xref: /netbsd-src/sys/external/bsd/ipf/netinet/ip_nat.c (revision e670fd5c413e99c2f6a37901bb21c537fcd322d2)
1 /*	$NetBSD: ip_nat.c,v 1.24 2021/05/26 14:48:02 christos Exp $	*/
2 
3 /*
4  * Copyright (C) 2012 by Darren Reed.
5  *
6  * See the IPFILTER.LICENCE file for details on licencing.
7  */
8 #if defined(KERNEL) || defined(_KERNEL)
9 # undef KERNEL
10 # undef KERNEL
11 # define        KERNEL	1
12 # define        KERNEL	1
13 #endif
14 #include <sys/errno.h>
15 #include <sys/types.h>
16 #include <sys/param.h>
17 #include <sys/time.h>
18 #include <sys/file.h>
19 #if defined(_KERNEL) && \
20     (defined(__NetBSD_Version) && (__NetBSD_Version >= 399002000))
21 # include <sys/kauth.h>
22 #endif
23 #if !defined(_KERNEL)
24 # include <stdio.h>
25 # include <string.h>
26 # include <stdlib.h>
27 # define KERNEL
28 # ifdef _OpenBSD__
29 struct file;
30 # endif
31 # include <sys/uio.h>
32 # undef KERNEL
33 #endif
34 #if defined(_KERNEL) && \
35     defined(__FreeBSD_version) && (__FreeBSD_version >= 220000)
36 # include <sys/filio.h>
37 # include <sys/fcntl.h>
38 #else
39 # include <sys/ioctl.h>
40 #endif
41 #if !defined(AIX)
42 # include <sys/fcntl.h>
43 #endif
44 #if !defined(linux)
45 # include <sys/protosw.h>
46 #endif
47 #include <sys/socket.h>
48 #if defined(_KERNEL)
49 # include <sys/systm.h>
50 # if !defined(__SVR4) && !defined(__svr4__)
51 #  include <sys/mbuf.h>
52 # endif
53 #endif
54 #if defined(__SVR4) || defined(__svr4__)
55 # include <sys/filio.h>
56 # include <sys/byteorder.h>
57 # ifdef KERNEL
58 #  include <sys/dditypes.h>
59 # endif
60 # include <sys/stream.h>
61 # include <sys/kmem.h>
62 #endif
63 #if _FreeBSD_version >= 300000
64 # include <sys/queue.h>
65 #endif
66 #include <net/if.h>
67 #if _FreeBSD_version >= 300000
68 # include <net/if_var.h>
69 #endif
70 #ifdef sun
71 # include <net/af.h>
72 #endif
73 #include <netinet/in.h>
74 #include <netinet/in_systm.h>
75 #include <netinet/ip.h>
76 
77 #ifdef RFC1825
78 # include <vpn/md5.h>
79 # include <vpn/ipsec.h>
80 extern struct ifnet vpnif;
81 #endif
82 
83 #if !defined(linux)
84 # include <netinet/ip_var.h>
85 #endif
86 #include <netinet/tcp.h>
87 #include <netinet/udp.h>
88 #include <netinet/ip_icmp.h>
89 #include "netinet/ip_compat.h"
90 #include "netinet/ipl.h"
91 #include "netinet/ip_fil.h"
92 #include "netinet/ip_nat.h"
93 #include "netinet/ip_frag.h"
94 #include "netinet/ip_state.h"
95 #include "netinet/ip_proxy.h"
96 #include "netinet/ip_lookup.h"
97 #include "netinet/ip_dstlist.h"
98 #include "netinet/ip_sync.h"
99 #if FREEBSD_GE_REV(300000)
100 # include <sys/malloc.h>
101 #endif
102 #ifdef HAS_SYS_MD5_H
103 # include <sys/md5.h>
104 #else
105 # include "md5.h"
106 #endif
107 /* END OF INCLUDES */
108 
109 #undef	SOCKADDR_IN
110 #define	SOCKADDR_IN	struct sockaddr_in
111 
112 #if !defined(lint)
113 #if defined(__NetBSD__)
114 #include <sys/cdefs.h>
115 __KERNEL_RCSID(0, "$NetBSD: ip_nat.c,v 1.24 2021/05/26 14:48:02 christos Exp $");
116 #else
117 static const char sccsid[] = "@(#)ip_nat.c	1.11 6/5/96 (C) 1995 Darren Reed";
118 static const char rcsid[] = "@(#)Id: ip_nat.c,v 1.1.1.2 2012/07/22 13:45:27 darrenr Exp";
119 #endif
120 #endif
121 
122 
123 #define	NATFSUM(n,v,f)	((v) == 4 ? (n)->f.in4.s_addr : (n)->f.i6[0] + \
124 			 (n)->f.i6[1] + (n)->f.i6[2] + (n)->f.i6[3])
125 #define	NBUMP(x)	softn->(x)++
126 #define	NBUMPD(x, y)	do { \
127 				softn->x.y++; \
128 				DT(y); \
129 			} while (0)
130 #define	NBUMPSIDE(y,x)	softn->ipf_nat_stats.ns_side[y].x++
131 #define	NBUMPSIDED(y,x)	do { softn->ipf_nat_stats.ns_side[y].x++; \
132 			     DT(x); } while (0)
133 #define	NBUMPSIDEX(y,x,z) \
134 			do { softn->ipf_nat_stats.ns_side[y].x++; \
135 			     DT(z); } while (0)
136 #define	NBUMPSIDEDF(y,x)do { softn->ipf_nat_stats.ns_side[y].x++; \
137 			     DT1(x, fr_info_t *, fin); } while (0)
138 
139 frentry_t	ipfnatblock;
140 
141 static const ipftuneable_t ipf_nat_tuneables[] = {
142 	/* nat */
143 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_lock) },
144 		"nat_lock",	0,	1,
145 		stsizeof(ipf_nat_softc_t, ipf_nat_lock),
146 		IPFT_RDONLY,		NULL,	NULL },
147 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_sz) },
148 		"nat_table_size", 1,	0x7fffffff,
149 		stsizeof(ipf_nat_softc_t, ipf_nat_table_sz),
150 		0,			NULL,	ipf_nat_rehash },
151 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_max) },
152 		"nat_table_max", 1,	0x7fffffff,
153 		stsizeof(ipf_nat_softc_t, ipf_nat_table_max),
154 		0,			NULL,	NULL },
155 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_maprules_sz) },
156 		"nat_rules_size", 1,	0x7fffffff,
157 		stsizeof(ipf_nat_softc_t, ipf_nat_maprules_sz),
158 		0,			NULL,	ipf_nat_rehash_rules },
159 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_rdrrules_sz) },
160 		"rdr_rules_size", 1,	0x7fffffff,
161 		stsizeof(ipf_nat_softc_t, ipf_nat_rdrrules_sz),
162 		0,			NULL,	ipf_nat_rehash_rules },
163 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_hostmap_sz) },
164 		"hostmap_size",	1,	0x7fffffff,
165 		stsizeof(ipf_nat_softc_t, ipf_nat_hostmap_sz),
166 		0,			NULL,	ipf_nat_hostmap_rehash },
167 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_maxbucket) },
168 		"nat_maxbucket",1,	0x7fffffff,
169 		stsizeof(ipf_nat_softc_t, ipf_nat_maxbucket),
170 		0,			NULL,	NULL },
171 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_logging) },
172 		"nat_logging",	0,	1,
173 		stsizeof(ipf_nat_softc_t, ipf_nat_logging),
174 		0,			NULL,	NULL },
175 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_doflush) },
176 		"nat_doflush",	0,	1,
177 		stsizeof(ipf_nat_softc_t, ipf_nat_doflush),
178 		0,			NULL,	NULL },
179 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_wm_low) },
180 		"nat_table_wm_low",	1,	99,
181 		stsizeof(ipf_nat_softc_t, ipf_nat_table_wm_low),
182 		0,			NULL,	NULL },
183 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_wm_high) },
184 		"nat_table_wm_high",	2,	100,
185 		stsizeof(ipf_nat_softc_t, ipf_nat_table_wm_high),
186 		0,			NULL,	NULL },
187 	{ { 0 },
188 		NULL,			0,	0,
189 		0,
190 		0,			NULL,	NULL }
191 };
192 
193 /* ======================================================================== */
194 /* How the NAT is organised and works.                                      */
195 /*                                                                          */
196 /* Inside (interface y) NAT       Outside (interface x)                     */
197 /* -------------------- -+- -------------------------------------           */
198 /* Packet going          |   out, processsed by ipf_nat_checkout() for x    */
199 /* ------------>         |   ------------>                                  */
200 /* src=10.1.1.1          |   src=192.1.1.1                                  */
201 /*                       |                                                  */
202 /*                       |   in, processed by ipf_nat_checkin() for x       */
203 /* <------------         |   <------------                                  */
204 /* dst=10.1.1.1          |   dst=192.1.1.1                                  */
205 /* -------------------- -+- -------------------------------------           */
206 /* ipf_nat_checkout() - changes ip_src and if required, sport               */
207 /*             - creates a new mapping, if required.                        */
208 /* ipf_nat_checkin()  - changes ip_dst and if required, dport               */
209 /*                                                                          */
210 /* In the NAT table, internal source is recorded as "in" and externally     */
211 /* seen as "out".                                                           */
212 /* ======================================================================== */
213 
214 
215 #if SOLARIS && !defined(INSTANCES)
216 extern	int		pfil_delayed_copy;
217 #endif
218 
219 static	int	ipf_nat_flush_entry(ipf_main_softc_t *, void *);
220 static	int	ipf_nat_getent(ipf_main_softc_t *, void *, int);
221 static	int	ipf_nat_getsz(ipf_main_softc_t *, void *, int);
222 static	int	ipf_nat_putent(ipf_main_softc_t *, void *, int);
223 static	void	ipf_nat_addmap(ipf_nat_softc_t *, ipnat_t *);
224 static	void	ipf_nat_addrdr(ipf_nat_softc_t *, ipnat_t *);
225 static	int	ipf_nat_builddivertmp(ipf_nat_softc_t *, ipnat_t *);
226 static	int	ipf_nat_clearlist(ipf_main_softc_t *, ipf_nat_softc_t *);
227 static	int	ipf_nat_cmp_rules(ipnat_t *, ipnat_t *);
228 static	int	ipf_nat_decap(fr_info_t *, nat_t *);
229 static	void	ipf_nat_delrule(ipf_main_softc_t *, ipf_nat_softc_t *,
230 				     ipnat_t *, int);
231 static	int	ipf_nat_extraflush(ipf_main_softc_t *, ipf_nat_softc_t *, int);
232 static	int	ipf_nat_finalise(fr_info_t *, nat_t *);
233 static	int	ipf_nat_flushtable(ipf_main_softc_t *, ipf_nat_softc_t *);
234 static	int	ipf_nat_getnext(ipf_main_softc_t *, ipftoken_t *,
235 				ipfgeniter_t *, ipfobj_t *);
236 static	int	ipf_nat_gettable(ipf_main_softc_t *, ipf_nat_softc_t *, char *);
237 static	hostmap_t *ipf_nat_hostmap(ipf_nat_softc_t *, ipnat_t *,
238 				   struct in_addr, struct in_addr,
239 				   struct in_addr, u_32_t);
240 static	int	ipf_nat_icmpquerytype(int);
241 static	int	ipf_nat_iterator(ipf_main_softc_t *, ipftoken_t *,
242 				 ipfgeniter_t *, ipfobj_t *);
243 static	int	ipf_nat_match(fr_info_t *, ipnat_t *);
244 static	int	ipf_nat_matcharray(nat_t *, int *, u_long);
245 static	int	ipf_nat_matchflush(ipf_main_softc_t *, ipf_nat_softc_t *,
246 				   void *);
247 static	void	ipf_nat_mssclamp(tcphdr_t *, u_32_t, fr_info_t *, u_short *);
248 static	int	ipf_nat_newmap(fr_info_t *, nat_t *, natinfo_t *);
249 static	int	ipf_nat_newdivert(fr_info_t *, nat_t *, natinfo_t *);
250 static	int	ipf_nat_newrdr(fr_info_t *, nat_t *, natinfo_t *);
251 static	int	ipf_nat_newrewrite(fr_info_t *, nat_t *, natinfo_t *);
252 static	int	ipf_nat_nextaddr(fr_info_t *, nat_addr_t *, u_32_t *, u_32_t *);
253 static	int	ipf_nat_nextaddrinit(ipf_main_softc_t *, char *,
254 				     nat_addr_t *, int, void *);
255 static	int	ipf_nat_resolverule(ipf_main_softc_t *, ipnat_t *);
256 static	int	ipf_nat_ruleaddrinit(ipf_main_softc_t *,
257 				     ipf_nat_softc_t *, ipnat_t *);
258 static	void	ipf_nat_rule_fini(ipf_main_softc_t *, ipnat_t *);
259 static	int	ipf_nat_rule_init(ipf_main_softc_t *, ipf_nat_softc_t *,
260 				       ipnat_t *);
261 static	int	ipf_nat_siocaddnat(ipf_main_softc_t *, ipf_nat_softc_t *,
262 					ipnat_t *, int);
263 static	void	ipf_nat_siocdelnat(ipf_main_softc_t *, ipf_nat_softc_t *,
264 					ipnat_t *, int);
265 static	void	ipf_nat_tabmove(ipf_nat_softc_t *, nat_t *);
266 
267 /* ------------------------------------------------------------------------ */
268 /* Function:    ipf_nat_main_load                                           */
269 /* Returns:     int - 0 == success, -1 == failure                           */
270 /* Parameters:  Nil                                                         */
271 /*                                                                          */
272 /* The only global NAT structure that needs to be initialised is the filter */
273 /* rule that is used with blocking packets.                                 */
274 /* ------------------------------------------------------------------------ */
275 int
276 ipf_nat_main_load(void)
277 {
278 	bzero((char *)&ipfnatblock, sizeof(ipfnatblock));
279 	ipfnatblock.fr_flags = FR_BLOCK|FR_QUICK;
280 	ipfnatblock.fr_ref = 1;
281 
282 	return 0;
283 }
284 
285 
286 /* ------------------------------------------------------------------------ */
287 /* Function:    ipf_nat_main_unload                                         */
288 /* Returns:     int - 0 == success, -1 == failure                           */
289 /* Parameters:  Nil                                                         */
290 /*                                                                          */
291 /* A null-op function that exists as a placeholder so that the flow in      */
292 /* other functions is obvious.                                              */
293 /* ------------------------------------------------------------------------ */
294 int
295 ipf_nat_main_unload(void)
296 {
297 	return 0;
298 }
299 
300 
301 /* ------------------------------------------------------------------------ */
302 /* Function:    ipf_nat_soft_create                                         */
303 /* Returns:     void * - NULL = failure, else pointer to NAT context        */
304 /* Parameters:  softc(I) - pointer to soft context main structure           */
305 /*                                                                          */
306 /* Allocate the initial soft context structure for NAT and populate it with */
307 /* some default values. Creating the tables is left until we call _init so  */
308 /* that sizes can be changed before we get under way.                       */
309 /* ------------------------------------------------------------------------ */
310 void *
311 ipf_nat_soft_create(ipf_main_softc_t *softc)
312 {
313 	ipf_nat_softc_t *softn;
314 
315 	KMALLOC(softn, ipf_nat_softc_t *);
316 	if (softn == NULL)
317 		return NULL;
318 
319 	bzero((char *)softn, sizeof(*softn));
320 
321 	softn->ipf_nat_tune = ipf_tune_array_copy(softn,
322 						  sizeof(ipf_nat_tuneables),
323 						  ipf_nat_tuneables);
324 	if (softn->ipf_nat_tune == NULL) {
325 		ipf_nat_soft_destroy(softc, softn);
326 		return NULL;
327 	}
328 	if (ipf_tune_array_link(softc, softn->ipf_nat_tune) == -1) {
329 		ipf_nat_soft_destroy(softc, softn);
330 		return NULL;
331 	}
332 
333 	softn->ipf_nat_list_tail = &softn->ipf_nat_list;
334 
335 	softn->ipf_nat_table_max = NAT_TABLE_MAX;
336 	softn->ipf_nat_table_sz = NAT_TABLE_SZ;
337 	softn->ipf_nat_maprules_sz = NAT_SIZE;
338 	softn->ipf_nat_rdrrules_sz = RDR_SIZE;
339 	softn->ipf_nat_hostmap_sz = HOSTMAP_SIZE;
340 	softn->ipf_nat_doflush = 0;
341 #ifdef  IPFILTER_LOG
342 	softn->ipf_nat_logging = 1;
343 #else
344 	softn->ipf_nat_logging = 0;
345 #endif
346 
347 	softn->ipf_nat_defage = DEF_NAT_AGE;
348 	softn->ipf_nat_defipage = IPF_TTLVAL(60);
349 	softn->ipf_nat_deficmpage = IPF_TTLVAL(3);
350 	softn->ipf_nat_table_wm_high = 99;
351 	softn->ipf_nat_table_wm_low = 90;
352 
353 	return softn;
354 }
355 
356 /* ------------------------------------------------------------------------ */
357 /* Function:    ipf_nat_soft_destroy                                        */
358 /* Returns:     Nil                                                         */
359 /* Parameters:  softc(I) - pointer to soft context main structure           */
360 /*                                                                          */
361 /* ------------------------------------------------------------------------ */
362 void
363 ipf_nat_soft_destroy(ipf_main_softc_t *softc, void *arg)
364 {
365 	ipf_nat_softc_t *softn = arg;
366 
367 	if (softn->ipf_nat_tune != NULL) {
368 		ipf_tune_array_unlink(softc, softn->ipf_nat_tune);
369 		KFREES(softn->ipf_nat_tune, sizeof(ipf_nat_tuneables));
370 		softn->ipf_nat_tune = NULL;
371 	}
372 
373 	KFREE(softn);
374 }
375 
376 
377 /* ------------------------------------------------------------------------ */
378 /* Function:    ipf_nat_init                                                */
379 /* Returns:     int - 0 == success, -1 == failure                           */
380 /* Parameters:  softc(I) - pointer to soft context main structure           */
381 /*                                                                          */
382 /* Initialise all of the NAT locks, tables and other structures.            */
383 /* ------------------------------------------------------------------------ */
384 int
385 ipf_nat_soft_init(ipf_main_softc_t *softc, void *arg)
386 {
387 	ipf_nat_softc_t *softn = arg;
388 	ipftq_t *tq;
389 	int i;
390 
391 	KMALLOCS(softn->ipf_nat_table[0], nat_t **, \
392 		 sizeof(nat_t *) * softn->ipf_nat_table_sz);
393 
394 	if (softn->ipf_nat_table[0] != NULL) {
395 		bzero((char *)softn->ipf_nat_table[0],
396 		      softn->ipf_nat_table_sz * sizeof(nat_t *));
397 	} else {
398 		return -1;
399 	}
400 
401 	KMALLOCS(softn->ipf_nat_table[1], nat_t **, \
402 		 sizeof(nat_t *) * softn->ipf_nat_table_sz);
403 
404 	if (softn->ipf_nat_table[1] != NULL) {
405 		bzero((char *)softn->ipf_nat_table[1],
406 		      softn->ipf_nat_table_sz * sizeof(nat_t *));
407 	} else {
408 		return -2;
409 	}
410 
411 	KMALLOCS(softn->ipf_nat_map_rules, ipnat_t **, \
412 		 sizeof(ipnat_t *) * softn->ipf_nat_maprules_sz);
413 
414 	if (softn->ipf_nat_map_rules != NULL) {
415 		bzero((char *)softn->ipf_nat_map_rules,
416 		      softn->ipf_nat_maprules_sz * sizeof(ipnat_t *));
417 	} else {
418 		return -3;
419 	}
420 
421 	KMALLOCS(softn->ipf_nat_rdr_rules, ipnat_t **, \
422 		 sizeof(ipnat_t *) * softn->ipf_nat_rdrrules_sz);
423 
424 	if (softn->ipf_nat_rdr_rules != NULL) {
425 		bzero((char *)softn->ipf_nat_rdr_rules,
426 		      softn->ipf_nat_rdrrules_sz * sizeof(ipnat_t *));
427 	} else {
428 		return -4;
429 	}
430 
431 	KMALLOCS(softn->ipf_hm_maptable, hostmap_t **, \
432 		 sizeof(hostmap_t *) * softn->ipf_nat_hostmap_sz);
433 
434 	if (softn->ipf_hm_maptable != NULL) {
435 		bzero((char *)softn->ipf_hm_maptable,
436 		      sizeof(hostmap_t *) * softn->ipf_nat_hostmap_sz);
437 	} else {
438 		return -5;
439 	}
440 	softn->ipf_hm_maplist = NULL;
441 
442 	KMALLOCS(softn->ipf_nat_stats.ns_side[0].ns_bucketlen, u_int *,
443 		 softn->ipf_nat_table_sz * sizeof(u_int));
444 
445 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen == NULL) {
446 		return -6;
447 	}
448 	bzero((char *)softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
449 	      softn->ipf_nat_table_sz * sizeof(u_int));
450 
451 	KMALLOCS(softn->ipf_nat_stats.ns_side[1].ns_bucketlen, u_int *,
452 		 softn->ipf_nat_table_sz * sizeof(u_int));
453 
454 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen == NULL) {
455 		return -7;
456 	}
457 
458 	bzero((char *)softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
459 	      softn->ipf_nat_table_sz * sizeof(u_int));
460 
461 	if (softn->ipf_nat_maxbucket == 0) {
462 		for (i = softn->ipf_nat_table_sz; i > 0; i >>= 1)
463 			softn->ipf_nat_maxbucket++;
464 		softn->ipf_nat_maxbucket *= 2;
465 	}
466 
467 	ipf_sttab_init(softc, softn->ipf_nat_tcptq);
468 	/*
469 	 * Increase this because we may have "keep state" following this too
470 	 * and packet storms can occur if this is removed too quickly.
471 	 */
472 	softn->ipf_nat_tcptq[IPF_TCPS_CLOSED].ifq_ttl = softc->ipf_tcplastack;
473 	softn->ipf_nat_tcptq[IPF_TCP_NSTATES - 1].ifq_next =
474 							&softn->ipf_nat_udptq;
475 
476 	IPFTQ_INIT(&softn->ipf_nat_udptq, softn->ipf_nat_defage,
477 		   "nat ipftq udp tab");
478 	softn->ipf_nat_udptq.ifq_next = &softn->ipf_nat_udpacktq;
479 
480 	IPFTQ_INIT(&softn->ipf_nat_udpacktq, softn->ipf_nat_defage,
481 		   "nat ipftq udpack tab");
482 	softn->ipf_nat_udpacktq.ifq_next = &softn->ipf_nat_icmptq;
483 
484 	IPFTQ_INIT(&softn->ipf_nat_icmptq, softn->ipf_nat_deficmpage,
485 		   "nat icmp ipftq tab");
486 	softn->ipf_nat_icmptq.ifq_next = &softn->ipf_nat_icmpacktq;
487 
488 	IPFTQ_INIT(&softn->ipf_nat_icmpacktq, softn->ipf_nat_defage,
489 		   "nat icmpack ipftq tab");
490 	softn->ipf_nat_icmpacktq.ifq_next = &softn->ipf_nat_iptq;
491 
492 	IPFTQ_INIT(&softn->ipf_nat_iptq, softn->ipf_nat_defipage,
493 		   "nat ip ipftq tab");
494 	softn->ipf_nat_iptq.ifq_next = &softn->ipf_nat_pending;
495 
496 	IPFTQ_INIT(&softn->ipf_nat_pending, 1, "nat pending ipftq tab");
497 	softn->ipf_nat_pending.ifq_next = NULL;
498 
499 	for (i = 0, tq = softn->ipf_nat_tcptq; i < IPF_TCP_NSTATES; i++, tq++) {
500 		if (tq->ifq_ttl < softn->ipf_nat_deficmpage)
501 			tq->ifq_ttl = softn->ipf_nat_deficmpage;
502 #ifdef LARGE_NAT
503 		else if (tq->ifq_ttl > softn->ipf_nat_defage)
504 			tq->ifq_ttl = softn->ipf_nat_defage;
505 #endif
506 	}
507 
508 	/*
509 	 * Increase this because we may have "keep state" following
510 	 * this too and packet storms can occur if this is removed
511 	 * too quickly.
512 	 */
513 	softn->ipf_nat_tcptq[IPF_TCPS_CLOSED].ifq_ttl = softc->ipf_tcplastack;
514 
515 	MUTEX_INIT(&softn->ipf_nat_new, "ipf nat new mutex");
516 	MUTEX_INIT(&softn->ipf_nat_io, "ipf nat io mutex");
517 
518 	softn->ipf_nat_inited = 1;
519 
520 	return 0;
521 }
522 
523 
524 /* ------------------------------------------------------------------------ */
525 /* Function:    ipf_nat_soft_fini                                           */
526 /* Returns:     Nil                                                         */
527 /* Parameters:  softc(I) - pointer to soft context main structure           */
528 /*                                                                          */
529 /* Free all memory used by NAT structures allocated at runtime.             */
530 /* ------------------------------------------------------------------------ */
531 int
532 ipf_nat_soft_fini(ipf_main_softc_t *softc, void *arg)
533 {
534 	ipf_nat_softc_t *softn = arg;
535 	ipftq_t *ifq, *ifqnext;
536 
537 	(void) ipf_nat_clearlist(softc, softn);
538 	(void) ipf_nat_flushtable(softc, softn);
539 
540 	/*
541 	 * Proxy timeout queues are not cleaned here because although they
542 	 * exist on the NAT list, ipf_proxy_unload is called after unload
543 	 * and the proxies actually are responsible for them being created.
544 	 * Should the proxy timeouts have their own list?  There's no real
545 	 * justification as this is the only complication.
546 	 */
547 	for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifqnext) {
548 		ifqnext = ifq->ifq_next;
549 		if (ipf_deletetimeoutqueue(ifq) == 0)
550 			ipf_freetimeoutqueue(softc, ifq);
551 	}
552 
553 	if (softn->ipf_nat_table[0] != NULL) {
554 		KFREES(softn->ipf_nat_table[0],
555 		       sizeof(nat_t *) * softn->ipf_nat_table_sz);
556 		softn->ipf_nat_table[0] = NULL;
557 	}
558 	if (softn->ipf_nat_table[1] != NULL) {
559 		KFREES(softn->ipf_nat_table[1],
560 		       sizeof(nat_t *) * softn->ipf_nat_table_sz);
561 		softn->ipf_nat_table[1] = NULL;
562 	}
563 	if (softn->ipf_nat_map_rules != NULL) {
564 		KFREES(softn->ipf_nat_map_rules,
565 		       sizeof(ipnat_t *) * softn->ipf_nat_maprules_sz);
566 		softn->ipf_nat_map_rules = NULL;
567 	}
568 	if (softn->ipf_nat_rdr_rules != NULL) {
569 		KFREES(softn->ipf_nat_rdr_rules,
570 		       sizeof(ipnat_t *) * softn->ipf_nat_rdrrules_sz);
571 		softn->ipf_nat_rdr_rules = NULL;
572 	}
573 	if (softn->ipf_hm_maptable != NULL) {
574 		KFREES(softn->ipf_hm_maptable,
575 		       sizeof(hostmap_t *) * softn->ipf_nat_hostmap_sz);
576 		softn->ipf_hm_maptable = NULL;
577 	}
578 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen != NULL) {
579 		KFREES(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
580 		       sizeof(u_int) * softn->ipf_nat_table_sz);
581 		softn->ipf_nat_stats.ns_side[0].ns_bucketlen = NULL;
582 	}
583 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen != NULL) {
584 		KFREES(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
585 		       sizeof(u_int) * softn->ipf_nat_table_sz);
586 		softn->ipf_nat_stats.ns_side[1].ns_bucketlen = NULL;
587 	}
588 
589 	if (softn->ipf_nat_inited == 1) {
590 		softn->ipf_nat_inited = 0;
591 		ipf_sttab_destroy(softn->ipf_nat_tcptq);
592 
593 		MUTEX_DESTROY(&softn->ipf_nat_new);
594 		MUTEX_DESTROY(&softn->ipf_nat_io);
595 
596 		MUTEX_DESTROY(&softn->ipf_nat_udptq.ifq_lock);
597 		MUTEX_DESTROY(&softn->ipf_nat_udpacktq.ifq_lock);
598 		MUTEX_DESTROY(&softn->ipf_nat_icmptq.ifq_lock);
599 		MUTEX_DESTROY(&softn->ipf_nat_icmpacktq.ifq_lock);
600 		MUTEX_DESTROY(&softn->ipf_nat_iptq.ifq_lock);
601 		MUTEX_DESTROY(&softn->ipf_nat_pending.ifq_lock);
602 	}
603 
604 	return 0;
605 }
606 
607 
608 /* ------------------------------------------------------------------------ */
609 /* Function:    ipf_nat_setlock                                             */
610 /* Returns:     Nil                                                         */
611 /* Parameters:  arg(I) - pointer to soft state information                  */
612 /*              tmp(I) - new lock value                                     */
613 /*                                                                          */
614 /* Set the "lock status" of NAT to the value in tmp.                        */
615 /* ------------------------------------------------------------------------ */
616 void
617 ipf_nat_setlock(void *arg, int tmp)
618 {
619 	ipf_nat_softc_t *softn = arg;
620 
621 	softn->ipf_nat_lock = tmp;
622 }
623 
624 
625 /* ------------------------------------------------------------------------ */
626 /* Function:    ipf_nat_addrdr                                              */
627 /* Returns:     Nil                                                         */
628 /* Parameters:  n(I) - pointer to NAT rule to add                           */
629 /*                                                                          */
630 /* Adds a redirect rule to the hash table of redirect rules and the list of */
631 /* loaded NAT rules.  Updates the bitmask indicating which netmasks are in  */
632 /* use by redirect rules.                                                   */
633 /* ------------------------------------------------------------------------ */
634 static void
635 ipf_nat_addrdr(ipf_nat_softc_t *softn, ipnat_t *n)
636 {
637 	ipnat_t **np;
638 	u_32_t j;
639 	u_int hv;
640 	u_int rhv;
641 	int k;
642 
643 	if (n->in_odstatype == FRI_NORMAL) {
644 		k = count4bits(n->in_odstmsk);
645 		ipf_inet_mask_add(k, &softn->ipf_nat_rdr_mask);
646 		j = (n->in_odstaddr & n->in_odstmsk);
647 		rhv = NAT_HASH_FN(j, 0, 0xffffffff);
648 	} else {
649 		ipf_inet_mask_add(0, &softn->ipf_nat_rdr_mask);
650 		j = 0;
651 		rhv = 0;
652 	}
653 	hv = rhv % softn->ipf_nat_rdrrules_sz;
654 	np = softn->ipf_nat_rdr_rules + hv;
655 	while (*np != NULL)
656 		np = &(*np)->in_rnext;
657 	n->in_rnext = NULL;
658 	n->in_prnext = np;
659 	n->in_hv[0] = hv;
660 	n->in_use++;
661 	*np = n;
662 }
663 
664 
665 /* ------------------------------------------------------------------------ */
666 /* Function:    ipf_nat_addmap                                              */
667 /* Returns:     Nil                                                         */
668 /* Parameters:  n(I) - pointer to NAT rule to add                           */
669 /*                                                                          */
670 /* Adds a NAT map rule to the hash table of rules and the list of  loaded   */
671 /* NAT rules.  Updates the bitmask indicating which netmasks are in use by  */
672 /* redirect rules.                                                          */
673 /* ------------------------------------------------------------------------ */
674 static void
675 ipf_nat_addmap(ipf_nat_softc_t *softn, ipnat_t *n)
676 {
677 	ipnat_t **np;
678 	u_32_t j;
679 	u_int hv;
680 	u_int rhv;
681 	int k;
682 
683 	if (n->in_osrcatype == FRI_NORMAL) {
684 		k = count4bits(n->in_osrcmsk);
685 		ipf_inet_mask_add(k, &softn->ipf_nat_map_mask);
686 		j = (n->in_osrcaddr & n->in_osrcmsk);
687 		rhv = NAT_HASH_FN(j, 0, 0xffffffff);
688 	} else {
689 		ipf_inet_mask_add(0, &softn->ipf_nat_map_mask);
690 		j = 0;
691 		rhv = 0;
692 	}
693 	hv = rhv % softn->ipf_nat_maprules_sz;
694 	np = softn->ipf_nat_map_rules + hv;
695 	while (*np != NULL)
696 		np = &(*np)->in_mnext;
697 	n->in_mnext = NULL;
698 	n->in_pmnext = np;
699 	n->in_hv[1] = rhv;
700 	n->in_use++;
701 	*np = n;
702 }
703 
704 
705 /* ------------------------------------------------------------------------ */
706 /* Function:    ipf_nat_delrdr                                              */
707 /* Returns:     Nil                                                         */
708 /* Parameters:  n(I) - pointer to NAT rule to delete                        */
709 /*                                                                          */
710 /* Removes a redirect rule from the hash table of redirect rules.           */
711 /* ------------------------------------------------------------------------ */
712 void
713 ipf_nat_delrdr(ipf_nat_softc_t *softn, ipnat_t *n)
714 {
715 	if (n->in_odstatype == FRI_NORMAL) {
716 		int k = count4bits(n->in_odstmsk);
717 		ipf_inet_mask_del(k, &softn->ipf_nat_rdr_mask);
718 	} else {
719 		ipf_inet_mask_del(0, &softn->ipf_nat_rdr_mask);
720 	}
721 	if (n->in_rnext)
722 		n->in_rnext->in_prnext = n->in_prnext;
723 	*n->in_prnext = n->in_rnext;
724 	n->in_use--;
725 }
726 
727 
728 /* ------------------------------------------------------------------------ */
729 /* Function:    ipf_nat_delmap                                              */
730 /* Returns:     Nil                                                         */
731 /* Parameters:  n(I) - pointer to NAT rule to delete                        */
732 /*                                                                          */
733 /* Removes a NAT map rule from the hash table of NAT map rules.             */
734 /* ------------------------------------------------------------------------ */
735 void
736 ipf_nat_delmap(ipf_nat_softc_t *softn, ipnat_t *n)
737 {
738 	if (n->in_osrcatype == FRI_NORMAL) {
739 		int k = count4bits(n->in_osrcmsk);
740 		ipf_inet_mask_del(k, &softn->ipf_nat_map_mask);
741 	} else {
742 		ipf_inet_mask_del(0, &softn->ipf_nat_map_mask);
743 	}
744 	if (n->in_mnext != NULL)
745 		n->in_mnext->in_pmnext = n->in_pmnext;
746 	*n->in_pmnext = n->in_mnext;
747 	n->in_use--;
748 }
749 
750 
751 /* ------------------------------------------------------------------------ */
752 /* Function:    ipf_nat_hostmap                                             */
753 /* Returns:     struct hostmap* - NULL if no hostmap could be created,      */
754 /*                                else a pointer to the hostmapping to use  */
755 /* Parameters:  np(I)   - pointer to NAT rule                               */
756 /*              real(I) - real IP address                                   */
757 /*              map(I)  - mapped IP address                                 */
758 /*              port(I) - destination port number                           */
759 /* Write Locks: ipf_nat                                                     */
760 /*                                                                          */
761 /* Check if an ip address has already been allocated for a given mapping    */
762 /* that is not doing port based translation.  If is not yet allocated, then */
763 /* create a new entry if a non-NULL NAT rule pointer has been supplied.     */
764 /* ------------------------------------------------------------------------ */
765 static struct hostmap *
766 ipf_nat_hostmap(ipf_nat_softc_t *softn, ipnat_t *np, struct in_addr src,
767     struct in_addr dst, struct in_addr map, u_32_t port)
768 {
769 	hostmap_t *hm;
770 	u_int hv, rhv;
771 
772 	hv = (src.s_addr ^ dst.s_addr);
773 	hv += src.s_addr;
774 	hv += dst.s_addr;
775 	rhv = hv;
776 	hv %= softn->ipf_nat_hostmap_sz;
777 	for (hm = softn->ipf_hm_maptable[hv]; hm; hm = hm->hm_hnext)
778 		if ((hm->hm_osrcip.s_addr == src.s_addr) &&
779 		    (hm->hm_odstip.s_addr == dst.s_addr) &&
780 		    ((np == NULL) || (np == hm->hm_ipnat)) &&
781 		    ((port == 0) || (port == hm->hm_port))) {
782 			softn->ipf_nat_stats.ns_hm_addref++;
783 			hm->hm_ref++;
784 			return hm;
785 		}
786 
787 	if (np == NULL) {
788 		softn->ipf_nat_stats.ns_hm_nullnp++;
789 		return NULL;
790 	}
791 
792 	KMALLOC(hm, hostmap_t *);
793 	if (hm) {
794 		hm->hm_next = softn->ipf_hm_maplist;
795 		hm->hm_pnext = &softn->ipf_hm_maplist;
796 		if (softn->ipf_hm_maplist != NULL)
797 			softn->ipf_hm_maplist->hm_pnext = &hm->hm_next;
798 		softn->ipf_hm_maplist = hm;
799 		hm->hm_hnext = softn->ipf_hm_maptable[hv];
800 		hm->hm_phnext = softn->ipf_hm_maptable + hv;
801 		if (softn->ipf_hm_maptable[hv] != NULL)
802 			softn->ipf_hm_maptable[hv]->hm_phnext = &hm->hm_hnext;
803 		softn->ipf_hm_maptable[hv] = hm;
804 		hm->hm_ipnat = np;
805 		np->in_use++;
806 		hm->hm_osrcip = src;
807 		hm->hm_odstip = dst;
808 		hm->hm_nsrcip = map;
809 		hm->hm_ndstip.s_addr = 0;
810 		hm->hm_ref = 1;
811 		hm->hm_port = port;
812 		hm->hm_hv = rhv;
813 		hm->hm_v = 4;
814 		softn->ipf_nat_stats.ns_hm_new++;
815 	} else {
816 		softn->ipf_nat_stats.ns_hm_newfail++;
817 	}
818 	return hm;
819 }
820 
821 
822 /* ------------------------------------------------------------------------ */
823 /* Function:    ipf_nat_hostmapdel                                          */
824 /* Returns:     Nil                                                         */
825 /* Parameters:  hmp(I) - pointer to hostmap structure pointer               */
826 /* Write Locks: ipf_nat                                                     */
827 /*                                                                          */
828 /* Decrement the references to this hostmap structure by one.  If this      */
829 /* reaches zero then remove it and free it.                                 */
830 /* ------------------------------------------------------------------------ */
831 void
832 ipf_nat_hostmapdel(ipf_main_softc_t *softc, struct hostmap **hmp)
833 {
834 	struct hostmap *hm;
835 
836 	hm = *hmp;
837 	*hmp = NULL;
838 
839 	hm->hm_ref--;
840 	if (hm->hm_ref == 0) {
841 		ipf_nat_rule_deref(softc, &hm->hm_ipnat);
842 		if (hm->hm_hnext)
843 			hm->hm_hnext->hm_phnext = hm->hm_phnext;
844 		*hm->hm_phnext = hm->hm_hnext;
845 		if (hm->hm_next)
846 			hm->hm_next->hm_pnext = hm->hm_pnext;
847 		*hm->hm_pnext = hm->hm_next;
848 		KFREE(hm);
849 	}
850 }
851 
852 
853 /* ------------------------------------------------------------------------ */
854 /* Function:    ipf_fix_outcksum                                            */
855 /* Returns:     Nil                                                         */
856 /* Parameters:  fin(I) - pointer to packet information                      */
857 /*              sp(I)  - location of 16bit checksum to update               */
858 /*              n((I)  - amount to adjust checksum by                       */
859 /*                                                                          */
860 /* Adjusts the 16bit checksum by "n" for packets going out.                 */
861 /* ------------------------------------------------------------------------ */
862 void
863 ipf_fix_outcksum(int cksum, u_short *sp, u_32_t n, u_32_t partial)
864 {
865 	u_short sumshort;
866 	u_32_t sum1;
867 
868 	if (n == 0)
869 		return;
870 
871 	if (cksum == 4) {
872 		*sp = 0;
873 		return;
874 	}
875 	if (cksum == 2) {
876 		sum1 = partial;
877 		sum1 = (sum1 & 0xffff) + (sum1 >> 16);
878 		*sp = htons(sum1);
879 		return;
880 	}
881 	sum1 = (~ntohs(*sp)) & 0xffff;
882 	sum1 += (n);
883 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
884 	/* Again */
885 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
886 	sumshort = ~(u_short)sum1;
887 	*(sp) = htons(sumshort);
888 }
889 
890 
891 /* ------------------------------------------------------------------------ */
892 /* Function:    ipf_fix_incksum                                             */
893 /* Returns:     Nil                                                         */
894 /* Parameters:  fin(I) - pointer to packet information                      */
895 /*              sp(I)  - location of 16bit checksum to update               */
896 /*              n((I)  - amount to adjust checksum by                       */
897 /*                                                                          */
898 /* Adjusts the 16bit checksum by "n" for packets going in.                  */
899 /* ------------------------------------------------------------------------ */
900 void
901 ipf_fix_incksum(int cksum, u_short *sp, u_32_t n, u_32_t partial)
902 {
903 	u_short sumshort;
904 	u_32_t sum1;
905 
906 	if (n == 0)
907 		return;
908 
909 	if (cksum == 4) {
910 		*sp = 0;
911 		return;
912 	}
913 	if (cksum == 2) {
914 		sum1 = partial;
915 		sum1 = (sum1 & 0xffff) + (sum1 >> 16);
916 		*sp = htons(sum1);
917 		return;
918 	}
919 
920 	sum1 = (~ntohs(*sp)) & 0xffff;
921 	sum1 += ~(n) & 0xffff;
922 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
923 	/* Again */
924 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
925 	sumshort = ~(u_short)sum1;
926 	*(sp) = htons(sumshort);
927 }
928 
929 
930 /* ------------------------------------------------------------------------ */
931 /* Function:    ipf_fix_datacksum                                           */
932 /* Returns:     Nil                                                         */
933 /* Parameters:  sp(I)  - location of 16bit checksum to update               */
934 /*              n((I)  - amount to adjust checksum by                       */
935 /*                                                                          */
936 /* Fix_datacksum is used *only* for the adjustments of checksums in the     */
937 /* data section of an IP packet.                                            */
