xref: /titanic_51/usr/src/uts/common/inet/ipf/ip_nat.c (revision 6f45ec7b0b964c3be967c4880e8867ac1e7763a5)
1 /*
2  * Copyright (C) 1995-2003 by Darren Reed.
3  *
4  * See the IPFILTER.LICENCE file for details on licencing.
5  *
6  * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
7  * Use is subject to license terms.
8  */
9 
10 #pragma ident	"%Z%%M%	%I%	%E% SMI"
11 
12 #if defined(KERNEL) || defined(_KERNEL)
13 # undef KERNEL
14 # undef _KERNEL
15 # define        KERNEL	1
16 # define        _KERNEL	1
17 #endif
18 #include <sys/errno.h>
19 #include <sys/types.h>
20 #include <sys/param.h>
21 #include <sys/time.h>
22 #include <sys/file.h>
23 #if defined(__NetBSD__) && (NetBSD >= 199905) && !defined(IPFILTER_LKM) && \
24     defined(_KERNEL)
25 # include "opt_ipfilter_log.h"
26 #endif
27 #if !defined(_KERNEL)
28 # include <stdio.h>
29 # include <string.h>
30 # include <stdlib.h>
31 # define _KERNEL
32 # ifdef __OpenBSD__
33 struct file;
34 # endif
35 # include <sys/uio.h>
36 # undef _KERNEL
37 #endif
38 #if defined(_KERNEL) && (__FreeBSD_version >= 220000)
39 # include <sys/filio.h>
40 # include <sys/fcntl.h>
41 #else
42 # include <sys/ioctl.h>
43 #endif
44 #if !defined(AIX)
45 # include <sys/fcntl.h>
46 #endif
47 #if !defined(linux)
48 # include <sys/protosw.h>
49 #endif
50 #include <sys/socket.h>
51 #if defined(_KERNEL)
52 # include <sys/systm.h>
53 # if !defined(__SVR4) && !defined(__svr4__)
54 #  include <sys/mbuf.h>
55 # endif
56 #endif
57 #if defined(__SVR4) || defined(__svr4__)
58 # include <sys/filio.h>
59 # include <sys/byteorder.h>
60 # ifdef _KERNEL
61 #  include <sys/dditypes.h>
62 # endif
63 # include <sys/stream.h>
64 # include <sys/kmem.h>
65 #endif
66 #if __FreeBSD_version >= 300000
67 # include <sys/queue.h>
68 #endif
69 #include <net/if.h>
70 #if __FreeBSD_version >= 300000
71 # include <net/if_var.h>
72 # if defined(_KERNEL) && !defined(IPFILTER_LKM)
73 #  include "opt_ipfilter.h"
74 # endif
75 #endif
76 #ifdef sun
77 # include <net/af.h>
78 #endif
79 #include <net/route.h>
80 #include <netinet/in.h>
81 #include <netinet/in_systm.h>
82 #include <netinet/ip.h>
83 
84 #ifdef RFC1825
85 # include <vpn/md5.h>
86 # include <vpn/ipsec.h>
87 extern struct ifnet vpnif;
88 #endif
89 
90 #if !defined(linux)
91 # include <netinet/ip_var.h>
92 #endif
93 #include <netinet/tcp.h>
94 #include <netinet/udp.h>
95 #include <netinet/ip_icmp.h>
96 #include "netinet/ip_compat.h"
97 #include <netinet/tcpip.h>
98 #include "netinet/ip_fil.h"
99 #include "netinet/ip_nat.h"
100 #include "netinet/ip_frag.h"
101 #include "netinet/ip_state.h"
102 #include "netinet/ip_proxy.h"
103 #include "netinet/ipf_stack.h"
104 #ifdef	IPFILTER_SYNC
105 #include "netinet/ip_sync.h"
106 #endif
107 #if (__FreeBSD_version >= 300000)
108 # include <sys/malloc.h>
109 #endif
110 /* END OF INCLUDES */
111 
112 #undef	SOCKADDR_IN
113 #define	SOCKADDR_IN	struct sockaddr_in
114 
115 #if !defined(lint)
116 static const char sccsid[] = "@(#)ip_nat.c	1.11 6/5/96 (C) 1995 Darren Reed";
117 static const char rcsid[] = "@(#)$Id: ip_nat.c,v 2.195.2.42 2005/08/11 19:51:36 darrenr Exp $";
118 #endif
119 
120 
121 /* ======================================================================== */
122 /* How the NAT is organised and works.                                      */
123 /*                                                                          */
124 /* Inside (interface y) NAT       Outside (interface x)                     */
125 /* -------------------- -+- -------------------------------------           */
126 /* Packet going          |   out, processsed by fr_checknatout() for x      */
127 /* ------------>         |   ------------>                                  */
128 /* src=10.1.1.1          |   src=192.1.1.1                                  */
129 /*                       |                                                  */
130 /*                       |   in, processed by fr_checknatin() for x         */
131 /* <------------         |   <------------                                  */
132 /* dst=10.1.1.1          |   dst=192.1.1.1                                  */
133 /* -------------------- -+- -------------------------------------           */
134 /* fr_checknatout() - changes ip_src and if required, sport                 */
135 /*             - creates a new mapping, if required.                        */
136 /* fr_checknatin()  - changes ip_dst and if required, dport                 */
137 /*                                                                          */
138 /* In the NAT table, internal source is recorded as "in" and externally     */
139 /* seen as "out".                                                           */
140 /* ======================================================================== */
141 
142 
143 
144 static	int	nat_flushtable __P((ipf_stack_t *));
145 static	int	nat_clearlist __P((ipf_stack_t *));
146 static	void	nat_addnat __P((struct ipnat *, ipf_stack_t *));
147 static	void	nat_addrdr __P((struct ipnat *, ipf_stack_t *));
148 static	void	nat_delete __P((struct nat *, int, ipf_stack_t *));
149 static	void	nat_delrdr __P((struct ipnat *));
150 static	void	nat_delnat __P((struct ipnat *));
151 static	int	fr_natgetent __P((caddr_t, ipf_stack_t *));
152 static	int	fr_natgetsz __P((caddr_t, ipf_stack_t *));
153 static	int	fr_natputent __P((caddr_t, int, ipf_stack_t *));
154 static	void	nat_tabmove __P((nat_t *, ipf_stack_t *));
155 static	int	nat_match __P((fr_info_t *, ipnat_t *));
156 static	INLINE	int nat_newmap __P((fr_info_t *, nat_t *, natinfo_t *));
157 static	INLINE	int nat_newrdr __P((fr_info_t *, nat_t *, natinfo_t *));
158 static	hostmap_t *nat_hostmap __P((ipnat_t *, struct in_addr,
159 				    struct in_addr, struct in_addr, u_32_t,
160 				    ipf_stack_t *));
161 static	void	nat_hostmapdel __P((struct hostmap *));
162 static	INLINE	int nat_icmpquerytype4 __P((int));
163 static	int	nat_siocaddnat __P((ipnat_t *, ipnat_t **, int,
164 				    ipf_stack_t *));
165 static	void	nat_siocdelnat __P((ipnat_t *, ipnat_t **, int,
166 				    ipf_stack_t *));
167 static	INLINE	int nat_icmperrortype4 __P((int));
168 static	INLINE	int nat_finalise __P((fr_info_t *, nat_t *, natinfo_t *,
169 				      tcphdr_t *, nat_t **, int));
170 static	INLINE	void nat_resolverule __P((ipnat_t *, ipf_stack_t *));
171 static	nat_t	*fr_natclone __P((fr_info_t *, nat_t *));
172 static	void	nat_mssclamp __P((tcphdr_t *, u_32_t, u_short *));
173 static	INLINE	int nat_wildok __P((nat_t *, int, int, int, int));
174 static	int	nat_getnext __P((ipftoken_t *, ipfgeniter_t *, ipf_stack_t *));
175 static	int	nat_iterator __P((ipftoken_t *, ipfgeniter_t *, ipf_stack_t *));
176 
177 
178 /* ------------------------------------------------------------------------ */
179 /* Function:    fr_natinit                                                  */
180 /* Returns:     int - 0 == success, -1 == failure                           */
181 /* Parameters:  Nil                                                         */
182 /*                                                                          */
183 /* Initialise all of the NAT locks, tables and other structures.            */
184 /* ------------------------------------------------------------------------ */
185 int fr_natinit(ifs)
186 ipf_stack_t *ifs;
187 {
188 	int i;
189 
190 	ifs->ifs_ipf_nattable_sz = NAT_TABLE_SZ;
191 	ifs->ifs_ipf_nattable_max = NAT_TABLE_MAX;
192 	ifs->ifs_ipf_natrules_sz = NAT_SIZE;
193 	ifs->ifs_ipf_rdrrules_sz = RDR_SIZE;
194 	ifs->ifs_ipf_hostmap_sz = HOSTMAP_SIZE;
195 	ifs->ifs_fr_nat_maxbucket_reset = 1;
196 #ifdef  IPFILTER_LOG
197 	ifs->ifs_nat_logging = 1;
198 #else
199 	ifs->ifs_nat_logging = 0;
200 #endif
201 	ifs->ifs_fr_defnatage = DEF_NAT_AGE;
202 	ifs->ifs_fr_defnatipage = 120;		/* 60 seconds */
203 	ifs->ifs_fr_defnaticmpage = 6;		/* 3 seconds */
204 
205 	KMALLOCS(ifs->ifs_nat_table[0], nat_t **,
206 		 sizeof(nat_t *) * ifs->ifs_ipf_nattable_sz);
207 	if (ifs->ifs_nat_table[0] != NULL)
208 		bzero((char *)ifs->ifs_nat_table[0],
209 		      ifs->ifs_ipf_nattable_sz * sizeof(nat_t *));
210 	else
211 		return -1;
212 
213 	KMALLOCS(ifs->ifs_nat_table[1], nat_t **,
214 		 sizeof(nat_t *) * ifs->ifs_ipf_nattable_sz);
215 	if (ifs->ifs_nat_table[1] != NULL)
216 		bzero((char *)ifs->ifs_nat_table[1],
217 		      ifs->ifs_ipf_nattable_sz * sizeof(nat_t *));
218 	else
219 		return -2;
220 
221 	KMALLOCS(ifs->ifs_nat_rules, ipnat_t **,
222 		 sizeof(ipnat_t *) * ifs->ifs_ipf_natrules_sz);
223 	if (ifs->ifs_nat_rules != NULL)
224 		bzero((char *)ifs->ifs_nat_rules,
225 		      ifs->ifs_ipf_natrules_sz * sizeof(ipnat_t *));
226 	else
227 		return -3;
228 
229 	KMALLOCS(ifs->ifs_rdr_rules, ipnat_t **,
230 		 sizeof(ipnat_t *) * ifs->ifs_ipf_rdrrules_sz);
231 	if (ifs->ifs_rdr_rules != NULL)
232 		bzero((char *)ifs->ifs_rdr_rules,
233 		      ifs->ifs_ipf_rdrrules_sz * sizeof(ipnat_t *));
234 	else
235 		return -4;
236 
237 	KMALLOCS(ifs->ifs_maptable, hostmap_t **,
238 		 sizeof(hostmap_t *) * ifs->ifs_ipf_hostmap_sz);
239 	if (ifs->ifs_maptable != NULL)
240 		bzero((char *)ifs->ifs_maptable,
241 		      sizeof(hostmap_t *) * ifs->ifs_ipf_hostmap_sz);
242 	else
243 		return -5;
244 
245 	ifs->ifs_ipf_hm_maplist = NULL;
246 
247 	KMALLOCS(ifs->ifs_nat_stats.ns_bucketlen[0], u_long *,
248 		 ifs->ifs_ipf_nattable_sz * sizeof(u_long));
249 	if (ifs->ifs_nat_stats.ns_bucketlen[0] == NULL)
250 		return -1;
251 	bzero((char *)ifs->ifs_nat_stats.ns_bucketlen[0],
252 	      ifs->ifs_ipf_nattable_sz * sizeof(u_long));
253 
254 	KMALLOCS(ifs->ifs_nat_stats.ns_bucketlen[1], u_long *,
255 		 ifs->ifs_ipf_nattable_sz * sizeof(u_long));
256 	if (ifs->ifs_nat_stats.ns_bucketlen[1] == NULL)
257 		return -1;
258 	bzero((char *)ifs->ifs_nat_stats.ns_bucketlen[1],
259 	      ifs->ifs_ipf_nattable_sz * sizeof(u_long));
260 
261 	if (ifs->ifs_fr_nat_maxbucket == 0) {
262 		for (i = ifs->ifs_ipf_nattable_sz; i > 0; i >>= 1)
263 			ifs->ifs_fr_nat_maxbucket++;
264 		ifs->ifs_fr_nat_maxbucket *= 2;
265 	}
266 
267 	fr_sttab_init(ifs->ifs_nat_tqb, ifs);
268 	/*
269 	 * Increase this because we may have "keep state" following this too
270 	 * and packet storms can occur if this is removed too quickly.
271 	 */
272 	ifs->ifs_nat_tqb[IPF_TCPS_CLOSED].ifq_ttl = ifs->ifs_fr_tcplastack;
273 	ifs->ifs_nat_tqb[IPF_TCP_NSTATES - 1].ifq_next = &ifs->ifs_nat_udptq;
274 	ifs->ifs_nat_udptq.ifq_ttl = ifs->ifs_fr_defnatage;
275 	ifs->ifs_nat_udptq.ifq_ref = 1;
276 	ifs->ifs_nat_udptq.ifq_head = NULL;
277 	ifs->ifs_nat_udptq.ifq_tail = &ifs->ifs_nat_udptq.ifq_head;
278 	MUTEX_INIT(&ifs->ifs_nat_udptq.ifq_lock, "nat ipftq udp tab");
279 	ifs->ifs_nat_udptq.ifq_next = &ifs->ifs_nat_icmptq;
280 	ifs->ifs_nat_icmptq.ifq_ttl = ifs->ifs_fr_defnaticmpage;
281 	ifs->ifs_nat_icmptq.ifq_ref = 1;
282 	ifs->ifs_nat_icmptq.ifq_head = NULL;
283 	ifs->ifs_nat_icmptq.ifq_tail = &ifs->ifs_nat_icmptq.ifq_head;
284 	MUTEX_INIT(&ifs->ifs_nat_icmptq.ifq_lock, "nat icmp ipftq tab");
285 	ifs->ifs_nat_icmptq.ifq_next = &ifs->ifs_nat_iptq;
286 	ifs->ifs_nat_iptq.ifq_ttl = ifs->ifs_fr_defnatipage;
287 	ifs->ifs_nat_iptq.ifq_ref = 1;
288 	ifs->ifs_nat_iptq.ifq_head = NULL;
289 	ifs->ifs_nat_iptq.ifq_tail = &ifs->ifs_nat_iptq.ifq_head;
290 	MUTEX_INIT(&ifs->ifs_nat_iptq.ifq_lock, "nat ip ipftq tab");
291 	ifs->ifs_nat_iptq.ifq_next = NULL;
292 
293 	for (i = 0; i < IPF_TCP_NSTATES; i++) {
294 		if (ifs->ifs_nat_tqb[i].ifq_ttl < ifs->ifs_fr_defnaticmpage)
295 			ifs->ifs_nat_tqb[i].ifq_ttl = ifs->ifs_fr_defnaticmpage;
296 #ifdef LARGE_NAT
297 		else if (ifs->ifs_nat_tqb[i].ifq_ttl > ifs->ifs_fr_defnatage)
298 			ifs->ifs_nat_tqb[i].ifq_ttl = ifs->ifs_fr_defnatage;
299 #endif
300 	}
301 
302 	/*
303 	 * Increase this because we may have "keep state" following
304 	 * this too and packet storms can occur if this is removed
305 	 * too quickly.
306 	 */
307 	ifs->ifs_nat_tqb[IPF_TCPS_CLOSED].ifq_ttl =
308 	    ifs->ifs_nat_tqb[IPF_TCPS_LAST_ACK].ifq_ttl;
309 
310 	RWLOCK_INIT(&ifs->ifs_ipf_nat, "ipf IP NAT rwlock");
311 	RWLOCK_INIT(&ifs->ifs_ipf_natfrag, "ipf IP NAT-Frag rwlock");
312 	MUTEX_INIT(&ifs->ifs_ipf_nat_new, "ipf nat new mutex");
313 	MUTEX_INIT(&ifs->ifs_ipf_natio, "ipf nat io mutex");
314 
315 	ifs->ifs_fr_nat_init = 1;
316 
317 	return 0;
318 }
319 
320 
321 /* ------------------------------------------------------------------------ */
322 /* Function:    nat_addrdr                                                  */
323 /* Returns:     Nil                                                         */
324 /* Parameters:  n(I) - pointer to NAT rule to add                           */
325 /*                                                                          */
326 /* Adds a redirect rule to the hash table of redirect rules and the list of */
327 /* loaded NAT rules.  Updates the bitmask indicating which netmasks are in  */
328 /* use by redirect rules.                                                   */
329 /* ------------------------------------------------------------------------ */
330 static void nat_addrdr(n, ifs)
331 ipnat_t *n;
332 ipf_stack_t *ifs;
333 {
334 	ipnat_t **np;
335 	u_32_t j;
336 	u_int hv;
337 	int k;
338 
339 	k = count4bits(n->in_outmsk);
340 	if ((k >= 0) && (k != 32))
341 		ifs->ifs_rdr_masks |= 1 << k;
342 	j = (n->in_outip & n->in_outmsk);
343 	hv = NAT_HASH_FN(j, 0, ifs->ifs_ipf_rdrrules_sz);
344 	np = ifs->ifs_rdr_rules + hv;
345 	while (*np != NULL)
346 		np = &(*np)->in_rnext;
347 	n->in_rnext = NULL;
348 	n->in_prnext = np;
349 	n->in_hv = hv;
350 	*np = n;
351 }
352 
353 
354 /* ------------------------------------------------------------------------ */
355 /* Function:    nat_addnat                                                  */
356 /* Returns:     Nil                                                         */
357 /* Parameters:  n(I) - pointer to NAT rule to add                           */
358 /*                                                                          */
359 /* Adds a NAT map rule to the hash table of rules and the list of  loaded   */
360 /* NAT rules.  Updates the bitmask indicating which netmasks are in use by  */
361 /* redirect rules.                                                          */
362 /* ------------------------------------------------------------------------ */
363 static void nat_addnat(n, ifs)
364 ipnat_t *n;
365 ipf_stack_t *ifs;
366 {
367 	ipnat_t **np;
368 	u_32_t j;
369 	u_int hv;
370 	int k;
371 
372 	k = count4bits(n->in_inmsk);
373 	if ((k >= 0) && (k != 32))
374 		ifs->ifs_nat_masks |= 1 << k;
375 	j = (n->in_inip & n->in_inmsk);
376 	hv = NAT_HASH_FN(j, 0, ifs->ifs_ipf_natrules_sz);
377 	np = ifs->ifs_nat_rules + hv;
378 	while (*np != NULL)
379 		np = &(*np)->in_mnext;
380 	n->in_mnext = NULL;
381 	n->in_pmnext = np;
382 	n->in_hv = hv;
383 	*np = n;
384 }
385 
386 
387 /* ------------------------------------------------------------------------ */
388 /* Function:    nat_delrdr                                                  */
389 /* Returns:     Nil                                                         */
390 /* Parameters:  n(I) - pointer to NAT rule to delete                        */
391 /*                                                                          */
392 /* Removes a redirect rule from the hash table of redirect rules.           */
393 /* ------------------------------------------------------------------------ */
394 static void nat_delrdr(n)
395 ipnat_t *n;
396 {
397 	if (n->in_rnext)
398 		n->in_rnext->in_prnext = n->in_prnext;
399 	*n->in_prnext = n->in_rnext;
400 }
401 
402 
403 /* ------------------------------------------------------------------------ */
404 /* Function:    nat_delnat                                                  */
405 /* Returns:     Nil                                                         */
406 /* Parameters:  n(I) - pointer to NAT rule to delete                        */
407 /*                                                                          */
408 /* Removes a NAT map rule from the hash table of NAT map rules.             */
409 /* ------------------------------------------------------------------------ */
410 static void nat_delnat(n)
411 ipnat_t *n;
412 {
413 	if (n->in_mnext != NULL)
414 		n->in_mnext->in_pmnext = n->in_pmnext;
415 	*n->in_pmnext = n->in_mnext;
416 }
417 
418 
419 /* ------------------------------------------------------------------------ */
420 /* Function:    nat_hostmap                                                 */
421 /* Returns:     struct hostmap* - NULL if no hostmap could be created,      */
422 /*                                else a pointer to the hostmapping to use  */
423 /* Parameters:  np(I)   - pointer to NAT rule                               */
424 /*              real(I) - real IP address                                   */
425 /*              map(I)  - mapped IP address                                 */
426 /*              port(I) - destination port number                           */
427 /* Write Locks: ipf_nat                                                     */
428 /*                                                                          */
429 /* Check if an ip address has already been allocated for a given mapping    */
430 /* that is not doing port based translation.  If is not yet allocated, then */
431 /* create a new entry if a non-NULL NAT rule pointer has been supplied.     */
432 /* ------------------------------------------------------------------------ */
433 static struct hostmap *nat_hostmap(np, src, dst, map, port, ifs)
434 ipnat_t *np;
435 struct in_addr src;
436 struct in_addr dst;
437 struct in_addr map;
438 u_32_t port;
439 ipf_stack_t *ifs;
440 {
441 	hostmap_t *hm;
442 	u_int hv;
443 
444 	hv = (src.s_addr ^ dst.s_addr);
445 	hv += src.s_addr;
446 	hv += dst.s_addr;
447 	hv %= HOSTMAP_SIZE;
448 	for (hm = ifs->ifs_maptable[hv]; hm; hm = hm->hm_next)
449 		if ((hm->hm_srcip.s_addr == src.s_addr) &&
450 		    (hm->hm_dstip.s_addr == dst.s_addr) &&
451 		    ((np == NULL) || (np == hm->hm_ipnat)) &&
452 		    ((port == 0) || (port == hm->hm_port))) {
453 			hm->hm_ref++;
454 			return hm;
455 		}
456 
457 	if (np == NULL)
458 		return NULL;
459 
460 	KMALLOC(hm, hostmap_t *);
461 	if (hm) {
462 		hm->hm_hnext = ifs->ifs_ipf_hm_maplist;
463 		hm->hm_phnext = &ifs->ifs_ipf_hm_maplist;
464 		if (ifs->ifs_ipf_hm_maplist != NULL)
465 			ifs->ifs_ipf_hm_maplist->hm_phnext = &hm->hm_hnext;
466 		ifs->ifs_ipf_hm_maplist = hm;
467 
468 		hm->hm_next = ifs->ifs_maptable[hv];
469 		hm->hm_pnext = ifs->ifs_maptable + hv;
470 		if (ifs->ifs_maptable[hv] != NULL)
471 			ifs->ifs_maptable[hv]->hm_pnext = &hm->hm_next;
472 		ifs->ifs_maptable[hv] = hm;
473 		hm->hm_ipnat = np;
474 		hm->hm_srcip = src;
475 		hm->hm_dstip = dst;
476 		hm->hm_mapip = map;
477 		hm->hm_ref = 1;
478 		hm->hm_port = port;
479 	}
480 	return hm;
481 }
482 
483 
484 /* ------------------------------------------------------------------------ */
485 /* Function:    nat_hostmapdel                                              */
486 /* Returns:     Nil                                                         */
487 /* Parameters:  hm(I) - pointer to hostmap structure                        */
488 /* Write Locks: ipf_nat                                                     */
489 /*                                                                          */
490 /* Decrement the references to this hostmap structure by one.  If this      */
491 /* reaches zero then remove it and free it.                                 */
492 /* ------------------------------------------------------------------------ */
493 static void nat_hostmapdel(hm)
494 struct hostmap *hm;
495 {
496 	hm->hm_ref--;
497 	if (hm->hm_ref == 0) {
498 		if (hm->hm_next)
499 			hm->hm_next->hm_pnext = hm->hm_pnext;
500 		*hm->hm_pnext = hm->hm_next;
501 		if (hm->hm_hnext)
502 			hm->hm_hnext->hm_phnext = hm->hm_phnext;
503 		*hm->hm_phnext = hm->hm_hnext;
504 		KFREE(hm);
505 	}
506 }
507 
508 void fr_hostmapderef(hmp)
509 struct hostmap **hmp;
510 {
511 	struct hostmap *hm;
512 
513 	hm = *hmp;
514 	*hmp = NULL;
515 	hm->hm_ref--;
516 	if (hm->hm_ref == 0)
517 		nat_hostmapdel(hm);
518 }
519 
520 
521 /* ------------------------------------------------------------------------ */
522 /* Function:    fix_outcksum                                                */
523 /* Returns:     Nil                                                         */
524 /* Parameters:  sp(I)  - location of 16bit checksum to update               */
525 /*              n((I)  - amount to adjust checksum by                       */
526 /*                                                                          */
527 /* Adjusts the 16bit checksum by "n" for packets going out.                 */
528 /* ------------------------------------------------------------------------ */
529 void fix_outcksum(sp, n)
530 u_short *sp;
531 u_32_t n;
532 {
533 	u_short sumshort;
534 	u_32_t sum1;
535 
536 	if (n == 0)
537 		return;
538 
539 	sum1 = (~ntohs(*sp)) & 0xffff;
540 	sum1 += (n);
541 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
542 	/* Again */
543 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
544 	sumshort = ~(u_short)sum1;
545 	*(sp) = htons(sumshort);
546 }
547 
548 
549 /* ------------------------------------------------------------------------ */
550 /* Function:    fix_incksum                                                 */
551 /* Returns:     Nil                                                         */
552 /* Parameters:  sp(I)  - location of 16bit checksum to update               */
553 /*              n((I)  - amount to adjust checksum by                       */
554 /*                                                                          */
555 /* Adjusts the 16bit checksum by "n" for packets going in.                  */
556 /* ------------------------------------------------------------------------ */
557 void fix_incksum(sp, n)
558 u_short *sp;
559 u_32_t n;
560 {
561 	u_short sumshort;
562 	u_32_t sum1;
563 
564 	if (n == 0)
565 		return;
566 
567 	sum1 = (~ntohs(*sp)) & 0xffff;
568 	sum1 += ~(n) & 0xffff;
569 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
570 	/* Again */
571 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
572 	sumshort = ~(u_short)sum1;
573 	*(sp) = htons(sumshort);
574 }
575 
576 
577 /* ------------------------------------------------------------------------ */
578 /* Function:    fix_datacksum                                               */
