
2752 			}
2753 		}
2754 
2755 		if (np->in_flags & IPN_SIPRANGE) {
2756 			if (np->in_snip > ntohl(np->in_nsrcmsk))
2757 				np->in_snip = ntohl(np->in_nsrcaddr);
2758 		} else {
2759 			if ((np->in_nsrcmsk != 0xffffffff) &&
2760 			    ((np->in_snip + 1) & ntohl(np->in_nsrcmsk)) >
2761 			    ntohl(np->in_nsrcaddr))
2762 				np->in_snip = ntohl(np->in_nsrcaddr) + 1;
2763 		}
2764 
2765 		if ((port == 0) && (flags & (IPN_TCPUDPICMP|IPN_ICMPQUERY)))
2766 			port = sport;
2767 
2768 		/*
2769 		 * Here we do a lookup of the connection as seen from
2770 		 * the outside.  If an IP# pair already exists, try
2771 		 * again.  So if you have A->B becomes C->B, you can
2772 		 * also have D->E become C->E but not D->B causing
2773 		 * another C->B.  Also take protocol and ports into
2774 		 * account when determining whether a pre-existing
2775 		 * NAT setup will cause an external conflict where
2776 		 * this is appropriate.
2777 		 */
2778 		inb.s_addr = htonl(in.s_addr);
2779 		sp = fin->fin_data[0];
2780 		dp = fin->fin_data[1];
2781 		fin->fin_data[0] = fin->fin_data[1];
2782 		fin->fin_data[1] = ntohs(port);
2783 		natl = ipf_nat_inlookup(fin, flags & ~(SI_WILDP|NAT_SEARCH),
2784 					(u_int)fin->fin_p, fin->fin_dst, inb);
2785 		fin->fin_data[0] = sp;
2786 		fin->fin_data[1] = dp;
2787 
2788 		/*
2789 		 * Has the search wrapped around and come back to the
2790 		 * start ?
2791 		 */
2792 		if ((natl != NULL) &&
2793 		    (np->in_spnext != 0) && (st_port == np->in_spnext) &&
2794 		    (np->in_snip != 0) && (st_ip == np->in_snip)) {
2795 			NBUMPSIDED(1, ns_wrap);
2796 			return -1;
2797 		}
2798 		l++;
2799 	} while (natl != NULL);
2800 
2801 	/* Setup the NAT table */
2802 	nat->nat_osrcip = fin->fin_src;
2803 	nat->nat_nsrcaddr = htonl(in.s_addr);
2804 	nat->nat_odstip = fin->fin_dst;
2805 	nat->nat_ndstip = fin->fin_dst;
2806 	if (nat->nat_hm == NULL)
2807 		nat->nat_hm = ipf_nat_hostmap(softn, np, fin->fin_src,
2808 					      fin->fin_dst, nat->nat_nsrcip,
2809 					      0);
2810 
2811 	if (flags & IPN_TCPUDP) {
2812 		nat->nat_osport = sport;
2813 		nat->nat_nsport = port;	/* sport */
2814 		nat->nat_odport = dport;
2815 		nat->nat_ndport = dport;
2816 		((tcphdr_t *)fin->fin_dp)->th_sport = port;
2817 	} else if (flags & IPN_ICMPQUERY) {
2818 		nat->nat_oicmpid = fin->fin_data[1];
2819 		((icmphdr_t *)fin->fin_dp)->icmp_id = port;
2820 		nat->nat_nicmpid = port;
2821 	}
2822 	return 0;
2823 }
2824 
2825 
2826 /* ------------------------------------------------------------------------ */
2827 /* Function:    ipf_nat_newrdr                                              */
2828 /* Returns:     int - -1 == error, 0 == success (no move), 1 == success and */
2829 /*                    allow rule to be moved if IPN_ROUNDR is set.          */
2830 /* Parameters:  fin(I) - pointer to packet information                      */
2831 /*              nat(I) - pointer to NAT entry                               */
2832 /*              ni(I)  - pointer to structure with misc. information needed */
2833 /*                       to create new NAT entry.                           */
2834 /*                                                                          */
2835 /* ni.nai_ip is passed in uninitialised and must be set, in host byte order,*/
2836 /* to the new IP address for the translation.                               */
2837 /* ------------------------------------------------------------------------ */
2838 static int
2839 ipf_nat_newrdr(fr_info_t *fin, nat_t *nat, natinfo_t *ni)
2840 {
2841 	ipf_main_softc_t *softc = fin->fin_main_soft;
2842 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2843 	u_short nport, dport, sport;
2844 	struct in_addr in, inb;
2845 	u_short sp, dp;
2846 	hostmap_t *hm;
2847 	u_32_t flags;
2848 	ipnat_t *np;
2849 	nat_t *natl;
2850 	int move;
2851 
2852 	move = 1;
2853 	hm = NULL;
2854 	in.s_addr = 0;
2855 	np = ni->nai_np;
2856 	flags = nat->nat_flags;
2857 
2858 	if (flags & IPN_ICMPQUERY) {
2859 		dport = fin->fin_data[1];
2860 		sport = 0;
2861 	} else {
2862 		sport = htons(fin->fin_data[0]);
2863 		dport = htons(fin->fin_data[1]);
2864 	}
2865 
2866 	/* TRACE sport, dport */
2867 
2868 
2869 	/*
2870 	 * If the matching rule has IPN_STICKY set, then we want to have the
2871 	 * same rule kick in as before.  Why would this happen?  If you have
2872 	 * a collection of rdr rules with "round-robin sticky", the current
2873 	 * packet might match a different one to the previous connection but
2874 	 * we want the same destination to be used.
2875 	 */
2876 	if (((np->in_flags & (IPN_ROUNDR|IPN_SPLIT)) != 0) &&
2877 	    ((np->in_flags & IPN_STICKY) != 0)) {
2878 		hm = ipf_nat_hostmap(softn, NULL, fin->fin_src, fin->fin_dst,
2879 				     in, (u_32_t)dport);
2880 		if (hm != NULL) {
2881 			in.s_addr = ntohl(hm->hm_ndstip.s_addr);
2882 			np = hm->hm_ipnat;
2883 			ni->nai_np = np;
2884 			move = 0;
2885 			ipf_nat_hostmapdel(softc, &hm);
2886 		}
2887 	}
2888 
2889 	/*
2890 	 * Otherwise, it's an inbound packet. Most likely, we don't
2891 	 * want to rewrite source ports and source addresses. Instead,
2892 	 * we want to rewrite to a fixed internal address and fixed
2893 	 * internal port.
2894 	 */
2895 	if (np->in_flags & IPN_SPLIT) {
2896 		in.s_addr = np->in_dnip;
2897 		inb.s_addr = htonl(in.s_addr);
2898 
2899 		if ((np->in_flags & (IPN_ROUNDR|IPN_STICKY)) == IPN_STICKY) {
2900 			hm = ipf_nat_hostmap(softn, NULL, fin->fin_src,
2901 					     fin->fin_dst, inb, (u_32_t)dport);
2902 			if (hm != NULL) {
2903 				in.s_addr = hm->hm_ndstip.s_addr;
2904 				move = 0;
2905 			}
2906 		}
2907 
2908 		if (hm == NULL || hm->hm_ref == 1) {
2909 			if (np->in_ndstaddr == htonl(in.s_addr)) {
2910 				np->in_dnip = ntohl(np->in_ndstmsk);
2911 				move = 0;
2912 			} else {
2913 				np->in_dnip = ntohl(np->in_ndstaddr);
2914 			}
2915 		}
2916 		if (hm != NULL)
2917 			ipf_nat_hostmapdel(softc, &hm);
2918 
2919 	} else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0xffffffff)) {
2920 		i6addr_t in6;
2921 
2922 		/*
2923 		 * 0/32 - use the interface's IP address.
2924 		 */
2925 		if (ipf_ifpaddr(softc, 4, FRI_NORMAL, fin->fin_ifp,
2926 			       &in6, NULL) == -1) {
2927 			NBUMPSIDEX(0, ns_new_ifpaddr, ns_new_ifpaddr_2);
2928 			return -1;
2929 		}
2930 		in.s_addr = ntohl(in6.in4.s_addr);
2931 
2932 	} else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk== 0)) {
2933 		/*
2934 		 * 0/0 - use the original destination address/port.
2935 		 */
2936 		in.s_addr = ntohl(fin->fin_daddr);
2937 
2938 	} else if (np->in_redir == NAT_BIMAP &&
2939 		   np->in_ndstmsk == np->in_odstmsk) {
2940 		/*
2941 		 * map the address block in a 1:1 fashion
2942 		 */
2943 		in.s_addr = np->in_ndstaddr;
2944 		in.s_addr |= fin->fin_daddr & ~np->in_ndstmsk;
2945 		in.s_addr = ntohl(in.s_addr);
2946 	} else {
2947 		in.s_addr = ntohl(np->in_ndstaddr);
2948 	}
2949 
2950 	if ((np->in_dpnext == 0) || ((flags & NAT_NOTRULEPORT) != 0))
2951 		nport = dport;
2952 	else {
2953 		/*
2954 		 * Whilst not optimized for the case where
2955 		 * pmin == pmax, the gain is not significant.
2956 		 */
2957 		if (((np->in_flags & IPN_FIXEDDPORT) == 0) &&
2958 		    (np->in_odport != np->in_dtop)) {
2959 			nport = ntohs(dport) - np->in_odport + np->in_dpmax;
2960 			nport = htons(nport);
2961 		} else {
2962 			nport = htons(np->in_dpnext);
2963 			np->in_dpnext++;
2964 			if (np->in_dpnext > np->in_dpmax)
2965 				np->in_dpnext = np->in_dpmin;
2966 		}
2967 	}
2968 
2969 	/*
2970 	 * When the redirect-to address is set to 0.0.0.0, just
2971 	 * assume a blank `forwarding' of the packet.  We don't
2972 	 * setup any translation for this either.
2973 	 */
2974 	if (in.s_addr == 0) {
2975 		if (nport == dport) {
2976 			NBUMPSIDED(0, ns_xlate_null);
2977 			return -1;
2978 		}
2979 		in.s_addr = ntohl(fin->fin_daddr);
2980 	}
2981 
2982 	/*
2983 	 * Check to see if this redirect mapping already exists and if
2984 	 * it does, return "failure" (allowing it to be created will just
2985 	 * cause one or both of these "connections" to stop working.)
2986 	 */
2987 	inb.s_addr = htonl(in.s_addr);
2988 	sp = fin->fin_data[0];
2989 	dp = fin->fin_data[1];
2990 	fin->fin_data[1] = fin->fin_data[0];
2991 	fin->fin_data[0] = ntohs(nport);
2992 	natl = ipf_nat_outlookup(fin, flags & ~(SI_WILDP|NAT_SEARCH),
2993 			     (u_int)fin->fin_p, inb, fin->fin_src);
2994 	fin->fin_data[0] = sp;
2995 	fin->fin_data[1] = dp;
2996 	if (natl != NULL) {
2997 		DT2(ns_new_xlate_exists, fr_info_t *, fin, nat_t *, natl);
2998 		NBUMPSIDE(0, ns_xlate_exists);
2999 		return -1;
3000 	}
3001 
3002 	nat->nat_ndstaddr = htonl(in.s_addr);
3003 	nat->nat_odstip = fin->fin_dst;
3004 	nat->nat_nsrcip = fin->fin_src;
3005 	nat->nat_osrcip = fin->fin_src;
3006 	if ((nat->nat_hm == NULL) && ((np->in_flags & IPN_STICKY) != 0))
3007 		nat->nat_hm = ipf_nat_hostmap(softn, np, fin->fin_src,
3008 					      fin->fin_dst, inb, (u_32_t)dport);
3009 
3010 	if (flags & IPN_TCPUDP) {
3011 		nat->nat_odport = dport;
3012 		nat->nat_ndport = nport;
3013 		nat->nat_osport = sport;
3014 		nat->nat_nsport = sport;
3015 		((tcphdr_t *)fin->fin_dp)->th_dport = nport;
3016 	} else if (flags & IPN_ICMPQUERY) {
3017 		nat->nat_oicmpid = fin->fin_data[1];
3018 		((icmphdr_t *)fin->fin_dp)->icmp_id = nport;
3019 		nat->nat_nicmpid = nport;
3020 	}
3021 
3022 	return move;
3023 }
3024 
3025 /* ------------------------------------------------------------------------ */
3026 /* Function:    ipf_nat_add                                                 */
3027 /* Returns:     nat_t* - NULL == failure to create new NAT structure,       */
3028 /*                       else pointer to new NAT structure                  */
3029 /* Parameters:  fin(I)       - pointer to packet information                */
3030 /*              np(I)        - pointer to NAT rule                          */
3031 /*              natsave(I)   - pointer to where to store NAT struct pointer */
3032 /*              flags(I)     - flags describing the current packet          */
3033 /*              direction(I) - direction of packet (in/out)                 */
3034 /* Write Lock:  ipf_nat                                                     */
3035 /*                                                                          */
3036 /* Attempts to create a new NAT entry.  Does not actually change the packet */
3037 /* in any way.                                                              */
3038 /*                                                                          */
3039 /* This function is in three main parts: (1) deal with creating a new NAT   */
3040 /* structure for a "MAP" rule (outgoing NAT translation); (2) deal with     */
3041 /* creating a new NAT structure for a "RDR" rule (incoming NAT translation) */
3042 /* and (3) building that structure and putting it into the NAT table(s).    */
3043 /*                                                                          */
3044 /* NOTE: natsave should NOT be used to point back to an ipstate_t struct    */
3045 /*       as it can result in memory being corrupted.                        */
3046 /* ------------------------------------------------------------------------ */
3047 nat_t *
3048 ipf_nat_add(fr_info_t *fin, ipnat_t *np, nat_t **natsave, u_int flags,
3049     int direction)
3050 {
3051 	ipf_main_softc_t *softc = fin->fin_main_soft;
3052 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3053 	hostmap_t *hm = NULL;
3054 	nat_t *nat, *natl;
3055 	natstat_t *nsp;
3056 	u_int nflags;
3057 	natinfo_t ni;
3058 	int move;
3059 
3060 	memset(&ni, 0, sizeof ni);	/* XXX gcc */
3061 	nsp = &softn->ipf_nat_stats;
3062 
3063 	if ((nsp->ns_active * 100 / softn->ipf_nat_table_max) >
3064 	    softn->ipf_nat_table_wm_high) {
3065 		softn->ipf_nat_doflush = 1;
3066 	}
3067 
3068 	if (nsp->ns_active >= softn->ipf_nat_table_max) {
3069 		NBUMPSIDED(fin->fin_out, ns_table_max);
3070 		return NULL;
3071 	}
3072 
3073 	move = 1;
3074 	nflags = np->in_flags & flags;
3075 	nflags &= NAT_FROMRULE;
3076 
3077 	ni.nai_np = np;
3078 	ni.nai_dport = 0;
3079 	ni.nai_sport = 0;
3080 
3081 	/* Give me a new nat */
3082 	KMALLOC(nat, nat_t *);
3083 	if (nat == NULL) {
3084 		NBUMPSIDED(fin->fin_out, ns_memfail);
3085 		/*
3086 		 * Try to automatically tune the max # of entries in the
3087 		 * table allowed to be less than what will cause kmem_alloc()
3088 		 * to fail and try to eliminate panics due to out of memory
3089 		 * conditions arising.
3090 		 */
3091 		if ((softn->ipf_nat_table_max > softn->ipf_nat_table_sz) &&
3092 		    (nsp->ns_active > 100)) {
3093 			softn->ipf_nat_table_max = nsp->ns_active - 100;
3094 			printf("table_max reduced to %d\n",
3095 				softn->ipf_nat_table_max);
3096 		}
3097 		return NULL;
3098 	}
3099 
3100 	if (flags & IPN_ICMPQUERY) {
3101 		/*
3102 		 * In the ICMP query NAT code, we translate the ICMP id fields
3103 		 * to make them unique. This is indepedent of the ICMP type
3104 		 * (e.g. in the unlikely event that a host sends an echo and
3105 		 * an tstamp request with the same id, both packets will have
3106 		 * their ip address/id field changed in the same way).
3107 		 */
3108 		/* The icmp_id field is used by the sender to identify the
3109 		 * process making the icmp request. (the receiver justs
3110 		 * copies it back in its response). So, it closely matches
3111 		 * the concept of source port. We overlay sport, so we can
3112 		 * maximally reuse the existing code.
3113 		 */
3114 		ni.nai_sport = fin->fin_data[1];
3115 		ni.nai_dport = 0;
3116 	}
3117 
3118 	bzero((char *)nat, sizeof(*nat));
3119 	nat->nat_flags = flags;
3120 	nat->nat_redir = np->in_redir;
3121 	nat->nat_dir = direction;
3122 	nat->nat_pr[0] = fin->fin_p;
3123 	nat->nat_pr[1] = fin->fin_p;
3124 
3125 	/*
3126 	 * Search the current table for a match and create a new mapping
3127 	 * if there is none found.
3128 	 */
3129 	if (np->in_redir & NAT_DIVERTUDP) {
3130 		move = ipf_nat_newdivert(fin, nat, &ni);
3131 
3132 	} else if (np->in_redir & NAT_REWRITE) {
3133 		move = ipf_nat_newrewrite(fin, nat, &ni);
3134 
3135 	} else if (direction == NAT_OUTBOUND) {
3136 		/*
3137 		 * We can now arrange to call this for the same connection
3138 		 * because ipf_nat_new doesn't protect the code path into
3139 		 * this function.
3140 		 */
3141 		natl = ipf_nat_outlookup(fin, nflags, (u_int)fin->fin_p,
3142 				     fin->fin_src, fin->fin_dst);
3143 		if (natl != NULL) {
3144 			KFREE(nat);
3145 			nat = natl;
3146 			goto done;
3147 		}
3148 
3149 		move = ipf_nat_newmap(fin, nat, &ni);
3150 	} else {
3151 		/*
3152 		 * NAT_INBOUND is used for redirects rules
3153 		 */
3154 		natl = ipf_nat_inlookup(fin, nflags, (u_int)fin->fin_p,
3155 					fin->fin_src, fin->fin_dst);
3156 		if (natl != NULL) {
3157 			KFREE(nat);
3158 			nat = natl;
3159 			goto done;
3160 		}
3161 
3162 		move = ipf_nat_newrdr(fin, nat, &ni);
3163 	}
3164 	if (move == -1)
3165 		goto badnat;
3166 
3167 	np = ni.nai_np;
3168 
3169 	nat->nat_mssclamp = np->in_mssclamp;
3170 	nat->nat_me = natsave;
3171 	nat->nat_fr = fin->fin_fr;
3172 	nat->nat_rev = fin->fin_rev;
3173 	nat->nat_ptr = np;
3174 	nat->nat_dlocal = np->in_dlocal;
3175 
3176 	if ((np->in_apr != NULL) && ((nat->nat_flags & NAT_SLAVE) == 0)) {
3177 		if (ipf_proxy_new(fin, nat) == -1) {
3178 			NBUMPSIDED(fin->fin_out, ns_appr_fail);
3179 			goto badnat;
3180 		}
3181 	}
3182 
3183 	nat->nat_ifps[0] = np->in_ifps[0];
3184 	if (np->in_ifps[0] != NULL) {
3185 		COPYIFNAME(np->in_v[0], np->in_ifps[0], nat->nat_ifnames[0]);
3186 	}
3187 
3188 	nat->nat_ifps[1] = np->in_ifps[1];
3189 	if (np->in_ifps[1] != NULL) {
3190 		COPYIFNAME(np->in_v[1], np->in_ifps[1], nat->nat_ifnames[1]);
3191 	}
3192 
3193 	if (ipf_nat_finalise(fin, nat) == -1) {
3194 		goto badnat;
3195 	}
3196 
3197 	np->in_use++;
3198 
3199 	if ((move == 1) && (np->in_flags & IPN_ROUNDR)) {
3200 		if ((np->in_redir & (NAT_REDIRECT|NAT_MAP)) == NAT_REDIRECT) {
3201 			ipf_nat_delrdr(softn, np);
3202 			ipf_nat_addrdr(softn, np);
3203 		} else if ((np->in_redir & (NAT_REDIRECT|NAT_MAP)) == NAT_MAP) {
3204 			ipf_nat_delmap(softn, np);
3205 			ipf_nat_addmap(softn, np);
3206 		}
3207 	}
3208 
3209 	if (flags & SI_WILDP)
3210 		nsp->ns_wilds++;
3211 	nsp->ns_proto[nat->nat_pr[0]]++;
3212 
3213 	goto done;
3214 badnat:
3215 	DT2(ns_badnatnew, fr_info_t *, fin, nat_t *, nat);
3216 	NBUMPSIDE(fin->fin_out, ns_badnatnew);
3217 	if ((hm = nat->nat_hm) != NULL)
3218 		ipf_nat_hostmapdel(softc, &hm);
3219 	KFREE(nat);
3220 	nat = NULL;
3221 done:
3222 	if (nat != NULL && np != NULL)
3223 		np->in_hits++;
3224 	if (natsave != NULL)
3225 		*natsave = nat;
3226 	return nat;
3227 }
3228 
3229 
3230 /* ------------------------------------------------------------------------ */
3231 /* Function:    ipf_nat_finalise                                            */
3232 /* Returns:     int - 0 == sucess, -1 == failure                            */
3233 /* Parameters:  fin(I) - pointer to packet information                      */
3234 /*              nat(I) - pointer to NAT entry                               */
3235 /* Write Lock:  ipf_nat                                                     */
3236 /*                                                                          */
3237 /* This is the tail end of constructing a new NAT entry and is the same     */
3238 /* for both IPv4 and IPv6.                                                  */
3239 /* ------------------------------------------------------------------------ */
3240 /*ARGSUSED*/
3241 static int
3242 ipf_nat_finalise(fr_info_t *fin, nat_t *nat)
3243 {
3244 	ipf_main_softc_t *softc = fin->fin_main_soft;
3245 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3246 	u_32_t sum1, sum2, sumd;
3247 	frentry_t *fr;
3248 #if SOLARIS && defined(_KERNEL) && (SOLARIS2 >= 6) && defined(ICK_M_CTL_MAGIC)
3249 	qpktinfo_t *qpi = fin->fin_qpi;
3250 #endif
3251 
3252 	switch (nat->nat_pr[0])
3253 	{
3254 	case IPPROTO_ICMP :
3255 		sum1 = LONG_SUM(ntohs(nat->nat_oicmpid));
3256 		sum2 = LONG_SUM(ntohs(nat->nat_nicmpid));
3257 		CALC_SUMD(sum1, sum2, sumd);
3258 		nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
3259 
3260 		break;
3261 
3262 	default :
3263 		sum1 = LONG_SUM(ntohl(nat->nat_osrcaddr) + \
3264 				ntohs(nat->nat_osport));
3265 		sum2 = LONG_SUM(ntohl(nat->nat_nsrcaddr) + \
3266 				ntohs(nat->nat_nsport));
3267 		CALC_SUMD(sum1, sum2, sumd);
3268 		nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
3269 
3270 		sum1 = LONG_SUM(ntohl(nat->nat_odstaddr) + \
3271 				ntohs(nat->nat_odport));
3272 		sum2 = LONG_SUM(ntohl(nat->nat_ndstaddr) + \
3273 				ntohs(nat->nat_ndport));
3274 		CALC_SUMD(sum1, sum2, sumd);
3275 		nat->nat_sumd[0] += (sumd & 0xffff) + (sumd >> 16);
3276 		break;
3277 	}
3278 
3279 	/*
3280 	 * Compute the partial checksum, just in case.
3281 	 * This is only ever placed into outbound packets so care needs
3282 	 * to be taken over which pair of addresses are used.
3283 	 */
3284 	if (nat->nat_dir == NAT_OUTBOUND) {
3285 		sum1 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
3286 		sum1 += LONG_SUM(ntohl(nat->nat_ndstaddr));
3287 	} else {
3288 		sum1 = LONG_SUM(ntohl(nat->nat_osrcaddr));
3289 		sum1 += LONG_SUM(ntohl(nat->nat_odstaddr));
3290 	}
3291 	sum1 += nat->nat_pr[1];
3292 	nat->nat_sumd[1] = (sum1 & 0xffff) + (sum1 >> 16);
3293 
3294 	sum1 = LONG_SUM(ntohl(nat->nat_osrcaddr));
3295 	sum2 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
3296 	CALC_SUMD(sum1, sum2, sumd);
3297 	nat->nat_ipsumd = (sumd & 0xffff) + (sumd >> 16);
3298 
3299 	sum1 = LONG_SUM(ntohl(nat->nat_odstaddr));
3300 	sum2 = LONG_SUM(ntohl(nat->nat_ndstaddr));
3301 	CALC_SUMD(sum1, sum2, sumd);
3302 	nat->nat_ipsumd += (sumd & 0xffff) + (sumd >> 16);
3303 
3304 	nat->nat_v[0] = 4;
3305 	nat->nat_v[1] = 4;
3306 
3307 	if ((nat->nat_ifps[0] != NULL) && (nat->nat_ifps[0] != (void *)-1)) {
3308 		nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
3309 	}
3310 
3311 	if ((nat->nat_ifps[1] != NULL) && (nat->nat_ifps[1] != (void *)-1)) {
3312 		nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
3313 	}
3314 
3315 	if ((nat->nat_flags & SI_CLONE) == 0)
3316 		nat->nat_sync = ipf_sync_new(softc, SMC_NAT, fin, nat);
3317 
3318 	if (ipf_nat_insert(softc, softn, nat) == 0) {
3319 		if (softn->ipf_nat_logging)
3320 			ipf_nat_log(softc, softn, nat, NL_NEW);
3321 		fr = nat->nat_fr;
3322 		if (fr != NULL) {
3323 			MUTEX_ENTER(&fr->fr_lock);
3324 			fr->fr_ref++;
3325 			MUTEX_EXIT(&fr->fr_lock);
3326 		}
3327 		return 0;
3328 	}
3329 
3330 	NBUMPSIDED(fin->fin_out, ns_unfinalised);
3331 	/*
3332 	 * nat_insert failed, so cleanup time...
3333 	 */
3334 	if (nat->nat_sync != NULL)
3335 		ipf_sync_del_nat(softc->ipf_sync_soft, nat->nat_sync);
3336 	return -1;
3337 }
3338 
3339 
3340 /* ------------------------------------------------------------------------ */
3341 /* Function:    ipf_nat_insert                                              */
3342 /* Returns:     int - 0 == sucess, -1 == failure                            */
3343 /* Parameters:  softc(I) - pointer to soft context main structure           */
3344 /*              softn(I) - pointer to NAT context structure                 */
3345 /*              nat(I) - pointer to NAT structure                           */
3346 /* Write Lock:  ipf_nat                                                     */
3347 /*                                                                          */
3348 /* Insert a NAT entry into the hash tables for searching and add it to the  */
3349 /* list of active NAT entries.  Adjust global counters when complete.       */
3350 /* ------------------------------------------------------------------------ */
3351 int
3352 ipf_nat_insert(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, nat_t *nat)
3353 {
3354 	u_int hv0, hv1;
3355 	u_int sp, dp;
3356 	ipnat_t *in;
3357 	int ret;
3358 
3359 	/*
3360 	 * Try and return an error as early as possible, so calculate the hash
3361 	 * entry numbers first and then proceed.