938 /*                                                                          */
939 /* The only situation in which you need to do this is when NAT'ing an       */
940 /* ICMP error message. Such a message, contains in its body the IP header   */
941 /* of the original IP packet, that causes the error.                        */
942 /*                                                                          */
943 /* You can't use fix_incksum or fix_outcksum in that case, because for the  */
944 /* kernel the data section of the ICMP error is just data, and no special   */
945 /* processing like hardware cksum or ntohs processing have been done by the */
946 /* kernel on the data section.                                              */
947 /* ------------------------------------------------------------------------ */
948 void
949 ipf_fix_datacksum(u_short *sp, u_32_t n)
950 {
951 	u_short sumshort;
952 	u_32_t sum1;
953 
954 	if (n == 0)
955 		return;
956 
957 	sum1 = (~ntohs(*sp)) & 0xffff;
958 	sum1 += (n);
959 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
960 	/* Again */
961 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
962 	sumshort = ~(u_short)sum1;
963 	*(sp) = htons(sumshort);
964 }
965 
966 
967 /* ------------------------------------------------------------------------ */
968 /* Function:    ipf_nat_ioctl                                               */
969 /* Returns:     int - 0 == success, != 0 == failure                         */
970 /* Parameters:  softc(I) - pointer to soft context main structure           */
971 /*              data(I)  - pointer to ioctl data                            */
972 /*              cmd(I)   - ioctl command integer                            */
973 /*              mode(I)  - file mode bits used with open                    */
974 /*              uid(I)   - uid of calling process                           */
975 /*              ctx(I)   - pointer used as key for finding context          */
976 /*                                                                          */
977 /* Processes an ioctl call made to operate on the IP Filter NAT device.     */
978 /* ------------------------------------------------------------------------ */
979 int
980 ipf_nat_ioctl(ipf_main_softc_t *softc, void *data, ioctlcmd_t cmd, int mode,
981    int uid, void *ctx)
982 {
983 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
984 	int error = 0, ret, arg, getlock;
985 	ipnat_t *nat, *nt, *n;
986 	ipnat_t *natd = NULL;
987 	SPL_INT(s);
988 
989 #if BSD_GE_YEAR(199306) && defined(_KERNEL)
990 # if NETBSD_GE_REV(399002000)
991 	if ((mode & FWRITE) &&
992 	     kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_FIREWALL,
993 				     KAUTH_REQ_NETWORK_FIREWALL_FW,
994 				     NULL, NULL, NULL))
995 # else
996 #  if defined(__FreeBSD_version) && (__FreeBSD_version >= 500034)
997 	if (securelevel_ge(curthread->td_ucred, 3) && (mode & FWRITE))
998 #  else
999 	if ((securelevel >= 3) && (mode & FWRITE))
1000 #  endif
1001 # endif
1002 	{
1003 		IPFERROR(60001);
1004 		return EPERM;
1005 	}
1006 #endif
1007 
1008 #if defined(__osf__) && defined(_KERNEL)
1009 	getlock = 0;
1010 #else
1011 	getlock = (mode & NAT_LOCKHELD) ? 0 : 1;
1012 #endif
1013 
1014 	n = NULL;
1015 	nt = NULL;
1016 	nat = NULL;
1017 
1018 	if ((cmd == (ioctlcmd_t)SIOCADNAT) || (cmd == (ioctlcmd_t)SIOCRMNAT) ||
1019 	    (cmd == (ioctlcmd_t)SIOCPURGENAT)) {
1020 		KMALLOC(natd, ipnat_t *);
1021 		if (natd == NULL) {
1022 			error = ENOMEM;
1023 			goto done;
1024 		}
1025 		if (mode & NAT_SYSSPACE) {
1026 			bcopy(data, natd, sizeof(*natd));
1027 			nat = natd;
1028 			error = 0;
1029 		} else {
1030 			bzero(natd, sizeof(*natd));
1031 			error = ipf_inobj(softc, data, NULL, natd,
1032 					  IPFOBJ_IPNAT);
1033 			if (error != 0)
1034 				goto done;
1035 
1036 			if (natd->in_size < sizeof(ipnat_t)) {
1037 				error = EINVAL;
1038 				goto done;
1039 			}
1040 			KMALLOCS(nt, ipnat_t *, natd->in_size);
1041 			if (nt == NULL) {
1042 				IPFERROR(60070);
1043 				error = ENOMEM;
1044 				goto done;
1045 			}
1046 			bzero(nt, natd->in_size);
1047 			error = ipf_inobjsz(softc, data, nt, IPFOBJ_IPNAT,
1048 					    natd->in_size);
1049 			if (error)
1050 				goto done;
1051 			nat = nt;
1052 		}
1053 
1054 		/*
1055 		 * For add/delete, look to see if the NAT entry is
1056 		 * already present
1057 		 */
1058 		nat->in_flags &= IPN_USERFLAGS;
1059 		if ((nat->in_redir & NAT_MAPBLK) == 0) {
1060 			if (nat->in_osrcatype == FRI_NORMAL ||
1061 			    nat->in_osrcatype == FRI_NONE)
1062 				nat->in_osrcaddr &= nat->in_osrcmsk;
1063 			if (nat->in_odstatype == FRI_NORMAL ||
1064 			    nat->in_odstatype == FRI_NONE)
1065 				nat->in_odstaddr &= nat->in_odstmsk;
1066 			if ((nat->in_flags & (IPN_SPLIT|IPN_SIPRANGE)) == 0) {
1067 				if (nat->in_nsrcatype == FRI_NORMAL)
1068 					nat->in_nsrcaddr &= nat->in_nsrcmsk;
1069 				if (nat->in_ndstatype == FRI_NORMAL)
1070 					nat->in_ndstaddr &= nat->in_ndstmsk;
1071 			}
1072 		}
1073 
1074 		error = ipf_nat_rule_init(softc, softn, nat);
1075 		if (error != 0)
1076 			goto done;
1077 
1078 		MUTEX_ENTER(&softn->ipf_nat_io);
1079 		for (n = softn->ipf_nat_list; n != NULL; n = n->in_next)
1080 			if (ipf_nat_cmp_rules(nat, n) == 0)
1081 				break;
1082 	}
1083 
1084 	switch (cmd)
1085 	{
1086 #ifdef  IPFILTER_LOG
1087 	case SIOCIPFFB :
1088 	{
1089 		int tmp;
1090 
1091 		if (!(mode & FWRITE)) {
1092 			IPFERROR(60002);
1093 			error = EPERM;
1094 		} else {
1095 			tmp = ipf_log_clear(softc, IPL_LOGNAT);
1096 			error = BCOPYOUT(&tmp, data, sizeof(tmp));
1097 			if (error != 0) {
1098 				IPFERROR(60057);
1099 				error = EFAULT;
1100 			}
1101 		}
1102 		break;
1103 	}
1104 
1105 	case SIOCSETLG :
1106 		if (!(mode & FWRITE)) {
1107 			IPFERROR(60003);
1108 			error = EPERM;
1109 		} else {
1110 			error = BCOPYIN(data, &softn->ipf_nat_logging,
1111 					sizeof(softn->ipf_nat_logging));
1112 			if (error != 0)
1113 				error = EFAULT;
1114 		}
1115 		break;
1116 
1117 	case SIOCGETLG :
1118 		error = BCOPYOUT(&softn->ipf_nat_logging, data,
1119 				 sizeof(softn->ipf_nat_logging));
1120 		if (error != 0) {
1121 			IPFERROR(60004);
1122 			error = EFAULT;
1123 		}
1124 		break;
1125 
1126 	case FIONREAD :
1127 		arg = ipf_log_bytesused(softc, IPL_LOGNAT);
1128 		error = BCOPYOUT(&arg, data, sizeof(arg));
1129 		if (error != 0) {
1130 			IPFERROR(60005);
1131 			error = EFAULT;
1132 		}
1133 		break;
1134 #endif
1135 	case SIOCADNAT :
1136 		if (!(mode & FWRITE)) {
1137 			IPFERROR(60006);
1138 			error = EPERM;
1139 		} else if (n != NULL) {
1140 			KMALLOC(natd, ipnat_t *);
1141 			if (natd == NULL) {
1142 				error = ENOMEM;
1143 				goto done;
1144 			}
1145 			bzero(natd, sizeof(*natd));
1146 			natd->in_flineno = n->in_flineno;
1147 			(void) ipf_outobj(softc, data, &natd, IPFOBJ_IPNAT);
1148 			IPFERROR(60007);
1149 			error = EEXIST;
1150 		} else if (nt == NULL) {
1151 			IPFERROR(60008);
1152 			error = ENOMEM;
1153 		}
1154 		if (error != 0) {
1155 			MUTEX_EXIT(&softn->ipf_nat_io);
1156 			break;
1157 		}
1158 		if (nat != nt)
1159 			bcopy((char *)nat, (char *)nt, sizeof(*n));
1160 		error = ipf_nat_siocaddnat(softc, softn, nt, getlock);
1161 		MUTEX_EXIT(&softn->ipf_nat_io);
1162 		if (error == 0) {
1163 			nat = NULL;
1164 			nt = NULL;
1165 		}
1166 		break;
1167 
1168 	case SIOCRMNAT :
1169 	case SIOCPURGENAT :
1170 		if (!(mode & FWRITE)) {
1171 			IPFERROR(60009);
1172 			error = EPERM;
1173 			n = NULL;
1174 		} else if (n == NULL) {
1175 			IPFERROR(60010);
1176 			error = ESRCH;
1177 		}
1178 
1179 		if (error != 0) {
1180 			MUTEX_EXIT(&softn->ipf_nat_io);
1181 			break;
1182 		}
1183 		if (cmd == (ioctlcmd_t)SIOCPURGENAT) {
1184 			error = ipf_outobjsz(softc, data, n, IPFOBJ_IPNAT,
1185 					     n->in_size);
1186 			if (error) {
1187 				MUTEX_EXIT(&softn->ipf_nat_io);
1188 				goto done;
1189 			}
1190 			n->in_flags |= IPN_PURGE;
1191 		}
1192 		ipf_nat_siocdelnat(softc, softn, n, getlock);
1193 
1194 		MUTEX_EXIT(&softn->ipf_nat_io);
1195 		n = NULL;
1196 		break;
1197 
1198 	case SIOCGNATS :
1199 	    {
1200 		natstat_t *nsp = &softn->ipf_nat_stats;
1201 
1202 		nsp->ns_side[0].ns_table = softn->ipf_nat_table[0];
1203 		nsp->ns_side[1].ns_table = softn->ipf_nat_table[1];
1204 		nsp->ns_list = softn->ipf_nat_list;
1205 		nsp->ns_maptable = softn->ipf_hm_maptable;
1206 		nsp->ns_maplist = softn->ipf_hm_maplist;
1207 		nsp->ns_nattab_sz = softn->ipf_nat_table_sz;
1208 		nsp->ns_nattab_max = softn->ipf_nat_table_max;
1209 		nsp->ns_rultab_sz = softn->ipf_nat_maprules_sz;
1210 		nsp->ns_rdrtab_sz = softn->ipf_nat_rdrrules_sz;
1211 		nsp->ns_hostmap_sz = softn->ipf_nat_hostmap_sz;
1212 		nsp->ns_instances = softn->ipf_nat_instances;
1213 		nsp->ns_ticks = softc->ipf_ticks;
1214 #ifdef IPFILTER_LOGGING
1215 		nsp->ns_log_ok = ipf_log_logok(softc, IPF_LOGNAT);
1216 		nsp->ns_log_fail = ipf_log_failures(softc, IPF_LOGNAT);
1217 #else
1218 		nsp->ns_log_ok = 0;
1219 		nsp->ns_log_fail = 0;
1220 #endif
1221 		error = ipf_outobj(softc, data, nsp, IPFOBJ_NATSTAT);
1222 		break;
1223 	    }
1224 
1225 	case SIOCGNATL :
1226 	    {
1227 		natlookup_t nl;
1228 
1229 		error = ipf_inobj(softc, data, NULL, &nl, IPFOBJ_NATLOOKUP);
1230 		if (error == 0) {
1231 			void *ptr;
1232 
1233 			if (getlock) {
1234 				READ_ENTER(&softc->ipf_nat);
1235 			}
1236 
1237 			switch (nl.nl_v)
1238 			{
1239 			case 4 :
1240 				ptr = ipf_nat_lookupredir(softc, &nl);
1241 				break;
1242 #ifdef USE_INET6
1243 			case 6 :
1244 				ptr = ipf_nat6_lookupredir(softc, &nl);
1245 				break;
1246 #endif
1247 			default:
1248 				ptr = NULL;
1249 				break;
1250 			}
1251 
1252 			if (getlock) {
1253 				RWLOCK_EXIT(&softc->ipf_nat);
1254 			}
1255 			if (ptr != NULL) {
1256 				error = ipf_outobj(softc, data, &nl,
1257 						   IPFOBJ_NATLOOKUP);
1258 			} else {
1259 				IPFERROR(60011);
1260 				error = ESRCH;
1261 			}
1262 		}
1263 		break;
1264 	    }
1265 
1266 	case SIOCIPFFL :	/* old SIOCFLNAT & SIOCCNATL */
1267 		if (!(mode & FWRITE)) {
1268 			IPFERROR(60012);
1269 			error = EPERM;
1270 			break;
1271 		}
1272 		if (getlock) {
1273 			WRITE_ENTER(&softc->ipf_nat);
1274 		}
1275 
1276 		error = BCOPYIN(data, &arg, sizeof(arg));
1277 		if (error != 0) {
1278 			IPFERROR(60013);
1279 			error = EFAULT;
1280 		} else {
1281 			if (arg == 0)
1282 				ret = ipf_nat_flushtable(softc, softn);
1283 			else if (arg == 1)
1284 				ret = ipf_nat_clearlist(softc, softn);
1285 			else
1286 				ret = ipf_nat_extraflush(softc, softn, arg);
1287 			ipf_proxy_flush(softc->ipf_proxy_soft, arg);
1288 		}
1289 
1290 		if (getlock) {
1291 			RWLOCK_EXIT(&softc->ipf_nat);
1292 		}
1293 		if (error == 0) {
1294 			error = BCOPYOUT(&ret, data, sizeof(ret));
1295 		}
1296 		break;
1297 
1298 	case SIOCMATCHFLUSH :
1299 		if (!(mode & FWRITE)) {
1300 			IPFERROR(60014);
1301 			error = EPERM;
1302 			break;
1303 		}
1304 		if (getlock) {
1305 			WRITE_ENTER(&softc->ipf_nat);
1306 		}
1307 
1308 		error = ipf_nat_matchflush(softc, softn, data);
1309 
1310 		if (getlock) {
1311 			RWLOCK_EXIT(&softc->ipf_nat);
1312 		}
1313 		break;
1314 
1315 	case SIOCPROXY :
1316 		error = ipf_proxy_ioctl(softc, data, cmd, mode, ctx);
1317 		break;
1318 
1319 	case SIOCSTLCK :
1320 		if (!(mode & FWRITE)) {
1321 			IPFERROR(60015);
1322 			error = EPERM;
1323 		} else {
1324 			error = ipf_lock(data, &softn->ipf_nat_lock);
1325 		}
1326 		break;
1327 
1328 	case SIOCSTPUT :
1329 		if ((mode & FWRITE) != 0) {
1330 			error = ipf_nat_putent(softc, data, getlock);
1331 		} else {
1332 			IPFERROR(60016);
1333 			error = EACCES;
1334 		}
1335 		break;
1336 
1337 	case SIOCSTGSZ :
1338 		if (softn->ipf_nat_lock) {
1339 			error = ipf_nat_getsz(softc, data, getlock);
1340 		} else {
1341 			IPFERROR(60017);
1342 			error = EACCES;
1343 		}
1344 		break;
1345 
1346 	case SIOCSTGET :
1347 		if (softn->ipf_nat_lock) {
1348 			error = ipf_nat_getent(softc, data, getlock);
1349 		} else {
1350 			IPFERROR(60018);
1351 			error = EACCES;
1352 		}
1353 		break;
1354 
1355 	case SIOCGENITER :
1356 	    {
1357 		ipfgeniter_t iter;
1358 		ipftoken_t *token;
1359 		ipfobj_t obj;
1360 
1361 		error = ipf_inobj(softc, data, &obj, &iter, IPFOBJ_GENITER);
1362 		if (error != 0)
1363 			break;
1364 
1365 		SPL_SCHED(s);
1366 		token = ipf_token_find(softc, iter.igi_type, uid, ctx);
1367 		if (token != NULL) {
1368 			error  = ipf_nat_iterator(softc, token, &iter, &obj);
1369 			WRITE_ENTER(&softc->ipf_tokens);
1370 			ipf_token_deref(softc, token);
1371 			RWLOCK_EXIT(&softc->ipf_tokens);
1372 		}
1373 		SPL_X(s);
1374 		break;
1375 	    }
1376 
1377 	case SIOCIPFDELTOK :
1378 		error = BCOPYIN(data, &arg, sizeof(arg));
1379 		if (error == 0) {
1380 			SPL_SCHED(s);
1381 			error = ipf_token_del(softc, arg, uid, ctx);
1382 			SPL_X(s);
1383 		} else {
1384 			IPFERROR(60019);
1385 			error = EFAULT;
1386 		}
1387 		break;
1388 
1389 	case SIOCGTQTAB :
1390 		error = ipf_outobj(softc, data, softn->ipf_nat_tcptq,
1391 				   IPFOBJ_STATETQTAB);
1392 		break;
1393 
1394 	case SIOCGTABL :
1395 		error = ipf_nat_gettable(softc, softn, data);
1396 		break;
1397 
1398 	default :
1399 		IPFERROR(60020);
1400 		error = EINVAL;
1401 		break;
1402 	}
1403 done:
1404 	if (nat != NULL)
1405 		ipf_nat_rule_fini(softc, nat);
1406 	if (natd != NULL)
1407 		KFREE(natd);
1408 	if (nt != NULL)
1409 		KFREES(nt, nt->in_size);
1410 	return error;
1411 }
1412 
1413 
1414 /* ------------------------------------------------------------------------ */
1415 /* Function:    ipf_nat_siocaddnat                                          */
1416 /* Returns:     int - 0 == success, != 0 == failure                         */
1417 /* Parameters:  softc(I) - pointer to soft context main structure           */
1418 /*              softn(I) - pointer to NAT context structure                 */
1419 /*              n(I)       - pointer to new NAT rule                        */
1420 /*              np(I)      - pointer to where to insert new NAT rule        */
1421 /*              getlock(I) - flag indicating if lock on  is held            */
1422 /* Mutex Locks: ipf_nat_io                                                  */
1423 /*                                                                          */
1424 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
1425 /* from information passed to the kernel, then add it  to the appropriate   */
1426 /* NAT rule table(s).                                                       */
1427 /* ------------------------------------------------------------------------ */
1428 static int
1429 ipf_nat_siocaddnat(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *n,
1430 	int getlock)
1431 {
1432 	int error = 0;
1433 
1434 	if (ipf_nat_resolverule(softc, n) != 0) {
1435 		IPFERROR(60022);
1436 		return ENOENT;
1437 	}
1438 
1439 	if ((n->in_age[0] == 0) && (n->in_age[1] != 0)) {
1440 		IPFERROR(60023);
1441 		return EINVAL;
1442 	}
1443 
1444 	if (n->in_redir == (NAT_DIVERTUDP|NAT_MAP)) {
1445 		/*
1446 		 * Prerecord whether or not the destination of the divert
1447 		 * is local or not to the interface the packet is going
1448 		 * to be sent out.
1449 		 */
1450 		n->in_dlocal = ipf_deliverlocal(softc, n->in_v[1],
1451 						n->in_ifps[1], &n->in_ndstip6);
1452 	}
1453 
1454 	if (getlock) {
1455 		WRITE_ENTER(&softc->ipf_nat);
1456 	}
1457 	n->in_next = NULL;
1458 	n->in_pnext = softn->ipf_nat_list_tail;
1459 	*n->in_pnext = n;
1460 	softn->ipf_nat_list_tail = &n->in_next;
1461 	n->in_use++;
1462 
1463 	if (n->in_redir & NAT_REDIRECT) {
1464 		n->in_flags &= ~IPN_NOTDST;
1465 		switch (n->in_v[0])
1466 		{
1467 		case 4 :
1468 			ipf_nat_addrdr(softn, n);
1469 			break;
1470 #ifdef USE_INET6
1471 		case 6 :
1472 			ipf_nat6_addrdr(softn, n);
1473 			break;
1474 #endif
1475 		default :
1476 			break;
1477 		}
1478 		ATOMIC_INC32(softn->ipf_nat_stats.ns_rules_rdr);
1479 	}
1480 
1481 	if (n->in_redir & (NAT_MAP|NAT_MAPBLK)) {
1482 		n->in_flags &= ~IPN_NOTSRC;
1483 		switch (n->in_v[0])
1484 		{
1485 		case 4 :
1486 			ipf_nat_addmap(softn, n);
1487 			break;
1488 #ifdef USE_INET6
1489 		case 6 :
1490 			ipf_nat6_addmap(softn, n);
1491 			break;
1492 #endif
1493 		default :
1494 			break;
1495 		}
1496 		ATOMIC_INC32(softn->ipf_nat_stats.ns_rules_map);
1497 	}
1498 
1499 	if (n->in_age[0] != 0)
1500 		n->in_tqehead[0] = ipf_addtimeoutqueue(softc,
1501 						       &softn->ipf_nat_utqe,
1502 						       n->in_age[0]);
1503 
1504 	if (n->in_age[1] != 0)
1505 		n->in_tqehead[1] = ipf_addtimeoutqueue(softc,
1506 						       &softn->ipf_nat_utqe,
1507 						       n->in_age[1]);
1508 
1509 	MUTEX_INIT(&n->in_lock, "ipnat rule lock");
1510 
1511 	n = NULL;
1512 	ATOMIC_INC32(softn->ipf_nat_stats.ns_rules);
1513 #if SOLARIS && !defined(INSTANCES)
1514 	pfil_delayed_copy = 0;
1515 #endif
1516 	if (getlock) {
1517 		RWLOCK_EXIT(&softc->ipf_nat);			/* WRITE */
1518 	}
1519 
1520 	return error;
1521 }
1522 
1523 
1524 /* ------------------------------------------------------------------------ */
1525 /* Function:    ipf_nat_ruleaddrinit                                        */
1526 /* Parameters:  softc(I) - pointer to soft context main structure           */
1527 /*              softn(I) - pointer to NAT context structure                 */
1528 /*              n(I)     - pointer to NAT rule                              */
1529 /*                                                                          */
1530 /* Initialise all of the NAT address structures in a NAT rule.              */
1531 /* ------------------------------------------------------------------------ */
1532 static int
1533 ipf_nat_ruleaddrinit(ipf_main_softc_t *softc, ipf_nat_softc_t *softn,
1534     ipnat_t *n)
1535 {
1536 	int idx, error;
1537 
1538 	if ((n->in_ndst.na_atype == FRI_LOOKUP) &&
1539 	    (n->in_ndst.na_type != IPLT_DSTLIST)) {
1540 		IPFERROR(60071);
1541 		return EINVAL;
1542 	}
1543 	if ((n->in_nsrc.na_atype == FRI_LOOKUP) &&
1544 	    (n->in_nsrc.na_type != IPLT_DSTLIST)) {
1545 		IPFERROR(60069);
1546 		return EINVAL;
1547 	}
1548 
1549 	if (n->in_redir == NAT_BIMAP) {
1550 		n->in_ndstaddr = n->in_osrcaddr;
1551 		n->in_ndstmsk = n->in_osrcmsk;
1552 		n->in_odstaddr = n->in_nsrcaddr;
1553 		n->in_odstmsk = n->in_nsrcmsk;
1554 
1555 	}
1556 
1557 	if (n->in_redir & NAT_REDIRECT)
1558 		idx = 1;
1559 	else
1560 		idx = 0;
1561 	/*
1562 	 * Initialise all of the address fields.
1563 	 */
1564 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_osrc, 1,
1565 				     n->in_ifps[idx]);
1566 	if (error != 0)
1567 		return error;
1568 
1569 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_odst, 1,
1570 				     n->in_ifps[idx]);
1571 	if (error != 0)
1572 		return error;
1573 
1574 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_nsrc, 1,
1575 				     n->in_ifps[idx]);
1576 	if (error != 0)
1577 		return error;
1578 
1579 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_ndst, 1,
1580 				     n->in_ifps[idx]);
1581 	if (error != 0)
1582 		return error;
1583 
1584 	if (n->in_redir & NAT_DIVERTUDP)
1585 		ipf_nat_builddivertmp(softn, n);
1586 
1587 	return 0;
1588 }
1589 
1590 
1591 /* ------------------------------------------------------------------------ */
1592 /* Function:    ipf_nat_resolvrule                                          */
1593 /* Returns:     Nil                                                         */
1594 /* Parameters:  softc(I) - pointer to soft context main structure           */
1595 /*              n(I)     - pointer to NAT rule                              */
1596 /*                                                                          */
1597 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
1598 /* from information passed to the kernel, then add it  to the appropriate   */
1599 /* NAT rule table(s).                                                       */
1600 /* ------------------------------------------------------------------------ */
1601 static int
1602 ipf_nat_resolverule(ipf_main_softc_t *softc, ipnat_t *n)
1603 {
1604 	char *base;
1605 
1606 	base = n->in_names;
1607 
1608 	n->in_ifps[0] = ipf_resolvenic(softc, base + n->in_ifnames[0],
1609 				       n->in_v[0]);
1610 
1611 	if (n->in_ifnames[1] == -1) {
1612 		n->in_ifnames[1] = n->in_ifnames[0];
1613 		n->in_ifps[1] = n->in_ifps[0];
1614 	} else {
1615 		n->in_ifps[1] = ipf_resolvenic(softc, base + n->in_ifnames[1],
1616 					       n->in_v[1]);
1617 	}
1618 
1619 	if (n->in_plabel != -1) {
1620 		if (n->in_redir & NAT_REDIRECT)
1621 			n->in_apr = ipf_proxy_lookup(softc->ipf_proxy_soft,
1622 						     n->in_pr[0],
1623 						     base + n->in_plabel);
1624 		else
1625 			n->in_apr = ipf_proxy_lookup(softc->ipf_proxy_soft,
1626 						     n->in_pr[1],
1627 						     base + n->in_plabel);
1628 		if (n->in_apr == NULL)
1629 			return -1;
1630 	}
1631 	return 0;
1632 }
1633 
1634 
1635 /* ------------------------------------------------------------------------ */
1636 /* Function:    ipf_nat_siocdelnat                                          */
1637 /* Returns:     int - 0 == success, != 0 == failure                         */
1638 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1639 /*              softn(I)   - pointer to NAT context structure               */
1640 /*              n(I)       - pointer to new NAT rule                        */
1641 /*              getlock(I) - flag indicating if lock on  is held            */
1642 /* Mutex Locks: ipf_nat_io                                                  */
1643 /*                                                                          */
1644 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
1645 /* from information passed to the kernel, then add it  to the appropriate   */
1646 /* NAT rule table(s).                                                       */
1647 /* ------------------------------------------------------------------------ */
1648 static void
1649 ipf_nat_siocdelnat(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *n,
1650 	int getlock)
1651 {
1652 #ifdef IPF_NAT6
1653 	int i;
1654 #endif
1655 
1656 	if (getlock) {
1657 		WRITE_ENTER(&softc->ipf_nat);
1658 	}
1659 
1660 	ipf_nat_delrule(softc, softn, n, 1);
1661 
1662 	if (getlock) {
1663 		RWLOCK_EXIT(&softc->ipf_nat);			/* READ/WRITE */
1664 	}
1665 }
1666 
1667 
1668 /* ------------------------------------------------------------------------ */
1669 /* Function:    ipf_nat_getsz                                               */
1670 /* Returns:     int - 0 == success, != 0 is the error value.                */
1671 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1672 /*              data(I)    - pointer to natget structure with kernel        */
1673 /*                           pointer get the size of.                       */
1674 /*              getlock(I) - flag indicating whether or not the caller      */
1675 /*                           holds a lock on ipf_nat                        */
1676 /*                                                                          */
1677 /* Handle SIOCSTGSZ.                                                        */
1678 /* Return the size of the nat list entry to be copied back to user space.   */
1679 /* The size of the entry is stored in the ng_sz field and the enture natget */
1680 /* structure is copied back to the user.                                    */
1681 /* ------------------------------------------------------------------------ */
1682 static int
1683 ipf_nat_getsz(ipf_main_softc_t *softc, void *data, int getlock)
1684 {
1685 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1686 	ap_session_t *aps;
1687 	nat_t *nat, *n;
1688 	natget_t ng;
1689 	int error;
1690 
1691 	error = BCOPYIN(data, &ng, sizeof(ng));
1692 	if (error != 0) {
1693 		IPFERROR(60024);
1694 		return EFAULT;
1695 	}
1696 
1697 	if (getlock) {
1698 		READ_ENTER(&softc->ipf_nat);
1699 	}
1700 
1701 	nat = ng.ng_ptr;
1702 	if (!nat) {
1703 		nat = softn->ipf_nat_instances;
1704 		ng.ng_sz = 0;
1705 		/*
1706 		 * Empty list so the size returned is 0.  Simple.
1707 		 */
1708 		if (nat == NULL) {
1709 			if (getlock) {
1710 				RWLOCK_EXIT(&softc->ipf_nat);
1711 			}
1712 			error = BCOPYOUT(&ng, data, sizeof(ng));
1713 			if (error != 0) {
1714 				IPFERROR(60025);
1715 				return EFAULT;
1716 			}
1717 			return 0;
1718 		}
1719 	} else {
1720 		/*
1721 		 * Make sure the pointer we're copying from exists in the
1722 		 * current list of entries.  Security precaution to prevent
1723 		 * copying of random kernel data.
1724 		 */
1725 		for (n = softn->ipf_nat_instances; n; n = n->nat_next)
1726 			if (n == nat)
1727 				break;
1728 		if (n == NULL) {
1729 			if (getlock) {
1730 				RWLOCK_EXIT(&softc->ipf_nat);
1731 			}
1732 			IPFERROR(60026);
1733 			return ESRCH;
1734 		}
1735 	}
1736 
1737 	/*
1738 	 * Incluse any space required for proxy data structures.
1739 	 */
1740 	ng.ng_sz = sizeof(nat_save_t);
1741 	aps = nat->nat_aps;
1742 	if (aps != NULL) {
1743 		ng.ng_sz += sizeof(ap_session_t) - 4;
1744 		if (aps->aps_data != 0)
1745 			ng.ng_sz += aps->aps_psiz;
1746 	}
1747 	if (getlock) {
1748 		RWLOCK_EXIT(&softc->ipf_nat);
1749 	}
1750 
1751 	error = BCOPYOUT(&ng, data, sizeof(ng));
1752 	if (error != 0) {
1753 		IPFERROR(60027);
1754 		return EFAULT;
1755 	}
1756 	return 0;
1757 }
1758 
1759 
1760 /* ------------------------------------------------------------------------ */
1761 /* Function:    ipf_nat_getent                                              */
1762 /* Returns:     int - 0 == success, != 0 is the error value.                */
1763 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1764 /*              data(I)    - pointer to natget structure with kernel pointer*/
1765 /*                           to NAT structure to copy out.                  */
1766 /*              getlock(I) - flag indicating whether or not the caller      */
1767 /*                           holds a lock on ipf_nat                        */
1768 /*                                                                          */
1769 /* Handle SIOCSTGET.                                                        */
1770 /* Copies out NAT entry to user space.  Any additional data held for a      */
1771 /* proxy is also copied, as to is the NAT rule which was responsible for it */
1772 /* ------------------------------------------------------------------------ */
1773 static int
1774 ipf_nat_getent(ipf_main_softc_t *softc, void *data, int getlock)
1775 {
1776 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1777 	int error, outsize;
1778 	ap_session_t *aps;
1779 	nat_save_t *ipn, ipns;
1780 	nat_t *n, *nat;
1781 
1782 	error = ipf_inobj(softc, data, NULL, &ipns, IPFOBJ_NATSAVE);
1783 	if (error != 0)
1784 		return error;
1785 
1786 	if ((ipns.ipn_dsize < sizeof(ipns)) || (ipns.ipn_dsize > 81920)) {
1787 		IPFERROR(60028);
1788 		return EINVAL;
1789 	}
1790 
1791 	KMALLOCS(ipn, nat_save_t *, ipns.ipn_dsize);
1792 	if (ipn == NULL) {
1793 		IPFERROR(60029);
1794 		return ENOMEM;
1795 	}
1796 
1797 	if (getlock) {
1798 		READ_ENTER(&softc->ipf_nat);
1799 	}
1800 
1801 	ipn->ipn_dsize = ipns.ipn_dsize;
1802 	nat = ipns.ipn_next;
1803 	if (nat == NULL) {
1804 		nat = softn->ipf_nat_instances;
1805 		if (nat == NULL) {
1806 			if (softn->ipf_nat_instances == NULL) {
1807 				IPFERROR(60030);
1808 				error = ENOENT;
1809 			}
1810 			goto finished;
1811 		}
1812 	} else {
1813 		/*
1814 		 * Make sure the pointer we're copying from exists in the
1815 		 * current list of entries.  Security precaution to prevent
1816 		 * copying of random kernel data.
1817 		 */
1818 		for (n = softn->ipf_nat_instances; n; n = n->nat_next)
1819 			if (n == nat)
1820 				break;
1821 		if (n == NULL) {
1822 			IPFERROR(60031);
1823 			error = ESRCH;
1824 			goto finished;
1825 		}
1826 	}
1827 	ipn->ipn_next = nat->nat_next;
1828 
1829 	/*
1830 	 * Copy the NAT structure.
1831 	 */
1832 	bcopy((char *)nat, &ipn->ipn_nat, sizeof(*nat));
1833 
1834 	/*
1835 	 * If we have a pointer to the NAT rule it belongs to, save that too.
1836 	 */
1837 	if (nat->nat_ptr != NULL)
1838 		bcopy((char *)nat->nat_ptr, (char *)&ipn->ipn_ipnat,
1839 		      ipn->ipn_ipnat.in_size);
1840 
1841 	/*
1842 	 * If we also know the NAT entry has an associated filter rule,
1843 	 * save that too.
1844 	 */
1845 	if (nat->nat_fr != NULL)
1846 		bcopy((char *)nat->nat_fr, (char *)&ipn->ipn_fr,
1847 		      sizeof(ipn->ipn_fr));
1848 
1849 	/*
1850 	 * Last but not least, if there is an application proxy session set
1851 	 * up for this NAT entry, then copy that out too, including any
1852 	 * private data saved along side it by the proxy.
1853 	 */
1854 	aps = nat->nat_aps;
1855 	outsize = ipn->ipn_dsize - sizeof(*ipn) + sizeof(ipn->ipn_data);
1856 	if (aps != NULL) {
1857 		char *s;
1858 
1859 		if (outsize < sizeof(*aps)) {
1860 			IPFERROR(60032);
1861 			error = ENOBUFS;
1862 			goto finished;
1863 		}
1864 
1865 		s = ipn->ipn_data;
1866 		bcopy((char *)aps, s, sizeof(*aps));
1867 		s += sizeof(*aps);
1868 		outsize -= sizeof(*aps);
1869 		if ((aps->aps_data != NULL) && (outsize >= aps->aps_psiz))
1870 			bcopy(aps->aps_data, s, aps->aps_psiz);
1871 		else {
1872 			IPFERROR(60033);
1873 			error = ENOBUFS;
1874 		}
1875 	}
1876 	if (error == 0) {
1877 		if (getlock) {
1878 			READ_ENTER(&softc->ipf_nat);
1879 			getlock = 0;
1880 		}
1881 		error = ipf_outobjsz(softc, data, ipn, IPFOBJ_NATSAVE,
1882 				     ipns.ipn_dsize);
1883 	}
1884 
1885 finished:
1886 	if (getlock) {
1887 		READ_ENTER(&softc->ipf_nat);
1888 	}
1889 	if (ipn != NULL) {
1890 		KFREES(ipn, ipns.ipn_dsize);
1891 	}
1892 	return error;
1893 }
1894 
1895 
1896 /* ------------------------------------------------------------------------ */
1897 /* Function:    ipf_nat_putent                                              */
1898 /* Returns:     int - 0 == success, != 0 is the error value.                */
1899 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1900 /*              data(I)    - pointer to natget structure with NAT           */
1901 /*                           structure information to load into the kernel  */
1902 /*              getlock(I) - flag indicating whether or not a write lock    */
1903 /*                           on is already held.                            */
1904 /*                                                                          */
1905 /* Handle SIOCSTPUT.                                                        */
1906 /* Loads a NAT table entry from user space, including a NAT rule, proxy and */
1907 /* firewall rule data structures, if pointers to them indicate so.          */
1908 /* ------------------------------------------------------------------------ */
1909 static int
1910 ipf_nat_putent(ipf_main_softc_t *softc, void *data, int getlock)
1911 {
1912 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1913 	nat_save_t *ipn, *ipnn;
1914 	ap_session_t *aps;
1915 	nat_t *n, *nat;
1916 	frentry_t *fr;
1917 	fr_info_t *fin;
1918 	ipnat_t *in;
1919 	int error;
1920 
1921 	KMALLOC(ipn, nat_save_t *);
1922 	if (ipn == NULL)
1923 		return ENOMEM;
1924 	error = ipf_inobj(softc, data, NULL, ipn, IPFOBJ_NATSAVE);
1925 	if (error != 0)
1926 		return error;
1927 
1928 	/*
1929 	 * Initialise early because of code at junkput label.
1930 	 */
1931 	n = NULL;
1932 	in = NULL;
1933 	aps = NULL;
1934 	nat = NULL;
1935 	ipnn = NULL;
1936 	fin = NULL;
1937 	fr = NULL;
1938 
1939 	/*
1940 	 * New entry, copy in the rest of the NAT entry if it's size is more
1941 	 * than just the nat_t structure.
1942 	 */
1943 	if (ipn->ipn_dsize > sizeof(*ipn)) {
1944 		if (ipn->ipn_dsize > 81920) {
1945 			IPFERROR(60034);
1946 			error = ENOMEM;
1947 			goto junkput;
1948 		}
1949 
1950 		KMALLOCS(ipnn, nat_save_t *, ipn->ipn_dsize);
1951 		if (ipnn == NULL) {
1952 			IPFERROR(60035);
1953 			return ENOMEM;
1954 		}
1955 
1956 		bzero(ipnn, ipn->ipn_dsize);
1957 		error = ipf_inobjsz(softc, data, ipnn, IPFOBJ_NATSAVE,
1958 				    ipn->ipn_dsize);
1959 		if (error != 0) {
1960 			goto junkput;
1961 		}
1962 	} else
1963 		ipnn = ipn;
1964 
1965 	KMALLOC(nat, nat_t *);
1966 	if (nat == NULL) {
1967 		IPFERROR(60037);
1968 		error = ENOMEM;
1969 		goto junkput;
1970 	}
1971 
1972 	bcopy((char *)&ipnn->ipn_nat, (char *)nat, sizeof(*nat));
1973 
1974 	switch (nat->nat_v[0])
1975 	{
1976 	case 4:
1977 #ifdef USE_IENT6
1978 	case 6 :
1979 #endif
1980 		break;
1981 	default :
1982 		IPFERROR(60061);
1983 		error = EPROTONOSUPPORT;
1984 		goto junkput;
1985 		/*NOTREACHED*/
1986 	}
1987 
1988 	/*
1989 	 * Initialize all these so that ipf_nat_delete() doesn't cause a crash.
1990 	 */
1991 	bzero((char *)nat, offsetof(struct nat, nat_tqe));
1992 	nat->nat_tqe.tqe_pnext = NULL;
1993 	nat->nat_tqe.tqe_next = NULL;
1994 	nat->nat_tqe.tqe_ifq = NULL;
1995 	nat->nat_tqe.tqe_parent = nat;
1996 
1997 	/*
1998 	 * Restore the rule associated with this nat session
1999 	 */
2000 	in = ipnn->ipn_nat.nat_ptr;
2001 	if (in != NULL) {
2002 		KMALLOCS(in, ipnat_t *, ipnn->ipn_ipnat.in_size);
2003 		nat->nat_ptr = in;
2004 		if (in == NULL) {
2005 			IPFERROR(60038);
2006 			error = ENOMEM;
2007 			goto junkput;
2008 		}
2009 		bcopy((char *)&ipnn->ipn_ipnat, (char *)in,
2010 		      ipnn->ipn_ipnat.in_size);
2011 		in->in_use = 1;
2012 		in->in_flags |= IPN_DELETE;
2013 
2014 		ATOMIC_INC32(softn->ipf_nat_stats.ns_rules);
2015 
2016 		if (ipf_nat_resolverule(softc, in) != 0) {
2017 			IPFERROR(60039);
2018 			error = ESRCH;
2019 			goto junkput;
2020 		}
2021 	}
2022 
2023 	/*
2024 	 * Check that the NAT entry doesn't already exist in the kernel.
2025 	 *
2026 	 * For NAT_OUTBOUND, we're lookup for a duplicate MAP entry.  To do
2027 	 * this, we check to see if the inbound combination of addresses and
2028 	 * ports is already known.  Similar logic is applied for NAT_INBOUND.
2029 	 *
2030 	 */
2031 	KMALLOC(fin, fr_info_t *);
2032 	if (fin == NULL) {
2033 		error = ENOMEM;
2034 		goto junkput;
2035 	}
2036 	bzero(fin, sizeof(*fin));
2037 	fin->fin_v = nat->nat_v[0];
2038 	fin->fin_p = nat->nat_pr[0];
2039 	fin->fin_rev = nat->nat_rev;
2040 	fin->fin_ifp = nat->nat_ifps[0];
2041 	fin->fin_data[0] = ntohs(nat->nat_ndport);
2042 	fin->fin_data[1] = ntohs(nat->nat_nsport);
2043 
2044 	switch (nat->nat_dir)
2045 	{
2046 	case NAT_OUTBOUND :
2047 	case NAT_DIVERTOUT :
2048 		if (getlock) {
2049 			READ_ENTER(&softc->ipf_nat);
2050 		}
2051 
2052 		fin->fin_v = nat->nat_v[1];
2053 		if (nat->nat_v[1] == 4) {
2054 			n = ipf_nat_inlookup(fin, nat->nat_flags, fin->fin_p,
2055 					     nat->nat_ndstip, nat->nat_nsrcip);
2056 #ifdef USE_INET6
2057 		} else if (nat->nat_v[1] == 6) {
2058 			n = ipf_nat6_inlookup(fin, nat->nat_flags, fin->fin_p,
2059 					      &nat->nat_ndst6.in6,
2060 					      &nat->nat_nsrc6.in6);
2061 #endif
2062 		}
2063 
2064 		if (getlock) {
2065 			RWLOCK_EXIT(&softc->ipf_nat);
2066 		}
2067 		if (n != NULL) {
2068 			IPFERROR(60040);
2069 			error = EEXIST;
2070 			goto junkput;
2071 		}
2072 		break;
2073 
2074 	case NAT_INBOUND :
2075 	case NAT_DIVERTIN :
2076 		if (getlock) {
2077 			READ_ENTER(&softc->ipf_nat);
2078 		}
2079 
2080 		if (fin->fin_v == 4) {
2081 			n = ipf_nat_outlookup(fin, nat->nat_flags, fin->fin_p,
2082 					      nat->nat_ndstip,
2083 					      nat->nat_nsrcip);
2084 #ifdef USE_INET6
2085 		} else if (fin->fin_v == 6) {
2086 			n = ipf_nat6_outlookup(fin, nat->nat_flags, fin->fin_p,
2087 					       &nat->nat_ndst6.in6,
2088 					       &nat->nat_nsrc6.in6);
2089 #endif
2090 		}
2091 
2092 		if (getlock) {
2093 			RWLOCK_EXIT(&softc->ipf_nat);
2094 		}
2095 		if (n != NULL) {
2096 			IPFERROR(60041);
2097 			error = EEXIST;
2098 			goto junkput;
2099 		}
2100 		break;
2101 
2102 	default :
2103 		IPFERROR(60042);
2104 		error = EINVAL;
2105 		goto junkput;
2106 	}
2107 
2108 	/*
2109 	 * Restore ap_session_t structure.  Include the private data allocated
2110 	 * if it was there.
2111 	 */
2112 	aps = nat->nat_aps;
2113 	if (aps != NULL) {
2114 		KMALLOC(aps, ap_session_t *);
2115 		nat->nat_aps = aps;
2116 		if (aps == NULL) {
2117 			IPFERROR(60043);
2118 			error = ENOMEM;
2119 			goto junkput;
2120 		}
2121 		bcopy(ipnn->ipn_data, (char *)aps, sizeof(*aps));
2122 		if (in != NULL)
2123 			aps->aps_apr = in->in_apr;
2124 		else
2125 			aps->aps_apr = NULL;
2126 		if (aps->aps_psiz != 0) {
2127 			if (aps->aps_psiz > 81920) {
2128 				IPFERROR(60044);
2129 				error = ENOMEM;
2130 				goto junkput;
2131 			}
2132 			KMALLOCS(aps->aps_data, void *, aps->aps_psiz);
2133 			if (aps->aps_data == NULL) {
2134 				IPFERROR(60045);
2135 				error = ENOMEM;
2136 				goto junkput;
2137 			}
2138 			bcopy(ipnn->ipn_data + sizeof(*aps), aps->aps_data,
2139 			      aps->aps_psiz);
2140 		} else {
2141 			aps->aps_psiz = 0;
2142 			aps->aps_data = NULL;
2143 		}
2144 	}
2145 
2146 	/*
2147 	 * If there was a filtering rule associated with this entry then
2148 	 * build up a new one.