579 /* Returns:     Nil                                                         */
580 /* Parameters:  sp(I)  - location of 16bit checksum to update               */
581 /*              n((I)  - amount to adjust checksum by                       */
582 /*                                                                          */
583 /* Fix_datacksum is used *only* for the adjustments of checksums in the     */
584 /* data section of an IP packet.                                            */
585 /*                                                                          */
586 /* The only situation in which you need to do this is when NAT'ing an       */
587 /* ICMP error message. Such a message, contains in its body the IP header   */
588 /* of the original IP packet, that causes the error.                        */
589 /*                                                                          */
590 /* You can't use fix_incksum or fix_outcksum in that case, because for the  */
591 /* kernel the data section of the ICMP error is just data, and no special   */
592 /* processing like hardware cksum or ntohs processing have been done by the */
593 /* kernel on the data section.                                              */
594 /* ------------------------------------------------------------------------ */
595 void fix_datacksum(sp, n)
596 u_short *sp;
597 u_32_t n;
598 {
599 	u_short sumshort;
600 	u_32_t sum1;
601 
602 	if (n == 0)
603 		return;
604 
605 	sum1 = (~ntohs(*sp)) & 0xffff;
606 	sum1 += (n);
607 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
608 	/* Again */
609 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
610 	sumshort = ~(u_short)sum1;
611 	*(sp) = htons(sumshort);
612 }
613 
614 
615 /* ------------------------------------------------------------------------ */
616 /* Function:    fr_nat_ioctl                                                */
617 /* Returns:     int - 0 == success, != 0 == failure                         */
618 /* Parameters:  data(I) - pointer to ioctl data                             */
619 /*              cmd(I)  - ioctl command integer                             */
620 /*              mode(I) - file mode bits used with open                     */
621 /*                                                                          */
622 /* Processes an ioctl call made to operate on the IP Filter NAT device.     */
623 /* ------------------------------------------------------------------------ */
624 int fr_nat_ioctl(data, cmd, mode, uid, ctx, ifs)
625 ioctlcmd_t cmd;
626 caddr_t data;
627 int mode, uid;
628 void *ctx;
629 ipf_stack_t *ifs;
630 {
631 	ipnat_t *nat, *nt, *n = NULL, **np = NULL;
632 	int error = 0, ret, arg, getlock;
633 	ipnat_t natd;
634 
635 #if (BSD >= 199306) && defined(_KERNEL)
636 	if ((securelevel >= 2) && (mode & FWRITE))
637 		return EPERM;
638 #endif
639 
640 #if defined(__osf__) && defined(_KERNEL)
641 	getlock = 0;
642 #else
643 	getlock = (mode & NAT_LOCKHELD) ? 0 : 1;
644 #endif
645 
646 	nat = NULL;     /* XXX gcc -Wuninitialized */
647 	if (cmd == (ioctlcmd_t)SIOCADNAT) {
648 		KMALLOC(nt, ipnat_t *);
649 	} else {
650 		nt = NULL;
651 	}
652 
653 	if ((cmd == (ioctlcmd_t)SIOCADNAT) || (cmd == (ioctlcmd_t)SIOCRMNAT)) {
654 		if (mode & NAT_SYSSPACE) {
655 			bcopy(data, (char *)&natd, sizeof(natd));
656 			error = 0;
657 		} else {
658 			error = fr_inobj(data, &natd, IPFOBJ_IPNAT);
659 		}
660 
661 	} else if (cmd == (ioctlcmd_t)SIOCIPFFL) { /* SIOCFLNAT & SIOCCNATL */
662 		BCOPYIN(data, &arg, sizeof(arg));
663 	}
664 
665 	if (error != 0)
666 		goto done;
667 
668 	/*
669 	 * For add/delete, look to see if the NAT entry is already present
670 	 */
671 	if ((cmd == (ioctlcmd_t)SIOCADNAT) || (cmd == (ioctlcmd_t)SIOCRMNAT)) {
672 		nat = &natd;
673 		if (nat->in_v == 0)	/* For backward compat. */
674 			nat->in_v = 4;
675 		nat->in_flags &= IPN_USERFLAGS;
676 		if ((nat->in_redir & NAT_MAPBLK) == 0) {
677 			if ((nat->in_flags & IPN_SPLIT) == 0)
678 				nat->in_inip &= nat->in_inmsk;
679 			if ((nat->in_flags & IPN_IPRANGE) == 0)
680 				nat->in_outip &= nat->in_outmsk;
681 		}
682 		MUTEX_ENTER(&ifs->ifs_ipf_natio);
683 		for (np = &ifs->ifs_nat_list; ((n = *np) != NULL);
684 		     np = &n->in_next)
685 			if (!bcmp((char *)&nat->in_flags, (char *)&n->in_flags,
686 					IPN_CMPSIZ))
687 				break;
688 	}
689 
690 	switch (cmd)
691 	{
692 	case SIOCGENITER :
693 	    {
694 		ipfgeniter_t iter;
695 		ipftoken_t *token;
696 
697 		error = fr_inobj(data, &iter, IPFOBJ_GENITER);
698 		if (error != 0)
699 			break;
700 
701 		token = ipf_findtoken(iter.igi_type, uid, ctx, ifs);
702 		if (token != NULL)
703 			error  = nat_iterator(token, &iter, ifs);
704 		else
705 			error = ESRCH;
706 		RWLOCK_EXIT(&ifs->ifs_ipf_tokens);
707 		break;
708 	    }
709 #ifdef  IPFILTER_LOG
710 	case SIOCIPFFB :
711 	{
712 		int tmp;
713 
714 		if (!(mode & FWRITE))
715 			error = EPERM;
716 		else {
717 			tmp = ipflog_clear(IPL_LOGNAT, ifs);
718 			BCOPYOUT((char *)&tmp, (char *)data, sizeof(tmp));
719 		}
720 		break;
721 	}
722 	case SIOCSETLG :
723 		if (!(mode & FWRITE))
724 			error = EPERM;
725 		else {
726 			BCOPYIN((char *)data,
727 				       (char *)&ifs->ifs_nat_logging,
728 				sizeof(ifs->ifs_nat_logging));
729 		}
730 		break;
731 	case SIOCGETLG :
732 		BCOPYOUT((char *)&ifs->ifs_nat_logging, (char *)data,
733 			sizeof(ifs->ifs_nat_logging));
734 		break;
735 	case FIONREAD :
736 		arg = ifs->ifs_iplused[IPL_LOGNAT];
737 		BCOPYOUT(&arg, data, sizeof(arg));
738 		break;
739 #endif
740 	case SIOCADNAT :
741 		if (!(mode & FWRITE)) {
742 			error = EPERM;
743 		} else if (n != NULL) {
744 			error = EEXIST;
745 		} else if (nt == NULL) {
746 			error = ENOMEM;
747 		}
748 		if (error != 0) {
749 			MUTEX_EXIT(&ifs->ifs_ipf_natio);
750 			break;
751 		}
752 		bcopy((char *)nat, (char *)nt, sizeof(*n));
753 		error = nat_siocaddnat(nt, np, getlock, ifs);
754 		MUTEX_EXIT(&ifs->ifs_ipf_natio);
755 		if (error == 0)
756 			nt = NULL;
757 		break;
758 	case SIOCRMNAT :
759 		if (!(mode & FWRITE)) {
760 			error = EPERM;
761 			n = NULL;
762 		} else if (n == NULL) {
763 			error = ESRCH;
764 		}
765 
766 		if (error != 0) {
767 			MUTEX_EXIT(&ifs->ifs_ipf_natio);
768 			break;
769 		}
770 		nat_siocdelnat(n, np, getlock, ifs);
771 
772 		MUTEX_EXIT(&ifs->ifs_ipf_natio);
773 		n = NULL;
774 		break;
775 	case SIOCGNATS :
776 		ifs->ifs_nat_stats.ns_table[0] = ifs->ifs_nat_table[0];
777 		ifs->ifs_nat_stats.ns_table[1] = ifs->ifs_nat_table[1];
778 		ifs->ifs_nat_stats.ns_list = ifs->ifs_nat_list;
779 		ifs->ifs_nat_stats.ns_maptable = ifs->ifs_maptable;
780 		ifs->ifs_nat_stats.ns_maplist = ifs->ifs_ipf_hm_maplist;
781 		ifs->ifs_nat_stats.ns_nattab_max = ifs->ifs_ipf_nattable_max;
782 		ifs->ifs_nat_stats.ns_nattab_sz = ifs->ifs_ipf_nattable_sz;
783 		ifs->ifs_nat_stats.ns_rultab_sz = ifs->ifs_ipf_natrules_sz;
784 		ifs->ifs_nat_stats.ns_rdrtab_sz = ifs->ifs_ipf_rdrrules_sz;
785 		ifs->ifs_nat_stats.ns_hostmap_sz = ifs->ifs_ipf_hostmap_sz;
786 		ifs->ifs_nat_stats.ns_instances = ifs->ifs_nat_instances;
787 		ifs->ifs_nat_stats.ns_apslist = ifs->ifs_ap_sess_list;
788 		error = fr_outobj(data, &ifs->ifs_nat_stats, IPFOBJ_NATSTAT);
789 		break;
790 	case SIOCGNATL :
791 	    {
792 		natlookup_t nl;
793 
794 		if (getlock) {
795 			READ_ENTER(&ifs->ifs_ipf_nat);
796 		}
797 		error = fr_inobj(data, &nl, IPFOBJ_NATLOOKUP);
798 		if (error == 0) {
799 			if (nat_lookupredir(&nl, ifs) != NULL) {
800 				error = fr_outobj(data, &nl, IPFOBJ_NATLOOKUP);
801 			} else {
802 				error = ESRCH;
803 			}
804 		}
805 		if (getlock) {
806 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
807 		}
808 		break;
809 	    }
810 	case SIOCIPFFL :	/* old SIOCFLNAT & SIOCCNATL */
811 		if (!(mode & FWRITE)) {
812 			error = EPERM;
813 			break;
814 		}
815 		if (getlock) {
816 			WRITE_ENTER(&ifs->ifs_ipf_nat);
817 		}
818 		error = 0;
819 		if (arg == 0)
820 			ret = nat_flushtable(ifs);
821 		else if (arg == 1)
822 			ret = nat_clearlist(ifs);
823 		else
824 			error = EINVAL;
825 		if (getlock) {
826 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
827 		}
828 		if (error == 0) {
829 			BCOPYOUT(&ret, data, sizeof(ret));
830 		}
831 		break;
832 	case SIOCPROXY :
833 		error = appr_ioctl(data, cmd, mode, ifs);
834 		break;
835 	case SIOCSTLCK :
836 		if (!(mode & FWRITE)) {
837 			error = EPERM;
838 		} else {
839 			fr_lock(data, &ifs->ifs_fr_nat_lock);
840 		}
841 		break;
842 	case SIOCSTPUT :
843 		if ((mode & FWRITE) != 0) {
844 			error = fr_natputent(data, getlock, ifs);
845 		} else {
846 			error = EACCES;
847 		}
848 		break;
849 	case SIOCSTGSZ :
850 		if (ifs->ifs_fr_nat_lock) {
851 			if (getlock) {
852 				READ_ENTER(&ifs->ifs_ipf_nat);
853 			}
854 			error = fr_natgetsz(data, ifs);
855 			if (getlock) {
856 				RWLOCK_EXIT(&ifs->ifs_ipf_nat);
857 			}
858 		} else
859 			error = EACCES;
860 		break;
861 	case SIOCSTGET :
862 		if (ifs->ifs_fr_nat_lock) {
863 			if (getlock) {
864 				READ_ENTER(&ifs->ifs_ipf_nat);
865 			}
866 			error = fr_natgetent(data, ifs);
867 			if (getlock) {
868 				RWLOCK_EXIT(&ifs->ifs_ipf_nat);
869 			}
870 		} else
871 			error = EACCES;
872 		break;
873 	case SIOCIPFDELTOK :
874 		(void) BCOPYIN((caddr_t)data, (caddr_t)&arg, sizeof(arg));
875 		error = ipf_deltoken(arg, uid, ctx, ifs);
876 		break;
877 	default :
878 		error = EINVAL;
879 		break;
880 	}
881 done:
882 	if (nt)
883 		KFREE(nt);
884 	return error;
885 }
886 
887 
888 /* ------------------------------------------------------------------------ */
889 /* Function:    nat_siocaddnat                                              */
890 /* Returns:     int - 0 == success, != 0 == failure                         */
891 /* Parameters:  n(I)       - pointer to new NAT rule                        */
892 /*              np(I)      - pointer to where to insert new NAT rule        */
893 /*              getlock(I) - flag indicating if lock on ipf_nat is held     */
894 /* Mutex Locks: ipf_natio                                                   */
895 /*                                                                          */
896 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
897 /* from information passed to the kernel, then add it  to the appropriate   */
898 /* NAT rule table(s).                                                       */
899 /* ------------------------------------------------------------------------ */
900 static int nat_siocaddnat(n, np, getlock, ifs)
901 ipnat_t *n, **np;
902 int getlock;
903 ipf_stack_t *ifs;
904 {
905 	int error = 0, i, j;
906 
907 	nat_resolverule(n, ifs);
908 	if (n->in_plabel[0] != '\0') {
909 		if (n->in_apr == NULL)
910 			return ENOENT;
911 	}
912 
913 	if ((n->in_age[0] == 0) && (n->in_age[1] != 0))
914 		return EINVAL;
915 
916 	n->in_use = 0;
917 	if (n->in_redir & NAT_MAPBLK)
918 		n->in_space = USABLE_PORTS * ~ntohl(n->in_outmsk);
919 	else if (n->in_flags & IPN_AUTOPORTMAP)
920 		n->in_space = USABLE_PORTS * ~ntohl(n->in_inmsk);
921 	else if (n->in_flags & IPN_IPRANGE)
922 		n->in_space = ntohl(n->in_outmsk) - ntohl(n->in_outip);
923 	else if (n->in_flags & IPN_SPLIT)
924 		n->in_space = 2;
925 	else if (n->in_outmsk != 0)
926 		n->in_space = ~ntohl(n->in_outmsk);
927 	else
928 		n->in_space = 1;
929 
930 	/*
931 	 * Calculate the number of valid IP addresses in the output
932 	 * mapping range.  In all cases, the range is inclusive of
933 	 * the start and ending IP addresses.
934 	 * If to a CIDR address, lose 2: broadcast + network address
935 	 *                               (so subtract 1)
936 	 * If to a range, add one.
937 	 * If to a single IP address, set to 1.
938 	 */
939 	if (n->in_space) {
940 		if ((n->in_flags & IPN_IPRANGE) != 0)
941 			n->in_space += 1;
942 		else
943 			n->in_space -= 1;
944 	} else
945 		n->in_space = 1;
946 
947 	if ((n->in_outmsk != 0xffffffff) && (n->in_outmsk != 0) &&
948 	    ((n->in_flags & (IPN_IPRANGE|IPN_SPLIT)) == 0))
949 		n->in_nip = ntohl(n->in_outip) + 1;
950 	else if ((n->in_flags & IPN_SPLIT) &&
951 		 (n->in_redir & NAT_REDIRECT))
952 		n->in_nip = ntohl(n->in_inip);
953 	else
954 		n->in_nip = ntohl(n->in_outip);
955 	if (n->in_redir & NAT_MAP) {
956 		n->in_pnext = ntohs(n->in_pmin);
957 		/*
958 		 * Multiply by the number of ports made available.
959 		 */
960 		if (ntohs(n->in_pmax) >= ntohs(n->in_pmin)) {
961 			n->in_space *= (ntohs(n->in_pmax) -
962 					ntohs(n->in_pmin) + 1);
963 			/*
964 			 * Because two different sources can map to
965 			 * different destinations but use the same
966 			 * local IP#/port #.
967 			 * If the result is smaller than in_space, then
968 			 * we may have wrapped around 32bits.
969 			 */
970 			i = n->in_inmsk;
971 			if ((i != 0) && (i != 0xffffffff)) {
972 				j = n->in_space * (~ntohl(i) + 1);
973 				if (j >= n->in_space)
974 					n->in_space = j;
975 				else
976 					n->in_space = 0xffffffff;
977 			}
978 		}
979 		/*
980 		 * If no protocol is specified, multiple by 256 to allow for
981 		 * at least one IP:IP mapping per protocol.
982 		 */
983 		if ((n->in_flags & IPN_TCPUDPICMP) == 0) {
984 				j = n->in_space * 256;
985 				if (j >= n->in_space)
986 					n->in_space = j;
987 				else
988 					n->in_space = 0xffffffff;
989 		}
990 	}
991 
992 	/* Otherwise, these fields are preset */
993 
994 	if (getlock) {
995 		WRITE_ENTER(&ifs->ifs_ipf_nat);
996 	}
997 	n->in_next = NULL;
998 	*np = n;
999 
1000 	if (n->in_age[0] != 0)
1001 	    n->in_tqehead[0] = fr_addtimeoutqueue(&ifs->ifs_nat_utqe,
1002 						  n->in_age[0], ifs);
1003 
1004 	if (n->in_age[1] != 0)
1005 	    n->in_tqehead[1] = fr_addtimeoutqueue(&ifs->ifs_nat_utqe,
1006 						  n->in_age[1], ifs);
1007 
1008 	if (n->in_redir & NAT_REDIRECT) {
1009 		n->in_flags &= ~IPN_NOTDST;
1010 		nat_addrdr(n, ifs);
1011 	}
1012 	if (n->in_redir & (NAT_MAP|NAT_MAPBLK)) {
1013 		n->in_flags &= ~IPN_NOTSRC;
1014 		nat_addnat(n, ifs);
1015 	}
1016 	n = NULL;
1017 	ifs->ifs_nat_stats.ns_rules++;
1018 	if (getlock) {
1019 		RWLOCK_EXIT(&ifs->ifs_ipf_nat);			/* WRITE */
1020 	}
1021 
1022 	return error;
1023 }
1024 
1025 
1026 /* ------------------------------------------------------------------------ */
1027 /* Function:    nat_resolvrule                                              */
1028 /* Returns:     Nil                                                         */
1029 /* Parameters:  n(I)  - pointer to NAT rule                                 */
1030 /*                                                                          */
1031 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
1032 /* from information passed to the kernel, then add it  to the appropriate   */
1033 /* NAT rule table(s).                                                       */
1034 /* ------------------------------------------------------------------------ */
1035 static void nat_resolverule(n, ifs)
1036 ipnat_t *n;
1037 ipf_stack_t *ifs;
1038 {
1039 	n->in_ifnames[0][LIFNAMSIZ - 1] = '\0';
1040 	n->in_ifps[0] = fr_resolvenic(n->in_ifnames[0], 4, ifs);
1041 
1042 	n->in_ifnames[1][LIFNAMSIZ - 1] = '\0';
1043 	if (n->in_ifnames[1][0] == '\0') {
1044 		(void) strncpy(n->in_ifnames[1], n->in_ifnames[0], LIFNAMSIZ);
1045 		n->in_ifps[1] = n->in_ifps[0];
1046 	} else {
1047 		n->in_ifps[1] = fr_resolvenic(n->in_ifnames[0], 4, ifs);
1048 	}
1049 
1050 	if (n->in_plabel[0] != '\0') {
1051 		n->in_apr = appr_lookup(n->in_p, n->in_plabel, ifs);
1052 	}
1053 }
1054 
1055 
1056 /* ------------------------------------------------------------------------ */
1057 /* Function:    nat_siocdelnat                                              */
1058 /* Returns:     int - 0 == success, != 0 == failure                         */
1059 /* Parameters:  n(I)       - pointer to new NAT rule                        */
1060 /*              np(I)      - pointer to where to insert new NAT rule        */
1061 /*              getlock(I) - flag indicating if lock on ipf_nat is held     */
1062 /* Mutex Locks: ipf_natio                                                   */
1063 /*                                                                          */
1064 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
1065 /* from information passed to the kernel, then add it  to the appropriate   */
1066 /* NAT rule table(s).                                                       */
1067 /* ------------------------------------------------------------------------ */
1068 static void nat_siocdelnat(n, np, getlock, ifs)
1069 ipnat_t *n, **np;
1070 int getlock;
1071 ipf_stack_t *ifs;
1072 {
1073 	if (getlock) {
1074 		WRITE_ENTER(&ifs->ifs_ipf_nat);
1075 	}
1076 	if (n->in_redir & NAT_REDIRECT)
1077 		nat_delrdr(n);
1078 	if (n->in_redir & (NAT_MAPBLK|NAT_MAP))
1079 		nat_delnat(n);
1080 	if (ifs->ifs_nat_list == NULL) {
1081 		ifs->ifs_nat_masks = 0;
1082 		ifs->ifs_rdr_masks = 0;
1083 	}
1084 
1085 	if (n->in_tqehead[0] != NULL) {
1086 		if (fr_deletetimeoutqueue(n->in_tqehead[0]) == 0) {
1087 			fr_freetimeoutqueue(n->in_tqehead[1], ifs);
1088 		}
1089 	}
1090 
1091 	if (n->in_tqehead[1] != NULL) {
1092 		if (fr_deletetimeoutqueue(n->in_tqehead[1]) == 0) {
1093 			fr_freetimeoutqueue(n->in_tqehead[1], ifs);
1094 		}
1095 	}
1096 
1097 	*np = n->in_next;
1098 
1099 	if (n->in_use == 0) {
1100 		if (n->in_apr)
1101 			appr_free(n->in_apr);
1102 		KFREE(n);
1103 		ifs->ifs_nat_stats.ns_rules--;
1104 	} else {
1105 		n->in_flags |= IPN_DELETE;
1106 		n->in_next = NULL;
1107 	}
1108 	if (getlock) {
1109 		RWLOCK_EXIT(&ifs->ifs_ipf_nat);			/* READ/WRITE */
1110 	}
1111 }
1112 
1113 
1114 /* ------------------------------------------------------------------------ */
1115 /* Function:    fr_natgetsz                                                 */
1116 /* Returns:     int - 0 == success, != 0 is the error value.                */
1117 /* Parameters:  data(I) - pointer to natget structure with kernel pointer   */
1118 /*                        get the size of.                                  */
1119 /*                                                                          */
1120 /* Handle SIOCSTGSZ.                                                        */
1121 /* Return the size of the nat list entry to be copied back to user space.   */
1122 /* The size of the entry is stored in the ng_sz field and the enture natget */
1123 /* structure is copied back to the user.                                    */
1124 /* ------------------------------------------------------------------------ */
1125 static int fr_natgetsz(data, ifs)
1126 caddr_t data;
1127 ipf_stack_t *ifs;
1128 {
1129 	ap_session_t *aps;
1130 	nat_t *nat, *n;
1131 	natget_t ng;
1132 
1133 	BCOPYIN(data, &ng, sizeof(ng));
1134 
1135 	nat = ng.ng_ptr;
1136 	if (!nat) {
1137 		nat = ifs->ifs_nat_instances;
1138 		ng.ng_sz = 0;
1139 		/*
1140 		 * Empty list so the size returned is 0.  Simple.
1141 		 */
1142 		if (nat == NULL) {
1143 			BCOPYOUT(&ng, data, sizeof(ng));
1144 			return 0;
1145 		}
1146 	} else {
1147 		/*
1148 		 * Make sure the pointer we're copying from exists in the
1149 		 * current list of entries.  Security precaution to prevent
1150 		 * copying of random kernel data.
1151 		 */
1152 		for (n = ifs->ifs_nat_instances; n; n = n->nat_next)
1153 			if (n == nat)
1154 				break;
1155 		if (!n)
1156 			return ESRCH;
1157 	}
1158 
1159 	/*
1160 	 * Incluse any space required for proxy data structures.
1161 	 */
1162 	ng.ng_sz = sizeof(nat_save_t);
1163 	aps = nat->nat_aps;
1164 	if (aps != NULL) {
1165 		ng.ng_sz += sizeof(ap_session_t) - 4;
1166 		if (aps->aps_data != 0)
1167 			ng.ng_sz += aps->aps_psiz;
1168 	}
1169 
1170 	BCOPYOUT(&ng, data, sizeof(ng));
1171 	return 0;
1172 }
1173 
1174 
1175 /* ------------------------------------------------------------------------ */
1176 /* Function:    fr_natgetent                                                */
1177 /* Returns:     int - 0 == success, != 0 is the error value.                */
1178 /* Parameters:  data(I) - pointer to natget structure with kernel pointer   */
1179 /*                        to NAT structure to copy out.                     */
1180 /*                                                                          */
1181 /* Handle SIOCSTGET.                                                        */
1182 /* Copies out NAT entry to user space.  Any additional data held for a      */
1183 /* proxy is also copied, as to is the NAT rule which was responsible for it */
1184 /* ------------------------------------------------------------------------ */
1185 static int fr_natgetent(data, ifs)
1186 caddr_t data;
1187 ipf_stack_t *ifs;
1188 {
1189 	int error, outsize;
1190 	ap_session_t *aps;
1191 	nat_save_t *ipn, ipns;
1192 	nat_t *n, *nat;
1193 
1194 	error = fr_inobj(data, &ipns, IPFOBJ_NATSAVE);
1195 	if (error != 0)
1196 		return error;
1197 
1198 	if ((ipns.ipn_dsize < sizeof(ipns)) || (ipns.ipn_dsize > 81920))
1199 		return EINVAL;
1200 
1201 	KMALLOCS(ipn, nat_save_t *, ipns.ipn_dsize);
1202 	if (ipn == NULL)
1203 		return ENOMEM;
1204 
1205 	ipn->ipn_dsize = ipns.ipn_dsize;
1206 	nat = ipns.ipn_next;
1207 	if (nat == NULL) {
1208 		nat = ifs->ifs_nat_instances;
1209 		if (nat == NULL) {
1210 			if (ifs->ifs_nat_instances == NULL)
1211 				error = ENOENT;
1212 			goto finished;
1213 		}
1214 	} else {
1215 		/*
1216 		 * Make sure the pointer we're copying from exists in the
1217 		 * current list of entries.  Security precaution to prevent
1218 		 * copying of random kernel data.
1219 		 */
1220 		for (n = ifs->ifs_nat_instances; n; n = n->nat_next)
1221 			if (n == nat)
1222 				break;
1223 		if (n == NULL) {
1224 			error = ESRCH;
1225 			goto finished;
1226 		}
1227 	}
1228 	ipn->ipn_next = nat->nat_next;
1229 
1230 	/*
1231 	 * Copy the NAT structure.
1232 	 */
1233 	bcopy((char *)nat, &ipn->ipn_nat, sizeof(*nat));
1234 
1235 	/*
1236 	 * If we have a pointer to the NAT rule it belongs to, save that too.
1237 	 */
1238 	if (nat->nat_ptr != NULL)
1239 		bcopy((char *)nat->nat_ptr, (char *)&ipn->ipn_ipnat,
1240 		      sizeof(ipn->ipn_ipnat));
1241 
1242 	/*
1243 	 * If we also know the NAT entry has an associated filter rule,
1244 	 * save that too.
1245 	 */
1246 	if (nat->nat_fr != NULL)
1247 		bcopy((char *)nat->nat_fr, (char *)&ipn->ipn_fr,
1248 		      sizeof(ipn->ipn_fr));
1249 
1250 	/*
1251 	 * Last but not least, if there is an application proxy session set
1252 	 * up for this NAT entry, then copy that out too, including any
1253 	 * private data saved along side it by the proxy.