3362 	 */
3363 	if ((nat->nat_flags & (SI_W_SPORT|SI_W_DPORT)) == 0) {
3364 		if ((nat->nat_flags & IPN_TCPUDP) != 0) {
3365 			sp = nat->nat_osport;
3366 			dp = nat->nat_odport;
3367 		} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
3368 			sp = 0;
3369 			dp = nat->nat_oicmpid;
3370 		} else {
3371 			sp = 0;
3372 			dp = 0;
3373 		}
3374 		hv0 = NAT_HASH_FN(nat->nat_osrcaddr, sp, 0xffffffff);
3375 		hv0 = NAT_HASH_FN(nat->nat_odstaddr, hv0 + dp, 0xffffffff);
3376 		/*
3377 		 * TRACE nat_osrcaddr, nat_osport, nat_odstaddr,
3378 		 * nat_odport, hv0
3379 		 */
3380 
3381 		if ((nat->nat_flags & IPN_TCPUDP) != 0) {
3382 			sp = nat->nat_nsport;
3383 			dp = nat->nat_ndport;
3384 		} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
3385 			sp = 0;
3386 			dp = nat->nat_nicmpid;
3387 		} else {
3388 			sp = 0;
3389 			dp = 0;
3390 		}
3391 		hv1 = NAT_HASH_FN(nat->nat_nsrcaddr, sp, 0xffffffff);
3392 		hv1 = NAT_HASH_FN(nat->nat_ndstaddr, hv1 + dp, 0xffffffff);
3393 		/*
3394 		 * TRACE nat_nsrcaddr, nat_nsport, nat_ndstaddr,
3395 		 * nat_ndport, hv1
3396 		 */
3397 	} else {
3398 		hv0 = NAT_HASH_FN(nat->nat_osrcaddr, 0, 0xffffffff);
3399 		hv0 = NAT_HASH_FN(nat->nat_odstaddr, hv0, 0xffffffff);
3400 		/* TRACE nat_osrcaddr, nat_odstaddr, hv0 */
3401 
3402 		hv1 = NAT_HASH_FN(nat->nat_nsrcaddr, 0, 0xffffffff);
3403 		hv1 = NAT_HASH_FN(nat->nat_ndstaddr, hv1, 0xffffffff);
3404 		/* TRACE nat_nsrcaddr, nat_ndstaddr, hv1 */
3405 	}
3406 
3407 	if ((nat->nat_dir & NAT_OUTBOUND) == NAT_OUTBOUND) {
3408 		nat->nat_hv[0] = hv0;
3409 		nat->nat_hv[1] = hv1;
3410 	} else {
3411 		nat->nat_hv[0] = hv1;
3412 		nat->nat_hv[1] = hv0;
3413 	}
3414 
3415 	MUTEX_INIT(&nat->nat_lock, "nat entry lock");
3416 
3417 	in = nat->nat_ptr;
3418 	nat->nat_ref = nat->nat_me ? 2 : 1;
3419 
3420 	nat->nat_ifnames[0][LIFNAMSIZ - 1] = '\0';
3421 	nat->nat_ifps[0] = ipf_resolvenic(softc, nat->nat_ifnames[0], 4);
3422 
3423 	if (nat->nat_ifnames[1][0] != '\0') {
3424 		nat->nat_ifnames[1][LIFNAMSIZ - 1] = '\0';
3425 		nat->nat_ifps[1] = ipf_resolvenic(softc,
3426 						  nat->nat_ifnames[1], 4);
3427 	} else if (in->in_ifnames[1] != -1) {
3428 		char *name;
3429 
3430 		name = in->in_names + in->in_ifnames[1];
3431 		if (name[1] != '\0' && name[0] != '-' && name[0] != '*') {
3432 			(void) strncpy(nat->nat_ifnames[1],
3433 				       nat->nat_ifnames[0], LIFNAMSIZ);
3434 			nat->nat_ifnames[1][LIFNAMSIZ - 1] = '\0';
3435 			nat->nat_ifps[1] = nat->nat_ifps[0];
3436 		}
3437 	}
3438 	if ((nat->nat_ifps[0] != NULL) && (nat->nat_ifps[0] != (void *)-1)) {
3439 		nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
3440 	}
3441 	if ((nat->nat_ifps[1] != NULL) && (nat->nat_ifps[1] != (void *)-1)) {
3442 		nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
3443 	}
3444 
3445 	ret = ipf_nat_hashtab_add(softc, softn, nat);
3446 	if (ret == -1)
3447 		MUTEX_DESTROY(&nat->nat_lock);
3448 	return ret;
3449 }
3450 
3451 
3452 /* ------------------------------------------------------------------------ */
3453 /* Function:    ipf_nat_hashtab_add                                         */
3454 /* Returns:     int - 0 == sucess, -1 == failure                            */
3455 /* Parameters:  softc(I) - pointer to soft context main structure           */
3456 /*              softn(I) - pointer to NAT context structure                 */
3457 /*              nat(I) - pointer to NAT structure                           */
3458 /* Write Lock:  ipf_nat                                                     */
3459 /*                                                                          */
3460 /* Handle the insertion of a NAT entry into the table/list.                 */
3461 /* ------------------------------------------------------------------------ */
3462 int
3463 ipf_nat_hashtab_add(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, nat_t *nat)
3464 {
3465 	nat_t **natp;
3466 	u_int hv0;
3467 	u_int hv1;
3468 
3469 	hv0 = nat->nat_hv[0] % softn->ipf_nat_table_sz;
3470 	hv1 = nat->nat_hv[1] % softn->ipf_nat_table_sz;
3471 
3472 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen[hv0] >=
3473 	    softn->ipf_nat_maxbucket) {
3474 		DT1(ns_bucket_max_0, int,
3475 		    softn->ipf_nat_stats.ns_side[0].ns_bucketlen[hv0]);
3476 		NBUMPSIDE(0, ns_bucket_max);
3477 		return -1;
3478 	}
3479 
3480 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen[hv1] >=
3481 	    softn->ipf_nat_maxbucket) {
3482 		DT1(ns_bucket_max_1, int,
3483 		    softn->ipf_nat_stats.ns_side[1].ns_bucketlen[hv1]);
3484 		NBUMPSIDE(1, ns_bucket_max);
3485 		return -1;
3486 	}
3487 
3488 	/*
3489 	 * The ordering of operations in the list and hash table insertion
3490 	 * is very important.  The last operation for each task should be
3491 	 * to update the top of the list, after all the "nexts" have been
3492 	 * done so that walking the list while it is being done does not
3493 	 * find strange pointers.
3494 	 *
3495 	 * Global list of NAT instances
3496 	 */
3497 	nat->nat_next = softn->ipf_nat_instances;
3498 	nat->nat_pnext = &softn->ipf_nat_instances;
3499 	if (softn->ipf_nat_instances)
3500 		softn->ipf_nat_instances->nat_pnext = &nat->nat_next;
3501 	softn->ipf_nat_instances = nat;
3502 
3503 	/*
3504 	 * Inbound hash table.
3505 	 */
3506 	natp = &softn->ipf_nat_table[0][hv0];
3507 	nat->nat_phnext[0] = natp;
3508 	nat->nat_hnext[0] = *natp;
3509 	if (*natp) {
3510 		(*natp)->nat_phnext[0] = &nat->nat_hnext[0];
3511 	} else {
3512 		NBUMPSIDE(0, ns_inuse);
3513 	}
3514 	*natp = nat;
3515 	NBUMPSIDE(0, ns_bucketlen[hv0]);
3516 
3517 	/*
3518 	 * Outbound hash table.
3519 	 */
3520 	natp = &softn->ipf_nat_table[1][hv1];
3521 	nat->nat_phnext[1] = natp;
3522 	nat->nat_hnext[1] = *natp;
3523 	if (*natp)
3524 		(*natp)->nat_phnext[1] = &nat->nat_hnext[1];
3525 	else {
3526 		NBUMPSIDE(1, ns_inuse);
3527 	}
3528 	*natp = nat;
3529 	NBUMPSIDE(1, ns_bucketlen[hv1]);
3530 
3531 	ipf_nat_setqueue(softc, softn, nat);
3532 
3533 	if (nat->nat_dir & NAT_OUTBOUND) {
3534 		NBUMPSIDE(1, ns_added);
3535 	} else {
3536 		NBUMPSIDE(0, ns_added);
3537 	}
3538 	softn->ipf_nat_stats.ns_active++;
3539 	return 0;
3540 }
3541 
3542 
3543 /* ------------------------------------------------------------------------ */
3544 /* Function:    ipf_nat_icmperrorlookup                                     */
3545 /* Returns:     nat_t* - point to matching NAT structure                    */
3546 /* Parameters:  fin(I) - pointer to packet information                      */
3547 /*              dir(I) - direction of packet (in/out)                       */
3548 /*                                                                          */
3549 /* Check if the ICMP error message is related to an existing TCP, UDP or    */
3550 /* ICMP query nat entry.  It is assumed that the packet is already of the   */
3551 /* the required length.                                                     */
3552 /* ------------------------------------------------------------------------ */
3553 nat_t *
3554 ipf_nat_icmperrorlookup(fr_info_t *fin, int dir)
3555 {
3556 	ipf_main_softc_t *softc = fin->fin_main_soft;
3557 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3558 	int flags = 0, minlen;
3559 	icmphdr_t *orgicmp;
3560 	nat_stat_side_t *nside;
3561 	tcphdr_t *tcp = NULL;
3562 	u_short data[2];
3563 	nat_t *nat;
3564 	ip_t *oip;
3565 	u_int p;
3566 
3567 	nside = &softn->ipf_nat_stats.ns_side[fin->fin_out];
3568 	/*
3569 	 * Does it at least have the return (basic) IP header ?
3570 	 * Only a basic IP header (no options) should be with an ICMP error
3571 	 * header.  Also, if it's not an error type, then return.
3572 	 */
3573 	if ((fin->fin_hlen != sizeof(ip_t)) || !(fin->fin_flx & FI_ICMPERR)) {
3574 		ATOMIC_INCL(nside->ns_icmp_basic);
3575 		return NULL;
3576 	}
3577 
3578 	/*
3579 	 * Check packet size
3580 	 */
3581 	oip = (ip_t *)((char *)fin->fin_dp + 8);
3582 	minlen = IP_HL(oip) << 2;
3583 	if ((minlen < sizeof(ip_t)) ||
3584 	    (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen)) {
3585 		ATOMIC_INCL(nside->ns_icmp_size);
3586 		return NULL;
3587 	}
3588 
3589 	/*
3590 	 * Is the buffer big enough for all of it ?  It's the size of the IP
3591 	 * header claimed in the encapsulated part which is of concern.  It
3592 	 * may be too big to be in this buffer but not so big that it's
3593 	 * outside the ICMP packet, leading to TCP deref's causing problems.
3594 	 * This is possible because we don't know how big oip_hl is when we
3595 	 * do the pullup early in ipf_check() and thus can't gaurantee it is
3596 	 * all here now.
3597 	 */
3598 #ifdef  ipf_nat_KERNEL
3599 	{
3600 	mb_t *m;
3601 
3602 	m = fin->fin_m;
3603 # if defined(MENTAT)
3604 	if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN >
3605 	    (char *)m->b_wptr) {
3606 		ATOMIC_INCL(nside->ns_icmp_mbuf);
3607 		return NULL;
3608 	}
3609 # else
3610 	if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN >
3611 	    (char *)fin->fin_ip + M_LEN(m)) {
3612 		ATOMIC_INCL(nside->ns_icmp_mbuf);
3613 		return NULL;
3614 	}
3615 # endif
3616 	}
3617 #endif
3618 
3619 	if (fin->fin_daddr != oip->ip_src.s_addr) {
3620 		ATOMIC_INCL(nside->ns_icmp_address);
3621 		return NULL;
3622 	}
3623 
3624 	p = oip->ip_p;
3625 	if (p == IPPROTO_TCP)
3626 		flags = IPN_TCP;
3627 	else if (p == IPPROTO_UDP)
3628 		flags = IPN_UDP;
3629 	else if (p == IPPROTO_ICMP) {
3630 		orgicmp = (icmphdr_t *)((char *)oip + (IP_HL(oip) << 2));
3631 
3632 		/* see if this is related to an ICMP query */
3633 		if (ipf_nat_icmpquerytype(orgicmp->icmp_type)) {
3634 			data[0] = fin->fin_data[0];
3635 			data[1] = fin->fin_data[1];
3636 			fin->fin_data[0] = 0;
3637 			fin->fin_data[1] = orgicmp->icmp_id;
3638 
3639 			flags = IPN_ICMPERR|IPN_ICMPQUERY;
3640 			/*
3641 			 * NOTE : dir refers to the direction of the original
3642 			 *        ip packet. By definition the icmp error
3643 			 *        message flows in the opposite direction.
3644 			 */
3645 			if (dir == NAT_INBOUND)
3646 				nat = ipf_nat_inlookup(fin, flags, p,
3647 						       oip->ip_dst,
3648 						       oip->ip_src);
3649 			else
3650 				nat = ipf_nat_outlookup(fin, flags, p,
3651 							oip->ip_dst,
3652 							oip->ip_src);
3653 			fin->fin_data[0] = data[0];
3654 			fin->fin_data[1] = data[1];
3655 			return nat;
3656 		}
3657 	}
3658 
3659 	if (flags & IPN_TCPUDP) {
3660 		minlen += 8;		/* + 64bits of data to get ports */
3661 		/* TRACE (fin,minlen) */
3662 		if (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen) {
3663 			ATOMIC_INCL(nside->ns_icmp_short);
3664 			return NULL;
3665 		}
3666 
3667 		data[0] = fin->fin_data[0];
3668 		data[1] = fin->fin_data[1];
3669 		tcp = (tcphdr_t *)((char *)oip + (IP_HL(oip) << 2));
3670 		fin->fin_data[0] = ntohs(tcp->th_dport);
3671 		fin->fin_data[1] = ntohs(tcp->th_sport);
3672 
3673 		if (dir == NAT_INBOUND) {
3674 			nat = ipf_nat_inlookup(fin, flags, p, oip->ip_dst,
3675 					       oip->ip_src);
3676 		} else {
3677 			nat = ipf_nat_outlookup(fin, flags, p, oip->ip_dst,
3678 					    oip->ip_src);
3679 		}
3680 		fin->fin_data[0] = data[0];
3681 		fin->fin_data[1] = data[1];
3682 		return nat;
3683 	}
3684 	if (dir == NAT_INBOUND)
3685 		nat = ipf_nat_inlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
3686 	else
3687 		nat = ipf_nat_outlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
3688 
3689 	return nat;
3690 }
3691 
3692 
3693 /* ------------------------------------------------------------------------ */
3694 /* Function:    ipf_nat_icmperror                                           */
3695 /* Returns:     nat_t* - point to matching NAT structure                    */
3696 /* Parameters:  fin(I)    - pointer to packet information                   */
3697 /*              nflags(I) - NAT flags for this packet                       */
3698 /*              dir(I)    - direction of packet (in/out)                    */
3699 /*                                                                          */
3700 /* Fix up an ICMP packet which is an error message for an existing NAT      */
3701 /* session.  This will correct both packet header data and checksums.       */
3702 /*                                                                          */
3703 /* This should *ONLY* be used for incoming ICMP error packets to make sure  */
3704 /* a NAT'd ICMP packet gets correctly recognised.                           */
3705 /* ------------------------------------------------------------------------ */
3706 nat_t *
3707 ipf_nat_icmperror(fr_info_t *fin, u_int *nflags, int dir)
3708 {
3709 	ipf_main_softc_t *softc = fin->fin_main_soft;
3710 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3711 	u_32_t sum1, sum2, sumd, sumd2;
3712 	struct in_addr a1, a2, a3, a4;
3713 	int flags, dlen, odst;
3714 	icmphdr_t *icmp;
3715 	u_short *csump;
3716 	tcphdr_t *tcp;
3717 	nat_t *nat;
3718 	ip_t *oip;
3719 	void *dp;
3720 
3721 	if ((fin->fin_flx & (FI_SHORT|FI_FRAGBODY))) {
3722 		NBUMPSIDED(fin->fin_out, ns_icmp_short);
3723 		return NULL;
3724 	}
3725 
3726 	/*
3727 	 * ipf_nat_icmperrorlookup() will return NULL for `defective' packets.
3728 	 */
3729 	if ((fin->fin_v != 4) || !(nat = ipf_nat_icmperrorlookup(fin, dir))) {
3730 		NBUMPSIDED(fin->fin_out, ns_icmp_notfound);
3731 		return NULL;
3732 	}
3733 
3734 	tcp = NULL;
3735 	csump = NULL;
3736 	flags = 0;
3737 	sumd2 = 0;
3738 	*nflags = IPN_ICMPERR;
3739 	icmp = fin->fin_dp;
3740 	oip = (ip_t *)&icmp->icmp_ip;
3741 	dp = (((char *)oip) + (IP_HL(oip) << 2));
3742 	if (oip->ip_p == IPPROTO_TCP) {
3743 		tcp = (tcphdr_t *)dp;
3744 		csump = (u_short *)&tcp->th_sum;
3745 		flags = IPN_TCP;
3746 	} else if (oip->ip_p == IPPROTO_UDP) {
3747 		udphdr_t *udp;
3748 
3749 		udp = (udphdr_t *)dp;
3750 		tcp = (tcphdr_t *)dp;
3751 		csump = (u_short *)&udp->uh_sum;
3752 		flags = IPN_UDP;
3753 	} else if (oip->ip_p == IPPROTO_ICMP)
3754 		flags = IPN_ICMPQUERY;
3755 	dlen = fin->fin_plen - ((char *)dp - (char *)fin->fin_ip);
3756 
3757 	/*
3758 	 * Need to adjust ICMP header to include the real IP#'s and
3759 	 * port #'s.  Only apply a checksum change relative to the
3760 	 * IP address change as it will be modified again in ipf_nat_checkout
3761 	 * for both address and port.  Two checksum changes are
3762 	 * necessary for the two header address changes.  Be careful
3763 	 * to only modify the checksum once for the port # and twice
3764 	 * for the IP#.
3765 	 */
3766 
3767 	/*
3768 	 * Step 1
3769 	 * Fix the IP addresses in the offending IP packet. You also need
3770 	 * to adjust the IP header checksum of that offending IP packet.
3771 	 *
3772 	 * Normally, you would expect that the ICMP checksum of the
3773 	 * ICMP error message needs to be adjusted as well for the
3774 	 * IP address change in oip.
3775 	 * However, this is a NOP, because the ICMP checksum is
3776 	 * calculated over the complete ICMP packet, which includes the
3777 	 * changed oip IP addresses and oip->ip_sum. However, these
3778 	 * two changes cancel each other out (if the delta for
3779 	 * the IP address is x, then the delta for ip_sum is minus x),
3780 	 * so no change in the icmp_cksum is necessary.
3781 	 *
3782 	 * Inbound ICMP
3783 	 * ------------
3784 	 * MAP rule, SRC=a,DST=b -> SRC=c,DST=b
3785 	 * - response to outgoing packet (a,b)=>(c,b) (OIP_SRC=c,OIP_DST=b)
3786 	 * - OIP_SRC(c)=nat_newsrcip,          OIP_DST(b)=nat_newdstip
3787 	 *=> OIP_SRC(c)=nat_oldsrcip,          OIP_DST(b)=nat_olddstip
3788 	 *
3789 	 * RDR rule, SRC=a,DST=b -> SRC=a,DST=c
3790 	 * - response to outgoing packet (c,a)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3791 	 * - OIP_SRC(b)=nat_olddstip,          OIP_DST(a)=nat_oldsrcip
3792 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3793 	 *
3794 	 * REWRITE out rule, SRC=a,DST=b -> SRC=c,DST=d
3795 	 * - response to outgoing packet (a,b)=>(c,d) (OIP_SRC=c,OIP_DST=d)
3796 	 * - OIP_SRC(c)=nat_newsrcip,          OIP_DST(d)=nat_newdstip
3797 	 *=> OIP_SRC(c)=nat_oldsrcip,          OIP_DST(d)=nat_olddstip
3798 	 *
3799 	 * REWRITE in rule, SRC=a,DST=b -> SRC=c,DST=d
3800 	 * - response to outgoing packet (d,c)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3801 	 * - OIP_SRC(b)=nat_olddstip,          OIP_DST(a)=nat_oldsrcip
3802 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3803 	 *
3804 	 * Outbound ICMP
3805 	 * -------------
3806 	 * MAP rule, SRC=a,DST=b -> SRC=c,DST=b
3807 	 * - response to incoming packet (b,c)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3808 	 * - OIP_SRC(b)=nat_olddstip,          OIP_DST(a)=nat_oldsrcip
3809 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3810 	 *
3811 	 * RDR rule, SRC=a,DST=b -> SRC=a,DST=c
3812 	 * - response to incoming packet (a,b)=>(a,c) (OIP_SRC=a,OIP_DST=c)
3813 	 * - OIP_SRC(a)=nat_newsrcip,          OIP_DST(c)=nat_newdstip
3814 	 *=> OIP_SRC(a)=nat_oldsrcip,          OIP_DST(c)=nat_olddstip
3815 	 *
3816 	 * REWRITE out rule, SRC=a,DST=b -> SRC=c,DST=d
3817 	 * - response to incoming packet (d,c)=>(b,a) (OIP_SRC=c,OIP_DST=d)
3818 	 * - OIP_SRC(c)=nat_olddstip,          OIP_DST(d)=nat_oldsrcip
3819 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3820 	 *
3821 	 * REWRITE in rule, SRC=a,DST=b -> SRC=c,DST=d
3822 	 * - response to incoming packet (a,b)=>(c,d) (OIP_SRC=b,OIP_DST=a)
3823 	 * - OIP_SRC(b)=nat_newsrcip,          OIP_DST(a)=nat_newdstip
3824 	 *=> OIP_SRC(a)=nat_oldsrcip,          OIP_DST(c)=nat_olddstip
3825 	 */
3826 
3827 	if (((fin->fin_out == 0) && ((nat->nat_redir & NAT_MAP) != 0)) ||
3828 	    ((fin->fin_out == 1) && ((nat->nat_redir & NAT_REDIRECT) != 0))) {
3829 		a1.s_addr = ntohl(nat->nat_osrcaddr);
3830 		a4.s_addr = ntohl(oip->ip_src.s_addr);
3831 		a3.s_addr = ntohl(nat->nat_odstaddr);
3832 		a2.s_addr = ntohl(oip->ip_dst.s_addr);
3833 		oip->ip_src.s_addr = htonl(a1.s_addr);
3834 		oip->ip_dst.s_addr = htonl(a3.s_addr);
3835 		odst = 1;
3836 	} else {
3837 		a1.s_addr = ntohl(nat->nat_ndstaddr);
3838 		a2.s_addr = ntohl(oip->ip_dst.s_addr);
3839 		a3.s_addr = ntohl(nat->nat_nsrcaddr);
3840 		a4.s_addr = ntohl(oip->ip_src.s_addr);
3841 		oip->ip_dst.s_addr = htonl(a3.s_addr);
3842 		oip->ip_src.s_addr = htonl(a1.s_addr);
3843 		odst = 0;
3844 	}
3845 	sum1 = 0;
3846 	sum2 = 0;
3847 	sumd = 0;
3848 	CALC_SUMD(a2.s_addr, a3.s_addr, sum1);
3849 	CALC_SUMD(a4.s_addr, a1.s_addr, sum2);
3850 	sumd = sum2 + sum1;
3851 	if (sumd != 0)
3852 		ipf_fix_datacksum(&oip->ip_sum, sumd);
3853 
3854 	sumd2 = sumd;
3855 	sum1 = 0;
3856 	sum2 = 0;
3857 
3858 	/*
3859 	 * Fix UDP pseudo header checksum to compensate for the
3860 	 * IP address change.
3861 	 */
3862 	if (((flags & IPN_TCPUDP) != 0) && (dlen >= 4)) {
3863 		u_32_t sum3, sum4, sumt;
3864 
3865 		/*
3866 		 * Step 2 :
3867 		 * For offending TCP/UDP IP packets, translate the ports as
3868 		 * well, based on the NAT specification. Of course such
3869 		 * a change may be reflected in the ICMP checksum as well.
3870 		 *
3871 		 * Since the port fields are part of the TCP/UDP checksum
3872 		 * of the offending IP packet, you need to adjust that checksum
3873 		 * as well... except that the change in the port numbers should
3874 		 * be offset by the checksum change.  However, the TCP/UDP
3875 		 * checksum will also need to change if there has been an
3876 		 * IP address change.
3877 		 */
3878 		if (odst == 1) {
3879 			sum1 = ntohs(nat->nat_osport);
3880 			sum4 = ntohs(tcp->th_sport);
3881 			sum3 = ntohs(nat->nat_odport);
3882 			sum2 = ntohs(tcp->th_dport);
3883 
3884 			tcp->th_sport = htons(sum1);
3885 			tcp->th_dport = htons(sum3);
3886 		} else {
3887 			sum1 = ntohs(nat->nat_ndport);
3888 			sum2 = ntohs(tcp->th_dport);
3889 			sum3 = ntohs(nat->nat_nsport);
3890 			sum4 = ntohs(tcp->th_sport);
3891 
3892 			tcp->th_dport = htons(sum3);
3893 			tcp->th_sport = htons(sum1);
3894 		}
3895 		CALC_SUMD(sum4, sum1, sumt);
3896 		sumd += sumt;
3897 		CALC_SUMD(sum2, sum3, sumt);
3898 		sumd += sumt;
3899 
3900 		if (sumd != 0 || sumd2 != 0) {
3901 			/*
3902 			 * At this point, sumd is the delta to apply to the
3903 			 * TCP/UDP header, given the changes in both the IP
3904 			 * address and the ports and sumd2 is the delta to
3905 			 * apply to the ICMP header, given the IP address
3906 			 * change delta that may need to be applied to the
3907 			 * TCP/UDP checksum instead.
3908 			 *
3909 			 * If we will both the IP and TCP/UDP checksums
3910 			 * then the ICMP checksum changes by the address
3911 			 * delta applied to the TCP/UDP checksum.  If we
3912 			 * do not change the TCP/UDP checksum them we
3913 			 * apply the delta in ports to the ICMP checksum.
3914 			 */
3915 			if (oip->ip_p == IPPROTO_UDP) {
3916 				if ((dlen >= 8) && (*csump != 0)) {
3917 					ipf_fix_datacksum(csump, sumd);
3918 				} else {
3919 					CALC_SUMD(sum1, sum4, sumd2);
3920 					CALC_SUMD(sum3, sum2, sumt);
3921 					sumd2 += sumt;
3922 				}
3923 			} else if (oip->ip_p == IPPROTO_TCP) {
3924 				if (dlen >= 18) {
3925 					ipf_fix_datacksum(csump, sumd);
3926 				} else {
3927 					CALC_SUMD(sum1, sum4, sumd2);
3928 					CALC_SUMD(sum3, sum2, sumt);
3929 					sumd2 += sumt;
3930 				}
3931 			}
3932 			if (sumd2 != 0) {
3933 				sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3934 				sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3935 				sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3936 				ipf_fix_incksum(0, &icmp->icmp_cksum, sumd2, 0);
3937 			}
3938 		}
3939 	} else if (((flags & IPN_ICMPQUERY) != 0) && (dlen >= 8)) {
3940 		icmphdr_t *orgicmp;
3941 
3942 		/*
3943 		 * XXX - what if this is bogus hl and we go off the end ?
3944 		 * In this case, ipf_nat_icmperrorlookup() will have
3945 		 * returned NULL.
3946 		 */
3947 		orgicmp = (icmphdr_t *)dp;
3948 
3949 		if (odst == 1) {
3950 			if (orgicmp->icmp_id != nat->nat_osport) {
3951 
3952 				/*
3953 				 * Fix ICMP checksum (of the offening ICMP
3954 				 * query packet) to compensate the change
3955 				 * in the ICMP id of the offending ICMP
3956 				 * packet.
3957 				 *
3958 				 * Since you modify orgicmp->icmp_id with
3959 				 * a delta (say x) and you compensate that
3960 				 * in origicmp->icmp_cksum with a delta
3961 				 * minus x, you don't have to adjust the
3962 				 * overall icmp->icmp_cksum
3963 				 */
3964 				sum1 = ntohs(orgicmp->icmp_id);
3965 				sum2 = ntohs(nat->nat_oicmpid);
3966 				CALC_SUMD(sum1, sum2, sumd);
3967 				orgicmp->icmp_id = nat->nat_oicmpid;
3968 				ipf_fix_datacksum(&orgicmp->icmp_cksum, sumd);
3969 			}
3970 		} /* nat_dir == NAT_INBOUND is impossible for icmp queries */
3971 	}
3972 	return nat;
3973 }
3974 
3975 
3976 /*
3977  *       MAP-IN    MAP-OUT   RDR-IN   RDR-OUT
3978  * osrc    X       == src    == src      X
3979  * odst    X       == dst    == dst      X
3980  * nsrc  == dst      X         X      == dst
3981  * ndst  == src      X         X      == src
3982  * MAP = NAT_OUTBOUND, RDR = NAT_INBOUND
3983  */
3984 /*
3985  * NB: these lookups don't lock access to the list, it assumed that it has
3986  * already been done!
3987  */
3988 /* ------------------------------------------------------------------------ */
3989 /* Function:    ipf_nat_inlookup                                            */
3990 /* Returns:     nat_t* - NULL == no match,                                  */
3991 /*                       else pointer to matching NAT entry                 */
3992 /* Parameters:  fin(I)    - pointer to packet information                   */
3993 /*              flags(I)  - NAT flags for this packet                       */
3994 /*              p(I)      - protocol for this packet                        */
3995 /*              src(I)    - source IP address                               */
3996 /*              mapdst(I) - destination IP address                          */
3997 /*                                                                          */
3998 /* Lookup a nat entry based on the mapped destination ip address/port and   */
3999 /* real source address/port.  We use this lookup when receiving a packet,   */
4000 /* we're looking for a table entry, based on the destination address.       */
4001 /*                                                                          */
4002 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY.         */
4003 /*                                                                          */
4004 /* NOTE: IT IS ASSUMED THAT  IS ONLY HELD WITH A READ LOCK WHEN             */
4005 /*       THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags.             */
4006 /*                                                                          */
4007 /* flags   -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if   */
4008 /*            the packet is of said protocol                                */
4009 /* ------------------------------------------------------------------------ */
4010 nat_t *
4011 ipf_nat_inlookup(fr_info_t *fin, u_int flags, u_int p, struct in_addr src,
4012     struct in_addr mapdst)
4013 {
4014 	ipf_main_softc_t *softc = fin->fin_main_soft;
4015 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4016 	u_short sport, dport;
4017 	ipnat_t *ipn;
4018 	nat_t *nat;
4019 	int nflags;
4020 	u_32_t dst;
4021 	void *ifp;
4022 	u_int hv, rhv;
4023 
4024 	ifp = fin->fin_ifp;
4025 	dst = mapdst.s_addr;
4026 
4027 	switch (p)
4028 	{
4029 	case IPPROTO_TCP :
4030 	case IPPROTO_UDP :
4031 		sport = htons(fin->fin_data[0]);
4032 		dport = htons(fin->fin_data[1]);
4033 		break;
4034 	case IPPROTO_ICMP :
4035 		if (flags & IPN_ICMPERR) {
4036 			sport = fin->fin_data[1];
4037 			dport = 0;
4038 		} else {
4039 			dport = fin->fin_data[1];
4040 			sport = 0;
4041 		}
4042 		break;
4043 	default :
4044 		sport = 0;
4045 		dport = 0;
4046 		break;
4047 	}
4048 
4049 
4050 	if ((flags & SI_WILDP) != 0)
4051 		goto find_in_wild_ports;
4052 
4053 	rhv = NAT_HASH_FN(dst, dport, 0xffffffff);
4054 	rhv = NAT_HASH_FN(src.s_addr, rhv + sport, 0xffffffff);
4055 	hv = rhv % softn->ipf_nat_table_sz;
4056 	nat = softn->ipf_nat_table[1][hv];
4057 	/* TRACE dst, dport, src, sport, hv, nat */
4058 
4059 	for (; nat; nat = nat->nat_hnext[1]) {
4060 		if (nat->nat_ifps[0] != NULL) {
4061 			if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
4062 				continue;
4063 		}
4064 
4065 		if (nat->nat_pr[0] != p)
4066 			continue;
4067 
4068 		switch (nat->nat_dir)
4069 		{
4070 		case NAT_INBOUND :
4071 		case NAT_DIVERTIN :
4072 			if (nat->nat_v[0] != 4)
4073 				continue;
4074 			if (nat->nat_osrcaddr != src.s_addr ||
4075 			    nat->nat_odstaddr != dst)
4076 				continue;
4077 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4078 				if (nat->nat_osport != sport)
4079 					continue;
4080 				if (nat->nat_odport != dport)
4081 					continue;
4082 
4083 			} else if (p == IPPROTO_ICMP) {
4084 				if (nat->nat_osport != dport) {
4085 					continue;
4086 				}
4087 			}
4088 			break;
4089 		case NAT_DIVERTOUT :
4090 			if (nat->nat_dlocal)
4091 				continue;
4092 		case NAT_OUTBOUND :
4093 			if (nat->nat_v[1] != 4)
4094 				continue;
4095 			if (nat->nat_dlocal)
4096 				continue;
4097 			if (nat->nat_dlocal)
4098 				continue;
4099 			if (nat->nat_ndstaddr != src.s_addr ||
4100 			    nat->nat_nsrcaddr != dst)
4101 				continue;
4102 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4103 				if (nat->nat_ndport != sport)
4104 					continue;
4105 				if (nat->nat_nsport != dport)
4106 					continue;
4107 
4108 			} else if (p == IPPROTO_ICMP) {
4109 				if (nat->nat_osport != dport) {
4110 					continue;
4111 				}
4112 			}
4113 			break;
4114 		}
4115 
4116 
4117 		if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4118 			ipn = nat->nat_ptr;
4119 			if ((ipn != NULL) && (nat->nat_aps != NULL))
4120 				if (ipf_proxy_match(fin, nat) != 0)
4121 					continue;
4122 		}
4123 		if ((nat->nat_ifps[0] == NULL) && (ifp != NULL)) {
4124 			nat->nat_ifps[0] = ifp;
4125 			nat->nat_mtu[0] = GETIFMTU_4(ifp);
4126 		}
4127 		return nat;
4128 	}
4129 
4130 	/*
4131 	 * So if we didn't find it but there are wildcard members in the hash
4132 	 * table, go back and look for them.  We do this search and update here
4133 	 * because it is modifying the NAT table and we want to do this only
4134 	 * for the first packet that matches.  The exception, of course, is
4135 	 * for "dummy" (FI_IGNORE) lookups.