2149 	 */
2150 	fr = nat->nat_fr;
2151 	if (fr != NULL) {
2152 		if ((nat->nat_flags & SI_NEWFR) != 0) {
2153 			KMALLOC(fr, frentry_t *);
2154 			nat->nat_fr = fr;
2155 			if (fr == NULL) {
2156 				IPFERROR(60046);
2157 				error = ENOMEM;
2158 				goto junkput;
2159 			}
2160 			ipnn->ipn_nat.nat_fr = fr;
2161 			fr->fr_ref = 1;
2162 			(void) ipf_outobj(softc, data, ipnn, IPFOBJ_NATSAVE);
2163 			bcopy((char *)&ipnn->ipn_fr, (char *)fr, sizeof(*fr));
2164 
2165 			fr->fr_ref = 1;
2166 			fr->fr_dsize = 0;
2167 			fr->fr_data = NULL;
2168 			fr->fr_type = FR_T_NONE;
2169 
2170 			MUTEX_NUKE(&fr->fr_lock);
2171 			MUTEX_INIT(&fr->fr_lock, "nat-filter rule lock");
2172 		} else {
2173 			if (getlock) {
2174 				READ_ENTER(&softc->ipf_nat);
2175 			}
2176 			for (n = softn->ipf_nat_instances; n; n = n->nat_next)
2177 				if (n->nat_fr == fr)
2178 					break;
2179 
2180 			if (n != NULL) {
2181 				MUTEX_ENTER(&fr->fr_lock);
2182 				fr->fr_ref++;
2183 				MUTEX_EXIT(&fr->fr_lock);
2184 			}
2185 			if (getlock) {
2186 				RWLOCK_EXIT(&softc->ipf_nat);
2187 			}
2188 
2189 			if (n == NULL) {
2190 				IPFERROR(60047);
2191 				error = ESRCH;
2192 				goto junkput;
2193 			}
2194 		}
2195 	}
2196 
2197 	if (ipnn != ipn) {
2198 		KFREES(ipnn, ipn->ipn_dsize);
2199 		ipnn = NULL;
2200 	}
2201 
2202 	if (getlock) {
2203 		WRITE_ENTER(&softc->ipf_nat);
2204 	}
2205 
2206 	if (fin->fin_v == 4)
2207 		error = ipf_nat_finalise(fin, nat);
2208 #ifdef USE_INET6
2209 	else
2210 		error = ipf_nat6_finalise(fin, nat);
2211 #endif
2212 
2213 	if (getlock) {
2214 		RWLOCK_EXIT(&softc->ipf_nat);
2215 	}
2216 
2217 	if (error == 0)
2218 		return 0;
2219 
2220 	IPFERROR(60048);
2221 	error = ENOMEM;
2222 
2223 junkput:
2224 	if (fin != NULL)
2225 		KFREE(fin);
2226 	if (fr != NULL) {
2227 		(void) ipf_derefrule(softc, &fr);
2228 	}
2229 
2230 	if ((ipnn != NULL) && (ipnn != ipn)) {
2231 		KFREES(ipnn, ipn->ipn_dsize);
2232 	}
2233 	if (ipn != NULL)
2234 		KFREE(ipn);
2235 	if (nat != NULL) {
2236 		if (aps != NULL) {
2237 			if (aps->aps_data != NULL) {
2238 				KFREES(aps->aps_data, aps->aps_psiz);
2239 			}
2240 			KFREE(aps);
2241 		}
2242 		if (in != NULL) {
2243 			if (in->in_apr)
2244 				ipf_proxy_deref(in->in_apr);
2245 			KFREES(in, in->in_size);
2246 		}
2247 		KFREE(nat);
2248 	}
2249 	return error;
2250 }
2251 
2252 
2253 /* ------------------------------------------------------------------------ */
2254 /* Function:    ipf_nat_delete                                              */
2255 /* Returns:     Nil                                                         */
2256 /* Parameters:  softc(I)   - pointer to soft context main structure         */
2257 /*              nat(I)     - pointer to NAT structure to delete             */
2258 /*              logtype(I) - type of LOG record to create before deleting   */
2259 /* Write Lock:  ipf_nat                                                     */
2260 /*                                                                          */
2261 /* Delete a nat entry from the various lists and table.  If NAT logging is  */
2262 /* enabled then generate a NAT log record for this event.                   */
2263 /* ------------------------------------------------------------------------ */
2264 void
2265 ipf_nat_delete(ipf_main_softc_t *softc, struct nat *nat, int logtype)
2266 {
2267 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2268 	int madeorphan = 0, removed = 0;
2269 	u_int bkt;
2270 	nat_stat_side_t *nss;
2271 	struct ipnat *ipn;
2272 
2273 	if (logtype != 0 && softn->ipf_nat_logging != 0)
2274 		ipf_nat_log(softc, softn, nat, logtype);
2275 
2276 	/*
2277 	 * Take it as a general indication that all the pointers are set if
2278 	 * nat_pnext is set.
2279 	 */
2280 	if (nat->nat_pnext != NULL) {
2281 		removed = 1;
2282 
2283 		bkt = nat->nat_hv[0] % softn->ipf_nat_table_sz;
2284 		nss = &softn->ipf_nat_stats.ns_side[0];
2285 		ASSERT(nss->ns_bucketlen[bkt] > 0);
2286 		nss->ns_bucketlen[bkt]--;
2287 		if (nss->ns_bucketlen[bkt] == 0) {
2288 			nss->ns_inuse--;
2289 		}
2290 
2291 		bkt = nat->nat_hv[1] % softn->ipf_nat_table_sz;
2292 		nss = &softn->ipf_nat_stats.ns_side[1];
2293 		ASSERT(nss->ns_bucketlen[bkt] > 0);
2294 		nss->ns_bucketlen[bkt]--;
2295 		if (nss->ns_bucketlen[bkt] == 0) {
2296 			nss->ns_inuse--;
2297 		}
2298 
2299 		*nat->nat_pnext = nat->nat_next;
2300 		if (nat->nat_next != NULL) {
2301 			nat->nat_next->nat_pnext = nat->nat_pnext;
2302 			nat->nat_next = NULL;
2303 		}
2304 		nat->nat_pnext = NULL;
2305 
2306 		*nat->nat_phnext[0] = nat->nat_hnext[0];
2307 		if (nat->nat_hnext[0] != NULL) {
2308 			nat->nat_hnext[0]->nat_phnext[0] = nat->nat_phnext[0];
2309 			nat->nat_hnext[0] = NULL;
2310 		}
2311 		nat->nat_phnext[0] = NULL;
2312 
2313 		*nat->nat_phnext[1] = nat->nat_hnext[1];
2314 		if (nat->nat_hnext[1] != NULL) {
2315 			nat->nat_hnext[1]->nat_phnext[1] = nat->nat_phnext[1];
2316 			nat->nat_hnext[1] = NULL;
2317 		}
2318 		nat->nat_phnext[1] = NULL;
2319 
2320 		if ((nat->nat_flags & SI_WILDP) != 0) {
2321 			ATOMIC_DEC32(softn->ipf_nat_stats.ns_wilds);
2322 		}
2323 		madeorphan = 1;
2324 	}
2325 
2326 	if (nat->nat_me != NULL) {
2327 		*nat->nat_me = NULL;
2328 		nat->nat_me = NULL;
2329 		nat->nat_ref--;
2330 		ASSERT(nat->nat_ref >= 0);
2331 	}
2332 
2333 	if (nat->nat_tqe.tqe_ifq != NULL) {
2334 		/*
2335 		 * No call to ipf_freetimeoutqueue() is made here, they are
2336 		 * garbage collected in ipf_nat_expire().
2337 		 */
2338 		(void) ipf_deletequeueentry(&nat->nat_tqe);
2339 	}
2340 
2341 	if (nat->nat_sync) {
2342 		ipf_sync_del_nat(softc->ipf_sync_soft, nat->nat_sync);
2343 		nat->nat_sync = NULL;
2344 	}
2345 
2346 	if (logtype == NL_EXPIRE)
2347 		softn->ipf_nat_stats.ns_expire++;
2348 
2349 	MUTEX_ENTER(&nat->nat_lock);
2350 	/*
2351 	 * NL_DESTROY should only be passed in when we've got nat_ref >= 2.
2352 	 * This happens when a nat'd packet is blocked and we want to throw
2353 	 * away the NAT session.
2354 	 */
2355 	if (logtype == NL_DESTROY) {
2356 		if (nat->nat_ref > 2) {
2357 			nat->nat_ref -= 2;
2358 			MUTEX_EXIT(&nat->nat_lock);
2359 			if (removed)
2360 				softn->ipf_nat_stats.ns_orphans++;
2361 			return;
2362 		}
2363 	} else if (nat->nat_ref > 1) {
2364 		nat->nat_ref--;
2365 		MUTEX_EXIT(&nat->nat_lock);
2366 		if (madeorphan == 1)
2367 			softn->ipf_nat_stats.ns_orphans++;
2368 		return;
2369 	}
2370 	ASSERT(nat->nat_ref >= 0);
2371 	MUTEX_EXIT(&nat->nat_lock);
2372 
2373 	nat->nat_ref = 0;
2374 
2375 	if (madeorphan == 0)
2376 		softn->ipf_nat_stats.ns_orphans--;
2377 
2378 	/*
2379 	 * At this point, nat_ref can be either 0 or -1
2380 	 */
2381 	softn->ipf_nat_stats.ns_proto[nat->nat_pr[0]]--;
2382 
2383 	if (nat->nat_fr != NULL) {
2384 		(void) ipf_derefrule(softc, &nat->nat_fr);
2385 	}
2386 
2387 	if (nat->nat_hm != NULL) {
2388 		ipf_nat_hostmapdel(softc, &nat->nat_hm);
2389 	}
2390 
2391 	/*
2392 	 * If there is an active reference from the nat entry to its parent
2393 	 * rule, decrement the rule's reference count and free it too if no
2394 	 * longer being used.
2395 	 */
2396 	ipn = nat->nat_ptr;
2397 	nat->nat_ptr = NULL;
2398 
2399 	if (ipn != NULL) {
2400 		ipn->in_space++;
2401 		ipf_nat_rule_deref(softc, &ipn);
2402 	}
2403 
2404 	if (nat->nat_aps != NULL) {
2405 		ipf_proxy_free(softc, nat->nat_aps);
2406 		nat->nat_aps = NULL;
2407 	}
2408 
2409 	MUTEX_DESTROY(&nat->nat_lock);
2410 
2411 	softn->ipf_nat_stats.ns_active--;
2412 
2413 	/*
2414 	 * If there's a fragment table entry too for this nat entry, then
2415 	 * dereference that as well.  This is after nat_lock is released
2416 	 * because of Tru64.
2417 	 */
2418 	ipf_frag_natforget(softc, (void *)nat);
2419 
2420 	KFREE(nat);
2421 }
2422 
2423 
2424 /* ------------------------------------------------------------------------ */
2425 /* Function:    ipf_nat_flushtable                                          */
2426 /* Returns:     int - number of NAT rules deleted                           */
2427 /* Parameters:  softc(I) - pointer to soft context main structure           */
2428 /*              softn(I) - pointer to NAT context structure                 */
2429 /* Write Lock:  ipf_nat                                                     */
2430 /*                                                                          */
2431 /* Deletes all currently active NAT sessions.  In deleting each NAT entry a */
2432 /* log record should be emitted in ipf_nat_delete() if NAT logging is       */
2433 /* enabled.                                                                 */
2434 /* ------------------------------------------------------------------------ */
2435 /*
2436  * nat_flushtable - clear the NAT table of all mapping entries.
2437  */
2438 static int
2439 ipf_nat_flushtable(ipf_main_softc_t *softc, ipf_nat_softc_t *softn)
2440 {
2441 	nat_t *nat;
2442 	int j = 0;
2443 
2444 	/*
2445 	 * ALL NAT mappings deleted, so lets just make the deletions
2446 	 * quicker.
2447 	 */
2448 	if (softn->ipf_nat_table[0] != NULL)
2449 		bzero((char *)softn->ipf_nat_table[0],
2450 		      sizeof(softn->ipf_nat_table[0]) *
2451 		      softn->ipf_nat_table_sz);
2452 	if (softn->ipf_nat_table[1] != NULL)
2453 		bzero((char *)softn->ipf_nat_table[1],
2454 		      sizeof(softn->ipf_nat_table[1]) *
2455 		      softn->ipf_nat_table_sz);
2456 
2457 	while ((nat = softn->ipf_nat_instances) != NULL) {
2458 		ipf_nat_delete(softc, nat, NL_FLUSH);
2459 		j++;
2460 	}
2461 
2462 	return j;
2463 }
2464 
2465 
2466 /* ------------------------------------------------------------------------ */
2467 /* Function:    ipf_nat_clearlist                                           */
2468 /* Returns:     int - number of NAT/RDR rules deleted                       */
2469 /* Parameters:  softc(I) - pointer to soft context main structure           */
2470 /*              softn(I) - pointer to NAT context structure                 */
2471 /*                                                                          */
2472 /* Delete all rules in the current list of rules.  There is nothing elegant */
2473 /* about this cleanup: simply free all entries on the list of rules and     */
2474 /* clear out the tables used for hashed NAT rule lookups.                   */
2475 /* ------------------------------------------------------------------------ */
2476 static int
2477 ipf_nat_clearlist(ipf_main_softc_t *softc, ipf_nat_softc_t *softn)
2478 {
2479 	ipnat_t *n;
2480 	int i = 0;
2481 
2482 	if (softn->ipf_nat_map_rules != NULL) {
2483 		bzero((char *)softn->ipf_nat_map_rules,
2484 		      sizeof(*softn->ipf_nat_map_rules) *
2485 		      softn->ipf_nat_maprules_sz);
2486 	}
2487 	if (softn->ipf_nat_rdr_rules != NULL) {
2488 		bzero((char *)softn->ipf_nat_rdr_rules,
2489 		      sizeof(*softn->ipf_nat_rdr_rules) *
2490 		      softn->ipf_nat_rdrrules_sz);
2491 	}
2492 
2493 	while ((n = softn->ipf_nat_list) != NULL) {
2494 		ipf_nat_delrule(softc, softn, n, 0);
2495 		i++;
2496 	}
2497 #if SOLARIS && !defined(INSTANCES)
2498 	pfil_delayed_copy = 1;
2499 #endif
2500 	return i;
2501 }
2502 
2503 
2504 /* ------------------------------------------------------------------------ */
2505 /* Function:    ipf_nat_delrule                                             */
2506 /* Returns:     Nil                                                         */
2507 /* Parameters:  softc(I) - pointer to soft context main structure           */
2508 /*              softn(I) - pointer to NAT context structure                 */
2509 /*              np(I)    - pointer to NAT rule to delete                    */
2510 /*              purge(I) - 1 == allow purge, 0 == prevent purge             */
2511 /* Locks:       WRITE(ipf_nat)                                              */
2512 /*                                                                          */
2513 /* Preventing "purge" from occuring is allowed because when all of the NAT  */
2514 /* rules are being removed, allowing the "purge" to walk through the list   */
2515 /* of NAT sessions, possibly multiple times, would be a large performance   */
2516 /* hit, on the order of O(N^2).                                             */
2517 /* ------------------------------------------------------------------------ */
2518 static void
2519 ipf_nat_delrule(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *np,
2520 	int purge)
2521 {
2522 
2523 	if (np->in_pnext != NULL) {
2524 		*np->in_pnext = np->in_next;
2525 		if (np->in_next != NULL)
2526 			np->in_next->in_pnext = np->in_pnext;
2527 		if (softn->ipf_nat_list_tail == &np->in_next)
2528 			softn->ipf_nat_list_tail = np->in_pnext;
2529 	}
2530 
2531 	if ((purge == 1) && ((np->in_flags & IPN_PURGE) != 0)) {
2532 		nat_t *next;
2533 		nat_t *nat;
2534 
2535 		for (next = softn->ipf_nat_instances; (nat = next) != NULL;) {
2536 			next = nat->nat_next;
2537 			if (nat->nat_ptr == np)
2538 				ipf_nat_delete(softc, nat, NL_PURGE);
2539 		}
2540 	}
2541 
2542 	if ((np->in_flags & IPN_DELETE) == 0) {
2543 		if (np->in_redir & NAT_REDIRECT) {
2544 			switch (np->in_v[0])
2545 			{
2546 			case 4 :
2547 				ipf_nat_delrdr(softn, np);
2548 				break;
2549 #ifdef USE_INET6
2550 			case 6 :
2551 				ipf_nat6_delrdr(softn, np);
2552 				break;
2553 #endif
2554 			}
2555 		}
2556 		if (np->in_redir & (NAT_MAPBLK|NAT_MAP)) {
2557 			switch (np->in_v[0])
2558 			{
2559 			case 4 :
2560 				ipf_nat_delmap(softn, np);
2561 				break;
2562 #ifdef USE_INET6
2563 			case 6 :
2564 				ipf_nat6_delmap(softn, np);
2565 				break;
2566 #endif
2567 			}
2568 		}
2569 	}
2570 
2571 	np->in_flags |= IPN_DELETE;
2572 	ipf_nat_rule_deref(softc, &np);
2573 }
2574 
2575 
2576 /* ------------------------------------------------------------------------ */
2577 /* Function:    ipf_nat_newmap                                              */
2578 /* Returns:     int - -1 == error, 0 == success                             */
2579 /* Parameters:  fin(I) - pointer to packet information                      */
2580 /*              nat(I) - pointer to NAT entry                               */
2581 /*              ni(I)  - pointer to structure with misc. information needed */
2582 /*                       to create new NAT entry.                           */
2583 /*                                                                          */
2584 /* Given an empty NAT structure, populate it with new information about a   */
2585 /* new NAT session, as defined by the matching NAT rule.                    */
2586 /* ni.nai_ip is passed in uninitialised and must be set, in host byte order,*/
2587 /* to the new IP address for the translation.                               */
2588 /* ------------------------------------------------------------------------ */
2589 static int
2590 ipf_nat_newmap(fr_info_t *fin, nat_t *nat, natinfo_t *ni)
2591 {
2592 	ipf_main_softc_t *softc = fin->fin_main_soft;
2593 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2594 	u_short st_port, dport, sport, port, sp, dp;
2595 	struct in_addr in, inb;
2596 	hostmap_t *hm;
2597 	u_32_t flags;
2598 	u_32_t st_ip;
2599 	ipnat_t *np;
2600 	nat_t *natl;
2601 	int l;
2602 
2603 	/*
2604 	 * If it's an outbound packet which doesn't match any existing
2605 	 * record, then create a new port
2606 	 */
2607 	l = 0;
2608 	hm = NULL;
2609 	np = ni->nai_np;
2610 	st_ip = np->in_snip;
2611 	st_port = np->in_spnext;
2612 	flags = nat->nat_flags;
2613 
2614 	if (flags & IPN_ICMPQUERY) {
2615 		sport = fin->fin_data[1];
2616 		dport = 0;
2617 	} else {
2618 		sport = htons(fin->fin_data[0]);
2619 		dport = htons(fin->fin_data[1]);
2620 	}
2621 
2622 	/*
2623 	 * Do a loop until we either run out of entries to try or we find
2624 	 * a NAT mapping that isn't currently being used.  This is done
2625 	 * because the change to the source is not (usually) being fixed.
2626 	 */
2627 	do {
2628 		port = 0;
2629 		in.s_addr = htonl(np->in_snip);
2630 		if (l == 0) {
2631 			/*
2632 			 * Check to see if there is an existing NAT
2633 			 * setup for this IP address pair.
2634 			 */
2635 			hm = ipf_nat_hostmap(softn, np, fin->fin_src,
2636 					     fin->fin_dst, in, 0);
2637 			if (hm != NULL)
2638 				in.s_addr = hm->hm_nsrcip.s_addr;
2639 		} else if ((l == 1) && (hm != NULL)) {
2640 			ipf_nat_hostmapdel(softc, &hm);
2641 		}
2642 		in.s_addr = ntohl(in.s_addr);
2643 
2644 		nat->nat_hm = hm;
2645 
2646 		if ((np->in_nsrcmsk == 0xffffffff) && (np->in_spnext == 0)) {
2647 			if (l > 0) {
2648 				NBUMPSIDEX(1, ns_exhausted, ns_exhausted_1);
2649 				return -1;
2650 			}
2651 		}
2652 
2653 		if (np->in_redir == NAT_BIMAP &&
2654 		    np->in_osrcmsk == np->in_nsrcmsk) {
2655 			/*
2656 			 * map the address block in a 1:1 fashion
2657 			 */
2658 			in.s_addr = np->in_nsrcaddr;
2659 			in.s_addr |= fin->fin_saddr & ~np->in_osrcmsk;
2660 			in.s_addr = ntohl(in.s_addr);
2661 
2662 		} else if (np->in_redir & NAT_MAPBLK) {
2663 			if ((l >= np->in_ppip) || ((l > 0) &&
2664 			     !(flags & IPN_TCPUDP))) {
2665 				NBUMPSIDEX(1, ns_exhausted, ns_exhausted_2);
2666 				return -1;
2667 			}
2668 			/*
2669 			 * map-block - Calculate destination address.
2670 			 */
2671 			in.s_addr = ntohl(fin->fin_saddr);
2672 			in.s_addr &= ntohl(~np->in_osrcmsk);
2673 			inb.s_addr = in.s_addr;
2674 			in.s_addr /= np->in_ippip;
2675 			in.s_addr &= ntohl(~np->in_nsrcmsk);
2676 			in.s_addr += ntohl(np->in_nsrcaddr);
2677 			/*
2678 			 * Calculate destination port.
2679 			 */
2680 			if ((flags & IPN_TCPUDP) &&
2681 			    (np->in_ppip != 0)) {
2682 				port = ntohs(sport) + l;
2683 				port %= np->in_ppip;
2684 				port += np->in_ppip *
2685 					(inb.s_addr % np->in_ippip);
2686 				port += MAPBLK_MINPORT;
2687 				port = htons(port);
2688 			}
2689 
2690 		} else if ((np->in_nsrcaddr == 0) &&
2691 			   (np->in_nsrcmsk == 0xffffffff)) {
2692 			i6addr_t in6;
2693 
2694 			/*
2695 			 * 0/32 - use the interface's IP address.
2696 			 */
2697 			if ((l > 0) ||
2698 			    ipf_ifpaddr(softc, 4, FRI_NORMAL, fin->fin_ifp,
2699 				       &in6, NULL) == -1) {
2700 				NBUMPSIDEX(1, ns_new_ifpaddr, ns_new_ifpaddr_1);
2701 				return -1;
2702 			}
2703 			in.s_addr = ntohl(in6.in4.s_addr);
2704 
2705 		} else if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0)) {
2706 			/*
2707 			 * 0/0 - use the original source address/port.
2708 			 */
2709 			if (l > 0) {
2710 				NBUMPSIDEX(1, ns_exhausted, ns_exhausted_3);
2711 				return -1;
2712 			}
2713 			in.s_addr = ntohl(fin->fin_saddr);
2714 
2715 		} else if ((np->in_nsrcmsk != 0xffffffff) &&
2716 			   (np->in_spnext == 0) && ((l > 0) || (hm == NULL)))
2717 			np->in_snip++;
2718 
2719 		natl = NULL;
2720 
2721 		if ((flags & IPN_TCPUDP) &&
2722 		    ((np->in_redir & NAT_MAPBLK) == 0) &&
2723 		    (np->in_flags & IPN_AUTOPORTMAP)) {
2724 			/*
2725 			 * "ports auto" (without map-block)
2726 			 */
2727 			if ((l > 0) && np->in_ppip && (l % np->in_ppip == 0)) {
2728 				if (l > np->in_space) {
2729 					return -1;
2730 				} else if ((l > np->in_ppip) &&
2731 					   np->in_nsrcmsk != 0xffffffff)
2732 					np->in_snip++;
2733 			}
2734 			if (np->in_ppip != 0) {
2735 				port = ntohs(sport);
2736 				port += (l % np->in_ppip);
2737 				port %= np->in_ppip;
2738 				port += np->in_ppip *
2739 					(ntohl(fin->fin_saddr) %
2740 					 np->in_ippip);
2741 				port += MAPBLK_MINPORT;
2742 				port = htons(port);
2743 			}
2744 
2745 		} else if (((np->in_redir & NAT_MAPBLK) == 0) &&
2746 			   (flags & IPN_TCPUDPICMP) && (np->in_spnext != 0)) {
2747 			/*
2748 			 * Standard port translation.  Select next port.
2749 			 */
2750 			if (np->in_flags & IPN_SEQUENTIAL) {
2751 				port = np->in_spnext;
2752 			} else {
2753 				port = ipf_random() % (np->in_spmax -
2754 						       np->in_spmin + 1);
2755 				port += np->in_spmin;
2756 			}
2757 			port = htons(port);
2758 			np->in_spnext++;
2759 
2760 			if (np->in_spnext > np->in_spmax) {
2761 				np->in_spnext = np->in_spmin;
2762 				if (np->in_nsrcmsk != 0xffffffff)
2763 					np->in_snip++;
2764 			}
2765 		}
2766 
2767 		if (np->in_flags & IPN_SIPRANGE) {
2768 			if (np->in_snip > ntohl(np->in_nsrcmsk))
2769 				np->in_snip = ntohl(np->in_nsrcaddr);
2770 		} else {
2771 			if ((np->in_nsrcmsk != 0xffffffff) &&
2772 			    ((np->in_snip + 1) & ntohl(np->in_nsrcmsk)) >
2773 			    ntohl(np->in_nsrcaddr))
2774 				np->in_snip = ntohl(np->in_nsrcaddr) + 1;
2775 		}
2776 
2777 		if ((port == 0) && (flags & (IPN_TCPUDPICMP|IPN_ICMPQUERY)))
2778 			port = sport;
2779 
2780 		/*
2781 		 * Here we do a lookup of the connection as seen from
2782 		 * the outside.  If an IP# pair already exists, try
2783 		 * again.  So if you have A->B becomes C->B, you can
2784 		 * also have D->E become C->E but not D->B causing
2785 		 * another C->B.  Also take protocol and ports into
2786 		 * account when determining whether a pre-existing
2787 		 * NAT setup will cause an external conflict where
2788 		 * this is appropriate.
2789 		 */
2790 		inb.s_addr = htonl(in.s_addr);
2791 		sp = fin->fin_data[0];
2792 		dp = fin->fin_data[1];
2793 		fin->fin_data[0] = fin->fin_data[1];
2794 		fin->fin_data[1] = ntohs(port);
2795 		natl = ipf_nat_inlookup(fin, flags & ~(SI_WILDP|NAT_SEARCH),
2796 					(u_int)fin->fin_p, fin->fin_dst, inb);
2797 		fin->fin_data[0] = sp;
2798 		fin->fin_data[1] = dp;
2799 
2800 		/*
2801 		 * Has the search wrapped around and come back to the
2802 		 * start ?
2803 		 */
2804 		if ((natl != NULL) &&
2805 		    (np->in_spnext != 0) && (st_port == np->in_spnext) &&
2806 		    (np->in_snip != 0) && (st_ip == np->in_snip)) {
2807 			NBUMPSIDED(1, ns_wrap);
2808 			return -1;
2809 		}
2810 		l++;
2811 	} while (natl != NULL);
2812 
2813 	/* Setup the NAT table */
2814 	nat->nat_osrcip = fin->fin_src;
2815 	nat->nat_nsrcaddr = htonl(in.s_addr);
2816 	nat->nat_odstip = fin->fin_dst;
2817 	nat->nat_ndstip = fin->fin_dst;
2818 	if (nat->nat_hm == NULL)
2819 		nat->nat_hm = ipf_nat_hostmap(softn, np, fin->fin_src,
2820 					      fin->fin_dst, nat->nat_nsrcip,
2821 					      0);
2822 
2823 	if (flags & IPN_TCPUDP) {
2824 		nat->nat_osport = sport;
2825 		nat->nat_nsport = port;	/* sport */
2826 		nat->nat_odport = dport;
2827 		nat->nat_ndport = dport;
2828 		((tcphdr_t *)fin->fin_dp)->th_sport = port;
2829 	} else if (flags & IPN_ICMPQUERY) {
2830 		nat->nat_oicmpid = fin->fin_data[1];
2831 		((icmphdr_t *)fin->fin_dp)->icmp_id = port;
2832 		nat->nat_nicmpid = port;
2833 	}
2834 	return 0;
2835 }
2836 
2837 
2838 /* ------------------------------------------------------------------------ */
2839 /* Function:    ipf_nat_newrdr                                              */
2840 /* Returns:     int - -1 == error, 0 == success (no move), 1 == success and */
2841 /*                    allow rule to be moved if IPN_ROUNDR is set.          */
2842 /* Parameters:  fin(I) - pointer to packet information                      */
2843 /*              nat(I) - pointer to NAT entry                               */
2844 /*              ni(I)  - pointer to structure with misc. information needed */
2845 /*                       to create new NAT entry.                           */
2846 /*                                                                          */
2847 /* ni.nai_ip is passed in uninitialised and must be set, in host byte order,*/
2848 /* to the new IP address for the translation.                               */
2849 /* ------------------------------------------------------------------------ */
2850 static int
2851 ipf_nat_newrdr(fr_info_t *fin, nat_t *nat, natinfo_t *ni)
2852 {
2853 	ipf_main_softc_t *softc = fin->fin_main_soft;
2854 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2855 	u_short nport, dport, sport;
2856 	struct in_addr in, inb;
2857 	u_short sp, dp;
2858 	hostmap_t *hm;
2859 	u_32_t flags;
2860 	ipnat_t *np;
2861 	nat_t *natl;
2862 	int move;
2863 
2864 	move = 1;
2865 	hm = NULL;
2866 	in.s_addr = 0;
2867 	np = ni->nai_np;
2868 	flags = nat->nat_flags;
2869 
2870 	if (flags & IPN_ICMPQUERY) {
2871 		dport = fin->fin_data[1];
2872 		sport = 0;
2873 	} else {
2874 		sport = htons(fin->fin_data[0]);
2875 		dport = htons(fin->fin_data[1]);
2876 	}
2877 
2878 	/* TRACE sport, dport */
2879 
2880 
2881 	/*
2882 	 * If the matching rule has IPN_STICKY set, then we want to have the
2883 	 * same rule kick in as before.  Why would this happen?  If you have
2884 	 * a collection of rdr rules with "round-robin sticky", the current
2885 	 * packet might match a different one to the previous connection but
2886 	 * we want the same destination to be used.
2887 	 */
2888 	if (((np->in_flags & (IPN_ROUNDR|IPN_SPLIT)) != 0) &&
2889 	    ((np->in_flags & IPN_STICKY) != 0)) {
2890 		hm = ipf_nat_hostmap(softn, NULL, fin->fin_src, fin->fin_dst,
2891 				     in, (u_32_t)dport);
2892 		if (hm != NULL) {
2893 			in.s_addr = ntohl(hm->hm_ndstip.s_addr);
2894 			np = hm->hm_ipnat;
2895 			ni->nai_np = np;
2896 			move = 0;
2897 			ipf_nat_hostmapdel(softc, &hm);
2898 		}
2899 	}
2900 
2901 	/*
2902 	 * Otherwise, it's an inbound packet. Most likely, we don't
2903 	 * want to rewrite source ports and source addresses. Instead,
2904 	 * we want to rewrite to a fixed internal address and fixed
2905 	 * internal port.
2906 	 */
2907 	if (np->in_flags & IPN_SPLIT) {
2908 		in.s_addr = np->in_dnip;
2909 		inb.s_addr = htonl(in.s_addr);
2910 
2911 		if ((np->in_flags & (IPN_ROUNDR|IPN_STICKY)) == IPN_STICKY) {
2912 			hm = ipf_nat_hostmap(softn, NULL, fin->fin_src,
2913 					     fin->fin_dst, inb, (u_32_t)dport);
2914 			if (hm != NULL) {
2915 				in.s_addr = hm->hm_ndstip.s_addr;
2916 				move = 0;
2917 			}
2918 		}
2919 
2920 		if (hm == NULL || hm->hm_ref == 1) {
2921 			if (np->in_ndstaddr == htonl(in.s_addr)) {
2922 				np->in_dnip = ntohl(np->in_ndstmsk);
2923 				move = 0;
2924 			} else {
2925 				np->in_dnip = ntohl(np->in_ndstaddr);
2926 			}
2927 		}
2928 		if (hm != NULL)
2929 			ipf_nat_hostmapdel(softc, &hm);
2930 
2931 	} else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0xffffffff)) {
2932 		i6addr_t in6;
2933 
2934 		/*
2935 		 * 0/32 - use the interface's IP address.
2936 		 */
2937 		if (ipf_ifpaddr(softc, 4, FRI_NORMAL, fin->fin_ifp,
2938 			       &in6, NULL) == -1) {
2939 			NBUMPSIDEX(0, ns_new_ifpaddr, ns_new_ifpaddr_2);
2940 			return -1;
2941 		}
2942 		in.s_addr = ntohl(in6.in4.s_addr);
2943 
2944 	} else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk== 0)) {
2945 		/*
2946 		 * 0/0 - use the original destination address/port.
2947 		 */
2948 		in.s_addr = ntohl(fin->fin_daddr);
2949 
2950 	} else if (np->in_redir == NAT_BIMAP &&
2951 		   np->in_ndstmsk == np->in_odstmsk) {
2952 		/*
2953 		 * map the address block in a 1:1 fashion
2954 		 */
2955 		in.s_addr = np->in_ndstaddr;
2956 		in.s_addr |= fin->fin_daddr & ~np->in_ndstmsk;
2957 		in.s_addr = ntohl(in.s_addr);
2958 	} else {
2959 		in.s_addr = ntohl(np->in_ndstaddr);
2960 	}
2961 
2962 	if ((np->in_dpnext == 0) || ((flags & NAT_NOTRULEPORT) != 0))
2963 		nport = dport;
2964 	else {
2965 		/*
2966 		 * Whilst not optimized for the case where
2967 		 * pmin == pmax, the gain is not significant.
2968 		 */
2969 		if (((np->in_flags & IPN_FIXEDDPORT) == 0) &&
2970 		    (np->in_odport != np->in_dtop)) {
2971 			nport = ntohs(dport) - np->in_odport + np->in_dpmax;
2972 			nport = htons(nport);
2973 		} else {
2974 			nport = htons(np->in_dpnext);
2975 			np->in_dpnext++;
2976 			if (np->in_dpnext > np->in_dpmax)
2977 				np->in_dpnext = np->in_dpmin;
2978 		}
2979 	}
2980 
2981 	/*
2982 	 * When the redirect-to address is set to 0.0.0.0, just
2983 	 * assume a blank `forwarding' of the packet.  We don't
2984 	 * setup any translation for this either.
2985 	 */
2986 	if (in.s_addr == 0) {
2987 		if (nport == dport) {
2988 			NBUMPSIDED(0, ns_xlate_null);
2989 			return -1;
2990 		}
2991 		in.s_addr = ntohl(fin->fin_daddr);
2992 	}
2993 
2994 	/*
2995 	 * Check to see if this redirect mapping already exists and if
2996 	 * it does, return "failure" (allowing it to be created will just
2997 	 * cause one or both of these "connections" to stop working.)
2998 	 */
2999 	inb.s_addr = htonl(in.s_addr);
3000 	sp = fin->fin_data[0];
3001 	dp = fin->fin_data[1];
3002 	fin->fin_data[1] = fin->fin_data[0];
3003 	fin->fin_data[0] = ntohs(nport);
3004 	natl = ipf_nat_outlookup(fin, flags & ~(SI_WILDP|NAT_SEARCH),
3005 			     (u_int)fin->fin_p, inb, fin->fin_src);
3006 	fin->fin_data[0] = sp;
3007 	fin->fin_data[1] = dp;
3008 	if (natl != NULL) {
3009 		DT2(ns_new_xlate_exists, fr_info_t *, fin, nat_t *, natl);
3010 		NBUMPSIDE(0, ns_xlate_exists);
3011 		return -1;
3012 	}
3013 
3014 	nat->nat_ndstaddr = htonl(in.s_addr);
3015 	nat->nat_odstip = fin->fin_dst;
3016 	nat->nat_nsrcip = fin->fin_src;
3017 	nat->nat_osrcip = fin->fin_src;
3018 	if ((nat->nat_hm == NULL) && ((np->in_flags & IPN_STICKY) != 0))
3019 		nat->nat_hm = ipf_nat_hostmap(softn, np, fin->fin_src,
3020 					      fin->fin_dst, inb, (u_32_t)dport);
3021 
3022 	if (flags & IPN_TCPUDP) {
3023 		nat->nat_odport = dport;
3024 		nat->nat_ndport = nport;
3025 		nat->nat_osport = sport;
3026 		nat->nat_nsport = sport;
3027 		((tcphdr_t *)fin->fin_dp)->th_dport = nport;
3028 	} else if (flags & IPN_ICMPQUERY) {
3029 		nat->nat_oicmpid = fin->fin_data[1];
3030 		((icmphdr_t *)fin->fin_dp)->icmp_id = nport;
3031 		nat->nat_nicmpid = nport;
3032 	}
3033 
3034 	return move;
3035 }
3036 
3037 /* ------------------------------------------------------------------------ */
3038 /* Function:    ipf_nat_add                                                 */
3039 /* Returns:     nat_t* - NULL == failure to create new NAT structure,       */
3040 /*                       else pointer to new NAT structure                  */
3041 /* Parameters:  fin(I)       - pointer to packet information                */
3042 /*              np(I)        - pointer to NAT rule                          */
3043 /*              natsave(I)   - pointer to where to store NAT struct pointer */
3044 /*              flags(I)     - flags describing the current packet          */
3045 /*              direction(I) - direction of packet (in/out)                 */
3046 /* Write Lock:  ipf_nat                                                     */
3047 /*                                                                          */
3048 /* Attempts to create a new NAT entry.  Does not actually change the packet */
3049 /* in any way.                                                              */
3050 /*                                                                          */
3051 /* This function is in three main parts: (1) deal with creating a new NAT   */
3052 /* structure for a "MAP" rule (outgoing NAT translation); (2) deal with     */
3053 /* creating a new NAT structure for a "RDR" rule (incoming NAT translation) */
3054 /* and (3) building that structure and putting it into the NAT table(s).    */
3055 /*                                                                          */
3056 /* NOTE: natsave should NOT be used to point back to an ipstate_t struct    */
3057 /*       as it can result in memory being corrupted.                        */
3058 /* ------------------------------------------------------------------------ */
3059 nat_t *
3060 ipf_nat_add(fr_info_t *fin, ipnat_t *np, nat_t **natsave, u_int flags,
3061     int direction)
3062 {
3063 	ipf_main_softc_t *softc = fin->fin_main_soft;
3064 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3065 	hostmap_t *hm = NULL;
3066 	nat_t *nat, *natl;
3067 	natstat_t *nsp;
3068 	u_int nflags;
3069 	natinfo_t ni;
3070 	int move;
3071 
3072 	memset(&ni, 0, sizeof ni);	/* XXX gcc */
3073 	nsp = &softn->ipf_nat_stats;
3074 
3075 	if ((nsp->ns_active * 100 / softn->ipf_nat_table_max) >
3076 	    softn->ipf_nat_table_wm_high) {
3077 		softn->ipf_nat_doflush = 1;
3078 	}
3079 
3080 	if (nsp->ns_active >= softn->ipf_nat_table_max) {
3081 		NBUMPSIDED(fin->fin_out, ns_table_max);
3082 		return NULL;
3083 	}
3084 
3085 	move = 1;
3086 	nflags = np->in_flags & flags;
3087 	nflags &= NAT_FROMRULE;
3088 
3089 	ni.nai_np = np;
3090 	ni.nai_dport = 0;
3091 	ni.nai_sport = 0;
3092 
3093 	/* Give me a new nat */
3094 	KMALLOC(nat, nat_t *);
3095 	if (nat == NULL) {
3096 		NBUMPSIDED(fin->fin_out, ns_memfail);
3097 		/*
3098 		 * Try to automatically tune the max # of entries in the
3099 		 * table allowed to be less than what will cause kmem_alloc()
3100 		 * to fail and try to eliminate panics due to out of memory
3101 		 * conditions arising.
3102 		 */
3103 		if ((softn->ipf_nat_table_max > softn->ipf_nat_table_sz) &&
3104 		    (nsp->ns_active > 100)) {
3105 			softn->ipf_nat_table_max = nsp->ns_active - 100;
3106 			printf("table_max reduced to %d\n",
3107 				softn->ipf_nat_table_max);
3108 		}
3109 		return NULL;
3110 	}
3111 
3112 	if (flags & IPN_ICMPQUERY) {
3113 		/*
3114 		 * In the ICMP query NAT code, we translate the ICMP id fields
3115 		 * to make them unique. This is indepedent of the ICMP type
3116 		 * (e.g. in the unlikely event that a host sends an echo and
3117 		 * an tstamp request with the same id, both packets will have
3118 		 * their ip address/id field changed in the same way).
3119 		 */
3120 		/* The icmp_id field is used by the sender to identify the
3121 		 * process making the icmp request. (the receiver justs
3122 		 * copies it back in its response). So, it closely matches
3123 		 * the concept of source port. We overlay sport, so we can
3124 		 * maximally reuse the existing code.
3125 		 */
3126 		ni.nai_sport = fin->fin_data[1];
3127 		ni.nai_dport = 0;
3128 	}
3129 
3130 	bzero((char *)nat, sizeof(*nat));
3131 	nat->nat_flags = flags;
3132 	nat->nat_redir = np->in_redir;
3133 	nat->nat_dir = direction;
3134 	nat->nat_pr[0] = fin->fin_p;
3135 	nat->nat_pr[1] = fin->fin_p;
3136 
3137 	/*
3138 	 * Search the current table for a match and create a new mapping
3139 	 * if there is none found.
3140 	 */
3141 	if (np->in_redir & NAT_DIVERTUDP) {
3142 		move = ipf_nat_newdivert(fin, nat, &ni);
3143 
3144 	} else if (np->in_redir & NAT_REWRITE) {
3145 		move = ipf_nat_newrewrite(fin, nat, &ni);
3146 
3147 	} else if (direction == NAT_OUTBOUND) {
3148 		/*
3149 		 * We can now arrange to call this for the same connection
3150 		 * because ipf_nat_new doesn't protect the code path into
3151 		 * this function.