1254 	 */
1255 	aps = nat->nat_aps;
1256 	outsize = ipn->ipn_dsize - sizeof(*ipn) + sizeof(ipn->ipn_data);
1257 	if (aps != NULL) {
1258 		char *s;
1259 
1260 		if (outsize < sizeof(*aps)) {
1261 			error = ENOBUFS;
1262 			goto finished;
1263 		}
1264 
1265 		s = ipn->ipn_data;
1266 		bcopy((char *)aps, s, sizeof(*aps));
1267 		s += sizeof(*aps);
1268 		outsize -= sizeof(*aps);
1269 		if ((aps->aps_data != NULL) && (outsize >= aps->aps_psiz))
1270 			bcopy(aps->aps_data, s, aps->aps_psiz);
1271 		else
1272 			error = ENOBUFS;
1273 	}
1274 	if (error == 0) {
1275 		error = fr_outobjsz(data, ipn, IPFOBJ_NATSAVE, ipns.ipn_dsize);
1276 	}
1277 
1278 finished:
1279 	if (ipn != NULL) {
1280 		KFREES(ipn, ipns.ipn_dsize);
1281 	}
1282 	return error;
1283 }
1284 
1285 
1286 /* ------------------------------------------------------------------------ */
1287 /* Function:    fr_natputent                                                */
1288 /* Returns:     int - 0 == success, != 0 is the error value.                */
1289 /* Parameters:  data(I) -     pointer to natget structure with NAT          */
1290 /*                            structure information to load into the kernel */
1291 /*              getlock(I) - flag indicating whether or not a write lock    */
1292 /*                           on ipf_nat is already held.                    */
1293 /*                                                                          */
1294 /* Handle SIOCSTPUT.                                                        */
1295 /* Loads a NAT table entry from user space, including a NAT rule, proxy and */
1296 /* firewall rule data structures, if pointers to them indicate so.          */
1297 /* ------------------------------------------------------------------------ */
1298 static int fr_natputent(data, getlock, ifs)
1299 caddr_t data;
1300 int getlock;
1301 ipf_stack_t *ifs;
1302 {
1303 	nat_save_t ipn, *ipnn;
1304 	ap_session_t *aps;
1305 	nat_t *n, *nat;
1306 	frentry_t *fr;
1307 	fr_info_t fin;
1308 	ipnat_t *in;
1309 	int error;
1310 
1311 	error = fr_inobj(data, &ipn, IPFOBJ_NATSAVE);
1312 	if (error != 0)
1313 		return error;
1314 
1315 	/*
1316 	 * Initialise early because of code at junkput label.
1317 	 */
1318 	in = NULL;
1319 	aps = NULL;
1320 	nat = NULL;
1321 	ipnn = NULL;
1322 
1323 	/*
1324 	 * New entry, copy in the rest of the NAT entry if it's size is more
1325 	 * than just the nat_t structure.
1326 	 */
1327 	fr = NULL;
1328 	if (ipn.ipn_dsize > sizeof(ipn)) {
1329 		if (ipn.ipn_dsize > 81920) {
1330 			error = ENOMEM;
1331 			goto junkput;
1332 		}
1333 
1334 		KMALLOCS(ipnn, nat_save_t *, ipn.ipn_dsize);
1335 		if (ipnn == NULL)
1336 			return ENOMEM;
1337 
1338 		error = fr_inobjsz(data, ipnn, IPFOBJ_NATSAVE, ipn.ipn_dsize);
1339 		if (error != 0) {
1340 			error = EFAULT;
1341 			goto junkput;
1342 		}
1343 	} else
1344 		ipnn = &ipn;
1345 
1346 	KMALLOC(nat, nat_t *);
1347 	if (nat == NULL) {
1348 		error = ENOMEM;
1349 		goto junkput;
1350 	}
1351 
1352 	bcopy((char *)&ipnn->ipn_nat, (char *)nat, sizeof(*nat));
1353 	/*
1354 	 * Initialize all these so that nat_delete() doesn't cause a crash.
1355 	 */
1356 	bzero((char *)nat, offsetof(struct nat, nat_tqe));
1357 	nat->nat_tqe.tqe_pnext = NULL;
1358 	nat->nat_tqe.tqe_next = NULL;
1359 	nat->nat_tqe.tqe_ifq = NULL;
1360 	nat->nat_tqe.tqe_parent = nat;
1361 
1362 	/*
1363 	 * Restore the rule associated with this nat session
1364 	 */
1365 	in = ipnn->ipn_nat.nat_ptr;
1366 	if (in != NULL) {
1367 		KMALLOC(in, ipnat_t *);
1368 		nat->nat_ptr = in;
1369 		if (in == NULL) {
1370 			error = ENOMEM;
1371 			goto junkput;
1372 		}
1373 		bzero((char *)in, offsetof(struct ipnat, in_next6));
1374 		bcopy((char *)&ipnn->ipn_ipnat, (char *)in, sizeof(*in));
1375 		in->in_use = 1;
1376 		in->in_flags |= IPN_DELETE;
1377 
1378 		ATOMIC_INC(ifs->ifs_nat_stats.ns_rules);
1379 
1380 		nat_resolverule(in, ifs);
1381 	}
1382 
1383 	/*
1384 	 * Check that the NAT entry doesn't already exist in the kernel.
1385 	 */
1386 	bzero((char *)&fin, sizeof(fin));
1387 	fin.fin_p = nat->nat_p;
1388 	if (nat->nat_dir == NAT_OUTBOUND) {
1389 		fin.fin_data[0] = ntohs(nat->nat_oport);
1390 		fin.fin_data[1] = ntohs(nat->nat_outport);
1391 		fin.fin_ifp = nat->nat_ifps[0];
1392 		if (getlock) {
1393 			READ_ENTER(&ifs->ifs_ipf_nat);
1394 		}
1395 		n = nat_inlookup(&fin, nat->nat_flags, fin.fin_p,
1396 			nat->nat_oip, nat->nat_outip);
1397 		if (getlock) {
1398 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
1399 		}
1400 		if (n != NULL) {
1401 			error = EEXIST;
1402 			goto junkput;
1403 		}
1404 	} else if (nat->nat_dir == NAT_INBOUND) {
1405 		fin.fin_data[0] = ntohs(nat->nat_inport);
1406 		fin.fin_data[1] = ntohs(nat->nat_oport);
1407 		fin.fin_ifp = nat->nat_ifps[1];
1408 		if (getlock) {
1409 			READ_ENTER(&ifs->ifs_ipf_nat);
1410 		}
1411 		n = nat_outlookup(&fin, nat->nat_flags, fin.fin_p,
1412 			nat->nat_inip, nat->nat_oip);
1413 		if (getlock) {
1414 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
1415 		}
1416 		if (n != NULL) {
1417 			error = EEXIST;
1418 			goto junkput;
1419 		}
1420 	} else {
1421 		error = EINVAL;
1422 		goto junkput;
1423 	}
1424 
1425 	/*
1426 	 * Restore ap_session_t structure.  Include the private data allocated
1427 	 * if it was there.
1428 	 */
1429 	aps = nat->nat_aps;
1430 	if (aps != NULL) {
1431 		KMALLOC(aps, ap_session_t *);
1432 		nat->nat_aps = aps;
1433 		if (aps == NULL) {
1434 			error = ENOMEM;
1435 			goto junkput;
1436 		}
1437 		bcopy(ipnn->ipn_data, (char *)aps, sizeof(*aps));
1438 		if (in != NULL)
1439 			aps->aps_apr = in->in_apr;
1440 		else
1441 			aps->aps_apr = NULL;
1442 		if (aps->aps_psiz != 0) {
1443 			if (aps->aps_psiz > 81920) {
1444 				error = ENOMEM;
1445 				goto junkput;
1446 			}
1447 			KMALLOCS(aps->aps_data, void *, aps->aps_psiz);
1448 			if (aps->aps_data == NULL) {
1449 				error = ENOMEM;
1450 				goto junkput;
1451 			}
1452 			bcopy(ipnn->ipn_data + sizeof(*aps), aps->aps_data,
1453 			      aps->aps_psiz);
1454 		} else {
1455 			aps->aps_psiz = 0;
1456 			aps->aps_data = NULL;
1457 		}
1458 	}
1459 
1460 	/*
1461 	 * If there was a filtering rule associated with this entry then
1462 	 * build up a new one.
1463 	 */
1464 	fr = nat->nat_fr;
1465 	if (fr != NULL) {
1466 		if ((nat->nat_flags & SI_NEWFR) != 0) {
1467 			KMALLOC(fr, frentry_t *);
1468 			nat->nat_fr = fr;
1469 			if (fr == NULL) {
1470 				error = ENOMEM;
1471 				goto junkput;
1472 			}
1473 			ipnn->ipn_nat.nat_fr = fr;
1474 			fr->fr_ref = 1;
1475 			(void) fr_outobj(data, ipnn, IPFOBJ_NATSAVE);
1476 			bcopy((char *)&ipnn->ipn_fr, (char *)fr, sizeof(*fr));
1477 			MUTEX_NUKE(&fr->fr_lock);
1478 			MUTEX_INIT(&fr->fr_lock, "nat-filter rule lock");
1479 		} else {
1480 			READ_ENTER(&ifs->ifs_ipf_nat);
1481 			for (n = ifs->ifs_nat_instances; n; n = n->nat_next)
1482 				if (n->nat_fr == fr)
1483 					break;
1484 
1485 			if (n != NULL) {
1486 				MUTEX_ENTER(&fr->fr_lock);
1487 				fr->fr_ref++;
1488 				MUTEX_EXIT(&fr->fr_lock);
1489 			}
1490 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
1491 
1492 			if (!n) {
1493 				error = ESRCH;
1494 				goto junkput;
1495 			}
1496 		}
1497 	}
1498 
1499 	if (ipnn != &ipn) {
1500 		KFREES(ipnn, ipn.ipn_dsize);
1501 		ipnn = NULL;
1502 	}
1503 
1504 	if (getlock) {
1505 		WRITE_ENTER(&ifs->ifs_ipf_nat);
1506 	}
1507 	error = nat_insert(nat, nat->nat_rev, ifs);
1508 	if ((error == 0) && (aps != NULL)) {
1509 		aps->aps_next = ifs->ifs_ap_sess_list;
1510 		ifs->ifs_ap_sess_list = aps;
1511 	}
1512 	if (getlock) {
1513 		RWLOCK_EXIT(&ifs->ifs_ipf_nat);
1514 	}
1515 
1516 	if (error == 0)
1517 		return 0;
1518 
1519 	error = ENOMEM;
1520 
1521 junkput:
1522 	if (fr != NULL)
1523 		(void) fr_derefrule(&fr, ifs);
1524 
1525 	if ((ipnn != NULL) && (ipnn != &ipn)) {
1526 		KFREES(ipnn, ipn.ipn_dsize);
1527 	}
1528 	if (nat != NULL) {
1529 		if (aps != NULL) {
1530 			if (aps->aps_data != NULL) {
1531 				KFREES(aps->aps_data, aps->aps_psiz);
1532 			}
1533 			KFREE(aps);
1534 		}
1535 		if (in != NULL) {
1536 			if (in->in_apr)
1537 				appr_free(in->in_apr);
1538 			KFREE(in);
1539 		}
1540 		KFREE(nat);
1541 	}
1542 	return error;
1543 }
1544 
1545 
1546 /* ------------------------------------------------------------------------ */
1547 /* Function:    nat_delete                                                  */
1548 /* Returns:     Nil                                                         */
1549 /* Parameters:  natd(I)    - pointer to NAT structure to delete             */
1550 /*              logtype(I) - type of LOG record to create before deleting   */
1551 /* Write Lock:  ipf_nat                                                     */
1552 /*                                                                          */
1553 /* Delete a nat entry from the various lists and table.  If NAT logging is  */
1554 /* enabled then generate a NAT log record for this event.                   */
1555 /* ------------------------------------------------------------------------ */
1556 static void nat_delete(nat, logtype, ifs)
1557 struct nat *nat;
1558 int logtype;
1559 ipf_stack_t *ifs;
1560 {
1561 	struct ipnat *ipn;
1562 
1563 	if (logtype != 0 && ifs->ifs_nat_logging != 0)
1564 		nat_log(nat, logtype, ifs);
1565 
1566 	MUTEX_ENTER(&ifs->ifs_ipf_nat_new);
1567 
1568 	/*
1569 	 * Take it as a general indication that all the pointers are set if
1570 	 * nat_pnext is set.
1571 	 */
1572 	if (nat->nat_pnext != NULL) {
1573 		ifs->ifs_nat_stats.ns_bucketlen[0][nat->nat_hv[0]]--;
1574 		ifs->ifs_nat_stats.ns_bucketlen[1][nat->nat_hv[1]]--;
1575 
1576 		*nat->nat_pnext = nat->nat_next;
1577 		if (nat->nat_next != NULL) {
1578 			nat->nat_next->nat_pnext = nat->nat_pnext;
1579 			nat->nat_next = NULL;
1580 		}
1581 		nat->nat_pnext = NULL;
1582 
1583 		*nat->nat_phnext[0] = nat->nat_hnext[0];
1584 		if (nat->nat_hnext[0] != NULL) {
1585 			nat->nat_hnext[0]->nat_phnext[0] = nat->nat_phnext[0];
1586 			nat->nat_hnext[0] = NULL;
1587 		}
1588 		nat->nat_phnext[0] = NULL;
1589 
1590 		*nat->nat_phnext[1] = nat->nat_hnext[1];
1591 		if (nat->nat_hnext[1] != NULL) {
1592 			nat->nat_hnext[1]->nat_phnext[1] = nat->nat_phnext[1];
1593 			nat->nat_hnext[1] = NULL;
1594 		}
1595 		nat->nat_phnext[1] = NULL;
1596 
1597 		if ((nat->nat_flags & SI_WILDP) != 0)
1598 			ifs->ifs_nat_stats.ns_wilds--;
1599 	}
1600 
1601 	if (nat->nat_me != NULL) {
1602 		*nat->nat_me = NULL;
1603 		nat->nat_me = NULL;
1604 	}
1605 
1606 	fr_deletequeueentry(&nat->nat_tqe);
1607 
1608 	nat->nat_ref--;
1609 	if (nat->nat_ref > 0) {
1610 		MUTEX_EXIT(&ifs->ifs_ipf_nat_new);
1611 		return;
1612 	}
1613 
1614 #ifdef	IPFILTER_SYNC
1615 	if (nat->nat_sync)
1616 		ipfsync_del(nat->nat_sync);
1617 #endif
1618 
1619 	if (nat->nat_fr != NULL)
1620 		(void)fr_derefrule(&nat->nat_fr, ifs);
1621 
1622 	if (nat->nat_hm != NULL)
1623 		nat_hostmapdel(nat->nat_hm);
1624 
1625 	/*
1626 	 * If there is an active reference from the nat entry to its parent
1627 	 * rule, decrement the rule's reference count and free it too if no
1628 	 * longer being used.
1629 	 */
1630 	ipn = nat->nat_ptr;
1631 	if (ipn != NULL) {
1632 		ipn->in_space++;
1633 		ipn->in_use--;
1634 		if (ipn->in_use == 0 && (ipn->in_flags & IPN_DELETE)) {
1635 			if (ipn->in_apr)
1636 				appr_free(ipn->in_apr);
1637 			KFREE(ipn);
1638 			ifs->ifs_nat_stats.ns_rules--;
1639 		}
1640 	}
1641 
1642 	MUTEX_DESTROY(&nat->nat_lock);
1643 
1644 	aps_free(nat->nat_aps, ifs);
1645 	ifs->ifs_nat_stats.ns_inuse--;
1646 	MUTEX_EXIT(&ifs->ifs_ipf_nat_new);
1647 
1648 	/*
1649 	 * If there's a fragment table entry too for this nat entry, then
1650 	 * dereference that as well.  This is after nat_lock is released
1651 	 * because of Tru64.
1652 	 */
1653 	fr_forgetnat((void *)nat, ifs);
1654 
1655 	KFREE(nat);
1656 }
1657 
1658 
1659 /* ------------------------------------------------------------------------ */
1660 /* Function:    nat_flushtable                                              */
1661 /* Returns:     int - number of NAT rules deleted                           */
1662 /* Parameters:  Nil                                                         */
1663 /*                                                                          */
1664 /* Deletes all currently active NAT sessions.  In deleting each NAT entry a */
1665 /* log record should be emitted in nat_delete() if NAT logging is enabled.  */
1666 /* ------------------------------------------------------------------------ */
1667 /*
1668  * nat_flushtable - clear the NAT table of all mapping entries.
1669  */
1670 static int nat_flushtable(ifs)
1671 ipf_stack_t *ifs;
1672 {
1673 	nat_t *nat;
1674 	int j = 0;
1675 
1676 	/*
1677 	 * ALL NAT mappings deleted, so lets just make the deletions
1678 	 * quicker.
1679 	 */
1680 	if (ifs->ifs_nat_table[0] != NULL)
1681 		bzero((char *)ifs->ifs_nat_table[0],
1682 		      sizeof(ifs->ifs_nat_table[0]) * ifs->ifs_ipf_nattable_sz);
1683 	if (ifs->ifs_nat_table[1] != NULL)
1684 		bzero((char *)ifs->ifs_nat_table[1],
1685 		      sizeof(ifs->ifs_nat_table[1]) * ifs->ifs_ipf_nattable_sz);
1686 
1687 	while ((nat = ifs->ifs_nat_instances) != NULL) {
1688 		nat_delete(nat, NL_FLUSH, ifs);
1689 		j++;
1690 	}
1691 
1692 	ifs->ifs_nat_stats.ns_inuse = 0;
1693 	return j;
1694 }
1695 
1696 
1697 /* ------------------------------------------------------------------------ */
1698 /* Function:    nat_clearlist                                               */
1699 /* Returns:     int - number of NAT/RDR rules deleted                       */
1700 /* Parameters:  Nil                                                         */
1701 /*                                                                          */
1702 /* Delete all rules in the current list of rules.  There is nothing elegant */
1703 /* about this cleanup: simply free all entries on the list of rules and     */
1704 /* clear out the tables used for hashed NAT rule lookups.                   */
1705 /* ------------------------------------------------------------------------ */
1706 static int nat_clearlist(ifs)
1707 ipf_stack_t *ifs;
1708 {
1709 	ipnat_t *n, **np = &ifs->ifs_nat_list;
1710 	int i = 0;
1711 
1712 	if (ifs->ifs_nat_rules != NULL)
1713 		bzero((char *)ifs->ifs_nat_rules,
1714 		      sizeof(*ifs->ifs_nat_rules) * ifs->ifs_ipf_natrules_sz);
1715 	if (ifs->ifs_rdr_rules != NULL)
1716 		bzero((char *)ifs->ifs_rdr_rules,
1717 		      sizeof(*ifs->ifs_rdr_rules) * ifs->ifs_ipf_rdrrules_sz);
1718 
1719 	while ((n = *np) != NULL) {
1720 		*np = n->in_next;
1721 		if (n->in_use == 0) {
1722 			if (n->in_apr != NULL)
1723 				appr_free(n->in_apr);
1724 			KFREE(n);
1725 			ifs->ifs_nat_stats.ns_rules--;
1726 		} else {
1727 			n->in_flags |= IPN_DELETE;
1728 			n->in_next = NULL;
1729 		}
1730 		i++;
1731 	}
1732 	ifs->ifs_nat_masks = 0;
1733 	ifs->ifs_rdr_masks = 0;
1734 	return i;
1735 }
1736 
1737 
1738 /* ------------------------------------------------------------------------ */
1739 /* Function:    nat_newmap                                                  */
1740 /* Returns:     int - -1 == error, 0 == success                             */
1741 /* Parameters:  fin(I) - pointer to packet information                      */
1742 /*              nat(I) - pointer to NAT entry                               */
1743 /*              ni(I)  - pointer to structure with misc. information needed */
1744 /*                       to create new NAT entry.                           */
1745 /*                                                                          */
1746 /* Given an empty NAT structure, populate it with new information about a   */
1747 /* new NAT session, as defined by the matching NAT rule.                    */
1748 /* ni.nai_ip is passed in uninitialised and must be set, in host byte order,*/
1749 /* to the new IP address for the translation.                               */
1750 /* ------------------------------------------------------------------------ */
1751 static INLINE int nat_newmap(fin, nat, ni)
1752 fr_info_t *fin;
1753 nat_t *nat;
1754 natinfo_t *ni;
1755 {
1756 	u_short st_port, dport, sport, port, sp, dp;
1757 	struct in_addr in, inb;
1758 	hostmap_t *hm;
1759 	u_32_t flags;
1760 	u_32_t st_ip;
1761 	ipnat_t *np;
1762 	nat_t *natl;
1763 	int l;
1764 	ipf_stack_t *ifs = fin->fin_ifs;
1765 
1766 	/*
1767 	 * If it's an outbound packet which doesn't match any existing
1768 	 * record, then create a new port
1769 	 */
1770 	l = 0;
1771 	hm = NULL;
1772 	np = ni->nai_np;
1773 	st_ip = np->in_nip;
1774 	st_port = np->in_pnext;
1775 	flags = ni->nai_flags;
1776 	sport = ni->nai_sport;
1777 	dport = ni->nai_dport;
1778 
1779 	/*
1780 	 * Do a loop until we either run out of entries to try or we find
1781 	 * a NAT mapping that isn't currently being used.  This is done
1782 	 * because the change to the source is not (usually) being fixed.
1783 	 */
1784 	do {
1785 		port = 0;
1786 		in.s_addr = htonl(np->in_nip);
1787 		if (l == 0) {
1788 			/*
1789 			 * Check to see if there is an existing NAT
1790 			 * setup for this IP address pair.
1791 			 */
1792 			hm = nat_hostmap(np, fin->fin_src, fin->fin_dst,
1793 					 in, 0, ifs);
1794 			if (hm != NULL)
1795 				in.s_addr = hm->hm_mapip.s_addr;
1796 		} else if ((l == 1) && (hm != NULL)) {
1797 			nat_hostmapdel(hm);
1798 			hm = NULL;
1799 		}
1800 		in.s_addr = ntohl(in.s_addr);
1801 
1802 		nat->nat_hm = hm;
1803 
1804 		if ((np->in_outmsk == 0xffffffff) && (np->in_pnext == 0)) {
1805 			if (l > 0)
1806 				return -1;
1807 		}
1808 
1809 		if (np->in_redir == NAT_BIMAP &&
1810 		    np->in_inmsk == np->in_outmsk) {
1811 			/*
1812 			 * map the address block in a 1:1 fashion
1813 			 */
1814 			in.s_addr = np->in_outip;
1815 			in.s_addr |= fin->fin_saddr & ~np->in_inmsk;
1816 			in.s_addr = ntohl(in.s_addr);
1817 
1818 		} else if (np->in_redir & NAT_MAPBLK) {
1819 			if ((l >= np->in_ppip) || ((l > 0) &&
1820 			     !(flags & IPN_TCPUDP)))
1821 				return -1;
1822 			/*
1823 			 * map-block - Calculate destination address.
1824 			 */
1825 			in.s_addr = ntohl(fin->fin_saddr);
1826 			in.s_addr &= ntohl(~np->in_inmsk);
1827 			inb.s_addr = in.s_addr;
1828 			in.s_addr /= np->in_ippip;
1829 			in.s_addr &= ntohl(~np->in_outmsk);
1830 			in.s_addr += ntohl(np->in_outip);
1831 			/*
1832 			 * Calculate destination port.
1833 			 */
1834 			if ((flags & IPN_TCPUDP) &&
1835 			    (np->in_ppip != 0)) {
1836 				port = ntohs(sport) + l;
1837 				port %= np->in_ppip;
1838 				port += np->in_ppip *
1839 					(inb.s_addr % np->in_ippip);
1840 				port += MAPBLK_MINPORT;
1841 				port = htons(port);
1842 			}
1843 
1844 		} else if ((np->in_outip == 0) &&
1845 			   (np->in_outmsk == 0xffffffff)) {
1846 			/*
1847 			 * 0/32 - use the interface's IP address.
1848 			 */
1849 			if ((l > 0) ||
1850 			    fr_ifpaddr(4, FRI_NORMAL, fin->fin_ifp,
1851 				       &in, NULL, fin->fin_ifs) == -1)
1852 				return -1;
1853 			in.s_addr = ntohl(in.s_addr);
1854 
1855 		} else if ((np->in_outip == 0) && (np->in_outmsk == 0)) {
1856 			/*
1857 			 * 0/0 - use the original source address/port.
1858 			 */
1859 			if (l > 0)
1860 				return -1;
1861 			in.s_addr = ntohl(fin->fin_saddr);
1862 
1863 		} else if ((np->in_outmsk != 0xffffffff) &&
1864 			   (np->in_pnext == 0) && ((l > 0) || (hm == NULL)))
1865 			np->in_nip++;
1866 
1867 		natl = NULL;
1868 
1869 		if ((flags & IPN_TCPUDP) &&
1870 		    ((np->in_redir & NAT_MAPBLK) == 0) &&
1871 		    (np->in_flags & IPN_AUTOPORTMAP)) {
1872 			/*
1873 			 * "ports auto" (without map-block)
1874 			 */
1875 			if ((l > 0) && (l % np->in_ppip == 0)) {
1876 				if (l > np->in_space) {
1877 					return -1;
1878 				} else if ((l > np->in_ppip) &&
1879 					   np->in_outmsk != 0xffffffff)
1880 					np->in_nip++;
1881 			}
1882 			if (np->in_ppip != 0) {
1883 				port = ntohs(sport);
1884 				port += (l % np->in_ppip);
1885 				port %= np->in_ppip;
1886 				port += np->in_ppip *
1887 					(ntohl(fin->fin_saddr) %
1888 					 np->in_ippip);
1889 				port += MAPBLK_MINPORT;
1890 				port = htons(port);
1891 			}
1892 
1893 		} else if (((np->in_redir & NAT_MAPBLK) == 0) &&
1894 			   (flags & IPN_TCPUDPICMP) && (np->in_pnext != 0)) {
1895 			/*
1896 			 * Standard port translation.  Select next port.
1897 			 */
1898 			port = htons(np->in_pnext++);
1899 
1900 			if (np->in_pnext > ntohs(np->in_pmax)) {
1901 				np->in_pnext = ntohs(np->in_pmin);
1902 				if (np->in_outmsk != 0xffffffff)
1903 					np->in_nip++;
1904 			}
1905 		}
1906 
1907 		if (np->in_flags & IPN_IPRANGE) {
1908 			if (np->in_nip > ntohl(np->in_outmsk))
1909 				np->in_nip = ntohl(np->in_outip);
1910 		} else {
1911 			if ((np->in_outmsk != 0xffffffff) &&
1912 			    ((np->in_nip + 1) & ntohl(np->in_outmsk)) >
1913 			    ntohl(np->in_outip))
1914 				np->in_nip = ntohl(np->in_outip) + 1;
1915 		}
1916 
1917 		if ((port == 0) && (flags & (IPN_TCPUDPICMP|IPN_ICMPQUERY)))
1918 			port = sport;
1919 
1920 		/*
1921 		 * Here we do a lookup of the connection as seen from
1922 		 * the outside.  If an IP# pair already exists, try
1923 		 * again.  So if you have A->B becomes C->B, you can
1924 		 * also have D->E become C->E but not D->B causing
1925 		 * another C->B.  Also take protocol and ports into
1926 		 * account when determining whether a pre-existing
1927 		 * NAT setup will cause an external conflict where
1928 		 * this is appropriate.
1929 		 */
1930 		inb.s_addr = htonl(in.s_addr);
1931 		sp = fin->fin_data[0];
1932 		dp = fin->fin_data[1];
1933 		fin->fin_data[0] = fin->fin_data[1];
1934 		fin->fin_data[1] = htons(port);
1935 		natl = nat_inlookup(fin, flags & ~(SI_WILDP|NAT_SEARCH),
1936 				    (u_int)fin->fin_p, fin->fin_dst, inb);
1937 		fin->fin_data[0] = sp;
1938 		fin->fin_data[1] = dp;
1939 
1940 		/*
1941 		 * Has the search wrapped around and come back to the
1942 		 * start ?
