xref: /freebsd/sys/netpfil/ipfilter/netinet/ip_nat.c (revision 6535e9308a26e17023831fe68fb71d2febf2a002)
1 
2 /*
3  * Copyright (C) 2012 by Darren Reed.
4  *
5  * See the IPFILTER.LICENCE file for details on licencing.
6  */
7 #if defined(KERNEL) || defined(_KERNEL)
8 # undef KERNEL
9 # undef _KERNEL
10 # define        KERNEL	1
11 # define        _KERNEL	1
12 #endif
13 #include <sys/errno.h>
14 #include <sys/types.h>
15 #include <sys/param.h>
16 #include <sys/time.h>
17 #include <sys/file.h>
18 #if defined(_KERNEL) && \
19     (defined(__NetBSD_Version) && (__NetBSD_Version >= 399002000))
20 # include <sys/kauth.h>
21 #endif
22 #if !defined(_KERNEL)
23 # include <stdio.h>
24 # include <string.h>
25 # include <stdlib.h>
26 # define KERNEL
27 # ifdef _OpenBSD__
28 struct file;
29 # endif
30 # include <sys/uio.h>
31 # undef KERNEL
32 #endif
33 #if defined(_KERNEL) && defined(__FreeBSD__)
34 # include <sys/filio.h>
35 # include <sys/fcntl.h>
36 #else
37 # include <sys/ioctl.h>
38 #endif
39 # include <sys/fcntl.h>
40 # include <sys/protosw.h>
41 #include <sys/socket.h>
42 #if defined(_KERNEL)
43 # include <sys/systm.h>
44 # if defined(__FreeBSD__)
45 #  include <sys/jail.h>
46 # endif
47 # if !defined(__SVR4)
48 #  include <sys/mbuf.h>
49 # endif
50 #endif
51 #if defined(__SVR4)
52 # include <sys/filio.h>
53 # include <sys/byteorder.h>
54 # ifdef KERNEL
55 #  include <sys/dditypes.h>
56 # endif
57 # include <sys/stream.h>
58 # include <sys/kmem.h>
59 #endif
60 #if defined(__FreeBSD__)
61 # include <sys/queue.h>
62 #endif
63 #include <net/if.h>
64 #if defined(__FreeBSD__)
65 # include <net/if_var.h>
66 #endif
67 #ifdef sun
68 # include <net/af.h>
69 #endif
70 #include <netinet/in.h>
71 #include <netinet/in_systm.h>
72 #include <netinet/ip.h>
73 
74 #ifdef RFC1825
75 # include <vpn/md5.h>
76 # include <vpn/ipsec.h>
77 extern struct ifnet vpnif;
78 #endif
79 
80 # include <netinet/ip_var.h>
81 #include <netinet/tcp.h>
82 #include <netinet/udp.h>
83 #include <netinet/ip_icmp.h>
84 #include "netinet/ip_compat.h"
85 #include <netinet/tcpip.h>
86 #include "netinet/ipl.h"
87 #include "netinet/ip_fil.h"
88 #include "netinet/ip_nat.h"
89 #include "netinet/ip_frag.h"
90 #include "netinet/ip_state.h"
91 #include "netinet/ip_proxy.h"
92 #include "netinet/ip_lookup.h"
93 #include "netinet/ip_dstlist.h"
94 #include "netinet/ip_sync.h"
95 #if defined(__FreeBSD__)
96 # include <sys/malloc.h>
97 #endif
98 #ifdef HAS_SYS_MD5_H
99 # include <sys/md5.h>
100 #else
101 # include "md5.h"
102 #endif
103 /* END OF INCLUDES */
104 
105 #undef	SOCKADDR_IN
106 #define	SOCKADDR_IN	struct sockaddr_in
107 
108 
109 
110 #define	NATFSUM(n,v,f)	((v) == 4 ? (n)->f.in4.s_addr : (n)->f.i6[0] + \
111 			 (n)->f.i6[1] + (n)->f.i6[2] + (n)->f.i6[3])
112 #define	NBUMP(x)	softn->(x)++
113 #define	NBUMPD(x, y)	do { \
114 				softn->x.y++; \
115 				DT(y); \
116 			} while (0)
117 #define	NBUMPSIDE(y,x)	softn->ipf_nat_stats.ns_side[y].x++
118 #define	NBUMPSIDED(y,x)	do { softn->ipf_nat_stats.ns_side[y].x++; \
119 			     DT(x); } while (0)