3152 		 */
3153 		natl = ipf_nat_outlookup(fin, nflags, (u_int)fin->fin_p,
3154 				     fin->fin_src, fin->fin_dst);
3155 		if (natl != NULL) {
3156 			KFREE(nat);
3157 			nat = natl;
3158 			goto done;
3159 		}
3160 
3161 		move = ipf_nat_newmap(fin, nat, &ni);
3162 	} else {
3163 		/*
3164 		 * NAT_INBOUND is used for redirects rules
3165 		 */
3166 		natl = ipf_nat_inlookup(fin, nflags, (u_int)fin->fin_p,
3167 					fin->fin_src, fin->fin_dst);
3168 		if (natl != NULL) {
3169 			KFREE(nat);
3170 			nat = natl;
3171 			goto done;
3172 		}
3173 
3174 		move = ipf_nat_newrdr(fin, nat, &ni);
3175 	}
3176 	if (move == -1)
3177 		goto badnat;
3178 
3179 	np = ni.nai_np;
3180 
3181 	nat->nat_mssclamp = np->in_mssclamp;
3182 	nat->nat_me = natsave;
3183 	nat->nat_fr = fin->fin_fr;
3184 	nat->nat_rev = fin->fin_rev;
3185 	nat->nat_ptr = np;
3186 	nat->nat_dlocal = np->in_dlocal;
3187 
3188 	if ((np->in_apr != NULL) && ((nat->nat_flags & NAT_SLAVE) == 0)) {
3189 		if (ipf_proxy_new(fin, nat) == -1) {
3190 			NBUMPSIDED(fin->fin_out, ns_appr_fail);
3191 			goto badnat;
3192 		}
3193 	}
3194 
3195 	nat->nat_ifps[0] = np->in_ifps[0];
3196 	if (np->in_ifps[0] != NULL) {
3197 		COPYIFNAME(np->in_v[0], np->in_ifps[0], nat->nat_ifnames[0]);
3198 	}
3199 
3200 	nat->nat_ifps[1] = np->in_ifps[1];
3201 	if (np->in_ifps[1] != NULL) {
3202 		COPYIFNAME(np->in_v[1], np->in_ifps[1], nat->nat_ifnames[1]);
3203 	}
3204 
3205 	if (ipf_nat_finalise(fin, nat) == -1) {
3206 		goto badnat;
3207 	}
3208 
3209 	np->in_use++;
3210 
3211 	if ((move == 1) && (np->in_flags & IPN_ROUNDR)) {
3212 		if ((np->in_redir & (NAT_REDIRECT|NAT_MAP)) == NAT_REDIRECT) {
3213 			ipf_nat_delrdr(softn, np);
3214 			ipf_nat_addrdr(softn, np);
3215 		} else if ((np->in_redir & (NAT_REDIRECT|NAT_MAP)) == NAT_MAP) {
3216 			ipf_nat_delmap(softn, np);
3217 			ipf_nat_addmap(softn, np);
3218 		}
3219 	}
3220 
3221 	if (flags & SI_WILDP)
3222 		nsp->ns_wilds++;
3223 	nsp->ns_proto[nat->nat_pr[0]]++;
3224 
3225 	goto done;
3226 badnat:
3227 	DT2(ns_badnatnew, fr_info_t *, fin, nat_t *, nat);
3228 	NBUMPSIDE(fin->fin_out, ns_badnatnew);
3229 	if ((hm = nat->nat_hm) != NULL)
3230 		ipf_nat_hostmapdel(softc, &hm);
3231 	KFREE(nat);
3232 	nat = NULL;
3233 done:
3234 	if (nat != NULL && np != NULL)
3235 		np->in_hits++;
3236 	if (natsave != NULL)
3237 		*natsave = nat;
3238 	return nat;
3239 }
3240 
3241 
3242 /* ------------------------------------------------------------------------ */
3243 /* Function:    ipf_nat_finalise                                            */
3244 /* Returns:     int - 0 == sucess, -1 == failure                            */
3245 /* Parameters:  fin(I) - pointer to packet information                      */
3246 /*              nat(I) - pointer to NAT entry                               */
3247 /* Write Lock:  ipf_nat                                                     */
3248 /*                                                                          */
3249 /* This is the tail end of constructing a new NAT entry and is the same     */
3250 /* for both IPv4 and IPv6.                                                  */
3251 /* ------------------------------------------------------------------------ */
3252 /*ARGSUSED*/
3253 static int
3254 ipf_nat_finalise(fr_info_t *fin, nat_t *nat)
3255 {
3256 	ipf_main_softc_t *softc = fin->fin_main_soft;
3257 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3258 	u_32_t sum1, sum2, sumd;
3259 	frentry_t *fr;
3260 #if SOLARIS && defined(_KERNEL) && (SOLARIS2 >= 6) && defined(ICK_M_CTL_MAGIC)
3261 	qpktinfo_t *qpi = fin->fin_qpi;
3262 #endif
3263 
3264 	switch (nat->nat_pr[0])
3265 	{
3266 	case IPPROTO_ICMP :
3267 		sum1 = LONG_SUM(ntohs(nat->nat_oicmpid));
3268 		sum2 = LONG_SUM(ntohs(nat->nat_nicmpid));
3269 		CALC_SUMD(sum1, sum2, sumd);
3270 		nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
3271 
3272 		break;
3273 
3274 	default :
3275 		sum1 = LONG_SUM(ntohl(nat->nat_osrcaddr) + \
3276 				ntohs(nat->nat_osport));
3277 		sum2 = LONG_SUM(ntohl(nat->nat_nsrcaddr) + \
3278 				ntohs(nat->nat_nsport));
3279 		CALC_SUMD(sum1, sum2, sumd);
3280 		nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
3281 
3282 		sum1 = LONG_SUM(ntohl(nat->nat_odstaddr) + \
3283 				ntohs(nat->nat_odport));
3284 		sum2 = LONG_SUM(ntohl(nat->nat_ndstaddr) + \
3285 				ntohs(nat->nat_ndport));
3286 		CALC_SUMD(sum1, sum2, sumd);
3287 		nat->nat_sumd[0] += (sumd & 0xffff) + (sumd >> 16);
3288 		break;
3289 	}
3290 
3291 	/*
3292 	 * Compute the partial checksum, just in case.
3293 	 * This is only ever placed into outbound packets so care needs
3294 	 * to be taken over which pair of addresses are used.
3295 	 */
3296 	if (nat->nat_dir == NAT_OUTBOUND) {
3297 		sum1 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
3298 		sum1 += LONG_SUM(ntohl(nat->nat_ndstaddr));
3299 	} else {
3300 		sum1 = LONG_SUM(ntohl(nat->nat_osrcaddr));
3301 		sum1 += LONG_SUM(ntohl(nat->nat_odstaddr));
3302 	}
3303 	sum1 += nat->nat_pr[1];
3304 	nat->nat_sumd[1] = (sum1 & 0xffff) + (sum1 >> 16);
3305 
3306 	sum1 = LONG_SUM(ntohl(nat->nat_osrcaddr));
3307 	sum2 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
3308 	CALC_SUMD(sum1, sum2, sumd);
3309 	nat->nat_ipsumd = (sumd & 0xffff) + (sumd >> 16);
3310 
3311 	sum1 = LONG_SUM(ntohl(nat->nat_odstaddr));
3312 	sum2 = LONG_SUM(ntohl(nat->nat_ndstaddr));
3313 	CALC_SUMD(sum1, sum2, sumd);
3314 	nat->nat_ipsumd += (sumd & 0xffff) + (sumd >> 16);
3315 
3316 	nat->nat_v[0] = 4;
3317 	nat->nat_v[1] = 4;
3318 
3319 	if ((nat->nat_ifps[0] != NULL) && (nat->nat_ifps[0] != (void *)-1)) {
3320 		nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
3321 	}
3322 
3323 	if ((nat->nat_ifps[1] != NULL) && (nat->nat_ifps[1] != (void *)-1)) {
3324 		nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
3325 	}
3326 
3327 	if ((nat->nat_flags & SI_CLONE) == 0)
3328 		nat->nat_sync = ipf_sync_new(softc, SMC_NAT, fin, nat);
3329 
3330 	if (ipf_nat_insert(softc, softn, nat) == 0) {
3331 		if (softn->ipf_nat_logging)
3332 			ipf_nat_log(softc, softn, nat, NL_NEW);
3333 		fr = nat->nat_fr;
3334 		if (fr != NULL) {
3335 			MUTEX_ENTER(&fr->fr_lock);
3336 			fr->fr_ref++;
3337 			MUTEX_EXIT(&fr->fr_lock);
3338 		}
3339 		return 0;
3340 	}
3341 
3342 	NBUMPSIDED(fin->fin_out, ns_unfinalised);
3343 	/*
3344 	 * nat_insert failed, so cleanup time...
3345 	 */
3346 	if (nat->nat_sync != NULL)
3347 		ipf_sync_del_nat(softc->ipf_sync_soft, nat->nat_sync);
3348 	return -1;
3349 }
3350 
3351 
3352 /* ------------------------------------------------------------------------ */
3353 /* Function:    ipf_nat_insert                                              */
3354 /* Returns:     int - 0 == sucess, -1 == failure                            */
3355 /* Parameters:  softc(I) - pointer to soft context main structure           */
3356 /*              softn(I) - pointer to NAT context structure                 */
3357 /*              nat(I) - pointer to NAT structure                           */
3358 /* Write Lock:  ipf_nat                                                     */
3359 /*                                                                          */
3360 /* Insert a NAT entry into the hash tables for searching and add it to the  */
3361 /* list of active NAT entries.  Adjust global counters when complete.       */
3362 /* ------------------------------------------------------------------------ */
3363 int
3364 ipf_nat_insert(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, nat_t *nat)
3365 {
3366 	u_int hv0, hv1;
3367 	u_int sp, dp;
3368 	ipnat_t *in;
3369 	int ret;
3370 
3371 	/*
3372 	 * Try and return an error as early as possible, so calculate the hash
3373 	 * entry numbers first and then proceed.
3374 	 */
3375 	if ((nat->nat_flags & (SI_W_SPORT|SI_W_DPORT)) == 0) {
3376 		if ((nat->nat_flags & IPN_TCPUDP) != 0) {
3377 			sp = nat->nat_osport;
3378 			dp = nat->nat_odport;
3379 		} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
3380 			sp = 0;
3381 			dp = nat->nat_oicmpid;
3382 		} else {
3383 			sp = 0;
3384 			dp = 0;
3385 		}
3386 		hv0 = NAT_HASH_FN(nat->nat_osrcaddr, sp, 0xffffffff);
3387 		hv0 = NAT_HASH_FN(nat->nat_odstaddr, hv0 + dp, 0xffffffff);
3388 		/*
3389 		 * TRACE nat_osrcaddr, nat_osport, nat_odstaddr,
3390 		 * nat_odport, hv0
3391 		 */
3392 
3393 		if ((nat->nat_flags & IPN_TCPUDP) != 0) {
3394 			sp = nat->nat_nsport;
3395 			dp = nat->nat_ndport;
3396 		} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
3397 			sp = 0;
3398 			dp = nat->nat_nicmpid;
3399 		} else {
3400 			sp = 0;
3401 			dp = 0;
3402 		}
3403 		hv1 = NAT_HASH_FN(nat->nat_nsrcaddr, sp, 0xffffffff);
3404 		hv1 = NAT_HASH_FN(nat->nat_ndstaddr, hv1 + dp, 0xffffffff);
3405 		/*
3406 		 * TRACE nat_nsrcaddr, nat_nsport, nat_ndstaddr,
3407 		 * nat_ndport, hv1
3408 		 */
3409 	} else {
3410 		hv0 = NAT_HASH_FN(nat->nat_osrcaddr, 0, 0xffffffff);
3411 		hv0 = NAT_HASH_FN(nat->nat_odstaddr, hv0, 0xffffffff);
3412 		/* TRACE nat_osrcaddr, nat_odstaddr, hv0 */
3413 
3414 		hv1 = NAT_HASH_FN(nat->nat_nsrcaddr, 0, 0xffffffff);
3415 		hv1 = NAT_HASH_FN(nat->nat_ndstaddr, hv1, 0xffffffff);
3416 		/* TRACE nat_nsrcaddr, nat_ndstaddr, hv1 */
3417 	}
3418 
3419 	if ((nat->nat_dir & NAT_OUTBOUND) == NAT_OUTBOUND) {
3420 		nat->nat_hv[0] = hv0;
3421 		nat->nat_hv[1] = hv1;
3422 	} else {
3423 		nat->nat_hv[0] = hv1;
3424 		nat->nat_hv[1] = hv0;
3425 	}
3426 
3427 	MUTEX_INIT(&nat->nat_lock, "nat entry lock");
3428 
3429 	in = nat->nat_ptr;
3430 	nat->nat_ref = nat->nat_me ? 2 : 1;
3431 
3432 	nat->nat_ifnames[0][LIFNAMSIZ - 1] = '\0';
3433 	nat->nat_ifps[0] = ipf_resolvenic(softc, nat->nat_ifnames[0], 4);
3434 
3435 	if (nat->nat_ifnames[1][0] != '\0') {
3436 		nat->nat_ifnames[1][LIFNAMSIZ - 1] = '\0';
3437 		nat->nat_ifps[1] = ipf_resolvenic(softc,
3438 						  nat->nat_ifnames[1], 4);
3439 	} else if (in->in_ifnames[1] != -1) {
3440 		char *name;
3441 
3442 		name = in->in_names + in->in_ifnames[1];
3443 		if (name[1] != '\0' && name[0] != '-' && name[0] != '*') {
3444 			(void) strncpy(nat->nat_ifnames[1],
3445 				       nat->nat_ifnames[0], LIFNAMSIZ);
3446 			nat->nat_ifnames[1][LIFNAMSIZ - 1] = '\0';
3447 			nat->nat_ifps[1] = nat->nat_ifps[0];
3448 		}
3449 	}
3450 	if ((nat->nat_ifps[0] != NULL) && (nat->nat_ifps[0] != (void *)-1)) {
3451 		nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
3452 	}
3453 	if ((nat->nat_ifps[1] != NULL) && (nat->nat_ifps[1] != (void *)-1)) {
3454 		nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
3455 	}
3456 
3457 	ret = ipf_nat_hashtab_add(softc, softn, nat);
3458 	if (ret != 0)
3459 		MUTEX_DESTROY(&nat->nat_lock);
3460 	return ret;
3461 }
3462 
3463 
3464 /* ------------------------------------------------------------------------ */
3465 /* Function:    ipf_nat_hashtab_add                                         */
3466 /* Returns:     int - 0 == sucess, -1 == failure                            */
3467 /* Parameters:  softc(I) - pointer to soft context main structure           */
3468 /*              softn(I) - pointer to NAT context structure                 */
3469 /*              nat(I) - pointer to NAT structure                           */
3470 /* Write Lock:  ipf_nat                                                     */
3471 /*                                                                          */
3472 /* Handle the insertion of a NAT entry into the table/list.                 */
3473 /* ------------------------------------------------------------------------ */
3474 int
3475 ipf_nat_hashtab_add(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, nat_t *nat)
3476 {
3477 	nat_t **natp;
3478 	u_int hv0;
3479 	u_int hv1;
3480 
3481 	hv0 = nat->nat_hv[0] % softn->ipf_nat_table_sz;
3482 	hv1 = nat->nat_hv[1] % softn->ipf_nat_table_sz;
3483 
3484 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen[hv0] >=
3485 	    softn->ipf_nat_maxbucket) {
3486 		DT1(ns_bucket_max_0, int,
3487 		    softn->ipf_nat_stats.ns_side[0].ns_bucketlen[hv0]);
3488 		NBUMPSIDE(0, ns_bucket_max);
3489 		return -1;
3490 	}
3491 
3492 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen[hv1] >=
3493 	    softn->ipf_nat_maxbucket) {
3494 		DT1(ns_bucket_max_1, int,
3495 		    softn->ipf_nat_stats.ns_side[1].ns_bucketlen[hv1]);
3496 		NBUMPSIDE(1, ns_bucket_max);
3497 		return -1;
3498 	}
3499 
3500 	/*
3501 	 * The ordering of operations in the list and hash table insertion
3502 	 * is very important.  The last operation for each task should be
3503 	 * to update the top of the list, after all the "nexts" have been
3504 	 * done so that walking the list while it is being done does not
3505 	 * find strange pointers.
3506 	 *
3507 	 * Global list of NAT instances
3508 	 */
3509 	nat->nat_next = softn->ipf_nat_instances;
3510 	nat->nat_pnext = &softn->ipf_nat_instances;
3511 	if (softn->ipf_nat_instances)
3512 		softn->ipf_nat_instances->nat_pnext = &nat->nat_next;
3513 	softn->ipf_nat_instances = nat;
3514 
3515 	/*
3516 	 * Inbound hash table.
3517 	 */
3518 	natp = &softn->ipf_nat_table[0][hv0];
3519 	nat->nat_phnext[0] = natp;
3520 	nat->nat_hnext[0] = *natp;
3521 	if (*natp) {
3522 		(*natp)->nat_phnext[0] = &nat->nat_hnext[0];
3523 	} else {
3524 		NBUMPSIDE(0, ns_inuse);
3525 	}
3526 	*natp = nat;
3527 	NBUMPSIDE(0, ns_bucketlen[hv0]);
3528 
3529 	/*
3530 	 * Outbound hash table.
3531 	 */
3532 	natp = &softn->ipf_nat_table[1][hv1];
3533 	nat->nat_phnext[1] = natp;
3534 	nat->nat_hnext[1] = *natp;
3535 	if (*natp)
3536 		(*natp)->nat_phnext[1] = &nat->nat_hnext[1];
3537 	else {
3538 		NBUMPSIDE(1, ns_inuse);
3539 	}
3540 	*natp = nat;
3541 	NBUMPSIDE(1, ns_bucketlen[hv1]);
3542 
3543 	ipf_nat_setqueue(softc, softn, nat);
3544 
3545 	if (nat->nat_dir & NAT_OUTBOUND) {
3546 		NBUMPSIDE(1, ns_added);
3547 	} else {
3548 		NBUMPSIDE(0, ns_added);
3549 	}
3550 	softn->ipf_nat_stats.ns_active++;
3551 	return 0;
3552 }
3553 
3554 
3555 /* ------------------------------------------------------------------------ */
3556 /* Function:    ipf_nat_icmperrorlookup                                     */
3557 /* Returns:     nat_t* - point to matching NAT structure                    */
3558 /* Parameters:  fin(I) - pointer to packet information                      */
3559 /*              dir(I) - direction of packet (in/out)                       */
3560 /*                                                                          */
3561 /* Check if the ICMP error message is related to an existing TCP, UDP or    */
3562 /* ICMP query nat entry.  It is assumed that the packet is already of the   */
3563 /* the required length.                                                     */
3564 /* ------------------------------------------------------------------------ */
3565 nat_t *
3566 ipf_nat_icmperrorlookup(fr_info_t *fin, int dir)
3567 {
3568 	ipf_main_softc_t *softc = fin->fin_main_soft;
3569 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3570 	int flags = 0, minlen;
3571 	icmphdr_t *orgicmp;
3572 	nat_stat_side_t *nside;
3573 	tcphdr_t *tcp = NULL;
3574 	u_short data[2];
3575 	nat_t *nat;
3576 	ip_t *oip;
3577 	u_int p;
3578 
3579 	nside = &softn->ipf_nat_stats.ns_side[fin->fin_out];
3580 	/*
3581 	 * Does it at least have the return (basic) IP header ?
3582 	 * Only a basic IP header (no options) should be with an ICMP error
3583 	 * header.  Also, if it's not an error type, then return.
3584 	 */
3585 	if ((fin->fin_hlen != sizeof(ip_t)) || !(fin->fin_flx & FI_ICMPERR)) {
3586 		ATOMIC_INCL(nside->ns_icmp_basic);
3587 		return NULL;
3588 	}
3589 
3590 	/*
3591 	 * Check packet size
3592 	 */
3593 	oip = (ip_t *)((char *)fin->fin_dp + 8);
3594 	minlen = IP_HL(oip) << 2;
3595 	if ((minlen < sizeof(ip_t)) ||
3596 	    (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen)) {
3597 		ATOMIC_INCL(nside->ns_icmp_size);
3598 		return NULL;
3599 	}
3600 
3601 	/*
3602 	 * Is the buffer big enough for all of it ?  It's the size of the IP
3603 	 * header claimed in the encapsulated part which is of concern.  It
3604 	 * may be too big to be in this buffer but not so big that it's
3605 	 * outside the ICMP packet, leading to TCP deref's causing problems.
3606 	 * This is possible because we don't know how big oip_hl is when we
3607 	 * do the pullup early in ipf_check() and thus can't gaurantee it is
3608 	 * all here now.
3609 	 */
3610 #ifdef  ipf_nat_KERNEL
3611 	{
3612 	mb_t *m;
3613 
3614 	m = fin->fin_m;
3615 # if defined(MENTAT)
3616 	if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN >
3617 	    (char *)m->b_wptr) {
3618 		ATOMIC_INCL(nside->ns_icmp_mbuf);
3619 		return NULL;
3620 	}
3621 # else
3622 	if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN >
3623 	    (char *)fin->fin_ip + M_LEN(m)) {
3624 		ATOMIC_INCL(nside->ns_icmp_mbuf);
3625 		return NULL;
3626 	}
3627 # endif
3628 	}
3629 #endif
3630 
3631 	if (fin->fin_daddr != oip->ip_src.s_addr) {
3632 		ATOMIC_INCL(nside->ns_icmp_address);
3633 		return NULL;
3634 	}
3635 
3636 	p = oip->ip_p;
3637 	if (p == IPPROTO_TCP)
3638 		flags = IPN_TCP;
3639 	else if (p == IPPROTO_UDP)
3640 		flags = IPN_UDP;
3641 	else if (p == IPPROTO_ICMP) {
3642 		orgicmp = (icmphdr_t *)((char *)oip + (IP_HL(oip) << 2));
3643 
3644 		/* see if this is related to an ICMP query */
3645 		if (ipf_nat_icmpquerytype(orgicmp->icmp_type)) {
3646 			data[0] = fin->fin_data[0];
3647 			data[1] = fin->fin_data[1];
3648 			fin->fin_data[0] = 0;
3649 			fin->fin_data[1] = orgicmp->icmp_id;
3650 
3651 			flags = IPN_ICMPERR|IPN_ICMPQUERY;
3652 			/*
3653 			 * NOTE : dir refers to the direction of the original
3654 			 *        ip packet. By definition the icmp error
3655 			 *        message flows in the opposite direction.
3656 			 */
3657 			if (dir == NAT_INBOUND)
3658 				nat = ipf_nat_inlookup(fin, flags, p,
3659 						       oip->ip_dst,
3660 						       oip->ip_src);
3661 			else
3662 				nat = ipf_nat_outlookup(fin, flags, p,
3663 							oip->ip_dst,
3664 							oip->ip_src);
3665 			fin->fin_data[0] = data[0];
3666 			fin->fin_data[1] = data[1];
3667 			return nat;
3668 		}
3669 	}
3670 
3671 	if (flags & IPN_TCPUDP) {
3672 		minlen += 8;		/* + 64bits of data to get ports */
3673 		/* TRACE (fin,minlen) */
3674 		if (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen) {
3675 			ATOMIC_INCL(nside->ns_icmp_short);
3676 			return NULL;
3677 		}
3678 
3679 		data[0] = fin->fin_data[0];
3680 		data[1] = fin->fin_data[1];
3681 		tcp = (tcphdr_t *)((char *)oip + (IP_HL(oip) << 2));
3682 		fin->fin_data[0] = ntohs(tcp->th_dport);
3683 		fin->fin_data[1] = ntohs(tcp->th_sport);
3684 
3685 		if (dir == NAT_INBOUND) {
3686 			nat = ipf_nat_inlookup(fin, flags, p, oip->ip_dst,
3687 					       oip->ip_src);
3688 		} else {
3689 			nat = ipf_nat_outlookup(fin, flags, p, oip->ip_dst,
3690 					    oip->ip_src);
3691 		}
3692 		fin->fin_data[0] = data[0];
3693 		fin->fin_data[1] = data[1];
3694 		return nat;
3695 	}
3696 	if (dir == NAT_INBOUND)
3697 		nat = ipf_nat_inlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
3698 	else
3699 		nat = ipf_nat_outlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
3700 
3701 	return nat;
3702 }
3703 
3704 
3705 /* ------------------------------------------------------------------------ */
3706 /* Function:    ipf_nat_icmperror                                           */
3707 /* Returns:     nat_t* - point to matching NAT structure                    */
3708 /* Parameters:  fin(I)    - pointer to packet information                   */
3709 /*              nflags(I) - NAT flags for this packet                       */
3710 /*              dir(I)    - direction of packet (in/out)                    */
3711 /*                                                                          */
3712 /* Fix up an ICMP packet which is an error message for an existing NAT      */
3713 /* session.  This will correct both packet header data and checksums.       */
3714 /*                                                                          */
3715 /* This should *ONLY* be used for incoming ICMP error packets to make sure  */
3716 /* a NAT'd ICMP packet gets correctly recognised.                           */
3717 /* ------------------------------------------------------------------------ */
3718 nat_t *
3719 ipf_nat_icmperror(fr_info_t *fin, u_int *nflags, int dir)
3720 {
3721 	ipf_main_softc_t *softc = fin->fin_main_soft;
3722 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3723 	u_32_t sum1, sum2, sumd, sumd2;
3724 	struct in_addr a1, a2, a3, a4;
3725 	int flags, dlen, odst;
3726 	icmphdr_t *icmp;
3727 	u_short *csump;
3728 	tcphdr_t *tcp;
3729 	nat_t *nat;
3730 	ip_t *oip;
3731 	void *dp;
3732 
3733 	if ((fin->fin_flx & (FI_SHORT|FI_FRAGBODY))) {
3734 		NBUMPSIDED(fin->fin_out, ns_icmp_short);
3735 		return NULL;
3736 	}
3737 
3738 	/*
3739 	 * ipf_nat_icmperrorlookup() will return NULL for `defective' packets.
3740 	 */
3741 	if ((fin->fin_v != 4) || !(nat = ipf_nat_icmperrorlookup(fin, dir))) {
3742 		NBUMPSIDED(fin->fin_out, ns_icmp_notfound);
3743 		return NULL;
3744 	}
3745 
3746 	tcp = NULL;
3747 	csump = NULL;
3748 	flags = 0;
3749 	sumd2 = 0;
3750 	*nflags = IPN_ICMPERR;
3751 	icmp = fin->fin_dp;
3752 	oip = (ip_t *)&icmp->icmp_ip;
3753 	dp = (((char *)oip) + (IP_HL(oip) << 2));
3754 	if (oip->ip_p == IPPROTO_TCP) {
3755 		tcp = (tcphdr_t *)dp;
3756 		csump = (u_short *)&tcp->th_sum;
3757 		flags = IPN_TCP;
3758 	} else if (oip->ip_p == IPPROTO_UDP) {
3759 		udphdr_t *udp;
3760 
3761 		udp = (udphdr_t *)dp;
3762 		tcp = (tcphdr_t *)dp;
3763 		csump = (u_short *)&udp->uh_sum;
3764 		flags = IPN_UDP;
3765 	} else if (oip->ip_p == IPPROTO_ICMP)
3766 		flags = IPN_ICMPQUERY;
3767 	dlen = fin->fin_plen - ((char *)dp - (char *)fin->fin_ip);
3768 
3769 	/*
3770 	 * Need to adjust ICMP header to include the real IP#'s and
3771 	 * port #'s.  Only apply a checksum change relative to the
3772 	 * IP address change as it will be modified again in ipf_nat_checkout
3773 	 * for both address and port.  Two checksum changes are
3774 	 * necessary for the two header address changes.  Be careful
3775 	 * to only modify the checksum once for the port # and twice
3776 	 * for the IP#.
3777 	 */
3778 
3779 	/*
3780 	 * Step 1
3781 	 * Fix the IP addresses in the offending IP packet. You also need
3782 	 * to adjust the IP header checksum of that offending IP packet.
3783 	 *
3784 	 * Normally, you would expect that the ICMP checksum of the
3785 	 * ICMP error message needs to be adjusted as well for the
3786 	 * IP address change in oip.
3787 	 * However, this is a NOP, because the ICMP checksum is
3788 	 * calculated over the complete ICMP packet, which includes the
3789 	 * changed oip IP addresses and oip->ip_sum. However, these
3790 	 * two changes cancel each other out (if the delta for
3791 	 * the IP address is x, then the delta for ip_sum is minus x),
3792 	 * so no change in the icmp_cksum is necessary.
3793 	 *
3794 	 * Inbound ICMP
3795 	 * ------------
3796 	 * MAP rule, SRC=a,DST=b -> SRC=c,DST=b
3797 	 * - response to outgoing packet (a,b)=>(c,b) (OIP_SRC=c,OIP_DST=b)
3798 	 * - OIP_SRC(c)=nat_newsrcip,          OIP_DST(b)=nat_newdstip
3799 	 *=> OIP_SRC(c)=nat_oldsrcip,          OIP_DST(b)=nat_olddstip
3800 	 *
3801 	 * RDR rule, SRC=a,DST=b -> SRC=a,DST=c
3802 	 * - response to outgoing packet (c,a)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3803 	 * - OIP_SRC(b)=nat_olddstip,          OIP_DST(a)=nat_oldsrcip
3804 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3805 	 *
3806 	 * REWRITE out rule, SRC=a,DST=b -> SRC=c,DST=d
3807 	 * - response to outgoing packet (a,b)=>(c,d) (OIP_SRC=c,OIP_DST=d)
3808 	 * - OIP_SRC(c)=nat_newsrcip,          OIP_DST(d)=nat_newdstip
3809 	 *=> OIP_SRC(c)=nat_oldsrcip,          OIP_DST(d)=nat_olddstip
3810 	 *
3811 	 * REWRITE in rule, SRC=a,DST=b -> SRC=c,DST=d
3812 	 * - response to outgoing packet (d,c)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3813 	 * - OIP_SRC(b)=nat_olddstip,          OIP_DST(a)=nat_oldsrcip
3814 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3815 	 *
3816 	 * Outbound ICMP
3817 	 * -------------
3818 	 * MAP rule, SRC=a,DST=b -> SRC=c,DST=b
3819 	 * - response to incoming packet (b,c)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3820 	 * - OIP_SRC(b)=nat_olddstip,          OIP_DST(a)=nat_oldsrcip
3821 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3822 	 *
3823 	 * RDR rule, SRC=a,DST=b -> SRC=a,DST=c
3824 	 * - response to incoming packet (a,b)=>(a,c) (OIP_SRC=a,OIP_DST=c)
3825 	 * - OIP_SRC(a)=nat_newsrcip,          OIP_DST(c)=nat_newdstip
3826 	 *=> OIP_SRC(a)=nat_oldsrcip,          OIP_DST(c)=nat_olddstip
3827 	 *
3828 	 * REWRITE out rule, SRC=a,DST=b -> SRC=c,DST=d
3829 	 * - response to incoming packet (d,c)=>(b,a) (OIP_SRC=c,OIP_DST=d)
3830 	 * - OIP_SRC(c)=nat_olddstip,          OIP_DST(d)=nat_oldsrcip
3831 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3832 	 *
3833 	 * REWRITE in rule, SRC=a,DST=b -> SRC=c,DST=d
3834 	 * - response to incoming packet (a,b)=>(c,d) (OIP_SRC=b,OIP_DST=a)
3835 	 * - OIP_SRC(b)=nat_newsrcip,          OIP_DST(a)=nat_newdstip
3836 	 *=> OIP_SRC(a)=nat_oldsrcip,          OIP_DST(c)=nat_olddstip
3837 	 */
3838 
3839 	if (((fin->fin_out == 0) && ((nat->nat_redir & NAT_MAP) != 0)) ||
3840 	    ((fin->fin_out == 1) && ((nat->nat_redir & NAT_REDIRECT) != 0))) {
3841 		a1.s_addr = ntohl(nat->nat_osrcaddr);
3842 		a4.s_addr = ntohl(oip->ip_src.s_addr);
3843 		a3.s_addr = ntohl(nat->nat_odstaddr);
3844 		a2.s_addr = ntohl(oip->ip_dst.s_addr);
3845 		oip->ip_src.s_addr = htonl(a1.s_addr);
3846 		oip->ip_dst.s_addr = htonl(a3.s_addr);
3847 		odst = 1;
3848 	} else {
3849 		a1.s_addr = ntohl(nat->nat_ndstaddr);
3850 		a2.s_addr = ntohl(oip->ip_dst.s_addr);
3851 		a3.s_addr = ntohl(nat->nat_nsrcaddr);
3852 		a4.s_addr = ntohl(oip->ip_src.s_addr);
3853 		oip->ip_dst.s_addr = htonl(a3.s_addr);
3854 		oip->ip_src.s_addr = htonl(a1.s_addr);
3855 		odst = 0;
3856 	}
3857 	sum1 = 0;
3858 	sum2 = 0;
3859 	sumd = 0;
3860 	CALC_SUMD(a2.s_addr, a3.s_addr, sum1);
3861 	CALC_SUMD(a4.s_addr, a1.s_addr, sum2);
3862 	sumd = sum2 + sum1;
3863 	if (sumd != 0)
3864 		ipf_fix_datacksum(&oip->ip_sum, sumd);
3865 
3866 	sumd2 = sumd;
3867 	sum1 = 0;
3868 	sum2 = 0;
3869 
3870 	/*
3871 	 * Fix UDP pseudo header checksum to compensate for the
3872 	 * IP address change.
3873 	 */
3874 	if (((flags & IPN_TCPUDP) != 0) && (dlen >= 4)) {
3875 		u_32_t sum3, sum4, sumt;
3876 
3877 		/*
3878 		 * Step 2 :
3879 		 * For offending TCP/UDP IP packets, translate the ports as
3880 		 * well, based on the NAT specification. Of course such
3881 		 * a change may be reflected in the ICMP checksum as well.
3882 		 *
3883 		 * Since the port fields are part of the TCP/UDP checksum
3884 		 * of the offending IP packet, you need to adjust that checksum
3885 		 * as well... except that the change in the port numbers should
3886 		 * be offset by the checksum change.  However, the TCP/UDP
3887 		 * checksum will also need to change if there has been an
3888 		 * IP address change.
3889 		 */
3890 		if (odst == 1) {
3891 			sum1 = ntohs(nat->nat_osport);
3892 			sum4 = ntohs(tcp->th_sport);
3893 			sum3 = ntohs(nat->nat_odport);
3894 			sum2 = ntohs(tcp->th_dport);
3895 
3896 			tcp->th_sport = htons(sum1);
3897 			tcp->th_dport = htons(sum3);
3898 		} else {
3899 			sum1 = ntohs(nat->nat_ndport);
3900 			sum2 = ntohs(tcp->th_dport);
3901 			sum3 = ntohs(nat->nat_nsport);
3902 			sum4 = ntohs(tcp->th_sport);
3903 
3904 			tcp->th_dport = htons(sum3);
3905 			tcp->th_sport = htons(sum1);
3906 		}
3907 		CALC_SUMD(sum4, sum1, sumt);
3908 		sumd += sumt;
3909 		CALC_SUMD(sum2, sum3, sumt);
3910 		sumd += sumt;
3911 
3912 		if (sumd != 0 || sumd2 != 0) {
3913 			/*
3914 			 * At this point, sumd is the delta to apply to the
3915 			 * TCP/UDP header, given the changes in both the IP
3916 			 * address and the ports and sumd2 is the delta to
3917 			 * apply to the ICMP header, given the IP address
3918 			 * change delta that may need to be applied to the
3919 			 * TCP/UDP checksum instead.
3920 			 *
3921 			 * If we will both the IP and TCP/UDP checksums
3922 			 * then the ICMP checksum changes by the address
3923 			 * delta applied to the TCP/UDP checksum.  If we
3924 			 * do not change the TCP/UDP checksum them we
3925 			 * apply the delta in ports to the ICMP checksum.
3926 			 */
3927 			if (oip->ip_p == IPPROTO_UDP) {
3928 				if ((dlen >= 8) && (*csump != 0)) {
3929 					ipf_fix_datacksum(csump, sumd);
3930 				} else {
3931 					CALC_SUMD(sum1, sum4, sumd2);
3932 					CALC_SUMD(sum3, sum2, sumt);
3933 					sumd2 += sumt;
3934 				}
3935 			} else if (oip->ip_p == IPPROTO_TCP) {
3936 				if (dlen >= 18) {
3937 					ipf_fix_datacksum(csump, sumd);
3938 				} else {
3939 					CALC_SUMD(sum1, sum4, sumd2);
3940 					CALC_SUMD(sum3, sum2, sumt);
3941 					sumd2 += sumt;
3942 				}
3943 			}
3944 			if (sumd2 != 0) {
3945 				sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3946 				sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3947 				sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3948 				ipf_fix_incksum(0, &icmp->icmp_cksum, sumd2, 0);
3949 			}
3950 		}
3951 	} else if (((flags & IPN_ICMPQUERY) != 0) && (dlen >= 8)) {
3952 		icmphdr_t *orgicmp;
3953 
3954 		/*
3955 		 * XXX - what if this is bogus hl and we go off the end ?
3956 		 * In this case, ipf_nat_icmperrorlookup() will have
3957 		 * returned NULL.
3958 		 */
3959 		orgicmp = (icmphdr_t *)dp;
3960 
3961 		if (odst == 1) {
3962 			if (orgicmp->icmp_id != nat->nat_osport) {
3963 
3964 				/*
3965 				 * Fix ICMP checksum (of the offening ICMP
3966 				 * query packet) to compensate the change
3967 				 * in the ICMP id of the offending ICMP
3968 				 * packet.
3969 				 *
3970 				 * Since you modify orgicmp->icmp_id with
3971 				 * a delta (say x) and you compensate that
3972 				 * in origicmp->icmp_cksum with a delta
3973 				 * minus x, you don't have to adjust the
3974 				 * overall icmp->icmp_cksum
3975 				 */
3976 				sum1 = ntohs(orgicmp->icmp_id);
3977 				sum2 = ntohs(nat->nat_oicmpid);
3978 				CALC_SUMD(sum1, sum2, sumd);
3979 				orgicmp->icmp_id = nat->nat_oicmpid;
3980 				ipf_fix_datacksum(&orgicmp->icmp_cksum, sumd);
3981 			}
3982 		} /* nat_dir == NAT_INBOUND is impossible for icmp queries */
3983 	}
3984 	return nat;
3985 }
3986 
3987 
3988 /*
3989  *       MAP-IN    MAP-OUT   RDR-IN   RDR-OUT
3990  * osrc    X       == src    == src      X
3991  * odst    X       == dst    == dst      X
3992  * nsrc  == dst      X         X      == dst
3993  * ndst  == src      X         X      == src
3994  * MAP = NAT_OUTBOUND, RDR = NAT_INBOUND
3995  */
3996 /*
3997  * NB: these lookups don't lock access to the list, it assumed that it has
3998  * already been done!
3999  */
4000 /* ------------------------------------------------------------------------ */
4001 /* Function:    ipf_nat_inlookup                                            */
4002 /* Returns:     nat_t* - NULL == no match,                                  */
4003 /*                       else pointer to matching NAT entry                 */
4004 /* Parameters:  fin(I)    - pointer to packet information                   */
4005 /*              flags(I)  - NAT flags for this packet                       */
4006 /*              p(I)      - protocol for this packet                        */
4007 /*              src(I)    - source IP address                               */
4008 /*              mapdst(I) - destination IP address                          */
4009 /*                                                                          */
4010 /* Lookup a nat entry based on the mapped destination ip address/port and   */
4011 /* real source address/port.  We use this lookup when receiving a packet,   */
4012 /* we're looking for a table entry, based on the destination address.       */
4013 /*                                                                          */
4014 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY.         */
4015 /*                                                                          */
4016 /* NOTE: IT IS ASSUMED THAT  IS ONLY HELD WITH A READ LOCK WHEN             */
4017 /*       THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags.             */
4018 /*                                                                          */
4019 /* flags   -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if   */
4020 /*            the packet is of said protocol                                */
4021 /* ------------------------------------------------------------------------ */
4022 nat_t *
4023 ipf_nat_inlookup(fr_info_t *fin, u_int flags, u_int p, struct in_addr src,
4024     struct in_addr mapdst)
4025 {
4026 	ipf_main_softc_t *softc = fin->fin_main_soft;
4027 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4028 	u_short sport, dport;
4029 	ipnat_t *ipn;
4030 	nat_t *nat;
4031 	int nflags;
4032 	u_32_t dst;
4033 	void *ifp;
4034 	u_int hv, rhv;
4035 
4036 	ifp = fin->fin_ifp;
4037 	dst = mapdst.s_addr;
4038 
4039 	switch (p)
4040 	{
4041 	case IPPROTO_TCP :
4042 	case IPPROTO_UDP :
4043 		sport = htons(fin->fin_data[0]);
4044 		dport = htons(fin->fin_data[1]);
4045 		break;
4046 	case IPPROTO_ICMP :
4047 		sport = 0;
4048 		dport = fin->fin_data[1];
4049 		break;
4050 	default :
4051 		sport = 0;
4052 		dport = 0;
4053 		break;
4054 	}
4055 
4056 
4057 	if ((flags & SI_WILDP) != 0)
4058 		goto find_in_wild_ports;
4059 
4060 	rhv = NAT_HASH_FN(dst, dport, 0xffffffff);
4061 	rhv = NAT_HASH_FN(src.s_addr, rhv + sport, 0xffffffff);
4062 	hv = rhv % softn->ipf_nat_table_sz;
4063 	nat = softn->ipf_nat_table[1][hv];
4064 	/* TRACE dst, dport, src, sport, hv, nat */
4065 
4066 	for (; nat; nat = nat->nat_hnext[1]) {
4067 		if (nat->nat_ifps[0] != NULL) {
4068 			if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
4069 				continue;
4070 		}
4071 
4072 		if (nat->nat_pr[0] != p)
4073 			continue;
4074 
4075 		switch (nat->nat_dir)
4076 		{
4077 		case NAT_INBOUND :
4078 		case NAT_DIVERTIN :
4079 			if (nat->nat_v[0] != 4)
4080 				continue;
4081 			if (nat->nat_osrcaddr != src.s_addr ||
4082 			    nat->nat_odstaddr != dst)
4083 				continue;
4084 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4085 				if (nat->nat_osport != sport)
4086 					continue;
4087 				if (nat->nat_odport != dport)
4088 					continue;
4089 
4090 			} else if (p == IPPROTO_ICMP) {
4091 				if (nat->nat_oicmpid != dport) {
4092 					continue;
4093 				}
4094 			}
4095 			break;
4096 		case NAT_DIVERTOUT :
4097 			if (nat->nat_dlocal)
4098 				continue;
4099 			/* FALLTHROUGH */
4100 		case NAT_OUTBOUND :
4101 			if (nat->nat_v[1] != 4)
4102 				continue;
4103 			if (nat->nat_dlocal)
4104 				continue;
4105 			if (nat->nat_dlocal)
4106 				continue;
4107 			if (nat->nat_ndstaddr != src.s_addr ||
4108 			    nat->nat_nsrcaddr != dst)
4109 				continue;
4110 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4111 				if (nat->nat_ndport != sport)
4112 					continue;
4113 				if (nat->nat_nsport != dport)
4114 					continue;
4115 
4116 			} else if (p == IPPROTO_ICMP) {
4117 				if (nat->nat_nicmpid != dport) {
4118 					continue;
4119 				}
4120 			}
4121 			break;
4122 		}
4123 
4124 
4125 		if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4126 			ipn = nat->nat_ptr;
4127 			if ((ipn != NULL) && (nat->nat_aps != NULL))
4128 				if (ipf_proxy_match(fin, nat) != 0)
4129 					continue;
4130 		}
4131 		if ((nat->nat_ifps[0] == NULL) && (ifp != NULL)) {
4132 			nat->nat_ifps[0] = ifp;
4133 			nat->nat_mtu[0] = GETIFMTU_4(ifp);
4134 		}
4135 		return nat;
4136 	}
4137 
4138 	/*
4139 	 * So if we didn't find it but there are wildcard members in the hash
4140 	 * table, go back and look for them.  We do this search and update here
4141 	 * because it is modifying the NAT table and we want to do this only
4142 	 * for the first packet that matches.  The exception, of course, is
4143 	 * for "dummy" (FI_IGNORE) lookups.