1943 		 */
1944 		if ((natl != NULL) &&
1945 		    (np->in_pnext != 0) && (st_port == np->in_pnext) &&
1946 		    (np->in_nip != 0) && (st_ip == np->in_nip))
1947 			return -1;
1948 		l++;
1949 	} while (natl != NULL);
1950 
1951 	if (np->in_space > 0)
1952 		np->in_space--;
1953 
1954 	/* Setup the NAT table */
1955 	nat->nat_inip = fin->fin_src;
1956 	nat->nat_outip.s_addr = htonl(in.s_addr);
1957 	nat->nat_oip = fin->fin_dst;
1958 	if (nat->nat_hm == NULL)
1959 		nat->nat_hm = nat_hostmap(np, fin->fin_src, fin->fin_dst,
1960 					  nat->nat_outip, 0, ifs);
1961 
1962 	/*
1963 	 * The ICMP checksum does not have a pseudo header containing
1964 	 * the IP addresses
1965 	 */
1966 	ni->nai_sum1 = LONG_SUM(ntohl(fin->fin_saddr));
1967 	ni->nai_sum2 = LONG_SUM(in.s_addr);
1968 	if ((flags & IPN_TCPUDP)) {
1969 		ni->nai_sum1 += ntohs(sport);
1970 		ni->nai_sum2 += ntohs(port);
1971 	}
1972 
1973 	if (flags & IPN_TCPUDP) {
1974 		nat->nat_inport = sport;
1975 		nat->nat_outport = port;	/* sport */
1976 		nat->nat_oport = dport;
1977 		((tcphdr_t *)fin->fin_dp)->th_sport = port;
1978 	} else if (flags & IPN_ICMPQUERY) {
1979 		((icmphdr_t *)fin->fin_dp)->icmp_id = port;
1980 		nat->nat_inport = port;
1981 		nat->nat_outport = port;
1982 	}
1983 
1984 	ni->nai_ip.s_addr = in.s_addr;
1985 	ni->nai_port = port;
1986 	ni->nai_nport = dport;
1987 	return 0;
1988 }
1989 
1990 
1991 /* ------------------------------------------------------------------------ */
1992 /* Function:    nat_newrdr                                                  */
1993 /* Returns:     int - -1 == error, 0 == success (no move), 1 == success and */
1994 /*                    allow rule to be moved if IPN_ROUNDR is set.          */
1995 /* Parameters:  fin(I) - pointer to packet information                      */
1996 /*              nat(I) - pointer to NAT entry                               */
1997 /*              ni(I)  - pointer to structure with misc. information needed */
1998 /*                       to create new NAT entry.                           */
1999 /*                                                                          */
2000 /* ni.nai_ip is passed in uninitialised and must be set, in host byte order,*/
2001 /* to the new IP address for the translation.                               */
2002 /* ------------------------------------------------------------------------ */
2003 static INLINE int nat_newrdr(fin, nat, ni)
2004 fr_info_t *fin;
2005 nat_t *nat;
2006 natinfo_t *ni;
2007 {
2008 	u_short nport, dport, sport;
2009 	struct in_addr in;
2010 	hostmap_t *hm;
2011 	u_32_t flags;
2012 	ipnat_t *np;
2013 	int move;
2014 	ipf_stack_t *ifs = fin->fin_ifs;
2015 
2016 	move = 1;
2017 	hm = NULL;
2018 	in.s_addr = 0;
2019 	np = ni->nai_np;
2020 	flags = ni->nai_flags;
2021 	sport = ni->nai_sport;
2022 	dport = ni->nai_dport;
2023 
2024 	/*
2025 	 * If the matching rule has IPN_STICKY set, then we want to have the
2026 	 * same rule kick in as before.  Why would this happen?  If you have
2027 	 * a collection of rdr rules with "round-robin sticky", the current
2028 	 * packet might match a different one to the previous connection but
2029 	 * we want the same destination to be used.
2030 	 */
2031 	if ((np->in_flags & (IPN_ROUNDR|IPN_STICKY)) ==
2032 	    (IPN_ROUNDR|IPN_STICKY)) {
2033 		hm = nat_hostmap(NULL, fin->fin_src, fin->fin_dst, in,
2034 				 (u_32_t)dport, ifs);
2035 		if (hm != NULL) {
2036 			in.s_addr = ntohl(hm->hm_mapip.s_addr);
2037 			np = hm->hm_ipnat;
2038 			ni->nai_np = np;
2039 			move = 0;
2040 		}
2041 	}
2042 
2043 	/*
2044 	 * Otherwise, it's an inbound packet. Most likely, we don't
2045 	 * want to rewrite source ports and source addresses. Instead,
2046 	 * we want to rewrite to a fixed internal address and fixed
2047 	 * internal port.
2048 	 */
2049 	if (np->in_flags & IPN_SPLIT) {
2050 		in.s_addr = np->in_nip;
2051 
2052 		if ((np->in_flags & (IPN_ROUNDR|IPN_STICKY)) == IPN_STICKY) {
2053 			hm = nat_hostmap(np, fin->fin_src, fin->fin_dst,
2054 					 in, (u_32_t)dport, ifs);
2055 			if (hm != NULL) {
2056 				in.s_addr = hm->hm_mapip.s_addr;
2057 				move = 0;
2058 			}
2059 		}
2060 
2061 		if (hm == NULL || hm->hm_ref == 1) {
2062 			if (np->in_inip == htonl(in.s_addr)) {
2063 				np->in_nip = ntohl(np->in_inmsk);
2064 				move = 0;
2065 			} else {
2066 				np->in_nip = ntohl(np->in_inip);
2067 			}
2068 		}
2069 
2070 	} else if ((np->in_inip == 0) && (np->in_inmsk == 0xffffffff)) {
2071 		/*
2072 		 * 0/32 - use the interface's IP address.
2073 		 */
2074 		if (fr_ifpaddr(4, FRI_NORMAL, fin->fin_ifp, &in, NULL,
2075 			   fin->fin_ifs) == -1)
2076 			return -1;
2077 		in.s_addr = ntohl(in.s_addr);
2078 
2079 	} else if ((np->in_inip == 0) && (np->in_inmsk== 0)) {
2080 		/*
2081 		 * 0/0 - use the original destination address/port.
2082 		 */
2083 		in.s_addr = ntohl(fin->fin_daddr);
2084 
2085 	} else if (np->in_redir == NAT_BIMAP &&
2086 		   np->in_inmsk == np->in_outmsk) {
2087 		/*
2088 		 * map the address block in a 1:1 fashion
2089 		 */
2090 		in.s_addr = np->in_inip;
2091 		in.s_addr |= fin->fin_daddr & ~np->in_inmsk;
2092 		in.s_addr = ntohl(in.s_addr);
2093 	} else {
2094 		in.s_addr = ntohl(np->in_inip);
2095 	}
2096 
2097 	if ((np->in_pnext == 0) || ((flags & NAT_NOTRULEPORT) != 0))
2098 		nport = dport;
2099 	else {
2100 		/*
2101 		 * Whilst not optimized for the case where
2102 		 * pmin == pmax, the gain is not significant.
2103 		 */
2104 		if (((np->in_flags & IPN_FIXEDDPORT) == 0) &&
2105 		    (np->in_pmin != np->in_pmax)) {
2106 			nport = ntohs(dport) - ntohs(np->in_pmin) +
2107 				ntohs(np->in_pnext);
2108 			nport = htons(nport);
2109 		} else
2110 			nport = np->in_pnext;
2111 	}
2112 
2113 	/*
2114 	 * When the redirect-to address is set to 0.0.0.0, just
2115 	 * assume a blank `forwarding' of the packet.  We don't
2116 	 * setup any translation for this either.
2117 	 */
2118 	if (in.s_addr == 0) {
2119 		if (nport == dport)
2120 			return -1;
2121 		in.s_addr = ntohl(fin->fin_daddr);
2122 	}
2123 
2124 	nat->nat_inip.s_addr = htonl(in.s_addr);
2125 	nat->nat_outip = fin->fin_dst;
2126 	nat->nat_oip = fin->fin_src;
2127 
2128 	ni->nai_sum1 = LONG_SUM(ntohl(fin->fin_daddr)) + ntohs(dport);
2129 	ni->nai_sum2 = LONG_SUM(in.s_addr) + ntohs(nport);
2130 
2131 	ni->nai_ip.s_addr = in.s_addr;
2132 	ni->nai_nport = nport;
2133 	ni->nai_port = sport;
2134 
2135 	if (flags & IPN_TCPUDP) {
2136 		nat->nat_inport = nport;
2137 		nat->nat_outport = dport;
2138 		nat->nat_oport = sport;
2139 		((tcphdr_t *)fin->fin_dp)->th_dport = nport;
2140 	} else if (flags & IPN_ICMPQUERY) {
2141 		((icmphdr_t *)fin->fin_dp)->icmp_id = nport;
2142 		nat->nat_inport = nport;
2143 		nat->nat_outport = nport;
2144 	}
2145 
2146 	return move;
2147 }
2148 
2149 /* ------------------------------------------------------------------------ */
2150 /* Function:    nat_new                                                     */
2151 /* Returns:     nat_t* - NULL == failure to create new NAT structure,       */
2152 /*                       else pointer to new NAT structure                  */
2153 /* Parameters:  fin(I)       - pointer to packet information                */
2154 /*              np(I)        - pointer to NAT rule                          */
2155 /*              natsave(I)   - pointer to where to store NAT struct pointer */
2156 /*              flags(I)     - flags describing the current packet          */
2157 /*              direction(I) - direction of packet (in/out)                 */
2158 /* Write Lock:  ipf_nat                                                     */
2159 /*                                                                          */
2160 /* Attempts to create a new NAT entry.  Does not actually change the packet */
2161 /* in any way.                                                              */
2162 /*                                                                          */
2163 /* This fucntion is in three main parts: (1) deal with creating a new NAT   */
2164 /* structure for a "MAP" rule (outgoing NAT translation); (2) deal with     */
2165 /* creating a new NAT structure for a "RDR" rule (incoming NAT translation) */
2166 /* and (3) building that structure and putting it into the NAT table(s).    */
2167 /* ------------------------------------------------------------------------ */
2168 nat_t *nat_new(fin, np, natsave, flags, direction)
2169 fr_info_t *fin;
2170 ipnat_t *np;
2171 nat_t **natsave;
2172 u_int flags;
2173 int direction;
2174 {
2175 	u_short port = 0, sport = 0, dport = 0, nport = 0;
2176 	tcphdr_t *tcp = NULL;
2177 	hostmap_t *hm = NULL;
2178 	struct in_addr in;
2179 	nat_t *nat, *natl;
2180 	u_int nflags;
2181 	natinfo_t ni;
2182 	u_32_t sumd;
2183 	int move;
2184 	ipf_stack_t *ifs = fin->fin_ifs;
2185 
2186 	if (ifs->ifs_nat_stats.ns_inuse >= ifs->ifs_ipf_nattable_max) {
2187 		ifs->ifs_nat_stats.ns_memfail++;
2188 		return NULL;
2189 	}
2190 
2191 	move = 1;
2192 	nflags = np->in_flags & flags;
2193 	nflags &= NAT_FROMRULE;
2194 
2195 	ni.nai_np = np;
2196 	ni.nai_nflags = nflags;
2197 	ni.nai_flags = flags;
2198 
2199 	/* Give me a new nat */
2200 	KMALLOC(nat, nat_t *);
2201 	if (nat == NULL) {
2202 		ifs->ifs_nat_stats.ns_memfail++;
2203 		/*
2204 		 * Try to automatically tune the max # of entries in the
2205 		 * table allowed to be less than what will cause kmem_alloc()
2206 		 * to fail and try to eliminate panics due to out of memory
2207 		 * conditions arising.
2208 		 */
2209 		if (ifs->ifs_ipf_nattable_max > ifs->ifs_ipf_nattable_sz) {
2210 			ifs->ifs_ipf_nattable_max = ifs->ifs_nat_stats.ns_inuse - 100;
2211 			printf("ipf_nattable_max reduced to %d\n",
2212 				ifs->ifs_ipf_nattable_max);
2213 		}
2214 		return NULL;
2215 	}
2216 
2217 	if (flags & IPN_TCPUDP) {
2218 		tcp = fin->fin_dp;
2219 		ni.nai_sport = htons(fin->fin_sport);
2220 		ni.nai_dport = htons(fin->fin_dport);
2221 	} else if (flags & IPN_ICMPQUERY) {
2222 		/*
2223 		 * In the ICMP query NAT code, we translate the ICMP id fields
2224 		 * to make them unique. This is indepedent of the ICMP type
2225 		 * (e.g. in the unlikely event that a host sends an echo and
2226 		 * an tstamp request with the same id, both packets will have
2227 		 * their ip address/id field changed in the same way).
2228 		 */
2229 		/* The icmp_id field is used by the sender to identify the
2230 		 * process making the icmp request. (the receiver justs
2231 		 * copies it back in its response). So, it closely matches
2232 		 * the concept of source port. We overlay sport, so we can
2233 		 * maximally reuse the existing code.
2234 		 */
2235 		ni.nai_sport = ((icmphdr_t *)fin->fin_dp)->icmp_id;
2236 		ni.nai_dport = ni.nai_sport;
2237 	}
2238 
2239 	bzero((char *)nat, sizeof(*nat));
2240 	nat->nat_flags = flags;
2241 	nat->nat_redir = np->in_redir;
2242 
2243 	if ((flags & NAT_SLAVE) == 0) {
2244 		MUTEX_ENTER(&ifs->ifs_ipf_nat_new);
2245 	}
2246 
2247 	/*
2248 	 * Search the current table for a match.
2249 	 */
2250 	if (direction == NAT_OUTBOUND) {
2251 		/*
2252 		 * We can now arrange to call this for the same connection
2253 		 * because ipf_nat_new doesn't protect the code path into
2254 		 * this function.
2255 		 */
2256 		natl = nat_outlookup(fin, nflags, (u_int)fin->fin_p,
2257 				     fin->fin_src, fin->fin_dst);
2258 		if (natl != NULL) {
2259 			nat = natl;
2260 			goto done;
2261 		}
2262 
2263 		move = nat_newmap(fin, nat, &ni);
2264 		if (move == -1)
2265 			goto badnat;
2266 
2267 		np = ni.nai_np;
2268 		in = ni.nai_ip;
2269 	} else {
2270 		/*
2271 		 * NAT_INBOUND is used only for redirects rules
2272 		 */
2273 		natl = nat_inlookup(fin, nflags, (u_int)fin->fin_p,
2274 				    fin->fin_src, fin->fin_dst);
2275 		if (natl != NULL) {
2276 			nat = natl;
2277 			goto done;
2278 		}
2279 
2280 		move = nat_newrdr(fin, nat, &ni);
2281 		if (move == -1)
2282 			goto badnat;
2283 
2284 		np = ni.nai_np;
2285 		in = ni.nai_ip;
2286 	}
2287 	port = ni.nai_port;
2288 	nport = ni.nai_nport;
2289 
2290 	if ((move == 1) && (np->in_flags & IPN_ROUNDR)) {
2291 		if (np->in_redir == NAT_REDIRECT) {
2292 			nat_delrdr(np);
2293 			nat_addrdr(np, ifs);
2294 		} else if (np->in_redir == NAT_MAP) {
2295 			nat_delnat(np);
2296 			nat_addnat(np, ifs);
2297 		}
2298 	}
2299 
2300 	if (flags & IPN_TCPUDP) {
2301 		sport = ni.nai_sport;
2302 		dport = ni.nai_dport;
2303 	} else if (flags & IPN_ICMPQUERY) {
2304 		sport = ni.nai_sport;
2305 		dport = 0;
2306 	}
2307 
2308 	/*
2309 	 * nat_sumd[0] stores adjustment value including both IP address and
2310 	 * port number changes. nat_sumd[1] stores adjustment value only for
2311 	 * IP address changes, to be used for pseudo header adjustment, in
2312 	 * case hardware partial checksum offload is offered.
2313 	 */
2314 	CALC_SUMD(ni.nai_sum1, ni.nai_sum2, sumd);
2315 	nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
2316 #if SOLARIS && defined(_KERNEL) && (SOLARIS2 >= 6)
2317 	if (flags & IPN_TCPUDP) {
2318 		ni.nai_sum1 = LONG_SUM(in.s_addr);
2319 		if (direction == NAT_OUTBOUND)
2320 			ni.nai_sum2 = LONG_SUM(ntohl(fin->fin_saddr));
2321 		else
2322 			ni.nai_sum2 = LONG_SUM(ntohl(fin->fin_daddr));
2323 
2324 		CALC_SUMD(ni.nai_sum1, ni.nai_sum2, sumd);
2325 		nat->nat_sumd[1] = (sumd & 0xffff) + (sumd >> 16);
2326 	} else
2327 #endif
2328 		nat->nat_sumd[1] = nat->nat_sumd[0];
2329 
2330 	if ((flags & IPN_TCPUDPICMP) && ((sport != port) || (dport != nport))) {
2331 		if (direction == NAT_OUTBOUND)
2332 			ni.nai_sum1 = LONG_SUM(ntohl(fin->fin_saddr));
2333 		else
2334 			ni.nai_sum1 = LONG_SUM(ntohl(fin->fin_daddr));
2335 
2336 		ni.nai_sum2 = LONG_SUM(in.s_addr);
2337 
2338 		CALC_SUMD(ni.nai_sum1, ni.nai_sum2, sumd);
2339 		nat->nat_ipsumd = (sumd & 0xffff) + (sumd >> 16);
2340 	} else {
2341 		nat->nat_ipsumd = nat->nat_sumd[0];
2342 		if (!(flags & IPN_TCPUDPICMP)) {
2343 			nat->nat_sumd[0] = 0;
2344 			nat->nat_sumd[1] = 0;
2345 		}
2346 	}
2347 
2348 	if (nat_finalise(fin, nat, &ni, tcp, natsave, direction) == -1) {
2349 		goto badnat;
2350 	}
2351 	if (flags & SI_WILDP)
2352 		ifs->ifs_nat_stats.ns_wilds++;
2353 	goto done;
2354 badnat:
2355 	ifs->ifs_nat_stats.ns_badnat++;
2356 	if ((hm = nat->nat_hm) != NULL)
2357 		nat_hostmapdel(hm);
2358 	KFREE(nat);
2359 	nat = NULL;
2360 done:
2361 	if ((flags & NAT_SLAVE) == 0) {
2362 		MUTEX_EXIT(&ifs->ifs_ipf_nat_new);
2363 	}
2364 	return nat;
2365 }
2366 
2367 
2368 /* ------------------------------------------------------------------------ */
2369 /* Function:    nat_finalise                                                */
2370 /* Returns:     int - 0 == sucess, -1 == failure                            */
2371 /* Parameters:  fin(I) - pointer to packet information                      */
2372 /*              nat(I) - pointer to NAT entry                               */
2373 /*              ni(I)  - pointer to structure with misc. information needed */
2374 /*                       to create new NAT entry.                           */
2375 /* Write Lock:  ipf_nat                                                     */
2376 /*                                                                          */
2377 /* This is the tail end of constructing a new NAT entry and is the same     */
2378 /* for both IPv4 and IPv6.                                                  */
2379 /* ------------------------------------------------------------------------ */
2380 /*ARGSUSED*/
2381 static INLINE int nat_finalise(fin, nat, ni, tcp, natsave, direction)
2382 fr_info_t *fin;
2383 nat_t *nat;
2384 natinfo_t *ni;
2385 tcphdr_t *tcp;
2386 nat_t **natsave;
2387 int direction;
2388 {
2389 	frentry_t *fr;
2390 	ipnat_t *np;
2391 	ipf_stack_t *ifs = fin->fin_ifs;
2392 
2393 	np = ni->nai_np;
2394 
2395 	COPYIFNAME(fin->fin_ifp, nat->nat_ifnames[0], fin->fin_v);
2396 
2397 #ifdef	IPFILTER_SYNC
2398 	if ((nat->nat_flags & SI_CLONE) == 0)
2399 		nat->nat_sync = ipfsync_new(SMC_NAT, fin, nat);
2400 #endif
2401 
2402 	nat->nat_me = natsave;
2403 	nat->nat_dir = direction;
2404 	nat->nat_ifps[0] = np->in_ifps[0];
2405 	nat->nat_ifps[1] = np->in_ifps[1];
2406 	nat->nat_ptr = np;
2407 	nat->nat_p = fin->fin_p;
2408 	nat->nat_mssclamp = np->in_mssclamp;
2409 	fr = fin->fin_fr;
2410 	nat->nat_fr = fr;
2411 
2412 	if ((np->in_apr != NULL) && ((ni->nai_flags & NAT_SLAVE) == 0))
2413 		if (appr_new(fin, nat) == -1)
2414 			return -1;
2415 
2416 	if (nat_insert(nat, fin->fin_rev, ifs) == 0) {
2417 		if (ifs->ifs_nat_logging)
2418 			nat_log(nat, (u_int)np->in_redir, ifs);
2419 		np->in_use++;
2420 		if (fr != NULL) {
2421 			MUTEX_ENTER(&fr->fr_lock);
2422 			fr->fr_ref++;
2423 			MUTEX_EXIT(&fr->fr_lock);
2424 		}
2425 		return 0;
2426 	}
2427 
2428 	/*
2429 	 * nat_insert failed, so cleanup time...
2430 	 */
2431 	return -1;
2432 }
2433 
2434 
2435 /* ------------------------------------------------------------------------ */
2436 /* Function:   nat_insert                                                   */
2437 /* Returns:    int - 0 == sucess, -1 == failure                             */
2438 /* Parameters: nat(I) - pointer to NAT structure                            */
2439 /*             rev(I) - flag indicating forward/reverse direction of packet */
2440 /* Write Lock: ipf_nat                                                      */
2441 /*                                                                          */
2442 /* Insert a NAT entry into the hash tables for searching and add it to the  */
2443 /* list of active NAT entries.  Adjust global counters when complete.       */
2444 /* ------------------------------------------------------------------------ */
2445 int	nat_insert(nat, rev, ifs)
2446 nat_t	*nat;
2447 int	rev;
2448 ipf_stack_t *ifs;
2449 {
2450 	u_int hv1, hv2;
2451 	nat_t **natp;
2452 
2453 	/*
2454 	 * Try and return an error as early as possible, so calculate the hash
2455 	 * entry numbers first and then proceed.
2456 	 */
2457 	if ((nat->nat_flags & (SI_W_SPORT|SI_W_DPORT)) == 0) {
2458 		hv1 = NAT_HASH_FN(nat->nat_inip.s_addr, nat->nat_inport,
2459 				  0xffffffff);
2460 		hv1 = NAT_HASH_FN(nat->nat_oip.s_addr, hv1 + nat->nat_oport,
2461 				  ifs->ifs_ipf_nattable_sz);
2462 		hv2 = NAT_HASH_FN(nat->nat_outip.s_addr, nat->nat_outport,
2463 				  0xffffffff);
2464 		hv2 = NAT_HASH_FN(nat->nat_oip.s_addr, hv2 + nat->nat_oport,
2465 				  ifs->ifs_ipf_nattable_sz);
2466 	} else {
2467 		hv1 = NAT_HASH_FN(nat->nat_inip.s_addr, 0, 0xffffffff);
2468 		hv1 = NAT_HASH_FN(nat->nat_oip.s_addr, hv1,
2469 				  ifs->ifs_ipf_nattable_sz);
2470 		hv2 = NAT_HASH_FN(nat->nat_outip.s_addr, 0, 0xffffffff);
2471 		hv2 = NAT_HASH_FN(nat->nat_oip.s_addr, hv2,
2472 				  ifs->ifs_ipf_nattable_sz);
2473 	}
2474 
2475 	if (ifs->ifs_nat_stats.ns_bucketlen[0][hv1] >= ifs->ifs_fr_nat_maxbucket ||
2476 	    ifs->ifs_nat_stats.ns_bucketlen[1][hv2] >= ifs->ifs_fr_nat_maxbucket) {
2477 		return -1;
2478 	}
2479 
2480 	nat->nat_hv[0] = hv1;
2481 	nat->nat_hv[1] = hv2;
2482 
2483 	MUTEX_INIT(&nat->nat_lock, "nat entry lock");
2484 
2485 	nat->nat_rev = rev;
2486 	nat->nat_ref = 1;
2487 	nat->nat_bytes[0] = 0;
2488 	nat->nat_pkts[0] = 0;
2489 	nat->nat_bytes[1] = 0;
2490 	nat->nat_pkts[1] = 0;
2491 
2492 	nat->nat_ifnames[0][LIFNAMSIZ - 1] = '\0';
2493 	nat->nat_ifps[0] = fr_resolvenic(nat->nat_ifnames[0], 4, ifs);
2494 
2495 	if (nat->nat_ifnames[1][0] !='\0') {
2496 		nat->nat_ifnames[1][LIFNAMSIZ - 1] = '\0';
2497 		nat->nat_ifps[1] = fr_resolvenic(nat->nat_ifnames[1], 4, ifs);
2498 	} else {
2499 		(void) strncpy(nat->nat_ifnames[1], nat->nat_ifnames[0],
2500 			       LIFNAMSIZ);
2501 		nat->nat_ifnames[1][LIFNAMSIZ - 1] = '\0';
2502 		nat->nat_ifps[1] = nat->nat_ifps[0];
2503 	}
2504 
2505 	nat->nat_next = ifs->ifs_nat_instances;
2506 	nat->nat_pnext = &ifs->ifs_nat_instances;
2507 	if (ifs->ifs_nat_instances)
2508 		ifs->ifs_nat_instances->nat_pnext = &nat->nat_next;
2509 	ifs->ifs_nat_instances = nat;
2510 
2511 	natp = &ifs->ifs_nat_table[0][hv1];
2512 	if (*natp)
2513 		(*natp)->nat_phnext[0] = &nat->nat_hnext[0];
2514 	nat->nat_phnext[0] = natp;
2515 	nat->nat_hnext[0] = *natp;
2516 	*natp = nat;
2517 	ifs->ifs_nat_stats.ns_bucketlen[0][hv1]++;
2518 
2519 	natp = &ifs->ifs_nat_table[1][hv2];
2520 	if (*natp)
2521 		(*natp)->nat_phnext[1] = &nat->nat_hnext[1];
2522 	nat->nat_phnext[1] = natp;
2523 	nat->nat_hnext[1] = *natp;
2524 	*natp = nat;
2525 	ifs->ifs_nat_stats.ns_bucketlen[1][hv2]++;
2526 
2527 	fr_setnatqueue(nat, rev, ifs);
2528 
2529 	ifs->ifs_nat_stats.ns_added++;
2530 	ifs->ifs_nat_stats.ns_inuse++;
2531 	return 0;
2532 }
2533 
2534 
2535 /* ------------------------------------------------------------------------ */
2536 /* Function:    nat_icmperrorlookup                                         */
2537 /* Returns:     nat_t* - point to matching NAT structure                    */
2538 /* Parameters:  fin(I) - pointer to packet information                      */
2539 /*              dir(I) - direction of packet (in/out)                       */
2540 /*                                                                          */
2541 /* Check if the ICMP error message is related to an existing TCP, UDP or    */
2542 /* ICMP query nat entry.  It is assumed that the packet is already of the   */
2543 /* the required length.                                                     */
2544 /* ------------------------------------------------------------------------ */
2545 nat_t *nat_icmperrorlookup(fin, dir)
2546 fr_info_t *fin;
2547 int dir;
2548 {
2549 	int flags = 0, minlen;
2550 	icmphdr_t *orgicmp;
2551 	tcphdr_t *tcp = NULL;
2552 	u_short data[2];
2553 	nat_t *nat;
2554 	ip_t *oip;
2555 	u_int p;
2556 
2557 	/*
2558 	 * Does it at least have the return (basic) IP header ?