4136 	 */
4137 find_in_wild_ports:
4138 	if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH)) {
4139 		NBUMPSIDEX(0, ns_lookup_miss, ns_lookup_miss_0);
4140 		return NULL;
4141 	}
4142 	if (softn->ipf_nat_stats.ns_wilds == 0 || (fin->fin_flx & FI_NOWILD)) {
4143 		NBUMPSIDEX(0, ns_lookup_nowild, ns_lookup_nowild_0);
4144 		return NULL;
4145 	}
4146 
4147 	RWLOCK_EXIT(&softc->ipf_nat);
4148 
4149 	hv = NAT_HASH_FN(dst, 0, 0xffffffff);
4150 	hv = NAT_HASH_FN(src.s_addr, hv, softn->ipf_nat_table_sz);
4151 	WRITE_ENTER(&softc->ipf_nat);
4152 
4153 	nat = softn->ipf_nat_table[1][hv];
4154 	/* TRACE dst, src, hv, nat */
4155 	for (; nat; nat = nat->nat_hnext[1]) {
4156 		if (nat->nat_ifps[0] != NULL) {
4157 			if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
4158 				continue;
4159 		}
4160 
4161 		if (nat->nat_pr[0] != fin->fin_p)
4162 			continue;
4163 
4164 		switch (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND))
4165 		{
4166 		case NAT_INBOUND :
4167 			if (nat->nat_v[0] != 4)
4168 				continue;
4169 			if (nat->nat_osrcaddr != src.s_addr ||
4170 			    nat->nat_odstaddr != dst)
4171 				continue;
4172 			break;
4173 		case NAT_OUTBOUND :
4174 			if (nat->nat_v[1] != 4)
4175 				continue;
4176 			if (nat->nat_ndstaddr != src.s_addr ||
4177 			    nat->nat_nsrcaddr != dst)
4178 				continue;
4179 			break;
4180 		}
4181 
4182 		nflags = nat->nat_flags;
4183 		if (!(nflags & (NAT_TCPUDP|SI_WILDP)))
4184 			continue;
4185 
4186 		if (ipf_nat_wildok(nat, (int)sport, (int)dport, nflags,
4187 				   NAT_INBOUND) == 1) {
4188 			if ((fin->fin_flx & FI_IGNORE) != 0)
4189 				break;
4190 			if ((nflags & SI_CLONE) != 0) {
4191 				nat = ipf_nat_clone(fin, nat);
4192 				if (nat == NULL)
4193 					break;
4194 			} else {
4195 				MUTEX_ENTER(&softn->ipf_nat_new);
4196 				softn->ipf_nat_stats.ns_wilds--;
4197 				MUTEX_EXIT(&softn->ipf_nat_new);
4198 			}
4199 
4200 			if (nat->nat_dir == NAT_INBOUND) {
4201 				if (nat->nat_osport == 0) {
4202 					nat->nat_osport = sport;
4203 					nat->nat_nsport = sport;
4204 				}
4205 				if (nat->nat_odport == 0) {
4206 					nat->nat_odport = dport;
4207 					nat->nat_ndport = dport;
4208 				}
4209 			} else if (nat->nat_dir == NAT_OUTBOUND) {
4210 				if (nat->nat_osport == 0) {
4211 					nat->nat_osport = dport;
4212 					nat->nat_nsport = dport;
4213 				}
4214 				if (nat->nat_odport == 0) {
4215 					nat->nat_odport = sport;
4216 					nat->nat_ndport = sport;
4217 				}
4218 			}
4219 			if ((nat->nat_ifps[0] == NULL) && (ifp != NULL)) {
4220 				nat->nat_ifps[0] = ifp;
4221 				nat->nat_mtu[0] = GETIFMTU_4(ifp);
4222 			}
4223 			nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
4224 			ipf_nat_tabmove(softn, nat);
4225 			break;
4226 		}
4227 	}
4228 
4229 	MUTEX_DOWNGRADE(&softc->ipf_nat);
4230 
4231 	if (nat == NULL) {
4232 		NBUMPSIDE(0, ns_lookup_miss);
4233 	}
4234 	return nat;
4235 }
4236 
4237 
4238 /* ------------------------------------------------------------------------ */
4239 /* Function:    ipf_nat_tabmove                                             */
4240 /* Returns:     Nil                                                         */
4241 /* Parameters:  softn(I) - pointer to NAT context structure                 */
4242 /*              nat(I)   - pointer to NAT structure                         */
4243 /* Write Lock:  ipf_nat                                                     */
4244 /*                                                                          */
4245 /* This function is only called for TCP/UDP NAT table entries where the     */
4246 /* original was placed in the table without hashing on the ports and we now */
4247 /* want to include hashing on port numbers.                                 */
4248 /* ------------------------------------------------------------------------ */
4249 static void
4250 ipf_nat_tabmove(ipf_nat_softc_t *softn, nat_t *nat)
4251 {
4252 	u_int hv0, hv1, rhv0, rhv1;
4253 	natstat_t *nsp;
4254 	nat_t **natp;
4255 
4256 	if (nat->nat_flags & SI_CLONE)
4257 		return;
4258 
4259 	nsp = &softn->ipf_nat_stats;
4260 	/*
4261 	 * Remove the NAT entry from the old location
4262 	 */
4263 	if (nat->nat_hnext[0])
4264 		nat->nat_hnext[0]->nat_phnext[0] = nat->nat_phnext[0];
4265 	*nat->nat_phnext[0] = nat->nat_hnext[0];
4266 	hv0 = nat->nat_hv[0] % softn->ipf_nat_table_sz;
4267 	hv1 = nat->nat_hv[1] % softn->ipf_nat_table_sz;
4268 
4269 	ASSERT(nsp->ns_side[0].ns_bucketlen[hv0] > 0);
4270 	nsp->ns_side[0].ns_bucketlen[hv0]--;
4271 
4272 	if (nat->nat_hnext[1])
4273 		nat->nat_hnext[1]->nat_phnext[1] = nat->nat_phnext[1];
4274 	*nat->nat_phnext[1] = nat->nat_hnext[1];
4275 	ASSERT(nsp->ns_side[1].ns_bucketlen[hv1] > 0);
4276 	nsp->ns_side[1].ns_bucketlen[hv1]--;
4277 
4278 	/*
4279 	 * Add into the NAT table in the new position
4280 	 */
4281 	rhv0 = NAT_HASH_FN(nat->nat_osrcaddr, nat->nat_osport, 0xffffffff);
4282 	rhv0 = NAT_HASH_FN(nat->nat_odstaddr, rhv0 + nat->nat_odport,
4283 			   0xffffffff);
4284 	rhv1 = NAT_HASH_FN(nat->nat_nsrcaddr, nat->nat_nsport, 0xffffffff);
4285 	rhv1 = NAT_HASH_FN(nat->nat_ndstaddr, rhv1 + nat->nat_ndport,
4286 			   0xffffffff);
4287 
4288 	if ((nat->nat_dir & NAT_OUTBOUND) == NAT_OUTBOUND) {
4289 		nat->nat_hv[0] = rhv0;
4290 		nat->nat_hv[1] = rhv1;
4291 	} else {
4292 		nat->nat_hv[0] = rhv1;
4293 		nat->nat_hv[1] = rhv0;
4294 	}
4295 
4296 	hv0 = nat->nat_hv[0] % softn->ipf_nat_table_sz;
4297 	hv1 = nat->nat_hv[1] % softn->ipf_nat_table_sz;
4298 
4299 	/* TRACE nat_osrcaddr, nat_osport, nat_odstaddr, nat_odport, hv0 */
4300 	/* TRACE nat_nsrcaddr, nat_nsport, nat_ndstaddr, nat_ndport, hv1 */
4301 
4302 	natp = &softn->ipf_nat_table[0][hv0];
4303 	if (*natp)
4304 		(*natp)->nat_phnext[0] = &nat->nat_hnext[0];
4305 	nat->nat_phnext[0] = natp;
4306 	nat->nat_hnext[0] = *natp;
4307 	*natp = nat;
4308 	nsp->ns_side[0].ns_bucketlen[hv0]++;
4309 
4310 	natp = &softn->ipf_nat_table[1][hv1];
4311 	if (*natp)
4312 		(*natp)->nat_phnext[1] = &nat->nat_hnext[1];
4313 	nat->nat_phnext[1] = natp;
4314 	nat->nat_hnext[1] = *natp;
4315 	*natp = nat;
4316 	nsp->ns_side[1].ns_bucketlen[hv1]++;
4317 }
4318 
4319 
4320 /* ------------------------------------------------------------------------ */
4321 /* Function:    ipf_nat_outlookup                                           */
4322 /* Returns:     nat_t* - NULL == no match,                                  */
4323 /*                       else pointer to matching NAT entry                 */
4324 /* Parameters:  fin(I)   - pointer to packet information                    */
4325 /*              flags(I) - NAT flags for this packet                        */
4326 /*              p(I)     - protocol for this packet                         */
4327 /*              src(I)   - source IP address                                */
4328 /*              dst(I)   - destination IP address                           */
4329 /*              rw(I)    - 1 == write lock on  held, 0 == read lock.        */
4330 /*                                                                          */
4331 /* Lookup a nat entry based on the source 'real' ip address/port and        */
4332 /* destination address/port.  We use this lookup when sending a packet out, */
4333 /* we're looking for a table entry, based on the source address.            */
4334 /*                                                                          */
4335 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY.         */
4336 /*                                                                          */
4337 /* NOTE: IT IS ASSUMED THAT  IS ONLY HELD WITH A READ LOCK WHEN             */
4338 /*       THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags.             */
4339 /*                                                                          */
4340 /* flags   -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if   */
4341 /*            the packet is of said protocol                                */
4342 /* ------------------------------------------------------------------------ */
4343 nat_t *
4344 ipf_nat_outlookup(fr_info_t *fin, u_int flags, u_int p, struct in_addr src,
4345    struct in_addr dst)
4346 {
4347 	ipf_main_softc_t *softc = fin->fin_main_soft;
4348 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4349 	u_short sport, dport;
4350 	ipnat_t *ipn;
4351 	nat_t *nat;
4352 	void *ifp;
4353 	u_int hv;
4354 
4355 	ifp = fin->fin_ifp;
4356 	sport = 0;
4357 	dport = 0;
4358 
4359 	switch (p)
4360 	{
4361 	case IPPROTO_TCP :
4362 	case IPPROTO_UDP :
4363 		sport = htons(fin->fin_data[0]);
4364 		dport = htons(fin->fin_data[1]);
4365 		break;
4366 	case IPPROTO_ICMP :
4367 		if (flags & IPN_ICMPERR)
4368 			sport = fin->fin_data[1];
4369 		else
4370 			dport = fin->fin_data[1];
4371 		break;
4372 	default :
4373 		break;
4374 	}
4375 
4376 	if ((flags & SI_WILDP) != 0)
4377 		goto find_out_wild_ports;
4378 
4379 	hv = NAT_HASH_FN(src.s_addr, sport, 0xffffffff);
4380 	hv = NAT_HASH_FN(dst.s_addr, hv + dport, softn->ipf_nat_table_sz);
4381 	nat = softn->ipf_nat_table[0][hv];
4382 
4383 	/* TRACE src, sport, dst, dport, hv, nat */
4384 
4385 	for (; nat; nat = nat->nat_hnext[0]) {
4386 		if (nat->nat_ifps[1] != NULL) {
4387 			if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
4388 				continue;
4389 		}
4390 
4391 		if (nat->nat_pr[1] != p)
4392 			continue;
4393 
4394 		switch (nat->nat_dir)
4395 		{
4396 		case NAT_INBOUND :
4397 		case NAT_DIVERTIN :
4398 			if (nat->nat_v[1] != 4)
4399 				continue;
4400 			if (nat->nat_ndstaddr != src.s_addr ||
4401 			    nat->nat_nsrcaddr != dst.s_addr)
4402 				continue;
4403 
4404 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4405 				if (nat->nat_ndport != sport)
4406 					continue;
4407 				if (nat->nat_nsport != dport)
4408 					continue;
4409 
4410 			} else if (p == IPPROTO_ICMP) {
4411 				if (nat->nat_osport != dport) {
4412 					continue;
4413 				}
4414 			}
4415 			break;
4416 		case NAT_OUTBOUND :
4417 		case NAT_DIVERTOUT :
4418 			if (nat->nat_v[0] != 4)
4419 				continue;
4420 			if (nat->nat_osrcaddr != src.s_addr ||
4421 			    nat->nat_odstaddr != dst.s_addr)
4422 				continue;
4423 
4424 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4425 				if (nat->nat_odport != dport)
4426 					continue;
4427 				if (nat->nat_osport != sport)
4428 					continue;
4429 
4430 			} else if (p == IPPROTO_ICMP) {
4431 				if (nat->nat_osport != dport) {
4432 					continue;
4433 				}
4434 			}
4435 			break;
4436 		}
4437 
4438 		ipn = nat->nat_ptr;
4439 		if ((ipn != NULL) && (nat->nat_aps != NULL))
4440 			if (ipf_proxy_match(fin, nat) != 0)
4441 				continue;
4442 
4443 		if ((nat->nat_ifps[1] == NULL) && (ifp != NULL)) {
4444 			nat->nat_ifps[1] = ifp;
4445 			nat->nat_mtu[1] = GETIFMTU_4(ifp);
4446 		}
4447 		return nat;
4448 	}
4449 
4450 	/*
4451 	 * So if we didn't find it but there are wildcard members in the hash
4452 	 * table, go back and look for them.  We do this search and update here
4453 	 * because it is modifying the NAT table and we want to do this only
4454 	 * for the first packet that matches.  The exception, of course, is
4455 	 * for "dummy" (FI_IGNORE) lookups.
4456 	 */
4457 find_out_wild_ports:
4458 	if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH)) {
4459 		NBUMPSIDEX(1, ns_lookup_miss, ns_lookup_miss_1);
4460 		return NULL;
4461 	}
4462 	if (softn->ipf_nat_stats.ns_wilds == 0 || (fin->fin_flx & FI_NOWILD)) {
4463 		NBUMPSIDEX(1, ns_lookup_nowild, ns_lookup_nowild_1);
4464 		return NULL;
4465 	}
4466 
4467 	RWLOCK_EXIT(&softc->ipf_nat);
4468 
4469 	hv = NAT_HASH_FN(src.s_addr, 0, 0xffffffff);
4470 	hv = NAT_HASH_FN(dst.s_addr, hv, softn->ipf_nat_table_sz);
4471 
4472 	WRITE_ENTER(&softc->ipf_nat);
4473 
4474 	nat = softn->ipf_nat_table[0][hv];
4475 	for (; nat; nat = nat->nat_hnext[0]) {
4476 		if (nat->nat_ifps[1] != NULL) {
4477 			if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
4478 				continue;
4479 		}
4480 
4481 		if (nat->nat_pr[1] != fin->fin_p)
4482 			continue;
4483 
4484 		switch (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND))
4485 		{
4486 		case NAT_INBOUND :
4487 			if (nat->nat_v[1] != 4)
4488 				continue;
4489 			if (nat->nat_ndstaddr != src.s_addr ||
4490 			    nat->nat_nsrcaddr != dst.s_addr)
4491 				continue;
4492 			break;
4493 		case NAT_OUTBOUND :
4494 			if (nat->nat_v[0] != 4)
4495 				continue;
4496 			if (nat->nat_osrcaddr != src.s_addr ||
4497 			    nat->nat_odstaddr != dst.s_addr)
4498 				continue;
4499 			break;
4500 		}
4501 
4502 		if (!(nat->nat_flags & (NAT_TCPUDP|SI_WILDP)))
4503 			continue;
4504 
4505 		if (ipf_nat_wildok(nat, (int)sport, (int)dport, nat->nat_flags,
4506 				   NAT_OUTBOUND) == 1) {
4507 			if ((fin->fin_flx & FI_IGNORE) != 0)
4508 				break;
4509 			if ((nat->nat_flags & SI_CLONE) != 0) {
4510 				nat = ipf_nat_clone(fin, nat);
4511 				if (nat == NULL)
4512 					break;
4513 			} else {
4514 				MUTEX_ENTER(&softn->ipf_nat_new);
4515 				softn->ipf_nat_stats.ns_wilds--;
4516 				MUTEX_EXIT(&softn->ipf_nat_new);
4517 			}
4518 
4519 			if (nat->nat_dir == NAT_OUTBOUND) {
4520 				if (nat->nat_osport == 0) {
4521 					nat->nat_osport = sport;
4522 					nat->nat_nsport = sport;
4523 				}
4524 				if (nat->nat_odport == 0) {
4525 					nat->nat_odport = dport;
4526 					nat->nat_ndport = dport;
4527 				}
4528 			} else if (nat->nat_dir == NAT_INBOUND) {
4529 				if (nat->nat_osport == 0) {
4530 					nat->nat_osport = dport;
4531 					nat->nat_nsport = dport;
4532 				}
4533 				if (nat->nat_odport == 0) {
4534 					nat->nat_odport = sport;
4535 					nat->nat_ndport = sport;
4536 				}
4537 			}
4538 			if ((nat->nat_ifps[1] == NULL) && (ifp != NULL)) {
4539 				nat->nat_ifps[1] = ifp;
4540 				nat->nat_mtu[1] = GETIFMTU_4(ifp);
4541 			}
4542 			nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
4543 			ipf_nat_tabmove(softn, nat);
4544 			break;
4545 		}
4546 	}
4547 
4548 	MUTEX_DOWNGRADE(&softc->ipf_nat);
4549 
4550 	if (nat == NULL) {
4551 		NBUMPSIDE(1, ns_lookup_miss);
4552 	}
4553 	return nat;
4554 }
4555 
4556 
4557 /* ------------------------------------------------------------------------ */
4558 /* Function:    ipf_nat_lookupredir                                         */
4559 /* Returns:     nat_t* - NULL == no match,                                  */
4560 /*                       else pointer to matching NAT entry                 */
4561 /* Parameters:  np(I) - pointer to description of packet to find NAT table  */
4562 /*                      entry for.                                          */
4563 /*                                                                          */
4564 /* Lookup the NAT tables to search for a matching redirect                  */
4565 /* The contents of natlookup_t should imitate those found in a packet that  */
4566 /* would be translated - ie a packet coming in for RDR or going out for MAP.*/
4567 /* We can do the lookup in one of two ways, imitating an inbound or         */
4568 /* outbound  packet.  By default we assume outbound, unless IPN_IN is set.  */
4569 /* For IN, the fields are set as follows:                                   */
4570 /*     nl_real* = source information                                        */
4571 /*     nl_out* = destination information (translated)                       */
4572 /* For an out packet, the fields are set like this:                         */
4573 /*     nl_in* = source information (untranslated)                           */
4574 /*     nl_out* = destination information (translated)                       */
4575 /* ------------------------------------------------------------------------ */
4576 nat_t *
4577 ipf_nat_lookupredir(natlookup_t *np)
4578 {
4579 	fr_info_t fi;
4580 	nat_t *nat;
4581 
4582 	bzero((char *)&fi, sizeof(fi));
4583 	if (np->nl_flags & IPN_IN) {
4584 		fi.fin_data[0] = ntohs(np->nl_realport);
4585 		fi.fin_data[1] = ntohs(np->nl_outport);
4586 	} else {
4587 		fi.fin_data[0] = ntohs(np->nl_inport);
4588 		fi.fin_data[1] = ntohs(np->nl_outport);
4589 	}
4590 	if (np->nl_flags & IPN_TCP)
4591 		fi.fin_p = IPPROTO_TCP;
4592 	else if (np->nl_flags & IPN_UDP)
4593 		fi.fin_p = IPPROTO_UDP;
4594 	else if (np->nl_flags & (IPN_ICMPERR|IPN_ICMPQUERY))
4595 		fi.fin_p = IPPROTO_ICMP;
4596 
4597 	/*
4598 	 * We can do two sorts of lookups:
4599 	 * - IPN_IN: we have the `real' and `out' address, look for `in'.
4600 	 * - default: we have the `in' and `out' address, look for `real'.
4601 	 */
4602 	if (np->nl_flags & IPN_IN) {
4603 		if ((nat = ipf_nat_inlookup(&fi, np->nl_flags, fi.fin_p,
4604 					    np->nl_realip, np->nl_outip))) {
4605 			np->nl_inip = nat->nat_odstip;
4606 			np->nl_inport = nat->nat_odport;
4607 		}
4608 	} else {
4609 		/*
4610 		 * If nl_inip is non null, this is a lookup based on the real
4611 		 * ip address. Else, we use the fake.
4612 		 */
4613 		if ((nat = ipf_nat_outlookup(&fi, np->nl_flags, fi.fin_p,
4614 					 np->nl_inip, np->nl_outip))) {
4615 
4616 			if ((np->nl_flags & IPN_FINDFORWARD) != 0) {
4617 				fr_info_t fin;
4618 				bzero((char *)&fin, sizeof(fin));
4619 				fin.fin_p = nat->nat_pr[0];
4620 				fin.fin_data[0] = ntohs(nat->nat_ndport);
4621 				fin.fin_data[1] = ntohs(nat->nat_nsport);
4622 				if (ipf_nat_inlookup(&fin, np->nl_flags,
4623 						     fin.fin_p, nat->nat_ndstip,
4624 						     nat->nat_nsrcip) != NULL) {
4625 					np->nl_flags &= ~IPN_FINDFORWARD;
4626 				}
4627 			}
4628 
4629 			np->nl_realip = nat->nat_ndstip;
4630 			np->nl_realport = nat->nat_ndport;
4631 		}
4632  	}
4633 
4634 	return nat;
4635 }
4636 
4637 
4638 /* ------------------------------------------------------------------------ */
4639 /* Function:    ipf_nat_match                                               */
4640 /* Returns:     int - 0 == no match, 1 == match                             */
4641 /* Parameters:  fin(I)   - pointer to packet information                    */
4642 /*              np(I)    - pointer to NAT rule                              */
4643 /*                                                                          */
4644 /* Pull the matching of a packet against a NAT rule out of that complex     */
4645 /* loop inside ipf_nat_checkin() and lay it out properly in its own function. */
4646 /* ------------------------------------------------------------------------ */
4647 static int
4648 ipf_nat_match(fr_info_t *fin, ipnat_t *np)
4649 {
4650 	ipf_main_softc_t *softc = fin->fin_main_soft;
4651 	frtuc_t *ft;
4652 	int match;
4653 
4654 	match = 0;
4655 	switch (np->in_osrcatype)
4656 	{
4657 	case FRI_NORMAL :
4658 		match = ((fin->fin_saddr & np->in_osrcmsk) != np->in_osrcaddr);
4659 		break;
4660 	case FRI_LOOKUP :
4661 		match = (*np->in_osrcfunc)(softc, np->in_osrcptr,
4662 					   4, &fin->fin_saddr, fin->fin_plen);
4663 		break;
4664 	}
4665 	match ^= ((np->in_flags & IPN_NOTSRC) != 0);
4666 	if (match)
4667 		return 0;
4668 
4669 	match = 0;
4670 	switch (np->in_odstatype)
4671 	{
4672 	case FRI_NORMAL :
4673 		match = ((fin->fin_daddr & np->in_odstmsk) != np->in_odstaddr);
4674 		break;
4675 	case FRI_LOOKUP :
4676 		match = (*np->in_odstfunc)(softc, np->in_odstptr,
4677 					   4, &fin->fin_daddr, fin->fin_plen);
4678 		break;
4679 	}
4680 
4681 	match ^= ((np->in_flags & IPN_NOTDST) != 0);
4682 	if (match)
4683 		return 0;
4684 
4685 	ft = &np->in_tuc;
4686 	if (!(fin->fin_flx & FI_TCPUDP) ||
4687 	    (fin->fin_flx & (FI_SHORT|FI_FRAGBODY))) {
4688 		if (ft->ftu_scmp || ft->ftu_dcmp)
4689 			return 0;
4690 		return 1;
4691 	}
4692 
4693 	return ipf_tcpudpchk(&fin->fin_fi, ft);
4694 }
4695 
4696 
4697 /* ------------------------------------------------------------------------ */
4698 /* Function:    ipf_nat_update                                              */
4699 /* Returns:     Nil                                                         */
4700 /* Parameters:  fin(I) - pointer to packet information                      */
4701 /*              nat(I) - pointer to NAT structure                           */
4702 /*                                                                          */
4703 /* Updates the lifetime of a NAT table entry for non-TCP packets.  Must be  */
4704 /* called with fin_rev updated - i.e. after calling ipf_nat_proto().        */
4705 /*                                                                          */
4706 /* This *MUST* be called after ipf_nat_proto() as it expects fin_rev to     */
4707 /* already be set.                                                          */
4708 /* ------------------------------------------------------------------------ */
4709 void
4710 ipf_nat_update(fr_info_t *fin, nat_t *nat)
4711 {
4712 	ipf_main_softc_t *softc = fin->fin_main_soft;
4713 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4714 	ipftq_t *ifq, *ifq2;
4715 	ipftqent_t *tqe;
4716 	ipnat_t *np = nat->nat_ptr;
4717 
4718 	tqe = &nat->nat_tqe;
4719 	ifq = tqe->tqe_ifq;
4720 
4721 	/*
4722 	 * We allow over-riding of NAT timeouts from NAT rules, even for
4723 	 * TCP, however, if it is TCP and there is no rule timeout set,
4724 	 * then do not update the timeout here.