4144 	 */
4145 find_in_wild_ports:
4146 	if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH)) {
4147 		NBUMPSIDEX(0, ns_lookup_miss, ns_lookup_miss_0);
4148 		return NULL;
4149 	}
4150 	if (softn->ipf_nat_stats.ns_wilds == 0 || (fin->fin_flx & FI_NOWILD)) {
4151 		NBUMPSIDEX(0, ns_lookup_nowild, ns_lookup_nowild_0);
4152 		return NULL;
4153 	}
4154 
4155 	RWLOCK_EXIT(&softc->ipf_nat);
4156 
4157 	hv = NAT_HASH_FN(dst, 0, 0xffffffff);
4158 	hv = NAT_HASH_FN(src.s_addr, hv, softn->ipf_nat_table_sz);
4159 	WRITE_ENTER(&softc->ipf_nat);
4160 
4161 	nat = softn->ipf_nat_table[1][hv];
4162 	/* TRACE dst, src, hv, nat */
4163 	for (; nat; nat = nat->nat_hnext[1]) {
4164 		if (nat->nat_ifps[0] != NULL) {
4165 			if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
4166 				continue;
4167 		}
4168 
4169 		if (nat->nat_pr[0] != fin->fin_p)
4170 			continue;
4171 
4172 		switch (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND))
4173 		{
4174 		case NAT_INBOUND :
4175 			if (nat->nat_v[0] != 4)
4176 				continue;
4177 			if (nat->nat_osrcaddr != src.s_addr ||
4178 			    nat->nat_odstaddr != dst)
4179 				continue;
4180 			break;
4181 		case NAT_OUTBOUND :
4182 			if (nat->nat_v[1] != 4)
4183 				continue;
4184 			if (nat->nat_ndstaddr != src.s_addr ||
4185 			    nat->nat_nsrcaddr != dst)
4186 				continue;
4187 			break;
4188 		}
4189 
4190 		nflags = nat->nat_flags;
4191 		if (!(nflags & (NAT_TCPUDP|SI_WILDP)))
4192 			continue;
4193 
4194 		if (ipf_nat_wildok(nat, (int)sport, (int)dport, nflags,
4195 				   NAT_INBOUND) == 1) {
4196 			if ((fin->fin_flx & FI_IGNORE) != 0)
4197 				break;
4198 			if ((nflags & SI_CLONE) != 0) {
4199 				nat = ipf_nat_clone(fin, nat);
4200 				if (nat == NULL)
4201 					break;
4202 			} else {
4203 				MUTEX_ENTER(&softn->ipf_nat_new);
4204 				softn->ipf_nat_stats.ns_wilds--;
4205 				MUTEX_EXIT(&softn->ipf_nat_new);
4206 			}
4207 
4208 			if (nat->nat_dir == NAT_INBOUND) {
4209 				if (nat->nat_osport == 0) {
4210 					nat->nat_osport = sport;
4211 					nat->nat_nsport = sport;
4212 				}
4213 				if (nat->nat_odport == 0) {
4214 					nat->nat_odport = dport;
4215 					nat->nat_ndport = dport;
4216 				}
4217 			} else if (nat->nat_dir == NAT_OUTBOUND) {
4218 				if (nat->nat_osport == 0) {
4219 					nat->nat_osport = dport;
4220 					nat->nat_nsport = dport;
4221 				}
4222 				if (nat->nat_odport == 0) {
4223 					nat->nat_odport = sport;
4224 					nat->nat_ndport = sport;
4225 				}
4226 			}
4227 			if ((nat->nat_ifps[0] == NULL) && (ifp != NULL)) {
4228 				nat->nat_ifps[0] = ifp;
4229 				nat->nat_mtu[0] = GETIFMTU_4(ifp);
4230 			}
4231 			nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
4232 			ipf_nat_tabmove(softn, nat);
4233 			break;
4234 		}
4235 	}
4236 
4237 	MUTEX_DOWNGRADE(&softc->ipf_nat);
4238 
4239 	if (nat == NULL) {
4240 		NBUMPSIDE(0, ns_lookup_miss);
4241 	}
4242 	return nat;
4243 }
4244 
4245 
4246 /* ------------------------------------------------------------------------ */
4247 /* Function:    ipf_nat_tabmove                                             */
4248 /* Returns:     Nil                                                         */
4249 /* Parameters:  softn(I) - pointer to NAT context structure                 */
4250 /*              nat(I)   - pointer to NAT structure                         */
4251 /* Write Lock:  ipf_nat                                                     */
4252 /*                                                                          */
4253 /* This function is only called for TCP/UDP NAT table entries where the     */
4254 /* original was placed in the table without hashing on the ports and we now */
4255 /* want to include hashing on port numbers.                                 */
4256 /* ------------------------------------------------------------------------ */
4257 static void
4258 ipf_nat_tabmove(ipf_nat_softc_t *softn, nat_t *nat)
4259 {
4260 	u_int hv0, hv1, rhv0, rhv1;
4261 	natstat_t *nsp;
4262 	nat_t **natp;
4263 
4264 	if (nat->nat_flags & SI_CLONE)
4265 		return;
4266 
4267 	nsp = &softn->ipf_nat_stats;
4268 	/*
4269 	 * Remove the NAT entry from the old location
4270 	 */
4271 	if (nat->nat_hnext[0])
4272 		nat->nat_hnext[0]->nat_phnext[0] = nat->nat_phnext[0];
4273 	*nat->nat_phnext[0] = nat->nat_hnext[0];
4274 	hv0 = nat->nat_hv[0] % softn->ipf_nat_table_sz;
4275 	hv1 = nat->nat_hv[1] % softn->ipf_nat_table_sz;
4276 
4277 	ASSERT(nsp->ns_side[0].ns_bucketlen[hv0] > 0);
4278 	nsp->ns_side[0].ns_bucketlen[hv0]--;
4279 
4280 	if (nat->nat_hnext[1])
4281 		nat->nat_hnext[1]->nat_phnext[1] = nat->nat_phnext[1];
4282 	*nat->nat_phnext[1] = nat->nat_hnext[1];
4283 	ASSERT(nsp->ns_side[1].ns_bucketlen[hv1] > 0);
4284 	nsp->ns_side[1].ns_bucketlen[hv1]--;
4285 
4286 	/*
4287 	 * Add into the NAT table in the new position
4288 	 */
4289 	rhv0 = NAT_HASH_FN(nat->nat_osrcaddr, nat->nat_osport, 0xffffffff);
4290 	rhv0 = NAT_HASH_FN(nat->nat_odstaddr, rhv0 + nat->nat_odport,
4291 			   0xffffffff);
4292 	rhv1 = NAT_HASH_FN(nat->nat_nsrcaddr, nat->nat_nsport, 0xffffffff);
4293 	rhv1 = NAT_HASH_FN(nat->nat_ndstaddr, rhv1 + nat->nat_ndport,
4294 			   0xffffffff);
4295 
4296 	if ((nat->nat_dir & NAT_OUTBOUND) == NAT_OUTBOUND) {
4297 		nat->nat_hv[0] = rhv0;
4298 		nat->nat_hv[1] = rhv1;
4299 	} else {
4300 		nat->nat_hv[0] = rhv1;
4301 		nat->nat_hv[1] = rhv0;
4302 	}
4303 
4304 	hv0 = nat->nat_hv[0] % softn->ipf_nat_table_sz;
4305 	hv1 = nat->nat_hv[1] % softn->ipf_nat_table_sz;
4306 
4307 	/* TRACE nat_osrcaddr, nat_osport, nat_odstaddr, nat_odport, hv0 */
4308 	/* TRACE nat_nsrcaddr, nat_nsport, nat_ndstaddr, nat_ndport, hv1 */
4309 
4310 	natp = &softn->ipf_nat_table[0][hv0];
4311 	if (*natp)
4312 		(*natp)->nat_phnext[0] = &nat->nat_hnext[0];
4313 	nat->nat_phnext[0] = natp;
4314 	nat->nat_hnext[0] = *natp;
4315 	*natp = nat;
4316 	nsp->ns_side[0].ns_bucketlen[hv0]++;
4317 
4318 	natp = &softn->ipf_nat_table[1][hv1];
4319 	if (*natp)
4320 		(*natp)->nat_phnext[1] = &nat->nat_hnext[1];
4321 	nat->nat_phnext[1] = natp;
4322 	nat->nat_hnext[1] = *natp;
4323 	*natp = nat;
4324 	nsp->ns_side[1].ns_bucketlen[hv1]++;
4325 }
4326 
4327 
4328 /* ------------------------------------------------------------------------ */
4329 /* Function:    ipf_nat_outlookup                                           */
4330 /* Returns:     nat_t* - NULL == no match,                                  */
4331 /*                       else pointer to matching NAT entry                 */
4332 /* Parameters:  fin(I)   - pointer to packet information                    */
4333 /*              flags(I) - NAT flags for this packet                        */
4334 /*              p(I)     - protocol for this packet                         */
4335 /*              src(I)   - source IP address                                */
4336 /*              dst(I)   - destination IP address                           */
4337 /*              rw(I)    - 1 == write lock on  held, 0 == read lock.        */
4338 /*                                                                          */
4339 /* Lookup a nat entry based on the source 'real' ip address/port and        */
4340 /* destination address/port.  We use this lookup when sending a packet out, */
4341 /* we're looking for a table entry, based on the source address.            */
4342 /*                                                                          */
4343 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY.         */
4344 /*                                                                          */
4345 /* NOTE: IT IS ASSUMED THAT  IS ONLY HELD WITH A READ LOCK WHEN             */
4346 /*       THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags.             */
4347 /*                                                                          */
4348 /* flags   -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if   */
4349 /*            the packet is of said protocol                                */
4350 /* ------------------------------------------------------------------------ */
4351 nat_t *
4352 ipf_nat_outlookup(fr_info_t *fin, u_int flags, u_int p, struct in_addr src,
4353    struct in_addr dst)
4354 {
4355 	ipf_main_softc_t *softc = fin->fin_main_soft;
4356 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4357 	u_short sport, dport;
4358 	ipnat_t *ipn;
4359 	nat_t *nat;
4360 	void *ifp;
4361 	u_int hv;
4362 
4363 	ifp = fin->fin_ifp;
4364 
4365 	switch (p)
4366 	{
4367 	case IPPROTO_TCP :
4368 	case IPPROTO_UDP :
4369 		sport = htons(fin->fin_data[0]);
4370 		dport = htons(fin->fin_data[1]);
4371 		break;
4372 	case IPPROTO_ICMP :
4373 		sport = 0;
4374 		dport = fin->fin_data[1];
4375 		break;
4376 	default :
4377 		sport = 0;
4378 		dport = 0;
4379 		break;
4380 	}
4381 
4382 	if ((flags & SI_WILDP) != 0)
4383 		goto find_out_wild_ports;
4384 
4385 	hv = NAT_HASH_FN(src.s_addr, sport, 0xffffffff);
4386 	hv = NAT_HASH_FN(dst.s_addr, hv + dport, softn->ipf_nat_table_sz);
4387 	nat = softn->ipf_nat_table[0][hv];
4388 
4389 	/* TRACE src, sport, dst, dport, hv, nat */
4390 
4391 	for (; nat; nat = nat->nat_hnext[0]) {
4392 		if (nat->nat_ifps[1] != NULL) {
4393 			if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
4394 				continue;
4395 		}
4396 
4397 		if (nat->nat_pr[1] != p)
4398 			continue;
4399 
4400 		switch (nat->nat_dir)
4401 		{
4402 		case NAT_INBOUND :
4403 		case NAT_DIVERTIN :
4404 			if (nat->nat_v[1] != 4)
4405 				continue;
4406 			if (nat->nat_ndstaddr != src.s_addr ||
4407 			    nat->nat_nsrcaddr != dst.s_addr)
4408 				continue;
4409 
4410 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4411 				if (nat->nat_ndport != sport)
4412 					continue;
4413 				if (nat->nat_nsport != dport)
4414 					continue;
4415 
4416 			} else if (p == IPPROTO_ICMP) {
4417 				if (nat->nat_nicmpid != dport) {
4418 					continue;
4419 				}
4420 			}
4421 			break;
4422 		case NAT_OUTBOUND :
4423 		case NAT_DIVERTOUT :
4424 			if (nat->nat_v[0] != 4)
4425 				continue;
4426 			if (nat->nat_osrcaddr != src.s_addr ||
4427 			    nat->nat_odstaddr != dst.s_addr)
4428 				continue;
4429 
4430 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4431 				if (nat->nat_odport != dport)
4432 					continue;
4433 				if (nat->nat_osport != sport)
4434 					continue;
4435 
4436 			} else if (p == IPPROTO_ICMP) {
4437 				if (nat->nat_oicmpid != dport) {
4438 					continue;
4439 				}
4440 			}
4441 			break;
4442 		}
4443 
4444 		ipn = nat->nat_ptr;
4445 		if ((ipn != NULL) && (nat->nat_aps != NULL))
4446 			if (ipf_proxy_match(fin, nat) != 0)
4447 				continue;
4448 
4449 		if ((nat->nat_ifps[1] == NULL) && (ifp != NULL)) {
4450 			nat->nat_ifps[1] = ifp;
4451 			nat->nat_mtu[1] = GETIFMTU_4(ifp);
4452 		}
4453 		return nat;
4454 	}
4455 
4456 	/*
4457 	 * So if we didn't find it but there are wildcard members in the hash
4458 	 * table, go back and look for them.  We do this search and update here
4459 	 * because it is modifying the NAT table and we want to do this only
4460 	 * for the first packet that matches.  The exception, of course, is
4461 	 * for "dummy" (FI_IGNORE) lookups.
4462 	 */
4463 find_out_wild_ports:
4464 	if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH)) {
4465 		NBUMPSIDEX(1, ns_lookup_miss, ns_lookup_miss_1);
4466 		return NULL;
4467 	}
4468 	if (softn->ipf_nat_stats.ns_wilds == 0 || (fin->fin_flx & FI_NOWILD)) {
4469 		NBUMPSIDEX(1, ns_lookup_nowild, ns_lookup_nowild_1);
4470 		return NULL;
4471 	}
4472 
4473 	RWLOCK_EXIT(&softc->ipf_nat);
4474 
4475 	hv = NAT_HASH_FN(src.s_addr, 0, 0xffffffff);
4476 	hv = NAT_HASH_FN(dst.s_addr, hv, softn->ipf_nat_table_sz);
4477 
4478 	WRITE_ENTER(&softc->ipf_nat);
4479 
4480 	nat = softn->ipf_nat_table[0][hv];
4481 	for (; nat; nat = nat->nat_hnext[0]) {
4482 		if (nat->nat_ifps[1] != NULL) {
4483 			if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
4484 				continue;
4485 		}
4486 
4487 		if (nat->nat_pr[1] != fin->fin_p)
4488 			continue;
4489 
4490 		switch (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND))
4491 		{
4492 		case NAT_INBOUND :
4493 			if (nat->nat_v[1] != 4)
4494 				continue;
4495 			if (nat->nat_ndstaddr != src.s_addr ||
4496 			    nat->nat_nsrcaddr != dst.s_addr)
4497 				continue;
4498 			break;
4499 		case NAT_OUTBOUND :
4500 			if (nat->nat_v[0] != 4)
4501 				continue;
4502 			if (nat->nat_osrcaddr != src.s_addr ||
4503 			    nat->nat_odstaddr != dst.s_addr)
4504 				continue;
4505 			break;
4506 		}
4507 
4508 		if (!(nat->nat_flags & (NAT_TCPUDP|SI_WILDP)))
4509 			continue;
4510 
4511 		if (ipf_nat_wildok(nat, (int)sport, (int)dport, nat->nat_flags,
4512 				   NAT_OUTBOUND) == 1) {
4513 			if ((fin->fin_flx & FI_IGNORE) != 0)
4514 				break;
4515 			if ((nat->nat_flags & SI_CLONE) != 0) {
4516 				nat = ipf_nat_clone(fin, nat);
4517 				if (nat == NULL)
4518 					break;
4519 			} else {
4520 				MUTEX_ENTER(&softn->ipf_nat_new);
4521 				softn->ipf_nat_stats.ns_wilds--;
4522 				MUTEX_EXIT(&softn->ipf_nat_new);
4523 			}
4524 
4525 			if (nat->nat_dir == NAT_OUTBOUND) {
4526 				if (nat->nat_osport == 0) {
4527 					nat->nat_osport = sport;
4528 					nat->nat_nsport = sport;
4529 				}
4530 				if (nat->nat_odport == 0) {
4531 					nat->nat_odport = dport;
4532 					nat->nat_ndport = dport;
4533 				}
4534 			} else if (nat->nat_dir == NAT_INBOUND) {
4535 				if (nat->nat_osport == 0) {
4536 					nat->nat_osport = dport;
4537 					nat->nat_nsport = dport;
4538 				}
4539 				if (nat->nat_odport == 0) {
4540 					nat->nat_odport = sport;
4541 					nat->nat_ndport = sport;
4542 				}
4543 			}
4544 			if ((nat->nat_ifps[1] == NULL) && (ifp != NULL)) {
4545 				nat->nat_ifps[1] = ifp;
4546 				nat->nat_mtu[1] = GETIFMTU_4(ifp);
4547 			}
4548 			nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
4549 			ipf_nat_tabmove(softn, nat);
4550 			break;
4551 		}
4552 	}
4553 
4554 	MUTEX_DOWNGRADE(&softc->ipf_nat);
4555 
4556 	if (nat == NULL) {
4557 		NBUMPSIDE(1, ns_lookup_miss);
4558 	}
4559 	return nat;
4560 }
4561 
4562 
4563 /* ------------------------------------------------------------------------ */
4564 /* Function:    ipf_nat_lookupredir                                         */
4565 /* Returns:     nat_t* - NULL == no match,                                  */
4566 /*                       else pointer to matching NAT entry                 */
4567 /* Parameters:  softc(I) - pointer to soft context main structure           */
4568 /*              np(I)    - pointer to description of packet to find NAT     */
4569 /*                         table entry for.                                 */
4570 /*                                                                          */
4571 /* Lookup the NAT tables to search for a matching redirect                  */
4572 /* The contents of natlookup_t should imitate those found in a packet that  */
4573 /* would be translated - ie a packet coming in for RDR or going out for MAP.*/
4574 /* We can do the lookup in one of two ways, imitating an inbound or         */
4575 /* outbound  packet.  By default we assume outbound, unless IPN_IN is set.  */
4576 /* For IN, the fields are set as follows:                                   */
4577 /*     nl_real* = source information                                        */
4578 /*     nl_out* = destination information (translated)                       */
4579 /* For an out packet, the fields are set like this:                         */
4580 /*     nl_in* = source information (untranslated)                           */
4581 /*     nl_out* = destination information (translated)                       */
4582 /* ------------------------------------------------------------------------ */
4583 nat_t *
4584 ipf_nat_lookupredir(ipf_main_softc_t *softc, natlookup_t *np)
4585 {
4586 	fr_info_t fi;
4587 	nat_t *nat;
4588 
4589 	bzero((char *)&fi, sizeof(fi));
4590 	fi.fin_main_soft = softc;
4591 	if (np->nl_flags & IPN_IN) {
4592 		fi.fin_data[0] = ntohs(np->nl_realport);
4593 		fi.fin_data[1] = ntohs(np->nl_outport);
4594 	} else {
4595 		fi.fin_data[0] = ntohs(np->nl_inport);
4596 		fi.fin_data[1] = ntohs(np->nl_outport);
4597 	}
4598 	if (np->nl_flags & IPN_TCP)
4599 		fi.fin_p = IPPROTO_TCP;
4600 	else if (np->nl_flags & IPN_UDP)
4601 		fi.fin_p = IPPROTO_UDP;
4602 	else if (np->nl_flags & (IPN_ICMPERR|IPN_ICMPQUERY))
4603 		fi.fin_p = IPPROTO_ICMP;
4604 
4605 	/*
4606 	 * We can do two sorts of lookups:
4607 	 * - IPN_IN: we have the `real' and `out' address, look for `in'.
4608 	 * - default: we have the `in' and `out' address, look for `real'.
4609 	 */
4610 	if (np->nl_flags & IPN_IN) {
4611 		if ((nat = ipf_nat_inlookup(&fi, np->nl_flags, fi.fin_p,
4612 					    np->nl_realip, np->nl_outip))) {
4613 			np->nl_inip = nat->nat_odstip;
4614 			np->nl_inport = nat->nat_odport;
4615 		}
4616 	} else {
4617 		/*
4618 		 * If nl_inip is non null, this is a lookup based on the real
4619 		 * ip address. Else, we use the fake.
4620 		 */
4621 		if ((nat = ipf_nat_outlookup(&fi, np->nl_flags, fi.fin_p,
4622 					 np->nl_inip, np->nl_outip))) {
4623 
4624 			if ((np->nl_flags & IPN_FINDFORWARD) != 0) {
4625 				fr_info_t fin;
4626 				bzero((char *)&fin, sizeof(fin));
4627 				fin.fin_p = nat->nat_pr[0];
4628 				fin.fin_data[0] = ntohs(nat->nat_ndport);
4629 				fin.fin_data[1] = ntohs(nat->nat_nsport);
4630 				if (ipf_nat_inlookup(&fin, np->nl_flags,
4631 						     fin.fin_p, nat->nat_ndstip,
4632 						     nat->nat_nsrcip) != NULL) {
4633 					np->nl_flags &= ~IPN_FINDFORWARD;
4634 				}
4635 			}
4636 
4637 			np->nl_realip = nat->nat_odstip;
4638 			np->nl_realport = nat->nat_odport;
4639 		}
4640  	}
4641 
4642 	return nat;
4643 }
4644 
4645 
4646 /* ------------------------------------------------------------------------ */
4647 /* Function:    ipf_nat_match                                               */
4648 /* Returns:     int - 0 == no match, 1 == match                             */
4649 /* Parameters:  fin(I)   - pointer to packet information                    */
4650 /*              np(I)    - pointer to NAT rule                              */
4651 /*                                                                          */
4652 /* Pull the matching of a packet against a NAT rule out of that complex     */
4653 /* loop inside ipf_nat_checkin() and lay it out properly in its own function. */
4654 /* ------------------------------------------------------------------------ */
4655 static int
4656 ipf_nat_match(fr_info_t *fin, ipnat_t *np)
4657 {
4658 	ipf_main_softc_t *softc = fin->fin_main_soft;
4659 	frtuc_t *ft;
4660 	int match;
4661 
4662 	match = 0;
4663 	switch (np->in_osrcatype)
4664 	{
4665 	case FRI_NORMAL :
4666 		match = ((fin->fin_saddr & np->in_osrcmsk) != np->in_osrcaddr);
4667 		break;
4668 	case FRI_LOOKUP :
4669 		match = (*np->in_osrcfunc)(softc, np->in_osrcptr,
4670 					   4, &fin->fin_saddr, fin->fin_plen);
4671 		break;
4672 	}
4673 	match ^= ((np->in_flags & IPN_NOTSRC) != 0);
4674 	if (match)
4675 		return 0;
4676 
4677 	match = 0;
4678 	switch (np->in_odstatype)
4679 	{
4680 	case FRI_NORMAL :
4681 		match = ((fin->fin_daddr & np->in_odstmsk) != np->in_odstaddr);
4682 		break;
4683 	case FRI_LOOKUP :
4684 		match = (*np->in_odstfunc)(softc, np->in_odstptr,
4685 					   4, &fin->fin_daddr, fin->fin_plen);
4686 		break;
4687 	}
4688 
4689 	match ^= ((np->in_flags & IPN_NOTDST) != 0);
4690 	if (match)
4691 		return 0;
4692 
4693 	ft = &np->in_tuc;
4694 	if (!(fin->fin_flx & FI_TCPUDP) ||
4695 	    (fin->fin_flx & (FI_SHORT|FI_FRAGBODY))) {
4696 		if (ft->ftu_scmp || ft->ftu_dcmp)
4697 			return 0;
4698 		return 1;
4699 	}
4700 
4701 	return ipf_tcpudpchk(&fin->fin_fi, ft);
4702 }
4703 
4704 
4705 /* ------------------------------------------------------------------------ */
4706 /* Function:    ipf_nat_update                                              */
4707 /* Returns:     Nil                                                         */
4708 /* Parameters:  fin(I) - pointer to packet information                      */
4709 /*              nat(I) - pointer to NAT structure                           */
4710 /*                                                                          */
4711 /* Updates the lifetime of a NAT table entry for non-TCP packets.  Must be  */
4712 /* called with fin_rev updated - i.e. after calling ipf_nat_proto().        */
4713 /*                                                                          */
4714 /* This *MUST* be called after ipf_nat_proto() as it expects fin_rev to     */
4715 /* already be set.                                                          */
4716 /* ------------------------------------------------------------------------ */
4717 void
4718 ipf_nat_update(fr_info_t *fin, nat_t *nat)
4719 {
4720 	ipf_main_softc_t *softc = fin->fin_main_soft;
4721 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4722 	ipftq_t *ifq, *ifq2;
4723 	ipftqent_t *tqe;
4724 	ipnat_t *np = nat->nat_ptr;
4725 
4726 	tqe = &nat->nat_tqe;
4727 	ifq = tqe->tqe_ifq;
4728 
4729 	/*
4730 	 * We allow over-riding of NAT timeouts from NAT rules, even for
4731 	 * TCP, however, if it is TCP and there is no rule timeout set,
4732 	 * then do not update the timeout here.
4733 	 */
4734 	if (np != NULL) {
4735 		np->in_bytes[fin->fin_rev] += fin->fin_plen;
4736 		ifq2 = np->in_tqehead[fin->fin_rev];
4737 	} else {
4738 		ifq2 = NULL;
4739 	}
4740 
4741 	if (nat->nat_pr[0] == IPPROTO_TCP && ifq2 == NULL) {
4742 		(void) ipf_tcp_age(&nat->nat_tqe, fin, softn->ipf_nat_tcptq,
4743 				   0, 2);
4744 	} else {
4745 		if (ifq2 == NULL) {
4746 			if (nat->nat_pr[0] == IPPROTO_UDP)
4747 				ifq2 = fin->fin_rev ? &softn->ipf_nat_udpacktq :
4748 						      &softn->ipf_nat_udptq;
4749 			else if (nat->nat_pr[0] == IPPROTO_ICMP ||
4750 				 nat->nat_pr[0] == IPPROTO_ICMPV6)
4751 				ifq2 = fin->fin_rev ? &softn->ipf_nat_icmpacktq:
4752 						      &softn->ipf_nat_icmptq;
4753 			else
4754 				ifq2 = &softn->ipf_nat_iptq;
4755 		}
4756 
4757 		ipf_movequeue(softc->ipf_ticks, tqe, ifq, ifq2);
4758 	}
4759 }
4760 
4761 
4762 /* ------------------------------------------------------------------------ */
4763 /* Function:    ipf_nat_checkout                                            */
4764 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
4765 /*                     0 == no packet translation occurred,                 */
4766 /*                     1 == packet was successfully translated.             */
4767 /* Parameters:  fin(I)   - pointer to packet information                    */
4768 /*              passp(I) - pointer to filtering result flags                */
4769 /*                                                                          */
4770 /* Check to see if an outcoming packet should be changed.  ICMP packets are */
4771 /* first checked to see if they match an existing entry (if an error),      */
4772 /* otherwise a search of the current NAT table is made.  If neither results */
4773 /* in a match then a search for a matching NAT rule is made.  Create a new  */
4774 /* NAT entry if a we matched a NAT rule.  Lastly, actually change the       */
4775 /* packet header(s) as required.                                            */
4776 /* ------------------------------------------------------------------------ */
4777 int
4778 ipf_nat_checkout(fr_info_t *fin, u_32_t *passp)
4779 {
4780 	ipnat_t *np = NULL, *npnext;
4781 	struct ifnet *ifp, *sifp;
4782 	ipf_main_softc_t *softc;
4783 	ipf_nat_softc_t *softn;
4784 	tcphdr_t *tcp = NULL;
4785 	int rval, natfailed;
4786 	u_int nflags = 0;
4787 	u_32_t ipa, iph;
4788 	int natadd = 1;
4789 	frentry_t *fr;
4790 	nat_t *nat;
4791 
4792 	if (fin->fin_v == 6) {
4793 #ifdef USE_INET6
4794 		return ipf_nat6_checkout(fin, passp);
4795 #else
4796 		return 0;
4797 #endif
4798 	}
4799 
4800 	softc = fin->fin_main_soft;
4801 	softn = softc->ipf_nat_soft;
4802 
4803 	if (softn->ipf_nat_lock != 0)
4804 		return 0;
4805 	if (softn->ipf_nat_stats.ns_rules == 0 &&
4806 	    softn->ipf_nat_instances == NULL)
4807 		return 0;
4808 
4809 	natfailed = 0;
4810 	fr = fin->fin_fr;
4811 	sifp = fin->fin_ifp;
4812 	if (fr != NULL) {
4813 		ifp = fr->fr_tifs[fin->fin_rev].fd_ptr;
4814 		if ((ifp != NULL) && (ifp != (void *)-1))
4815 			fin->fin_ifp = ifp;
4816 	}
4817 	ifp = fin->fin_ifp;
4818 
4819 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
4820 		switch (fin->fin_p)
4821 		{
4822 		case IPPROTO_TCP :
4823 			nflags = IPN_TCP;
4824 			break;
4825 		case IPPROTO_UDP :
4826 			nflags = IPN_UDP;
4827 			break;
4828 		case IPPROTO_ICMP :
4829 			/*
4830 			 * This is an incoming packet, so the destination is
4831 			 * the icmp_id and the source port equals 0
4832 			 */
4833 			if ((fin->fin_flx & FI_ICMPQUERY) != 0)
4834 				nflags = IPN_ICMPQUERY;
4835 			break;
4836 		default :
4837 			break;
4838 		}
4839 
4840 		if ((nflags & IPN_TCPUDP))
4841 			tcp = fin->fin_dp;
4842 	}
4843 
4844 	ipa = fin->fin_saddr;
4845 
4846 	READ_ENTER(&softc->ipf_nat);
4847 
4848 	if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
4849 	    (nat = ipf_nat_icmperror(fin, &nflags, NAT_OUTBOUND)))
4850 		/*EMPTY*/;
4851 	else if ((fin->fin_flx & FI_FRAG) && (nat = ipf_frag_natknown(fin)))
4852 		natadd = 0;
4853 	else if ((nat = ipf_nat_outlookup(fin, nflags|NAT_SEARCH,
4854 				      (u_int)fin->fin_p, fin->fin_src,
4855 				      fin->fin_dst))) {
4856 		nflags = nat->nat_flags;
4857 	} else if (fin->fin_off == 0) {
4858 		u_32_t hv, msk, nmsk = 0;
4859 
4860 		/*
4861 		 * If there is no current entry in the nat table for this IP#,
4862 		 * create one for it (if there is a matching rule).
4863 		 */
4864 maskloop:
4865 		msk = softn->ipf_nat_map_active_masks[nmsk];
4866 		iph = ipa & msk;
4867 		hv = NAT_HASH_FN(iph, 0, softn->ipf_nat_maprules_sz);
4868 retry_roundrobin:
4869 		for (np = softn->ipf_nat_map_rules[hv]; np; np = npnext) {
4870 			npnext = np->in_mnext;
4871 			if ((np->in_ifps[1] && (np->in_ifps[1] != ifp)))
4872 				continue;
4873 			if (np->in_v[0] != 4)
4874 				continue;
4875 			if (np->in_pr[1] && (np->in_pr[1] != fin->fin_p))
4876 				continue;
4877 			if ((np->in_flags & IPN_RF) &&
4878 			    !(np->in_flags & nflags))
4879 				continue;
4880 			if (np->in_flags & IPN_FILTER) {
4881 				switch (ipf_nat_match(fin, np))
4882 				{
4883 				case 0 :
4884 					continue;
4885 				case -1 :
4886 					rval = -1;
4887 					goto outmatchfail;
4888 				case 1 :
4889 				default :
4890 					break;
4891 				}
4892 			} else if ((ipa & np->in_osrcmsk) != np->in_osrcaddr)
4893 				continue;
4894 
4895 			if ((fr != NULL) &&
4896 			    !ipf_matchtag(&np->in_tag, &fr->fr_nattag))
4897 				continue;
4898 
4899 			if (np->in_plabel != -1) {
4900 				if (((np->in_flags & IPN_FILTER) == 0) &&
4901 				    (np->in_odport != fin->fin_data[1]))
4902 					continue;
4903 				if (ipf_proxy_ok(fin, tcp, np) == 0)
4904 					continue;
4905 			}
4906 
4907 			if (np->in_flags & IPN_NO) {
4908 				np->in_hits++;
4909 				break;
4910 			}
4911 			MUTEX_ENTER(&softn->ipf_nat_new);
4912 			/*
4913 			 * If we've matched a round-robin rule but it has
4914 			 * moved in the list since we got it, start over as
4915 			 * this is now no longer correct.
4916 			 */
4917 			if (npnext != np->in_mnext) {
4918 				if ((np->in_flags & IPN_ROUNDR) != 0) {
4919 					MUTEX_EXIT(&softn->ipf_nat_new);
4920 					goto retry_roundrobin;
4921 				}
4922 				npnext = np->in_mnext;
4923 			}
4924 
4925 			nat = ipf_nat_add(fin, np, NULL, nflags, NAT_OUTBOUND);
4926 			MUTEX_EXIT(&softn->ipf_nat_new);
4927 			if (nat != NULL) {
4928 				natfailed = 0;
4929 				break;
4930 			}
4931 			natfailed = -1;
4932 		}
4933 		if ((np == NULL) && (nmsk < softn->ipf_nat_map_max)) {
4934 			nmsk++;
4935 			goto maskloop;
4936 		}
4937 	}
4938 
4939 	if (nat != NULL) {
4940 		rval = ipf_nat_out(fin, nat, natadd, nflags);
4941 		if (rval == 1) {
4942 			MUTEX_ENTER(&nat->nat_lock);
4943 			ipf_nat_update(fin, nat);
4944 			nat->nat_bytes[1] += fin->fin_plen;
4945 			nat->nat_pkts[1]++;
4946 			fin->fin_pktnum = nat->nat_pkts[1];
4947 			MUTEX_EXIT(&nat->nat_lock);
4948 		}
4949 	} else
4950 		rval = natfailed;
4951 outmatchfail:
4952 	RWLOCK_EXIT(&softc->ipf_nat);
4953 
4954 	switch (rval)
4955 	{
4956 	case -1 :
4957 		if (passp != NULL) {
4958 			DT1(frb_natv4out, fr_info_t *, fin);
4959 			NBUMPSIDED(1, ns_drop);
4960 			*passp = FR_BLOCK;
4961 			fin->fin_reason = FRB_NATV4;
4962 		}
4963 		fin->fin_flx |= FI_BADNAT;
4964 		NBUMPSIDED(1, ns_badnat);
4965 		break;
4966 	case 0 :
4967 		NBUMPSIDE(1, ns_ignored);
4968 		break;
4969 	case 1 :
4970 		NBUMPSIDE(1, ns_translated);
4971 		break;
4972 	}
4973 	fin->fin_ifp = sifp;
4974 	return rval;
4975 }
4976 
4977 /* ------------------------------------------------------------------------ */
4978 /* Function:    ipf_nat_out                                                 */
4979 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
4980 /*                     1 == packet was successfully translated.             */
4981 /* Parameters:  fin(I)    - pointer to packet information                   */
4982 /*              nat(I)    - pointer to NAT structure                        */
4983 /*              natadd(I) - flag indicating if it is safe to add frag cache */
4984 /*              nflags(I) - NAT flags set for this packet                   */
4985 /*                                                                          */
4986 /* Translate a packet coming "out" on an interface.                         */
4987 /* ------------------------------------------------------------------------ */
4988 int
4989 ipf_nat_out(fr_info_t *fin, nat_t *nat, int natadd, u_32_t nflags)
4990 {
4991 	ipf_main_softc_t *softc = fin->fin_main_soft;
4992 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4993 	icmphdr_t *icmp;
4994 	tcphdr_t *tcp;
4995 	ipnat_t *np;
4996 	int skip;
4997 	int i;
4998 
4999 	tcp = NULL;
5000 	icmp = NULL;
5001 	np = nat->nat_ptr;
5002 
5003 	if ((natadd != 0) && (fin->fin_flx & FI_FRAG) && (np != NULL))
5004 		(void) ipf_frag_natnew(softc, fin, 0, nat);
5005 
5006 	/*
5007 	 * Fix up checksums, not by recalculating them, but
5008 	 * simply computing adjustments.
5009 	 * This is only done for STREAMS based IP implementations where the
5010 	 * checksum has already been calculated by IP.  In all other cases,
5011 	 * IPFilter is called before the checksum needs calculating so there
5012 	 * is no call to modify whatever is in the header now.
5013 	 */
5014 	if (nflags == IPN_ICMPERR) {
5015 		u_32_t s1, s2, sumd, msumd;
5016 
5017 		s1 = LONG_SUM(ntohl(fin->fin_saddr));
5018 		if (nat->nat_dir == NAT_OUTBOUND) {
5019 			s2 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
5020 		} else {
5021 			s2 = LONG_SUM(ntohl(nat->nat_odstaddr));
5022 		}
5023 		CALC_SUMD(s1, s2, sumd);
5024 		msumd = sumd;
5025 
5026 		s1 = LONG_SUM(ntohl(fin->fin_daddr));
5027 		if (nat->nat_dir == NAT_OUTBOUND) {
5028 			s2 = LONG_SUM(ntohl(nat->nat_ndstaddr));
5029 		} else {
5030 			s2 = LONG_SUM(ntohl(nat->nat_osrcaddr));
5031 		}
5032 		CALC_SUMD(s1, s2, sumd);
5033 		msumd += sumd;
5034 
5035 		ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, msumd, 0);
5036 	}
5037 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5038     defined(linux)
5039 	else {
5040 		/*
5041 		 * Strictly speaking, this isn't necessary on BSD
5042 		 * kernels because they do checksum calculation after
5043 		 * this code has run BUT if ipfilter is being used
5044 		 * to do NAT as a bridge, that code doesn't exist.
5045 		 */
5046 		switch (nat->nat_dir)
5047 		{
5048 		case NAT_OUTBOUND :
5049 			ipf_fix_outcksum(fin->fin_cksum & FI_CK_L4PART,
5050 					 &fin->fin_ip->ip_sum,
5051 					 nat->nat_ipsumd, 0);
5052 			break;
5053 
5054 		case NAT_INBOUND :
5055 			ipf_fix_incksum(fin->fin_cksum & FI_CK_L4PART,
5056 					&fin->fin_ip->ip_sum,
5057 					nat->nat_ipsumd, 0);
5058 			break;
5059 
5060 		default :
5061 			break;
5062 		}
5063 	}
5064 #endif
5065 
5066 	/*
5067 	 * Address assignment is after the checksum modification because
5068 	 * we are using the address in the packet for determining the
5069 	 * correct checksum offset (the ICMP error could be coming from
5070 	 * anyone...)
5071 	 */
5072 	switch (nat->nat_dir)
5073 	{
5074 	case NAT_OUTBOUND :
5075 		fin->fin_ip->ip_src = nat->nat_nsrcip;
5076 		fin->fin_saddr = nat->nat_nsrcaddr;
5077 		fin->fin_ip->ip_dst = nat->nat_ndstip;
5078 		fin->fin_daddr = nat->nat_ndstaddr;
5079 		break;
5080 
5081 	case NAT_INBOUND :
5082 		fin->fin_ip->ip_src = nat->nat_odstip;
5083 		fin->fin_saddr = nat->nat_ndstaddr;
5084 		fin->fin_ip->ip_dst = nat->nat_osrcip;
5085 		fin->fin_daddr = nat->nat_nsrcaddr;
5086 		break;
5087 
5088 	case NAT_DIVERTIN :
5089 	    {
5090 		mb_t *m;
5091 
5092 		skip = ipf_nat_decap(fin, nat);
5093 		if (skip <= 0) {
5094 			NBUMPSIDED(1, ns_decap_fail);
5095 			return -1;
5096 		}
5097 
5098 		m = fin->fin_m;
5099 
5100 #if defined(MENTAT) && defined(_KERNEL)
5101 		m->b_rptr += skip;
5102 #else
5103 		m->m_data += skip;
5104 		m->m_len -= skip;
5105 
5106 # ifdef M_PKTHDR
5107 		if (m->m_flags & M_PKTHDR)
5108 			m->m_pkthdr.len -= skip;
5109 # endif
5110 #endif
5111 
5112 		MUTEX_ENTER(&nat->nat_lock);
5113 		ipf_nat_update(fin, nat);
5114 		MUTEX_EXIT(&nat->nat_lock);
5115 		fin->fin_flx |= FI_NATED;
5116 		if (np != NULL && np->in_tag.ipt_num[0] != 0)
5117 			fin->fin_nattag = &np->in_tag;
5118 		return 1;
5119 		/* NOTREACHED */
5120 	    }
5121 
5122 	case NAT_DIVERTOUT :
5123 	    {
5124 		u_32_t s1, s2, sumd;
5125 		udphdr_t *uh;
5126 		ip_t *ip;
5127 		mb_t *m;
5128 
5129 		m = M_DUP(np->in_divmp);
5130 		if (m == NULL) {
5131 			NBUMPSIDED(1, ns_divert_dup);
5132 			return -1;
5133 		}
5134 
5135 		ip = MTOD(m, ip_t *);
5136 		ip->ip_id = htons(ipf_nextipid(fin));
5137 		s2 = ntohs(ip->ip_id);
5138 
5139 		s1 = ip->ip_len;
5140 		ip->ip_len = ntohs(ip->ip_len);
5141 		ip->ip_len += fin->fin_plen;
5142 		ip->ip_len = htons(ip->ip_len);
5143 		s2 += ntohs(ip->ip_len);
5144 		CALC_SUMD(s1, s2, sumd);
5145 
5146 		uh = (udphdr_t *)(ip + 1);
5147 		uh->uh_ulen += fin->fin_plen;
5148 		uh->uh_ulen = htons(uh->uh_ulen);
5149 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5150     defined(linux)
5151 		ipf_fix_outcksum(0, &ip->ip_sum, sumd, 0);
5152 #endif
5153 
5154 		PREP_MB_T(fin, m);
5155 
5156 		fin->fin_src = ip->ip_src;
5157 		fin->fin_dst = ip->ip_dst;
5158 		fin->fin_ip = ip;
5159 		fin->fin_plen += sizeof(ip_t) + 8;	/* UDP + IPv4 hdr */
5160 		fin->fin_dlen += sizeof(ip_t) + 8;	/* UDP + IPv4 hdr */
5161 
5162 		nflags &= ~IPN_TCPUDPICMP;
5163 
5164 		break;
5165 	    }
5166 
5167 	default :
5168 		break;
5169 	}
5170 
5171 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5172 		u_short *csump;
5173 
5174 		if ((nat->nat_nsport != 0) && (nflags & IPN_TCPUDP)) {
5175 			tcp = fin->fin_dp;
5176 
5177 			switch (nat->nat_dir)
5178 			{
5179 			case NAT_OUTBOUND :
5180 				tcp->th_sport = nat->nat_nsport;
5181 				fin->fin_data[0] = ntohs(nat->nat_nsport);
5182 				tcp->th_dport = nat->nat_ndport;
5183 				fin->fin_data[1] = ntohs(nat->nat_ndport);
5184 				break;
5185 
5186 			case NAT_INBOUND :
5187 				tcp->th_sport = nat->nat_odport;
5188 				fin->fin_data[0] = ntohs(nat->nat_odport);
5189 				tcp->th_dport = nat->nat_osport;
5190 				fin->fin_data[1] = ntohs(nat->nat_osport);
5191 				break;
5192 			}
5193 		}
5194 
5195 		if ((nat->nat_oicmpid != 0) && (nflags & IPN_ICMPQUERY)) {
5196 			icmp = fin->fin_dp;
5197 
5198 			switch (nat->nat_dir)
5199 			{
5200 			case NAT_OUTBOUND :
5201 				icmp->icmp_id = nat->nat_nicmpid;
5202 				break;
5203 			case NAT_INBOUND :
5204 				icmp->icmp_id = nat->nat_oicmpid;
5205 				break;
5206 			}
5207 		}
5208 
5209 		csump = ipf_nat_proto(fin, nat, nflags);
5210 
5211 		/*
5212 		 * The above comments do not hold for layer 4 (or higher)
5213 		 * checksums...