2559 	 * Only a basic IP header (no options) should be with an ICMP error
2560 	 * header.  Also, if it's not an error type, then return.
2561 	 */
2562 	if ((fin->fin_hlen != sizeof(ip_t)) || !(fin->fin_flx & FI_ICMPERR))
2563 		return NULL;
2564 
2565 	/*
2566 	 * Check packet size
2567 	 */
2568 	oip = (ip_t *)((char *)fin->fin_dp + 8);
2569 	minlen = IP_HL(oip) << 2;
2570 	if ((minlen < sizeof(ip_t)) ||
2571 	    (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen))
2572 		return NULL;
2573 	/*
2574 	 * Is the buffer big enough for all of it ?  It's the size of the IP
2575 	 * header claimed in the encapsulated part which is of concern.  It
2576 	 * may be too big to be in this buffer but not so big that it's
2577 	 * outside the ICMP packet, leading to TCP deref's causing problems.
2578 	 * This is possible because we don't know how big oip_hl is when we
2579 	 * do the pullup early in fr_check() and thus can't gaurantee it is
2580 	 * all here now.
2581 	 */
2582 #ifdef  _KERNEL
2583 	{
2584 	mb_t *m;
2585 
2586 	m = fin->fin_m;
2587 # if defined(MENTAT)
2588 	if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN > (char *)m->b_wptr)
2589 		return NULL;
2590 # else
2591 	if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN >
2592 	    (char *)fin->fin_ip + M_LEN(m))
2593 		return NULL;
2594 # endif
2595 	}
2596 #endif
2597 
2598 	if (fin->fin_daddr != oip->ip_src.s_addr)
2599 		return NULL;
2600 
2601 	p = oip->ip_p;
2602 	if (p == IPPROTO_TCP)
2603 		flags = IPN_TCP;
2604 	else if (p == IPPROTO_UDP)
2605 		flags = IPN_UDP;
2606 	else if (p == IPPROTO_ICMP) {
2607 		orgicmp = (icmphdr_t *)((char *)oip + (IP_HL(oip) << 2));
2608 
2609 		/* see if this is related to an ICMP query */
2610 		if (nat_icmpquerytype4(orgicmp->icmp_type)) {
2611 			data[0] = fin->fin_data[0];
2612 			data[1] = fin->fin_data[1];
2613 			fin->fin_data[0] = 0;
2614 			fin->fin_data[1] = orgicmp->icmp_id;
2615 
2616 			flags = IPN_ICMPERR|IPN_ICMPQUERY;
2617 			/*
2618 			 * NOTE : dir refers to the direction of the original
2619 			 *        ip packet. By definition the icmp error
2620 			 *        message flows in the opposite direction.
2621 			 */
2622 			if (dir == NAT_INBOUND)
2623 				nat = nat_inlookup(fin, flags, p, oip->ip_dst,
2624 						   oip->ip_src);
2625 			else
2626 				nat = nat_outlookup(fin, flags, p, oip->ip_dst,
2627 						    oip->ip_src);
2628 			fin->fin_data[0] = data[0];
2629 			fin->fin_data[1] = data[1];
2630 			return nat;
2631 		}
2632 	}
2633 
2634 	if (flags & IPN_TCPUDP) {
2635 		minlen += 8;		/* + 64bits of data to get ports */
2636 		if (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen)
2637 			return NULL;
2638 
2639 		data[0] = fin->fin_data[0];
2640 		data[1] = fin->fin_data[1];
2641 		tcp = (tcphdr_t *)((char *)oip + (IP_HL(oip) << 2));
2642 		fin->fin_data[0] = ntohs(tcp->th_dport);
2643 		fin->fin_data[1] = ntohs(tcp->th_sport);
2644 
2645 		if (dir == NAT_INBOUND) {
2646 			nat = nat_inlookup(fin, flags, p, oip->ip_dst,
2647 					   oip->ip_src);
2648 		} else {
2649 			nat = nat_outlookup(fin, flags, p, oip->ip_dst,
2650 					    oip->ip_src);
2651 		}
2652 		fin->fin_data[0] = data[0];
2653 		fin->fin_data[1] = data[1];
2654 		return nat;
2655 	}
2656 	if (dir == NAT_INBOUND)
2657 		return nat_inlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
2658 	else
2659 		return nat_outlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
2660 }
2661 
2662 
2663 /* ------------------------------------------------------------------------ */
2664 /* Function:    nat_icmperror                                               */
2665 /* Returns:     nat_t* - point to matching NAT structure                    */
2666 /* Parameters:  fin(I)    - pointer to packet information                   */
2667 /*              nflags(I) - NAT flags for this packet                       */
2668 /*              dir(I)    - direction of packet (in/out)                    */
2669 /*                                                                          */
2670 /* Fix up an ICMP packet which is an error message for an existing NAT      */
2671 /* session.  This will correct both packet header data and checksums.       */
2672 /*                                                                          */
2673 /* This should *ONLY* be used for incoming ICMP error packets to make sure  */
2674 /* a NAT'd ICMP packet gets correctly recognised.                           */
2675 /* ------------------------------------------------------------------------ */
2676 nat_t *nat_icmperror(fin, nflags, dir)
2677 fr_info_t *fin;
2678 u_int *nflags;
2679 int dir;
2680 {
2681 	u_32_t sum1, sum2, sumd, psum1, psum2, psumd, sumd2;
2682 	struct in_addr in;
2683 	icmphdr_t *icmp, *orgicmp;
2684 	int dlen;
2685 	udphdr_t *udp;
2686 	tcphdr_t *tcp;
2687 	nat_t *nat;
2688 	ip_t *oip;
2689 	if ((fin->fin_flx & (FI_SHORT|FI_FRAGBODY)))
2690 		return NULL;
2691 
2692 	/*
2693 	 * nat_icmperrorlookup() looks up nat entry associated with the
2694 	 * offending IP packet and returns pointer to the entry, or NULL
2695 	 * if packet wasn't natted or for `defective' packets.
2696 	 */
2697 
2698 	if ((fin->fin_v != 4) || !(nat = nat_icmperrorlookup(fin, dir)))
2699 		return NULL;
2700 
2701 	sumd2 = 0;
2702 	*nflags = IPN_ICMPERR;
2703 	icmp = fin->fin_dp;
2704 	oip = (ip_t *)&icmp->icmp_ip;
2705 	udp = (udphdr_t *)((((char *)oip) + (IP_HL(oip) << 2)));
2706 	tcp = (tcphdr_t *)udp;
2707 	dlen = fin->fin_plen - ((char *)udp - (char *)fin->fin_ip);
2708 
2709 	/*
2710 	 * Need to adjust ICMP header to include the real IP#'s and
2711 	 * port #'s.  There are three steps required.
2712 	 *
2713 	 * Step 1
2714 	 * Fix the IP addresses in the offending IP packet and update
2715 	 * ip header checksum to compensate for the change.
2716 	 *
2717 	 * No update needed here for icmp_cksum because the ICMP checksum
2718 	 * is calculated over the complete ICMP packet, which includes the
2719 	 * changed oip IP addresses and oip->ip_sum.  These two changes
2720 	 * cancel each other out (if the delta for the IP address is x,
2721 	 * then the delta for ip_sum is minus x).
2722 	 */
2723 
2724 	if (oip->ip_dst.s_addr == nat->nat_oip.s_addr) {
2725 		sum1 = LONG_SUM(ntohl(oip->ip_src.s_addr));
2726 		in = nat->nat_inip;
2727 		oip->ip_src = in;
2728 	} else {
2729 		sum1 = LONG_SUM(ntohl(oip->ip_dst.s_addr));
2730 		in = nat->nat_outip;
2731 		oip->ip_dst = in;
2732 	}
2733 
2734 	sum2 = LONG_SUM(ntohl(in.s_addr));
2735 	CALC_SUMD(sum1, sum2, sumd);
2736 	fix_datacksum(&oip->ip_sum, sumd);
2737 
2738 	/*
2739 	 * Step 2
2740 	 * Perform other adjustments based on protocol of offending packet.
2741 	 */
2742 
2743 	switch (oip->ip_p) {
2744 		case IPPROTO_TCP :
2745 		case IPPROTO_UDP :
2746 
2747 			/*
2748 			* For offending TCP/UDP IP packets, translate the ports
2749 			* based on the NAT specification.
2750 			*
2751 			* Advance notice : Now it becomes complicated :-)
2752 			*
2753 			* Since the port and IP addresse fields are both part
2754 			* of the TCP/UDP checksum of the offending IP packet,
2755 			* we need to adjust that checksum as well.
2756 			*
2757 			* To further complicate things, the TCP/UDP checksum
2758 			* may not be present.  We must check to see if the
2759 			* length of the data portion is big enough to hold
2760 			* the checksum.  In the UDP case, a test to determine
2761 			* if the checksum is even set is also required.
2762 			*
2763 			* Any changes to an IP address, port or checksum within
2764 			* the ICMP packet requires a change to icmp_cksum.
2765 			*
2766 			* Be extremely careful here ... The change is dependent
2767 			* upon whether or not the TCP/UPD checksum is present.
2768 			*
2769 			* If TCP/UPD checksum is present, the icmp_cksum must
2770 			* compensate for checksum modification resulting from
2771 			* IP address change only.  Port change and resulting
2772 			* data checksum adjustments cancel each other out.
2773 			*
2774 			* If TCP/UDP checksum is not present, icmp_cksum must
2775 			* compensate for port change only.  The IP address
2776 			* change does not modify anything else in this case.
2777 			*/
2778 
2779 			psum1 = 0;
2780 			psum2 = 0;
2781 			psumd = 0;
2782 
2783 			if ((tcp->th_dport == nat->nat_oport) &&
2784 			    (tcp->th_sport != nat->nat_inport)) {
2785 
2786 				/*
2787 				 * Translate the source port.
2788 				 */
2789 
2790 				psum1 = ntohs(tcp->th_sport);
2791 				psum2 = ntohs(nat->nat_inport);
2792 				tcp->th_sport = nat->nat_inport;
2793 
2794 			} else if ((tcp->th_sport == nat->nat_oport) &&
2795 				    (tcp->th_dport != nat->nat_outport)) {
2796 
2797 				/*
2798 				 * Translate the destination port.
2799 				 */
2800 
2801 				psum1 = ntohs(tcp->th_dport);
2802 				psum2 = ntohs(nat->nat_outport);
2803 				tcp->th_dport = nat->nat_outport;
2804 			}
2805 
2806 			if ((oip->ip_p == IPPROTO_TCP) && (dlen >= 18)) {
2807 
2808 				/*
2809 				 * TCP checksum present.
2810 				 *
2811 				 * Adjust data checksum and icmp checksum to
2812 				 * compensate for any IP address change.
2813 				 */
2814 
2815 				sum1 = ntohs(tcp->th_sum);
2816 				fix_datacksum(&tcp->th_sum, sumd);
2817 				sum2 = ntohs(tcp->th_sum);
2818 				sumd2 = sumd << 1;
2819 				CALC_SUMD(sum1, sum2, sumd);
2820 				sumd2 += sumd;
2821 
2822 				/*
2823 				 * Also make data checksum adjustment to
2824 				 * compensate for any port change.
2825 				 */
2826 
2827 				if (psum1 != psum2) {
2828 					CALC_SUMD(psum1, psum2, psumd);
2829 					fix_datacksum(&tcp->th_sum, psumd);
2830 				}
2831 
2832 			} else if ((oip->ip_p == IPPROTO_UDP) &&
2833 				   (dlen >= 8) && (udp->uh_sum != 0)) {
2834 
2835 				/*
2836 				 * The UDP checksum is present and set.
2837 				 *
2838 				 * Adjust data checksum and icmp checksum to
2839 				 * compensate for any IP address change.
2840 				 */
2841 
2842 				sum1 = ntohs(udp->uh_sum);
2843 				fix_datacksum(&udp->uh_sum, sumd);
2844 				sum2 = ntohs(udp->uh_sum);
2845 				sumd2 = sumd << 1;
2846 				CALC_SUMD(sum1, sum2, sumd);
2847 				sumd2 += sumd;
2848 
2849 				/*
2850 				 * Also make data checksum adjustment to
2851 				 * compensate for any port change.
2852 				 */
2853 
2854 				if (psum1 != psum2) {
2855 					CALC_SUMD(psum1, psum2, psumd);
2856 					fix_datacksum(&udp->uh_sum, psumd);
2857 				}
2858 
2859 			} else {
2860 
2861 				/*
2862 				 * Data checksum was not present.
2863 				 *
2864 				 * Compensate for any port change.
2865 				 */
2866 
2867 				CALC_SUMD(psum2, psum1, psumd);
2868 				sumd2 += psumd;
2869 			}
2870 			break;
2871 
2872 		case IPPROTO_ICMP :
2873 
2874 			orgicmp = (icmphdr_t *)udp;
2875 
2876 			if ((nat->nat_dir == NAT_OUTBOUND) &&
2877 			    (orgicmp->icmp_id != nat->nat_inport) &&
2878 			    (dlen >= 8)) {
2879 
2880 				/*
2881 				 * Fix ICMP checksum (of the offening ICMP
2882 				 * query packet) to compensate the change
2883 				 * in the ICMP id of the offending ICMP
2884 				 * packet.
2885 				 *
2886 				 * Since you modify orgicmp->icmp_id with
2887 				 * a delta (say x) and you compensate that
2888 				 * in origicmp->icmp_cksum with a delta
2889 				 * minus x, you don't have to adjust the
2890 				 * overall icmp->icmp_cksum
2891 				 */
2892 
2893 				sum1 = ntohs(orgicmp->icmp_id);
2894 				sum2 = ntohs(nat->nat_inport);
2895 				CALC_SUMD(sum1, sum2, sumd);
2896 				orgicmp->icmp_id = nat->nat_inport;
2897 				fix_datacksum(&orgicmp->icmp_cksum, sumd);
2898 
2899 			} /* nat_dir can't be NAT_INBOUND for icmp queries */
2900 
2901 			break;
2902 
2903 		default :
2904 
2905 			break;
2906 
2907 	} /* switch (oip->ip_p) */
2908 
2909 	/*
2910 	 * Step 3
2911 	 * Make the adjustments to icmp checksum.
2912 	 */
2913 
2914 	if (sumd2 != 0) {
2915 		sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
2916 		sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
2917 		fix_incksum(&icmp->icmp_cksum, sumd2);
2918 	}
2919 	return nat;
2920 }
2921 
2922 
2923 /*
2924  * NB: these lookups don't lock access to the list, it assumed that it has
2925  * already been done!
2926  */
2927 
2928 /* ------------------------------------------------------------------------ */
2929 /* Function:    nat_inlookup                                                */
2930 /* Returns:     nat_t* - NULL == no match,                                  */
2931 /*                       else pointer to matching NAT entry                 */
2932 /* Parameters:  fin(I)    - pointer to packet information                   */
2933 /*              flags(I)  - NAT flags for this packet                       */
2934 /*              p(I)      - protocol for this packet                        */
2935 /*              src(I)    - source IP address                               */
2936 /*              mapdst(I) - destination IP address                          */
2937 /*                                                                          */
2938 /* Lookup a nat entry based on the mapped destination ip address/port and   */
2939 /* real source address/port.  We use this lookup when receiving a packet,   */
2940 /* we're looking for a table entry, based on the destination address.       */
2941 /*                                                                          */
2942 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY.         */
2943 /*                                                                          */
2944 /* NOTE: IT IS ASSUMED THAT ipf_nat IS ONLY HELD WITH A READ LOCK WHEN      */
2945 /*       THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags.             */
2946 /*                                                                          */
2947 /* flags   -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if   */
2948 /*            the packet is of said protocol                                */
2949 /* ------------------------------------------------------------------------ */
2950 nat_t *nat_inlookup(fin, flags, p, src, mapdst)
2951 fr_info_t *fin;
2952 u_int flags, p;
2953 struct in_addr src , mapdst;
2954 {
2955 	u_short sport, dport;
2956 	ipnat_t *ipn;
2957 	u_int sflags;
2958 	nat_t *nat;
2959 	int nflags;
2960 	u_32_t dst;
2961 	void *ifp;
2962 	u_int hv;
2963 	ipf_stack_t *ifs = fin->fin_ifs;
2964 
2965 	if (fin != NULL)
2966 		ifp = fin->fin_ifp;
2967 	else
2968 		ifp = NULL;
2969 	sport = 0;
2970 	dport = 0;
2971 	dst = mapdst.s_addr;
2972 	sflags = flags & NAT_TCPUDPICMP;
2973 
2974 	switch (p)
2975 	{
2976 	case IPPROTO_TCP :
2977 	case IPPROTO_UDP :
2978 		sport = htons(fin->fin_data[0]);
2979 		dport = htons(fin->fin_data[1]);
2980 		break;
2981 	case IPPROTO_ICMP :
2982 		if (flags & IPN_ICMPERR)
2983 			sport = fin->fin_data[1];
2984 		else
2985 			dport = fin->fin_data[1];
2986 		break;
2987 	default :
2988 		break;
2989 	}
2990 
2991 
2992 	if ((flags & SI_WILDP) != 0)
2993 		goto find_in_wild_ports;
2994 
2995 	hv = NAT_HASH_FN(dst, dport, 0xffffffff);
2996 	hv = NAT_HASH_FN(src.s_addr, hv + sport, ifs->ifs_ipf_nattable_sz);
2997 	nat = ifs->ifs_nat_table[1][hv];
2998 	for (; nat; nat = nat->nat_hnext[1]) {
2999 		if (nat->nat_ifps[0] != NULL) {
3000 			if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
3001 				continue;
3002 		} else if (ifp != NULL)
3003 			nat->nat_ifps[0] = ifp;
3004 
3005 		nflags = nat->nat_flags;
3006 
3007 		if (nat->nat_oip.s_addr == src.s_addr &&
3008 		    nat->nat_outip.s_addr == dst &&
3009 		    (((p == 0) &&
3010 		      (sflags == (nat->nat_flags & IPN_TCPUDPICMP)))
3011 		     || (p == nat->nat_p))) {
3012 			switch (p)
3013 			{
3014 #if 0
3015 			case IPPROTO_GRE :
3016 				if (nat->nat_call[1] != fin->fin_data[0])
3017 					continue;
3018 				break;
3019 #endif
3020 			case IPPROTO_ICMP :
3021 				if ((flags & IPN_ICMPERR) != 0) {
3022 					if (nat->nat_outport != sport)
3023 						continue;
3024 				} else {
3025 					if (nat->nat_outport != dport)
3026 						continue;
3027 				}
3028 				break;
3029 			case IPPROTO_TCP :
3030 			case IPPROTO_UDP :
3031 				if (nat->nat_oport != sport)
3032 					continue;
3033 				if (nat->nat_outport != dport)
3034 					continue;
3035 				break;
3036 			default :
3037 				break;
3038 			}
3039 
3040 			ipn = nat->nat_ptr;
3041 			if ((ipn != NULL) && (nat->nat_aps != NULL))
3042 				if (appr_match(fin, nat) != 0)
3043 					continue;
3044 			return nat;
3045 		}
3046 	}
3047 
3048 	/*
3049 	 * So if we didn't find it but there are wildcard members in the hash
3050 	 * table, go back and look for them.  We do this search and update here
3051 	 * because it is modifying the NAT table and we want to do this only
3052 	 * for the first packet that matches.  The exception, of course, is
3053 	 * for "dummy" (FI_IGNORE) lookups.
3054 	 */
3055 find_in_wild_ports:
3056 	if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH))
3057 		return NULL;
3058 	if (ifs->ifs_nat_stats.ns_wilds == 0)
3059 		return NULL;
3060 
3061 	RWLOCK_EXIT(&ifs->ifs_ipf_nat);
3062 
3063 	hv = NAT_HASH_FN(dst, 0, 0xffffffff);
3064 	hv = NAT_HASH_FN(src.s_addr, hv, ifs->ifs_ipf_nattable_sz);
3065 
3066 	WRITE_ENTER(&ifs->ifs_ipf_nat);
3067 
3068 	nat = ifs->ifs_nat_table[1][hv];
3069 	for (; nat; nat = nat->nat_hnext[1]) {
3070 		if (nat->nat_ifps[0] != NULL) {
3071 			if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
3072 				continue;
3073 		} else if (ifp != NULL)
3074 			nat->nat_ifps[0] = ifp;
3075 
3076 		if (nat->nat_p != fin->fin_p)
3077 			continue;
3078 		if (nat->nat_oip.s_addr != src.s_addr ||
3079 		    nat->nat_outip.s_addr != dst)
3080 			continue;
3081 
3082 		nflags = nat->nat_flags;
3083 		if (!(nflags & (NAT_TCPUDP|SI_WILDP)))
3084 			continue;
3085 
3086 		if (nat_wildok(nat, (int)sport, (int)dport, nflags,
3087 			       NAT_INBOUND) == 1) {
3088 			if ((fin->fin_flx & FI_IGNORE) != 0)
3089 				break;
3090 			if ((nflags & SI_CLONE) != 0) {
3091 				nat = fr_natclone(fin, nat);
3092 				if (nat == NULL)
3093 					break;
3094 			} else {
3095 				MUTEX_ENTER(&ifs->ifs_ipf_nat_new);
3096 				ifs->ifs_nat_stats.ns_wilds--;
3097 				MUTEX_EXIT(&ifs->ifs_ipf_nat_new);
3098 			}
3099 			nat->nat_oport = sport;
3100 			nat->nat_outport = dport;
3101 			nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
3102 			nat_tabmove(nat, ifs);
3103 			break;
3104 		}
3105 	}
3106 
3107 	MUTEX_DOWNGRADE(&ifs->ifs_ipf_nat);
3108 
3109 	return nat;
3110 }
3111 
3112 
3113 /* ------------------------------------------------------------------------ */
3114 /* Function:    nat_tabmove                                                 */
3115 /* Returns:     Nil                                                         */
3116 /* Parameters:  nat(I) - pointer to NAT structure                           */
3117 /* Write Lock:  ipf_nat                                                     */
3118 /*                                                                          */
3119 /* This function is only called for TCP/UDP NAT table entries where the     */
3120 /* original was placed in the table without hashing on the ports and we now */
3121 /* want to include hashing on port numbers.                                 */
3122 /* ------------------------------------------------------------------------ */
3123 static void nat_tabmove(nat, ifs)
3124 nat_t *nat;
3125 ipf_stack_t *ifs;
3126 {
3127 	nat_t **natp;
3128 	u_int hv;
3129 
3130 	if (nat->nat_flags & SI_CLONE)
3131 		return;
3132 
3133 	/*
3134 	 * Remove the NAT entry from the old location
3135 	 */
3136 	if (nat->nat_hnext[0])
3137 		nat->nat_hnext[0]->nat_phnext[0] = nat->nat_phnext[0];
3138 	*nat->nat_phnext[0] = nat->nat_hnext[0];
3139 	ifs->ifs_nat_stats.ns_bucketlen[0][nat->nat_hv[0]]--;
3140 
3141 	if (nat->nat_hnext[1])
3142 		nat->nat_hnext[1]->nat_phnext[1] = nat->nat_phnext[1];
3143 	*nat->nat_phnext[1] = nat->nat_hnext[1];
3144 	ifs->ifs_nat_stats.ns_bucketlen[1][nat->nat_hv[1]]--;
3145 
3146 	/*
3147 	 * Add into the NAT table in the new position
3148 	 */
3149 	hv = NAT_HASH_FN(nat->nat_inip.s_addr, nat->nat_inport, 0xffffffff);
3150 	hv = NAT_HASH_FN(nat->nat_oip.s_addr, hv + nat->nat_oport,
3151 			 ifs->ifs_ipf_nattable_sz);
3152 	nat->nat_hv[0] = hv;
3153 	natp = &ifs->ifs_nat_table[0][hv];
3154 	if (*natp)
3155 		(*natp)->nat_phnext[0] = &nat->nat_hnext[0];
3156 	nat->nat_phnext[0] = natp;
3157 	nat->nat_hnext[0] = *natp;
3158 	*natp = nat;
3159 	ifs->ifs_nat_stats.ns_bucketlen[0][hv]++;
3160 
3161 	hv = NAT_HASH_FN(nat->nat_outip.s_addr, nat->nat_outport, 0xffffffff);
3162 	hv = NAT_HASH_FN(nat->nat_oip.s_addr, hv + nat->nat_oport,
3163 			 ifs->ifs_ipf_nattable_sz);
3164 	nat->nat_hv[1] = hv;
3165 	natp = &ifs->ifs_nat_table[1][hv];
3166 	if (*natp)
3167 		(*natp)->nat_phnext[1] = &nat->nat_hnext[1];
3168 	nat->nat_phnext[1] = natp;
3169 	nat->nat_hnext[1] = *natp;
3170 	*natp = nat;
3171 	ifs->ifs_nat_stats.ns_bucketlen[1][hv]++;
3172 }
3173 
3174 
3175 /* ------------------------------------------------------------------------ */
3176 /* Function:    nat_outlookup                                               */
3177 /* Returns:     nat_t* - NULL == no match,                                  */
3178 /*                       else pointer to matching NAT entry                 */
3179 /* Parameters:  fin(I)   - pointer to packet information                    */
3180 /*              flags(I) - NAT flags for this packet                        */
3181 /*              p(I)     - protocol for this packet                         */
3182 /*              src(I)   - source IP address                                */
3183 /*              dst(I)   - destination IP address                           */
3184 /*              rw(I)    - 1 == write lock on ipf_nat held, 0 == read lock. */
3185 /*                                                                          */
3186 /* Lookup a nat entry based on the source 'real' ip address/port and        */
3187 /* destination address/port.  We use this lookup when sending a packet out, */
3188 /* we're looking for a table entry, based on the source address.            */
3189 /*                                                                          */
3190 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY.         */
3191 /*                                                                          */
3192 /* NOTE: IT IS ASSUMED THAT ipf_nat IS ONLY HELD WITH A READ LOCK WHEN      */
3193 /*       THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags.             */
3194 /*                                                                          */
3195 /* flags   -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if   */
3196 /*            the packet is of said protocol                                */
3197 /* ------------------------------------------------------------------------ */
3198 nat_t *nat_outlookup(fin, flags, p, src, dst)
3199 fr_info_t *fin;
3200 u_int flags, p;
3201 struct in_addr src , dst;
3202 {
3203 	u_short sport, dport;
3204 	u_int sflags;
3205 	ipnat_t *ipn;
3206 	u_32_t srcip;
3207 	nat_t *nat;
3208 	int nflags;
3209 	void *ifp;
3210 	u_int hv;
3211 	frentry_t *fr;
3212 	ipf_stack_t *ifs = fin->fin_ifs;
3213 
3214 	fr = fin->fin_fr;
3215 
3216 	if ((fr != NULL) && !(fr->fr_flags & FR_DUP) &&
3217 	    fr->fr_tif.fd_ifp && fr->fr_tif.fd_ifp != (void *)-1)
3218 		ifp = fr->fr_tif.fd_ifp;
3219 	else
3220 		ifp = fin->fin_ifp;
3221 
3222 	srcip = src.s_addr;
3223 	sflags = flags & IPN_TCPUDPICMP;
3224 	sport = 0;
3225 	dport = 0;
3226 
3227 	switch (p)
3228 	{
3229 	case IPPROTO_TCP :
3230 	case IPPROTO_UDP :
3231 		sport = htons(fin->fin_data[0]);
3232 		dport = htons(fin->fin_data[1]);
3233 		break;
3234 	case IPPROTO_ICMP :
3235 		if (flags & IPN_ICMPERR)
3236 			sport = fin->fin_data[1];
3237 		else
3238 			dport = fin->fin_data[1];
3239 		break;
3240 	default :
3241 		break;
3242 	}
3243 
3244 	if ((flags & SI_WILDP) != 0)
3245 		goto find_out_wild_ports;
3246 
3247 	hv = NAT_HASH_FN(srcip, sport, 0xffffffff);
3248 	hv = NAT_HASH_FN(dst.s_addr, hv + dport, ifs->ifs_ipf_nattable_sz);
3249 	nat = ifs->ifs_nat_table[0][hv];
3250 	for (; nat; nat = nat->nat_hnext[0]) {
3251 		if (nat->nat_ifps[1] != NULL) {
3252 			if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
3253 				continue;
3254 		} else if (ifp != NULL)
3255 			nat->nat_ifps[1] = ifp;
3256 
3257 		nflags = nat->nat_flags;
3258 
3259 		if (nat->nat_inip.s_addr == srcip &&
3260 		    nat->nat_oip.s_addr == dst.s_addr &&
3261 		    (((p == 0) && (sflags == (nflags & NAT_TCPUDPICMP)))
3262 		     || (p == nat->nat_p))) {
3263 			switch (p)
3264 			{
3265 #if 0
3266 			case IPPROTO_GRE :
3267 				if (nat->nat_call[1] != fin->fin_data[0])
3268 					continue;
3269 				break;
3270 #endif
3271 			case IPPROTO_TCP :
3272 			case IPPROTO_UDP :
3273 				if (nat->nat_oport != dport)
3274 					continue;
3275 				if (nat->nat_inport != sport)
3276 					continue;
3277 				break;
3278 			default :
3279 				break;
3280 			}
3281 
3282 			ipn = nat->nat_ptr;
3283 			if ((ipn != NULL) && (nat->nat_aps != NULL))
3284 				if (appr_match(fin, nat) != 0)
3285 					continue;
3286 			return nat;
3287 		}
3288 	}
3289 
3290 	/*
3291 	 * So if we didn't find it but there are wildcard members in the hash
3292 	 * table, go back and look for them.  We do this search and update here
3293 	 * because it is modifying the NAT table and we want to do this only
3294 	 * for the first packet that matches.  The exception, of course, is
3295 	 * for "dummy" (FI_IGNORE) lookups.