4725 	 */
4726 	if (np != NULL) {
4727 		np->in_bytes[fin->fin_rev] += fin->fin_plen;
4728 		ifq2 = np->in_tqehead[fin->fin_rev];
4729 	} else {
4730 		ifq2 = NULL;
4731 	}
4732 
4733 	if (nat->nat_pr[0] == IPPROTO_TCP && ifq2 == NULL) {
4734 		(void) ipf_tcp_age(&nat->nat_tqe, fin, softn->ipf_nat_tcptq,
4735 				   0, 2);
4736 	} else {
4737 		if (ifq2 == NULL) {
4738 			if (nat->nat_pr[0] == IPPROTO_UDP)
4739 				ifq2 = fin->fin_rev ? &softn->ipf_nat_udpacktq :
4740 						      &softn->ipf_nat_udptq;
4741 			else if (nat->nat_pr[0] == IPPROTO_ICMP ||
4742 				 nat->nat_pr[0] == IPPROTO_ICMPV6)
4743 				ifq2 = fin->fin_rev ? &softn->ipf_nat_icmpacktq:
4744 						      &softn->ipf_nat_icmptq;
4745 			else
4746 				ifq2 = &softn->ipf_nat_iptq;
4747 		}
4748 
4749 		ipf_movequeue(softc->ipf_ticks, tqe, ifq, ifq2);
4750 	}
4751 }
4752 
4753 
4754 /* ------------------------------------------------------------------------ */
4755 /* Function:    ipf_nat_checkout                                            */
4756 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
4757 /*                     0 == no packet translation occurred,                 */
4758 /*                     1 == packet was successfully translated.             */
4759 /* Parameters:  fin(I)   - pointer to packet information                    */
4760 /*              passp(I) - pointer to filtering result flags                */
4761 /*                                                                          */
4762 /* Check to see if an outcoming packet should be changed.  ICMP packets are */
4763 /* first checked to see if they match an existing entry (if an error),      */
4764 /* otherwise a search of the current NAT table is made.  If neither results */
4765 /* in a match then a search for a matching NAT rule is made.  Create a new  */
4766 /* NAT entry if a we matched a NAT rule.  Lastly, actually change the       */
4767 /* packet header(s) as required.                                            */
4768 /* ------------------------------------------------------------------------ */
4769 int
4770 ipf_nat_checkout(fr_info_t *fin, u_32_t *passp)
4771 {
4772 	ipnat_t *np = NULL, *npnext;
4773 	struct ifnet *ifp, *sifp;
4774 	ipf_main_softc_t *softc;
4775 	ipf_nat_softc_t *softn;
4776 	tcphdr_t *tcp = NULL;
4777 	int rval, natfailed;
4778 	u_int nflags = 0;
4779 	u_32_t ipa, iph;
4780 	int natadd = 1;
4781 	frentry_t *fr;
4782 	nat_t *nat;
4783 
4784 	if (fin->fin_v == 6) {
4785 #ifdef USE_INET6
4786 		return ipf_nat6_checkout(fin, passp);
4787 #else
4788 		return 0;
4789 #endif
4790 	}
4791 
4792 	softc = fin->fin_main_soft;
4793 	softn = softc->ipf_nat_soft;
4794 
4795 	if (softn->ipf_nat_lock != 0)
4796 		return 0;
4797 	if (softn->ipf_nat_stats.ns_rules == 0 &&
4798 	    softn->ipf_nat_instances == NULL)
4799 		return 0;
4800 
4801 	natfailed = 0;
4802 	fr = fin->fin_fr;
4803 	sifp = fin->fin_ifp;
4804 	if (fr != NULL) {
4805 		ifp = fr->fr_tifs[fin->fin_rev].fd_ptr;
4806 		if ((ifp != NULL) && (ifp != (void *)-1))
4807 			fin->fin_ifp = ifp;
4808 	}
4809 	ifp = fin->fin_ifp;
4810 
4811 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
4812 		switch (fin->fin_p)
4813 		{
4814 		case IPPROTO_TCP :
4815 			nflags = IPN_TCP;
4816 			break;
4817 		case IPPROTO_UDP :
4818 			nflags = IPN_UDP;
4819 			break;
4820 		case IPPROTO_ICMP :
4821 
4822 			/*
4823 			 * This is an incoming packet, so the destination is
4824 			 * the icmp_id and the source port equals 0
4825 			 */
4826 			if ((fin->fin_flx & FI_ICMPQUERY) != 0)
4827 				nflags = IPN_ICMPQUERY;
4828 			break;
4829 		default :
4830 			break;
4831 		}
4832 
4833 		if ((nflags & IPN_TCPUDP))
4834 			tcp = fin->fin_dp;
4835 	}
4836 
4837 	ipa = fin->fin_saddr;
4838 
4839 	READ_ENTER(&softc->ipf_nat);
4840 
4841 	if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
4842 	    (nat = ipf_nat_icmperror(fin, &nflags, NAT_OUTBOUND)))
4843 		/*EMPTY*/;
4844 	else if ((fin->fin_flx & FI_FRAG) && (nat = ipf_frag_natknown(fin)))
4845 		natadd = 0;
4846 	else if ((nat = ipf_nat_outlookup(fin, nflags|NAT_SEARCH,
4847 				      (u_int)fin->fin_p, fin->fin_src,
4848 				      fin->fin_dst))) {
4849 		nflags = nat->nat_flags;
4850 	} else if (fin->fin_off == 0) {
4851 		u_32_t hv, msk, nmsk = 0;
4852 
4853 		/*
4854 		 * If there is no current entry in the nat table for this IP#,
4855 		 * create one for it (if there is a matching rule).
4856 		 */
4857 maskloop:
4858 		msk = softn->ipf_nat_map_active_masks[nmsk];
4859 		iph = ipa & msk;
4860 		hv = NAT_HASH_FN(iph, 0, softn->ipf_nat_maprules_sz);
4861 retry_roundrobin:
4862 		for (np = softn->ipf_nat_map_rules[hv]; np; np = npnext) {
4863 			npnext = np->in_mnext;
4864 			if ((np->in_ifps[1] && (np->in_ifps[1] != ifp)))
4865 				continue;
4866 			if (np->in_v[0] != 4)
4867 				continue;
4868 			if (np->in_pr[1] && (np->in_pr[1] != fin->fin_p))
4869 				continue;
4870 			if ((np->in_flags & IPN_RF) &&
4871 			    !(np->in_flags & nflags))
4872 				continue;
4873 			if (np->in_flags & IPN_FILTER) {
4874 				switch (ipf_nat_match(fin, np))
4875 				{
4876 				case 0 :
4877 					continue;
4878 				case -1 :
4879 					rval = -1;
4880 					goto outmatchfail;
4881 				case 1 :
4882 				default :
4883 					break;
4884 				}
4885 			} else if ((ipa & np->in_osrcmsk) != np->in_osrcaddr)
4886 				continue;
4887 
4888 			if ((fr != NULL) &&
4889 			    !ipf_matchtag(&np->in_tag, &fr->fr_nattag))
4890 				continue;
4891 
4892 			if (np->in_plabel != -1) {
4893 				if (((np->in_flags & IPN_FILTER) == 0) &&
4894 				    (np->in_odport != fin->fin_data[1]))
4895 					continue;
4896 				if (ipf_proxy_ok(fin, tcp, np) == 0)
4897 					continue;
4898 			}
4899 
4900 			if (np->in_flags & IPN_NO) {
4901 				np->in_hits++;
4902 				break;
4903 			}
4904 			MUTEX_ENTER(&softn->ipf_nat_new);
4905 			/*
4906 			 * If we've matched a round-robin rule but it has
4907 			 * moved in the list since we got it, start over as
4908 			 * this is now no longer correct.
4909 			 */
4910 			if (npnext != np->in_mnext) {
4911 				if ((np->in_flags & IPN_ROUNDR) != 0) {
4912 					MUTEX_EXIT(&softn->ipf_nat_new);
4913 					goto retry_roundrobin;
4914 				}
4915 				npnext = np->in_mnext;
4916 			}
4917 
4918 			nat = ipf_nat_add(fin, np, NULL, nflags, NAT_OUTBOUND);
4919 			MUTEX_EXIT(&softn->ipf_nat_new);
4920 			if (nat != NULL) {
4921 				natfailed = 0;
4922 				break;
4923 			}
4924 			natfailed = -1;
4925 		}
4926 		if ((np == NULL) && (nmsk < softn->ipf_nat_map_max)) {
4927 			nmsk++;
4928 			goto maskloop;
4929 		}
4930 	}
4931 
4932 	if (nat != NULL) {
4933 		rval = ipf_nat_out(fin, nat, natadd, nflags);
4934 		if (rval == 1) {
4935 			MUTEX_ENTER(&nat->nat_lock);
4936 			ipf_nat_update(fin, nat);
4937 			nat->nat_bytes[1] += fin->fin_plen;
4938 			nat->nat_pkts[1]++;
4939 			fin->fin_pktnum = nat->nat_pkts[1];
4940 			MUTEX_EXIT(&nat->nat_lock);
4941 		}
4942 	} else
4943 		rval = natfailed;
4944 outmatchfail:
4945 	RWLOCK_EXIT(&softc->ipf_nat);
4946 
4947 	switch (rval)
4948 	{
4949 	case -1 :
4950 		if (passp != NULL) {
4951 			DT1(frb_natv4out, fr_info_t *, fin);
4952 			NBUMPSIDED(1, ns_drop);
4953 			*passp = FR_BLOCK;
4954 			fin->fin_reason = FRB_NATV4;
4955 		}
4956 		fin->fin_flx |= FI_BADNAT;
4957 		NBUMPSIDED(1, ns_badnat);
4958 		break;
4959 	case 0 :
4960 		NBUMPSIDE(1, ns_ignored);
4961 		break;
4962 	case 1 :
4963 		NBUMPSIDE(1, ns_translated);
4964 		break;
4965 	}
4966 	fin->fin_ifp = sifp;
4967 	return rval;
4968 }
4969 
4970 /* ------------------------------------------------------------------------ */
4971 /* Function:    ipf_nat_out                                                 */
4972 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
4973 /*                     1 == packet was successfully translated.             */
4974 /* Parameters:  fin(I)    - pointer to packet information                   */
4975 /*              nat(I)    - pointer to NAT structure                        */
4976 /*              natadd(I) - flag indicating if it is safe to add frag cache */
4977 /*              nflags(I) - NAT flags set for this packet                   */
4978 /*                                                                          */
4979 /* Translate a packet coming "out" on an interface.                         */
4980 /* ------------------------------------------------------------------------ */
4981 int
4982 ipf_nat_out(fr_info_t *fin, nat_t *nat, int natadd, u_32_t nflags)
4983 {
4984 	ipf_main_softc_t *softc = fin->fin_main_soft;
4985 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4986 	icmphdr_t *icmp;
4987 	tcphdr_t *tcp;
4988 	ipnat_t *np;
4989 	int skip;
4990 	int i;
4991 
4992 	tcp = NULL;
4993 	icmp = NULL;
4994 	np = nat->nat_ptr;
4995 
4996 	if ((natadd != 0) && (fin->fin_flx & FI_FRAG) && (np != NULL))
4997 		(void) ipf_frag_natnew(softc, fin, 0, nat);
4998 
4999 	/*
5000 	 * Fix up checksums, not by recalculating them, but
5001 	 * simply computing adjustments.
5002 	 * This is only done for STREAMS based IP implementations where the
5003 	 * checksum has already been calculated by IP.  In all other cases,
5004 	 * IPFilter is called before the checksum needs calculating so there
5005 	 * is no call to modify whatever is in the header now.
5006 	 */
5007 	if (nflags == IPN_ICMPERR) {
5008 		u_32_t s1, s2, sumd, msumd;
5009 
5010 		s1 = LONG_SUM(ntohl(fin->fin_saddr));
5011 		if (nat->nat_dir == NAT_OUTBOUND) {
5012 			s2 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
5013 		} else {
5014 			s2 = LONG_SUM(ntohl(nat->nat_odstaddr));
5015 		}
5016 		CALC_SUMD(s1, s2, sumd);
5017 		msumd = sumd;
5018 
5019 		s1 = LONG_SUM(ntohl(fin->fin_daddr));
5020 		if (nat->nat_dir == NAT_OUTBOUND) {
5021 			s2 = LONG_SUM(ntohl(nat->nat_ndstaddr));
5022 		} else {
5023 			s2 = LONG_SUM(ntohl(nat->nat_osrcaddr));
5024 		}
5025 		CALC_SUMD(s1, s2, sumd);
5026 		msumd += sumd;
5027 
5028 		ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, msumd, 0);
5029 	}
5030 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5031     defined(linux) || defined(BRIDGE_IPF)
5032 	else {
5033 		/*
5034 		 * Strictly speaking, this isn't necessary on BSD
5035 		 * kernels because they do checksum calculation after
5036 		 * this code has run BUT if ipfilter is being used
5037 		 * to do NAT as a bridge, that code doesn't exist.
5038 		 */
5039 		switch (nat->nat_dir)
5040 		{
5041 		case NAT_OUTBOUND :
5042 			ipf_fix_outcksum(fin->fin_cksum & FI_CK_L4PART,
5043 					 &fin->fin_ip->ip_sum,
5044 					 nat->nat_ipsumd, 0);
5045 			break;
5046 
5047 		case NAT_INBOUND :
5048 			ipf_fix_incksum(fin->fin_cksum & FI_CK_L4PART,
5049 					&fin->fin_ip->ip_sum,
5050 					nat->nat_ipsumd, 0);
5051 			break;
5052 
5053 		default :
5054 			break;
5055 		}
5056 	}
5057 #endif
5058 
5059 	/*
5060 	 * Address assignment is after the checksum modification because
5061 	 * we are using the address in the packet for determining the
5062 	 * correct checksum offset (the ICMP error could be coming from
5063 	 * anyone...)
5064 	 */
5065 	switch (nat->nat_dir)
5066 	{
5067 	case NAT_OUTBOUND :
5068 		fin->fin_ip->ip_src = nat->nat_nsrcip;
5069 		fin->fin_saddr = nat->nat_nsrcaddr;
5070 		fin->fin_ip->ip_dst = nat->nat_ndstip;
5071 		fin->fin_daddr = nat->nat_ndstaddr;
5072 		break;
5073 
5074 	case NAT_INBOUND :
5075 		fin->fin_ip->ip_src = nat->nat_odstip;
5076 		fin->fin_saddr = nat->nat_ndstaddr;
5077 		fin->fin_ip->ip_dst = nat->nat_osrcip;
5078 		fin->fin_daddr = nat->nat_nsrcaddr;
5079 		break;
5080 
5081 	case NAT_DIVERTIN :
5082 	    {
5083 		mb_t *m;
5084 
5085 		skip = ipf_nat_decap(fin, nat);
5086 		if (skip <= 0) {
5087 			NBUMPSIDED(1, ns_decap_fail);
5088 			return -1;
5089 		}
5090 
5091 		m = fin->fin_m;
5092 
5093 #if defined(MENTAT) && defined(_KERNEL)
5094 		m->b_rptr += skip;
5095 #else
5096 		m->m_data += skip;
5097 		m->m_len -= skip;
5098 
5099 # ifdef M_PKTHDR
5100 		if (m->m_flags & M_PKTHDR)
5101 			m->m_pkthdr.len -= skip;
5102 # endif
5103 #endif
5104 
5105 		MUTEX_ENTER(&nat->nat_lock);
5106 		ipf_nat_update(fin, nat);
5107 		MUTEX_EXIT(&nat->nat_lock);
5108 		fin->fin_flx |= FI_NATED;
5109 		if (np != NULL && np->in_tag.ipt_num[0] != 0)
5110 			fin->fin_nattag = &np->in_tag;
5111 		return 1;
5112 		/* NOTREACHED */
5113 	    }
5114 
5115 	case NAT_DIVERTOUT :
5116 	    {
5117 		u_32_t s1, s2, sumd;
5118 		udphdr_t *uh;
5119 		ip_t *ip;
5120 		mb_t *m;
5121 
5122 		m = M_DUP(np->in_divmp);
5123 		if (m == NULL) {
5124 			NBUMPSIDED(1, ns_divert_dup);
5125 			return -1;
5126 		}
5127 
5128 		ip = MTOD(m, ip_t *);
5129 		ip->ip_id = htons(ipf_nextipid(fin));
5130 		s2 = ntohs(ip->ip_id);
5131 
5132 		s1 = ip->ip_len;
5133 		ip->ip_len = ntohs(ip->ip_len);
5134 		ip->ip_len += fin->fin_plen;
5135 		ip->ip_len = htons(ip->ip_len);
5136 		s2 += ntohs(ip->ip_len);
5137 		CALC_SUMD(s1, s2, sumd);
5138 
5139 		uh = (udphdr_t *)(ip + 1);
5140 		uh->uh_ulen += fin->fin_plen;
5141 		uh->uh_ulen = htons(uh->uh_ulen);
5142 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5143     defined(linux) || defined(BRIDGE_IPF)
5144 		ipf_fix_outcksum(0, &ip->ip_sum, sumd, 0);
5145 #endif
5146 
5147 		PREP_MB_T(fin, m);
5148 
5149 		fin->fin_src = ip->ip_src;
5150 		fin->fin_dst = ip->ip_dst;
5151 		fin->fin_ip = ip;
5152 		fin->fin_plen += sizeof(ip_t) + 8;	/* UDP + IPv4 hdr */
5153 		fin->fin_dlen += sizeof(ip_t) + 8;	/* UDP + IPv4 hdr */
5154 
5155 		nflags &= ~IPN_TCPUDPICMP;
5156 
5157 		break;
5158 	    }
5159 
5160 	default :
5161 		break;
5162 	}
5163 
5164 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5165 		u_short *csump;
5166 
5167 		if ((nat->nat_nsport != 0) && (nflags & IPN_TCPUDP)) {
5168 			tcp = fin->fin_dp;
5169 
5170 			switch (nat->nat_dir)
5171 			{
5172 			case NAT_OUTBOUND :
5173 				tcp->th_sport = nat->nat_nsport;
5174 				fin->fin_data[0] = ntohs(nat->nat_nsport);
5175 				tcp->th_dport = nat->nat_ndport;
5176 				fin->fin_data[1] = ntohs(nat->nat_ndport);
5177 				break;
5178 
5179 			case NAT_INBOUND :
5180 				tcp->th_sport = nat->nat_odport;
5181 				fin->fin_data[0] = ntohs(nat->nat_odport);
5182 				tcp->th_dport = nat->nat_osport;
5183 				fin->fin_data[1] = ntohs(nat->nat_osport);
5184 				break;
5185 			}
5186 		}
5187 
5188 		if ((nat->nat_nsport != 0) && (nflags & IPN_ICMPQUERY)) {
5189 			icmp = fin->fin_dp;
5190 			icmp->icmp_id = nat->nat_nicmpid;
5191 		}
5192 
5193 		csump = ipf_nat_proto(fin, nat, nflags);
5194 
5195 		/*
5196 		 * The above comments do not hold for layer 4 (or higher)
5197 		 * checksums...
5198 		 */
5199 		if (csump != NULL) {
5200 			if (nat->nat_dir == NAT_OUTBOUND)
5201 				ipf_fix_outcksum(fin->fin_cksum, csump,
5202 						 nat->nat_sumd[0],
5203 						 nat->nat_sumd[1] +
5204 						 fin->fin_dlen);
5205 			else
5206 				ipf_fix_incksum(fin->fin_cksum, csump,
5207 						nat->nat_sumd[0],
5208 						nat->nat_sumd[1] +
5209 						fin->fin_dlen);
5210 		}
5211 	}
5212 
5213 	ipf_sync_update(softc, SMC_NAT, fin, nat->nat_sync);
5214 	/* ------------------------------------------------------------- */
5215 	/* A few quick notes:                                            */
5216 	/*      Following are test conditions prior to calling the       */
5217 	/*      ipf_proxy_check routine.                                 */
5218 	/*                                                               */
5219 	/*      A NULL tcp indicates a non TCP/UDP packet.  When dealing */
5220 	/*      with a redirect rule, we attempt to match the packet's   */
5221 	/*      source port against in_dport, otherwise we'd compare the */
5222 	/*      packet's destination.                                    */
5223 	/* ------------------------------------------------------------- */
5224 	if ((np != NULL) && (np->in_apr != NULL)) {
5225 		i = ipf_proxy_check(fin, nat);
5226 		if (i == 0) {
5227 			i = 1;
5228 		} else if (i == -1) {
5229 			NBUMPSIDED(1, ns_ipf_proxy_fail);
5230 		}
5231 	} else {
5232 		i = 1;
5233 	}
5234 	fin->fin_flx |= FI_NATED;
5235 	return i;
5236 }
5237 
5238 
5239 /* ------------------------------------------------------------------------ */
5240 /* Function:    ipf_nat_checkin                                             */
5241 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
5242 /*                     0 == no packet translation occurred,                 */
5243 /*                     1 == packet was successfully translated.             */
5244 /* Parameters:  fin(I)   - pointer to packet information                    */
5245 /*              passp(I) - pointer to filtering result flags                */
5246 /*                                                                          */
5247 /* Check to see if an incoming packet should be changed.  ICMP packets are  */
5248 /* first checked to see if they match an existing entry (if an error),      */
5249 /* otherwise a search of the current NAT table is made.  If neither results */
5250 /* in a match then a search for a matching NAT rule is made.  Create a new  */
5251 /* NAT entry if a we matched a NAT rule.  Lastly, actually change the       */
5252 /* packet header(s) as required.                                            */
5253 /* ------------------------------------------------------------------------ */
5254 int
5255 ipf_nat_checkin(fr_info_t *fin, u_32_t *passp)
5256 {
5257 	ipf_main_softc_t *softc;
5258 	ipf_nat_softc_t *softn;
5259 	u_int nflags, natadd;
5260 	ipnat_t *np, *npnext;
5261 	int rval, natfailed;
5262 	struct ifnet *ifp;
5263 	struct in_addr in;
5264 	icmphdr_t *icmp;
5265 	tcphdr_t *tcp;
5266 	u_short dport;
5267 	nat_t *nat;
5268 	u_32_t iph;
5269 
5270 	softc = fin->fin_main_soft;
5271 	softn = softc->ipf_nat_soft;
5272 
5273 	if (softn->ipf_nat_lock != 0)
5274 		return 0;
5275 	if (softn->ipf_nat_stats.ns_rules == 0 &&
5276 	    softn->ipf_nat_instances == NULL)
5277 		return 0;
5278 
5279 	tcp = NULL;
5280 	icmp = NULL;
5281 	dport = 0;
5282 	natadd = 1;
5283 	nflags = 0;
5284 	natfailed = 0;
5285 	ifp = fin->fin_ifp;
5286 
5287 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5288 		switch (fin->fin_p)
5289 		{
5290 		case IPPROTO_TCP :
5291 			nflags = IPN_TCP;
5292 			break;
5293 		case IPPROTO_UDP :
5294 			nflags = IPN_UDP;
5295 			break;
5296 		case IPPROTO_ICMP :
5297 			icmp = fin->fin_dp;
5298 
5299 			/*
5300 			 * This is an incoming packet, so the destination is
5301 			 * the icmp_id and the source port equals 0
5302 			 */
5303 			if ((fin->fin_flx & FI_ICMPQUERY) != 0) {
5304 				nflags = IPN_ICMPQUERY;
5305 				dport = icmp->icmp_id;
5306 			} break;
5307 		default :
5308 			break;
5309 		}
5310 
5311 		if ((nflags & IPN_TCPUDP)) {
5312 			tcp = fin->fin_dp;
5313 			dport = fin->fin_data[1];
5314 		}
5315 	}
5316 
5317 	in = fin->fin_dst;
5318 
5319 	READ_ENTER(&softc->ipf_nat);
5320 
5321 	if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
5322 	    (nat = ipf_nat_icmperror(fin, &nflags, NAT_INBOUND)))
5323 		/*EMPTY*/;
5324 	else if ((fin->fin_flx & FI_FRAG) && (nat = ipf_frag_natknown(fin)))
5325 		natadd = 0;
5326 	else if ((nat = ipf_nat_inlookup(fin, nflags|NAT_SEARCH,
5327 					 (u_int)fin->fin_p,
5328 					 fin->fin_src, in))) {
5329 		nflags = nat->nat_flags;
5330 	} else if (fin->fin_off == 0) {
5331 		u_32_t hv, msk, rmsk = 0;
5332 
5333 		/*
5334 		 * If there is no current entry in the nat table for this IP#,
5335 		 * create one for it (if there is a matching rule).
5336 		 */
5337 maskloop:
5338 		msk = softn->ipf_nat_rdr_active_masks[rmsk];
5339 		iph = in.s_addr & msk;
5340 		hv = NAT_HASH_FN(iph, 0, softn->ipf_nat_rdrrules_sz);
5341 retry_roundrobin:
5342 		/* TRACE (iph,msk,rmsk,hv,softn->ipf_nat_rdrrules_sz) */
5343 		for (np = softn->ipf_nat_rdr_rules[hv]; np; np = npnext) {
5344 			npnext = np->in_rnext;
5345 			if (np->in_ifps[0] && (np->in_ifps[0] != ifp))
5346 				continue;
5347 			if (np->in_v[0] != 4)
5348 				continue;
5349 			if (np->in_pr[0] && (np->in_pr[0] != fin->fin_p))
5350 				continue;
5351 			if ((np->in_flags & IPN_RF) && !(np->in_flags & nflags))
5352 				continue;
5353 			if (np->in_flags & IPN_FILTER) {
5354 				switch (ipf_nat_match(fin, np))
5355 				{
5356 				case 0 :
5357 					continue;
5358 				case -1 :
5359 					rval = -1;
5360 					goto inmatchfail;
5361 				case 1 :
5362 				default :
5363 					break;
5364 				}
5365 			} else {
5366 				if ((in.s_addr & np->in_odstmsk) !=
5367 				    np->in_odstaddr)
5368 					continue;
5369 				if (np->in_odport &&
5370 				    ((np->in_dtop < dport) ||
5371 				     (dport < np->in_odport)))
5372 					continue;
5373 			}
5374 
5375 			if (np->in_plabel != -1) {
5376 				if (!ipf_proxy_ok(fin, tcp, np)) {
5377 					continue;
5378 				}
5379 			}
5380 
5381 			if (np->in_flags & IPN_NO) {
5382 				np->in_hits++;
5383 				break;
5384 			}
5385 
5386 			MUTEX_ENTER(&softn->ipf_nat_new);
5387 			/*
5388 			 * If we've matched a round-robin rule but it has
5389 			 * moved in the list since we got it, start over as
5390 			 * this is now no longer correct.
5391 			 */
5392 			if (npnext != np->in_rnext) {
5393 				if ((np->in_flags & IPN_ROUNDR) != 0) {
5394 					MUTEX_EXIT(&softn->ipf_nat_new);
5395 					goto retry_roundrobin;
5396 				}
5397 				npnext = np->in_rnext;
5398 			}
5399 
5400 			nat = ipf_nat_add(fin, np, NULL, nflags, NAT_INBOUND);
5401 			MUTEX_EXIT(&softn->ipf_nat_new);
5402 			if (nat != NULL) {
5403 				natfailed = 0;
5404 				break;
5405 			}
5406 			natfailed = -1;
5407 		}
5408 		if ((np == NULL) && (rmsk < softn->ipf_nat_rdr_max)) {
5409 			rmsk++;
5410 			goto maskloop;
5411 		}
5412 	}
5413 
5414 	if (nat != NULL) {
5415 		rval = ipf_nat_in(fin, nat, natadd, nflags);
5416 		if (rval == 1) {
5417 			MUTEX_ENTER(&nat->nat_lock);
5418 			ipf_nat_update(fin, nat);
5419 			nat->nat_bytes[0] += fin->fin_plen;
5420 			nat->nat_pkts[0]++;
5421 			fin->fin_pktnum = nat->nat_pkts[0];
5422 			MUTEX_EXIT(&nat->nat_lock);
5423 		}
5424 	} else
5425 		rval = natfailed;
5426 inmatchfail:
5427 	RWLOCK_EXIT(&softc->ipf_nat);
5428 
5429 	switch (rval)
5430 	{
5431 	case -1 :
5432 		if (passp != NULL) {
5433 			DT1(frb_natv4in, fr_info_t *, fin);
5434 			NBUMPSIDED(0, ns_drop);
5435 			*passp = FR_BLOCK;
5436 			fin->fin_reason = FRB_NATV4;
5437 		}
5438 		fin->fin_flx |= FI_BADNAT;
5439 		NBUMPSIDED(0, ns_badnat);
5440 		break;
5441 	case 0 :
5442 		NBUMPSIDE(0, ns_ignored);
5443 		break;
5444 	case 1 :
5445 		NBUMPSIDE(0, ns_translated);
5446 		break;
5447 	}
5448 	return rval;
5449 }
5450 
5451 
5452 /* ------------------------------------------------------------------------ */
5453 /* Function:    ipf_nat_in                                                  */
5454 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
5455 /*                     1 == packet was successfully translated.             */
5456 /* Parameters:  fin(I)    - pointer to packet information                   */
5457 /*              nat(I)    - pointer to NAT structure                        */
5458 /*              natadd(I) - flag indicating if it is safe to add frag cache */
5459 /*              nflags(I) - NAT flags set for this packet                   */
5460 /* Locks Held:  ipf_nat(READ)                                               */
5461 /*                                                                          */
5462 /* Translate a packet coming "in" on an interface.                          */
5463 /* ------------------------------------------------------------------------ */
5464 int
5465 ipf_nat_in(fr_info_t *fin, nat_t *nat, int natadd, u_32_t nflags)
5466 {
5467 	ipf_main_softc_t *softc = fin->fin_main_soft;
5468 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5469 	u_32_t sumd, ipsumd, sum1, sum2;
5470 	icmphdr_t *icmp;
5471 	tcphdr_t *tcp;
5472 	ipnat_t *np;
5473 	int skip;
5474 	int i;
5475 
5476 	tcp = NULL;
5477 	np = nat->nat_ptr;
5478 	fin->fin_fr = nat->nat_fr;
5479 
5480 	if (np != NULL) {
5481 		if ((natadd != 0) && (fin->fin_flx & FI_FRAG))
5482 			(void) ipf_frag_natnew(softc, fin, 0, nat);
5483 
5484 	/* ------------------------------------------------------------- */
5485 	/* A few quick notes:                                            */
5486 	/*      Following are test conditions prior to calling the       */
5487 	/*      ipf_proxy_check routine.                                 */
5488 	/*                                                               */
5489 	/*      A NULL tcp indicates a non TCP/UDP packet.  When dealing */
5490 	/*      with a map rule, we attempt to match the packet's        */
5491 	/*      source port against in_dport, otherwise we'd compare the */
5492 	/*      packet's destination.                                    */
5493 	/* ------------------------------------------------------------- */
5494 		if (np->in_apr != NULL) {
5495 			i = ipf_proxy_check(fin, nat);
5496 			if (i == -1) {
5497 				NBUMPSIDED(0, ns_ipf_proxy_fail);
5498 				return -1;
5499 			}
5500 		}
5501 	}
5502 
5503 	ipf_sync_update(softc, SMC_NAT, fin, nat->nat_sync);
5504 
5505 	ipsumd = nat->nat_ipsumd;
5506 	/*
5507 	 * Fix up checksums, not by recalculating them, but
5508 	 * simply computing adjustments.
5509 	 * Why only do this for some platforms on inbound packets ?
5510 	 * Because for those that it is done, IP processing is yet to happen
5511 	 * and so the IPv4 header checksum has not yet been evaluated.
5512 	 * Perhaps it should always be done for the benefit of things like
5513 	 * fast forwarding (so that it doesn't need to be recomputed) but with
5514 	 * header checksum offloading, perhaps it is a moot point.