5214 		 */
5215 		if (csump != NULL) {
5216 			if (nat->nat_dir == NAT_OUTBOUND)
5217 				ipf_fix_outcksum(fin->fin_cksum, csump,
5218 						 nat->nat_sumd[0],
5219 						 nat->nat_sumd[1] +
5220 						 fin->fin_dlen);
5221 			else
5222 				ipf_fix_incksum(fin->fin_cksum, csump,
5223 						nat->nat_sumd[0],
5224 						nat->nat_sumd[1] +
5225 						fin->fin_dlen);
5226 		}
5227 	}
5228 
5229 	ipf_sync_update(softc, SMC_NAT, fin, nat->nat_sync);
5230 	/* ------------------------------------------------------------- */
5231 	/* A few quick notes:                                            */
5232 	/*      Following are test conditions prior to calling the       */
5233 	/*      ipf_proxy_check routine.                                 */
5234 	/*                                                               */
5235 	/*      A NULL tcp indicates a non TCP/UDP packet.  When dealing */
5236 	/*      with a redirect rule, we attempt to match the packet's   */
5237 	/*      source port against in_dport, otherwise we'd compare the */
5238 	/*      packet's destination.                                    */
5239 	/* ------------------------------------------------------------- */
5240 	if ((np != NULL) && (np->in_apr != NULL)) {
5241 		i = ipf_proxy_check(fin, nat);
5242 		if (i == 0) {
5243 			i = 1;
5244 		} else if (i == -1) {
5245 			NBUMPSIDED(1, ns_ipf_proxy_fail);
5246 		}
5247 	} else {
5248 		i = 1;
5249 	}
5250 	fin->fin_flx |= FI_NATED;
5251 	return i;
5252 }
5253 
5254 
5255 /* ------------------------------------------------------------------------ */
5256 /* Function:    ipf_nat_checkin                                             */
5257 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
5258 /*                     0 == no packet translation occurred,                 */
5259 /*                     1 == packet was successfully translated.             */
5260 /* Parameters:  fin(I)   - pointer to packet information                    */
5261 /*              passp(I) - pointer to filtering result flags                */
5262 /*                                                                          */
5263 /* Check to see if an incoming packet should be changed.  ICMP packets are  */
5264 /* first checked to see if they match an existing entry (if an error),      */
5265 /* otherwise a search of the current NAT table is made.  If neither results */
5266 /* in a match then a search for a matching NAT rule is made.  Create a new  */
5267 /* NAT entry if a we matched a NAT rule.  Lastly, actually change the       */
5268 /* packet header(s) as required.                                            */
5269 /* ------------------------------------------------------------------------ */
5270 int
5271 ipf_nat_checkin(fr_info_t *fin, u_32_t *passp)
5272 {
5273 	ipf_main_softc_t *softc;
5274 	ipf_nat_softc_t *softn;
5275 	u_int nflags, natadd;
5276 	ipnat_t *np, *npnext;
5277 	int rval, natfailed;
5278 	struct ifnet *ifp;
5279 	struct in_addr in;
5280 	icmphdr_t *icmp;
5281 	tcphdr_t *tcp;
5282 	u_short dport;
5283 	nat_t *nat;
5284 	u_32_t iph;
5285 
5286 	softc = fin->fin_main_soft;
5287 	softn = softc->ipf_nat_soft;
5288 
5289 	if (softn->ipf_nat_lock != 0)
5290 		return 0;
5291 	if (softn->ipf_nat_stats.ns_rules == 0 &&
5292 	    softn->ipf_nat_instances == NULL)
5293 		return 0;
5294 
5295 	tcp = NULL;
5296 	icmp = NULL;
5297 	dport = 0;
5298 	natadd = 1;
5299 	nflags = 0;
5300 	natfailed = 0;
5301 	ifp = fin->fin_ifp;
5302 
5303 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5304 		switch (fin->fin_p)
5305 		{
5306 		case IPPROTO_TCP :
5307 			nflags = IPN_TCP;
5308 			break;
5309 		case IPPROTO_UDP :
5310 			nflags = IPN_UDP;
5311 			break;
5312 		case IPPROTO_ICMP :
5313 			icmp = fin->fin_dp;
5314 
5315 			/*
5316 			 * This is an incoming packet, so the destination is
5317 			 * the icmp_id and the source port equals 0
5318 			 */
5319 			if ((fin->fin_flx & FI_ICMPQUERY) != 0) {
5320 				nflags = IPN_ICMPQUERY;
5321 				dport = icmp->icmp_id;
5322 			} break;
5323 		default :
5324 			break;
5325 		}
5326 
5327 		if ((nflags & IPN_TCPUDP)) {
5328 			tcp = fin->fin_dp;
5329 			dport = fin->fin_data[1];
5330 		}
5331 	}
5332 
5333 	in = fin->fin_dst;
5334 
5335 	READ_ENTER(&softc->ipf_nat);
5336 
5337 	if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
5338 	    (nat = ipf_nat_icmperror(fin, &nflags, NAT_INBOUND)))
5339 		/*EMPTY*/;
5340 	else if ((fin->fin_flx & FI_FRAG) && (nat = ipf_frag_natknown(fin)))
5341 		natadd = 0;
5342 	else if ((nat = ipf_nat_inlookup(fin, nflags|NAT_SEARCH,
5343 					 (u_int)fin->fin_p,
5344 					 fin->fin_src, in))) {
5345 		nflags = nat->nat_flags;
5346 	} else if (fin->fin_off == 0) {
5347 		u_32_t hv, msk, rmsk = 0;
5348 
5349 		/*
5350 		 * If there is no current entry in the nat table for this IP#,
5351 		 * create one for it (if there is a matching rule).
5352 		 */
5353 maskloop:
5354 		msk = softn->ipf_nat_rdr_active_masks[rmsk];
5355 		iph = in.s_addr & msk;
5356 		hv = NAT_HASH_FN(iph, 0, softn->ipf_nat_rdrrules_sz);
5357 retry_roundrobin:
5358 		/* TRACE (iph,msk,rmsk,hv,softn->ipf_nat_rdrrules_sz) */
5359 		for (np = softn->ipf_nat_rdr_rules[hv]; np; np = npnext) {
5360 			npnext = np->in_rnext;
5361 			if (np->in_ifps[0] && (np->in_ifps[0] != ifp))
5362 				continue;
5363 			if (np->in_v[0] != 4)
5364 				continue;
5365 			if (np->in_pr[0] && (np->in_pr[0] != fin->fin_p))
5366 				continue;
5367 			if ((np->in_flags & IPN_RF) && !(np->in_flags & nflags))
5368 				continue;
5369 			if (np->in_flags & IPN_FILTER) {
5370 				switch (ipf_nat_match(fin, np))
5371 				{
5372 				case 0 :
5373 					continue;
5374 				case -1 :
5375 					rval = -1;
5376 					goto inmatchfail;
5377 				case 1 :
5378 				default :
5379 					break;
5380 				}
5381 			} else {
5382 				if ((in.s_addr & np->in_odstmsk) !=
5383 				    np->in_odstaddr)
5384 					continue;
5385 				if (np->in_odport &&
5386 				    ((np->in_dtop < dport) ||
5387 				     (dport < np->in_odport)))
5388 					continue;
5389 			}
5390 
5391 			if (np->in_plabel != -1) {
5392 				if (!ipf_proxy_ok(fin, tcp, np)) {
5393 					continue;
5394 				}
5395 			}
5396 
5397 			if (np->in_flags & IPN_NO) {
5398 				np->in_hits++;
5399 				break;
5400 			}
5401 
5402 			MUTEX_ENTER(&softn->ipf_nat_new);
5403 			/*
5404 			 * If we've matched a round-robin rule but it has
5405 			 * moved in the list since we got it, start over as
5406 			 * this is now no longer correct.
5407 			 */
5408 			if (npnext != np->in_rnext) {
5409 				if ((np->in_flags & IPN_ROUNDR) != 0) {
5410 					MUTEX_EXIT(&softn->ipf_nat_new);
5411 					goto retry_roundrobin;
5412 				}
5413 				npnext = np->in_rnext;
5414 			}
5415 
5416 			nat = ipf_nat_add(fin, np, NULL, nflags, NAT_INBOUND);
5417 			MUTEX_EXIT(&softn->ipf_nat_new);
5418 			if (nat != NULL) {
5419 				natfailed = 0;
5420 				break;
5421 			}
5422 			natfailed = -1;
5423 		}
5424 		if ((np == NULL) && (rmsk < softn->ipf_nat_rdr_max)) {
5425 			rmsk++;
5426 			goto maskloop;
5427 		}
5428 	}
5429 
5430 	if (nat != NULL) {
5431 		rval = ipf_nat_in(fin, nat, natadd, nflags);
5432 		if (rval == 1) {
5433 			MUTEX_ENTER(&nat->nat_lock);
5434 			ipf_nat_update(fin, nat);
5435 			nat->nat_bytes[0] += fin->fin_plen;
5436 			nat->nat_pkts[0]++;
5437 			fin->fin_pktnum = nat->nat_pkts[0];
5438 			MUTEX_EXIT(&nat->nat_lock);
5439 		}
5440 	} else
5441 		rval = natfailed;
5442 inmatchfail:
5443 	RWLOCK_EXIT(&softc->ipf_nat);
5444 
5445 	switch (rval)
5446 	{
5447 	case -1 :
5448 		if (passp != NULL) {
5449 			DT1(frb_natv4in, fr_info_t *, fin);
5450 			NBUMPSIDED(0, ns_drop);
5451 			*passp = FR_BLOCK;
5452 			fin->fin_reason = FRB_NATV4;
5453 		}
5454 		fin->fin_flx |= FI_BADNAT;
5455 		NBUMPSIDED(0, ns_badnat);
5456 		break;
5457 	case 0 :
5458 		NBUMPSIDE(0, ns_ignored);
5459 		break;
5460 	case 1 :
5461 		NBUMPSIDE(0, ns_translated);
5462 		break;
5463 	}
5464 	return rval;
5465 }
5466 
5467 
5468 /* ------------------------------------------------------------------------ */
5469 /* Function:    ipf_nat_in                                                  */
5470 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
5471 /*                     1 == packet was successfully translated.             */
5472 /* Parameters:  fin(I)    - pointer to packet information                   */
5473 /*              nat(I)    - pointer to NAT structure                        */
5474 /*              natadd(I) - flag indicating if it is safe to add frag cache */
5475 /*              nflags(I) - NAT flags set for this packet                   */
5476 /* Locks Held:  ipf_nat(READ)                                               */
5477 /*                                                                          */
5478 /* Translate a packet coming "in" on an interface.                          */
5479 /* ------------------------------------------------------------------------ */
5480 int
5481 ipf_nat_in(fr_info_t *fin, nat_t *nat, int natadd, u_32_t nflags)
5482 {
5483 	ipf_main_softc_t *softc = fin->fin_main_soft;
5484 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5485 	u_32_t sumd, ipsumd, sum1, sum2;
5486 	icmphdr_t *icmp;
5487 	tcphdr_t *tcp;
5488 	ipnat_t *np;
5489 	int skip;
5490 	int i;
5491 
5492 	tcp = NULL;
5493 	np = nat->nat_ptr;
5494 	fin->fin_fr = nat->nat_fr;
5495 
5496 	if (np != NULL) {
5497 		if ((natadd != 0) && (fin->fin_flx & FI_FRAG))
5498 			(void) ipf_frag_natnew(softc, fin, 0, nat);
5499 
5500 	/* ------------------------------------------------------------- */
5501 	/* A few quick notes:                                            */
5502 	/*      Following are test conditions prior to calling the       */
5503 	/*      ipf_proxy_check routine.                                 */
5504 	/*                                                               */
5505 	/*      A NULL tcp indicates a non TCP/UDP packet.  When dealing */
5506 	/*      with a map rule, we attempt to match the packet's        */
5507 	/*      source port against in_dport, otherwise we'd compare the */
5508 	/*      packet's destination.                                    */
5509 	/* ------------------------------------------------------------- */
5510 		if (np->in_apr != NULL) {
5511 			i = ipf_proxy_check(fin, nat);
5512 			if (i == -1) {
5513 				NBUMPSIDED(0, ns_ipf_proxy_fail);
5514 				return -1;
5515 			}
5516 		}
5517 	}
5518 
5519 	ipf_sync_update(softc, SMC_NAT, fin, nat->nat_sync);
5520 
5521 	ipsumd = nat->nat_ipsumd;
5522 	/*
5523 	 * Fix up checksums, not by recalculating them, but
5524 	 * simply computing adjustments.
5525 	 * Why only do this for some platforms on inbound packets ?
5526 	 * Because for those that it is done, IP processing is yet to happen
5527 	 * and so the IPv4 header checksum has not yet been evaluated.
5528 	 * Perhaps it should always be done for the benefit of things like
5529 	 * fast forwarding (so that it doesn't need to be recomputed) but with
5530 	 * header checksum offloading, perhaps it is a moot point.
5531 	 */
5532 
5533 	switch (nat->nat_dir)
5534 	{
5535 	case NAT_INBOUND :
5536 		if ((fin->fin_flx & FI_ICMPERR) == 0) {
5537 			fin->fin_ip->ip_src = nat->nat_nsrcip;
5538 			fin->fin_saddr = nat->nat_nsrcaddr;
5539 		} else {
5540 			sum1 = nat->nat_osrcaddr;
5541 			sum2 = nat->nat_nsrcaddr;
5542 			CALC_SUMD(sum1, sum2, sumd);
5543 			ipsumd -= sumd;
5544 		}
5545 		fin->fin_ip->ip_dst = nat->nat_ndstip;
5546 		fin->fin_daddr = nat->nat_ndstaddr;
5547 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5548      defined(__osf__) || defined(linux)
5549 		ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, ipsumd, 0);
5550 #endif
5551 		break;
5552 
5553 	case NAT_OUTBOUND :
5554 		if ((fin->fin_flx & FI_ICMPERR) == 0) {
5555 			fin->fin_ip->ip_src = nat->nat_odstip;
5556 			fin->fin_saddr = nat->nat_odstaddr;
5557 		} else {
5558 			sum1 = nat->nat_odstaddr;
5559 			sum2 = nat->nat_ndstaddr;
5560 			CALC_SUMD(sum1, sum2, sumd);
5561 			ipsumd -= sumd;
5562 		}
5563 		fin->fin_ip->ip_dst = nat->nat_osrcip;
5564 		fin->fin_daddr = nat->nat_osrcaddr;
5565 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5566      defined(__osf__) || defined(linux)
5567 		ipf_fix_incksum(0, &fin->fin_ip->ip_sum, ipsumd, 0);
5568 #endif
5569 		break;
5570 
5571 	case NAT_DIVERTIN :
5572 	    {
5573 		udphdr_t *uh;
5574 		ip_t *ip;
5575 		mb_t *m;
5576 
5577 		m = M_DUP(np->in_divmp);
5578 		if (m == NULL) {
5579 			NBUMPSIDED(0, ns_divert_dup);
5580 			return -1;
5581 		}
5582 
5583 		ip = MTOD(m, ip_t *);
5584 		ip->ip_id = htons(ipf_nextipid(fin));
5585 		sum1 = ntohs(ip->ip_len);
5586 		ip->ip_len = ntohs(ip->ip_len);
5587 		ip->ip_len += fin->fin_plen;
5588 		ip->ip_len = htons(ip->ip_len);
5589 
5590 		uh = (udphdr_t *)(ip + 1);
5591 		uh->uh_ulen += fin->fin_plen;
5592 		uh->uh_ulen = htons(uh->uh_ulen);
5593 
5594 		sum2 = ntohs(ip->ip_id) + ntohs(ip->ip_len);
5595 		sum2 += ntohs(ip->ip_off) & IP_DF;
5596 		CALC_SUMD(sum1, sum2, sumd);
5597 
5598 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
5599      defined(__osf__) || defined(linux)
5600 		ipf_fix_outcksum(0, &ip->ip_sum, sumd, 0);
5601 #endif
5602 		PREP_MB_T(fin, m);
5603 
5604 		fin->fin_ip = ip;
5605 		fin->fin_plen += sizeof(ip_t) + 8;	/* UDP + new IPv4 hdr */
5606 		fin->fin_dlen += sizeof(ip_t) + 8;	/* UDP + old IPv4 hdr */
5607 
5608 		nflags &= ~IPN_TCPUDPICMP;
5609 
5610 		break;
5611 	    }
5612 
5613 	case NAT_DIVERTOUT :
5614 	    {
5615 		mb_t *m;
5616 
5617 		skip = ipf_nat_decap(fin, nat);
5618 		if (skip <= 0) {
5619 			NBUMPSIDED(0, ns_decap_fail);
5620 			return -1;
5621 		}
5622 
5623 		m = fin->fin_m;
5624 
5625 #if defined(MENTAT) && defined(_KERNEL)
5626 		m->b_rptr += skip;
5627 #else
5628 		m->m_data += skip;
5629 		m->m_len -= skip;
5630 
5631 # ifdef M_PKTHDR
5632 		if (m->m_flags & M_PKTHDR)
5633 			m->m_pkthdr.len -= skip;
5634 # endif
5635 #endif
5636 
5637 		ipf_nat_update(fin, nat);
5638 		nflags &= ~IPN_TCPUDPICMP;
5639 		fin->fin_flx |= FI_NATED;
5640 		if (np != NULL && np->in_tag.ipt_num[0] != 0)
5641 			fin->fin_nattag = &np->in_tag;
5642 		return 1;
5643 		/* NOTREACHED */
5644 	    }
5645 	}
5646 	if (nflags & IPN_TCPUDP)
5647 		tcp = fin->fin_dp;
5648 
5649 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5650 		u_short *csump;
5651 
5652 		if ((nat->nat_odport != 0) && (nflags & IPN_TCPUDP)) {
5653 			switch (nat->nat_dir)
5654 			{
5655 			case NAT_INBOUND :
5656 				tcp->th_sport = nat->nat_nsport;
5657 				fin->fin_data[0] = ntohs(nat->nat_nsport);
5658 				tcp->th_dport = nat->nat_ndport;
5659 				fin->fin_data[1] = ntohs(nat->nat_ndport);
5660 				break;
5661 
5662 			case NAT_OUTBOUND :
5663 				tcp->th_sport = nat->nat_odport;
5664 				fin->fin_data[0] = ntohs(nat->nat_odport);
5665 				tcp->th_dport = nat->nat_osport;
5666 				fin->fin_data[1] = ntohs(nat->nat_osport);
5667 				break;
5668 			}
5669 		}
5670 
5671 
5672 		if ((nat->nat_oicmpid != 0) && (nflags & IPN_ICMPQUERY)) {
5673 			icmp = fin->fin_dp;
5674 
5675 			switch (nat->nat_dir)
5676 			{
5677 			case NAT_INBOUND :
5678 				icmp->icmp_id = nat->nat_nicmpid;
5679 				break;
5680 			case NAT_OUTBOUND :
5681 				icmp->icmp_id = nat->nat_oicmpid;
5682 				break;
5683 			}
5684 		}
5685 
5686 		csump = ipf_nat_proto(fin, nat, nflags);
5687 
5688 		/*
5689 		 * The above comments do not hold for layer 4 (or higher)
5690 		 * checksums...
5691 		 */
5692 		if (csump != NULL) {
5693 			if (nat->nat_dir == NAT_OUTBOUND)
5694 				ipf_fix_incksum(0, csump, nat->nat_sumd[0], 0);
5695 			else
5696 				ipf_fix_outcksum(0, csump, nat->nat_sumd[0], 0);
5697 		}
5698 	}
5699 
5700 	fin->fin_flx |= FI_NATED;
5701 	if (np != NULL && np->in_tag.ipt_num[0] != 0)
5702 		fin->fin_nattag = &np->in_tag;
5703 	return 1;
5704 }
5705 
5706 
5707 /* ------------------------------------------------------------------------ */
5708 /* Function:    ipf_nat_proto                                               */
5709 /* Returns:     u_short* - pointer to transport header checksum to update,  */
5710 /*                         NULL if the transport protocol is not recognised */
5711 /*                         as needing a checksum update.                    */
5712 /* Parameters:  fin(I)    - pointer to packet information                   */
5713 /*              nat(I)    - pointer to NAT structure                        */
5714 /*              nflags(I) - NAT flags set for this packet                   */
5715 /*                                                                          */
5716 /* Return the pointer to the checksum field for each protocol so understood.*/
5717 /* If support for making other changes to a protocol header is required,    */
5718 /* that is not strictly 'address' translation, such as clamping the MSS in  */
5719 /* TCP down to a specific value, then do it from here.                      */
5720 /* ------------------------------------------------------------------------ */
5721 u_short *
5722 ipf_nat_proto(fr_info_t *fin, nat_t *nat, u_int nflags)
5723 {
5724 	icmphdr_t *icmp;
5725 	u_short *csump;
5726 	tcphdr_t *tcp;
5727 	udphdr_t *udp;
5728 
5729 	csump = NULL;
5730 	if (fin->fin_out == 0) {
5731 		fin->fin_rev = (nat->nat_dir & NAT_OUTBOUND);
5732 	} else {
5733 		fin->fin_rev = ((nat->nat_dir & NAT_OUTBOUND) == 0);
5734 	}
5735 
5736 	switch (fin->fin_p)
5737 	{
5738 	case IPPROTO_TCP :
5739 		tcp = fin->fin_dp;
5740 
5741 		if ((nflags & IPN_TCP) != 0)
5742 			csump = &tcp->th_sum;
5743 
5744 		/*
5745 		 * Do a MSS CLAMPING on a SYN packet,
5746 		 * only deal IPv4 for now.
5747 		 */
5748 		if ((nat->nat_mssclamp != 0) && (tcp->th_flags & TH_SYN) != 0)
5749 			ipf_nat_mssclamp(tcp, nat->nat_mssclamp, fin, csump);
5750 
5751 		break;
5752 
5753 	case IPPROTO_UDP :
5754 		udp = fin->fin_dp;
5755 
5756 		if ((nflags & IPN_UDP) != 0) {
5757 			if (udp->uh_sum != 0)
5758 				csump = &udp->uh_sum;
5759 		}
5760 		break;
5761 
5762 	case IPPROTO_ICMP :
5763 		icmp = fin->fin_dp;
5764 
5765 		if ((nflags & IPN_ICMPQUERY) != 0) {
5766 			if (icmp->icmp_cksum != 0)
5767 				csump = &icmp->icmp_cksum;
5768 		}
5769 		break;
5770 
5771 #ifdef USE_INET6
5772 	case IPPROTO_ICMPV6 :
5773 	    {
5774 		struct icmp6_hdr *icmp6 = (struct icmp6_hdr *)fin->fin_dp;
5775 
5776 		icmp6 = fin->fin_dp;
5777 
5778 		if ((nflags & IPN_ICMPQUERY) != 0) {
5779 			if (icmp6->icmp6_cksum != 0)
5780 				csump = &icmp6->icmp6_cksum;
5781 		}
5782 		break;
5783 	    }
5784 #endif
5785 	}
5786 	return csump;
5787 }
5788 
5789 
5790 /* ------------------------------------------------------------------------ */
5791 /* Function:    ipf_nat_expire                                              */
5792 /* Returns:     Nil                                                         */
5793 /* Parameters:  softc(I) - pointer to soft context main structure           */
5794 /*                                                                          */
5795 /* Check all of the timeout queues for entries at the top which need to be  */
5796 /* expired.                                                                 */
5797 /* ------------------------------------------------------------------------ */
5798 void
5799 ipf_nat_expire(ipf_main_softc_t *softc)
5800 {
5801 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5802 	ipftq_t *ifq, *ifqnext;
5803 	ipftqent_t *tqe, *tqn;
5804 	int i;
5805 	SPL_INT(s);
5806 
5807 	SPL_NET(s);
5808 	WRITE_ENTER(&softc->ipf_nat);
5809 	for (ifq = softn->ipf_nat_tcptq, i = 0; ifq != NULL;
5810 	     ifq = ifq->ifq_next) {
5811 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
5812 			if (tqe->tqe_die > softc->ipf_ticks)
5813 				break;
5814 			tqn = tqe->tqe_next;
5815 			ipf_nat_delete(softc, tqe->tqe_parent, NL_EXPIRE);
5816 		}
5817 	}
5818 
5819 	for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifq->ifq_next) {
5820 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
5821 			if (tqe->tqe_die > softc->ipf_ticks)
5822 				break;
5823 			tqn = tqe->tqe_next;
5824 			ipf_nat_delete(softc, tqe->tqe_parent, NL_EXPIRE);
5825 		}
5826 	}
5827 
5828 	for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifqnext) {
5829 		ifqnext = ifq->ifq_next;
5830 
5831 		if (((ifq->ifq_flags & IFQF_DELETE) != 0) &&
5832 		    (ifq->ifq_ref == 0)) {
5833 			ipf_freetimeoutqueue(softc, ifq);
5834 		}
5835 	}
5836 
5837 	if (softn->ipf_nat_doflush != 0) {
5838 		ipf_nat_extraflush(softc, softn, 2);
5839 		softn->ipf_nat_doflush = 0;
5840 	}
5841 
5842 	RWLOCK_EXIT(&softc->ipf_nat);
5843 	SPL_X(s);
5844 }
5845 
5846 
5847 /* ------------------------------------------------------------------------ */
5848 /* Function:    ipf_nat_sync                                                */
5849 /* Returns:     Nil                                                         */
5850 /* Parameters:  softc(I) - pointer to soft context main structure           */
5851 /*              ifp(I) - pointer to network interface                       */
5852 /*                                                                          */
5853 /* Walk through all of the currently active NAT sessions, looking for those */
5854 /* which need to have their translated address updated.                     */
5855 /* ------------------------------------------------------------------------ */
5856 void
5857 ipf_nat_sync(ipf_main_softc_t *softc, void *ifp)
5858 {
5859 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5860 	u_32_t sum1, sum2, sumd;
5861 	i6addr_t in;
5862 	ipnat_t *n;
5863 	nat_t *nat;
5864 	void *ifp2;
5865 	int idx;
5866 	SPL_INT(s);
5867 
5868 	if (softc->ipf_running <= 0)
5869 		return;
5870 
5871 	/*
5872 	 * Change IP addresses for NAT sessions for any protocol except TCP
5873 	 * since it will break the TCP connection anyway.  The only rules
5874 	 * which will get changed are those which are "map ... -> 0/32",
5875 	 * where the rule specifies the address is taken from the interface.
5876 	 */
5877 	SPL_NET(s);
5878 	WRITE_ENTER(&softc->ipf_nat);
5879 
5880 	if (softc->ipf_running <= 0) {
5881 		RWLOCK_EXIT(&softc->ipf_nat);
5882 		return;
5883 	}
5884 
5885 	for (nat = softn->ipf_nat_instances; nat; nat = nat->nat_next) {
5886 		if ((nat->nat_flags & IPN_TCP) != 0)
5887 			continue;
5888 
5889 		n = nat->nat_ptr;
5890 		if (n != NULL) {
5891 			if (n->in_v[1] == 4) {
5892 				if (n->in_redir & NAT_MAP) {
5893 					if ((n->in_nsrcaddr != 0) ||
5894 					    (n->in_nsrcmsk != 0xffffffff))
5895 						continue;
5896 				} else if (n->in_redir & NAT_REDIRECT) {
5897 					if ((n->in_ndstaddr != 0) ||
5898 					    (n->in_ndstmsk != 0xffffffff))
5899 						continue;
5900 				}
5901 			}
5902 #ifdef USE_INET6
5903 			if (n->in_v[1] == 4) {
5904 				if (n->in_redir & NAT_MAP) {
5905 					if (!IP6_ISZERO(&n->in_nsrcaddr) ||
5906 					    !IP6_ISONES(&n->in_nsrcmsk))
5907 						continue;
5908 				} else if (n->in_redir & NAT_REDIRECT) {
5909 					if (!IP6_ISZERO(&n->in_ndstaddr) ||
5910 					    !IP6_ISONES(&n->in_ndstmsk))
5911 						continue;
5912 				}
5913 			}
5914 #endif
5915 		}
5916 
5917 		if (((ifp == NULL) || (ifp == nat->nat_ifps[0]) ||
5918 		     (ifp == nat->nat_ifps[1]))) {
5919 			nat->nat_ifps[0] = GETIFP(nat->nat_ifnames[0],
5920 						  nat->nat_v[0]);
5921 			if ((nat->nat_ifps[0] != NULL) &&
5922 			    (nat->nat_ifps[0] != (void *)-1)) {
5923 				nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
5924 			}
5925 			if (nat->nat_ifnames[1][0] != '\0') {
5926 				nat->nat_ifps[1] = GETIFP(nat->nat_ifnames[1],
5927 							  nat->nat_v[1]);
5928 			} else {
5929 				nat->nat_ifps[1] = nat->nat_ifps[0];
5930 			}
5931 			if ((nat->nat_ifps[1] != NULL) &&
5932 			    (nat->nat_ifps[1] != (void *)-1)) {
5933 				nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
5934 			}
5935 			ifp2 = nat->nat_ifps[0];
5936 			if (ifp2 == NULL)
5937 				continue;
5938 
5939 			/*
5940 			 * Change the map-to address to be the same as the
5941 			 * new one.
5942 			 */
5943 			sum1 = NATFSUM(nat, nat->nat_v[1], nat_nsrc6);
5944 			if (ipf_ifpaddr(softc, nat->nat_v[0], FRI_NORMAL, ifp2,
5945 				       &in, NULL) != -1) {
5946 				if (nat->nat_v[0] == 4)
5947 					nat->nat_nsrcip = in.in4;
5948 			}
5949 			sum2 = NATFSUM(nat, nat->nat_v[1], nat_nsrc6);
5950 
5951 			if (sum1 == sum2)
5952 				continue;
5953 			/*
5954 			 * Readjust the checksum adjustment to take into
5955 			 * account the new IP#.
5956 			 */
5957 			CALC_SUMD(sum1, sum2, sumd);
5958 			/* XXX - dont change for TCP when solaris does
5959 			 * hardware checksumming.
5960 			 */
5961 			sumd += nat->nat_sumd[0];
5962 			nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
5963 			nat->nat_sumd[1] = nat->nat_sumd[0];
5964 		}
5965 	}
5966 
5967 	for (n = softn->ipf_nat_list; (n != NULL); n = n->in_next) {
5968 		char *base = n->in_names;
5969 
5970 		if ((ifp == NULL) || (n->in_ifps[0] == ifp))
5971 			n->in_ifps[0] = ipf_resolvenic(softc,
5972 						       base + n->in_ifnames[0],
5973 						       n->in_v[0]);
5974 		if ((ifp == NULL) || (n->in_ifps[1] == ifp))
5975 			n->in_ifps[1] = ipf_resolvenic(softc,
5976 						       base + n->in_ifnames[1],
5977 						       n->in_v[1]);
5978 
5979 		if (n->in_redir & NAT_REDIRECT)
5980 			idx = 1;
5981 		else
5982 			idx = 0;
5983 
5984 		if (((ifp == NULL) || (n->in_ifps[idx] == ifp)) &&
5985 		    (n->in_ifps[idx] != NULL &&
5986 		     n->in_ifps[idx] != (void *)-1)) {
5987 
5988 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_osrc,
5989 					     0, n->in_ifps[idx]);
5990 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_odst,
5991 					     0, n->in_ifps[idx]);
5992 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_nsrc,
5993 					     0, n->in_ifps[idx]);
5994 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_ndst,
5995 					     0, n->in_ifps[idx]);
5996 		}
5997 	}
5998 	RWLOCK_EXIT(&softc->ipf_nat);
5999 	SPL_X(s);
6000 }
6001 
6002 
6003 /* ------------------------------------------------------------------------ */
6004 /* Function:    ipf_nat_icmpquerytype                                       */
6005 /* Returns:     int - 1 == success, 0 == failure                            */
6006 /* Parameters:  icmptype(I) - ICMP type number                              */
6007 /*                                                                          */
6008 /* Tests to see if the ICMP type number passed is a query/response type or  */
6009 /* not.                                                                     */
6010 /* ------------------------------------------------------------------------ */
6011 static int
6012 ipf_nat_icmpquerytype(int icmptype)
6013 {
6014 
6015 	/*
6016 	 * For the ICMP query NAT code, it is essential that both the query
6017 	 * and the reply match on the NAT rule. Because the NAT structure
6018 	 * does not keep track of the icmptype, and a single NAT structure
6019 	 * is used for all icmp types with the same src, dest and id, we
6020 	 * simply define the replies as queries as well. The funny thing is,
6021 	 * although it seems silly to call a reply a query, this is exactly
6022 	 * as it is defined in the IPv4 specification
6023 	 */
6024 	switch (icmptype)
6025 	{
6026 	case ICMP_ECHOREPLY:
6027 	case ICMP_ECHO:
6028 	/* route advertisement/sollicitation is currently unsupported: */
6029 	/* it would require rewriting the ICMP data section            */
6030 	case ICMP_TSTAMP:
6031 	case ICMP_TSTAMPREPLY:
6032 	case ICMP_IREQ:
6033 	case ICMP_IREQREPLY:
6034 	case ICMP_MASKREQ:
6035 	case ICMP_MASKREPLY:
6036 		return 1;
6037 	default:
6038 		return 0;
6039 	}
6040 }
6041 
6042 
6043 /* ------------------------------------------------------------------------ */
6044 /* Function:    nat_log                                                     */
6045 /* Returns:     Nil                                                         */
6046 /* Parameters:  softc(I) - pointer to soft context main structure           */
6047 /*              softn(I) - pointer to NAT context structure                 */
6048 /*              nat(I)    - pointer to NAT structure                        */
6049 /*              action(I) - action related to NAT structure being performed */
6050 /*                                                                          */
6051 /* Creates a NAT log entry.                                                 */
6052 /* ------------------------------------------------------------------------ */
6053 void
6054 ipf_nat_log(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, struct nat *nat,
6055     u_int action)
6056 {
6057 #ifdef	IPFILTER_LOG
6058 # ifndef LARGE_NAT
6059 	struct ipnat *np;
6060 	int rulen;
6061 # endif
6062 	struct natlog natl;
6063 	void *items[1];
6064 	size_t sizes[1];
6065 	int types[1];
6066 
6067 	bcopy((char *)&nat->nat_osrc6, (char *)&natl.nl_osrcip,
6068 	      sizeof(natl.nl_osrcip));
6069 	bcopy((char *)&nat->nat_nsrc6, (char *)&natl.nl_nsrcip,
6070 	      sizeof(natl.nl_nsrcip));
6071 	bcopy((char *)&nat->nat_odst6, (char *)&natl.nl_odstip,
6072 	      sizeof(natl.nl_odstip));
6073 	bcopy((char *)&nat->nat_ndst6, (char *)&natl.nl_ndstip,
6074 	      sizeof(natl.nl_ndstip));
6075 
6076 	natl.nl_bytes[0] = nat->nat_bytes[0];
6077 	natl.nl_bytes[1] = nat->nat_bytes[1];
6078 	natl.nl_pkts[0] = nat->nat_pkts[0];
6079 	natl.nl_pkts[1] = nat->nat_pkts[1];
6080 	natl.nl_odstport = nat->nat_odport;
6081 	natl.nl_osrcport = nat->nat_osport;
6082 	natl.nl_nsrcport = nat->nat_nsport;
6083 	natl.nl_ndstport = nat->nat_ndport;
6084 	natl.nl_p[0] = nat->nat_pr[0];
6085 	natl.nl_p[1] = nat->nat_pr[1];
6086 	natl.nl_v[0] = nat->nat_v[0];
6087 	natl.nl_v[1] = nat->nat_v[1];
6088 	natl.nl_type = nat->nat_redir;
6089 	natl.nl_action = action;
6090 	natl.nl_rule = -1;
6091 
6092 	bcopy(nat->nat_ifnames[0], natl.nl_ifnames[0],
6093 	      sizeof(nat->nat_ifnames[0]));
6094 	bcopy(nat->nat_ifnames[1], natl.nl_ifnames[1],
6095 	      sizeof(nat->nat_ifnames[1]));
6096 
6097 # ifndef LARGE_NAT
6098 	if (nat->nat_ptr != NULL) {
6099 		for (rulen = 0, np = softn->ipf_nat_list; np != NULL;
6100 		     np = np->in_next, rulen++)
6101 			if (np == nat->nat_ptr) {
6102 				natl.nl_rule = rulen;
6103 				break;
6104 			}
6105 	}
6106 # endif
6107 	items[0] = &natl;
6108 	sizes[0] = sizeof(natl);
6109 	types[0] = 0;
6110 
6111 	(void) ipf_log_items(softc, IPL_LOGNAT, NULL, items, sizes, types, 1);
6112 #endif
6113 }
6114 
6115 
6116 #if defined(__OpenBSD__)
6117 /* ------------------------------------------------------------------------ */
6118 /* Function:    ipf_nat_ifdetach                                            */
6119 /* Returns:     Nil                                                         */
6120 /* Parameters:  ifp(I) - pointer to network interface                       */
6121 /*                                                                          */
6122 /* Compatibility interface for OpenBSD to trigger the correct updating of   */
6123 /* interface references within IPFilter.                                    */
6124 /* ------------------------------------------------------------------------ */
6125 void
6126 ipf_nat_ifdetach(ifp)
6127 	void *ifp;
6128 {
6129 	ipf_main_softc_t *softc;
6130 
6131 	softc = ipf_get_softc(0);
6132 
6133 	ipf_sync(ifp);
6134 	return;
6135 }
6136 #endif
6137 
6138 
6139 /* ------------------------------------------------------------------------ */
6140 /* Function:    ipf_nat_rule_deref                                          */
6141 /* Returns:     Nil                                                         */
6142 /* Parameters:  softc(I) - pointer to soft context main structure           */
6143 /*              inp(I)   - pointer to pointer to NAT rule                   */
6144 /* Write Locks: ipf_nat                                                     */
6145 /*                                                                          */
6146 /* Dropping the refernce count for a rule means that whatever held the      */
6147 /* pointer to this rule (*inp) is no longer interested in it and when the   */
6148 /* reference count drops to zero, any resources allocated for the rule can  */
6149 /* be released and the rule itself free'd.                                  */
6150 /* ------------------------------------------------------------------------ */
6151 void
6152 ipf_nat_rule_deref(ipf_main_softc_t *softc, ipnat_t **inp)
6153 {
6154 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6155 	ipnat_t *n;
6156 
6157 	n = *inp;
6158 	*inp = NULL;
6159 	n->in_use--;
6160 	if (n->in_use > 0)
6161 		return;
6162 
6163 	if (n->in_apr != NULL)
6164 		ipf_proxy_deref(n->in_apr);
6165 
6166 	ipf_nat_rule_fini(softc, n);
6167 
6168 	if (n->in_redir & NAT_REDIRECT) {
6169 		if ((n->in_flags & IPN_PROXYRULE) == 0) {
6170 			ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules_rdr);
6171 		}
6172 	}
6173 	if (n->in_redir & (NAT_MAP|NAT_MAPBLK)) {
6174 		if ((n->in_flags & IPN_PROXYRULE) == 0) {
6175 			ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules_map);
6176 		}
6177 	}
6178 
6179 	if (n->in_tqehead[0] != NULL) {
6180 		if (ipf_deletetimeoutqueue(n->in_tqehead[0]) == 0) {
6181 			ipf_freetimeoutqueue(softc, n->in_tqehead[0]);
6182 		}
6183 	}
6184 
6185 	if (n->in_tqehead[1] != NULL) {
6186 		if (ipf_deletetimeoutqueue(n->in_tqehead[1]) == 0) {
6187 			ipf_freetimeoutqueue(softc, n->in_tqehead[1]);
6188 		}
6189 	}
6190 
6191 	if ((n->in_flags & IPN_PROXYRULE) == 0) {
6192 		ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules);
6193 	}
6194 
6195 	MUTEX_DESTROY(&n->in_lock);
6196 
6197 	KFREES(n, n->in_size);
6198 
6199 #if SOLARIS && !defined(INSTANCES)
6200 	if (softn->ipf_nat_stats.ns_rules == 0)
6201 		pfil_delayed_copy = 1;
6202 #endif
6203 }
6204 
6205 
6206 /* ------------------------------------------------------------------------ */
6207 /* Function:    ipf_nat_deref                                               */
6208 /* Returns:     Nil                                                         */
6209 /* Parameters:  softc(I) - pointer to soft context main structure           */
6210 /*              natp(I)  - pointer to pointer to NAT table entry            */
6211 /*                                                                          */
6212 /* Decrement the reference counter for this NAT table entry and free it if  */
6213 /* there are no more things using it.                                       */
6214 /*                                                                          */
6215 /* IF nat_ref == 1 when this function is called, then we have an orphan nat */
6216 /* structure *because* it only gets called on paths _after_ nat_ref has been*/
6217 /* incremented.  If nat_ref == 1 then we shouldn't decrement it here        */
6218 /* because nat_delete() will do that and send nat_ref to -1.                */
6219 /*                                                                          */
6220 /* Holding the lock on nat_lock is required to serialise nat_delete() being */
6221 /* called from a NAT flush ioctl with a deref happening because of a packet.*/
6222 /* ------------------------------------------------------------------------ */
6223 void
6224 ipf_nat_deref(ipf_main_softc_t *softc, nat_t **natp)
6225 {
6226 	nat_t *nat;
6227 
6228 	nat = *natp;
6229 	*natp = NULL;
6230 
6231 	MUTEX_ENTER(&nat->nat_lock);
6232 	if (nat->nat_ref > 1) {
6233 		nat->nat_ref--;
6234 		ASSERT(nat->nat_ref >= 0);
6235 		MUTEX_EXIT(&nat->nat_lock);
6236 		return;
6237 	}
6238 	MUTEX_EXIT(&nat->nat_lock);
6239 
6240 	WRITE_ENTER(&softc->ipf_nat);
6241 	ipf_nat_delete(softc, nat, NL_EXPIRE);
6242 	RWLOCK_EXIT(&softc->ipf_nat);
6243 }
6244 
6245 
6246 /* ------------------------------------------------------------------------ */
6247 /* Function:    ipf_nat_clone                                               */
6248 /* Returns:     ipstate_t* - NULL == cloning failed,                        */
6249 /*                           else pointer to new state structure            */
6250 /* Parameters:  fin(I) - pointer to packet information                      */
6251 /*              is(I)  - pointer to master state structure                  */
6252 /* Write Lock:  ipf_nat                                                     */
6253 /*                                                                          */
6254 /* Create a "duplcate" state table entry from the master.                   */
6255 /* ------------------------------------------------------------------------ */
6256 nat_t *
6257 ipf_nat_clone(fr_info_t *fin, nat_t *nat)
6258 {
6259 	ipf_main_softc_t *softc = fin->fin_main_soft;
6260 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6261 	frentry_t *fr;
6262 	nat_t *clone;
6263 	ipnat_t *np;
6264 
6265 	KMALLOC(clone, nat_t *);
6266 	if (clone == NULL) {
6267 		NBUMPSIDED(fin->fin_out, ns_clone_nomem);
6268 		return NULL;
6269 	}
6270 	bcopy((char *)nat, (char *)clone, sizeof(*clone));
6271 
6272 	MUTEX_NUKE(&clone->nat_lock);
6273 
6274 	clone->nat_rev = fin->fin_rev;
6275 	clone->nat_aps = NULL;
6276 	/*
6277 	 * Initialize all these so that ipf_nat_delete() doesn't cause a crash.
6278 	 */
6279 	clone->nat_tqe.tqe_pnext = NULL;
6280 	clone->nat_tqe.tqe_next = NULL;
6281 	clone->nat_tqe.tqe_ifq = NULL;
6282 	clone->nat_tqe.tqe_parent = clone;
6283 
6284 	clone->nat_flags &= ~SI_CLONE;
6285 	clone->nat_flags |= SI_CLONED;
6286 
6287 	if (clone->nat_hm)
6288 		clone->nat_hm->hm_ref++;
6289 
6290 	if (ipf_nat_insert(softc, softn, clone) == -1) {
6291 		KFREE(clone);
6292 		NBUMPSIDED(fin->fin_out, ns_insert_fail);
6293 		return NULL;
6294 	}
6295 
6296 	np = clone->nat_ptr;
6297 	if (np != NULL) {
6298 		if (softn->ipf_nat_logging)
6299 			ipf_nat_log(softc, softn, clone, NL_CLONE);
6300 		np->in_use++;
6301 	}
6302 	fr = clone->nat_fr;
6303 	if (fr != NULL) {
6304 		MUTEX_ENTER(&fr->fr_lock);
6305 		fr->fr_ref++;
6306 		MUTEX_EXIT(&fr->fr_lock);
6307 	}
6308 
6309 
6310 	/*
6311 	 * Because the clone is created outside the normal loop of things and
6312 	 * TCP has special needs in terms of state, initialise the timeout
6313 	 * state of the new NAT from here.