3296 	 */
3297 find_out_wild_ports:
3298 	if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH))
3299 		return NULL;
3300 	if (ifs->ifs_nat_stats.ns_wilds == 0)
3301 		return NULL;
3302 
3303 	RWLOCK_EXIT(&ifs->ifs_ipf_nat);
3304 
3305 	hv = NAT_HASH_FN(srcip, 0, 0xffffffff);
3306 	hv = NAT_HASH_FN(dst.s_addr, hv, ifs->ifs_ipf_nattable_sz);
3307 
3308 	WRITE_ENTER(&ifs->ifs_ipf_nat);
3309 
3310 	nat = ifs->ifs_nat_table[0][hv];
3311 	for (; nat; nat = nat->nat_hnext[0]) {
3312 		if (nat->nat_ifps[1] != NULL) {
3313 			if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
3314 				continue;
3315 		} else if (ifp != NULL)
3316 			nat->nat_ifps[1] = ifp;
3317 
3318 		if (nat->nat_p != fin->fin_p)
3319 			continue;
3320 		if ((nat->nat_inip.s_addr != srcip) ||
3321 		    (nat->nat_oip.s_addr != dst.s_addr))
3322 			continue;
3323 
3324 		nflags = nat->nat_flags;
3325 		if (!(nflags & (NAT_TCPUDP|SI_WILDP)))
3326 			continue;
3327 
3328 		if (nat_wildok(nat, (int)sport, (int)dport, nflags,
3329 			       NAT_OUTBOUND) == 1) {
3330 			if ((fin->fin_flx & FI_IGNORE) != 0)
3331 				break;
3332 			if ((nflags & SI_CLONE) != 0) {
3333 				nat = fr_natclone(fin, nat);
3334 				if (nat == NULL)
3335 					break;
3336 			} else {
3337 				MUTEX_ENTER(&ifs->ifs_ipf_nat_new);
3338 				ifs->ifs_nat_stats.ns_wilds--;
3339 				MUTEX_EXIT(&ifs->ifs_ipf_nat_new);
3340 			}
3341 			nat->nat_inport = sport;
3342 			nat->nat_oport = dport;
3343 			if (nat->nat_outport == 0)
3344 				nat->nat_outport = sport;
3345 			nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
3346 			nat_tabmove(nat, ifs);
3347 			break;
3348 		}
3349 	}
3350 
3351 	MUTEX_DOWNGRADE(&ifs->ifs_ipf_nat);
3352 
3353 	return nat;
3354 }
3355 
3356 
3357 /* ------------------------------------------------------------------------ */
3358 /* Function:    nat_lookupredir                                             */
3359 /* Returns:     nat_t* - NULL == no match,                                  */
3360 /*                       else pointer to matching NAT entry                 */
3361 /* Parameters:  np(I) - pointer to description of packet to find NAT table  */
3362 /*                      entry for.                                          */
3363 /*                                                                          */
3364 /* Lookup the NAT tables to search for a matching redirect                  */
3365 /* ------------------------------------------------------------------------ */
3366 nat_t *nat_lookupredir(np, ifs)
3367 natlookup_t *np;
3368 ipf_stack_t *ifs;
3369 {
3370 	fr_info_t fi;
3371 	nat_t *nat;
3372 
3373 	bzero((char *)&fi, sizeof(fi));
3374 	if (np->nl_flags & IPN_IN) {
3375 		fi.fin_data[0] = ntohs(np->nl_realport);
3376 		fi.fin_data[1] = ntohs(np->nl_outport);
3377 	} else {
3378 		fi.fin_data[0] = ntohs(np->nl_inport);
3379 		fi.fin_data[1] = ntohs(np->nl_outport);
3380 	}
3381 	if (np->nl_flags & IPN_TCP)
3382 		fi.fin_p = IPPROTO_TCP;
3383 	else if (np->nl_flags & IPN_UDP)
3384 		fi.fin_p = IPPROTO_UDP;
3385 	else if (np->nl_flags & (IPN_ICMPERR|IPN_ICMPQUERY))
3386 		fi.fin_p = IPPROTO_ICMP;
3387 
3388 	fi.fin_ifs = ifs;
3389 	/*
3390 	 * We can do two sorts of lookups:
3391 	 * - IPN_IN: we have the `real' and `out' address, look for `in'.
3392 	 * - default: we have the `in' and `out' address, look for `real'.
3393 	 */
3394 	if (np->nl_flags & IPN_IN) {
3395 		if ((nat = nat_inlookup(&fi, np->nl_flags, fi.fin_p,
3396 					np->nl_realip, np->nl_outip))) {
3397 			np->nl_inip = nat->nat_inip;
3398 			np->nl_inport = nat->nat_inport;
3399 		}
3400 	} else {
3401 		/*
3402 		 * If nl_inip is non null, this is a lookup based on the real
3403 		 * ip address. Else, we use the fake.
3404 		 */
3405 		if ((nat = nat_outlookup(&fi, np->nl_flags, fi.fin_p,
3406 					 np->nl_inip, np->nl_outip))) {
3407 
3408 			if ((np->nl_flags & IPN_FINDFORWARD) != 0) {
3409 				fr_info_t fin;
3410 				bzero((char *)&fin, sizeof(fin));
3411 				fin.fin_p = nat->nat_p;
3412 				fin.fin_data[0] = ntohs(nat->nat_outport);
3413 				fin.fin_data[1] = ntohs(nat->nat_oport);
3414 				fin.fin_ifs = ifs;
3415 				if (nat_inlookup(&fin, np->nl_flags, fin.fin_p,
3416 						 nat->nat_outip,
3417 						 nat->nat_oip) != NULL) {
3418 					np->nl_flags &= ~IPN_FINDFORWARD;
3419 				}
3420 			}
3421 
3422 			np->nl_realip = nat->nat_outip;
3423 			np->nl_realport = nat->nat_outport;
3424 		}
3425  	}
3426 
3427 	return nat;
3428 }
3429 
3430 
3431 /* ------------------------------------------------------------------------ */
3432 /* Function:    nat_match                                                   */
3433 /* Returns:     int - 0 == no match, 1 == match                             */
3434 /* Parameters:  fin(I)   - pointer to packet information                    */
3435 /*              np(I)    - pointer to NAT rule                              */
3436 /*                                                                          */
3437 /* Pull the matching of a packet against a NAT rule out of that complex     */
3438 /* loop inside fr_checknatin() and lay it out properly in its own function. */
3439 /* ------------------------------------------------------------------------ */
3440 static int nat_match(fin, np)
3441 fr_info_t *fin;
3442 ipnat_t *np;
3443 {
3444 	frtuc_t *ft;
3445 
3446 	if (fin->fin_v != 4)
3447 		return 0;
3448 
3449 	if (np->in_p && fin->fin_p != np->in_p)
3450 		return 0;
3451 
3452 	if (fin->fin_out) {
3453 		if (!(np->in_redir & (NAT_MAP|NAT_MAPBLK)))
3454 			return 0;
3455 		if (((fin->fin_fi.fi_saddr & np->in_inmsk) != np->in_inip)
3456 		    ^ ((np->in_flags & IPN_NOTSRC) != 0))
3457 			return 0;
3458 		if (((fin->fin_fi.fi_daddr & np->in_srcmsk) != np->in_srcip)
3459 		    ^ ((np->in_flags & IPN_NOTDST) != 0))
3460 			return 0;
3461 	} else {
3462 		if (!(np->in_redir & NAT_REDIRECT))
3463 			return 0;
3464 		if (((fin->fin_fi.fi_saddr & np->in_srcmsk) != np->in_srcip)
3465 		    ^ ((np->in_flags & IPN_NOTSRC) != 0))
3466 			return 0;
3467 		if (((fin->fin_fi.fi_daddr & np->in_outmsk) != np->in_outip)
3468 		    ^ ((np->in_flags & IPN_NOTDST) != 0))
3469 			return 0;
3470 	}
3471 
3472 	ft = &np->in_tuc;
3473 	if (!(fin->fin_flx & FI_TCPUDP) ||
3474 	    (fin->fin_flx & (FI_SHORT|FI_FRAGBODY))) {
3475 		if (ft->ftu_scmp || ft->ftu_dcmp)
3476 			return 0;
3477 		return 1;
3478 	}
3479 
3480 	return fr_tcpudpchk(fin, ft);
3481 }
3482 
3483 
3484 /* ------------------------------------------------------------------------ */
3485 /* Function:    nat_update                                                  */
3486 /* Returns:     Nil                                                         */
3487 /* Parameters:  nat(I)    - pointer to NAT structure                        */
3488 /*              np(I)     - pointer to NAT rule                             */
3489 /*                                                                          */
3490 /* Updates the lifetime of a NAT table entry for non-TCP packets.  Must be  */
3491 /* called with fin_rev updated - i.e. after calling nat_proto().            */
3492 /* ------------------------------------------------------------------------ */
3493 void nat_update(fin, nat, np)
3494 fr_info_t *fin;
3495 nat_t *nat;
3496 ipnat_t *np;
3497 {
3498 	ipftq_t *ifq, *ifq2;
3499 	ipftqent_t *tqe;
3500 	ipf_stack_t *ifs = fin->fin_ifs;
3501 
3502 	MUTEX_ENTER(&nat->nat_lock);
3503 	tqe = &nat->nat_tqe;
3504 	ifq = tqe->tqe_ifq;
3505 
3506 	/*
3507 	 * We allow over-riding of NAT timeouts from NAT rules, even for
3508 	 * TCP, however, if it is TCP and there is no rule timeout set,
3509 	 * then do not update the timeout here.
3510 	 */
3511 	if (np != NULL)
3512 		ifq2 = np->in_tqehead[fin->fin_rev];
3513 	else
3514 		ifq2 = NULL;
3515 
3516 	if (nat->nat_p == IPPROTO_TCP && ifq2 == NULL) {
3517 		(void) fr_tcp_age(&nat->nat_tqe, fin, ifs->ifs_nat_tqb, 0);
3518 	} else {
3519 		if (ifq2 == NULL) {
3520 			if (nat->nat_p == IPPROTO_UDP)
3521 				ifq2 = &ifs->ifs_nat_udptq;
3522 			else if (nat->nat_p == IPPROTO_ICMP)
3523 				ifq2 = &ifs->ifs_nat_icmptq;
3524 			else
3525 				ifq2 = &ifs->ifs_nat_iptq;
3526 		}
3527 
3528 		fr_movequeue(tqe, ifq, ifq2, ifs);
3529 	}
3530 	MUTEX_EXIT(&nat->nat_lock);
3531 }
3532 
3533 
3534 /* ------------------------------------------------------------------------ */
3535 /* Function:    fr_checknatout                                              */
3536 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
3537 /*                     0 == no packet translation occurred,                 */
3538 /*                     1 == packet was successfully translated.             */
3539 /* Parameters:  fin(I)   - pointer to packet information                    */
3540 /*              passp(I) - pointer to filtering result flags                */
3541 /*                                                                          */
3542 /* Check to see if an outcoming packet should be changed.  ICMP packets are */
3543 /* first checked to see if they match an existing entry (if an error),      */
3544 /* otherwise a search of the current NAT table is made.  If neither results */
3545 /* in a match then a search for a matching NAT rule is made.  Create a new  */
3546 /* NAT entry if a we matched a NAT rule.  Lastly, actually change the       */
3547 /* packet header(s) as required.                                            */
3548 /* ------------------------------------------------------------------------ */
3549 int fr_checknatout(fin, passp)
3550 fr_info_t *fin;
3551 u_32_t *passp;
3552 {
3553 	struct ifnet *ifp, *sifp;
3554 	icmphdr_t *icmp = NULL;
3555 	tcphdr_t *tcp = NULL;
3556 	int rval, natfailed;
3557 	ipnat_t *np = NULL;
3558 	u_int nflags = 0;
3559 	u_32_t ipa, iph;
3560 	int natadd = 1;
3561 	frentry_t *fr;
3562 	nat_t *nat;
3563 	ipf_stack_t *ifs = fin->fin_ifs;
3564 
3565 	if (ifs->ifs_nat_stats.ns_rules == 0 || ifs->ifs_fr_nat_lock != 0)
3566 		return 0;
3567 
3568 	natfailed = 0;
3569 	fr = fin->fin_fr;
3570 	sifp = fin->fin_ifp;
3571 	if ((fr != NULL) && !(fr->fr_flags & FR_DUP) &&
3572 	    fr->fr_tif.fd_ifp && fr->fr_tif.fd_ifp != (void *)-1)
3573 		fin->fin_ifp = fr->fr_tif.fd_ifp;
3574 	ifp = fin->fin_ifp;
3575 
3576 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
3577 		switch (fin->fin_p)
3578 		{
3579 		case IPPROTO_TCP :
3580 			nflags = IPN_TCP;
3581 			break;
3582 		case IPPROTO_UDP :
3583 			nflags = IPN_UDP;
3584 			break;
3585 		case IPPROTO_ICMP :
3586 			icmp = fin->fin_dp;
3587 
3588 			/*
3589 			 * This is an incoming packet, so the destination is
3590 			 * the icmp_id and the source port equals 0
3591 			 */
3592 			if (nat_icmpquerytype4(icmp->icmp_type))
3593 				nflags = IPN_ICMPQUERY;
3594 			break;
3595 		default :
3596 			break;
3597 		}
3598 
3599 		if ((nflags & IPN_TCPUDP))
3600 			tcp = fin->fin_dp;
3601 	}
3602 
3603 	ipa = fin->fin_saddr;
3604 
3605 	READ_ENTER(&ifs->ifs_ipf_nat);
3606 
3607 	if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
3608 	    (nat = nat_icmperror(fin, &nflags, NAT_OUTBOUND)))
3609 		/*EMPTY*/;
3610 	else if ((fin->fin_flx & FI_FRAG) && (nat = fr_nat_knownfrag(fin)))
3611 		natadd = 0;
3612 	else if ((nat = nat_outlookup(fin, nflags|NAT_SEARCH, (u_int)fin->fin_p,
3613 				      fin->fin_src, fin->fin_dst))) {
3614 		nflags = nat->nat_flags;
3615 	} else {
3616 		u_32_t hv, msk, nmsk;
3617 
3618 		/*
3619 		 * If there is no current entry in the nat table for this IP#,
3620 		 * create one for it (if there is a matching rule).
3621 		 */
3622 		RWLOCK_EXIT(&ifs->ifs_ipf_nat);
3623 		msk = 0xffffffff;
3624 		nmsk = ifs->ifs_nat_masks;
3625 		WRITE_ENTER(&ifs->ifs_ipf_nat);
3626 maskloop:
3627 		iph = ipa & htonl(msk);
3628 		hv = NAT_HASH_FN(iph, 0, ifs->ifs_ipf_natrules_sz);
3629 		for (np = ifs->ifs_nat_rules[hv]; np; np = np->in_mnext)
3630 		{
3631 			if ((np->in_ifps[1] && (np->in_ifps[1] != ifp)))
3632 				continue;
3633 			if (np->in_v != fin->fin_v)
3634 				continue;
3635 			if (np->in_p && (np->in_p != fin->fin_p))
3636 				continue;
3637 			if ((np->in_flags & IPN_RF) && !(np->in_flags & nflags))
3638 				continue;
3639 			if (np->in_flags & IPN_FILTER) {
3640 				if (!nat_match(fin, np))
3641 					continue;
3642 			} else if ((ipa & np->in_inmsk) != np->in_inip)
3643 				continue;
3644 
3645 			if ((fr != NULL) &&
3646 			    !fr_matchtag(&np->in_tag, &fr->fr_nattag))
3647 				continue;
3648 
3649 			if (*np->in_plabel != '\0') {
3650 				if (((np->in_flags & IPN_FILTER) == 0) &&
3651 				    (np->in_dport != tcp->th_dport))
3652 					continue;
3653 				if (appr_ok(fin, tcp, np) == 0)
3654 					continue;
3655 			}
3656 
3657 			if ((nat = nat_new(fin, np, NULL, nflags,
3658 					   NAT_OUTBOUND))) {
3659 				np->in_hits++;
3660 				break;
3661 			} else
3662 				natfailed = -1;
3663 		}
3664 		if ((np == NULL) && (nmsk != 0)) {
3665 			while (nmsk) {
3666 				msk <<= 1;
3667 				if (nmsk & 0x80000000)
3668 					break;
3669 				nmsk <<= 1;
3670 			}
3671 			if (nmsk != 0) {
3672 				nmsk <<= 1;
3673 				goto maskloop;
3674 			}
3675 		}
3676 		MUTEX_DOWNGRADE(&ifs->ifs_ipf_nat);
3677 	}
3678 
3679 	if (nat != NULL) {
3680 		rval = fr_natout(fin, nat, natadd, nflags);
3681 		if (rval == 1) {
3682 			MUTEX_ENTER(&nat->nat_lock);
3683 			nat->nat_ref++;
3684 			MUTEX_EXIT(&nat->nat_lock);
3685 			fin->fin_nat = nat;
3686 		}
3687 	} else
3688 		rval = natfailed;
3689 	RWLOCK_EXIT(&ifs->ifs_ipf_nat);
3690 
3691 	if (rval == -1) {
3692 		if (passp != NULL)
3693 			*passp = FR_BLOCK;
3694 		fin->fin_flx |= FI_BADNAT;
3695 	}
3696 	fin->fin_ifp = sifp;
3697 	return rval;
3698 }
3699 
3700 /* ------------------------------------------------------------------------ */
3701 /* Function:    fr_natout                                                   */
3702 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
3703 /*                     1 == packet was successfully translated.             */
3704 /* Parameters:  fin(I)    - pointer to packet information                   */
3705 /*              nat(I)    - pointer to NAT structure                        */
3706 /*              natadd(I) - flag indicating if it is safe to add frag cache */
3707 /*              nflags(I) - NAT flags set for this packet                   */
3708 /*                                                                          */
3709 /* Translate a packet coming "out" on an interface.                         */
3710 /* ------------------------------------------------------------------------ */
3711 int fr_natout(fin, nat, natadd, nflags)
3712 fr_info_t *fin;
3713 nat_t *nat;
3714 int natadd;
3715 u_32_t nflags;
3716 {
3717 	icmphdr_t *icmp;
3718 	u_short *csump;
3719 	u_32_t sumd;
3720 	tcphdr_t *tcp;
3721 	ipnat_t *np;
3722 	int i;
3723 	ipf_stack_t *ifs = fin->fin_ifs;
3724 
3725 #if SOLARIS && defined(_KERNEL)
3726 	net_data_t net_data_p;
3727 	if (fin->fin_v == 4)
3728 		net_data_p = ifs->ifs_ipf_ipv4;
3729 	else
3730 		net_data_p = ifs->ifs_ipf_ipv6;
3731 #endif
3732 
3733 	tcp = NULL;
3734 	icmp = NULL;
3735 	csump = NULL;
3736 	np = nat->nat_ptr;
3737 
3738 	if ((natadd != 0) && (fin->fin_flx & FI_FRAG) && (np != NULL))
3739 		(void) fr_nat_newfrag(fin, 0, nat);
3740 
3741 	MUTEX_ENTER(&nat->nat_lock);
3742 	nat->nat_bytes[1] += fin->fin_plen;
3743 	nat->nat_pkts[1]++;
3744 	MUTEX_EXIT(&nat->nat_lock);
3745 
3746 	/*
3747 	 * Fix up checksums, not by recalculating them, but
3748 	 * simply computing adjustments.
3749 	 * This is only done for STREAMS based IP implementations where the
3750 	 * checksum has already been calculated by IP.  In all other cases,
3751 	 * IPFilter is called before the checksum needs calculating so there
3752 	 * is no call to modify whatever is in the header now.
3753 	 */
3754 	ASSERT(fin->fin_m != NULL);
3755 	if (fin->fin_v == 4 && !NET_IS_HCK_L3_FULL(net_data_p, fin->fin_m)) {
3756 		if (nflags == IPN_ICMPERR) {
3757 			u_32_t s1, s2;
3758 
3759 			s1 = LONG_SUM(ntohl(fin->fin_saddr));
3760 			s2 = LONG_SUM(ntohl(nat->nat_outip.s_addr));
3761 			CALC_SUMD(s1, s2, sumd);
3762 
3763 			fix_outcksum(&fin->fin_ip->ip_sum, sumd);
3764 		}
3765 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
3766     defined(linux) || defined(BRIDGE_IPF)
3767 		else {
3768 			/*
3769 			 * Strictly speaking, this isn't necessary on BSD
3770 			 * kernels because they do checksum calculation after
3771 			 * this code has run BUT if ipfilter is being used
3772 			 * to do NAT as a bridge, that code doesn't exist.
3773 			 */
3774 			if (nat->nat_dir == NAT_OUTBOUND)
3775 				fix_outcksum(&fin->fin_ip->ip_sum,
3776 					    nat->nat_ipsumd);
3777 			else
3778 				fix_incksum(&fin->fin_ip->ip_sum,
3779 				 	   nat->nat_ipsumd);
3780 		}
3781 #endif
3782 	}
3783 
3784 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
3785 		if ((nat->nat_outport != 0) && (nflags & IPN_TCPUDP)) {
3786 			tcp = fin->fin_dp;
3787 
3788 			tcp->th_sport = nat->nat_outport;
3789 			fin->fin_data[0] = ntohs(nat->nat_outport);
3790 		}
3791 
3792 		if ((nat->nat_outport != 0) && (nflags & IPN_ICMPQUERY)) {
3793 			icmp = fin->fin_dp;
3794 			icmp->icmp_id = nat->nat_outport;
3795 		}
3796 
3797 		csump = nat_proto(fin, nat, nflags);
3798 	}
3799 
3800 	fin->fin_ip->ip_src = nat->nat_outip;
3801 
3802 	nat_update(fin, nat, np);
3803 
3804 	/*
3805 	 * The above comments do not hold for layer 4 (or higher) checksums...