120 #define	NBUMPSIDEX(y,x,z) \
121 			do { softn->ipf_nat_stats.ns_side[y].x++; \
122 			     DT(z); } while (0)
123 #define	NBUMPSIDEDF(y,x)do { softn->ipf_nat_stats.ns_side[y].x++; \
124 			     DT1(x, fr_info_t *, fin); } while (0)
125 
126 static ipftuneable_t ipf_nat_tuneables[] = {
127 	/* nat */
128 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_lock) },
129 		"nat_lock",	0,	1,
130 		stsizeof(ipf_nat_softc_t, ipf_nat_lock),
131 		IPFT_RDONLY,		NULL,	NULL },
132 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_sz) },
133 		"nat_table_size", 1,	0x7fffffff,
134 		stsizeof(ipf_nat_softc_t, ipf_nat_table_sz),
135 		0,			NULL,	ipf_nat_rehash },
136 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_max) },
137 		"nat_table_max", 1,	0x7fffffff,
138 		stsizeof(ipf_nat_softc_t, ipf_nat_table_max),
139 		0,			NULL,	NULL },
140 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_maprules_sz) },
141 		"nat_rules_size", 1,	0x7fffffff,
142 		stsizeof(ipf_nat_softc_t, ipf_nat_maprules_sz),
143 		0,			NULL,	ipf_nat_rehash_rules },
144 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_rdrrules_sz) },
145 		"rdr_rules_size", 1,	0x7fffffff,
146 		stsizeof(ipf_nat_softc_t, ipf_nat_rdrrules_sz),
147 		0,			NULL,	ipf_nat_rehash_rules },
148 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_hostmap_sz) },
149 		"hostmap_size",	1,	0x7fffffff,
150 		stsizeof(ipf_nat_softc_t, ipf_nat_hostmap_sz),
151 		0,			NULL,	ipf_nat_hostmap_rehash },
152 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_maxbucket) },
153 		"nat_maxbucket",1,	0x7fffffff,
154 		stsizeof(ipf_nat_softc_t, ipf_nat_maxbucket),
155 		0,			NULL,	NULL },
156 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_logging) },
157 		"nat_logging",	0,	1,
158 		stsizeof(ipf_nat_softc_t, ipf_nat_logging),
159 		0,			NULL,	NULL },
160 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_doflush) },
161 		"nat_doflush",	0,	1,
162 		stsizeof(ipf_nat_softc_t, ipf_nat_doflush),
163 		0,			NULL,	NULL },
164 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_wm_low) },
165 		"nat_table_wm_low",	1,	99,
166 		stsizeof(ipf_nat_softc_t, ipf_nat_table_wm_low),
167 		0,			NULL,	NULL },
168 	{ { (void *)offsetof(ipf_nat_softc_t, ipf_nat_table_wm_high) },
169 		"nat_table_wm_high",	2,	100,
170 		stsizeof(ipf_nat_softc_t, ipf_nat_table_wm_high),
171 		0,			NULL,	NULL },
172 	{ { 0 },
173 		NULL,			0,	0,
174 		0,
175 		0,			NULL,	NULL }
176 };
177 
178 /* ======================================================================== */
179 /* How the NAT is organised and works.                                      */
180 /*                                                                          */
181 /* Inside (interface y) NAT       Outside (interface x)                     */
182 /* -------------------- -+- -------------------------------------           */
183 /* Packet going          |   out, processed by ipf_nat_checkout() for x     */
184 /* ------------>         |   ------------>                                  */
185 /* src=10.1.1.1          |   src=192.1.1.1                                  */