6314 	 */
6315 	if (clone->nat_pr[0] == IPPROTO_TCP) {
6316 		(void) ipf_tcp_age(&clone->nat_tqe, fin, softn->ipf_nat_tcptq,
6317 				   clone->nat_flags, 2);
6318 	}
6319 	clone->nat_sync = ipf_sync_new(softc, SMC_NAT, fin, clone);
6320 	if (softn->ipf_nat_logging)
6321 		ipf_nat_log(softc, softn, clone, NL_CLONE);
6322 	return clone;
6323 }
6324 
6325 
6326 /* ------------------------------------------------------------------------ */
6327 /* Function:   ipf_nat_wildok                                               */
6328 /* Returns:    int - 1 == packet's ports match wildcards                    */
6329 /*                   0 == packet's ports don't match wildcards              */
6330 /* Parameters: nat(I)   - NAT entry                                         */
6331 /*             sport(I) - source port                                       */
6332 /*             dport(I) - destination port                                  */
6333 /*             flags(I) - wildcard flags                                    */
6334 /*             dir(I)   - packet direction                                  */
6335 /*                                                                          */
6336 /* Use NAT entry and packet direction to determine which combination of     */
6337 /* wildcard flags should be used.                                           */
6338 /* ------------------------------------------------------------------------ */
6339 int
6340 ipf_nat_wildok(nat_t *nat, int sport, int dport, int flags, int dir)
6341 {
6342 	/*
6343 	 * When called by       dir is set to
6344 	 * nat_inlookup         NAT_INBOUND (0)
6345 	 * nat_outlookup        NAT_OUTBOUND (1)
6346 	 *
6347 	 * We simply combine the packet's direction in dir with the original
6348 	 * "intended" direction of that NAT entry in nat->nat_dir to decide
6349 	 * which combination of wildcard flags to allow.
6350 	 */
6351 	switch ((dir << 1) | (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND)))
6352 	{
6353 	case 3: /* outbound packet / outbound entry */
6354 		if (((nat->nat_osport == sport) ||
6355 		    (flags & SI_W_SPORT)) &&
6356 		    ((nat->nat_odport == dport) ||
6357 		    (flags & SI_W_DPORT)))
6358 			return 1;
6359 		break;
6360 	case 2: /* outbound packet / inbound entry */
6361 		if (((nat->nat_osport == dport) ||
6362 		    (flags & SI_W_SPORT)) &&
6363 		    ((nat->nat_odport == sport) ||
6364 		    (flags & SI_W_DPORT)))
6365 			return 1;
6366 		break;
6367 	case 1: /* inbound packet / outbound entry */
6368 		if (((nat->nat_osport == dport) ||
6369 		    (flags & SI_W_SPORT)) &&
6370 		    ((nat->nat_odport == sport) ||
6371 		    (flags & SI_W_DPORT)))
6372 			return 1;
6373 		break;
6374 	case 0: /* inbound packet / inbound entry */
6375 		if (((nat->nat_osport == sport) ||
6376 		    (flags & SI_W_SPORT)) &&
6377 		    ((nat->nat_odport == dport) ||
6378 		    (flags & SI_W_DPORT)))
6379 			return 1;
6380 		break;
6381 	default:
6382 		break;
6383 	}
6384 
6385 	return(0);
6386 }
6387 
6388 
6389 /* ------------------------------------------------------------------------ */
6390 /* Function:    nat_mssclamp                                                */
6391 /* Returns:     Nil                                                         */
6392 /* Parameters:  tcp(I)    - pointer to TCP header                           */
6393 /*              maxmss(I) - value to clamp the TCP MSS to                   */
6394 /*              fin(I)    - pointer to packet information                   */
6395 /*              csump(I)  - pointer to TCP checksum                         */
6396 /*                                                                          */
6397 /* Check for MSS option and clamp it if necessary.  If found and changed,   */
6398 /* then the TCP header checksum will be updated to reflect the change in    */
6399 /* the MSS.                                                                 */
6400 /* ------------------------------------------------------------------------ */
6401 static void
6402 ipf_nat_mssclamp(tcphdr_t *tcp, u_32_t maxmss, fr_info_t *fin, u_short *csump)
6403 {
6404 	u_char *cp, *ep, opt;
6405 	int hlen, advance;
6406 	u_32_t mss, sumd;
6407 
6408 	hlen = TCP_OFF(tcp) << 2;
6409 	if (hlen > sizeof(*tcp)) {
6410 		cp = (u_char *)tcp + sizeof(*tcp);
6411 		ep = (u_char *)tcp + hlen;
6412 
6413 		while (cp < ep) {
6414 			opt = cp[0];
6415 			if (opt == TCPOPT_EOL)
6416 				break;
6417 			else if (opt == TCPOPT_NOP) {
6418 				cp++;
6419 				continue;
6420 			}
6421 
6422 			if (cp + 1 >= ep)
6423 				break;
6424 			advance = cp[1];
6425 			if ((cp + advance > ep) || (advance <= 0))
6426 				break;
6427 			switch (opt)
6428 			{
6429 			case TCPOPT_MAXSEG:
6430 				if (advance != 4)
6431 					break;
6432 				mss = cp[2] * 256 + cp[3];
6433 				if (mss > maxmss) {
6434 					cp[2] = maxmss / 256;
6435 					cp[3] = maxmss & 0xff;
6436 					CALC_SUMD(mss, maxmss, sumd);
6437 					ipf_fix_outcksum(0, csump, sumd, 0);
6438 				}
6439 				break;
6440 			default:
6441 				/* ignore unknown options */
6442 				break;
6443 			}
6444 
6445 			cp += advance;
6446 		}
6447 	}
6448 }
6449 
6450 
6451 /* ------------------------------------------------------------------------ */
6452 /* Function:    ipf_nat_setqueue                                            */
6453 /* Returns:     Nil                                                         */
6454 /* Parameters:  softc(I) - pointer to soft context main structure           */
6455 /*              softn(I) - pointer to NAT context structure                 */
6456 /*              nat(I)- pointer to NAT structure                            */
6457 /* Locks:       ipf_nat (read or write)                                     */
6458 /*                                                                          */
6459 /* Put the NAT entry on its default queue entry, using rev as a helped in   */
6460 /* determining which queue it should be placed on.                          */
6461 /* ------------------------------------------------------------------------ */
6462 void
6463 ipf_nat_setqueue(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, nat_t *nat)
6464 {
6465 	ipftq_t *oifq, *nifq;
6466 	int rev = nat->nat_rev;
6467 
6468 	if (nat->nat_ptr != NULL)
6469 		nifq = nat->nat_ptr->in_tqehead[rev];
6470 	else
6471 		nifq = NULL;
6472 
6473 	if (nifq == NULL) {
6474 		switch (nat->nat_pr[0])
6475 		{
6476 		case IPPROTO_UDP :
6477 			nifq = &softn->ipf_nat_udptq;
6478 			break;
6479 		case IPPROTO_ICMP :
6480 			nifq = &softn->ipf_nat_icmptq;
6481 			break;
6482 		case IPPROTO_TCP :
6483 			nifq = softn->ipf_nat_tcptq +
6484 			       nat->nat_tqe.tqe_state[rev];
6485 			break;
6486 		default :
6487 			nifq = &softn->ipf_nat_iptq;
6488 			break;
6489 		}
6490 	}
6491 
6492 	oifq = nat->nat_tqe.tqe_ifq;
6493 	/*
6494 	 * If it's currently on a timeout queue, move it from one queue to
6495 	 * another, else put it on the end of the newly determined queue.
6496 	 */
6497 	if (oifq != NULL)
6498 		ipf_movequeue(softc->ipf_ticks, &nat->nat_tqe, oifq, nifq);
6499 	else
6500 		ipf_queueappend(softc->ipf_ticks, &nat->nat_tqe, nifq, nat);
6501 	return;
6502 }
6503 
6504 
6505 /* ------------------------------------------------------------------------ */
6506 /* Function:    nat_getnext                                                 */
6507 /* Returns:     int - 0 == ok, else error                                   */
6508 /* Parameters:  softc(I) - pointer to soft context main structure           */
6509 /*              t(I)   - pointer to ipftoken structure                      */
6510 /*              itp(I) - pointer to ipfgeniter_t structure                  */
6511 /*                                                                          */
6512 /* Fetch the next nat/ipnat structure pointer from the linked list and      */
6513 /* copy it out to the storage space pointed to by itp_data.  The next item  */
6514 /* in the list to look at is put back in the ipftoken struture.             */
6515 /* ------------------------------------------------------------------------ */
6516 static int
6517 ipf_nat_getnext(ipf_main_softc_t *softc, ipftoken_t *t, ipfgeniter_t *itp,
6518     ipfobj_t *objp)
6519 {
6520 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6521 	hostmap_t *hm, *nexthm = NULL, zerohm;
6522 	ipnat_t *ipn, *nextipnat = NULL, zeroipn;
6523 	nat_t *nat, *nextnat = NULL, zeronat;
6524 	int error = 0;
6525 	void *nnext;
6526 
6527 	if (itp->igi_nitems != 1) {
6528 		IPFERROR(60075);
6529 		return ENOSPC;
6530 	}
6531 
6532 	READ_ENTER(&softc->ipf_nat);
6533 
6534 	switch (itp->igi_type)
6535 	{
6536 	case IPFGENITER_HOSTMAP :
6537 		hm = t->ipt_data;
6538 		if (hm == NULL) {
6539 			nexthm = softn->ipf_hm_maplist;
6540 		} else {
6541 			nexthm = hm->hm_next;
6542 		}
6543 		if (nexthm != NULL) {
6544 			ATOMIC_INC32(nexthm->hm_ref);
6545 			t->ipt_data = nexthm;
6546 		} else {
6547 			bzero(&zerohm, sizeof(zerohm));
6548 			nexthm = &zerohm;
6549 			t->ipt_data = NULL;
6550 		}
6551 		nnext = nexthm->hm_next;
6552 		break;
6553 
6554 	case IPFGENITER_IPNAT :
6555 		ipn = t->ipt_data;
6556 		if (ipn == NULL) {
6557 			nextipnat = softn->ipf_nat_list;
6558 		} else {
6559 			nextipnat = ipn->in_next;
6560 		}
6561 		if (nextipnat != NULL) {
6562 			ATOMIC_INC32(nextipnat->in_use);
6563 			t->ipt_data = nextipnat;
6564 		} else {
6565 			bzero(&zeroipn, sizeof(zeroipn));
6566 			nextipnat = &zeroipn;
6567 			t->ipt_data = NULL;
6568 		}
6569 		nnext = nextipnat->in_next;
6570 		break;
6571 
6572 	case IPFGENITER_NAT :
6573 		nat = t->ipt_data;
6574 		if (nat == NULL) {
6575 			nextnat = softn->ipf_nat_instances;
6576 		} else {
6577 			nextnat = nat->nat_next;
6578 		}
6579 		if (nextnat != NULL) {
6580 			MUTEX_ENTER(&nextnat->nat_lock);
6581 			nextnat->nat_ref++;
6582 			MUTEX_EXIT(&nextnat->nat_lock);
6583 			t->ipt_data = nextnat;
6584 		} else {
6585 			bzero(&zeronat, sizeof(zeronat));
6586 			nextnat = &zeronat;
6587 			t->ipt_data = NULL;
6588 		}
6589 		nnext = nextnat->nat_next;
6590 		break;
6591 
6592 	default :
6593 		RWLOCK_EXIT(&softc->ipf_nat);
6594 		IPFERROR(60055);
6595 		return EINVAL;
6596 	}
6597 
6598 	RWLOCK_EXIT(&softc->ipf_nat);
6599 
6600 	objp->ipfo_ptr = itp->igi_data;
6601 
6602 	switch (itp->igi_type)
6603 	{
6604 	case IPFGENITER_HOSTMAP :
6605 		error = COPYOUT(nexthm, objp->ipfo_ptr, sizeof(*nexthm));
6606 		if (error != 0) {
6607 			IPFERROR(60049);
6608 			error = EFAULT;
6609 		}
6610 		if (hm != NULL) {
6611 			WRITE_ENTER(&softc->ipf_nat);
6612 			ipf_nat_hostmapdel(softc, &hm);
6613 			RWLOCK_EXIT(&softc->ipf_nat);
6614 		}
6615 		break;
6616 
6617 	case IPFGENITER_IPNAT :
6618 		objp->ipfo_size = nextipnat->in_size;
6619 		objp->ipfo_type = IPFOBJ_IPNAT;
6620 		error = ipf_outobjk(softc, objp, nextipnat);
6621 		if (ipn != NULL) {
6622 			WRITE_ENTER(&softc->ipf_nat);
6623 			ipf_nat_rule_deref(softc, &ipn);
6624 			RWLOCK_EXIT(&softc->ipf_nat);
6625 		}
6626 		break;
6627 
6628 	case IPFGENITER_NAT :
6629 		objp->ipfo_size = sizeof(nat_t);
6630 		objp->ipfo_type = IPFOBJ_NAT;
6631 		error = ipf_outobjk(softc, objp, nextnat);
6632 		if (nat != NULL)
6633 			ipf_nat_deref(softc, &nat);
6634 
6635 		break;
6636 	}
6637 
6638 	if (nnext == NULL)
6639 		ipf_token_mark_complete(t);
6640 
6641 	return error;
6642 }
6643 
6644 
6645 /* ------------------------------------------------------------------------ */
6646 /* Function:    nat_extraflush                                              */
6647 /* Returns:     int - 0 == success, -1 == failure                           */
6648 /* Parameters:  softc(I) - pointer to soft context main structure           */
6649 /*              softn(I) - pointer to NAT context structure                 */
6650 /*              which(I) - how to flush the active NAT table                */
6651 /* Write Locks: ipf_nat                                                     */
6652 /*                                                                          */
6653 /* Flush nat tables.  Three actions currently defined:                      */
6654 /* which == 0 : flush all nat table entries                                 */
6655 /* which == 1 : flush TCP connections which have started to close but are   */
6656 /*	      stuck for some reason.                                        */
6657 /* which == 2 : flush TCP connections which have been idle for a long time, */
6658 /*	      starting at > 4 days idle and working back in successive half-*/
6659 /*	      days to at most 12 hours old.  If this fails to free enough   */
6660 /*            slots then work backwards in half hour slots to 30 minutes.   */
6661 /*            If that too fails, then work backwards in 30 second intervals */
6662 /*            for the last 30 minutes to at worst 30 seconds idle.          */
6663 /* ------------------------------------------------------------------------ */
6664 static int
6665 ipf_nat_extraflush(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, int which)
6666 {
6667 	nat_t *nat, **natp;
6668 	ipftqent_t *tqn;
6669 	ipftq_t *ifq;
6670 	int removed;
6671 	SPL_INT(s);
6672 
6673 	removed = 0;
6674 
6675 	SPL_NET(s);
6676 	switch (which)
6677 	{
6678 	case 0 :
6679 		softn->ipf_nat_stats.ns_flush_all++;
6680 		/*
6681 		 * Style 0 flush removes everything...
6682 		 */
6683 		for (natp = &softn->ipf_nat_instances;
6684 		     ((nat = *natp) != NULL); ) {
6685 			ipf_nat_delete(softc, nat, NL_FLUSH);
6686 			removed++;
6687 		}
6688 		break;
6689 
6690 	case 1 :
6691 		softn->ipf_nat_stats.ns_flush_closing++;
6692 		/*
6693 		 * Since we're only interested in things that are closing,
6694 		 * we can start with the appropriate timeout queue.
6695 		 */
6696 		for (ifq = softn->ipf_nat_tcptq + IPF_TCPS_CLOSE_WAIT;
6697 		     ifq != NULL; ifq = ifq->ifq_next) {
6698 
6699 			for (tqn = ifq->ifq_head; tqn != NULL; ) {
6700 				nat = tqn->tqe_parent;
6701 				tqn = tqn->tqe_next;
6702 				if (nat->nat_pr[0] != IPPROTO_TCP ||
6703 				    nat->nat_pr[1] != IPPROTO_TCP)
6704 					break;
6705 				ipf_nat_delete(softc, nat, NL_EXPIRE);
6706 				removed++;
6707 			}
6708 		}
6709 
6710 		/*
6711 		 * Also need to look through the user defined queues.
6712 		 */
6713 		for (ifq = softn->ipf_nat_utqe; ifq != NULL;
6714 		     ifq = ifq->ifq_next) {
6715 			for (tqn = ifq->ifq_head; tqn != NULL; ) {
6716 				nat = tqn->tqe_parent;
6717 				tqn = tqn->tqe_next;
6718 				if (nat->nat_pr[0] != IPPROTO_TCP ||
6719 				    nat->nat_pr[1] != IPPROTO_TCP)
6720 					continue;
6721 
6722 				if ((nat->nat_tcpstate[0] >
6723 				     IPF_TCPS_ESTABLISHED) &&
6724 				    (nat->nat_tcpstate[1] >
6725 				     IPF_TCPS_ESTABLISHED)) {
6726 					ipf_nat_delete(softc, nat, NL_EXPIRE);
6727 					removed++;
6728 				}
6729 			}
6730 		}
6731 		break;
6732 
6733 		/*
6734 		 * Args 5-11 correspond to flushing those particular states
6735 		 * for TCP connections.
6736 		 */
6737 	case IPF_TCPS_CLOSE_WAIT :
6738 	case IPF_TCPS_FIN_WAIT_1 :
6739 	case IPF_TCPS_CLOSING :
6740 	case IPF_TCPS_LAST_ACK :
6741 	case IPF_TCPS_FIN_WAIT_2 :
6742 	case IPF_TCPS_TIME_WAIT :
6743 	case IPF_TCPS_CLOSED :
6744 		softn->ipf_nat_stats.ns_flush_state++;
6745 		tqn = softn->ipf_nat_tcptq[which].ifq_head;
6746 		while (tqn != NULL) {
6747 			nat = tqn->tqe_parent;
6748 			tqn = tqn->tqe_next;
6749 			ipf_nat_delete(softc, nat, NL_FLUSH);
6750 			removed++;
6751 		}
6752 		break;
6753 
6754 	default :
6755 		if (which < 30)
6756 			break;
6757 
6758 		softn->ipf_nat_stats.ns_flush_timeout++;
6759 		/*
6760 		 * Take a large arbitrary number to mean the number of seconds
6761 		 * for which which consider to be the maximum value we'll allow
6762 		 * the expiration to be.
6763 		 */
6764 		which = IPF_TTLVAL(which);
6765 		for (natp = &softn->ipf_nat_instances;
6766 		     ((nat = *natp) != NULL); ) {
6767 			if (softc->ipf_ticks - nat->nat_touched > which) {
6768 				ipf_nat_delete(softc, nat, NL_FLUSH);
6769 				removed++;
6770 			} else
6771 				natp = &nat->nat_next;
6772 		}
6773 		break;
6774 	}
6775 
6776 	if (which != 2) {
6777 		SPL_X(s);
6778 		return removed;
6779 	}
6780 
6781 	softn->ipf_nat_stats.ns_flush_queue++;
6782 
6783 	/*
6784 	 * Asked to remove inactive entries because the table is full, try
6785 	 * again, 3 times, if first attempt failed with a different criteria
6786 	 * each time.  The order tried in must be in decreasing age.
6787 	 * Another alternative is to implement random drop and drop N entries
6788 	 * at random until N have been freed up.
6789 	 */
6790 	if (softc->ipf_ticks - softn->ipf_nat_last_force_flush >
6791 	    IPF_TTLVAL(5)) {
6792 		softn->ipf_nat_last_force_flush = softc->ipf_ticks;
6793 
6794 		removed = ipf_queueflush(softc, ipf_nat_flush_entry,
6795 					 softn->ipf_nat_tcptq,
6796 					 softn->ipf_nat_utqe,
6797 					 &softn->ipf_nat_stats.ns_active,
6798 					 softn->ipf_nat_table_sz,
6799 					 softn->ipf_nat_table_wm_low);
6800 	}
6801 
6802 	SPL_X(s);
6803 	return removed;
6804 }
6805 
6806 
6807 /* ------------------------------------------------------------------------ */
6808 /* Function:    ipf_nat_flush_entry                                         */
6809 /* Returns:     0 - always succeeds                                         */
6810 /* Parameters:  softc(I) - pointer to soft context main structure           */
6811 /*              entry(I) - pointer to NAT entry                             */
6812 /* Write Locks: ipf_nat                                                     */
6813 /*                                                                          */
6814 /* This function is a stepping stone between ipf_queueflush() and           */
6815 /* nat_dlete().  It is used so we can provide a uniform interface via the   */
6816 /* ipf_queueflush() function.  Since the nat_delete() function returns void */
6817 /* we translate that to mean it always succeeds in deleting something.      */
6818 /* ------------------------------------------------------------------------ */
6819 static int
6820 ipf_nat_flush_entry(ipf_main_softc_t *softc, void *entry)
6821 {
6822 	ipf_nat_delete(softc, entry, NL_FLUSH);
6823 	return 0;
6824 }
6825 
6826 
6827 /* ------------------------------------------------------------------------ */
6828 /* Function:    ipf_nat_iterator                                            */
6829 /* Returns:     int - 0 == ok, else error                                   */
6830 /* Parameters:  softc(I) - pointer to soft context main structure           */
6831 /*              token(I) - pointer to ipftoken structure                    */
6832 /*              itp(I)   - pointer to ipfgeniter_t structure                */
6833 /*              obj(I)   - pointer to data description structure            */
6834 /*                                                                          */
6835 /* This function acts as a handler for the SIOCGENITER ioctls that use a    */
6836 /* generic structure to iterate through a list.  There are three different  */
6837 /* linked lists of NAT related information to go through: NAT rules, active */
6838 /* NAT mappings and the NAT fragment cache.                                 */
6839 /* ------------------------------------------------------------------------ */
6840 static int
6841 ipf_nat_iterator(ipf_main_softc_t *softc, ipftoken_t *token, ipfgeniter_t *itp,
6842     ipfobj_t *obj)
6843 {
6844 	int error;
6845 
6846 	if (itp->igi_data == NULL) {
6847 		IPFERROR(60052);
6848 		return EFAULT;
6849 	}
6850 
6851 	switch (itp->igi_type)
6852 	{
6853 	case IPFGENITER_HOSTMAP :
6854 	case IPFGENITER_IPNAT :
6855 	case IPFGENITER_NAT :
6856 		error = ipf_nat_getnext(softc, token, itp, obj);
6857 		break;
6858 
6859 	case IPFGENITER_NATFRAG :
6860 		error = ipf_frag_nat_next(softc, token, itp);
6861 		break;
6862 	default :
6863 		IPFERROR(60053);
6864 		error = EINVAL;
6865 		break;
6866 	}
6867 
6868 	return error;
6869 }
6870 
6871 
6872 /* ------------------------------------------------------------------------ */
6873 /* Function:    ipf_nat_setpending                                          */
6874 /* Returns:     Nil                                                         */
6875 /* Parameters:  softc(I) - pointer to soft context main structure           */
6876 /*              nat(I)   - pointer to NAT structure                         */
6877 /* Locks:       ipf_nat (read or write)                                     */
6878 /*                                                                          */
6879 /* Put the NAT entry on to the pending queue - this queue has a very short  */
6880 /* lifetime where items are put that can't be deleted straight away because */
6881 /* of locking issues but we want to delete them ASAP, anyway.  In calling   */
6882 /* this function, it is assumed that the owner (if there is one, as shown   */
6883 /* by nat_me) is no longer interested in it.                                */
6884 /* ------------------------------------------------------------------------ */
6885 void
6886 ipf_nat_setpending(ipf_main_softc_t *softc, nat_t *nat)
6887 {
6888 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6889 	ipftq_t *oifq;
6890 
6891 	oifq = nat->nat_tqe.tqe_ifq;
6892 	if (oifq != NULL)
6893 		ipf_movequeue(softc->ipf_ticks, &nat->nat_tqe, oifq,
6894 			      &softn->ipf_nat_pending);
6895 	else
6896 		ipf_queueappend(softc->ipf_ticks, &nat->nat_tqe,
6897 				&softn->ipf_nat_pending, nat);
6898 
6899 	if (nat->nat_me != NULL) {
6900 		*nat->nat_me = NULL;
6901 		nat->nat_me = NULL;
6902 		nat->nat_ref--;
6903 		ASSERT(nat->nat_ref >= 0);
6904 	}
6905 }
6906 
6907 
6908 /* ------------------------------------------------------------------------ */
6909 /* Function:    nat_newrewrite                                              */
6910 /* Returns:     int - -1 == error, 0 == success (no move), 1 == success and */
6911 /*                    allow rule to be moved if IPN_ROUNDR is set.          */
6912 /* Parameters:  fin(I) - pointer to packet information                      */
6913 /*              nat(I) - pointer to NAT entry                               */
6914 /*              ni(I)  - pointer to structure with misc. information needed */
6915 /*                       to create new NAT entry.                           */
6916 /* Write Lock:  ipf_nat                                                     */
6917 /*                                                                          */
6918 /* This function is responsible for setting up an active NAT session where  */
6919 /* we are changing both the source and destination parameters at the same   */
6920 /* time.  The loop in here works differently to elsewhere - each iteration  */
6921 /* is responsible for changing a single parameter that can be incremented.  */
6922 /* So one pass may increase the source IP#, next source port, next dest. IP#*/
6923 /* and the last destination port for a total of 4 iterations to try each.   */
6924 /* This is done to try and exhaustively use the translation space available.*/
6925 /* ------------------------------------------------------------------------ */
6926 static int
6927 ipf_nat_newrewrite(fr_info_t *fin, nat_t *nat, natinfo_t *nai)
6928 {
6929 	int src_search = 1;
6930 	int dst_search = 1;
6931 	fr_info_t frnat;
6932 	u_32_t flags;
6933 	u_short swap;
6934 	ipnat_t *np;
6935 	nat_t *natl;
6936 	int l = 0;
6937 	int changed;
6938 
6939 	natl = NULL;
6940 	changed = -1;
6941 	np = nai->nai_np;
6942 	flags = nat->nat_flags;
6943 	bcopy((char *)fin, (char *)&frnat, sizeof(*fin));
6944 
6945 	nat->nat_hm = NULL;
6946 
6947 	do {
6948 		changed = -1;
6949 		/* TRACE (l, src_search, dst_search, np) */
6950 
6951 		if ((src_search == 0) && (np->in_spnext == 0) &&
6952 		    (dst_search == 0) && (np->in_dpnext == 0)) {
6953 			if (l > 0)
6954 				return -1;
6955 		}
6956 
6957 		/*
6958 		 * Find a new source address
6959 		 */
6960 		if (ipf_nat_nextaddr(fin, &np->in_nsrc, &frnat.fin_saddr,
6961 				     &frnat.fin_saddr) == -1) {
6962 			return -1;
6963 		}
6964 
6965 		if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0xffffffff)) {
6966 			src_search = 0;
6967 			if (np->in_stepnext == 0)
6968 				np->in_stepnext = 1;
6969 
6970 		} else if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0)) {
6971 			src_search = 0;
6972 			if (np->in_stepnext == 0)
6973 				np->in_stepnext = 1;
6974 
6975 		} else if (np->in_nsrcmsk == 0xffffffff) {
6976 			src_search = 0;
6977 			if (np->in_stepnext == 0)
6978 				np->in_stepnext = 1;
6979 
6980 		} else if (np->in_nsrcmsk != 0xffffffff) {
6981 			if (np->in_stepnext == 0 && changed == -1) {
6982 				np->in_snip++;
6983 				np->in_stepnext++;
6984 				changed = 0;
6985 			}
6986 		}
6987 
6988 		if ((flags & IPN_TCPUDPICMP) != 0) {
6989 			if (np->in_spnext != 0)
6990 				frnat.fin_data[0] = np->in_spnext;
6991 
6992 			/*
6993 			 * Standard port translation.  Select next port.
6994 			 */
6995 			if ((flags & IPN_FIXEDSPORT) != 0) {
6996 				np->in_stepnext = 2;
6997 			} else if ((np->in_stepnext == 1) &&
6998 				   (changed == -1) && (natl != NULL)) {
6999 				np->in_spnext++;
7000 				np->in_stepnext++;
7001 				changed = 1;
7002 				if (np->in_spnext > np->in_spmax)
7003 					np->in_spnext = np->in_spmin;
7004 			}
7005 		} else {
7006 			np->in_stepnext = 2;
7007 		}
7008 		np->in_stepnext &= 0x3;
7009 
7010 		/*
7011 		 * Find a new destination address
7012 		 */
7013 		/* TRACE (fin, np, l, frnat) */
7014 
7015 		if (ipf_nat_nextaddr(fin, &np->in_ndst, &frnat.fin_daddr,
7016 				     &frnat.fin_daddr) == -1)
7017 			return -1;
7018 		if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0xffffffff)) {
7019 			dst_search = 0;
7020 			if (np->in_stepnext == 2)
7021 				np->in_stepnext = 3;
7022 
7023 		} else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0)) {
7024 			dst_search = 0;
7025 			if (np->in_stepnext == 2)
7026 				np->in_stepnext = 3;
7027 
7028 		} else if (np->in_ndstmsk == 0xffffffff) {
7029 			dst_search = 0;
7030 			if (np->in_stepnext == 2)
7031 				np->in_stepnext = 3;
7032 
7033 		} else if (np->in_ndstmsk != 0xffffffff) {
7034 			if ((np->in_stepnext == 2) && (changed == -1) &&
7035 			    (natl != NULL)) {
7036 				changed = 2;
7037 				np->in_stepnext++;
7038 				np->in_dnip++;
7039 			}
7040 		}
7041 
7042 		if ((flags & IPN_TCPUDPICMP) != 0) {
7043 			if (np->in_dpnext != 0)
7044 				frnat.fin_data[1] = np->in_dpnext;
7045 
7046 			/*
7047 			 * Standard port translation.  Select next port.
7048 			 */
7049 			if ((flags & IPN_FIXEDDPORT) != 0) {
7050 				np->in_stepnext = 0;
7051 			} else if (np->in_stepnext == 3 && changed == -1) {
7052 				np->in_dpnext++;
7053 				np->in_stepnext++;
7054 				changed = 3;
7055 				if (np->in_dpnext > np->in_dpmax)
7056 					np->in_dpnext = np->in_dpmin;
7057 			}
7058 		} else {
7059 			if (np->in_stepnext == 3)
7060 				np->in_stepnext = 0;
7061 		}
7062 
7063 		/* TRACE (frnat) */
7064 
7065 		/*
7066 		 * Here we do a lookup of the connection as seen from
7067 		 * the outside.  If an IP# pair already exists, try
7068 		 * again.  So if you have A->B becomes C->B, you can
7069 		 * also have D->E become C->E but not D->B causing
7070 		 * another C->B.  Also take protocol and ports into
7071 		 * account when determining whether a pre-existing
7072 		 * NAT setup will cause an external conflict where
7073 		 * this is appropriate.
7074 		 *
7075 		 * fin_data[] is swapped around because we are doing a
7076 		 * lookup of the packet is if it were moving in the opposite
7077 		 * direction of the one we are working with now.
7078 		 */
7079 		if (flags & IPN_TCPUDP) {
7080 			swap = frnat.fin_data[0];
7081 			frnat.fin_data[0] = frnat.fin_data[1];
7082 			frnat.fin_data[1] = swap;
7083 		}
7084 		if (fin->fin_out == 1) {
7085 			natl = ipf_nat_inlookup(&frnat,
7086 						flags & ~(SI_WILDP|NAT_SEARCH),
7087 						(u_int)frnat.fin_p,
7088 						frnat.fin_dst, frnat.fin_src);
7089 
7090 		} else {
7091 			natl = ipf_nat_outlookup(&frnat,
7092 						 flags & ~(SI_WILDP|NAT_SEARCH),
7093 						 (u_int)frnat.fin_p,
7094 						 frnat.fin_dst, frnat.fin_src);
7095 		}
7096 		if (flags & IPN_TCPUDP) {
7097 			swap = frnat.fin_data[0];
7098 			frnat.fin_data[0] = frnat.fin_data[1];
7099 			frnat.fin_data[1] = swap;
7100 		}
7101 
7102 		/* TRACE natl, in_stepnext, l */
7103 
7104 		if ((natl != NULL) && (l > 8))	/* XXX 8 is arbitrary */
7105 			return -1;
7106 
7107 		np->in_stepnext &= 0x3;
7108 
7109 		l++;
7110 		changed = -1;
7111 	} while (natl != NULL);
7112 
7113 	nat->nat_osrcip = fin->fin_src;
7114 	nat->nat_odstip = fin->fin_dst;
7115 	nat->nat_nsrcip = frnat.fin_src;
7116 	nat->nat_ndstip = frnat.fin_dst;
7117 
7118 	if ((flags & IPN_TCPUDP) != 0) {
7119 		nat->nat_osport = htons(fin->fin_data[0]);
7120 		nat->nat_odport = htons(fin->fin_data[1]);
7121 		nat->nat_nsport = htons(frnat.fin_data[0]);
7122 		nat->nat_ndport = htons(frnat.fin_data[1]);
7123 	} else if ((flags & IPN_ICMPQUERY) != 0) {
7124 		nat->nat_oicmpid = fin->fin_data[1];
7125 		nat->nat_nicmpid = frnat.fin_data[1];
7126 	}
7127 
7128 	return 0;
7129 }
7130 
7131 
7132 /* ------------------------------------------------------------------------ */
7133 /* Function:    nat_newdivert                                               */
7134 /* Returns:     int - -1 == error, 0 == success                             */
7135 /* Parameters:  fin(I) - pointer to packet information                      */
7136 /*              nat(I) - pointer to NAT entry                               */
7137 /*              ni(I)  - pointer to structure with misc. information needed */
7138 /*                       to create new NAT entry.                           */
7139 /* Write Lock:  ipf_nat                                                     */
7140 /*                                                                          */
7141 /* Create a new NAT  divert session as defined by the NAT rule.  This is    */
7142 /* somewhat different to other NAT session creation routines because we     */
7143 /* do not iterate through either port numbers or IP addresses, searching    */
7144 /* for a unique mapping, however, a complimentary duplicate check is made.  */
7145 /* ------------------------------------------------------------------------ */
7146 static int
7147 ipf_nat_newdivert(fr_info_t *fin, nat_t *nat, natinfo_t *nai)
7148 {
7149 	ipf_main_softc_t *softc = fin->fin_main_soft;
7150 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7151 	fr_info_t frnat;
7152 	ipnat_t *np;
7153 	nat_t *natl;
7154 	int p;
7155 
7156 	np = nai->nai_np;
7157 	bcopy((char *)fin, (char *)&frnat, sizeof(*fin));
7158 
7159 	nat->nat_pr[0] = 0;
7160 	nat->nat_osrcaddr = fin->fin_saddr;
7161 	nat->nat_odstaddr = fin->fin_daddr;
7162 	frnat.fin_saddr = htonl(np->in_snip);
7163 	frnat.fin_daddr = htonl(np->in_dnip);
7164 	if ((nat->nat_flags & IPN_TCPUDP) != 0) {
7165 		nat->nat_osport = htons(fin->fin_data[0]);
7166 		nat->nat_odport = htons(fin->fin_data[1]);
7167 	} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
7168 		nat->nat_oicmpid = fin->fin_data[1];
7169 	}
7170 
7171 	if (np->in_redir & NAT_DIVERTUDP) {
7172 		frnat.fin_data[0] = np->in_spnext;
7173 		frnat.fin_data[1] = np->in_dpnext;
7174 		frnat.fin_flx |= FI_TCPUDP;
7175 		p = IPPROTO_UDP;
7176 	} else {
7177 		frnat.fin_flx &= ~FI_TCPUDP;
7178 		p = IPPROTO_IPIP;
7179 	}
7180 
7181 	if (fin->fin_out == 1) {
7182 		natl = ipf_nat_inlookup(&frnat, 0, p,
7183 					frnat.fin_dst, frnat.fin_src);
7184 
7185 	} else {
7186 		natl = ipf_nat_outlookup(&frnat, 0, p,
7187 					 frnat.fin_dst, frnat.fin_src);
7188 	}
7189 
7190 	if (natl != NULL) {
7191 		NBUMPSIDED(fin->fin_out, ns_divert_exist);
7192 		return -1;
7193 	}
7194 
7195 	nat->nat_nsrcaddr = frnat.fin_saddr;
7196 	nat->nat_ndstaddr = frnat.fin_daddr;
7197 	if ((nat->nat_flags & IPN_TCPUDP) != 0) {
7198 		nat->nat_nsport = htons(frnat.fin_data[0]);
7199 		nat->nat_ndport = htons(frnat.fin_data[1]);
7200 	} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
7201 		nat->nat_nicmpid = frnat.fin_data[1];
7202 	}
7203 
7204 	nat->nat_pr[fin->fin_out] = fin->fin_p;
7205 	nat->nat_pr[1 - fin->fin_out] = p;
7206 
7207 	if (np->in_redir & NAT_REDIRECT)
7208 		nat->nat_dir = NAT_DIVERTIN;
7209 	else
7210 		nat->nat_dir = NAT_DIVERTOUT;
7211 
7212 	return 0;
7213 }
7214 
7215 
7216 /* ------------------------------------------------------------------------ */
7217 /* Function:    nat_builddivertmp                                           */
7218 /* Returns:     int - -1 == error, 0 == success                             */
7219 /* Parameters:  softn(I) - pointer to NAT context structure                 */
7220 /*              np(I)    - pointer to a NAT rule                            */
7221 /*                                                                          */
7222 /* For divert rules, a skeleton packet representing what will be prepended  */
7223 /* to the real packet is created.  Even though we don't have the full       */
7224 /* packet here, a checksum is calculated that we update later when we       */
7225 /* fill in the final details.  At present a 0 checksum for UDP is being set */
7226 /* here because it is expected that divert will be used for localhost.      */
7227 /* ------------------------------------------------------------------------ */
7228 static int
7229 ipf_nat_builddivertmp(ipf_nat_softc_t *softn, ipnat_t *np)
7230 {
7231 	udphdr_t *uh;
7232 	size_t len;
7233 	ip_t *ip;
7234 
7235 	if ((np->in_redir & NAT_DIVERTUDP) != 0)
7236 		len = sizeof(ip_t) + sizeof(udphdr_t);
7237 	else
7238 		len = sizeof(ip_t);
7239 
7240 	ALLOC_MB_T(np->in_divmp, len);
7241 	if (np->in_divmp == NULL) {
7242 		NBUMPD(ipf_nat_stats, ns_divert_build);
7243 		return -1;
7244 	}
7245 
7246 	/*
7247 	 * First, the header to get the packet diverted to the new destination
7248 	 */
7249 	ip = MTOD(np->in_divmp, ip_t *);
7250 	IP_V_A(ip, 4);
7251 	IP_HL_A(ip, 5);
7252 	ip->ip_tos = 0;
7253 	if ((np->in_redir & NAT_DIVERTUDP) != 0)
7254 		ip->ip_p = IPPROTO_UDP;
7255 	else
7256 		ip->ip_p = IPPROTO_IPIP;
7257 	ip->ip_ttl = 255;
7258 	ip->ip_off = 0;
7259 	ip->ip_sum = 0;
7260 	ip->ip_len = htons(len);
7261 	ip->ip_id = 0;
7262 	ip->ip_src.s_addr = htonl(np->in_snip);
7263 	ip->ip_dst.s_addr = htonl(np->in_dnip);
7264 	ip->ip_sum = ipf_cksum((u_short *)ip, sizeof(*ip));
7265 
7266 	if (np->in_redir & NAT_DIVERTUDP) {
7267 		uh = (udphdr_t *)(ip + 1);
7268 		uh->uh_sum = 0;
7269 		uh->uh_ulen = 8;
7270 		uh->uh_sport = htons(np->in_spnext);
7271 		uh->uh_dport = htons(np->in_dpnext);
7272 	}
7273 
7274 	return 0;
7275 }
7276 
7277 
7278 #define	MINDECAP	(sizeof(ip_t) + sizeof(udphdr_t) + sizeof(ip_t))
7279 
7280 /* ------------------------------------------------------------------------ */
7281 /* Function:    nat_decap                                                   */
7282 /* Returns:     int - -1 == error, 0 == success                             */
7283 /* Parameters:  fin(I) - pointer to packet information                      */
7284 /*              nat(I) - pointer to current NAT session                     */
7285 /*                                                                          */
7286 /* This function is responsible for undoing a packet's encapsulation in the */
7287 /* reverse of an encap/divert rule.  After removing the outer encapsulation */
7288 /* it is necessary to call ipf_makefrip() again so that the contents of 'fin'*/
7289 /* match the "new" packet as it may still be used by IPFilter elsewhere.    */
7290 /* We use "dir" here as the basis for some of the expectations about the    */
7291 /* outer header.  If we return an error, the goal is to leave the original  */
7292 /* packet information undisturbed - this falls short at the end where we'd  */
7293 /* need to back a backup copy of "fin" - expensive.                         */
7294 /* ------------------------------------------------------------------------ */
7295 static int
7296 ipf_nat_decap(fr_info_t *fin, nat_t *nat)
7297 {
7298 	ipf_main_softc_t *softc = fin->fin_main_soft;
7299 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7300 	char *hdr;
7301 	int hlen;
7302 	int skip;
7303 	mb_t *m;
7304 
7305 	if ((fin->fin_flx & FI_ICMPERR) != 0) {
7306 		/*
7307 		 * ICMP packets don't get decapsulated, instead what we need
7308 		 * to do is change the ICMP reply from including (in the data
7309 		 * portion for errors) the encapsulated packet that we sent
7310 		 * out to something that resembles the original packet prior
7311 		 * to encapsulation.  This isn't done here - all we're doing
7312 		 * here is changing the outer address to ensure that it gets
7313 		 * targetted back to the correct system.
7314 		 */
7315 
7316 		if (nat->nat_dir & NAT_OUTBOUND) {
7317 			u_32_t sum1, sum2, sumd;
7318 
7319 			sum1 = ntohl(fin->fin_daddr);
7320 			sum2 = ntohl(nat->nat_osrcaddr);
7321 			CALC_SUMD(sum1, sum2, sumd);
7322 			fin->fin_ip->ip_dst = nat->nat_osrcip;
7323 			fin->fin_daddr = nat->nat_osrcaddr;
7324 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
7325      defined(__osf__) || defined(linux)
7326 			ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, sumd, 0);
7327 #endif
7328 		}
7329 		return 0;
7330 	}
7331 
7332 	m = fin->fin_m;
7333 	skip = fin->fin_hlen;
7334 
7335 	switch (nat->nat_dir)
7336 	{
7337 	case NAT_DIVERTIN :
7338 	case NAT_DIVERTOUT :
7339 		if (fin->fin_plen < MINDECAP)
7340 			return -1;
7341 		skip += sizeof(udphdr_t);
7342 		break;
7343 
7344 	case NAT_ENCAPIN :
7345 	case NAT_ENCAPOUT :
7346 		if (fin->fin_plen < (skip + sizeof(ip_t)))
7347 			return -1;
7348 		break;
7349 	default :
7350 		return -1;
7351 		/* NOTREACHED */
7352 	}
7353 
7354 	/*
7355 	 * The aim here is to keep the original packet details in "fin" for
7356 	 * as long as possible so that returning with an error is for the
7357 	 * original packet and there is little undoing work to do.