5515 	 */
5516 
5517 	switch (nat->nat_dir)
5518 	{
5519 	case NAT_INBOUND :
5520 		if ((fin->fin_flx & FI_ICMPERR) == 0) {
5521 			fin->fin_ip->ip_src = nat->nat_nsrcip;
5522 			fin->fin_saddr = nat->nat_nsrcaddr;
5523 		} else {
5524 			sum1 = nat->nat_osrcaddr;
5525 			sum2 = nat->nat_nsrcaddr;
5526 			CALC_SUMD(sum1, sum2, sumd);
5527 			ipsumd -= sumd;
5528 		}
5529 		fin->fin_ip->ip_dst = nat->nat_ndstip;
5530 		fin->fin_daddr = nat->nat_ndstaddr;
5531 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5532      defined(__osf__) || defined(linux)
5533 		ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, ipsumd, 0);
5534 #endif
5535 		break;
5536 
5537 	case NAT_OUTBOUND :
5538 		if ((fin->fin_flx & FI_ICMPERR) == 0) {
5539 			fin->fin_ip->ip_src = nat->nat_odstip;
5540 			fin->fin_saddr = nat->nat_odstaddr;
5541 		} else {
5542 			sum1 = nat->nat_odstaddr;
5543 			sum2 = nat->nat_ndstaddr;
5544 			CALC_SUMD(sum1, sum2, sumd);
5545 			ipsumd -= sumd;
5546 		}
5547 		fin->fin_ip->ip_dst = nat->nat_osrcip;
5548 		fin->fin_daddr = nat->nat_osrcaddr;
5549 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5550      defined(__osf__) || defined(linux)
5551 		ipf_fix_incksum(0, &fin->fin_ip->ip_sum, ipsumd, 0);
5552 #endif
5553 		break;
5554 
5555 	case NAT_DIVERTIN :
5556 	    {
5557 		udphdr_t *uh;
5558 		ip_t *ip;
5559 		mb_t *m;
5560 
5561 		m = M_DUP(np->in_divmp);
5562 		if (m == NULL) {
5563 			NBUMPSIDED(0, ns_divert_dup);
5564 			return -1;
5565 		}
5566 
5567 		ip = MTOD(m, ip_t *);
5568 		ip->ip_id = htons(ipf_nextipid(fin));
5569 		sum1 = ntohs(ip->ip_len);
5570 		ip->ip_len = ntohs(ip->ip_len);
5571 		ip->ip_len += fin->fin_plen;
5572 		ip->ip_len = htons(ip->ip_len);
5573 
5574 		uh = (udphdr_t *)(ip + 1);
5575 		uh->uh_ulen += fin->fin_plen;
5576 		uh->uh_ulen = htons(uh->uh_ulen);
5577 
5578 		sum2 = ntohs(ip->ip_id) + ntohs(ip->ip_len);
5579 		sum2 += ntohs(ip->ip_off) & IP_DF;
5580 		CALC_SUMD(sum1, sum2, sumd);
5581 
5582 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5583      defined(__osf__) || defined(linux)
5584 		ipf_fix_outcksum(0, &ip->ip_sum, sumd, 0);
5585 #endif
5586 		PREP_MB_T(fin, m);
5587 
5588 		fin->fin_ip = ip;
5589 		fin->fin_plen += sizeof(ip_t) + 8;	/* UDP + new IPv4 hdr */
5590 		fin->fin_dlen += sizeof(ip_t) + 8;	/* UDP + old IPv4 hdr */
5591 
5592 		nflags &= ~IPN_TCPUDPICMP;
5593 
5594 		break;
5595 	    }
5596 
5597 	case NAT_DIVERTOUT :
5598 	    {
5599 		mb_t *m;
5600 
5601 		skip = ipf_nat_decap(fin, nat);
5602 		if (skip <= 0) {
5603 			NBUMPSIDED(0, ns_decap_fail);
5604 			return -1;
5605 		}
5606 
5607 		m = fin->fin_m;
5608 
5609 #if defined(MENTAT) && defined(_KERNEL)
5610 		m->b_rptr += skip;
5611 #else
5612 		m->m_data += skip;
5613 		m->m_len -= skip;
5614 
5615 # ifdef M_PKTHDR
5616 		if (m->m_flags & M_PKTHDR)
5617 			m->m_pkthdr.len -= skip;
5618 # endif
5619 #endif
5620 
5621 		ipf_nat_update(fin, nat);
5622 		nflags &= ~IPN_TCPUDPICMP;
5623 		fin->fin_flx |= FI_NATED;
5624 		if (np != NULL && np->in_tag.ipt_num[0] != 0)
5625 			fin->fin_nattag = &np->in_tag;
5626 		return 1;
5627 		/* NOTREACHED */
5628 	    }
5629 	}
5630 	if (nflags & IPN_TCPUDP)
5631 		tcp = fin->fin_dp;
5632 
5633 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5634 		u_short *csump;
5635 
5636 		if ((nat->nat_odport != 0) && (nflags & IPN_TCPUDP)) {
5637 			switch (nat->nat_dir)
5638 			{
5639 			case NAT_INBOUND :
5640 				tcp->th_sport = nat->nat_nsport;
5641 				fin->fin_data[0] = ntohs(nat->nat_nsport);
5642 				tcp->th_dport = nat->nat_ndport;
5643 				fin->fin_data[1] = ntohs(nat->nat_ndport);
5644 				break;
5645 
5646 			case NAT_OUTBOUND :
5647 				tcp->th_sport = nat->nat_odport;
5648 				fin->fin_data[0] = ntohs(nat->nat_odport);
5649 				tcp->th_dport = nat->nat_osport;
5650 				fin->fin_data[1] = ntohs(nat->nat_osport);
5651 				break;
5652 			}
5653 		}
5654 
5655 
5656 		if ((nat->nat_odport != 0) && (nflags & IPN_ICMPQUERY)) {
5657 			icmp = fin->fin_dp;
5658 
5659 			icmp->icmp_id = nat->nat_nicmpid;
5660 		}
5661 
5662 		csump = ipf_nat_proto(fin, nat, nflags);
5663 
5664 		/*
5665 		 * The above comments do not hold for layer 4 (or higher)
5666 		 * checksums...
5667 		 */
5668 		if (csump != NULL) {
5669 			if (nat->nat_dir == NAT_OUTBOUND)
5670 				ipf_fix_incksum(0, csump, nat->nat_sumd[0], 0);
5671 			else
5672 				ipf_fix_outcksum(0, csump, nat->nat_sumd[0], 0);
5673 		}
5674 	}
5675 
5676 	fin->fin_flx |= FI_NATED;
5677 	if (np != NULL && np->in_tag.ipt_num[0] != 0)
5678 		fin->fin_nattag = &np->in_tag;
5679 	return 1;
5680 }
5681 
5682 
5683 /* ------------------------------------------------------------------------ */
5684 /* Function:    ipf_nat_proto                                               */
5685 /* Returns:     u_short* - pointer to transport header checksum to update,  */
5686 /*                         NULL if the transport protocol is not recognised */
5687 /*                         as needing a checksum update.                    */
5688 /* Parameters:  fin(I)    - pointer to packet information                   */
5689 /*              nat(I)    - pointer to NAT structure                        */
5690 /*              nflags(I) - NAT flags set for this packet                   */
5691 /*                                                                          */
5692 /* Return the pointer to the checksum field for each protocol so understood.*/
5693 /* If support for making other changes to a protocol header is required,    */
5694 /* that is not strictly 'address' translation, such as clamping the MSS in  */
5695 /* TCP down to a specific value, then do it from here.                      */
5696 /* ------------------------------------------------------------------------ */
5697 u_short *
5698 ipf_nat_proto(fr_info_t *fin, nat_t *nat, u_int nflags)
5699 {
5700 	icmphdr_t *icmp;
5701 	u_short *csump;
5702 	tcphdr_t *tcp;
5703 	udphdr_t *udp;
5704 
5705 	csump = NULL;
5706 	if (fin->fin_out == 0) {
5707 		fin->fin_rev = (nat->nat_dir & NAT_OUTBOUND);
5708 	} else {
5709 		fin->fin_rev = ((nat->nat_dir & NAT_OUTBOUND) == 0);
5710 	}
5711 
5712 	switch (fin->fin_p)
5713 	{
5714 	case IPPROTO_TCP :
5715 		tcp = fin->fin_dp;
5716 
5717 		if ((nflags & IPN_TCP) != 0)
5718 			csump = &tcp->th_sum;
5719 
5720 		/*
5721 		 * Do a MSS CLAMPING on a SYN packet,
5722 		 * only deal IPv4 for now.
5723 		 */
5724 		if ((nat->nat_mssclamp != 0) && (tcp->th_flags & TH_SYN) != 0)
5725 			ipf_nat_mssclamp(tcp, nat->nat_mssclamp, fin, csump);
5726 
5727 		break;
5728 
5729 	case IPPROTO_UDP :
5730 		udp = fin->fin_dp;
5731 
5732 		if ((nflags & IPN_UDP) != 0) {
5733 			if (udp->uh_sum != 0)
5734 				csump = &udp->uh_sum;
5735 		}
5736 		break;
5737 
5738 	case IPPROTO_ICMP :
5739 		icmp = fin->fin_dp;
5740 
5741 		if ((nflags & IPN_ICMPQUERY) != 0) {
5742 			if (icmp->icmp_cksum != 0)
5743 				csump = &icmp->icmp_cksum;
5744 		}
5745 		break;
5746 
5747 #ifdef USE_INET6
5748 	case IPPROTO_ICMPV6 :
5749 	    {
5750 		struct icmp6_hdr *icmp6 = (struct icmp6_hdr *)fin->fin_dp;
5751 
5752 		icmp6 = fin->fin_dp;
5753 
5754 		if ((nflags & IPN_ICMPQUERY) != 0) {
5755 			if (icmp6->icmp6_cksum != 0)
5756 				csump = &icmp6->icmp6_cksum;
5757 		}
5758 		break;
5759 	    }
5760 #endif
5761 	}
5762 	return csump;
5763 }
5764 
5765 
5766 /* ------------------------------------------------------------------------ */
5767 /* Function:    ipf_nat_expire                                              */
5768 /* Returns:     Nil                                                         */
5769 /* Parameters:  softc(I) - pointer to soft context main structure           */
5770 /*                                                                          */
5771 /* Check all of the timeout queues for entries at the top which need to be  */
5772 /* expired.                                                                 */
5773 /* ------------------------------------------------------------------------ */
5774 void
5775 ipf_nat_expire(ipf_main_softc_t *softc)
5776 {
5777 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5778 	ipftq_t *ifq, *ifqnext;
5779 	ipftqent_t *tqe, *tqn;
5780 	int i;
5781 	SPL_INT(s);
5782 
5783 	SPL_NET(s);
5784 	WRITE_ENTER(&softc->ipf_nat);
5785 	for (ifq = softn->ipf_nat_tcptq, i = 0; ifq != NULL;
5786 	     ifq = ifq->ifq_next) {
5787 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
5788 			if (tqe->tqe_die > softc->ipf_ticks)
5789 				break;
5790 			tqn = tqe->tqe_next;
5791 			ipf_nat_delete(softc, tqe->tqe_parent, NL_EXPIRE);
5792 		}
5793 	}
5794 
5795 	for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifq->ifq_next) {
5796 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
5797 			if (tqe->tqe_die > softc->ipf_ticks)
5798 				break;
5799 			tqn = tqe->tqe_next;
5800 			ipf_nat_delete(softc, tqe->tqe_parent, NL_EXPIRE);
5801 		}
5802 	}
5803 
5804 	for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifqnext) {
5805 		ifqnext = ifq->ifq_next;
5806 
5807 		if (((ifq->ifq_flags & IFQF_DELETE) != 0) &&
5808 		    (ifq->ifq_ref == 0)) {
5809 			ipf_freetimeoutqueue(softc, ifq);
5810 		}
5811 	}
5812 
5813 	if (softn->ipf_nat_doflush != 0) {
5814 		ipf_nat_extraflush(softc, softn, 2);
5815 		softn->ipf_nat_doflush = 0;
5816 	}
5817 
5818 	RWLOCK_EXIT(&softc->ipf_nat);
5819 	SPL_X(s);
5820 }
5821 
5822 
5823 /* ------------------------------------------------------------------------ */
5824 /* Function:    ipf_nat_sync                                                */
5825 /* Returns:     Nil                                                         */
5826 /* Parameters:  softc(I) - pointer to soft context main structure           */
5827 /*              ifp(I) - pointer to network interface                       */
5828 /*                                                                          */
5829 /* Walk through all of the currently active NAT sessions, looking for those */
5830 /* which need to have their translated address updated.                     */
5831 /* ------------------------------------------------------------------------ */
5832 void
5833 ipf_nat_sync(ipf_main_softc_t *softc, void *ifp)
5834 {
5835 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5836 	u_32_t sum1, sum2, sumd;
5837 	i6addr_t in;
5838 	ipnat_t *n;
5839 	nat_t *nat;
5840 	void *ifp2;
5841 	int idx;
5842 	SPL_INT(s);
5843 
5844 	if (softc->ipf_running <= 0)
5845 		return;
5846 
5847 	/*
5848 	 * Change IP addresses for NAT sessions for any protocol except TCP
5849 	 * since it will break the TCP connection anyway.  The only rules
5850 	 * which will get changed are those which are "map ... -> 0/32",
5851 	 * where the rule specifies the address is taken from the interface.
5852 	 */
5853 	SPL_NET(s);
5854 	WRITE_ENTER(&softc->ipf_nat);
5855 
5856 	if (softc->ipf_running <= 0) {
5857 		RWLOCK_EXIT(&softc->ipf_nat);
5858 		return;
5859 	}
5860 
5861 	for (nat = softn->ipf_nat_instances; nat; nat = nat->nat_next) {
5862 		if ((nat->nat_flags & IPN_TCP) != 0)
5863 			continue;
5864 
5865 		n = nat->nat_ptr;
5866 		if (n != NULL) {
5867 			if (n->in_v[1] == 4) {
5868 				if (n->in_redir & NAT_MAP) {
5869 					if ((n->in_nsrcaddr != 0) ||
5870 					    (n->in_nsrcmsk != 0xffffffff))
5871 						continue;
5872 				} else if (n->in_redir & NAT_REDIRECT) {
5873 					if ((n->in_ndstaddr != 0) ||
5874 					    (n->in_ndstmsk != 0xffffffff))
5875 						continue;
5876 				}
5877 			}
5878 #ifdef USE_INET6
5879 			if (n->in_v[1] == 4) {
5880 				if (n->in_redir & NAT_MAP) {
5881 					if (!IP6_ISZERO(&n->in_nsrcaddr) ||
5882 					    !IP6_ISONES(&n->in_nsrcmsk))
5883 						continue;
5884 				} else if (n->in_redir & NAT_REDIRECT) {
5885 					if (!IP6_ISZERO(&n->in_ndstaddr) ||
5886 					    !IP6_ISONES(&n->in_ndstmsk))
5887 						continue;
5888 				}
5889 			}
5890 #endif
5891 		}
5892 
5893 		if (((ifp == NULL) || (ifp == nat->nat_ifps[0]) ||
5894 		     (ifp == nat->nat_ifps[1]))) {
5895 			nat->nat_ifps[0] = GETIFP(nat->nat_ifnames[0],
5896 						  nat->nat_v[0]);
5897 			if ((nat->nat_ifps[0] != NULL) &&
5898 			    (nat->nat_ifps[0] != (void *)-1)) {
5899 				nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
5900 			}
5901 			if (nat->nat_ifnames[1][0] != '\0') {
5902 				nat->nat_ifps[1] = GETIFP(nat->nat_ifnames[1],
5903 							  nat->nat_v[1]);
5904 			} else {
5905 				nat->nat_ifps[1] = nat->nat_ifps[0];
5906 			}
5907 			if ((nat->nat_ifps[1] != NULL) &&
5908 			    (nat->nat_ifps[1] != (void *)-1)) {
5909 				nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
5910 			}
5911 			ifp2 = nat->nat_ifps[0];
5912 			if (ifp2 == NULL)
5913 				continue;
5914 
5915 			/*
5916 			 * Change the map-to address to be the same as the
5917 			 * new one.
5918 			 */
5919 			sum1 = NATFSUM(nat, nat->nat_v[1], nat_nsrc6);
5920 			if (ipf_ifpaddr(softc, nat->nat_v[0], FRI_NORMAL, ifp2,
5921 				       &in, NULL) != -1) {
5922 				if (nat->nat_v[0] == 4)
5923 					nat->nat_nsrcip = in.in4;
5924 			}
5925 			sum2 = NATFSUM(nat, nat->nat_v[1], nat_nsrc6);
5926 
5927 			if (sum1 == sum2)
5928 				continue;
5929 			/*
5930 			 * Readjust the checksum adjustment to take into
5931 			 * account the new IP#.
5932 			 */
5933 			CALC_SUMD(sum1, sum2, sumd);
5934 			/* XXX - dont change for TCP when solaris does
5935 			 * hardware checksumming.
5936 			 */
5937 			sumd += nat->nat_sumd[0];
5938 			nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
5939 			nat->nat_sumd[1] = nat->nat_sumd[0];
5940 		}
5941 	}
5942 
5943 	for (n = softn->ipf_nat_list; (n != NULL); n = n->in_next) {
5944 		char *base = n->in_names;
5945 
5946 		if ((ifp == NULL) || (n->in_ifps[0] == ifp))
5947 			n->in_ifps[0] = ipf_resolvenic(softc,
5948 						       base + n->in_ifnames[0],
5949 						       n->in_v[0]);
5950 		if ((ifp == NULL) || (n->in_ifps[1] == ifp))
5951 			n->in_ifps[1] = ipf_resolvenic(softc,
5952 						       base + n->in_ifnames[1],
5953 						       n->in_v[1]);
5954 
5955 		if (n->in_redir & NAT_REDIRECT)
5956 			idx = 1;
5957 		else
5958 			idx = 0;
5959 
5960 		if (((ifp == NULL) || (n->in_ifps[idx] == ifp)) &&
5961 		    (n->in_ifps[idx] != NULL &&
5962 		     n->in_ifps[idx] != (void *)-1)) {
5963 
5964 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_osrc,
5965 					     0, n->in_ifps[idx]);
5966 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_odst,
5967 					     0, n->in_ifps[idx]);
5968 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_nsrc,
5969 					     0, n->in_ifps[idx]);
5970 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_ndst,
5971 					     0, n->in_ifps[idx]);
5972 		}
5973 	}
5974 	RWLOCK_EXIT(&softc->ipf_nat);
5975 	SPL_X(s);
5976 }
5977 
5978 
5979 /* ------------------------------------------------------------------------ */
5980 /* Function:    ipf_nat_icmpquerytype                                       */
5981 /* Returns:     int - 1 == success, 0 == failure                            */
5982 /* Parameters:  icmptype(I) - ICMP type number                              */
5983 /*                                                                          */
5984 /* Tests to see if the ICMP type number passed is a query/response type or  */
5985 /* not.                                                                     */
5986 /* ------------------------------------------------------------------------ */
5987 static int
5988 ipf_nat_icmpquerytype(int icmptype)
5989 {
5990 
5991 	/*
5992 	 * For the ICMP query NAT code, it is essential that both the query
5993 	 * and the reply match on the NAT rule. Because the NAT structure
5994 	 * does not keep track of the icmptype, and a single NAT structure
5995 	 * is used for all icmp types with the same src, dest and id, we
5996 	 * simply define the replies as queries as well. The funny thing is,
5997 	 * although it seems silly to call a reply a query, this is exactly
5998 	 * as it is defined in the IPv4 specification
5999 	 */
6000 	switch (icmptype)
6001 	{
6002 	case ICMP_ECHOREPLY:
6003 	case ICMP_ECHO:
6004 	/* route aedvertisement/solliciation is currently unsupported: */
6005 	/* it would require rewriting the ICMP data section            */
6006 	case ICMP_TSTAMP:
6007 	case ICMP_TSTAMPREPLY:
6008 	case ICMP_IREQ:
6009 	case ICMP_IREQREPLY:
6010 	case ICMP_MASKREQ:
6011 	case ICMP_MASKREPLY:
6012 		return 1;
6013 	default:
6014 		return 0;
6015 	}
6016 }
6017 
6018 
6019 /* ------------------------------------------------------------------------ */
6020 /* Function:    nat_log                                                     */
6021 /* Returns:     Nil                                                         */
6022 /* Parameters:  softc(I) - pointer to soft context main structure           */
6023 /*              softn(I) - pointer to NAT context structure                 */
6024 /*              nat(I)    - pointer to NAT structure                        */
6025 /*              action(I) - action related to NAT structure being performed */
6026 /*                                                                          */
6027 /* Creates a NAT log entry.                                                 */
6028 /* ------------------------------------------------------------------------ */
6029 void
6030 ipf_nat_log(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, struct nat *nat,
6031     u_int action)
6032 {
6033 #ifdef	IPFILTER_LOG
6034 # ifndef LARGE_NAT
6035 	struct ipnat *np;
6036 	int rulen;
6037 # endif
6038 	struct natlog natl;
6039 	void *items[1];
6040 	size_t sizes[1];
6041 	int types[1];
6042 
6043 	bcopy((char *)&nat->nat_osrc6, (char *)&natl.nl_osrcip,
6044 	      sizeof(natl.nl_osrcip));
6045 	bcopy((char *)&nat->nat_nsrc6, (char *)&natl.nl_nsrcip,
6046 	      sizeof(natl.nl_nsrcip));
6047 	bcopy((char *)&nat->nat_odst6, (char *)&natl.nl_odstip,
6048 	      sizeof(natl.nl_odstip));
6049 	bcopy((char *)&nat->nat_ndst6, (char *)&natl.nl_ndstip,
6050 	      sizeof(natl.nl_ndstip));
6051 
6052 	natl.nl_bytes[0] = nat->nat_bytes[0];
6053 	natl.nl_bytes[1] = nat->nat_bytes[1];
6054 	natl.nl_pkts[0] = nat->nat_pkts[0];
6055 	natl.nl_pkts[1] = nat->nat_pkts[1];
6056 	natl.nl_odstport = nat->nat_odport;
6057 	natl.nl_osrcport = nat->nat_osport;
6058 	natl.nl_nsrcport = nat->nat_nsport;
6059 	natl.nl_ndstport = nat->nat_ndport;
6060 	natl.nl_p[0] = nat->nat_pr[0];
6061 	natl.nl_p[1] = nat->nat_pr[1];
6062 	natl.nl_v[0] = nat->nat_v[0];
6063 	natl.nl_v[1] = nat->nat_v[1];
6064 	natl.nl_type = nat->nat_redir;
6065 	natl.nl_action = action;
6066 	natl.nl_rule = -1;
6067 
6068 	bcopy(nat->nat_ifnames[0], natl.nl_ifnames[0],
6069 	      sizeof(nat->nat_ifnames[0]));
6070 	bcopy(nat->nat_ifnames[1], natl.nl_ifnames[1],
6071 	      sizeof(nat->nat_ifnames[1]));
6072 
6073 # ifndef LARGE_NAT
6074 	if (nat->nat_ptr != NULL) {
6075 		for (rulen = 0, np = softn->ipf_nat_list; np != NULL;
6076 		     np = np->in_next, rulen++)
6077 			if (np == nat->nat_ptr) {
6078 				natl.nl_rule = rulen;
6079 				break;
6080 			}
6081 	}
6082 # endif
6083 	items[0] = &natl;
6084 	sizes[0] = sizeof(natl);
6085 	types[0] = 0;
6086 
6087 	(void) ipf_log_items(softc, IPL_LOGNAT, NULL, items, sizes, types, 1);
6088 #endif
6089 }
6090 
6091 
6092 #if defined(__OpenBSD__)
6093 /* ------------------------------------------------------------------------ */
6094 /* Function:    ipf_nat_ifdetach                                            */
6095 /* Returns:     Nil                                                         */
6096 /* Parameters:  ifp(I) - pointer to network interface                       */
6097 /*                                                                          */
6098 /* Compatibility interface for OpenBSD to trigger the correct updating of   */
6099 /* interface references within IPFilter.                                    */
6100 /* ------------------------------------------------------------------------ */
6101 void
6102 ipf_nat_ifdetach(ifp)
6103 	void *ifp;
6104 {
6105 	ipf_main_softc_t *softc;
6106 
6107 	softc = ipf_get_softc(0);
6108 
6109 	ipf_sync(ifp);
6110 	return;
6111 }
6112 #endif
6113 
6114 
6115 /* ------------------------------------------------------------------------ */
6116 /* Function:    ipf_nat_rule_deref                                          */
6117 /* Returns:     Nil                                                         */
6118 /* Parameters:  softc(I) - pointer to soft context main structure           */
6119 /*              inp(I)   - pointer to pointer to NAT rule                   */
6120 /* Write Locks: ipf_nat                                                     */
6121 /*                                                                          */
6122 /* Dropping the refernce count for a rule means that whatever held the      */
6123 /* pointer to this rule (*inp) is no longer interested in it and when the   */
6124 /* reference count drops to zero, any resources allocated for the rule can  */
6125 /* be released and the rule itself free'd.                                  */
6126 /* ------------------------------------------------------------------------ */
6127 void
6128 ipf_nat_rule_deref(ipf_main_softc_t *softc, ipnat_t **inp)
6129 {
6130 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6131 	ipnat_t *n;
6132 
6133 	n = *inp;
6134 	*inp = NULL;
6135 	n->in_use--;
6136 	if (n->in_use > 0)
6137 		return;
6138 
6139 	if (n->in_apr != NULL)
6140 		ipf_proxy_deref(n->in_apr);
6141 
6142 	ipf_nat_rule_fini(softc, n);
6143 
6144 	if (n->in_redir & NAT_REDIRECT) {
6145 		if ((n->in_flags & IPN_PROXYRULE) == 0) {
6146 			ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules_rdr);
6147 		}
6148 	}
6149 	if (n->in_redir & (NAT_MAP|NAT_MAPBLK)) {
6150 		if ((n->in_flags & IPN_PROXYRULE) == 0) {
6151 			ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules_map);
6152 		}
6153 	}
6154 
6155 	if (n->in_tqehead[0] != NULL) {
6156 		if (ipf_deletetimeoutqueue(n->in_tqehead[0]) == 0) {
6157 			ipf_freetimeoutqueue(softc, n->in_tqehead[1]);
6158 		}
6159 	}
6160 
6161 	if (n->in_tqehead[1] != NULL) {
6162 		if (ipf_deletetimeoutqueue(n->in_tqehead[1]) == 0) {
6163 			ipf_freetimeoutqueue(softc, n->in_tqehead[1]);
6164 		}
6165 	}
6166 
6167 	if ((n->in_flags & IPN_PROXYRULE) == 0) {
6168 		ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules);
6169 	}
6170 
6171 	MUTEX_DESTROY(&n->in_lock);
6172 
6173 	KFREES(n, n->in_size);
6174 
6175 #if SOLARIS && !defined(INSTANCES)
6176 	if (softn->ipf_nat_stats.ns_rules == 0)
6177 		pfil_delayed_copy = 1;
6178 #endif
6179 }
6180 
6181 
6182 /* ------------------------------------------------------------------------ */
6183 /* Function:    ipf_nat_deref                                               */
6184 /* Returns:     Nil                                                         */
6185 /* Parameters:  softc(I) - pointer to soft context main structure           */
6186 /*              natp(I)  - pointer to pointer to NAT table entry            */
6187 /*                                                                          */
6188 /* Decrement the reference counter for this NAT table entry and free it if  */
6189 /* there are no more things using it.                                       */
6190 /*                                                                          */
6191 /* IF nat_ref == 1 when this function is called, then we have an orphan nat */
6192 /* structure *because* it only gets called on paths _after_ nat_ref has been*/
6193 /* incremented.  If nat_ref == 1 then we shouldn't decrement it here        */
6194 /* because nat_delete() will do that and send nat_ref to -1.                */
6195 /*                                                                          */
6196 /* Holding the lock on nat_lock is required to serialise nat_delete() being */
6197 /* called from a NAT flush ioctl with a deref happening because of a packet.*/
6198 /* ------------------------------------------------------------------------ */
6199 void
6200 ipf_nat_deref(ipf_main_softc_t *softc, nat_t **natp)
6201 {
6202 	nat_t *nat;
6203 
6204 	nat = *natp;
6205 	*natp = NULL;
6206 
6207 	MUTEX_ENTER(&nat->nat_lock);
6208 	if (nat->nat_ref > 1) {
6209 		nat->nat_ref--;
6210 		ASSERT(nat->nat_ref >= 0);
6211 		MUTEX_EXIT(&nat->nat_lock);
6212 		return;
6213 	}
6214 	MUTEX_EXIT(&nat->nat_lock);
6215 
6216 	WRITE_ENTER(&softc->ipf_nat);
6217 	ipf_nat_delete(softc, nat, NL_EXPIRE);
6218 	RWLOCK_EXIT(&softc->ipf_nat);
6219 }
6220 
6221 
6222 /* ------------------------------------------------------------------------ */
6223 /* Function:    ipf_nat_clone                                               */
6224 /* Returns:     ipstate_t* - NULL == cloning failed,                        */
6225 /*                           else pointer to new state structure            */
6226 /* Parameters:  fin(I) - pointer to packet information                      */
6227 /*              is(I)  - pointer to master state structure                  */
6228 /* Write Lock:  ipf_nat                                                     */
6229 /*                                                                          */
6230 /* Create a "duplcate" state table entry from the master.                   */
6231 /* ------------------------------------------------------------------------ */
6232 nat_t *
6233 ipf_nat_clone(fr_info_t *fin, nat_t *nat)
6234 {
6235 	ipf_main_softc_t *softc = fin->fin_main_soft;
6236 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6237 	frentry_t *fr;
6238 	nat_t *clone;
6239 	ipnat_t *np;
6240 
6241 	KMALLOC(clone, nat_t *);
6242 	if (clone == NULL) {
6243 		NBUMPSIDED(fin->fin_out, ns_clone_nomem);
6244 		return NULL;
6245 	}
6246 	bcopy((char *)nat, (char *)clone, sizeof(*clone));
6247 
6248 	MUTEX_NUKE(&clone->nat_lock);
6249 
6250 	clone->nat_rev = fin->fin_rev;
6251 	clone->nat_aps = NULL;
6252 	/*
6253 	 * Initialize all these so that ipf_nat_delete() doesn't cause a crash.