3806 	 */
3807 	if (csump != NULL && !NET_IS_HCK_L4_FULL(net_data_p, fin->fin_m)) {
3808 		if (nflags & IPN_TCPUDP &&
3809 	   	    NET_IS_HCK_L4_PART(net_data_p, fin->fin_m))
3810 			sumd = nat->nat_sumd[1];
3811 		else
3812 			sumd = nat->nat_sumd[0];
3813 
3814 		if (nat->nat_dir == NAT_OUTBOUND)
3815 			fix_outcksum(csump, sumd);
3816 		else
3817 			fix_incksum(csump, sumd);
3818 	}
3819 #ifdef	IPFILTER_SYNC
3820 	ipfsync_update(SMC_NAT, fin, nat->nat_sync);
3821 #endif
3822 	/* ------------------------------------------------------------- */
3823 	/* A few quick notes:						 */
3824 	/*	Following are test conditions prior to calling the 	 */
3825 	/*	appr_check routine.					 */
3826 	/*								 */
3827 	/* 	A NULL tcp indicates a non TCP/UDP packet.  When dealing */
3828 	/*	with a redirect rule, we attempt to match the packet's	 */
3829 	/*	source port against in_dport, otherwise	we'd compare the */
3830 	/*	packet's destination.			 		 */
3831 	/* ------------------------------------------------------------- */
3832 	if ((np != NULL) && (np->in_apr != NULL)) {
3833 		i = appr_check(fin, nat);
3834 		if (i == 0)
3835 			i = 1;
3836 	} else
3837 		i = 1;
3838 	ATOMIC_INCL(ifs->ifs_nat_stats.ns_mapped[1]);
3839 	fin->fin_flx |= FI_NATED;
3840 	return i;
3841 }
3842 
3843 
3844 /* ------------------------------------------------------------------------ */
3845 /* Function:    fr_checknatin                                               */
3846 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
3847 /*                     0 == no packet translation occurred,                 */
3848 /*                     1 == packet was successfully translated.             */
3849 /* Parameters:  fin(I)   - pointer to packet information                    */
3850 /*              passp(I) - pointer to filtering result flags                */
3851 /*                                                                          */
3852 /* Check to see if an incoming packet should be changed.  ICMP packets are  */
3853 /* first checked to see if they match an existing entry (if an error),      */
3854 /* otherwise a search of the current NAT table is made.  If neither results */
3855 /* in a match then a search for a matching NAT rule is made.  Create a new  */
3856 /* NAT entry if a we matched a NAT rule.  Lastly, actually change the       */
3857 /* packet header(s) as required.                                            */
3858 /* ------------------------------------------------------------------------ */
3859 int fr_checknatin(fin, passp)
3860 fr_info_t *fin;
3861 u_32_t *passp;
3862 {
3863 	u_int nflags, natadd;
3864 	int rval, natfailed;
3865 	struct ifnet *ifp;
3866 	struct in_addr in;
3867 	icmphdr_t *icmp;
3868 	tcphdr_t *tcp;
3869 	u_short dport;
3870 	ipnat_t *np;
3871 	nat_t *nat;
3872 	u_32_t iph;
3873 	ipf_stack_t *ifs = fin->fin_ifs;
3874 
3875 	if (ifs->ifs_nat_stats.ns_rules == 0 || ifs->ifs_fr_nat_lock != 0)
3876 		return 0;
3877 
3878 	tcp = NULL;
3879 	icmp = NULL;
3880 	dport = 0;
3881 	natadd = 1;
3882 	nflags = 0;
3883 	natfailed = 0;
3884 	ifp = fin->fin_ifp;
3885 
3886 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
3887 		switch (fin->fin_p)
3888 		{
3889 		case IPPROTO_TCP :
3890 			nflags = IPN_TCP;
3891 			break;
3892 		case IPPROTO_UDP :
3893 			nflags = IPN_UDP;
3894 			break;
3895 		case IPPROTO_ICMP :
3896 			icmp = fin->fin_dp;
3897 
3898 			/*
3899 			 * This is an incoming packet, so the destination is
3900 			 * the icmp_id and the source port equals 0
3901 			 */
3902 			if (nat_icmpquerytype4(icmp->icmp_type)) {
3903 				nflags = IPN_ICMPQUERY;
3904 				dport = icmp->icmp_id;
3905 			} break;
3906 		default :
3907 			break;
3908 		}
3909 
3910 		if ((nflags & IPN_TCPUDP)) {
3911 			tcp = fin->fin_dp;
3912 			dport = tcp->th_dport;
3913 		}
3914 	}
3915 
3916 	in = fin->fin_dst;
3917 
3918 	READ_ENTER(&ifs->ifs_ipf_nat);
3919 
3920 	if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
3921 	    (nat = nat_icmperror(fin, &nflags, NAT_INBOUND)))
3922 		/*EMPTY*/;
3923 	else if ((fin->fin_flx & FI_FRAG) && (nat = fr_nat_knownfrag(fin)))
3924 		natadd = 0;
3925 	else if ((nat = nat_inlookup(fin, nflags|NAT_SEARCH, (u_int)fin->fin_p,
3926 				     fin->fin_src, in))) {
3927 		nflags = nat->nat_flags;
3928 	} else {
3929 		u_32_t hv, msk, rmsk;
3930 
3931 		RWLOCK_EXIT(&ifs->ifs_ipf_nat);
3932 		rmsk = ifs->ifs_rdr_masks;
3933 		msk = 0xffffffff;
3934 		WRITE_ENTER(&ifs->ifs_ipf_nat);
3935 		/*
3936 		 * If there is no current entry in the nat table for this IP#,
3937 		 * create one for it (if there is a matching rule).
3938 		 */
3939 maskloop:
3940 		iph = in.s_addr & htonl(msk);
3941 		hv = NAT_HASH_FN(iph, 0, ifs->ifs_ipf_rdrrules_sz);
3942 		for (np = ifs->ifs_rdr_rules[hv]; np; np = np->in_rnext) {
3943 			if (np->in_ifps[0] && (np->in_ifps[0] != ifp))
3944 				continue;
3945 			if (np->in_v != fin->fin_v)
3946 				continue;
3947 			if (np->in_p && (np->in_p != fin->fin_p))
3948 				continue;
3949 			if ((np->in_flags & IPN_RF) && !(np->in_flags & nflags))
3950 				continue;
3951 			if (np->in_flags & IPN_FILTER) {
3952 				if (!nat_match(fin, np))
3953 					continue;
3954 			} else {
3955 				if ((in.s_addr & np->in_outmsk) != np->in_outip)
3956 					continue;
3957 				if (np->in_pmin &&
3958 				    ((ntohs(np->in_pmax) < ntohs(dport)) ||
3959 				     (ntohs(dport) < ntohs(np->in_pmin))))
3960 					continue;
3961 			}
3962 
3963 			if (*np->in_plabel != '\0') {
3964 				if (!appr_ok(fin, tcp, np)) {
3965 					continue;
3966 				}
3967 			}
3968 
3969 			nat = nat_new(fin, np, NULL, nflags, NAT_INBOUND);
3970 			if (nat != NULL) {
3971 				np->in_hits++;
3972 				break;
3973 			} else
3974 				natfailed = -1;
3975 		}
3976 
3977 		if ((np == NULL) && (rmsk != 0)) {
3978 			while (rmsk) {
3979 				msk <<= 1;
3980 				if (rmsk & 0x80000000)
3981 					break;
3982 				rmsk <<= 1;
3983 			}
3984 			if (rmsk != 0) {
3985 				rmsk <<= 1;
3986 				goto maskloop;
3987 			}
3988 		}
3989 		MUTEX_DOWNGRADE(&ifs->ifs_ipf_nat);
3990 	}
3991 	if (nat != NULL) {
3992 		rval = fr_natin(fin, nat, natadd, nflags);
3993 		if (rval == 1) {
3994 			MUTEX_ENTER(&nat->nat_lock);
3995 			nat->nat_ref++;
3996 			MUTEX_EXIT(&nat->nat_lock);
3997 			fin->fin_nat = nat;
3998 			fin->fin_state = nat->nat_state;
3999 		}
4000 	} else
4001 		rval = natfailed;
4002 	RWLOCK_EXIT(&ifs->ifs_ipf_nat);
4003 
4004 	if (rval == -1) {
4005 		if (passp != NULL)
4006 			*passp = FR_BLOCK;
4007 		fin->fin_flx |= FI_BADNAT;
4008 	}
4009 	return rval;
4010 }
4011 
4012 
4013 /* ------------------------------------------------------------------------ */
4014 /* Function:    fr_natin                                                    */
4015 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
4016 /*                     1 == packet was successfully translated.             */
4017 /* Parameters:  fin(I)    - pointer to packet information                   */
4018 /*              nat(I)    - pointer to NAT structure                        */
4019 /*              natadd(I) - flag indicating if it is safe to add frag cache */
4020 /*              nflags(I) - NAT flags set for this packet                   */
4021 /* Locks Held:  ipf_nat (READ)                                              */
4022 /*                                                                          */
4023 /* Translate a packet coming "in" on an interface.                          */
4024 /* ------------------------------------------------------------------------ */
4025 int fr_natin(fin, nat, natadd, nflags)
4026 fr_info_t *fin;
4027 nat_t *nat;
4028 int natadd;
4029 u_32_t nflags;
4030 {
4031 	icmphdr_t *icmp;
4032 	u_short *csump, *csump1;
4033 	u_32_t sumd;
4034 	tcphdr_t *tcp;
4035 	ipnat_t *np;
4036 	int i;
4037 	ipf_stack_t *ifs = fin->fin_ifs;
4038 
4039 #if SOLARIS && defined(_KERNEL)
4040 	net_data_t net_data_p;
4041 	if (fin->fin_v == 4)
4042 		net_data_p = ifs->ifs_ipf_ipv4;
4043 	else
4044 		net_data_p = ifs->ifs_ipf_ipv6;
4045 #endif
4046 
4047 	tcp = NULL;
4048 	csump = NULL;
4049 	np = nat->nat_ptr;
4050 	fin->fin_fr = nat->nat_fr;
4051 
4052 	if (np != NULL) {
4053 		if ((natadd != 0) && (fin->fin_flx & FI_FRAG))
4054 			(void) fr_nat_newfrag(fin, 0, nat);
4055 
4056 	/* ------------------------------------------------------------- */
4057 	/* A few quick notes:						 */
4058 	/*	Following are test conditions prior to calling the 	 */
4059 	/*	appr_check routine.					 */
4060 	/*								 */
4061 	/* 	A NULL tcp indicates a non TCP/UDP packet.  When dealing */
4062 	/*	with a map rule, we attempt to match the packet's	 */
4063 	/*	source port against in_dport, otherwise	we'd compare the */
4064 	/*	packet's destination.			 		 */
4065 	/* ------------------------------------------------------------- */
4066 		if (np->in_apr != NULL) {
4067 			i = appr_check(fin, nat);
4068 			if (i == -1) {
4069 				return -1;
4070 			}
4071 		}
4072 	}
4073 
4074 #ifdef	IPFILTER_SYNC
4075 	ipfsync_update(SMC_NAT, fin, nat->nat_sync);
4076 #endif
4077 
4078 	MUTEX_ENTER(&nat->nat_lock);
4079 	nat->nat_bytes[0] += fin->fin_plen;
4080 	nat->nat_pkts[0]++;
4081 	MUTEX_EXIT(&nat->nat_lock);
4082 
4083 	fin->fin_ip->ip_dst = nat->nat_inip;
4084 	fin->fin_fi.fi_daddr = nat->nat_inip.s_addr;
4085 	if (nflags & IPN_TCPUDP)
4086 		tcp = fin->fin_dp;
4087 
4088 	/*
4089 	 * Fix up checksums, not by recalculating them, but
4090 	 * simply computing adjustments.
4091 	 * Why only do this for some platforms on inbound packets ?
4092 	 * Because for those that it is done, IP processing is yet to happen
4093 	 * and so the IPv4 header checksum has not yet been evaluated.
4094 	 * Perhaps it should always be done for the benefit of things like
4095 	 * fast forwarding (so that it doesn't need to be recomputed) but with
4096 	 * header checksum offloading, perhaps it is a moot point.
4097 	 */
4098 #if !defined(_KERNEL) || defined(MENTAT) || defined(__sgi) || \
4099      defined(__osf__) || defined(linux)
4100 	if (nat->nat_dir == NAT_OUTBOUND)
4101 		fix_incksum(&fin->fin_ip->ip_sum, nat->nat_ipsumd);
4102 	else
4103 		fix_outcksum(&fin->fin_ip->ip_sum, nat->nat_ipsumd);
4104 #endif
4105 
4106 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
4107 		if ((nat->nat_inport != 0) && (nflags & IPN_TCPUDP)) {
4108 			tcp->th_dport = nat->nat_inport;
4109 			fin->fin_data[1] = ntohs(nat->nat_inport);
4110 		}
4111 
4112 
4113 		if ((nat->nat_inport != 0) && (nflags & IPN_ICMPQUERY)) {
4114 			icmp = fin->fin_dp;
4115 
4116 			icmp->icmp_id = nat->nat_inport;
4117 		}
4118 
4119 		csump = nat_proto(fin, nat, nflags);
4120 	}
4121 
4122 	nat_update(fin, nat, np);
4123 
4124 #if SOLARIS && defined(_KERNEL)
4125 	if (nflags & IPN_TCPUDP &&
4126 	    NET_IS_HCK_L4_PART(net_data_p, fin->fin_m)) {
4127 		sumd = nat->nat_sumd[1];
4128 		csump1 = &(fin->fin_m->b_datap->db_struioun.cksum.cksum_val.u16);
4129 		if (csump1 != NULL) {
4130 			if (nat->nat_dir == NAT_OUTBOUND)
4131 				fix_incksum(csump1, sumd);
4132 			else
4133 				fix_outcksum(csump1, sumd);
4134 		}
4135 	} else
4136 #endif
4137 		sumd = nat->nat_sumd[0];
4138 
4139 	/*
4140 	 * Inbound packets always need to have their address adjusted in case
4141 	 * code following this validates it.
4142 	 */
4143 	if (csump != NULL) {
4144 		if (nat->nat_dir == NAT_OUTBOUND)
4145 			fix_incksum(csump, sumd);
4146 		else
4147 			fix_outcksum(csump, sumd);
4148 	}
4149 	ATOMIC_INCL(ifs->ifs_nat_stats.ns_mapped[0]);
4150 	fin->fin_flx |= FI_NATED;
4151 	if (np != NULL && np->in_tag.ipt_num[0] != 0)
4152 		fin->fin_nattag = &np->in_tag;
4153 	return 1;
4154 }
4155 
4156 
4157 /* ------------------------------------------------------------------------ */
4158 /* Function:    nat_proto                                                   */
4159 /* Returns:     u_short* - pointer to transport header checksum to update,  */
4160 /*                         NULL if the transport protocol is not recognised */
4161 /*                         as needing a checksum update.                    */
4162 /* Parameters:  fin(I)    - pointer to packet information                   */
4163 /*              nat(I)    - pointer to NAT structure                        */
4164 /*              nflags(I) - NAT flags set for this packet                   */
4165 /*                                                                          */
4166 /* Return the pointer to the checksum field for each protocol so understood.*/
4167 /* If support for making other changes to a protocol header is required,    */
4168 /* that is not strictly 'address' translation, such as clamping the MSS in  */
4169 /* TCP down to a specific value, then do it from here.                      */
4170 /* ------------------------------------------------------------------------ */
4171 u_short *nat_proto(fin, nat, nflags)
4172 fr_info_t *fin;
4173 nat_t *nat;
4174 u_int nflags;
4175 {
4176 	icmphdr_t *icmp;
4177 	u_short *csump;
4178 	tcphdr_t *tcp;
4179 	udphdr_t *udp;
4180 
4181 	csump = NULL;
4182 	if (fin->fin_out == 0) {
4183 		fin->fin_rev = (nat->nat_dir == NAT_OUTBOUND);
4184 	} else {
4185 		fin->fin_rev = (nat->nat_dir == NAT_INBOUND);
4186 	}
4187 
4188 	switch (fin->fin_p)
4189 	{
4190 	case IPPROTO_TCP :
4191 		tcp = fin->fin_dp;
4192 
4193 		csump = &tcp->th_sum;
4194 
4195 		/*
4196 		 * Do a MSS CLAMPING on a SYN packet,
4197 		 * only deal IPv4 for now.
4198 		 */
4199 		if ((nat->nat_mssclamp != 0) && (tcp->th_flags & TH_SYN) != 0)
4200 			nat_mssclamp(tcp, nat->nat_mssclamp, csump);
4201 
4202 		break;
4203 
4204 	case IPPROTO_UDP :
4205 		udp = fin->fin_dp;
4206 
4207 		if (udp->uh_sum)
4208 			csump = &udp->uh_sum;
4209 		break;
4210 
4211 	case IPPROTO_ICMP :
4212 		icmp = fin->fin_dp;
4213 
4214 		if ((nflags & IPN_ICMPQUERY) != 0) {
4215 			if (icmp->icmp_cksum != 0)
4216 				csump = &icmp->icmp_cksum;
4217 		}
4218 		break;
4219 	}
4220 	return csump;
4221 }
4222 
4223 
4224 /* ------------------------------------------------------------------------ */
4225 /* Function:    fr_natunload                                                */
4226 /* Returns:     Nil                                                         */
4227 /* Parameters:  Nil                                                         */
4228 /*                                                                          */
4229 /* Free all memory used by NAT structures allocated at runtime.             */
4230 /* ------------------------------------------------------------------------ */
4231 void fr_natunload(ifs)
4232 ipf_stack_t *ifs;
4233 {
4234 	ipftq_t *ifq, *ifqnext;
4235 
4236 	(void) nat_clearlist(ifs);
4237 	(void) nat_flushtable(ifs);
4238 
4239 	/*
4240 	 * Proxy timeout queues are not cleaned here because although they
4241 	 * exist on the NAT list, appr_unload is called after fr_natunload
4242 	 * and the proxies actually are responsible for them being created.
4243 	 * Should the proxy timeouts have their own list?  There's no real
4244 	 * justification as this is the only complication.
4245 	 */
4246 	for (ifq = ifs->ifs_nat_utqe; ifq != NULL; ifq = ifqnext) {
4247 		ifqnext = ifq->ifq_next;
4248 		if (((ifq->ifq_flags & IFQF_PROXY) == 0) &&
4249 		    (fr_deletetimeoutqueue(ifq) == 0))
4250 			fr_freetimeoutqueue(ifq, ifs);
4251 	}
4252 
4253 	if (ifs->ifs_nat_table[0] != NULL) {
4254 		KFREES(ifs->ifs_nat_table[0],
4255 		       sizeof(nat_t *) * ifs->ifs_ipf_nattable_sz);
4256 		ifs->ifs_nat_table[0] = NULL;
4257 	}
4258 	if (ifs->ifs_nat_table[1] != NULL) {
4259 		KFREES(ifs->ifs_nat_table[1],
4260 		       sizeof(nat_t *) * ifs->ifs_ipf_nattable_sz);
4261 		ifs->ifs_nat_table[1] = NULL;
4262 	}
4263 	if (ifs->ifs_nat_rules != NULL) {
4264 		KFREES(ifs->ifs_nat_rules,
4265 		       sizeof(ipnat_t *) * ifs->ifs_ipf_natrules_sz);
4266 		ifs->ifs_nat_rules = NULL;
4267 	}
4268 	if (ifs->ifs_rdr_rules != NULL) {
4269 		KFREES(ifs->ifs_rdr_rules,
4270 		       sizeof(ipnat_t *) * ifs->ifs_ipf_rdrrules_sz);
4271 		ifs->ifs_rdr_rules = NULL;
4272 	}
4273 	if (ifs->ifs_maptable != NULL) {
4274 		KFREES(ifs->ifs_maptable,
4275 		       sizeof(hostmap_t *) * ifs->ifs_ipf_hostmap_sz);
4276 		ifs->ifs_maptable = NULL;
4277 	}
4278 	if (ifs->ifs_nat_stats.ns_bucketlen[0] != NULL) {
4279 		KFREES(ifs->ifs_nat_stats.ns_bucketlen[0],
4280 		       sizeof(u_long *) * ifs->ifs_ipf_nattable_sz);
4281 		ifs->ifs_nat_stats.ns_bucketlen[0] = NULL;
4282 	}
4283 	if (ifs->ifs_nat_stats.ns_bucketlen[1] != NULL) {
4284 		KFREES(ifs->ifs_nat_stats.ns_bucketlen[1],
4285 		       sizeof(u_long *) * ifs->ifs_ipf_nattable_sz);
4286 		ifs->ifs_nat_stats.ns_bucketlen[1] = NULL;
4287 	}
4288 
4289 	if (ifs->ifs_fr_nat_maxbucket_reset == 1)
4290 		ifs->ifs_fr_nat_maxbucket = 0;
4291 
4292 	if (ifs->ifs_fr_nat_init == 1) {
4293 		ifs->ifs_fr_nat_init = 0;
4294 		fr_sttab_destroy(ifs->ifs_nat_tqb);
4295 
4296 		RW_DESTROY(&ifs->ifs_ipf_natfrag);
4297 		RW_DESTROY(&ifs->ifs_ipf_nat);
4298 
4299 		MUTEX_DESTROY(&ifs->ifs_ipf_nat_new);
4300 		MUTEX_DESTROY(&ifs->ifs_ipf_natio);
4301 
4302 		MUTEX_DESTROY(&ifs->ifs_nat_udptq.ifq_lock);
4303 		MUTEX_DESTROY(&ifs->ifs_nat_icmptq.ifq_lock);
4304 		MUTEX_DESTROY(&ifs->ifs_nat_iptq.ifq_lock);
4305 	}
4306 }
4307 
4308 
4309 /* ------------------------------------------------------------------------ */
4310 /* Function:    fr_natexpire                                                */
4311 /* Returns:     Nil                                                         */
4312 /* Parameters:  Nil                                                         */
4313 /*                                                                          */
4314 /* Check all of the timeout queues for entries at the top which need to be  */
4315 /* expired.                                                                 */
4316 /* ------------------------------------------------------------------------ */
4317 void fr_natexpire(ifs)
4318 ipf_stack_t *ifs;
4319 {
4320 	ipftq_t *ifq, *ifqnext;
4321 	ipftqent_t *tqe, *tqn;
4322 	int i;
4323 	SPL_INT(s);
4324 
4325 	SPL_NET(s);
4326 	WRITE_ENTER(&ifs->ifs_ipf_nat);
4327 	for (ifq = ifs->ifs_nat_tqb, i = 0; ifq != NULL; ifq = ifq->ifq_next) {
4328 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
4329 			if (tqe->tqe_die > ifs->ifs_fr_ticks)
4330 				break;
4331 			tqn = tqe->tqe_next;
4332 			nat_delete(tqe->tqe_parent, NL_EXPIRE, ifs);
4333 		}
4334 	}
4335 
4336 	for (ifq = ifs->ifs_nat_utqe; ifq != NULL; ifq = ifqnext) {
4337 		ifqnext = ifq->ifq_next;
4338 
4339 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
4340 			if (tqe->tqe_die > ifs->ifs_fr_ticks)
4341 				break;
4342 			tqn = tqe->tqe_next;
4343 			nat_delete(tqe->tqe_parent, NL_EXPIRE, ifs);
4344 		}
4345 	}
4346 
4347 	for (ifq = ifs->ifs_nat_utqe; ifq != NULL; ifq = ifqnext) {
4348 		ifqnext = ifq->ifq_next;
4349 
4350 		if (((ifq->ifq_flags & IFQF_DELETE) != 0) &&
4351 		    (ifq->ifq_ref == 0)) {
4352 			fr_freetimeoutqueue(ifq, ifs);
4353 		}
4354 	}
4355 
4356 	RWLOCK_EXIT(&ifs->ifs_ipf_nat);
4357 	SPL_X(s);
4358 }
4359 
4360 
4361 /* ------------------------------------------------------------------------ */
4362 /* Function:    fr_nataddrsync                                              */
4363 /* Returns:     Nil                                                         */
4364 /* Parameters:  ifp(I) -  pointer to network interface                      */
4365 /*              addr(I) - pointer to new network address                    */
4366 /*                                                                          */
4367 /* Walk through all of the currently active NAT sessions, looking for those */
4368 /* which need to have their translated address updated (where the interface */
4369 /* matches the one passed in) and change it, recalculating the checksum sum */
4370 /* difference too.                                                          */
4371 /* ------------------------------------------------------------------------ */
4372 void fr_nataddrsync(ifp, addr, ifs)
4373 void *ifp;
4374 struct in_addr *addr;
4375 ipf_stack_t *ifs;
4376 {
4377 	u_32_t sum1, sum2, sumd;
4378 	nat_t *nat;
4379 	ipnat_t *np;
4380 	SPL_INT(s);
4381 
4382 	if (ifs->ifs_fr_running <= 0)
4383 		return;
4384 
4385 	SPL_NET(s);
4386 	WRITE_ENTER(&ifs->ifs_ipf_nat);
4387 
4388 	if (ifs->ifs_fr_running <= 0) {
4389 		RWLOCK_EXIT(&ifs->ifs_ipf_nat);
4390 		return;
4391 	}
4392 
4393 	/*
4394 	 * Change IP addresses for NAT sessions for any protocol except TCP
4395 	 * since it will break the TCP connection anyway.  The only rules
4396 	 * which will get changed are those which are "map ... -> 0/32",
4397 	 * where the rule specifies the address is taken from the interface.
4398 	 */
4399 	for (nat = ifs->ifs_nat_instances; nat; nat = nat->nat_next) {
4400 		if (addr != NULL) {
4401 			if (((ifp != NULL) && ifp != (nat->nat_ifps[0])) ||
4402 			    ((nat->nat_flags & IPN_TCP) != 0))
4403 				continue;
4404 			if (((np = nat->nat_ptr) == NULL) ||
4405 			    (np->in_nip || (np->in_outmsk != 0xffffffff)))
4406 				continue;
4407 
4408 			/*
4409 			 * Change the map-to address to be the same as the
4410 			 * new one.
4411 			 */
4412 			sum1 = nat->nat_outip.s_addr;
4413 			nat->nat_outip = *addr;
4414 			sum2 = nat->nat_outip.s_addr;
4415 
4416 		} else if (((ifp == NULL) || (ifp == nat->nat_ifps[0])) &&
4417 		    !(nat->nat_flags & IPN_TCP) && (np = nat->nat_ptr) &&
4418 		    (np->in_outmsk == 0xffffffff) && !np->in_nip) {
4419 			struct in_addr in;
4420 
4421 			/*
4422 			 * Change the map-to address to be the same as the
4423 			 * new one.
4424 			 */
4425 			sum1 = nat->nat_outip.s_addr;
4426 			if (fr_ifpaddr(4, FRI_NORMAL, nat->nat_ifps[0],
4427 				       &in, NULL, ifs) != -1)
4428 				nat->nat_outip = in;
4429 			sum2 = nat->nat_outip.s_addr;
4430 		} else {
4431 			continue;
4432 		}
4433 
4434 		if (sum1 == sum2)
4435 			continue;
4436 		/*
4437 		 * Readjust the checksum adjustment to take into
4438 		 * account the new IP#.
4439 		 */
4440 		CALC_SUMD(sum1, sum2, sumd);
4441 		/* XXX - dont change for TCP when solaris does
4442 		 * hardware checksumming.