186 /*                       |                                                  */
187 /*                       |   in, processed by ipf_nat_checkin() for x       */
188 /* <------------         |   <------------                                  */
189 /* dst=10.1.1.1          |   dst=192.1.1.1                                  */
190 /* -------------------- -+- -------------------------------------           */
191 /* ipf_nat_checkout() - changes ip_src and if required, sport               */
192 /*             - creates a new mapping, if required.                        */
193 /* ipf_nat_checkin()  - changes ip_dst and if required, dport               */
194 /*                                                                          */
195 /* In the NAT table, internal source is recorded as "in" and externally     */
196 /* seen as "out".                                                           */
197 /* ======================================================================== */
198 
199 
200 #if SOLARIS && !defined(INSTANCES)
201 extern	int		pfil_delayed_copy;
202 #endif
203 
204 static	int	ipf_nat_flush_entry(ipf_main_softc_t *, void *);
205 static	int	ipf_nat_getent(ipf_main_softc_t *, caddr_t, int);
206 static	int	ipf_nat_getsz(ipf_main_softc_t *, caddr_t, int);
207 static	int	ipf_nat_putent(ipf_main_softc_t *, caddr_t, int);
208 static	void	ipf_nat_addmap(ipf_nat_softc_t *, ipnat_t *);
209 static	void	ipf_nat_addrdr(ipf_nat_softc_t *, ipnat_t *);
210 static	int	ipf_nat_builddivertmp(ipf_nat_softc_t *, ipnat_t *);
211 static	int	ipf_nat_clearlist(ipf_main_softc_t *, ipf_nat_softc_t *);
212 static	int	ipf_nat_cmp_rules(ipnat_t *, ipnat_t *);
213 static	int	ipf_nat_decap(fr_info_t *, nat_t *);
214 static	void	ipf_nat_delrule(ipf_main_softc_t *, ipf_nat_softc_t *,
215 				     ipnat_t *, int);
216 static	int	ipf_nat_extraflush(ipf_main_softc_t *, ipf_nat_softc_t *, int);
217 static	int	ipf_nat_finalise(fr_info_t *, nat_t *);
218 static	int	ipf_nat_flushtable(ipf_main_softc_t *, ipf_nat_softc_t *);
219 static	int	ipf_nat_getnext(ipf_main_softc_t *, ipftoken_t *,
220 				     ipfgeniter_t *, ipfobj_t *);
221 static	int	ipf_nat_gettable(ipf_main_softc_t *, ipf_nat_softc_t *,
222 				      char *);
223 static	hostmap_t *ipf_nat_hostmap(ipf_nat_softc_t *, ipnat_t *,
224 					struct in_addr, struct in_addr,
225 					struct in_addr, u_32_t);
226 static	int	ipf_nat_icmpquerytype(int);
227 static	int	ipf_nat_iterator(ipf_main_softc_t *, ipftoken_t *,
228 				      ipfgeniter_t *, ipfobj_t *);
229 static	int	ipf_nat_match(fr_info_t *, ipnat_t *);
230 static	int	ipf_nat_matcharray(nat_t *, int *, u_long);
231 static	int	ipf_nat_matchflush(ipf_main_softc_t *, ipf_nat_softc_t *,
232 					caddr_t);
233 static	void	ipf_nat_mssclamp(tcphdr_t *, u_32_t, fr_info_t *,
234 				      u_short *);
235 static	int	ipf_nat_newmap(fr_info_t *, nat_t *, natinfo_t *);
236 static	int	ipf_nat_newdivert(fr_info_t *, nat_t *, natinfo_t *);
237 static	int	ipf_nat_newrdr(fr_info_t *, nat_t *, natinfo_t *);