6254 	 */
6255 	clone->nat_tqe.tqe_pnext = NULL;
6256 	clone->nat_tqe.tqe_next = NULL;
6257 	clone->nat_tqe.tqe_ifq = NULL;
6258 	clone->nat_tqe.tqe_parent = clone;
6259 
6260 	clone->nat_flags &= ~SI_CLONE;
6261 	clone->nat_flags |= SI_CLONED;
6262 
6263 	if (clone->nat_hm)
6264 		clone->nat_hm->hm_ref++;
6265 
6266 	if (ipf_nat_insert(softc, softn, clone) == -1) {
6267 		KFREE(clone);
6268 		NBUMPSIDED(fin->fin_out, ns_insert_fail);
6269 		return NULL;
6270 	}
6271 
6272 	np = clone->nat_ptr;
6273 	if (np != NULL) {
6274 		if (softn->ipf_nat_logging)
6275 			ipf_nat_log(softc, softn, clone, NL_CLONE);
6276 		np->in_use++;
6277 	}
6278 	fr = clone->nat_fr;
6279 	if (fr != NULL) {
6280 		MUTEX_ENTER(&fr->fr_lock);
6281 		fr->fr_ref++;
6282 		MUTEX_EXIT(&fr->fr_lock);
6283 	}
6284 
6285 
6286 	/*
6287 	 * Because the clone is created outside the normal loop of things and
6288 	 * TCP has special needs in terms of state, initialise the timeout
6289 	 * state of the new NAT from here.
6290 	 */
6291 	if (clone->nat_pr[0] == IPPROTO_TCP) {
6292 		(void) ipf_tcp_age(&clone->nat_tqe, fin, softn->ipf_nat_tcptq,
6293 				   clone->nat_flags, 2);
6294 	}
6295 	clone->nat_sync = ipf_sync_new(softc, SMC_NAT, fin, clone);
6296 	if (softn->ipf_nat_logging)
6297 		ipf_nat_log(softc, softn, clone, NL_CLONE);
6298 	return clone;
6299 }
6300 
6301 
6302 /* ------------------------------------------------------------------------ */
6303 /* Function:   ipf_nat_wildok                                               */
6304 /* Returns:    int - 1 == packet's ports match wildcards                    */
6305 /*                   0 == packet's ports don't match wildcards              */
6306 /* Parameters: nat(I)   - NAT entry                                         */
6307 /*             sport(I) - source port                                       */
6308 /*             dport(I) - destination port                                  */
6309 /*             flags(I) - wildcard flags                                    */
6310 /*             dir(I)   - packet direction                                  */
6311 /*                                                                          */
6312 /* Use NAT entry and packet direction to determine which combination of     */
6313 /* wildcard flags should be used.                                           */
6314 /* ------------------------------------------------------------------------ */
6315 int
6316 ipf_nat_wildok(nat_t *nat, int sport, int dport, int flags, int dir)
6317 {
6318 	/*
6319 	 * When called by       dir is set to
6320 	 * nat_inlookup         NAT_INBOUND (0)
6321 	 * nat_outlookup        NAT_OUTBOUND (1)
6322 	 *
6323 	 * We simply combine the packet's direction in dir with the original
6324 	 * "intended" direction of that NAT entry in nat->nat_dir to decide
6325 	 * which combination of wildcard flags to allow.
6326 	 */
6327 	switch ((dir << 1) | (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND)))
6328 	{
6329 	case 3: /* outbound packet / outbound entry */
6330 		if (((nat->nat_osport == sport) ||
6331 		    (flags & SI_W_SPORT)) &&
6332 		    ((nat->nat_odport == dport) ||
6333 		    (flags & SI_W_DPORT)))
6334 			return 1;
6335 		break;
6336 	case 2: /* outbound packet / inbound entry */
6337 		if (((nat->nat_osport == dport) ||
6338 		    (flags & SI_W_SPORT)) &&
6339 		    ((nat->nat_odport == sport) ||
6340 		    (flags & SI_W_DPORT)))
6341 			return 1;
6342 		break;
6343 	case 1: /* inbound packet / outbound entry */
6344 		if (((nat->nat_osport == dport) ||
6345 		    (flags & SI_W_SPORT)) &&
6346 		    ((nat->nat_odport == sport) ||
6347 		    (flags & SI_W_DPORT)))
6348 			return 1;
6349 		break;
6350 	case 0: /* inbound packet / inbound entry */
6351 		if (((nat->nat_osport == sport) ||
6352 		    (flags & SI_W_SPORT)) &&
6353 		    ((nat->nat_odport == dport) ||
6354 		    (flags & SI_W_DPORT)))
6355 			return 1;
6356 		break;
6357 	default:
6358 		break;
6359 	}
6360 
6361 	return(0);
6362 }
6363 
6364 
6365 /* ------------------------------------------------------------------------ */
6366 /* Function:    nat_mssclamp                                                */
6367 /* Returns:     Nil                                                         */
6368 /* Parameters:  tcp(I)    - pointer to TCP header                           */
6369 /*              maxmss(I) - value to clamp the TCP MSS to                   */
6370 /*              fin(I)    - pointer to packet information                   */
6371 /*              csump(I)  - pointer to TCP checksum                         */
6372 /*                                                                          */
6373 /* Check for MSS option and clamp it if necessary.  If found and changed,   */
6374 /* then the TCP header checksum will be updated to reflect the change in    */
6375 /* the MSS.                                                                 */
6376 /* ------------------------------------------------------------------------ */
6377 static void
6378 ipf_nat_mssclamp(tcphdr_t *tcp, u_32_t maxmss, fr_info_t *fin, u_short *csump)
6379 {
6380 	u_char *cp, *ep, opt;
6381 	int hlen, advance;
6382 	u_32_t mss, sumd;
6383 
6384 	hlen = TCP_OFF(tcp) << 2;
6385 	if (hlen > sizeof(*tcp)) {
6386 		cp = (u_char *)tcp + sizeof(*tcp);
6387 		ep = (u_char *)tcp + hlen;
6388 
6389 		while (cp < ep) {
6390 			opt = cp[0];
6391 			if (opt == TCPOPT_EOL)
6392 				break;
6393 			else if (opt == TCPOPT_NOP) {
6394 				cp++;
6395 				continue;
6396 			}
6397 
6398 			if (cp + 1 >= ep)
6399 				break;
6400 			advance = cp[1];
6401 			if ((cp + advance > ep) || (advance <= 0))
6402 				break;
6403 			switch (opt)
6404 			{
6405 			case TCPOPT_MAXSEG:
6406 				if (advance != 4)
6407 					break;
6408 				mss = cp[2] * 256 + cp[3];
6409 				if (mss > maxmss) {
6410 					cp[2] = maxmss / 256;
6411 					cp[3] = maxmss & 0xff;
6412 					CALC_SUMD(mss, maxmss, sumd);
6413 					ipf_fix_outcksum(0, csump, sumd, 0);
6414 				}
6415 				break;
6416 			default:
6417 				/* ignore unknown options */
6418 				break;
6419 			}
6420 
6421 			cp += advance;
6422 		}
6423 	}
6424 }
6425 
6426 
6427 /* ------------------------------------------------------------------------ */
6428 /* Function:    ipf_nat_setqueue                                            */
6429 /* Returns:     Nil                                                         */
6430 /* Parameters:  softc(I) - pointer to soft context main structure           */
6431 /*              softn(I) - pointer to NAT context structure                 */
6432 /*              nat(I)- pointer to NAT structure                            */
6433 /* Locks:       ipf_nat (read or write)                                     */
6434 /*                                                                          */
6435 /* Put the NAT entry on its default queue entry, using rev as a helped in   */
6436 /* determining which queue it should be placed on.                          */
6437 /* ------------------------------------------------------------------------ */
6438 void
6439 ipf_nat_setqueue(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, nat_t *nat)
6440 {
6441 	ipftq_t *oifq, *nifq;
6442 	int rev = nat->nat_rev;
6443 
6444 	if (nat->nat_ptr != NULL)
6445 		nifq = nat->nat_ptr->in_tqehead[rev];
6446 	else
6447 		nifq = NULL;
6448 
6449 	if (nifq == NULL) {
6450 		switch (nat->nat_pr[0])
6451 		{
6452 		case IPPROTO_UDP :
6453 			nifq = &softn->ipf_nat_udptq;
6454 			break;
6455 		case IPPROTO_ICMP :
6456 			nifq = &softn->ipf_nat_icmptq;
6457 			break;
6458 		case IPPROTO_TCP :
6459 			nifq = softn->ipf_nat_tcptq +
6460 			       nat->nat_tqe.tqe_state[rev];
6461 			break;
6462 		default :
6463 			nifq = &softn->ipf_nat_iptq;
6464 			break;
6465 		}
6466 	}
6467 
6468 	oifq = nat->nat_tqe.tqe_ifq;
6469 	/*
6470 	 * If it's currently on a timeout queue, move it from one queue to
6471 	 * another, else put it on the end of the newly determined queue.
6472 	 */
6473 	if (oifq != NULL)
6474 		ipf_movequeue(softc->ipf_ticks, &nat->nat_tqe, oifq, nifq);
6475 	else
6476 		ipf_queueappend(softc->ipf_ticks, &nat->nat_tqe, nifq, nat);
6477 	return;
6478 }
6479 
6480 
6481 /* ------------------------------------------------------------------------ */
6482 /* Function:    nat_getnext                                                 */
6483 /* Returns:     int - 0 == ok, else error                                   */
6484 /* Parameters:  softc(I) - pointer to soft context main structure           */
6485 /*              t(I)   - pointer to ipftoken structure                      */
6486 /*              itp(I) - pointer to ipfgeniter_t structure                  */
6487 /*                                                                          */
6488 /* Fetch the next nat/ipnat structure pointer from the linked list and      */
6489 /* copy it out to the storage space pointed to by itp_data.  The next item  */
6490 /* in the list to look at is put back in the ipftoken struture.             */
6491 /* ------------------------------------------------------------------------ */
6492 static int
6493 ipf_nat_getnext(ipf_main_softc_t *softc, ipftoken_t *t, ipfgeniter_t *itp,
6494     ipfobj_t *objp)
6495 {
6496 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6497 	hostmap_t *hm, *nexthm = NULL, zerohm;
6498 	ipnat_t *ipn, *nextipnat = NULL, zeroipn;
6499 	nat_t *nat, *nextnat = NULL, zeronat;
6500 	int error = 0;
6501 	void *nnext;
6502 
6503 	if (itp->igi_nitems != 1) {
6504 		IPFERROR(60075);
6505 		return ENOSPC;
6506 	}
6507 
6508 	READ_ENTER(&softc->ipf_nat);
6509 
6510 	switch (itp->igi_type)
6511 	{
6512 	case IPFGENITER_HOSTMAP :
6513 		hm = t->ipt_data;
6514 		if (hm == NULL) {
6515 			nexthm = softn->ipf_hm_maplist;
6516 		} else {
6517 			nexthm = hm->hm_next;
6518 		}
6519 		if (nexthm != NULL) {
6520 			ATOMIC_INC32(nexthm->hm_ref);
6521 			t->ipt_data = nexthm;
6522 		} else {
6523 			bzero(&zerohm, sizeof(zerohm));
6524 			nexthm = &zerohm;
6525 			t->ipt_data = NULL;
6526 		}
6527 		nnext = nexthm->hm_next;
6528 		break;
6529 
6530 	case IPFGENITER_IPNAT :
6531 		ipn = t->ipt_data;
6532 		if (ipn == NULL) {
6533 			nextipnat = softn->ipf_nat_list;
6534 		} else {
6535 			nextipnat = ipn->in_next;
6536 		}
6537 		if (nextipnat != NULL) {
6538 			ATOMIC_INC32(nextipnat->in_use);
6539 			t->ipt_data = nextipnat;
6540 		} else {
6541 			bzero(&zeroipn, sizeof(zeroipn));
6542 			nextipnat = &zeroipn;
6543 			t->ipt_data = NULL;
6544 		}
6545 		nnext = nextipnat->in_next;
6546 		break;
6547 
6548 	case IPFGENITER_NAT :
6549 		nat = t->ipt_data;
6550 		if (nat == NULL) {
6551 			nextnat = softn->ipf_nat_instances;
6552 		} else {
6553 			nextnat = nat->nat_next;
6554 		}
6555 		if (nextnat != NULL) {
6556 			MUTEX_ENTER(&nextnat->nat_lock);
6557 			nextnat->nat_ref++;
6558 			MUTEX_EXIT(&nextnat->nat_lock);
6559 			t->ipt_data = nextnat;
6560 		} else {
6561 			bzero(&zeronat, sizeof(zeronat));
6562 			nextnat = &zeronat;
6563 			t->ipt_data = NULL;
6564 		}
6565 		nnext = nextnat->nat_next;
6566 		break;
6567 
6568 	default :
6569 		RWLOCK_EXIT(&softc->ipf_nat);
6570 		IPFERROR(60055);
6571 		return EINVAL;
6572 	}
6573 
6574 	RWLOCK_EXIT(&softc->ipf_nat);
6575 
6576 	objp->ipfo_ptr = itp->igi_data;
6577 
6578 	switch (itp->igi_type)
6579 	{
6580 	case IPFGENITER_HOSTMAP :
6581 		error = COPYOUT(nexthm, objp->ipfo_ptr, sizeof(*nexthm));
6582 		if (error != 0) {
6583 			IPFERROR(60049);
6584 			error = EFAULT;
6585 		}
6586 		if (hm != NULL) {
6587 			WRITE_ENTER(&softc->ipf_nat);
6588 			ipf_nat_hostmapdel(softc, &hm);
6589 			RWLOCK_EXIT(&softc->ipf_nat);
6590 		}
6591 		break;
6592 
6593 	case IPFGENITER_IPNAT :
6594 		objp->ipfo_size = nextipnat->in_size;
6595 		objp->ipfo_type = IPFOBJ_IPNAT;
6596 		error = ipf_outobjk(softc, objp, nextipnat);
6597 		if (ipn != NULL) {
6598 			WRITE_ENTER(&softc->ipf_nat);
6599 			ipf_nat_rule_deref(softc, &ipn);
6600 			RWLOCK_EXIT(&softc->ipf_nat);
6601 		}
6602 		break;
6603 
6604 	case IPFGENITER_NAT :
6605 		objp->ipfo_size = sizeof(nat_t);
6606 		objp->ipfo_type = IPFOBJ_NAT;
6607 		error = ipf_outobjk(softc, objp, nextnat);
6608 		if (nat != NULL)
6609 			ipf_nat_deref(softc, &nat);
6610 
6611 		break;
6612 	}
6613 
6614 	if (nnext == NULL)
6615 		ipf_token_mark_complete(t);
6616 
6617 	return error;
6618 }
6619 
6620 
6621 /* ------------------------------------------------------------------------ */
6622 /* Function:    nat_extraflush                                              */
6623 /* Returns:     int - 0 == success, -1 == failure                           */
6624 /* Parameters:  softc(I) - pointer to soft context main structure           */
6625 /*              softn(I) - pointer to NAT context structure                 */
6626 /*              which(I) - how to flush the active NAT table                */
6627 /* Write Locks: ipf_nat                                                     */
6628 /*                                                                          */
6629 /* Flush nat tables.  Three actions currently defined:                      */
6630 /* which == 0 : flush all nat table entries                                 */
6631 /* which == 1 : flush TCP connections which have started to close but are   */
6632 /*	      stuck for some reason.                                        */
6633 /* which == 2 : flush TCP connections which have been idle for a long time, */
6634 /*	      starting at > 4 days idle and working back in successive half-*/
6635 /*	      days to at most 12 hours old.  If this fails to free enough   */
6636 /*            slots then work backwards in half hour slots to 30 minutes.   */
6637 /*            If that too fails, then work backwards in 30 second intervals */
6638 /*            for the last 30 minutes to at worst 30 seconds idle.          */
6639 /* ------------------------------------------------------------------------ */
6640 static int
6641 ipf_nat_extraflush(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, int which)
6642 {
6643 	nat_t *nat, **natp;
6644 	ipftqent_t *tqn;
6645 	ipftq_t *ifq;
6646 	int removed;
6647 	SPL_INT(s);
6648 
6649 	removed = 0;
6650 
6651 	SPL_NET(s);
6652 	switch (which)
6653 	{
6654 	case 0 :
6655 		softn->ipf_nat_stats.ns_flush_all++;
6656 		/*
6657 		 * Style 0 flush removes everything...
6658 		 */
6659 		for (natp = &softn->ipf_nat_instances;
6660 		     ((nat = *natp) != NULL); ) {
6661 			ipf_nat_delete(softc, nat, NL_FLUSH);
6662 			removed++;
6663 		}
6664 		break;
6665 
6666 	case 1 :
6667 		softn->ipf_nat_stats.ns_flush_closing++;
6668 		/*
6669 		 * Since we're only interested in things that are closing,
6670 		 * we can start with the appropriate timeout queue.
6671 		 */
6672 		for (ifq = softn->ipf_nat_tcptq + IPF_TCPS_CLOSE_WAIT;
6673 		     ifq != NULL; ifq = ifq->ifq_next) {
6674 
6675 			for (tqn = ifq->ifq_head; tqn != NULL; ) {
6676 				nat = tqn->tqe_parent;
6677 				tqn = tqn->tqe_next;
6678 				if (nat->nat_pr[0] != IPPROTO_TCP ||
6679 				    nat->nat_pr[1] != IPPROTO_TCP)
6680 					break;
6681 				ipf_nat_delete(softc, nat, NL_EXPIRE);
6682 				removed++;
6683 			}
6684 		}
6685 
6686 		/*
6687 		 * Also need to look through the user defined queues.
6688 		 */
6689 		for (ifq = softn->ipf_nat_utqe; ifq != NULL;
6690 		     ifq = ifq->ifq_next) {
6691 			for (tqn = ifq->ifq_head; tqn != NULL; ) {
6692 				nat = tqn->tqe_parent;
6693 				tqn = tqn->tqe_next;
6694 				if (nat->nat_pr[0] != IPPROTO_TCP ||
6695 				    nat->nat_pr[1] != IPPROTO_TCP)
6696 					continue;
6697 
6698 				if ((nat->nat_tcpstate[0] >
6699 				     IPF_TCPS_ESTABLISHED) &&
6700 				    (nat->nat_tcpstate[1] >
6701 				     IPF_TCPS_ESTABLISHED)) {
6702 					ipf_nat_delete(softc, nat, NL_EXPIRE);
6703 					removed++;
6704 				}
6705 			}
6706 		}
6707 		break;
6708 
6709 		/*
6710 		 * Args 5-11 correspond to flushing those particular states
6711 		 * for TCP connections.
6712 		 */
6713 	case IPF_TCPS_CLOSE_WAIT :
6714 	case IPF_TCPS_FIN_WAIT_1 :
6715 	case IPF_TCPS_CLOSING :
6716 	case IPF_TCPS_LAST_ACK :
6717 	case IPF_TCPS_FIN_WAIT_2 :
6718 	case IPF_TCPS_TIME_WAIT :
6719 	case IPF_TCPS_CLOSED :
6720 		softn->ipf_nat_stats.ns_flush_state++;
6721 		tqn = softn->ipf_nat_tcptq[which].ifq_head;
6722 		while (tqn != NULL) {
6723 			nat = tqn->tqe_parent;
6724 			tqn = tqn->tqe_next;
6725 			ipf_nat_delete(softc, nat, NL_FLUSH);
6726 			removed++;
6727 		}
6728 		break;
6729 
6730 	default :
6731 		if (which < 30)
6732 			break;
6733 
6734 		softn->ipf_nat_stats.ns_flush_timeout++;
6735 		/*
6736 		 * Take a large arbitrary number to mean the number of seconds
6737 		 * for which which consider to be the maximum value we'll allow
6738 		 * the expiration to be.
6739 		 */
6740 		which = IPF_TTLVAL(which);
6741 		for (natp = &softn->ipf_nat_instances;
6742 		     ((nat = *natp) != NULL); ) {
6743 			if (softc->ipf_ticks - nat->nat_touched > which) {
6744 				ipf_nat_delete(softc, nat, NL_FLUSH);
6745 				removed++;
6746 			} else
6747 				natp = &nat->nat_next;
6748 		}
6749 		break;
6750 	}
6751 
6752 	if (which != 2) {
6753 		SPL_X(s);
6754 		return removed;
6755 	}
6756 
6757 	softn->ipf_nat_stats.ns_flush_queue++;
6758 
6759 	/*
6760 	 * Asked to remove inactive entries because the table is full, try
6761 	 * again, 3 times, if first attempt failed with a different criteria
6762 	 * each time.  The order tried in must be in decreasing age.
6763 	 * Another alternative is to implement random drop and drop N entries
6764 	 * at random until N have been freed up.
6765 	 */
6766 	if (softc->ipf_ticks - softn->ipf_nat_last_force_flush >
6767 	    IPF_TTLVAL(5)) {
6768 		softn->ipf_nat_last_force_flush = softc->ipf_ticks;
6769 
6770 		removed = ipf_queueflush(softc, ipf_nat_flush_entry,
6771 					 softn->ipf_nat_tcptq,
6772 					 softn->ipf_nat_utqe,
6773 					 &softn->ipf_nat_stats.ns_active,
6774 					 softn->ipf_nat_table_sz,
6775 					 softn->ipf_nat_table_wm_low);
6776 	}
6777 
6778 	SPL_X(s);
6779 	return removed;
6780 }
6781 
6782 
6783 /* ------------------------------------------------------------------------ */
6784 /* Function:    ipf_nat_flush_entry                                         */
6785 /* Returns:     0 - always succeeds                                         */
6786 /* Parameters:  softc(I) - pointer to soft context main structure           */
6787 /*              entry(I) - pointer to NAT entry                             */
6788 /* Write Locks: ipf_nat                                                     */
6789 /*                                                                          */
6790 /* This function is a stepping stone between ipf_queueflush() and           */
6791 /* nat_dlete().  It is used so we can provide a uniform interface via the   */
6792 /* ipf_queueflush() function.  Since the nat_delete() function returns void */
6793 /* we translate that to mean it always succeeds in deleting something.      */
6794 /* ------------------------------------------------------------------------ */
6795 static int
6796 ipf_nat_flush_entry(ipf_main_softc_t *softc, void *entry)
6797 {
6798 	ipf_nat_delete(softc, entry, NL_FLUSH);
6799 	return 0;
6800 }
6801 
6802 
6803 /* ------------------------------------------------------------------------ */
6804 /* Function:    ipf_nat_iterator                                            */
6805 /* Returns:     int - 0 == ok, else error                                   */
6806 /* Parameters:  softc(I) - pointer to soft context main structure           */
6807 /*              token(I) - pointer to ipftoken structure                    */
6808 /*              itp(I)   - pointer to ipfgeniter_t structure                */
6809 /*              obj(I)   - pointer to data description structure            */
6810 /*                                                                          */
6811 /* This function acts as a handler for the SIOCGENITER ioctls that use a    */
6812 /* generic structure to iterate through a list.  There are three different  */
6813 /* linked lists of NAT related information to go through: NAT rules, active */
6814 /* NAT mappings and the NAT fragment cache.                                 */
6815 /* ------------------------------------------------------------------------ */
6816 static int
6817 ipf_nat_iterator(ipf_main_softc_t *softc, ipftoken_t *token, ipfgeniter_t *itp,
6818     ipfobj_t *obj)
6819 {
6820 	int error;
6821 
6822 	if (itp->igi_data == NULL) {
6823 		IPFERROR(60052);
6824 		return EFAULT;
6825 	}
6826 
6827 	switch (itp->igi_type)
6828 	{
6829 	case IPFGENITER_HOSTMAP :
6830 	case IPFGENITER_IPNAT :
6831 	case IPFGENITER_NAT :
6832 		error = ipf_nat_getnext(softc, token, itp, obj);
6833 		break;
6834 
6835 	case IPFGENITER_NATFRAG :
6836 		error = ipf_frag_nat_next(softc, token, itp);
6837 		break;
6838 	default :
6839 		IPFERROR(60053);
6840 		error = EINVAL;
6841 		break;
6842 	}
6843 
6844 	return error;
6845 }
6846 
6847 
6848 /* ------------------------------------------------------------------------ */
6849 /* Function:    ipf_nat_setpending                                          */
6850 /* Returns:     Nil                                                         */
6851 /* Parameters:  softc(I) - pointer to soft context main structure           */
6852 /*              nat(I)   - pointer to NAT structure                         */
6853 /* Locks:       ipf_nat (read or write)                                     */
6854 /*                                                                          */
6855 /* Put the NAT entry on to the pending queue - this queue has a very short  */
6856 /* lifetime where items are put that can't be deleted straight away because */
6857 /* of locking issues but we want to delete them ASAP, anyway.  In calling   */
6858 /* this function, it is assumed that the owner (if there is one, as shown   */
6859 /* by nat_me) is no longer interested in it.                                */
6860 /* ------------------------------------------------------------------------ */
6861 void
6862 ipf_nat_setpending(ipf_main_softc_t *softc, nat_t *nat)
6863 {
6864 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6865 	ipftq_t *oifq;
6866 
6867 	oifq = nat->nat_tqe.tqe_ifq;
6868 	if (oifq != NULL)
6869 		ipf_movequeue(softc->ipf_ticks, &nat->nat_tqe, oifq,
6870 			      &softn->ipf_nat_pending);
6871 	else
6872 		ipf_queueappend(softc->ipf_ticks, &nat->nat_tqe,
6873 				&softn->ipf_nat_pending, nat);
6874 
6875 	if (nat->nat_me != NULL) {
6876 		*nat->nat_me = NULL;
6877 		nat->nat_me = NULL;
6878 		nat->nat_ref--;
6879 		ASSERT(nat->nat_ref >= 0);
6880 	}
6881 }
6882 
6883 
6884 /* ------------------------------------------------------------------------ */
6885 /* Function:    nat_newrewrite                                              */
6886 /* Returns:     int - -1 == error, 0 == success (no move), 1 == success and */
6887 /*                    allow rule to be moved if IPN_ROUNDR is set.          */
6888 /* Parameters:  fin(I) - pointer to packet information                      */
6889 /*              nat(I) - pointer to NAT entry                               */
6890 /*              ni(I)  - pointer to structure with misc. information needed */
6891 /*                       to create new NAT entry.                           */
6892 /* Write Lock:  ipf_nat                                                     */
6893 /*                                                                          */
6894 /* This function is responsible for setting up an active NAT session where  */
6895 /* we are changing both the source and destination parameters at the same   */
6896 /* time.  The loop in here works differently to elsewhere - each iteration  */
6897 /* is responsible for changing a single parameter that can be incremented.  */
6898 /* So one pass may increase the source IP#, next source port, next dest. IP#*/
6899 /* and the last destination port for a total of 4 iterations to try each.   */
6900 /* This is done to try and exhaustively use the translation space available.*/
6901 /* ------------------------------------------------------------------------ */
6902 static int
6903 ipf_nat_newrewrite(fr_info_t *fin, nat_t *nat, natinfo_t *nai)
6904 {
6905 	int src_search = 1;
6906 	int dst_search = 1;
6907 	fr_info_t frnat;
6908 	u_32_t flags;
6909 	u_short swap;
6910 	ipnat_t *np;
6911 	nat_t *natl;
6912 	int l = 0;
6913 	int changed;
6914 
6915 	natl = NULL;
6916 	changed = -1;
6917 	np = nai->nai_np;
6918 	flags = nat->nat_flags;
6919 	bcopy((char *)fin, (char *)&frnat, sizeof(*fin));
6920 
6921 	nat->nat_hm = NULL;
6922 
6923 	do {
6924 		changed = -1;
6925 		/* TRACE (l, src_search, dst_search, np) */
6926 
6927 		if ((src_search == 0) && (np->in_spnext == 0) &&
6928 		    (dst_search == 0) && (np->in_dpnext == 0)) {
6929 			if (l > 0)
6930 				return -1;
6931 		}
6932 
6933 		/*
6934 		 * Find a new source address
6935 		 */
6936 		if (ipf_nat_nextaddr(fin, &np->in_nsrc, &frnat.fin_saddr,
6937 				     &frnat.fin_saddr) == -1) {
6938 			return -1;
6939 		}
6940 
6941 		if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0xffffffff)) {
6942 			src_search = 0;
6943 			if (np->in_stepnext == 0)
6944 				np->in_stepnext = 1;
6945 
6946 		} else if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0)) {
6947 			src_search = 0;
6948 			if (np->in_stepnext == 0)
6949 				np->in_stepnext = 1;
6950 
6951 		} else if (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_nsrcmsk != 0xffffffff) {
6957 			if (np->in_stepnext == 0 && changed == -1) {
6958 				np->in_snip++;
6959 				np->in_stepnext++;
6960 				changed = 0;
6961 			}
6962 		}
6963 
6964 		if ((flags & IPN_TCPUDPICMP) != 0) {
6965 			if (np->in_spnext != 0)
6966 				frnat.fin_data[0] = np->in_spnext;
6967 
6968 			/*
6969 			 * Standard port translation.  Select next port.