4443 		 */
4444 		sumd += nat->nat_sumd[0];
4445 		nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
4446 		nat->nat_sumd[1] = nat->nat_sumd[0];
4447 	}
4448 
4449 	RWLOCK_EXIT(&ifs->ifs_ipf_nat);
4450 	SPL_X(s);
4451 }
4452 
4453 
4454 /* ------------------------------------------------------------------------ */
4455 /* Function:    fr_natifpsync                                               */
4456 /* Returns:     Nil                                                         */
4457 /* Parameters:  action(I) - how we are syncing                              */
4458 /*              ifp(I)    - pointer to network interface                    */
4459 /*              name(I)   - name of interface to sync to                    */
4460 /*                                                                          */
4461 /* This function is used to resync the mapping of interface names and their */
4462 /* respective 'pointers'.  For "action == IPFSYNC_RESYNC", resync all       */
4463 /* interfaces by doing a new lookup of name to 'pointer'.  For "action ==   */
4464 /* IPFSYNC_NEWIFP", treat ifp as the new pointer value associated with      */
4465 /* "name" and for "action == IPFSYNC_OLDIFP", ifp is a pointer for which    */
4466 /* there is no longer any interface associated with it.                     */
4467 /* ------------------------------------------------------------------------ */
4468 void fr_natifpsync(action, ifp, name, ifs)
4469 int action;
4470 void *ifp;
4471 char *name;
4472 ipf_stack_t *ifs;
4473 {
4474 #if defined(_KERNEL) && !defined(MENTAT) && defined(USE_SPL)
4475 	int s;
4476 #endif
4477 	nat_t *nat;
4478 	ipnat_t *n;
4479 
4480 	if (ifs->ifs_fr_running <= 0)
4481 		return;
4482 
4483 	SPL_NET(s);
4484 	WRITE_ENTER(&ifs->ifs_ipf_nat);
4485 
4486 	if (ifs->ifs_fr_running <= 0) {
4487 		RWLOCK_EXIT(&ifs->ifs_ipf_nat);
4488 		return;
4489 	}
4490 
4491 	switch (action)
4492 	{
4493 	case IPFSYNC_RESYNC :
4494 		for (nat = ifs->ifs_nat_instances; nat; nat = nat->nat_next) {
4495 			if ((ifp == nat->nat_ifps[0]) ||
4496 			    (nat->nat_ifps[0] == (void *)-1)) {
4497 				nat->nat_ifps[0] =
4498 				    fr_resolvenic(nat->nat_ifnames[0], 4, ifs);
4499 			}
4500 
4501 			if ((ifp == nat->nat_ifps[1]) ||
4502 			    (nat->nat_ifps[1] == (void *)-1)) {
4503 				nat->nat_ifps[1] =
4504 				    fr_resolvenic(nat->nat_ifnames[1], 4, ifs);
4505 			}
4506 		}
4507 
4508 		for (n = ifs->ifs_nat_list; (n != NULL); n = n->in_next) {
4509 			if (n->in_ifps[0] == ifp ||
4510 			    n->in_ifps[0] == (void *)-1) {
4511 				n->in_ifps[0] =
4512 				    fr_resolvenic(n->in_ifnames[0], 4, ifs);
4513 			}
4514 			if (n->in_ifps[1] == ifp ||
4515 			    n->in_ifps[1] == (void *)-1) {
4516 				n->in_ifps[1] =
4517 				    fr_resolvenic(n->in_ifnames[1], 4, ifs);
4518 			}
4519 		}
4520 		break;
4521 	case IPFSYNC_NEWIFP :
4522 		for (nat = ifs->ifs_nat_instances; nat; nat = nat->nat_next) {
4523 			if (!strncmp(name, nat->nat_ifnames[0],
4524 				     sizeof(nat->nat_ifnames[0])))
4525 				nat->nat_ifps[0] = ifp;
4526 			if (!strncmp(name, nat->nat_ifnames[1],
4527 				     sizeof(nat->nat_ifnames[1])))
4528 				nat->nat_ifps[1] = ifp;
4529 		}
4530 		for (n = ifs->ifs_nat_list; (n != NULL); n = n->in_next) {
4531 			if (!strncmp(name, n->in_ifnames[0],
4532 				     sizeof(n->in_ifnames[0])))
4533 				n->in_ifps[0] = ifp;
4534 			if (!strncmp(name, n->in_ifnames[1],
4535 				     sizeof(n->in_ifnames[1])))
4536 				n->in_ifps[1] = ifp;
4537 		}
4538 		break;
4539 	case IPFSYNC_OLDIFP :
4540 		for (nat = ifs->ifs_nat_instances; nat; nat = nat->nat_next) {
4541 			if (ifp == nat->nat_ifps[0])
4542 				nat->nat_ifps[0] = (void *)-1;
4543 			if (ifp == nat->nat_ifps[1])
4544 				nat->nat_ifps[1] = (void *)-1;
4545 		}
4546 		for (n = ifs->ifs_nat_list; (n != NULL); n = n->in_next) {
4547 			if (n->in_ifps[0] == ifp)
4548 				n->in_ifps[0] = (void *)-1;
4549 			if (n->in_ifps[1] == ifp)
4550 				n->in_ifps[1] = (void *)-1;
4551 		}
4552 		break;
4553 	}
4554 	RWLOCK_EXIT(&ifs->ifs_ipf_nat);
4555 	SPL_X(s);
4556 }
4557 
4558 
4559 /* ------------------------------------------------------------------------ */
4560 /* Function:    nat_icmpquerytype4                                          */
4561 /* Returns:     int - 1 == success, 0 == failure                            */
4562 /* Parameters:  icmptype(I) - ICMP type number                              */
4563 /*                                                                          */
4564 /* Tests to see if the ICMP type number passed is a query/response type or  */
4565 /* not.                                                                     */
4566 /* ------------------------------------------------------------------------ */
4567 static INLINE int nat_icmpquerytype4(icmptype)
4568 int icmptype;
4569 {
4570 
4571 	/*
4572 	 * For the ICMP query NAT code, it is essential that both the query
4573 	 * and the reply match on the NAT rule. Because the NAT structure
4574 	 * does not keep track of the icmptype, and a single NAT structure
4575 	 * is used for all icmp types with the same src, dest and id, we
4576 	 * simply define the replies as queries as well. The funny thing is,
4577 	 * altough it seems silly to call a reply a query, this is exactly
4578 	 * as it is defined in the IPv4 specification
4579 	 */
4580 
4581 	switch (icmptype)
4582 	{
4583 
4584 	case ICMP_ECHOREPLY:
4585 	case ICMP_ECHO:
4586 	/* route aedvertisement/solliciation is currently unsupported: */
4587 	/* it would require rewriting the ICMP data section            */
4588 	case ICMP_TSTAMP:
4589 	case ICMP_TSTAMPREPLY:
4590 	case ICMP_IREQ:
4591 	case ICMP_IREQREPLY:
4592 	case ICMP_MASKREQ:
4593 	case ICMP_MASKREPLY:
4594 		return 1;
4595 	default:
4596 		return 0;
4597 	}
4598 }
4599 
4600 
4601 /* ------------------------------------------------------------------------ */
4602 /* Function:    nat_log                                                     */
4603 /* Returns:     Nil                                                         */
4604 /* Parameters:  nat(I)  - pointer to NAT structure                          */
4605 /*              type(I) - type of log entry to create                       */
4606 /*                                                                          */
4607 /* Creates a NAT log entry.                                                 */
4608 /* ------------------------------------------------------------------------ */
4609 void nat_log(nat, type, ifs)
4610 struct nat *nat;
4611 u_int type;
4612 ipf_stack_t *ifs;
4613 {
4614 #ifdef	IPFILTER_LOG
4615 # ifndef LARGE_NAT
4616 	struct ipnat *np;
4617 	int rulen;
4618 # endif
4619 	struct natlog natl;
4620 	void *items[1];
4621 	size_t sizes[1];
4622 	int types[1];
4623 
4624 	natl.nl_inip = nat->nat_inip;
4625 	natl.nl_outip = nat->nat_outip;
4626 	natl.nl_origip = nat->nat_oip;
4627 	natl.nl_bytes[0] = nat->nat_bytes[0];
4628 	natl.nl_bytes[1] = nat->nat_bytes[1];
4629 	natl.nl_pkts[0] = nat->nat_pkts[0];
4630 	natl.nl_pkts[1] = nat->nat_pkts[1];
4631 	natl.nl_origport = nat->nat_oport;
4632 	natl.nl_inport = nat->nat_inport;
4633 	natl.nl_outport = nat->nat_outport;
4634 	natl.nl_p = nat->nat_p;
4635 	natl.nl_type = type;
4636 	natl.nl_rule = -1;
4637 # ifndef LARGE_NAT
4638 	if (nat->nat_ptr != NULL) {
4639 		for (rulen = 0, np = ifs->ifs_nat_list; np;
4640 		     np = np->in_next, rulen++)
4641 			if (np == nat->nat_ptr) {
4642 				natl.nl_rule = rulen;
4643 				break;
4644 			}
4645 	}
4646 # endif
4647 	items[0] = &natl;
4648 	sizes[0] = sizeof(natl);
4649 	types[0] = 0;
4650 
4651 	(void) ipllog(IPL_LOGNAT, NULL, items, sizes, types, 1, ifs);
4652 #endif
4653 }
4654 
4655 
4656 #if defined(__OpenBSD__)
4657 /* ------------------------------------------------------------------------ */
4658 /* Function:    nat_ifdetach                                                */
4659 /* Returns:     Nil                                                         */
4660 /* Parameters:  ifp(I) - pointer to network interface                       */
4661 /*                                                                          */
4662 /* Compatibility interface for OpenBSD to trigger the correct updating of   */
4663 /* interface references within IPFilter.                                    */
4664 /* ------------------------------------------------------------------------ */
4665 void nat_ifdetach(ifp, ifs)
4666 void *ifp;
4667 ipf_stack_t *ifs;
4668 {
4669 	frsync(ifp, ifs);
4670 	return;
4671 }
4672 #endif
4673 
4674 
4675 /* ------------------------------------------------------------------------ */
4676 /* Function:    fr_ipnatderef                                               */
4677 /* Returns:     Nil                                                         */
4678 /* Parameters:  isp(I) - pointer to pointer to NAT rule                     */
4679 /* Write Locks: ipf_nat                                                     */
4680 /*                                                                          */
4681 /* ------------------------------------------------------------------------ */
4682 void fr_ipnatderef(inp, ifs)
4683 ipnat_t **inp;
4684 ipf_stack_t *ifs;
4685 {
4686 	ipnat_t *in;
4687 
4688 	in = *inp;
4689 	*inp = NULL;
4690 	in->in_space++;
4691 	in->in_use--;
4692 	if (in->in_use == 0 && (in->in_flags & IPN_DELETE)) {
4693 		if (in->in_apr)
4694 			appr_free(in->in_apr);
4695 		KFREE(in);
4696 		ifs->ifs_nat_stats.ns_rules--;
4697 #ifdef notdef
4698 #if SOLARIS
4699 		if (ifs->ifs_nat_stats.ns_rules == 0)
4700 			ifs->ifs_pfil_delayed_copy = 1;
4701 #endif
4702 #endif
4703 	}
4704 }
4705 
4706 
4707 /* ------------------------------------------------------------------------ */
4708 /* Function:    fr_natderef                                                 */
4709 /* Returns:     Nil                                                         */
4710 /* Parameters:  isp(I) - pointer to pointer to NAT table entry              */
4711 /*                                                                          */
4712 /* Decrement the reference counter for this NAT table entry and free it if  */
4713 /* there are no more things using it.                                       */
4714 /* ------------------------------------------------------------------------ */
4715 void fr_natderef(natp, ifs)
4716 nat_t **natp;
4717 ipf_stack_t *ifs;
4718 {
4719 	nat_t *nat;
4720 
4721 	nat = *natp;
4722 	*natp = NULL;
4723 	WRITE_ENTER(&ifs->ifs_ipf_nat);
4724 	nat->nat_ref--;
4725 	if (nat->nat_ref == 0)
4726 	    nat_delete(nat, NL_EXPIRE, ifs);
4727 	RWLOCK_EXIT(&ifs->ifs_ipf_nat);
4728 }
4729 
4730 
4731 /* ------------------------------------------------------------------------ */
4732 /* Function:    fr_natclone                                                 */
4733 /* Returns:     ipstate_t* - NULL == cloning failed,                        */
4734 /*                           else pointer to new state structure            */
4735 /* Parameters:  fin(I) - pointer to packet information                      */
4736 /*              is(I)  - pointer to master state structure                  */
4737 /* Write Lock:  ipf_nat                                                     */
4738 /*                                                                          */
4739 /* Create a "duplcate" state table entry from the master.                   */
4740 /* ------------------------------------------------------------------------ */
4741 static nat_t *fr_natclone(fin, nat)
4742 fr_info_t *fin;
4743 nat_t *nat;
4744 {
4745 	frentry_t *fr;
4746 	nat_t *clone;
4747 	ipnat_t *np;
4748 	ipf_stack_t *ifs = fin->fin_ifs;
4749 
4750 	KMALLOC(clone, nat_t *);
4751 	if (clone == NULL)
4752 		return NULL;
4753 	bcopy((char *)nat, (char *)clone, sizeof(*clone));
4754 
4755 	MUTEX_NUKE(&clone->nat_lock);
4756 
4757 	clone->nat_aps = NULL;
4758 	/*
4759 	 * Initialize all these so that nat_delete() doesn't cause a crash.
4760 	 */
4761 	clone->nat_tqe.tqe_pnext = NULL;
4762 	clone->nat_tqe.tqe_next = NULL;
4763 	clone->nat_tqe.tqe_ifq = NULL;
4764 	clone->nat_tqe.tqe_parent = clone;
4765 
4766 	clone->nat_flags &= ~SI_CLONE;
4767 	clone->nat_flags |= SI_CLONED;
4768 
4769 	if (clone->nat_hm)
4770 		clone->nat_hm->hm_ref++;
4771 
4772 	if (nat_insert(clone, fin->fin_rev, ifs) == -1) {
4773 		KFREE(clone);
4774 		return NULL;
4775 	}
4776 	np = clone->nat_ptr;
4777 	if (np != NULL) {
4778 		if (ifs->ifs_nat_logging)
4779 			nat_log(clone, (u_int)np->in_redir, ifs);
4780 		np->in_use++;
4781 	}
4782 	fr = clone->nat_fr;
4783 	if (fr != NULL) {
4784 		MUTEX_ENTER(&fr->fr_lock);
4785 		fr->fr_ref++;
4786 		MUTEX_EXIT(&fr->fr_lock);
4787 	}
4788 
4789 	/*
4790 	 * Because the clone is created outside the normal loop of things and
4791 	 * TCP has special needs in terms of state, initialise the timeout
4792 	 * state of the new NAT from here.
4793 	 */
4794 	if (clone->nat_p == IPPROTO_TCP) {
4795 		(void) fr_tcp_age(&clone->nat_tqe, fin, ifs->ifs_nat_tqb,
4796 				  clone->nat_flags);
4797 	}
4798 #ifdef	IPFILTER_SYNC
4799 	clone->nat_sync = ipfsync_new(SMC_NAT, fin, clone);
4800 #endif
4801 	if (ifs->ifs_nat_logging)
4802 		nat_log(clone, NL_CLONE, ifs);
4803 	return clone;
4804 }
4805 
4806 
4807 /* ------------------------------------------------------------------------ */
4808 /* Function:   nat_wildok                                                   */
4809 /* Returns:    int - 1 == packet's ports match wildcards                    */
4810 /*                   0 == packet's ports don't match wildcards              */
4811 /* Parameters: nat(I)   - NAT entry                                         */
4812 /*             sport(I) - source port                                       */
4813 /*             dport(I) - destination port                                  */
4814 /*             flags(I) - wildcard flags                                    */
4815 /*             dir(I)   - packet direction                                  */
4816 /*                                                                          */
4817 /* Use NAT entry and packet direction to determine which combination of     */
4818 /* wildcard flags should be used.                                           */
4819 /* ------------------------------------------------------------------------ */
4820 static INLINE int nat_wildok(nat, sport, dport, flags, dir)
4821 nat_t *nat;
4822 int sport;
4823 int dport;
4824 int flags;
4825 int dir;
4826 {
4827 	/*
4828 	 * When called by       dir is set to
4829 	 * nat_inlookup         NAT_INBOUND (0)
4830 	 * nat_outlookup        NAT_OUTBOUND (1)
4831 	 *
4832 	 * We simply combine the packet's direction in dir with the original
4833 	 * "intended" direction of that NAT entry in nat->nat_dir to decide
4834 	 * which combination of wildcard flags to allow.
4835 	 */
4836 
4837 	switch ((dir << 1) | nat->nat_dir)
4838 	{
4839 	case 3: /* outbound packet / outbound entry */
4840 		if (((nat->nat_inport == sport) ||
4841 		    (flags & SI_W_SPORT)) &&
4842 		    ((nat->nat_oport == dport) ||
4843 		    (flags & SI_W_DPORT)))
4844 			return 1;
4845 		break;
4846 	case 2: /* outbound packet / inbound entry */
4847 		if (((nat->nat_outport == sport) ||
4848 		    (flags & SI_W_DPORT)) &&
4849 		    ((nat->nat_oport == dport) ||
4850 		    (flags & SI_W_SPORT)))
4851 			return 1;
4852 		break;
4853 	case 1: /* inbound packet / outbound entry */
4854 		if (((nat->nat_oport == sport) ||
4855 		    (flags & SI_W_DPORT)) &&
4856 		    ((nat->nat_outport == dport) ||
4857 		    (flags & SI_W_SPORT)))
4858 			return 1;
4859 		break;
4860 	case 0: /* inbound packet / inbound entry */
4861 		if (((nat->nat_oport == sport) ||
4862 		    (flags & SI_W_SPORT)) &&
4863 		    ((nat->nat_outport == dport) ||
4864 		    (flags & SI_W_DPORT)))
4865 			return 1;
4866 		break;
4867 	default:
4868 		break;
4869 	}
4870 
4871 	return(0);
4872 }
4873 
4874 
4875 /* ------------------------------------------------------------------------ */
4876 /* Function:    nat_mssclamp                                                */
4877 /* Returns:     Nil                                                         */
4878 /* Parameters:  tcp(I)    - pointer to TCP header                           */
4879 /*              maxmss(I) - value to clamp the TCP MSS to                   */
4880 /*              csump(I)  - pointer to TCP checksum                         */
4881 /*                                                                          */
4882 /* Check for MSS option and clamp it if necessary.  If found and changed,   */
4883 /* then the TCP header checksum will be updated to reflect the change in    */
4884 /* the MSS.                                                                 */
4885 /* ------------------------------------------------------------------------ */
4886 static void nat_mssclamp(tcp, maxmss, csump)
4887 tcphdr_t *tcp;
4888 u_32_t maxmss;
4889 u_short *csump;
4890 {
4891 	u_char *cp, *ep, opt;
4892 	int hlen, advance;
4893 	u_32_t mss, sumd;
4894 
4895 	hlen = TCP_OFF(tcp) << 2;
4896 	if (hlen > sizeof(*tcp)) {
4897 		cp = (u_char *)tcp + sizeof(*tcp);
4898 		ep = (u_char *)tcp + hlen;
4899 
4900 		while (cp < ep) {
4901 			opt = cp[0];
4902 			if (opt == TCPOPT_EOL)
4903 				break;
4904 			else if (opt == TCPOPT_NOP) {
4905 				cp++;
4906 				continue;
4907 			}
4908 
4909 			if (cp + 1 >= ep)
4910 				break;
4911 			advance = cp[1];
4912 			if ((cp + advance > ep) || (advance <= 0))
4913 				break;
4914 			switch (opt)
4915 			{
4916 			case TCPOPT_MAXSEG:
4917 				if (advance != 4)
4918 					break;
4919 				mss = cp[2] * 256 + cp[3];
4920 				if (mss > maxmss) {
4921 					cp[2] = maxmss / 256;
4922 					cp[3] = maxmss & 0xff;
4923 					CALC_SUMD(mss, maxmss, sumd);
4924 					fix_outcksum(csump, sumd);
4925 				}
4926 				break;
4927 			default:
4928 				/* ignore unknown options */
4929 				break;
4930 			}
4931 
4932 			cp += advance;
4933 		}
4934 	}
4935 }
4936 
4937 
4938 /* ------------------------------------------------------------------------ */
4939 /* Function:    fr_setnatqueue                                              */
4940 /* Returns:     Nil                                                         */
4941 /* Parameters:  nat(I)- pointer to NAT structure                            */
4942 /*              rev(I) - forward(0) or reverse(1) direction                 */
4943 /* Locks:       ipf_nat (read or write)                                     */
4944 /*                                                                          */
4945 /* Put the NAT entry on its default queue entry, using rev as a helped in   */
4946 /* determining which queue it should be placed on.                          */
4947 /* ------------------------------------------------------------------------ */
4948 void fr_setnatqueue(nat, rev, ifs)
4949 nat_t *nat;
4950 int rev;
4951 ipf_stack_t *ifs;
4952 {
4953 	ipftq_t *oifq, *nifq;
4954 
4955 	if (nat->nat_ptr != NULL)
4956 		nifq = nat->nat_ptr->in_tqehead[rev];
4957 	else
4958 		nifq = NULL;
4959 
4960 	if (nifq == NULL) {
4961 		switch (nat->nat_p)
4962 		{
4963 		case IPPROTO_UDP :
4964 			nifq = &ifs->ifs_nat_udptq;
4965 			break;
4966 		case IPPROTO_ICMP :
4967 			nifq = &ifs->ifs_nat_icmptq;
4968 			break;
4969 		case IPPROTO_TCP :
4970 			nifq = ifs->ifs_nat_tqb + nat->nat_tqe.tqe_state[rev];
4971 			break;
4972 		default :
4973 			nifq = &ifs->ifs_nat_iptq;
4974 			break;
4975 		}
4976 	}
4977 
4978 	oifq = nat->nat_tqe.tqe_ifq;
4979 	/*
4980 	 * If it's currently on a timeout queue, move it from one queue to
4981 	 * another, else put it on the end of the newly determined queue.
4982 	 */
4983 	if (oifq != NULL)
4984 		fr_movequeue(&nat->nat_tqe, oifq, nifq, ifs);
4985 	else
4986 		fr_queueappend(&nat->nat_tqe, nifq, nat, ifs);
4987 	return;
4988 }
4989 
4990 /* Function:    nat_getnext                                                 */
4991 /* Returns:     int - 0 == ok, else error                                   */
4992 /* Parameters:  t(I)   - pointer to ipftoken structure                      */
4993 /*              itp(I) - pointer to ipfgeniter_t structure                  */
4994 /*                                                                          */
4995 /* Fetch the next nat/ipnat structure pointer from the linked list and      */
4996 /* copy it out to the storage space pointed to by itp_data.  The next item  */
4997 /* in the list to look at is put back in the ipftoken struture.             */
4998 /* If we call ipf_freetoken, the accompanying pointer is set to NULL because*/
4999 /* ipf_freetoken will call a deref function for us and we dont want to call */
5000 /* that twice (second time would be in the second switch statement below.   */
5001 /* ------------------------------------------------------------------------ */
5002 static int nat_getnext(t, itp, ifs)
5003 ipftoken_t *t;
5004 ipfgeniter_t *itp;
5005 ipf_stack_t *ifs;
5006 {
5007 	hostmap_t *hm, *nexthm = NULL, zerohm;
5008 	ipnat_t *ipn, *nextipnat = NULL, zeroipn;
5009 	nat_t *nat, *nextnat = NULL, zeronat;
5010 	int error = 0;
5011 
5012 	READ_ENTER(&ifs->ifs_ipf_nat);
5013 	switch (itp->igi_type)
5014 	{
5015 	case IPFGENITER_HOSTMAP :
5016 		hm = t->ipt_data;
5017 		if (hm == NULL) {
5018 			nexthm = ifs->ifs_ipf_hm_maplist;
5019 		} else {
5020 			nexthm = hm->hm_hnext;
5021 		}
5022 		if (nexthm != NULL) {
5023 			if (nexthm->hm_hnext == NULL) {
5024 				t->ipt_alive = 0;
5025 				ipf_unlinktoken(t, ifs);
5026 				KFREE(t);
5027 			} else {
5028 				/*MUTEX_ENTER(&nexthm->hm_lock);*/
5029 				nexthm->hm_ref++;
5030 				/*MUTEX_EXIT(&nextipnat->hm_lock);*/
5031 			}
5032 
5033 		} else {
5034 			bzero(&zerohm, sizeof(zerohm));
5035 			nexthm = &zerohm;
5036 			ipf_freetoken(t, ifs);
5037 		}
5038 		break;
5039 
5040 	case IPFGENITER_IPNAT :
5041 		ipn = t->ipt_data;
5042 		if (ipn == NULL) {
5043 			nextipnat = ifs->ifs_nat_list;
5044 		} else {
5045 			nextipnat = ipn->in_next;
5046 		}
5047 		if (nextipnat != NULL) {
5048 			if (nextipnat->in_next == NULL) {
5049 				t->ipt_alive = 0;
5050 				ipf_unlinktoken(t, ifs);
5051 				KFREE(t);
5052 			} else {
5053 				/* MUTEX_ENTER(&nextipnat->in_lock); */
5054 				nextipnat->in_use++;
5055 				/* MUTEX_EXIT(&nextipnat->in_lock); */
5056 			}
5057 		} else {
5058 			bzero(&zeroipn, sizeof(zeroipn));
5059 			nextipnat = &zeroipn;
5060 			ipf_freetoken(t, ifs);
5061 		}
5062 		break;
5063 
5064 	case IPFGENITER_NAT :
5065 		nat = t->ipt_data;
5066 		if (nat == NULL) {
5067 			nextnat = ifs->ifs_nat_instances;
5068 		} else {
5069 			nextnat = nat->nat_next;
5070 		}
5071 		if (nextnat != NULL) {
5072 			if (nextnat->nat_next == NULL) {
5073 				t->ipt_alive = 0;
5074 				ipf_unlinktoken(t, ifs);
5075 				KFREE(t);
5076 			} else {
5077 				MUTEX_ENTER(&nextnat->nat_lock);
5078 				nextnat->nat_ref++;
5079 				MUTEX_EXIT(&nextnat->nat_lock);
5080 			}
5081 		} else {
5082 			bzero(&zeronat, sizeof(zeronat));
5083 			nextnat = &zeronat;
5084 			ipf_freetoken(t, ifs);
5085 		}
5086 		break;
5087 	}
5088 
5089 	RWLOCK_EXIT(&ifs->ifs_ipf_nat);
5090 
5091 	switch (itp->igi_type)
5092 	{
5093 	case IPFGENITER_HOSTMAP :
5094 		if (hm != NULL) {
5095 			WRITE_ENTER(&ifs->ifs_ipf_nat);
5096 			fr_hostmapderef(&hm);
5097 			RWLOCK_EXIT(&ifs->ifs_ipf_nat);
5098 		}
5099 		if (nexthm->hm_hnext != NULL)
5100 			t->ipt_data = nexthm;
5101 		error = COPYOUT(nexthm, itp->igi_data, sizeof(*nexthm));
5102 		if (error != 0)
5103 			error = EFAULT;
5104 		break;
5105 
5106 	case IPFGENITER_IPNAT :
5107 		if (ipn != NULL)
5108 			fr_ipnatderef(&ipn, ifs);
5109 		if (nextipnat->in_next != NULL)
5110 			t->ipt_data = nextipnat;
5111 		error = COPYOUT(nextipnat, itp->igi_data, sizeof(*nextipnat));
5112 		if (error != 0)
5113 			error = EFAULT;
5114 		break;
5115 
5116 	case IPFGENITER_NAT :
5117 		if (nat != NULL)
5118 			fr_natderef(&nat, ifs);
5119 		if (nextnat->nat_next != NULL)
5120 			t->ipt_data = nextnat;
5121 		error = COPYOUT(nextnat, itp->igi_data, sizeof(*nextnat));
5122 		if (error != 0)
5123 			error = EFAULT;
5124 		break;
5125 	}
5126 
5127 	return error;
5128 }
5129 
5130 
5131 /* ------------------------------------------------------------------------ */
5132 /* Function:    nat_iterator                                                */
5133 /* Returns:     int - 0 == ok, else error                                   */
5134 /* Parameters:  token(I) - pointer to ipftoken structure                    */
5135 /*              itp(I) - pointer to ipfgeniter_t structure                  */
5136 /*                                                                          */
5137 /* This function acts as a handler for the SIOCGENITER ioctls that use a    */
5138 /* generic structure to iterate through a list.  There are three different  */
5139 /* linked lists of NAT related information to go through: NAT rules, active */
5140 /* NAT mappings and the NAT fragment cache.                                 */
5141 /* ------------------------------------------------------------------------ */
5142 static int nat_iterator(token, itp, ifs)
5143 ipftoken_t *token;
5144 ipfgeniter_t *itp;
5145 ipf_stack_t *ifs;
5146 {
5147 	int error;
5148 
5149 	if (itp->igi_data == NULL)
5150 		return EFAULT;
5151 
5152 	token->ipt_subtype = itp->igi_type;
5153 
5154 	switch (itp->igi_type)
5155 	{
5156 	case IPFGENITER_HOSTMAP :
5157 	case IPFGENITER_IPNAT :
5158 	case IPFGENITER_NAT :
5159 		error = nat_getnext(token, itp, ifs);
5160 		break;
5161 	case IPFGENITER_NATFRAG :
5162 		error = fr_nextfrag(token, itp, &ifs->ifs_ipfr_natlist,
5163 				    &ifs->ifs_ipfr_nattail,
5164 				    &ifs->ifs_ipf_natfrag, ifs);
5165 		break;
5166 	default :
5167 		error = EINVAL;
5168 		break;
5169 	}
5170 
5171 	return error;
5172 }
5173