238 static	int	ipf_nat_newrewrite(fr_info_t *, nat_t *, natinfo_t *);
239 static	int	ipf_nat_nextaddr(fr_info_t *, nat_addr_t *, u_32_t *,
240 				      u_32_t *);
241 static	int	ipf_nat_nextaddrinit(ipf_main_softc_t *, char *,
242 					  nat_addr_t *, int, void *);
243 static	int	ipf_nat_resolverule(ipf_main_softc_t *, ipnat_t *);
244 static	int	ipf_nat_ruleaddrinit(ipf_main_softc_t *,
245 					  ipf_nat_softc_t *, ipnat_t *);
246 static	void	ipf_nat_rule_fini(ipf_main_softc_t *, ipnat_t *);
247 static	int	ipf_nat_rule_init(ipf_main_softc_t *, ipf_nat_softc_t *,
248 				       ipnat_t *);
249 static	int	ipf_nat_siocaddnat(ipf_main_softc_t *, ipf_nat_softc_t *,
250 					ipnat_t *, int);
251 static	void	ipf_nat_siocdelnat(ipf_main_softc_t *, ipf_nat_softc_t *,
252 					ipnat_t *, int);
253 static	void	ipf_nat_tabmove(ipf_nat_softc_t *, nat_t *);
254 
255 /* ------------------------------------------------------------------------ */
256 /* Function:    ipf_nat_main_load                                           */
257 /* Returns:     int - 0 == success, -1 == failure                           */
258 /* Parameters:  Nil                                                         */
259 /*                                                                          */
260 /* The only global NAT structure that needs to be initialised is the filter */
261 /* rule that is used with blocking packets.                                 */
262 /* ------------------------------------------------------------------------ */
263 int
ipf_nat_main_load(void)264 ipf_nat_main_load(void)
265 {
266 
267 	return (0);
268 }
269 
270 
271 /* ------------------------------------------------------------------------ */
272 /* Function:    ipf_nat_main_unload                                         */
273 /* Returns:     int - 0 == success, -1 == failure                           */
274 /* Parameters:  Nil                                                         */
275 /*                                                                          */
276 /* A null-op function that exists as a placeholder so that the flow in      */
277 /* other functions is obvious.                                              */
278 /* ------------------------------------------------------------------------ */
279 int
ipf_nat_main_unload(void)280 ipf_nat_main_unload(void)
281 {
282 	return (0);
283 }
284 
285 
286 /* ------------------------------------------------------------------------ */
287 /* Function:    ipf_nat_soft_create                                         */
288 /* Returns:     void * - NULL = failure, else pointer to NAT context        */
289 /* Parameters:  softc(I) - pointer to soft context main structure           */
290 /*                                                                          */
291 /* Allocate the initial soft context structure for NAT and populate it with */
292 /* some default values. Creating the tables is left until we call _init so  */
293 /* that sizes can be changed before we get under way.                       */