6970 			 */
6971 			if ((flags & IPN_FIXEDSPORT) != 0) {
6972 				np->in_stepnext = 2;
6973 			} else if ((np->in_stepnext == 1) &&
6974 				   (changed == -1) && (natl != NULL)) {
6975 				np->in_spnext++;
6976 				np->in_stepnext++;
6977 				changed = 1;
6978 				if (np->in_spnext > np->in_spmax)
6979 					np->in_spnext = np->in_spmin;
6980 			}
6981 		} else {
6982 			np->in_stepnext = 2;
6983 		}
6984 		np->in_stepnext &= 0x3;
6985 
6986 		/*
6987 		 * Find a new destination address
6988 		 */
6989 		/* TRACE (fin, np, l, frnat) */
6990 
6991 		if (ipf_nat_nextaddr(fin, &np->in_ndst, &frnat.fin_daddr,
6992 				     &frnat.fin_daddr) == -1)
6993 			return -1;
6994 		if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0xffffffff)) {
6995 			dst_search = 0;
6996 			if (np->in_stepnext == 2)
6997 				np->in_stepnext = 3;
6998 
6999 		} else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0)) {
7000 			dst_search = 0;
7001 			if (np->in_stepnext == 2)
7002 				np->in_stepnext = 3;
7003 
7004 		} else if (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_ndstmsk != 0xffffffff) {
7010 			if ((np->in_stepnext == 2) && (changed == -1) &&
7011 			    (natl != NULL)) {
7012 				changed = 2;
7013 				np->in_stepnext++;
7014 				np->in_dnip++;
7015 			}
7016 		}
7017 
7018 		if ((flags & IPN_TCPUDPICMP) != 0) {
7019 			if (np->in_dpnext != 0)
7020 				frnat.fin_data[1] = np->in_dpnext;
7021 
7022 			/*
7023 			 * Standard port translation.  Select next port.
7024 			 */
7025 			if ((flags & IPN_FIXEDDPORT) != 0) {
7026 				np->in_stepnext = 0;
7027 			} else if (np->in_stepnext == 3 && changed == -1) {
7028 				np->in_dpnext++;
7029 				np->in_stepnext++;
7030 				changed = 3;
7031 				if (np->in_dpnext > np->in_dpmax)
7032 					np->in_dpnext = np->in_dpmin;
7033 			}
7034 		} else {
7035 			if (np->in_stepnext == 3)
7036 				np->in_stepnext = 0;
7037 		}
7038 
7039 		/* TRACE (frnat) */
7040 
7041 		/*
7042 		 * Here we do a lookup of the connection as seen from
7043 		 * the outside.  If an IP# pair already exists, try
7044 		 * again.  So if you have A->B becomes C->B, you can
7045 		 * also have D->E become C->E but not D->B causing
7046 		 * another C->B.  Also take protocol and ports into
7047 		 * account when determining whether a pre-existing
7048 		 * NAT setup will cause an external conflict where
7049 		 * this is appropriate.
7050 		 *
7051 		 * fin_data[] is swapped around because we are doing a
7052 		 * lookup of the packet is if it were moving in the opposite
7053 		 * direction of the one we are working with now.
7054 		 */
7055 		if (flags & IPN_TCPUDP) {
7056 			swap = frnat.fin_data[0];
7057 			frnat.fin_data[0] = frnat.fin_data[1];
7058 			frnat.fin_data[1] = swap;
7059 		}
7060 		if (fin->fin_out == 1) {
7061 			natl = ipf_nat_inlookup(&frnat,
7062 						flags & ~(SI_WILDP|NAT_SEARCH),
7063 						(u_int)frnat.fin_p,
7064 						frnat.fin_dst, frnat.fin_src);
7065 
7066 		} else {
7067 			natl = ipf_nat_outlookup(&frnat,
7068 						 flags & ~(SI_WILDP|NAT_SEARCH),
7069 						 (u_int)frnat.fin_p,
7070 						 frnat.fin_dst, frnat.fin_src);
7071 		}
7072 		if (flags & IPN_TCPUDP) {
7073 			swap = frnat.fin_data[0];
7074 			frnat.fin_data[0] = frnat.fin_data[1];
7075 			frnat.fin_data[1] = swap;
7076 		}
7077 
7078 		/* TRACE natl, in_stepnext, l */
7079 
7080 		if ((natl != NULL) && (l > 8))	/* XXX 8 is arbitrary */
7081 			return -1;
7082 
7083 		np->in_stepnext &= 0x3;
7084 
7085 		l++;
7086 		changed = -1;
7087 	} while (natl != NULL);
7088 
7089 	nat->nat_osrcip = fin->fin_src;
7090 	nat->nat_odstip = fin->fin_dst;
7091 	nat->nat_nsrcip = frnat.fin_src;
7092 	nat->nat_ndstip = frnat.fin_dst;
7093 
7094 	if ((flags & IPN_TCPUDP) != 0) {
7095 		nat->nat_osport = htons(fin->fin_data[0]);
7096 		nat->nat_odport = htons(fin->fin_data[1]);
7097 		nat->nat_nsport = htons(frnat.fin_data[0]);
7098 		nat->nat_ndport = htons(frnat.fin_data[1]);
7099 	} else if ((flags & IPN_ICMPQUERY) != 0) {
7100 		nat->nat_oicmpid = fin->fin_data[1];
7101 		nat->nat_nicmpid = frnat.fin_data[1];
7102 	}
7103 
7104 	return 0;
7105 }
7106 
7107 
7108 /* ------------------------------------------------------------------------ */
7109 /* Function:    nat_newdivert                                               */
7110 /* Returns:     int - -1 == error, 0 == success                             */
7111 /* Parameters:  fin(I) - pointer to packet information                      */
7112 /*              nat(I) - pointer to NAT entry                               */
7113 /*              ni(I)  - pointer to structure with misc. information needed */
7114 /*                       to create new NAT entry.                           */
7115 /* Write Lock:  ipf_nat                                                     */
7116 /*                                                                          */
7117 /* Create a new NAT  divert session as defined by the NAT rule.  This is    */
7118 /* somewhat different to other NAT session creation routines because we     */
7119 /* do not iterate through either port numbers or IP addresses, searching    */
7120 /* for a unique mapping, however, a complimentary duplicate check is made.  */
7121 /* ------------------------------------------------------------------------ */
7122 static int
7123 ipf_nat_newdivert(fr_info_t *fin, nat_t *nat, natinfo_t *nai)
7124 {
7125 	ipf_main_softc_t *softc = fin->fin_main_soft;
7126 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7127 	fr_info_t frnat;
7128 	ipnat_t *np;
7129 	nat_t *natl;
7130 	int p;
7131 
7132 	np = nai->nai_np;
7133 	bcopy((char *)fin, (char *)&frnat, sizeof(*fin));
7134 
7135 	nat->nat_pr[0] = 0;
7136 	nat->nat_osrcaddr = fin->fin_saddr;
7137 	nat->nat_odstaddr = fin->fin_daddr;
7138 	frnat.fin_saddr = htonl(np->in_snip);
7139 	frnat.fin_daddr = htonl(np->in_dnip);
7140 	if ((nat->nat_flags & IPN_TCPUDP) != 0) {
7141 		nat->nat_osport = htons(fin->fin_data[0]);
7142 		nat->nat_odport = htons(fin->fin_data[1]);
7143 	} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
7144 		nat->nat_oicmpid = fin->fin_data[1];
7145 	}
7146 
7147 	if (np->in_redir & NAT_DIVERTUDP) {
7148 		frnat.fin_data[0] = np->in_spnext;
7149 		frnat.fin_data[1] = np->in_dpnext;
7150 		frnat.fin_flx |= FI_TCPUDP;
7151 		p = IPPROTO_UDP;
7152 	} else {
7153 		frnat.fin_flx &= ~FI_TCPUDP;
7154 		p = IPPROTO_IPIP;
7155 	}
7156 
7157 	if (fin->fin_out == 1) {
7158 		natl = ipf_nat_inlookup(&frnat, 0, p,
7159 					frnat.fin_dst, frnat.fin_src);
7160 
7161 	} else {
7162 		natl = ipf_nat_outlookup(&frnat, 0, p,
7163 					 frnat.fin_dst, frnat.fin_src);
7164 	}
7165 
7166 	if (natl != NULL) {
7167 		NBUMPSIDED(fin->fin_out, ns_divert_exist);
7168 		return -1;
7169 	}
7170 
7171 	nat->nat_nsrcaddr = frnat.fin_saddr;
7172 	nat->nat_ndstaddr = frnat.fin_daddr;
7173 	if ((nat->nat_flags & IPN_TCPUDP) != 0) {
7174 		nat->nat_nsport = htons(frnat.fin_data[0]);
7175 		nat->nat_ndport = htons(frnat.fin_data[1]);
7176 	} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
7177 		nat->nat_nicmpid = frnat.fin_data[1];
7178 	}
7179 
7180 	nat->nat_pr[fin->fin_out] = fin->fin_p;
7181 	nat->nat_pr[1 - fin->fin_out] = p;
7182 
7183 	if (np->in_redir & NAT_REDIRECT)
7184 		nat->nat_dir = NAT_DIVERTIN;
7185 	else
7186 		nat->nat_dir = NAT_DIVERTOUT;
7187 
7188 	return 0;
7189 }
7190 
7191 
7192 /* ------------------------------------------------------------------------ */
7193 /* Function:    nat_builddivertmp                                           */
7194 /* Returns:     int - -1 == error, 0 == success                             */
7195 /* Parameters:  softn(I) - pointer to NAT context structure                 */
7196 /*              np(I)    - pointer to a NAT rule                            */
7197 /*                                                                          */
7198 /* For divert rules, a skeleton packet representing what will be prepended  */
7199 /* to the real packet is created.  Even though we don't have the full       */
7200 /* packet here, a checksum is calculated that we update later when we       */
7201 /* fill in the final details.  At present a 0 checksum for UDP is being set */
7202 /* here because it is expected that divert will be used for localhost.      */
7203 /* ------------------------------------------------------------------------ */
7204 static int
7205 ipf_nat_builddivertmp(ipf_nat_softc_t *softn, ipnat_t *np)
7206 {
7207 	udphdr_t *uh;
7208 	size_t len;
7209 	ip_t *ip;
7210 
7211 	if ((np->in_redir & NAT_DIVERTUDP) != 0)
7212 		len = sizeof(ip_t) + sizeof(udphdr_t);
7213 	else
7214 		len = sizeof(ip_t);
7215 
7216 	ALLOC_MB_T(np->in_divmp, len);
7217 	if (np->in_divmp == NULL) {
7218 		NBUMPD(ipf_nat_stats, ns_divert_build);
7219 		return -1;
7220 	}
7221 
7222 	/*
7223 	 * First, the header to get the packet diverted to the new destination
7224 	 */
7225 	ip = MTOD(np->in_divmp, ip_t *);
7226 	IP_V_A(ip, 4);
7227 	IP_HL_A(ip, 5);
7228 	ip->ip_tos = 0;
7229 	if ((np->in_redir & NAT_DIVERTUDP) != 0)
7230 		ip->ip_p = IPPROTO_UDP;
7231 	else
7232 		ip->ip_p = IPPROTO_IPIP;
7233 	ip->ip_ttl = 255;
7234 	ip->ip_off = 0;
7235 	ip->ip_sum = 0;
7236 	ip->ip_len = htons(len);
7237 	ip->ip_id = 0;
7238 	ip->ip_src.s_addr = htonl(np->in_snip);
7239 	ip->ip_dst.s_addr = htonl(np->in_dnip);
7240 	ip->ip_sum = ipf_cksum((u_short *)ip, sizeof(*ip));
7241 
7242 	if (np->in_redir & NAT_DIVERTUDP) {
7243 		uh = (udphdr_t *)(ip + 1);
7244 		uh->uh_sum = 0;
7245 		uh->uh_ulen = 8;
7246 		uh->uh_sport = htons(np->in_spnext);
7247 		uh->uh_dport = htons(np->in_dpnext);
7248 	}
7249 
7250 	return 0;
7251 }
7252 
7253 
7254 #define	MINDECAP	(sizeof(ip_t) + sizeof(udphdr_t) + sizeof(ip_t))
7255 
7256 /* ------------------------------------------------------------------------ */
7257 /* Function:    nat_decap                                                   */
7258 /* Returns:     int - -1 == error, 0 == success                             */
7259 /* Parameters:  fin(I) - pointer to packet information                      */
7260 /*              nat(I) - pointer to current NAT session                     */
7261 /*                                                                          */
7262 /* This function is responsible for undoing a packet's encapsulation in the */
7263 /* reverse of an encap/divert rule.  After removing the outer encapsulation */
7264 /* it is necessary to call ipf_makefrip() again so that the contents of 'fin'*/
7265 /* match the "new" packet as it may still be used by IPFilter elsewhere.    */
7266 /* We use "dir" here as the basis for some of the expectations about the    */
7267 /* outer header.  If we return an error, the goal is to leave the original  */
7268 /* packet information undisturbed - this falls short at the end where we'd  */
7269 /* need to back a backup copy of "fin" - expensive.                         */
7270 /* ------------------------------------------------------------------------ */
7271 static int
7272 ipf_nat_decap(fr_info_t *fin, nat_t *nat)
7273 {
7274 	ipf_main_softc_t *softc = fin->fin_main_soft;
7275 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7276 	char *hdr;
7277 	int hlen;
7278 	int skip;
7279 	mb_t *m;
7280 
7281 	if ((fin->fin_flx & FI_ICMPERR) != 0) {
7282 		/*
7283 		 * ICMP packets don't get decapsulated, instead what we need
7284 		 * to do is change the ICMP reply from including (in the data
7285 		 * portion for errors) the encapsulated packet that we sent
7286 		 * out to something that resembles the original packet prior
7287 		 * to encapsulation.  This isn't done here - all we're doing
7288 		 * here is changing the outer address to ensure that it gets
7289 		 * targetted back to the correct system.
7290 		 */
7291 
7292 		if (nat->nat_dir & NAT_OUTBOUND) {
7293 			u_32_t sum1, sum2, sumd;
7294 
7295 			sum1 = ntohl(fin->fin_daddr);
7296 			sum2 = ntohl(nat->nat_osrcaddr);
7297 			CALC_SUMD(sum1, sum2, sumd);
7298 			fin->fin_ip->ip_dst = nat->nat_osrcip;
7299 			fin->fin_daddr = nat->nat_osrcaddr;
7300 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
7301      defined(__osf__) || defined(linux)
7302 			ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, sumd, 0);
7303 #endif
7304 		}
7305 		return 0;
7306 	}
7307 
7308 	m = fin->fin_m;
7309 	skip = fin->fin_hlen;
7310 
7311 	switch (nat->nat_dir)
7312 	{
7313 	case NAT_DIVERTIN :
7314 	case NAT_DIVERTOUT :
7315 		if (fin->fin_plen < MINDECAP)
7316 			return -1;
7317 		skip += sizeof(udphdr_t);
7318 		break;
7319 
7320 	case NAT_ENCAPIN :
7321 	case NAT_ENCAPOUT :
7322 		if (fin->fin_plen < (skip + sizeof(ip_t)))
7323 			return -1;
7324 		break;
7325 	default :
7326 		return -1;
7327 		/* NOTREACHED */
7328 	}
7329 
7330 	/*
7331 	 * The aim here is to keep the original packet details in "fin" for
7332 	 * as long as possible so that returning with an error is for the
7333 	 * original packet and there is little undoing work to do.
7334 	 */
7335 	if (M_LEN(m) < skip + sizeof(ip_t)) {
7336 		if (ipf_pr_pullup(fin, skip + sizeof(ip_t)) == -1)
7337 			return -1;
7338 	}
7339 
7340 	hdr = MTOD(fin->fin_m, char *);
7341 	fin->fin_ip = (ip_t *)(hdr + skip);
7342 	hlen = IP_HL(fin->fin_ip) << 2;
7343 
7344 	if (ipf_pr_pullup(fin, skip + hlen) == -1) {
7345 		NBUMPSIDED(fin->fin_out, ns_decap_pullup);
7346 		return -1;
7347 	}
7348 
7349 	fin->fin_hlen = hlen;
7350 	fin->fin_dlen -= skip;
7351 	fin->fin_plen -= skip;
7352 	fin->fin_ipoff += skip;
7353 
7354 	if (ipf_makefrip(hlen, (ip_t *)hdr, fin) == -1) {
7355 		NBUMPSIDED(fin->fin_out, ns_decap_bad);
7356 		return -1;
7357 	}
7358 
7359 	return skip;
7360 }
7361 
7362 
7363 /* ------------------------------------------------------------------------ */
7364 /* Function:    nat_nextaddr                                                */
7365 /* Returns:     int - -1 == bad input (no new address),                     */
7366 /*                     0 == success and dst has new address                 */
7367 /* Parameters:  fin(I) - pointer to packet information                      */
7368 /*              na(I)  - how to generate new address                        */
7369 /*              old(I) - original address being replaced                    */
7370 /*              dst(O) - where to put the new address                       */
7371 /* Write Lock:  ipf_nat                                                     */
7372 /*                                                                          */
7373 /* This function uses the contents of the "na" structure, in combination    */
7374 /* with "old" to produce a new address to store in "dst".  Not all of the   */
7375 /* possible uses of "na" will result in a new address.                      */
7376 /* ------------------------------------------------------------------------ */
7377 static int
7378 ipf_nat_nextaddr(fr_info_t *fin, nat_addr_t *na, u_32_t *old, u_32_t *dst)
7379 {
7380 	ipf_main_softc_t *softc = fin->fin_main_soft;
7381 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7382 	u_32_t amin, amax, new;
7383 	i6addr_t newip;
7384 	int error;
7385 
7386 	new = 0;
7387 	amin = na->na_addr[0].in4.s_addr;
7388 
7389 	switch (na->na_atype)
7390 	{
7391 	case FRI_RANGE :
7392 		amax = na->na_addr[1].in4.s_addr;
7393 		break;
7394 
7395 	case FRI_NETMASKED :
7396 	case FRI_DYNAMIC :
7397 	case FRI_NORMAL :
7398 		/*
7399 		 * Compute the maximum address by adding the inverse of the
7400 		 * netmask to the minimum address.
7401 		 */
7402 		amax = ~na->na_addr[1].in4.s_addr;
7403 		amax |= amin;
7404 		break;
7405 
7406 	case FRI_LOOKUP :
7407 		break;
7408 
7409 	case FRI_BROADCAST :
7410 	case FRI_PEERADDR :
7411 	case FRI_NETWORK :
7412 	default :
7413 		return -1;
7414 	}
7415 
7416 	error = -1;
7417 
7418 	if (na->na_atype == FRI_LOOKUP) {
7419 		if (na->na_type == IPLT_DSTLIST) {
7420 			error = ipf_dstlist_select_node(fin, na->na_ptr, dst,
7421 							NULL);
7422 		} else {
7423 			NBUMPSIDE(fin->fin_out, ns_badnextaddr);
7424 		}
7425 
7426 	} else if (na->na_atype == IPLT_NONE) {
7427 		/*
7428 		 * 0/0 as the new address means leave it alone.
7429 		 */
7430 		if (na->na_addr[0].in4.s_addr == 0 &&
7431 		    na->na_addr[1].in4.s_addr == 0) {
7432 			new = *old;
7433 
7434 		/*
7435 		 * 0/32 means get the interface's address
7436 		 */
7437 		} else if (na->na_addr[0].in4.s_addr == 0 &&
7438 			   na->na_addr[1].in4.s_addr == 0xffffffff) {
7439 			if (ipf_ifpaddr(softc, 4, na->na_atype,
7440 					fin->fin_ifp, &newip, NULL) == -1) {
7441 				NBUMPSIDED(fin->fin_out, ns_ifpaddrfail);
7442 				return -1;
7443 			}
7444 			new = newip.in4.s_addr;
7445 		} else {
7446 			new = htonl(na->na_nextip);
7447 		}
7448 		*dst = new;
7449 		error = 0;
7450 
7451 	} else {
7452 		NBUMPSIDE(fin->fin_out, ns_badnextaddr);
7453 	}
7454 
7455 	return error;
7456 }
7457 
7458 
7459 /* ------------------------------------------------------------------------ */
7460 /* Function:    nat_nextaddrinit                                            */
7461 /* Returns:     int - 0 == success, else error number                       */
7462 /* Parameters:  softc(I) - pointer to soft context main structure           */
7463 /*              na(I)      - NAT address information for generating new addr*/
7464 /*              initial(I) - flag indicating if it is the first call for    */
7465 /*                           this "na" structure.                           */
7466 /*              ifp(I)     - network interface to derive address            */
7467 /*                           information from.                              */
7468 /*                                                                          */
7469 /* This function is expected to be called in two scenarious: when a new NAT */
7470 /* rule is loaded into the kernel and when the list of NAT rules is sync'd  */
7471 /* up with the valid network interfaces (possibly due to them changing.)    */
7472 /* To distinguish between these, the "initial" parameter is used.  If it is */
7473 /* 1 then this indicates the rule has just been reloaded and 0 for when we  */
7474 /* are updating information.  This difference is important because in       */
7475 /* instances where we are not updating address information associated with  */
7476 /* a network interface, we don't want to disturb what the "next" address to */
7477 /* come out of ipf_nat_nextaddr() will be.                                  */
7478 /* ------------------------------------------------------------------------ */
7479 static int
7480 ipf_nat_nextaddrinit(ipf_main_softc_t *softc, char *base, nat_addr_t *na,
7481     int initial, void *ifp)
7482 {
7483 
7484 	switch (na->na_atype)
7485 	{
7486 	case FRI_LOOKUP :
7487 		if (na->na_subtype == 0) {
7488 			na->na_ptr = ipf_lookup_res_num(softc, IPL_LOGNAT,
7489 							na->na_type,
7490 							na->na_num,
7491 							&na->na_func);
7492 		} else if (na->na_subtype == 1) {
7493 			na->na_ptr = ipf_lookup_res_name(softc, IPL_LOGNAT,
7494 							 na->na_type,
7495 							 base + na->na_num,
7496 							 &na->na_func);
7497 		}
7498 		if (na->na_func == NULL) {
7499 			IPFERROR(60060);
7500 			return ESRCH;
7501 		}
7502 		if (na->na_ptr == NULL) {
7503 			IPFERROR(60056);
7504 			return ESRCH;
7505 		}
7506 		break;
7507 
7508 	case FRI_DYNAMIC :
7509 	case FRI_BROADCAST :
7510 	case FRI_NETWORK :
7511 	case FRI_NETMASKED :
7512 	case FRI_PEERADDR :
7513 		if (ifp != NULL)
7514 			(void )ipf_ifpaddr(softc, 4, na->na_atype, ifp,
7515 					   &na->na_addr[0], &na->na_addr[1]);
7516 		break;
7517 
7518 	case FRI_SPLIT :
7519 	case FRI_RANGE :
7520 		if (initial)
7521 			na->na_nextip = ntohl(na->na_addr[0].in4.s_addr);
7522 		break;
7523 
7524 	case FRI_NONE :
7525 		na->na_addr[0].in4.s_addr &= na->na_addr[1].in4.s_addr;
7526 		return 0;
7527 
7528 	case FRI_NORMAL :
7529 		na->na_addr[0].in4.s_addr &= na->na_addr[1].in4.s_addr;
7530 		break;
7531 
7532 	default :
7533 		IPFERROR(60054);
7534 		return EINVAL;
7535 	}
7536 
7537 	if (initial && (na->na_atype == FRI_NORMAL)) {
7538 		if (na->na_addr[0].in4.s_addr == 0) {
7539 			if ((na->na_addr[1].in4.s_addr == 0xffffffff) ||
7540 			    (na->na_addr[1].in4.s_addr == 0)) {
7541 				return 0;
7542 			}
7543 		}
7544 
7545 		if (na->na_addr[1].in4.s_addr == 0xffffffff) {
7546 			na->na_nextip = ntohl(na->na_addr[0].in4.s_addr);
7547 		} else {
7548 			na->na_nextip = ntohl(na->na_addr[0].in4.s_addr) + 1;
7549 		}
7550 	}
7551 
7552 	return 0;
7553 }
7554 
7555 
7556 /* ------------------------------------------------------------------------ */
7557 /* Function:    ipf_nat_matchflush                                          */
7558 /* Returns:     int - -1 == error, 0 == success                             */
7559 /* Parameters:  softc(I) - pointer to soft context main structure           */
7560 /*              softn(I) - pointer to NAT context structure                 */
7561 /*              nat(I)   - pointer to current NAT session                   */
7562 /*                                                                          */
7563 /* ------------------------------------------------------------------------ */
7564 static int
7565 ipf_nat_matchflush(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, void *data)
7566 {
7567 	int *array, flushed, error;
7568 	nat_t *nat, *natnext;
7569 	ipfobj_t obj;
7570 
7571 	error = ipf_matcharray_load(softc, data, &obj, &array);
7572 	if (error != 0)
7573 		return error;
7574 
7575 	flushed = 0;
7576 
7577 	for (nat = softn->ipf_nat_instances; nat != NULL; nat = natnext) {
7578 		natnext = nat->nat_next;
7579 		if (ipf_nat_matcharray(nat, array, softc->ipf_ticks) == 0) {
7580 			ipf_nat_delete(softc, nat, NL_FLUSH);
7581 			flushed++;
7582 		}
7583 	}
7584 
7585 	obj.ipfo_retval = flushed;
7586 	error = BCOPYOUT(&obj, data, sizeof(obj));
7587 
7588 	KFREES(array, array[0] * sizeof(*array));
7589 
7590 	return error;
7591 }
7592 
7593 
7594 /* ------------------------------------------------------------------------ */
7595 /* Function:    ipf_nat_matcharray                                          */
7596 /* Returns:     int - -1 == error, 0 == success                             */
7597 /* Parameters:  fin(I) - pointer to packet information                      */
7598 /*              nat(I) - pointer to current NAT session                     */
7599 /*                                                                          */
7600 /* ------------------------------------------------------------------------ */
7601 static int
7602 ipf_nat_matcharray(nat_t *nat, int *array, u_long ticks)
7603 {
7604 	int i, n, *x, e, p;
7605 
7606 	e = 0;
7607 	n = array[0];
7608 	x = array + 1;
7609 
7610 	for (; n > 0; x += 3 + x[2]) {
7611 		if (x[0] == IPF_EXP_END)
7612 			break;
7613 		e = 0;
7614 
7615 		n -= x[2] + 3;
7616 		if (n < 0)
7617 			break;
7618 
7619 		p = x[0] >> 16;
7620 		if (p != 0 && p != nat->nat_pr[1])
7621 			break;
7622 
7623 		switch (x[0])
7624 		{
7625 		case IPF_EXP_IP_PR :
7626 			for (i = 0; !e && i < x[2]; i++) {
7627 				e |= (nat->nat_pr[1] == x[i + 3]);
7628 			}
7629 			break;
7630 
7631 		case IPF_EXP_IP_SRCADDR :
7632 			if (nat->nat_v[0] == 4) {
7633 				for (i = 0; !e && i < x[2]; i++) {
7634 					e |= ((nat->nat_osrcaddr & x[i + 4]) ==
7635 					      x[i + 3]);
7636 				}
7637 			}
7638 			if (nat->nat_v[1] == 4) {
7639 				for (i = 0; !e && i < x[2]; i++) {
7640 					e |= ((nat->nat_nsrcaddr & x[i + 4]) ==
7641 					      x[i + 3]);
7642 				}
7643 			}
7644 			break;
7645 
7646 		case IPF_EXP_IP_DSTADDR :
7647 			if (nat->nat_v[0] == 4) {
7648 				for (i = 0; !e && i < x[2]; i++) {
7649 					e |= ((nat->nat_odstaddr & x[i + 4]) ==
7650 					      x[i + 3]);
7651 				}
7652 			}
7653 			if (nat->nat_v[1] == 4) {
7654 				for (i = 0; !e && i < x[2]; i++) {
7655 					e |= ((nat->nat_ndstaddr & x[i + 4]) ==
7656 					      x[i + 3]);
7657 				}
7658 			}
7659 			break;
7660 
7661 		case IPF_EXP_IP_ADDR :
7662 			for (i = 0; !e && i < x[2]; i++) {
7663 				if (nat->nat_v[0] == 4) {
7664 					e |= ((nat->nat_osrcaddr & x[i + 4]) ==
7665 					      x[i + 3]);
7666 				}
7667 				if (nat->nat_v[1] == 4) {
7668 					e |= ((nat->nat_nsrcaddr & x[i + 4]) ==
7669 					      x[i + 3]);
7670 				}
7671 				if (nat->nat_v[0] == 4) {
7672 					e |= ((nat->nat_odstaddr & x[i + 4]) ==
7673 					      x[i + 3]);
7674 				}
7675 				if (nat->nat_v[1] == 4) {
7676 					e |= ((nat->nat_ndstaddr & x[i + 4]) ==
7677 					      x[i + 3]);
7678 				}
7679 			}
7680 			break;
7681 
7682 #ifdef USE_INET6
7683 		case IPF_EXP_IP6_SRCADDR :
7684 			if (nat->nat_v[0] == 6) {
7685 				for (i = 0; !e && i < x[3]; i++) {
7686 					e |= IP6_MASKEQ(&nat->nat_osrc6,
7687 							x + i + 7, x + i + 3);
7688 				}
7689 			}
7690 			if (nat->nat_v[1] == 6) {
7691 				for (i = 0; !e && i < x[3]; i++) {
7692 					e |= IP6_MASKEQ(&nat->nat_nsrc6,
7693 							x + i + 7, x + i + 3);
7694 				}
7695 			}
7696 			break;
7697 
7698 		case IPF_EXP_IP6_DSTADDR :
7699 			if (nat->nat_v[0] == 6) {
7700 				for (i = 0; !e && i < x[3]; i++) {
7701 					e |= IP6_MASKEQ(&nat->nat_odst6,
7702 							x + i + 7,
7703 							x + i + 3);
7704 				}
7705 			}
7706 			if (nat->nat_v[1] == 6) {
7707 				for (i = 0; !e && i < x[3]; i++) {
7708 					e |= IP6_MASKEQ(&nat->nat_ndst6,
7709 							x + i + 7,
7710 							x + i + 3);
7711 				}
7712 			}
7713 			break;
7714 
7715 		case IPF_EXP_IP6_ADDR :
7716 			for (i = 0; !e && i < x[3]; i++) {
7717 				if (nat->nat_v[0] == 6) {
7718 					e |= IP6_MASKEQ(&nat->nat_osrc6,
7719 							x + i + 7,
7720 							x + i + 3);
7721 				}
7722 				if (nat->nat_v[0] == 6) {
7723 					e |= IP6_MASKEQ(&nat->nat_odst6,
7724 							x + i + 7,
7725 							x + i + 3);
7726 				}
7727 				if (nat->nat_v[1] == 6) {
7728 					e |= IP6_MASKEQ(&nat->nat_nsrc6,
7729 							x + i + 7,
7730 							x + i + 3);
7731 				}
7732 				if (nat->nat_v[1] == 6) {
7733 					e |= IP6_MASKEQ(&nat->nat_ndst6,
7734 							x + i + 7,
7735 							x + i + 3);
7736 				}
7737 			}
7738 			break;
7739 #endif
7740 
7741 		case IPF_EXP_UDP_PORT :
7742 		case IPF_EXP_TCP_PORT :
7743 			for (i = 0; !e && i < x[2]; i++) {
7744 				e |= (nat->nat_nsport == x[i + 3]) ||
7745 				     (nat->nat_ndport == x[i + 3]);
7746 			}
7747 			break;
7748 
7749 		case IPF_EXP_UDP_SPORT :
7750 		case IPF_EXP_TCP_SPORT :
7751 			for (i = 0; !e && i < x[2]; i++) {
7752 				e |= (nat->nat_nsport == x[i + 3]);
7753 			}
7754 			break;
7755 
7756 		case IPF_EXP_UDP_DPORT :
7757 		case IPF_EXP_TCP_DPORT :
7758 			for (i = 0; !e && i < x[2]; i++) {
7759 				e |= (nat->nat_ndport == x[i + 3]);
7760 			}
7761 			break;
7762 
7763 		case IPF_EXP_TCP_STATE :
7764 			for (i = 0; !e && i < x[2]; i++) {
7765 				e |= (nat->nat_tcpstate[0] == x[i + 3]) ||
7766 				     (nat->nat_tcpstate[1] == x[i + 3]);
7767 			}
7768 			break;
7769 
7770 		case IPF_EXP_IDLE_GT :
7771 			e |= (ticks - nat->nat_touched > x[3]);
7772 			break;
7773 		}
7774 		e ^= x[1];
7775 
7776 		if (!e)
7777 			break;
7778 	}
7779 
7780 	return e;
7781 }
7782 
7783 
7784 /* ------------------------------------------------------------------------ */
7785 /* Function:    ipf_nat_gettable                                            */
7786 /* Returns:     int     - 0 = success, else error                           */
7787 /* Parameters:  softc(I) - pointer to soft context main structure           */
7788 /*              softn(I) - pointer to NAT context structure                 */
7789 /*              data(I)  - pointer to ioctl data                            */
7790 /*                                                                          */
7791 /* This function handles ioctl requests for tables of nat information.      */
7792 /* At present the only table it deals with is the hash bucket statistics.   */
7793 /* ------------------------------------------------------------------------ */
7794 static int
7795 ipf_nat_gettable(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, char *data)
7796 {
7797 	ipftable_t table;
7798 	int error;
7799 
7800 	error = ipf_inobj(softc, data, NULL, &table, IPFOBJ_GTABLE);
7801 	if (error != 0)
7802 		return error;
7803 
7804 	switch (table.ita_type)
7805 	{
7806 	case IPFTABLE_BUCKETS_NATIN :
7807 		error = COPYOUT(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
7808 				table.ita_table,
7809 				softn->ipf_nat_table_sz * sizeof(u_int));
7810 		break;
7811 
7812 	case IPFTABLE_BUCKETS_NATOUT :
7813 		error = COPYOUT(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
7814 				table.ita_table,
7815 				softn->ipf_nat_table_sz * sizeof(u_int));
7816 		break;
7817 
7818 	default :
7819 		IPFERROR(60058);
7820 		return EINVAL;
7821 	}
7822 
7823 	if (error != 0) {
7824 		IPFERROR(60059);
7825 		error = EFAULT;
7826 	}
7827 	return error;
7828 }
7829 
7830 
7831 /* ------------------------------------------------------------------------ */
7832 /* Function:    ipf_nat_settimeout                                          */
7833 /* Returns:     int  - 0 = success, else failure			    */
7834 /* Parameters:  softc(I) - pointer to soft context main structure           */
7835 /*              t(I) - pointer to tunable                                   */
7836 /*              p(I) - pointer to new tuning data                           */
7837 /*                                                                          */
7838 /* Apply the timeout change to the NAT timeout queues.                      */
7839 /* ------------------------------------------------------------------------ */
7840 int
7841 ipf_nat_settimeout(struct ipf_main_softc_s *softc, ipftuneable_t *t,
7842     ipftuneval_t *p)
7843 {
7844 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7845 
7846 	if (!strncmp(t->ipft_name, "tcp_", 4))
7847 		return ipf_settimeout_tcp(t, p, softn->ipf_nat_tcptq);
7848 
7849 	if (!strcmp(t->ipft_name, "udp_timeout")) {
7850 		ipf_apply_timeout(&softn->ipf_nat_udptq, p->ipftu_int);
7851 	} else if (!strcmp(t->ipft_name, "udp_ack_timeout")) {
7852 		ipf_apply_timeout(&softn->ipf_nat_udpacktq, p->ipftu_int);
7853 	} else if (!strcmp(t->ipft_name, "icmp_timeout")) {
7854 		ipf_apply_timeout(&softn->ipf_nat_icmptq, p->ipftu_int);
7855 	} else if (!strcmp(t->ipft_name, "icmp_ack_timeout")) {
7856 		ipf_apply_timeout(&softn->ipf_nat_icmpacktq, p->ipftu_int);
7857 	} else if (!strcmp(t->ipft_name, "ip_timeout")) {
7858 		ipf_apply_timeout(&softn->ipf_nat_iptq, p->ipftu_int);
7859 	} else {
7860 		IPFERROR(60062);
7861 		return ESRCH;
7862 	}
7863 	return 0;
7864 }
7865 
7866 
7867 /* ------------------------------------------------------------------------ */
7868 /* Function:    ipf_nat_rehash                                              */
7869 /* Returns:     int  - 0 = success, else failure			    */
7870 /* Parameters:  softc(I) - pointer to soft context main structure           */
7871 /*              t(I) - pointer to tunable                                   */
7872 /*              p(I) - pointer to new tuning data                           */
7873 /*                                                                          */
7874 /* To change the size of the basic NAT table, we need to first allocate the */
7875 /* new tables (lest it fails and we've got nowhere to store all of the NAT  */
7876 /* sessions currently active) and then walk through the entire list and     */
7877 /* insert them into the table.  There are two tables here: an inbound one   */
7878 /* and an outbound one.  Each NAT entry goes into each table once.          */
7879 /* ------------------------------------------------------------------------ */
7880 int
7881 ipf_nat_rehash(ipf_main_softc_t *softc, ipftuneable_t *t, ipftuneval_t *p)
7882 {
7883 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7884 	nat_t **newtab[2], *nat, **natp;
7885 	u_int *bucketlens[2];
7886 	u_int maxbucket;
7887 	u_int newsize;
7888 	int error;
7889 	u_int hv;
7890 	int i;
7891 
7892 	newsize = p->ipftu_int;
7893 	/*
7894 	 * In case there is nothing to do...