7358 	 */
7359 	if (M_LEN(m) < skip + sizeof(ip_t)) {
7360 		if (ipf_pr_pullup(fin, skip + sizeof(ip_t)) == -1)
7361 			return -1;
7362 	}
7363 
7364 	hdr = MTOD(fin->fin_m, char *);
7365 	fin->fin_ip = (ip_t *)(hdr + skip);
7366 	hlen = IP_HL(fin->fin_ip) << 2;
7367 
7368 	if (ipf_pr_pullup(fin, skip + hlen) == -1) {
7369 		NBUMPSIDED(fin->fin_out, ns_decap_pullup);
7370 		return -1;
7371 	}
7372 
7373 	fin->fin_hlen = hlen;
7374 	fin->fin_dlen -= skip;
7375 	fin->fin_plen -= skip;
7376 	fin->fin_ipoff += skip;
7377 
7378 	if (ipf_makefrip(hlen, (ip_t *)hdr, fin) == -1) {
7379 		NBUMPSIDED(fin->fin_out, ns_decap_bad);
7380 		return -1;
7381 	}
7382 
7383 	return skip;
7384 }
7385 
7386 
7387 /* ------------------------------------------------------------------------ */
7388 /* Function:    nat_nextaddr                                                */
7389 /* Returns:     int - -1 == bad input (no new address),                     */
7390 /*                     0 == success and dst has new address                 */
7391 /* Parameters:  fin(I) - pointer to packet information                      */
7392 /*              na(I)  - how to generate new address                        */
7393 /*              old(I) - original address being replaced                    */
7394 /*              dst(O) - where to put the new address                       */
7395 /* Write Lock:  ipf_nat                                                     */
7396 /*                                                                          */
7397 /* This function uses the contents of the "na" structure, in combination    */
7398 /* with "old" to produce a new address to store in "dst".  Not all of the   */
7399 /* possible uses of "na" will result in a new address.                      */
7400 /* ------------------------------------------------------------------------ */
7401 static int
7402 ipf_nat_nextaddr(fr_info_t *fin, nat_addr_t *na, u_32_t *old, u_32_t *dst)
7403 {
7404 	ipf_main_softc_t *softc = fin->fin_main_soft;
7405 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7406 	u_32_t amin, amax, new;
7407 	i6addr_t newip;
7408 	int error;
7409 
7410 	new = 0;
7411 	amin = na->na_addr[0].in4.s_addr;
7412 
7413 	switch (na->na_atype)
7414 	{
7415 	case FRI_RANGE :
7416 		amax = na->na_addr[1].in4.s_addr;
7417 		break;
7418 
7419 	case FRI_NETMASKED :
7420 	case FRI_DYNAMIC :
7421 	case FRI_NORMAL :
7422 		/*
7423 		 * Compute the maximum address by adding the inverse of the
7424 		 * netmask to the minimum address.
7425 		 */
7426 		amax = ~na->na_addr[1].in4.s_addr;
7427 		amax |= amin;
7428 		break;
7429 
7430 	case FRI_LOOKUP :
7431 		break;
7432 
7433 	case FRI_BROADCAST :
7434 	case FRI_PEERADDR :
7435 	case FRI_NETWORK :
7436 	default :
7437 		return -1;
7438 	}
7439 
7440 	error = -1;
7441 
7442 	if (na->na_atype == FRI_LOOKUP) {
7443 		if (na->na_type == IPLT_DSTLIST) {
7444 			error = ipf_dstlist_select_node(fin, na->na_ptr, dst,
7445 							NULL);
7446 		} else {
7447 			NBUMPSIDE(fin->fin_out, ns_badnextaddr);
7448 		}
7449 
7450 	} else if (na->na_atype == IPLT_NONE) {
7451 		/*
7452 		 * 0/0 as the new address means leave it alone.
7453 		 */
7454 		if (na->na_addr[0].in4.s_addr == 0 &&
7455 		    na->na_addr[1].in4.s_addr == 0) {
7456 			new = *old;
7457 
7458 		/*
7459 		 * 0/32 means get the interface's address
7460 		 */
7461 		} else if (na->na_addr[0].in4.s_addr == 0 &&
7462 			   na->na_addr[1].in4.s_addr == 0xffffffff) {
7463 			if (ipf_ifpaddr(softc, 4, na->na_atype,
7464 					fin->fin_ifp, &newip, NULL) == -1) {
7465 				NBUMPSIDED(fin->fin_out, ns_ifpaddrfail);
7466 				return -1;
7467 			}
7468 			new = newip.in4.s_addr;
7469 		} else {
7470 			new = htonl(na->na_nextip);
7471 		}
7472 		*dst = new;
7473 		error = 0;
7474 
7475 	} else {
7476 		NBUMPSIDE(fin->fin_out, ns_badnextaddr);
7477 	}
7478 
7479 	return error;
7480 }
7481 
7482 
7483 /* ------------------------------------------------------------------------ */
7484 /* Function:    nat_nextaddrinit                                            */
7485 /* Returns:     int - 0 == success, else error number                       */
7486 /* Parameters:  softc(I) - pointer to soft context main structure           */
7487 /*              na(I)      - NAT address information for generating new addr*/
7488 /*              initial(I) - flag indicating if it is the first call for    */
7489 /*                           this "na" structure.                           */
7490 /*              ifp(I)     - network interface to derive address            */
7491 /*                           information from.                              */
7492 /*                                                                          */
7493 /* This function is expected to be called in two scenarious: when a new NAT */
7494 /* rule is loaded into the kernel and when the list of NAT rules is sync'd  */
7495 /* up with the valid network interfaces (possibly due to them changing.)    */
7496 /* To distinguish between these, the "initial" parameter is used.  If it is */
7497 /* 1 then this indicates the rule has just been reloaded and 0 for when we  */
7498 /* are updating information.  This difference is important because in       */
7499 /* instances where we are not updating address information associated with  */
7500 /* a network interface, we don't want to disturb what the "next" address to */
7501 /* come out of ipf_nat_nextaddr() will be.                                  */
7502 /* ------------------------------------------------------------------------ */
7503 static int
7504 ipf_nat_nextaddrinit(ipf_main_softc_t *softc, char *base, nat_addr_t *na,
7505     int initial, void *ifp)
7506 {
7507 
7508 	switch (na->na_atype)
7509 	{
7510 	case FRI_LOOKUP :
7511 		if (na->na_subtype == 0) {
7512 			na->na_ptr = ipf_lookup_res_num(softc, IPL_LOGNAT,
7513 							na->na_type,
7514 							na->na_num,
7515 							&na->na_func);
7516 		} else if (na->na_subtype == 1) {
7517 			na->na_ptr = ipf_lookup_res_name(softc, IPL_LOGNAT,
7518 							 na->na_type,
7519 							 base + na->na_num,
7520 							 &na->na_func);
7521 		}
7522 		if (na->na_func == NULL) {
7523 			IPFERROR(60060);
7524 			return ESRCH;
7525 		}
7526 		if (na->na_ptr == NULL) {
7527 			IPFERROR(60056);
7528 			return ESRCH;
7529 		}
7530 		break;
7531 
7532 	case FRI_DYNAMIC :
7533 	case FRI_BROADCAST :
7534 	case FRI_NETWORK :
7535 	case FRI_NETMASKED :
7536 	case FRI_PEERADDR :
7537 		if (ifp != NULL)
7538 			(void )ipf_ifpaddr(softc, 4, na->na_atype, ifp,
7539 					   &na->na_addr[0], &na->na_addr[1]);
7540 		break;
7541 
7542 	case FRI_SPLIT :
7543 	case FRI_RANGE :
7544 		if (initial)
7545 			na->na_nextip = ntohl(na->na_addr[0].in4.s_addr);
7546 		break;
7547 
7548 	case FRI_NONE :
7549 		na->na_addr[0].in4.s_addr &= na->na_addr[1].in4.s_addr;
7550 		return 0;
7551 
7552 	case FRI_NORMAL :
7553 		na->na_addr[0].in4.s_addr &= na->na_addr[1].in4.s_addr;
7554 		break;
7555 
7556 	default :
7557 		IPFERROR(60054);
7558 		return EINVAL;
7559 	}
7560 
7561 	if (initial && (na->na_atype == FRI_NORMAL)) {
7562 		if (na->na_addr[0].in4.s_addr == 0) {
7563 			if ((na->na_addr[1].in4.s_addr == 0xffffffff) ||
7564 			    (na->na_addr[1].in4.s_addr == 0)) {
7565 				return 0;
7566 			}
7567 		}
7568 
7569 		if (na->na_addr[1].in4.s_addr == 0xffffffff) {
7570 			na->na_nextip = ntohl(na->na_addr[0].in4.s_addr);
7571 		} else {
7572 			na->na_nextip = ntohl(na->na_addr[0].in4.s_addr) + 1;
7573 		}
7574 	}
7575 
7576 	return 0;
7577 }
7578 
7579 
7580 /* ------------------------------------------------------------------------ */
7581 /* Function:    ipf_nat_matchflush                                          */
7582 /* Returns:     int - -1 == error, 0 == success                             */
7583 /* Parameters:  softc(I) - pointer to soft context main structure           */
7584 /*              softn(I) - pointer to NAT context structure                 */
7585 /*              nat(I)   - pointer to current NAT session                   */
7586 /*                                                                          */
7587 /* ------------------------------------------------------------------------ */
7588 static int
7589 ipf_nat_matchflush(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, void *data)
7590 {
7591 	int *array, flushed, error;
7592 	nat_t *nat, *natnext;
7593 	ipfobj_t obj;
7594 
7595 	error = ipf_matcharray_load(softc, data, &obj, &array);
7596 	if (error != 0)
7597 		return error;
7598 
7599 	flushed = 0;
7600 
7601 	for (nat = softn->ipf_nat_instances; nat != NULL; nat = natnext) {
7602 		natnext = nat->nat_next;
7603 		if (ipf_nat_matcharray(nat, array, softc->ipf_ticks) == 0) {
7604 			ipf_nat_delete(softc, nat, NL_FLUSH);
7605 			flushed++;
7606 		}
7607 	}
7608 
7609 	obj.ipfo_retval = flushed;
7610 	error = BCOPYOUT(&obj, data, sizeof(obj));
7611 
7612 	KFREES(array, array[0] * sizeof(*array));
7613 
7614 	return error;
7615 }
7616 
7617 
7618 /* ------------------------------------------------------------------------ */
7619 /* Function:    ipf_nat_matcharray                                          */
7620 /* Returns:     int - -1 == error, 0 == success                             */
7621 /* Parameters:  fin(I) - pointer to packet information                      */
7622 /*              nat(I) - pointer to current NAT session                     */
7623 /*                                                                          */
7624 /* ------------------------------------------------------------------------ */
7625 static int
7626 ipf_nat_matcharray(nat_t *nat, int *array, u_long ticks)
7627 {
7628 	int i, n, *x, e, p;
7629 
7630 	e = 0;
7631 	n = array[0];
7632 	x = array + 1;
7633 
7634 	for (; n > 0; x += 3 + x[2]) {
7635 		if (x[0] == IPF_EXP_END)
7636 			break;
7637 		e = 0;
7638 
7639 		n -= x[2] + 3;
7640 		if (n < 0)
7641 			break;
7642 
7643 		p = x[0] >> 16;
7644 		if (p != 0 && p != nat->nat_pr[1])
7645 			break;
7646 
7647 		switch (x[0])
7648 		{
7649 		case IPF_EXP_IP_PR :
7650 			for (i = 0; !e && i < x[2]; i++) {
7651 				e |= (nat->nat_pr[1] == x[i + 3]);
7652 			}
7653 			break;
7654 
7655 		case IPF_EXP_IP_SRCADDR :
7656 			if (nat->nat_v[0] == 4) {
7657 				for (i = 0; !e && i < x[2]; i++) {
7658 					e |= ((nat->nat_osrcaddr & x[i + 4]) ==
7659 					      x[i + 3]);
7660 				}
7661 			}
7662 			if (nat->nat_v[1] == 4) {
7663 				for (i = 0; !e && i < x[2]; i++) {
7664 					e |= ((nat->nat_nsrcaddr & x[i + 4]) ==
7665 					      x[i + 3]);
7666 				}
7667 			}
7668 			break;
7669 
7670 		case IPF_EXP_IP_DSTADDR :
7671 			if (nat->nat_v[0] == 4) {
7672 				for (i = 0; !e && i < x[2]; i++) {
7673 					e |= ((nat->nat_odstaddr & x[i + 4]) ==
7674 					      x[i + 3]);
7675 				}
7676 			}
7677 			if (nat->nat_v[1] == 4) {
7678 				for (i = 0; !e && i < x[2]; i++) {
7679 					e |= ((nat->nat_ndstaddr & x[i + 4]) ==
7680 					      x[i + 3]);
7681 				}
7682 			}
7683 			break;
7684 
7685 		case IPF_EXP_IP_ADDR :
7686 			for (i = 0; !e && i < x[2]; i++) {
7687 				if (nat->nat_v[0] == 4) {
7688 					e |= ((nat->nat_osrcaddr & x[i + 4]) ==
7689 					      x[i + 3]);
7690 				}
7691 				if (nat->nat_v[1] == 4) {
7692 					e |= ((nat->nat_nsrcaddr & x[i + 4]) ==
7693 					      x[i + 3]);
7694 				}
7695 				if (nat->nat_v[0] == 4) {
7696 					e |= ((nat->nat_odstaddr & x[i + 4]) ==
7697 					      x[i + 3]);
7698 				}
7699 				if (nat->nat_v[1] == 4) {
7700 					e |= ((nat->nat_ndstaddr & x[i + 4]) ==
7701 					      x[i + 3]);
7702 				}
7703 			}
7704 			break;
7705 
7706 #ifdef USE_INET6
7707 		case IPF_EXP_IP6_SRCADDR :
7708 			if (nat->nat_v[0] == 6) {
7709 				for (i = 0; !e && i < x[3]; i++) {
7710 					e |= IP6_MASKEQ(&nat->nat_osrc6,
7711 							x + i + 7, x + i + 3);
7712 				}
7713 			}
7714 			if (nat->nat_v[1] == 6) {
7715 				for (i = 0; !e && i < x[3]; i++) {
7716 					e |= IP6_MASKEQ(&nat->nat_nsrc6,
7717 							x + i + 7, x + i + 3);
7718 				}
7719 			}
7720 			break;
7721 
7722 		case IPF_EXP_IP6_DSTADDR :
7723 			if (nat->nat_v[0] == 6) {
7724 				for (i = 0; !e && i < x[3]; i++) {
7725 					e |= IP6_MASKEQ(&nat->nat_odst6,
7726 							x + i + 7,
7727 							x + i + 3);
7728 				}
7729 			}
7730 			if (nat->nat_v[1] == 6) {
7731 				for (i = 0; !e && i < x[3]; i++) {
7732 					e |= IP6_MASKEQ(&nat->nat_ndst6,
7733 							x + i + 7,
7734 							x + i + 3);
7735 				}
7736 			}
7737 			break;
7738 
7739 		case IPF_EXP_IP6_ADDR :
7740 			for (i = 0; !e && i < x[3]; i++) {
7741 				if (nat->nat_v[0] == 6) {
7742 					e |= IP6_MASKEQ(&nat->nat_osrc6,
7743 							x + i + 7,
7744 							x + i + 3);
7745 				}
7746 				if (nat->nat_v[0] == 6) {
7747 					e |= IP6_MASKEQ(&nat->nat_odst6,
7748 							x + i + 7,
7749 							x + i + 3);
7750 				}
7751 				if (nat->nat_v[1] == 6) {
7752 					e |= IP6_MASKEQ(&nat->nat_nsrc6,
7753 							x + i + 7,
7754 							x + i + 3);
7755 				}
7756 				if (nat->nat_v[1] == 6) {
7757 					e |= IP6_MASKEQ(&nat->nat_ndst6,
7758 							x + i + 7,
7759 							x + i + 3);
7760 				}
7761 			}
7762 			break;
7763 #endif
7764 
7765 		case IPF_EXP_UDP_PORT :
7766 		case IPF_EXP_TCP_PORT :
7767 			for (i = 0; !e && i < x[2]; i++) {
7768 				e |= (nat->nat_nsport == x[i + 3]) ||
7769 				     (nat->nat_ndport == x[i + 3]);
7770 			}
7771 			break;
7772 
7773 		case IPF_EXP_UDP_SPORT :
7774 		case IPF_EXP_TCP_SPORT :
7775 			for (i = 0; !e && i < x[2]; i++) {
7776 				e |= (nat->nat_nsport == x[i + 3]);
7777 			}
7778 			break;
7779 
7780 		case IPF_EXP_UDP_DPORT :
7781 		case IPF_EXP_TCP_DPORT :
7782 			for (i = 0; !e && i < x[2]; i++) {
7783 				e |= (nat->nat_ndport == x[i + 3]);
7784 			}
7785 			break;
7786 
7787 		case IPF_EXP_TCP_STATE :
7788 			for (i = 0; !e && i < x[2]; i++) {
7789 				e |= (nat->nat_tcpstate[0] == x[i + 3]) ||
7790 				     (nat->nat_tcpstate[1] == x[i + 3]);
7791 			}
7792 			break;
7793 
7794 		case IPF_EXP_IDLE_GT :
7795 			e |= (ticks - nat->nat_touched > x[3]);
7796 			break;
7797 		}
7798 		e ^= x[1];
7799 
7800 		if (!e)
7801 			break;
7802 	}
7803 
7804 	return e;
7805 }
7806 
7807 
7808 /* ------------------------------------------------------------------------ */
7809 /* Function:    ipf_nat_gettable                                            */
7810 /* Returns:     int     - 0 = success, else error                           */
7811 /* Parameters:  softc(I) - pointer to soft context main structure           */
7812 /*              softn(I) - pointer to NAT context structure                 */
7813 /*              data(I)  - pointer to ioctl data                            */
7814 /*                                                                          */
7815 /* This function handles ioctl requests for tables of nat information.      */
7816 /* At present the only table it deals with is the hash bucket statistics.   */
7817 /* ------------------------------------------------------------------------ */
7818 static int
7819 ipf_nat_gettable(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, char *data)
7820 {
7821 	ipftable_t table;
7822 	int error;
7823 
7824 	error = ipf_inobj(softc, data, NULL, &table, IPFOBJ_GTABLE);
7825 	if (error != 0)
7826 		return error;
7827 
7828 	switch (table.ita_type)
7829 	{
7830 	case IPFTABLE_BUCKETS_NATIN :
7831 		error = COPYOUT(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
7832 				table.ita_table,
7833 				softn->ipf_nat_table_sz * sizeof(u_int));
7834 		break;
7835 
7836 	case IPFTABLE_BUCKETS_NATOUT :
7837 		error = COPYOUT(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
7838 				table.ita_table,
7839 				softn->ipf_nat_table_sz * sizeof(u_int));
7840 		break;
7841 
7842 	default :
7843 		IPFERROR(60058);
7844 		return EINVAL;
7845 	}
7846 
7847 	if (error != 0) {
7848 		IPFERROR(60059);
7849 		error = EFAULT;
7850 	}
7851 	return error;
7852 }
7853 
7854 
7855 /* ------------------------------------------------------------------------ */
7856 /* Function:    ipf_nat_settimeout                                          */
7857 /* Returns:     int  - 0 = success, else failure			    */
7858 /* Parameters:  softc(I) - pointer to soft context main structure           */
7859 /*              t(I) - pointer to tunable                                   */
7860 /*              p(I) - pointer to new tuning data                           */
7861 /*                                                                          */
7862 /* Apply the timeout change to the NAT timeout queues.                      */
7863 /* ------------------------------------------------------------------------ */
7864 int
7865 ipf_nat_settimeout(struct ipf_main_softc_s *softc, ipftuneable_t *t,
7866     ipftuneval_t *p)
7867 {
7868 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7869 
7870 	if (!strncmp(t->ipft_name, "tcp_", 4))
7871 		return ipf_settimeout_tcp(t, p, softn->ipf_nat_tcptq);
7872 
7873 	if (!strcmp(t->ipft_name, "udp_timeout")) {
7874 		ipf_apply_timeout(&softn->ipf_nat_udptq, p->ipftu_int);
7875 	} else if (!strcmp(t->ipft_name, "udp_ack_timeout")) {
7876 		ipf_apply_timeout(&softn->ipf_nat_udpacktq, p->ipftu_int);
7877 	} else if (!strcmp(t->ipft_name, "icmp_timeout")) {
7878 		ipf_apply_timeout(&softn->ipf_nat_icmptq, p->ipftu_int);
7879 	} else if (!strcmp(t->ipft_name, "icmp_ack_timeout")) {
7880 		ipf_apply_timeout(&softn->ipf_nat_icmpacktq, p->ipftu_int);
7881 	} else if (!strcmp(t->ipft_name, "ip_timeout")) {
7882 		ipf_apply_timeout(&softn->ipf_nat_iptq, p->ipftu_int);
7883 	} else {
7884 		IPFERROR(60062);
7885 		return ESRCH;
7886 	}
7887 	return 0;
7888 }
7889 
7890 
7891 /* ------------------------------------------------------------------------ */
7892 /* Function:    ipf_nat_rehash                                              */
7893 /* Returns:     int  - 0 = success, else failure			    */
7894 /* Parameters:  softc(I) - pointer to soft context main structure           */
7895 /*              t(I) - pointer to tunable                                   */
7896 /*              p(I) - pointer to new tuning data                           */
7897 /*                                                                          */
7898 /* To change the size of the basic NAT table, we need to first allocate the */
7899 /* new tables (lest it fails and we've got nowhere to store all of the NAT  */
7900 /* sessions currently active) and then walk through the entire list and     */
7901 /* insert them into the table.  There are two tables here: an inbound one   */
7902 /* and an outbound one.  Each NAT entry goes into each table once.          */
7903 /* ------------------------------------------------------------------------ */
7904 int
7905 ipf_nat_rehash(ipf_main_softc_t *softc, ipftuneable_t *t, ipftuneval_t *p)
7906 {
7907 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7908 	nat_t **newtab[2], *nat, **natp;
7909 	u_int *bucketlens[2];
7910 	u_int maxbucket;
7911 	u_int newsize;
7912 	int error;
7913 	u_int hv;
7914 	int i;
7915 
7916 	newsize = p->ipftu_int;
7917 	/*
7918 	 * In case there is nothing to do...
7919 	 */
7920 	if (newsize == softn->ipf_nat_table_sz)
7921 		return 0;
7922 
7923 	newtab[0] = NULL;
7924 	newtab[1] = NULL;
7925 	bucketlens[0] = NULL;
7926 	bucketlens[1] = NULL;
7927 	/*
7928 	 * 4 tables depend on the NAT table size: the inbound looking table,
7929 	 * the outbound lookup table and the hash chain length for each.
7930 	 */
7931 	KMALLOCS(newtab[0], nat_t **, newsize * sizeof(nat_t *));
7932 	if (newtab[0] == NULL) {
7933 		error = 60063;
7934 		goto badrehash;
7935 	}
7936 
7937 	KMALLOCS(newtab[1], nat_t **, newsize * sizeof(nat_t *));
7938 	if (newtab[1] == NULL) {
7939 		error = 60064;
7940 		goto badrehash;
7941 	}
7942 
7943 	KMALLOCS(bucketlens[0], u_int *, newsize * sizeof(u_int));
7944 	if (bucketlens[0] == NULL) {
7945 		error = 60065;
7946 		goto badrehash;
7947 	}
7948 
7949 	KMALLOCS(bucketlens[1], u_int *, newsize * sizeof(u_int));
7950 	if (bucketlens[1] == NULL) {
7951 		error = 60066;
7952 		goto badrehash;
7953 	}
7954 
7955 	/*
7956 	 * Recalculate the maximum length based on the new size.
7957 	 */
7958 	for (maxbucket = 0, i = newsize; i > 0; i >>= 1)
7959 		maxbucket++;
7960 	maxbucket *= 2;
7961 
7962 	bzero((char *)newtab[0], newsize * sizeof(nat_t *));
7963 	bzero((char *)newtab[1], newsize * sizeof(nat_t *));
7964 	bzero((char *)bucketlens[0], newsize * sizeof(u_int));
7965 	bzero((char *)bucketlens[1], newsize * sizeof(u_int));
7966 
7967 	WRITE_ENTER(&softc->ipf_nat);
7968 
7969 	if (softn->ipf_nat_table[0] != NULL) {
7970 		KFREES(softn->ipf_nat_table[0],
7971 		       softn->ipf_nat_table_sz *
7972 		       sizeof(*softn->ipf_nat_table[0]));
7973 	}
7974 	softn->ipf_nat_table[0] = newtab[0];
7975 
7976 	if (softn->ipf_nat_table[1] != NULL) {
7977 		KFREES(softn->ipf_nat_table[1],
7978 		       softn->ipf_nat_table_sz *
7979 		       sizeof(*softn->ipf_nat_table[1]));
7980 	}
7981 	softn->ipf_nat_table[1] = newtab[1];
7982 
7983 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen != NULL) {
7984 		KFREES(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
7985 		       softn->ipf_nat_table_sz * sizeof(u_int));
7986 	}
7987 	softn->ipf_nat_stats.ns_side[0].ns_bucketlen = bucketlens[0];
7988 
7989 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen != NULL) {
7990 		KFREES(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
7991 		       softn->ipf_nat_table_sz * sizeof(u_int));
7992 	}
7993 	softn->ipf_nat_stats.ns_side[1].ns_bucketlen = bucketlens[1];
7994 
7995 	softn->ipf_nat_maxbucket = maxbucket;
7996 	softn->ipf_nat_table_sz = newsize;
7997 	/*
7998 	 * Walk through the entire list of NAT table entries and put them
7999 	 * in the new NAT table, somewhere.  Because we have a new table,
8000 	 * we need to restart the counter of how many chains are in use.
8001 	 */
8002 	softn->ipf_nat_stats.ns_side[0].ns_inuse = 0;
8003 	softn->ipf_nat_stats.ns_side[1].ns_inuse = 0;
8004 
8005 	for (nat = softn->ipf_nat_instances; nat != NULL; nat = nat->nat_next) {
8006 		nat->nat_hnext[0] = NULL;
8007 		nat->nat_phnext[0] = NULL;
8008 		hv = nat->nat_hv[0] % softn->ipf_nat_table_sz;
8009 
8010 		natp = &softn->ipf_nat_table[0][hv];
8011 		if (*natp) {
8012 			(*natp)->nat_phnext[0] = &nat->nat_hnext[0];
8013 		} else {
8014 			NBUMPSIDE(0, ns_inuse);
8015 		}
8016 		nat->nat_phnext[0] = natp;
8017 		nat->nat_hnext[0] = *natp;
8018 		*natp = nat;
8019 		NBUMPSIDE(0, ns_bucketlen[hv]);
8020 
8021 		nat->nat_hnext[1] = NULL;
8022 		nat->nat_phnext[1] = NULL;
8023 		hv = nat->nat_hv[1] % softn->ipf_nat_table_sz;
8024 
8025 		natp = &softn->ipf_nat_table[1][hv];
8026 		if (*natp) {
8027 			(*natp)->nat_phnext[1] = &nat->nat_hnext[1];
8028 		} else {
8029 			NBUMPSIDE(1, ns_inuse);
8030 		}
8031 		nat->nat_phnext[1] = natp;
8032 		nat->nat_hnext[1] = *natp;
8033 		*natp = nat;
8034 		NBUMPSIDE(1, ns_bucketlen[hv]);
8035 	}
8036 	RWLOCK_EXIT(&softc->ipf_nat);
8037 
8038 	return 0;
8039 
8040 badrehash:
8041 	if (bucketlens[1] != NULL) {
8042 		KFREES(bucketlens[0], newsize * sizeof(u_int));
8043 	}
8044 	if (bucketlens[0] != NULL) {
8045 		KFREES(bucketlens[0], newsize * sizeof(u_int));
8046 	}
8047 	if (newtab[0] != NULL) {
8048 		KFREES(newtab[0], newsize * sizeof(nat_t *));
8049 	}
8050 	if (newtab[1] != NULL) {
8051 		KFREES(newtab[1], newsize * sizeof(nat_t *));
8052 	}
8053 	IPFERROR(error);
8054 	return ENOMEM;
8055 }
8056 
8057 
8058 /* ------------------------------------------------------------------------ */
8059 /* Function:    ipf_nat_rehash_rules                                        */
8060 /* Returns:     int  - 0 = success, else failure			    */
8061 /* Parameters:  softc(I) - pointer to soft context main structure           */
8062 /*              t(I) - pointer to tunable                                   */
8063 /*              p(I) - pointer to new tuning data                           */
8064 /*                                                                          */
8065 /* All of the NAT rules hang off of a hash table that is searched with a    */
8066 /* hash on address after the netmask is applied.  There is a different table*/
8067 /* for both inbound rules (rdr) and outbound (map.)  The resizing will only */
8068 /* affect one of these two tables.                                          */
8069 /* ------------------------------------------------------------------------ */
8070 int
8071 ipf_nat_rehash_rules(ipf_main_softc_t *softc, ipftuneable_t *t, ipftuneval_t *p)
8072 {
8073 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8074 	ipnat_t **newtab, *np, ***old, **npp;
8075 	u_int newsize;
8076 	u_int mask;
8077 	u_int hv;
8078 
8079 	newsize = p->ipftu_int;
8080 	/*
8081 	 * In case there is nothing to do...
8082 	 */
8083 	if (newsize == *t->ipft_pint)
8084 		return 0;
8085 
8086 	/*
8087 	 * All inbound rules have the NAT_REDIRECT bit set in in_redir and
8088 	 * all outbound rules have either NAT_MAP or MAT_MAPBLK set.
8089 	 * This if statement allows for some more generic code to be below,
8090 	 * rather than two huge gobs of code that almost do the same thing.
8091 	 */
8092 	if (t->ipft_pint == &softn->ipf_nat_rdrrules_sz) {
8093 		old = &softn->ipf_nat_rdr_rules;
8094 		mask = NAT_REDIRECT;
8095 	} else {
8096 		old = &softn->ipf_nat_map_rules;
8097 		mask = NAT_MAP|NAT_MAPBLK;
8098 	}
8099 
8100 	KMALLOCS(newtab, ipnat_t **, newsize * sizeof(ipnat_t *));
8101 	if (newtab == NULL) {
8102 		IPFERROR(60067);
8103 		return ENOMEM;
8104 	}
8105 
8106 	bzero((char *)newtab, newsize * sizeof(ipnat_t *));
8107 
8108 	WRITE_ENTER(&softc->ipf_nat);
8109 
8110 	if (*old != NULL) {
8111 		KFREES(*old, *t->ipft_pint * sizeof(ipnat_t **));
8112 	}
8113 	*old = newtab;
8114 	*t->ipft_pint = newsize;
8115 
8116 	for (np = softn->ipf_nat_list; np != NULL; np = np->in_next) {
8117 		if ((np->in_redir & mask) == 0)
8118 			continue;
8119 
8120 		if (np->in_redir & NAT_REDIRECT) {
8121 			np->in_rnext = NULL;
8122 			hv = np->in_hv[0] % newsize;
8123 			for (npp = newtab + hv; *npp != NULL; )
8124 				npp = &(*npp)->in_rnext;
8125 			np->in_prnext = npp;
8126 			*npp = np;
8127 		}
8128 		if (np->in_redir & NAT_MAP) {
8129 			np->in_mnext = NULL;
8130 			hv = np->in_hv[1] % newsize;
8131 			for (npp = newtab + hv; *npp != NULL; )
8132 				npp = &(*npp)->in_mnext;
8133 			np->in_pmnext = npp;
8134 			*npp = np;
8135 		}
8136 
8137 	}
8138 	RWLOCK_EXIT(&softc->ipf_nat);
8139 
8140 	return 0;
8141 }
8142 
8143 
8144 /* ------------------------------------------------------------------------ */
8145 /* Function:    ipf_nat_hostmap_rehash                                      */
8146 /* Returns:     int  - 0 = success, else failure			    */
8147 /* Parameters:  softc(I) - pointer to soft context main structure           */
8148 /*              t(I) - pointer to tunable                                   */
8149 /*              p(I) - pointer to new tuning data                           */
8150 /*                                                                          */
8151 /* Allocate and populate a new hash table that will contain a reference to  */
8152 /* all of the active IP# translations currently in place.                   */
8153 /* ------------------------------------------------------------------------ */
8154 int
8155 ipf_nat_hostmap_rehash(ipf_main_softc_t *softc, ipftuneable_t *t,
8156     ipftuneval_t *p)
8157 {
8158 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8159 	hostmap_t *hm, **newtab;
8160 	u_int newsize;
8161 	u_int hv;
8162 
8163 	newsize = p->ipftu_int;
8164 	/*
8165 	 * In case there is nothing to do...
8166 	 */
8167 	if (newsize == *t->ipft_pint)
8168 		return 0;
8169 
8170 	KMALLOCS(newtab, hostmap_t **, newsize * sizeof(hostmap_t *));
8171 	if (newtab == NULL) {
8172 		IPFERROR(60068);
8173 		return ENOMEM;
8174 	}
8175 
8176 	bzero((char *)newtab, newsize * sizeof(hostmap_t *));
8177 
8178 	WRITE_ENTER(&softc->ipf_nat);
8179 	if (softn->ipf_hm_maptable != NULL) {
8180 		KFREES(softn->ipf_hm_maptable,
8181 		       softn->ipf_nat_hostmap_sz * sizeof(hostmap_t *));
8182 	}
8183 	softn->ipf_hm_maptable = newtab;
8184 	softn->ipf_nat_hostmap_sz = newsize;
8185 
8186 	for (hm = softn->ipf_hm_maplist; hm != NULL; hm = hm->hm_next) {
8187 		hv = hm->hm_hv % softn->ipf_nat_hostmap_sz;
8188 		hm->hm_hnext = softn->ipf_hm_maptable[hv];
8189 		hm->hm_phnext = softn->ipf_hm_maptable + hv;
8190 		if (softn->ipf_hm_maptable[hv] != NULL)
8191 			softn->ipf_hm_maptable[hv]->hm_phnext = &hm->hm_hnext;
8192 		softn->ipf_hm_maptable[hv] = hm;
8193 	}
8194 	RWLOCK_EXIT(&softc->ipf_nat);
8195 
8196 	return 0;
8197 }
8198 
8199 
8200 /* ------------------------------------------------------------------------ */
8201 /* Function:    ipf_nat_add_tq                                              */
8202 /* Parameters:  softc(I) - pointer to soft context main structure           */
8203 /*                                                                          */
8204 /* ------------------------------------------------------------------------ */
8205 ipftq_t *
8206 ipf_nat_add_tq(ipf_main_softc_t *softc, int ttl)
8207 {
8208 	ipf_nat_softc_t *softs = softc->ipf_nat_soft;
8209 
8210 	return ipf_addtimeoutqueue(softc, &softs->ipf_nat_utqe, ttl);
8211 }
8212 
8213 /* ------------------------------------------------------------------------ */
8214 /* Function:    ipf_nat_uncreate                                            */
8215 /* Returns:     Nil                                                         */
8216 /* Parameters:  fin(I) - pointer to packet information                      */
8217 /*                                                                          */
8218 /* This function is used to remove a NAT entry from the NAT table when we   */
8219 /* decide that the create was actually in error. It is thus assumed that    */
8220 /* fin_flx will have both FI_NATED and FI_NATNEW set. Because we're dealing */
8221 /* with the translated packet (not the original), we have to reverse the    */
8222 /* lookup. Although doing the lookup is expensive (relatively speaking), it */
8223 /* is not anticipated that this will be a frequent occurance for normal     */
8224 /* traffic patterns.                                                        */
8225 /* ------------------------------------------------------------------------ */
8226 void
8227 ipf_nat_uncreate(fr_info_t *fin)
8228 {
8229 	ipf_main_softc_t *softc = fin->fin_main_soft;
8230 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8231 	int nflags;
8232 	nat_t *nat;
8233 
8234 	switch (fin->fin_p)
8235 	{
8236 	case IPPROTO_TCP :
8237 		nflags = IPN_TCP;
8238 		break;
8239 	case IPPROTO_UDP :
8240 		nflags = IPN_UDP;
8241 		break;
8242 	default :
8243 		nflags = 0;
8244 		break;
8245 	}
8246 
8247 	WRITE_ENTER(&softc->ipf_nat);
8248 
8249 	if (fin->fin_out == 0) {
8250 		nat = ipf_nat_outlookup(fin, nflags, (u_int)fin->fin_p,
8251 					fin->fin_dst, fin->fin_src);
8252 	} else {
8253 		nat = ipf_nat_inlookup(fin, nflags, (u_int)fin->fin_p,
8254 				       fin->fin_src, fin->fin_dst);
8255 	}
8256 
8257 	if (nat != NULL) {
8258 		NBUMPSIDE(fin->fin_out, ns_uncreate[0]);
8259 		ipf_nat_delete(softc, nat, NL_DESTROY);
8260 	} else {
8261 		NBUMPSIDE(fin->fin_out, ns_uncreate[1]);
8262 	}
8263 
8264 	RWLOCK_EXIT(&softc->ipf_nat);
8265 }
8266 
8267 
8268 /* ------------------------------------------------------------------------ */
8269 /* Function:    ipf_nat_cmp_rules                                           */
8270 /* Returns:     int   - 0 == success, else rules do not match.              */
8271 /* Parameters:  n1(I) - first rule to compare                               */
8272 /*              n2(I) - first rule to compare                               */
8273 /*                                                                          */
8274 /* Compare two rules using pointers to each rule. A straight bcmp will not  */
8275 /* work as some fields (such as in_dst, in_pkts) actually do change once    */
8276 /* the rule has been loaded into the kernel. Whilst this function returns   */
8277 /* various non-zero returns, they're strictly to aid in debugging. Use of   */
8278 /* this function should simply care if the result is zero or not.           */
8279 /* ------------------------------------------------------------------------ */
8280 static int
8281 ipf_nat_cmp_rules(ipnat_t *n1, ipnat_t *n2)
8282 {
8283 	if (n1->in_size != n2->in_size)
8284 		return 1;
8285 
8286 	if (bcmp((char *)&n1->in_v, (char *)&n2->in_v,
8287 		 offsetof(ipnat_t, in_ndst) - offsetof(ipnat_t, in_v)) != 0)
8288 		return 2;
8289 
8290 	if (bcmp((char *)&n1->in_tuc, (char *)&n2->in_tuc,
8291 		 n1->in_size - offsetof(ipnat_t, in_tuc)) != 0)
8292 		return 3;
8293 	if (n1->in_ndst.na_atype != n2->in_ndst.na_atype)
8294 		return 5;
8295 	if (n1->in_ndst.na_function != n2->in_ndst.na_function)
8296 		return 6;
8297 	if (bcmp((char *)&n1->in_ndst.na_addr, (char *)&n2->in_ndst.na_addr,
8298 		 sizeof(n1->in_ndst.na_addr)))
8299 		return 7;
8300 	if (n1->in_nsrc.na_atype != n2->in_nsrc.na_atype)
8301 		return 8;
8302 	if (n1->in_nsrc.na_function != n2->in_nsrc.na_function)
8303 		return 9;
8304 	if (bcmp((char *)&n1->in_nsrc.na_addr, (char *)&n2->in_nsrc.na_addr,
8305 		 sizeof(n1->in_nsrc.na_addr)))
8306 		return 10;
8307 	if (n1->in_odst.na_atype != n2->in_odst.na_atype)
8308 		return 11;
8309 	if (n1->in_odst.na_function != n2->in_odst.na_function)
8310 		return 12;
8311 	if (bcmp((char *)&n1->in_odst.na_addr, (char *)&n2->in_odst.na_addr,
8312 		 sizeof(n1->in_odst.na_addr)))
8313 		return 13;
8314 	if (n1->in_osrc.na_atype != n2->in_osrc.na_atype)
8315 		return 14;
8316 	if (n1->in_osrc.na_function != n2->in_osrc.na_function)
8317 		return 15;
8318 	if (bcmp((char *)&n1->in_osrc.na_addr, (char *)&n2->in_osrc.na_addr,
8319 		 sizeof(n1->in_osrc.na_addr)))
8320 		return 16;
8321 	return 0;
8322 }
8323 
8324 
8325 /* ------------------------------------------------------------------------ */
8326 /* Function:    ipf_nat_rule_init                                           */
8327 /* Returns:     int   - 0 == success, else rules do not match.              */
8328 /* Parameters:  softc(I) - pointer to soft context main structure           */
8329 /*              softn(I) - pointer to NAT context structure                 */
8330 /*              n(I)     - first rule to compare                            */
8331 /*                                                                          */
8332 /* ------------------------------------------------------------------------ */
8333 static int
8334 ipf_nat_rule_init(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *n)
8335 {
8336 	int error = 0;
8337 
8338 	if ((n->in_flags & IPN_SIPRANGE) != 0)
8339 		n->in_nsrcatype = FRI_RANGE;
8340 
8341 	if ((n->in_flags & IPN_DIPRANGE) != 0)
8342 		n->in_ndstatype = FRI_RANGE;
8343 
8344 	if ((n->in_flags & IPN_SPLIT) != 0)
8345 		n->in_ndstatype = FRI_SPLIT;
8346 
8347 	if ((n->in_redir & (NAT_MAP|NAT_REWRITE|NAT_DIVERTUDP)) != 0)
8348 		n->in_spnext = n->in_spmin;
8349 
8350 	if ((n->in_redir & (NAT_REWRITE|NAT_DIVERTUDP)) != 0) {
8351 		n->in_dpnext = n->in_dpmin;
8352 	} else if (n->in_redir == NAT_REDIRECT) {
8353 		n->in_dpnext = n->in_dpmin;
8354 	}
8355 
8356 	n->in_stepnext = 0;
8357 
8358 	switch (n->in_v[0])
8359 	{
8360 	case 4 :
8361 		error = ipf_nat_ruleaddrinit(softc, softn, n);
8362 		if (error != 0)
8363 			return error;
8364 		break;
8365 #ifdef USE_INET6
8366 	case 6 :
8367 		error = ipf_nat6_ruleaddrinit(softc, softn, n);
8368 		if (error != 0)
8369 			return error;
8370 		break;
8371 #endif
8372 	default :
8373 		break;
8374 	}
8375 
8376 	if (n->in_redir == (NAT_DIVERTUDP|NAT_MAP)) {
8377 		/*
8378 		 * Prerecord whether or not the destination of the divert
8379 		 * is local or not to the interface the packet is going
8380 		 * to be sent out.
8381 		 */
8382 		n->in_dlocal = ipf_deliverlocal(softc, n->in_v[1],
8383 						n->in_ifps[1], &n->in_ndstip6);
8384 	}
8385 
8386 	return error;
8387 }
8388 
8389 
8390 /* ------------------------------------------------------------------------ */
8391 /* Function:    ipf_nat_rule_fini                                           */
8392 /* Returns:     int   - 0 == success, else rules do not match.              */
8393 /* Parameters:  softc(I) - pointer to soft context main structure           */
8394 /*              n(I)     - rule to work on                                  */
8395 /*                                                                          */
8396 /* This function is used to release any objects that were referenced during */
8397 /* the rule initialisation. This is useful both when free'ing the rule and  */
8398 /* when handling ioctls that need to initialise these fields but not        */
8399 /* actually use them after the ioctl processing has finished.               */
8400 /* ------------------------------------------------------------------------ */
8401 static void
8402 ipf_nat_rule_fini(ipf_main_softc_t *softc, ipnat_t *n)
8403 {
8404 	if (n->in_odst.na_atype == FRI_LOOKUP && n->in_odst.na_ptr != NULL)
8405 		ipf_lookup_deref(softc, n->in_odst.na_type, n->in_odst.na_ptr);
8406 
8407 	if (n->in_osrc.na_atype == FRI_LOOKUP && n->in_osrc.na_ptr != NULL)
8408 		ipf_lookup_deref(softc, n->in_osrc.na_type, n->in_osrc.na_ptr);
8409 
8410 	if (n->in_ndst.na_atype == FRI_LOOKUP && n->in_ndst.na_ptr != NULL)
8411 		ipf_lookup_deref(softc, n->in_ndst.na_type, n->in_ndst.na_ptr);
8412 
8413 	if (n->in_nsrc.na_atype == FRI_LOOKUP && n->in_nsrc.na_ptr != NULL)
8414 		ipf_lookup_deref(softc, n->in_nsrc.na_type, n->in_nsrc.na_ptr);
8415 
8416 	if (n->in_divmp != NULL)
8417 		FREE_MB_T(n->in_divmp);
8418 }
8419