294 /* ------------------------------------------------------------------------ */
295 void *
ipf_nat_soft_create(ipf_main_softc_t * softc)296 ipf_nat_soft_create(ipf_main_softc_t *softc)
297 {
298 	ipf_nat_softc_t *softn;
299 
300 	KMALLOC(softn, ipf_nat_softc_t *);
301 	if (softn == NULL)
302 		return (NULL);
303 
304 	bzero((char *)softn, sizeof(*softn));
305 
306 	softn->ipf_nat_tune = ipf_tune_array_copy(softn,
307 						  sizeof(ipf_nat_tuneables),
308 						  ipf_nat_tuneables);
309 	if (softn->ipf_nat_tune == NULL) {
310 		ipf_nat_soft_destroy(softc, softn);
311 		return (NULL);
312 	}
313 	if (ipf_tune_array_link(softc, softn->ipf_nat_tune) == -1) {
314 		ipf_nat_soft_destroy(softc, softn);
315 		return (NULL);
316 	}
317 
318 	softn->ipf_nat_list_tail = &softn->ipf_nat_list;
319 
320 	if (softc->ipf_large_nat) {
321 	softn->ipf_nat_table_max = NAT_TABLE_MAX_LARGE;
322 	softn->ipf_nat_table_sz = NAT_TABLE_SZ_LARGE;
323 	softn->ipf_nat_maprules_sz = NAT_SIZE_LARGE;
324 	softn->ipf_nat_rdrrules_sz = RDR_SIZE_LARGE;
325 	softn->ipf_nat_hostmap_sz = HOSTMAP_SIZE_LARGE;
326 	} else {
327 	softn->ipf_nat_table_max = NAT_TABLE_MAX_NORMAL;
328 	softn->ipf_nat_table_sz = NAT_TABLE_SZ_NORMAL;
329 	softn->ipf_nat_maprules_sz = NAT_SIZE_NORMAL;
330 	softn->ipf_nat_rdrrules_sz = RDR_SIZE_NORMAL;
331 	softn->ipf_nat_hostmap_sz = HOSTMAP_SIZE_NORMAL;
332 	}
333 	softn->ipf_nat_doflush = 0;
334 #ifdef  IPFILTER_LOG
335 	softn->ipf_nat_logging = 1;
336 #else
337 	softn->ipf_nat_logging = 0;
338 #endif
339 
340 	softn->ipf_nat_defage = DEF_NAT_AGE;
341 	softn->ipf_nat_defipage = IPF_TTLVAL(60);
342 	softn->ipf_nat_deficmpage = IPF_TTLVAL(3);
343 	softn->ipf_nat_table_wm_high = 99;
344 	softn->ipf_nat_table_wm_low = 90;
345 
346 	return (softn);
347 }
348 
349 /* ------------------------------------------------------------------------ */
350 /* Function:    ipf_nat_soft_destroy                                        */
351 /* Returns:     Nil                                                         */
352 /* Parameters:  softc(I) - pointer to soft context main structure           */
353 /*                                                                          */
354 /* ------------------------------------------------------------------------ */
355 void
ipf_nat_soft_destroy(ipf_main_softc_t * softc,void * arg)356 ipf_nat_soft_destroy(ipf_main_softc_t *softc, void *arg)
357 {
358 	ipf_nat_softc_t *softn = arg;
359 
360 	if (softn->ipf_nat_tune != NULL) {
361 		ipf_tune_array_unlink(softc, softn->ipf_nat_tune);
362 		KFREES(softn->ipf_nat_tune, sizeof(ipf_nat_tuneables));
363 		softn->ipf_nat_tune = NULL;
364 	}
365 
366 	KFREE(softn);
367 }
368 
369 
370 /* ------------------------------------------------------------------------ */
371 /* Function:    ipf_nat_init                                                */
372 /* Returns:     int - 0 == success, -1 == failure                           */
373 /* Parameters:  softc(I) - pointer to soft context main structure           */
374 /*                                                                          */