7895 	 */
7896 	if (newsize == softn->ipf_nat_table_sz)
7897 		return 0;
7898 
7899 	newtab[0] = NULL;
7900 	newtab[1] = NULL;
7901 	bucketlens[0] = NULL;
7902 	bucketlens[1] = NULL;
7903 	/*
7904 	 * 4 tables depend on the NAT table size: the inbound looking table,
7905 	 * the outbound lookup table and the hash chain length for each.
7906 	 */
7907 	KMALLOCS(newtab[0], nat_t **, newsize * sizeof(nat_t *));
7908 	if (newtab[0] == NULL) {
7909 		error = 60063;
7910 		goto badrehash;
7911 	}
7912 
7913 	KMALLOCS(newtab[1], nat_t **, newsize * sizeof(nat_t *));
7914 	if (newtab[1] == NULL) {
7915 		error = 60064;
7916 		goto badrehash;
7917 	}
7918 
7919 	KMALLOCS(bucketlens[0], u_int *, newsize * sizeof(u_int));
7920 	if (bucketlens[0] == NULL) {
7921 		error = 60065;
7922 		goto badrehash;
7923 	}
7924 
7925 	KMALLOCS(bucketlens[1], u_int *, newsize * sizeof(u_int));
7926 	if (bucketlens[1] == NULL) {
7927 		error = 60066;
7928 		goto badrehash;
7929 	}
7930 
7931 	/*
7932 	 * Recalculate the maximum length based on the new size.
7933 	 */
7934 	for (maxbucket = 0, i = newsize; i > 0; i >>= 1)
7935 		maxbucket++;
7936 	maxbucket *= 2;
7937 
7938 	bzero((char *)newtab[0], newsize * sizeof(nat_t *));
7939 	bzero((char *)newtab[1], newsize * sizeof(nat_t *));
7940 	bzero((char *)bucketlens[0], newsize * sizeof(u_int));
7941 	bzero((char *)bucketlens[1], newsize * sizeof(u_int));
7942 
7943 	WRITE_ENTER(&softc->ipf_nat);
7944 
7945 	if (softn->ipf_nat_table[0] != NULL) {
7946 		KFREES(softn->ipf_nat_table[0],
7947 		       softn->ipf_nat_table_sz *
7948 		       sizeof(*softn->ipf_nat_table[0]));
7949 	}
7950 	softn->ipf_nat_table[0] = newtab[0];
7951 
7952 	if (softn->ipf_nat_table[1] != NULL) {
7953 		KFREES(softn->ipf_nat_table[1],
7954 		       softn->ipf_nat_table_sz *
7955 		       sizeof(*softn->ipf_nat_table[1]));
7956 	}
7957 	softn->ipf_nat_table[1] = newtab[1];
7958 
7959 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen != NULL) {
7960 		KFREES(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
7961 		       softn->ipf_nat_table_sz * sizeof(u_int));
7962 	}
7963 	softn->ipf_nat_stats.ns_side[0].ns_bucketlen = bucketlens[0];
7964 
7965 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen != NULL) {
7966 		KFREES(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
7967 		       softn->ipf_nat_table_sz * sizeof(u_int));
7968 	}
7969 	softn->ipf_nat_stats.ns_side[1].ns_bucketlen = bucketlens[1];
7970 
7971 	softn->ipf_nat_maxbucket = maxbucket;
7972 	softn->ipf_nat_table_sz = newsize;
7973 	/*
7974 	 * Walk through the entire list of NAT table entries and put them
7975 	 * in the new NAT table, somewhere.  Because we have a new table,
7976 	 * we need to restart the counter of how many chains are in use.
7977 	 */
7978 	softn->ipf_nat_stats.ns_side[0].ns_inuse = 0;
7979 	softn->ipf_nat_stats.ns_side[1].ns_inuse = 0;
7980 
7981 	for (nat = softn->ipf_nat_instances; nat != NULL; nat = nat->nat_next) {
7982 		nat->nat_hnext[0] = NULL;
7983 		nat->nat_phnext[0] = NULL;
7984 		hv = nat->nat_hv[0] % softn->ipf_nat_table_sz;
7985 
7986 		natp = &softn->ipf_nat_table[0][hv];
7987 		if (*natp) {
7988 			(*natp)->nat_phnext[0] = &nat->nat_hnext[0];
7989 		} else {
7990 			NBUMPSIDE(0, ns_inuse);
7991 		}
7992 		nat->nat_phnext[0] = natp;
7993 		nat->nat_hnext[0] = *natp;
7994 		*natp = nat;
7995 		NBUMPSIDE(0, ns_bucketlen[hv]);
7996 
7997 		nat->nat_hnext[1] = NULL;
7998 		nat->nat_phnext[1] = NULL;
7999 		hv = nat->nat_hv[1] % softn->ipf_nat_table_sz;
8000 
8001 		natp = &softn->ipf_nat_table[1][hv];
8002 		if (*natp) {
8003 			(*natp)->nat_phnext[1] = &nat->nat_hnext[1];
8004 		} else {
8005 			NBUMPSIDE(1, ns_inuse);
8006 		}
8007 		nat->nat_phnext[1] = natp;
8008 		nat->nat_hnext[1] = *natp;
8009 		*natp = nat;
8010 		NBUMPSIDE(1, ns_bucketlen[hv]);
8011 	}
8012 	RWLOCK_EXIT(&softc->ipf_nat);
8013 
8014 	return 0;
8015 
8016 badrehash:
8017 	if (bucketlens[1] != NULL) {
8018 		KFREES(bucketlens[0], newsize * sizeof(u_int));
8019 	}
8020 	if (bucketlens[0] != NULL) {
8021 		KFREES(bucketlens[0], newsize * sizeof(u_int));
8022 	}
8023 	if (newtab[0] != NULL) {
8024 		KFREES(newtab[0], newsize * sizeof(nat_t *));
8025 	}
8026 	if (newtab[1] != NULL) {
8027 		KFREES(newtab[1], newsize * sizeof(nat_t *));
8028 	}
8029 	IPFERROR(error);
8030 	return ENOMEM;
8031 }
8032 
8033 
8034 /* ------------------------------------------------------------------------ */
8035 /* Function:    ipf_nat_rehash_rules                                        */
8036 /* Returns:     int  - 0 = success, else failure			    */
8037 /* Parameters:  softc(I) - pointer to soft context main structure           */
8038 /*              t(I) - pointer to tunable                                   */
8039 /*              p(I) - pointer to new tuning data                           */
8040 /*                                                                          */
8041 /* All of the NAT rules hang off of a hash table that is searched with a    */
8042 /* hash on address after the netmask is applied.  There is a different table*/
8043 /* for both inbound rules (rdr) and outbound (map.)  The resizing will only */
8044 /* affect one of these two tables.                                          */
8045 /* ------------------------------------------------------------------------ */
8046 int
8047 ipf_nat_rehash_rules(ipf_main_softc_t *softc, ipftuneable_t *t, ipftuneval_t *p)
8048 {
8049 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8050 	ipnat_t **newtab, *np, ***old, **npp;
8051 	u_int newsize;
8052 	u_int mask;
8053 	u_int hv;
8054 
8055 	newsize = p->ipftu_int;
8056 	/*
8057 	 * In case there is nothing to do...
8058 	 */
8059 	if (newsize == *t->ipft_pint)
8060 		return 0;
8061 
8062 	/*
8063 	 * All inbound rules have the NAT_REDIRECT bit set in in_redir and
8064 	 * all outbound rules have either NAT_MAP or MAT_MAPBLK set.
8065 	 * This if statement allows for some more generic code to be below,
8066 	 * rather than two huge gobs of code that almost do the same thing.
8067 	 */
8068 	if (t->ipft_pint == &softn->ipf_nat_rdrrules_sz) {
8069 		old = &softn->ipf_nat_rdr_rules;
8070 		mask = NAT_REDIRECT;
8071 	} else {
8072 		old = &softn->ipf_nat_map_rules;
8073 		mask = NAT_MAP|NAT_MAPBLK;
8074 	}
8075 
8076 	KMALLOCS(newtab, ipnat_t **, newsize * sizeof(ipnat_t *));
8077 	if (newtab == NULL) {
8078 		IPFERROR(60067);
8079 		return ENOMEM;
8080 	}
8081 
8082 	bzero((char *)newtab, newsize * sizeof(ipnat_t *));
8083 
8084 	WRITE_ENTER(&softc->ipf_nat);
8085 
8086 	if (*old != NULL) {
8087 		KFREES(*old, *t->ipft_pint * sizeof(ipnat_t **));
8088 	}
8089 	*old = newtab;
8090 	*t->ipft_pint = newsize;
8091 
8092 	for (np = softn->ipf_nat_list; np != NULL; np = np->in_next) {
8093 		if ((np->in_redir & mask) == 0)
8094 			continue;
8095 
8096 		if (np->in_redir & NAT_REDIRECT) {
8097 			np->in_rnext = NULL;
8098 			hv = np->in_hv[0] % newsize;
8099 			for (npp = newtab + hv; *npp != NULL; )
8100 				npp = &(*npp)->in_rnext;
8101 			np->in_prnext = npp;
8102 			*npp = np;
8103 		}
8104 		if (np->in_redir & NAT_MAP) {
8105 			np->in_mnext = NULL;
8106 			hv = np->in_hv[1] % newsize;
8107 			for (npp = newtab + hv; *npp != NULL; )
8108 				npp = &(*npp)->in_mnext;
8109 			np->in_pmnext = npp;
8110 			*npp = np;
8111 		}
8112 
8113 	}
8114 	RWLOCK_EXIT(&softc->ipf_nat);
8115 
8116 	return 0;
8117 }
8118 
8119 
8120 /* ------------------------------------------------------------------------ */
8121 /* Function:    ipf_nat_hostmap_rehash                                      */
8122 /* Returns:     int  - 0 = success, else failure			    */
8123 /* Parameters:  softc(I) - pointer to soft context main structure           */
8124 /*              t(I) - pointer to tunable                                   */
8125 /*              p(I) - pointer to new tuning data                           */
8126 /*                                                                          */
8127 /* Allocate and populate a new hash table that will contain a reference to  */
8128 /* all of the active IP# translations currently in place.                   */
8129 /* ------------------------------------------------------------------------ */
8130 int
8131 ipf_nat_hostmap_rehash(ipf_main_softc_t *softc, ipftuneable_t *t,
8132     ipftuneval_t *p)
8133 {
8134 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8135 	hostmap_t *hm, **newtab;
8136 	u_int newsize;
8137 	u_int hv;
8138 
8139 	newsize = p->ipftu_int;
8140 	/*
8141 	 * In case there is nothing to do...
8142 	 */
8143 	if (newsize == *t->ipft_pint)
8144 		return 0;
8145 
8146 	KMALLOCS(newtab, hostmap_t **, newsize * sizeof(hostmap_t *));
8147 	if (newtab == NULL) {
8148 		IPFERROR(60068);
8149 		return ENOMEM;
8150 	}
8151 
8152 	bzero((char *)newtab, newsize * sizeof(hostmap_t *));
8153 
8154 	WRITE_ENTER(&softc->ipf_nat);
8155 	if (softn->ipf_hm_maptable != NULL) {
8156 		KFREES(softn->ipf_hm_maptable,
8157 		       softn->ipf_nat_hostmap_sz * sizeof(hostmap_t *));
8158 	}
8159 	softn->ipf_hm_maptable = newtab;
8160 	softn->ipf_nat_hostmap_sz = newsize;
8161 
8162 	for (hm = softn->ipf_hm_maplist; hm != NULL; hm = hm->hm_next) {
8163 		hv = hm->hm_hv % softn->ipf_nat_hostmap_sz;
8164 		hm->hm_hnext = softn->ipf_hm_maptable[hv];
8165 		hm->hm_phnext = softn->ipf_hm_maptable + hv;
8166 		if (softn->ipf_hm_maptable[hv] != NULL)
8167 			softn->ipf_hm_maptable[hv]->hm_phnext = &hm->hm_hnext;
8168 		softn->ipf_hm_maptable[hv] = hm;
8169 	}
8170 	RWLOCK_EXIT(&softc->ipf_nat);
8171 
8172 	return 0;
8173 }
8174 
8175 
8176 /* ------------------------------------------------------------------------ */
8177 /* Function:    ipf_nat_add_tq                                              */
8178 /* Parameters:  softc(I) - pointer to soft context main structure           */
8179 /*                                                                          */
8180 /* ------------------------------------------------------------------------ */
8181 ipftq_t *
8182 ipf_nat_add_tq(ipf_main_softc_t *softc, int ttl)
8183 {
8184 	ipf_nat_softc_t *softs = softc->ipf_nat_soft;
8185 
8186 	return ipf_addtimeoutqueue(softc, &softs->ipf_nat_utqe, ttl);
8187 }
8188 
8189 /* ------------------------------------------------------------------------ */
8190 /* Function:    ipf_nat_uncreate                                            */
8191 /* Returns:     Nil                                                         */
8192 /* Parameters:  fin(I) - pointer to packet information                      */
8193 /*                                                                          */
8194 /* This function is used to remove a NAT entry from the NAT table when we   */
8195 /* decide that the create was actually in error. It is thus assumed that    */
8196 /* fin_flx will have both FI_NATED and FI_NATNEW set. Because we're dealing */
8197 /* with the translated packet (not the original), we have to reverse the    */
8198 /* lookup. Although doing the lookup is expensive (relatively speaking), it */
8199 /* is not anticipated that this will be a frequent occurance for normal     */
8200 /* traffic patterns.                                                        */
8201 /* ------------------------------------------------------------------------ */
8202 void
8203 ipf_nat_uncreate(fr_info_t *fin)
8204 {
8205 	ipf_main_softc_t *softc = fin->fin_main_soft;
8206 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8207 	int nflags;
8208 	nat_t *nat;
8209 
8210 	switch (fin->fin_p)
8211 	{
8212 	case IPPROTO_TCP :
8213 		nflags = IPN_TCP;
8214 		break;
8215 	case IPPROTO_UDP :
8216 		nflags = IPN_UDP;
8217 		break;
8218 	default :
8219 		nflags = 0;
8220 		break;
8221 	}
8222 
8223 	WRITE_ENTER(&softc->ipf_nat);
8224 
8225 	if (fin->fin_out == 0) {
8226 		nat = ipf_nat_outlookup(fin, nflags, (u_int)fin->fin_p,
8227 					fin->fin_dst, fin->fin_src);
8228 	} else {
8229 		nat = ipf_nat_inlookup(fin, nflags, (u_int)fin->fin_p,
8230 				       fin->fin_src, fin->fin_dst);
8231 	}
8232 
8233 	if (nat != NULL) {
8234 		NBUMPSIDE(fin->fin_out, ns_uncreate[0]);
8235 		ipf_nat_delete(softc, nat, NL_DESTROY);
8236 	} else {
8237 		NBUMPSIDE(fin->fin_out, ns_uncreate[1]);
8238 	}
8239 
8240 	RWLOCK_EXIT(&softc->ipf_nat);
8241 }
8242 
8243 
8244 /* ------------------------------------------------------------------------ */
8245 /* Function:    ipf_nat_cmp_rules                                           */
8246 /* Returns:     int   - 0 == success, else rules do not match.              */
8247 /* Parameters:  n1(I) - first rule to compare                               */
8248 /*              n2(I) - first rule to compare                               */
8249 /*                                                                          */
8250 /* Compare two rules using pointers to each rule. A straight bcmp will not  */
8251 /* work as some fields (such as in_dst, in_pkts) actually do change once    */
8252 /* the rule has been loaded into the kernel. Whilst this function returns   */
8253 /* various non-zero returns, they're strictly to aid in debugging. Use of   */
8254 /* this function should simply care if the result is zero or not.           */
8255 /* ------------------------------------------------------------------------ */
8256 static int
8257 ipf_nat_cmp_rules(ipnat_t *n1, ipnat_t *n2)
8258 {
8259 	if (n1->in_size != n2->in_size)
8260 		return 1;
8261 
8262 	if (bcmp((char *)&n1->in_v, (char *)&n2->in_v,
8263 		 offsetof(ipnat_t, in_ndst) - offsetof(ipnat_t, in_v)) != 0)
8264 		return 2;
8265 
8266 	if (bcmp((char *)&n1->in_tuc, (char *)&n2->in_tuc,
8267 		 n1->in_size - offsetof(ipnat_t, in_tuc)) != 0)
8268 		return 3;
8269 	if (n1->in_ndst.na_atype != n2->in_ndst.na_atype)
8270 		return 5;
8271 	if (n1->in_ndst.na_function != n2->in_ndst.na_function)
8272 		return 6;
8273 	if (bcmp((char *)&n1->in_ndst.na_addr, (char *)&n2->in_ndst.na_addr,
8274 		 sizeof(n1->in_ndst.na_addr)))
8275 		return 7;
8276 	if (n1->in_nsrc.na_atype != n2->in_nsrc.na_atype)
8277 		return 8;
8278 	if (n1->in_nsrc.na_function != n2->in_nsrc.na_function)
8279 		return 9;
8280 	if (bcmp((char *)&n1->in_nsrc.na_addr, (char *)&n2->in_nsrc.na_addr,
8281 		 sizeof(n1->in_nsrc.na_addr)))
8282 		return 10;
8283 	if (n1->in_odst.na_atype != n2->in_odst.na_atype)
8284 		return 11;
8285 	if (n1->in_odst.na_function != n2->in_odst.na_function)
8286 		return 12;
8287 	if (bcmp((char *)&n1->in_odst.na_addr, (char *)&n2->in_odst.na_addr,
8288 		 sizeof(n1->in_odst.na_addr)))
8289 		return 13;
8290 	if (n1->in_osrc.na_atype != n2->in_osrc.na_atype)
8291 		return 14;
8292 	if (n1->in_osrc.na_function != n2->in_osrc.na_function)
8293 		return 15;
8294 	if (bcmp((char *)&n1->in_osrc.na_addr, (char *)&n2->in_osrc.na_addr,
8295 		 sizeof(n1->in_osrc.na_addr)))
8296 		return 16;
8297 	return 0;
8298 }
8299 
8300 
8301 /* ------------------------------------------------------------------------ */
8302 /* Function:    ipf_nat_rule_init                                           */
8303 /* Returns:     int   - 0 == success, else rules do not match.              */
8304 /* Parameters:  softc(I) - pointer to soft context main structure           */
8305 /*              softn(I) - pointer to NAT context structure                 */
8306 /*              n(I)     - first rule to compare                            */
8307 /*                                                                          */
8308 /* ------------------------------------------------------------------------ */
8309 static int
8310 ipf_nat_rule_init(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *n)
8311 {
8312 	int error = 0;
8313 
8314 	if ((n->in_flags & IPN_SIPRANGE) != 0)
8315 		n->in_nsrcatype = FRI_RANGE;
8316 
8317 	if ((n->in_flags & IPN_DIPRANGE) != 0)
8318 		n->in_ndstatype = FRI_RANGE;
8319 
8320 	if ((n->in_flags & IPN_SPLIT) != 0)
8321 		n->in_ndstatype = FRI_SPLIT;
8322 
8323 	if ((n->in_redir & (NAT_MAP|NAT_REWRITE|NAT_DIVERTUDP)) != 0)
8324 		n->in_spnext = n->in_spmin;
8325 
8326 	if ((n->in_redir & (NAT_REWRITE|NAT_DIVERTUDP)) != 0) {
8327 		n->in_dpnext = n->in_dpmin;
8328 	} else if (n->in_redir == NAT_REDIRECT) {
8329 		n->in_dpnext = n->in_dpmin;
8330 	}
8331 
8332 	n->in_stepnext = 0;
8333 
8334 	switch (n->in_v[0])
8335 	{
8336 	case 4 :
8337 		error = ipf_nat_ruleaddrinit(softc, softn, n);
8338 		if (error != 0)
8339 			return error;
8340 		break;
8341 #ifdef USE_INET6
8342 	case 6 :
8343 		error = ipf_nat6_ruleaddrinit(softc, softn, n);
8344 		if (error != 0)
8345 			return error;
8346 		break;
8347 #endif
8348 	default :
8349 		break;
8350 	}
8351 
8352 	if (n->in_redir == (NAT_DIVERTUDP|NAT_MAP)) {
8353 		/*
8354 		 * Prerecord whether or not the destination of the divert
8355 		 * is local or not to the interface the packet is going
8356 		 * to be sent out.
8357 		 */
8358 		n->in_dlocal = ipf_deliverlocal(softc, n->in_v[1],
8359 						n->in_ifps[1], &n->in_ndstip6);
8360 	}
8361 
8362 	return error;
8363 }
8364 
8365 
8366 /* ------------------------------------------------------------------------ */
8367 /* Function:    ipf_nat_rule_fini                                           */
8368 /* Returns:     int   - 0 == success, else rules do not match.              */
8369 /* Parameters:  softc(I) - pointer to soft context main structure           */
8370 /*              n(I)     - rule to work on                                  */
8371 /*                                                                          */
8372 /* This function is used to release any objects that were referenced during */
8373 /* the rule initialisation. This is useful both when free'ing the rule and  */
8374 /* when handling ioctls that need to initialise these fields but not        */
8375 /* actually use them after the ioctl processing has finished.               */
8376 /* ------------------------------------------------------------------------ */
8377 static void
8378 ipf_nat_rule_fini(ipf_main_softc_t *softc, ipnat_t *n)
8379 {
8380 	if (n->in_odst.na_atype == FRI_LOOKUP && n->in_odst.na_ptr != NULL)
8381 		ipf_lookup_deref(softc, n->in_odst.na_type, n->in_odst.na_ptr);
8382 
8383 	if (n->in_osrc.na_atype == FRI_LOOKUP && n->in_osrc.na_ptr != NULL)
8384 		ipf_lookup_deref(softc, n->in_osrc.na_type, n->in_osrc.na_ptr);
8385 
8386 	if (n->in_ndst.na_atype == FRI_LOOKUP && n->in_ndst.na_ptr != NULL)
8387 		ipf_lookup_deref(softc, n->in_ndst.na_type, n->in_ndst.na_ptr);
8388 
8389 	if (n->in_nsrc.na_atype == FRI_LOOKUP && n->in_nsrc.na_ptr != NULL)
8390 		ipf_lookup_deref(softc, n->in_nsrc.na_type, n->in_nsrc.na_ptr);
8391 
8392 	if (n->in_divmp != NULL)
8393 		FREE_MB_T(n->in_divmp);
8394 }
8395