375 /* Initialise all of the NAT locks, tables and other structures.            */
376 /* ------------------------------------------------------------------------ */
377 int
ipf_nat_soft_init(ipf_main_softc_t * softc,void * arg)378 ipf_nat_soft_init(ipf_main_softc_t *softc, void *arg)
379 {
380 	ipf_nat_softc_t *softn = arg;
381 	ipftq_t *tq;
382 	int i;
383 
384 	KMALLOCS(softn->ipf_nat_table[0], nat_t **, \
385 		 sizeof(nat_t *) * softn->ipf_nat_table_sz);
386 
387 	if (softn->ipf_nat_table[0] != NULL) {
388 		bzero((char *)softn->ipf_nat_table[0],
389 		      softn->ipf_nat_table_sz * sizeof(nat_t *));
390 	} else {
391 		return (-1);
392 	}
393 
394 	KMALLOCS(softn->ipf_nat_table[1], nat_t **, \
395 		 sizeof(nat_t *) * softn->ipf_nat_table_sz);
396 
397 	if (softn->ipf_nat_table[1] != NULL) {
398 		bzero((char *)softn->ipf_nat_table[1],
399 		      softn->ipf_nat_table_sz * sizeof(nat_t *));
400 	} else {
401 		return (-2);
402 	}
403 
404 	KMALLOCS(softn->ipf_nat_map_rules, ipnat_t **, \
405 		 sizeof(ipnat_t *) * softn->ipf_nat_maprules_sz);
406 
407 	if (softn->ipf_nat_map_rules != NULL) {
408 		bzero((char *)softn->ipf_nat_map_rules,
409 		      softn->ipf_nat_maprules_sz * sizeof(ipnat_t *));
410 	} else {
411 		return (-3);
412 	}
413 
414 	KMALLOCS(softn->ipf_nat_rdr_rules, ipnat_t **, \
415 		 sizeof(ipnat_t *) * softn->ipf_nat_rdrrules_sz);
416 
417 	if (softn->ipf_nat_rdr_rules != NULL) {
418 		bzero((char *)softn->ipf_nat_rdr_rules,
419 		      softn->ipf_nat_rdrrules_sz * sizeof(ipnat_t *));
420 	} else {
421 		return (-4);
422 	}
423 
424 	KMALLOCS(softn->ipf_hm_maptable, hostmap_t **, \
425 		 sizeof(hostmap_t *) * softn->ipf_nat_hostmap_sz);
426 
427 	if (softn->ipf_hm_maptable != NULL) {
428 		bzero((char *)softn->ipf_hm_maptable,
429 		      sizeof(hostmap_t *) * softn->ipf_nat_hostmap_sz);
430 	} else {
431 		return (-5);
432 	}
433 	softn->ipf_hm_maplist = NULL;
434 
435 	KMALLOCS(softn->ipf_nat_stats.ns_side[0].ns_bucketlen, u_int *,
436 		 softn->ipf_nat_table_sz * sizeof(u_int));
437 
438 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen == NULL) {
439 		return (-6);
440 	}
441 	bzero((char *)softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
442 	      softn->ipf_nat_table_sz * sizeof(u_int));
443 
444 	KMALLOCS(softn->ipf_nat_stats.ns_side[1].ns_bucketlen, u_int *,
445 		 softn->ipf_nat_table_sz * sizeof(u_int));
446 
447 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen == NULL) {
448 		return (-7);
449 	}
450 
451 	bzero((char *)softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
452 	      softn->ipf_nat_table_sz * sizeof(u_int));
453 
454 	if (softn->ipf_nat_maxbucket == 0) {
455 		for (i = softn->ipf_nat_table_sz; i > 0; i >>= 1)
456 			softn->ipf_nat_maxbucket++;
457 		softn->ipf_nat_maxbucket *= 2;
458 	}
459 
460 	ipf_sttab_init(softc, softn->ipf_nat_tcptq);
461 	/*
462 	 * Increase this because we may have "keep state" following this too
463 	 * and packet storms can occur if this is removed too quickly.
464 	 */
465 	softn->ipf_nat_tcptq[IPF_TCPS_CLOSED].ifq_ttl = softc->ipf_tcplastack;
466 	softn->ipf_nat_tcptq[IPF_TCP_NSTATES - 1].ifq_next =
467 							&softn->ipf_nat_udptq;
468 
469 	IPFTQ_INIT(&softn->ipf_nat_udptq, softn->ipf_nat_defage,
470 		   "nat ipftq udp tab");
471 	softn->ipf_nat_udptq.ifq_next = &softn->ipf_nat_udpacktq;
472 
473 	IPFTQ_INIT(&softn->ipf_nat_udpacktq, softn->ipf_nat_defage,
474 		   "nat ipftq udpack tab");
475 	softn->ipf_nat_udpacktq.ifq_next = &softn->ipf_nat_icmptq;
476 
477 	IPFTQ_INIT(&softn->ipf_nat_icmptq, softn->ipf_nat_deficmpage,
478 		   "nat icmp ipftq tab");
479 	softn->ipf_nat_icmptq.ifq_next = &softn->ipf_nat_icmpacktq;
480 
481 	IPFTQ_INIT(&softn->ipf_nat_icmpacktq, softn->ipf_nat_defage,
482 		   "nat icmpack ipftq tab");
483 	softn->ipf_nat_icmpacktq.ifq_next = &softn->ipf_nat_iptq;
484 
485 	IPFTQ_INIT(&softn->ipf_nat_iptq, softn->ipf_nat_defipage,
486 		   "nat ip ipftq tab");
487 	softn->ipf_nat_iptq.ifq_next = &softn->ipf_nat_pending;
488 
489 	IPFTQ_INIT(&softn->ipf_nat_pending, 1, "nat pending ipftq tab");
490 	softn->ipf_nat_pending.ifq_next = NULL;
491 
492 	for (i = 0, tq = softn->ipf_nat_tcptq; i < IPF_TCP_NSTATES; i++, tq++) {
493 		if (tq->ifq_ttl < softn->ipf_nat_deficmpage)
494 			tq->ifq_ttl = softn->ipf_nat_deficmpage;
495 		else if (tq->ifq_ttl > softn->ipf_nat_defage && softc->ipf_large_nat)
496 			tq->ifq_ttl = softn->ipf_nat_defage;
497 	}
498 
499 	/*
500 	 * Increase this because we may have "keep state" following
501 	 * this too and packet storms can occur if this is removed
502 	 * too quickly.
503 	 */
504 	softn->ipf_nat_tcptq[IPF_TCPS_CLOSED].ifq_ttl = softc->ipf_tcplastack;
505 
506 	MUTEX_INIT(&softn->ipf_nat_new, "ipf nat new mutex");
507 	MUTEX_INIT(&softn->ipf_nat_io, "ipf nat io mutex");
508 
509 	softn->ipf_nat_inited = 1;
510 
511 	return (0);
512 }
513 
514 
515 /* ------------------------------------------------------------------------ */
516 /* Function:    ipf_nat_soft_fini                                           */
517 /* Returns:     Nil                                                         */
518 /* Parameters:  softc(I) - pointer to soft context main structure           */
519 /*                                                                          */
520 /* Free all memory used by NAT structures allocated at runtime.             */
521 /* ------------------------------------------------------------------------ */
522 int
ipf_nat_soft_fini(ipf_main_softc_t * softc,void * arg)523 ipf_nat_soft_fini(ipf_main_softc_t *softc, void *arg)
524 {
525 	ipf_nat_softc_t *softn = arg;
526 	ipftq_t *ifq, *ifqnext;
527 
528 	(void) ipf_nat_clearlist(softc, softn);
529 	(void) ipf_nat_flushtable(softc, softn);
530 
531 	/*
532 	 * Proxy timeout queues are not cleaned here because although they
533 	 * exist on the NAT list, ipf_proxy_unload is called after unload
534 	 * and the proxies actually are responsible for them being created.
535 	 * Should the proxy timeouts have their own list?  There's no real
536 	 * justification as this is the only complication.
537 	 */
538 	for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifqnext) {
539 		ifqnext = ifq->ifq_next;
540 		if (ipf_deletetimeoutqueue(ifq) == 0)
541 			ipf_freetimeoutqueue(softc, ifq);
542 	}
543 
544 	if (softn->ipf_nat_table[0] != NULL) {
545 		KFREES(softn->ipf_nat_table[0],
546 		       sizeof(nat_t *) * softn->ipf_nat_table_sz);
547 		softn->ipf_nat_table[0] = NULL;
548 	}
549 	if (softn->ipf_nat_table[1] != NULL) {
550 		KFREES(softn->ipf_nat_table[1],
551 		       sizeof(nat_t *) * softn->ipf_nat_table_sz);
552 		softn->ipf_nat_table[1] = NULL;
553 	}
554 	if (softn->ipf_nat_map_rules != NULL) {
555 		KFREES(softn->ipf_nat_map_rules,
556 		       sizeof(ipnat_t *) * softn->ipf_nat_maprules_sz);
557 		softn->ipf_nat_map_rules = NULL;
558 	}
559 	if (softn->ipf_nat_rdr_rules != NULL) {
560 		KFREES(softn->ipf_nat_rdr_rules,
561 		       sizeof(ipnat_t *) * softn->ipf_nat_rdrrules_sz);
562 		softn->ipf_nat_rdr_rules = NULL;
563 	}
564 	if (softn->ipf_hm_maptable != NULL) {
565 		KFREES(softn->ipf_hm_maptable,
566 		       sizeof(hostmap_t *) * softn->ipf_nat_hostmap_sz);
567 		softn->ipf_hm_maptable = NULL;
568 	}
569 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen != NULL) {
570 		KFREES(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
571 		       sizeof(u_int) * softn->ipf_nat_table_sz);
572 		softn->ipf_nat_stats.ns_side[0].ns_bucketlen = NULL;
573 	}
574 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen != NULL) {
575 		KFREES(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
576 		       sizeof(u_int) * softn->ipf_nat_table_sz);
577 		softn->ipf_nat_stats.ns_side[1].ns_bucketlen = NULL;
578 	}
579 
580 	if (softn->ipf_nat_inited == 1) {
581 		softn->ipf_nat_inited = 0;
582 		ipf_sttab_destroy(softn->ipf_nat_tcptq);
583 
584 		MUTEX_DESTROY(&softn->ipf_nat_new);
585 		MUTEX_DESTROY(&softn->ipf_nat_io);
586 
587 		MUTEX_DESTROY(&softn->ipf_nat_udptq.ifq_lock);
588 		MUTEX_DESTROY(&softn->ipf_nat_udpacktq.ifq_lock);
589 		MUTEX_DESTROY(&softn->ipf_nat_icmptq.ifq_lock);
590 		MUTEX_DESTROY(&softn->ipf_nat_icmpacktq.ifq_lock);
591 		MUTEX_DESTROY(&softn->ipf_nat_iptq.ifq_lock);
592 		MUTEX_DESTROY(&softn->ipf_nat_pending.ifq_lock);
593 	}
594 
595 	return (0);
596 }
597 
598 
599 /* ------------------------------------------------------------------------ */
600 /* Function:    ipf_nat_setlock                                             */
601 /* Returns:     Nil                                                         */
602 /* Parameters:  arg(I) - pointer to soft state information                  */
603 /*              tmp(I) - new lock value                                     */
604 /*                                                                          */
605 /* Set the "lock status" of NAT to the value in tmp.                        */
606 /* ------------------------------------------------------------------------ */
607 void
ipf_nat_setlock(void * arg,int tmp)608 ipf_nat_setlock(void *arg, int tmp)
609 {
610 	ipf_nat_softc_t *softn = arg;
611 
612 	softn->ipf_nat_lock = tmp;
613 }
614 
615 
616 /* ------------------------------------------------------------------------ */
617 /* Function:    ipf_nat_addrdr                                              */
618 /* Returns:     Nil                                                         */
619 /* Parameters:  n(I) - pointer to NAT rule to add                           */
620 /*                                                                          */
621 /* Adds a redirect rule to the hash table of redirect rules and the list of */
622 /* loaded NAT rules.  Updates the bitmask indicating which netmasks are in  */
623 /* use by redirect rules.                                                   */
624 /* ------------------------------------------------------------------------ */
625 static void
ipf_nat_addrdr(ipf_nat_softc_t * softn,ipnat_t * n)626 ipf_nat_addrdr(ipf_nat_softc_t *softn, ipnat_t *n)
627 {
628 	ipnat_t **np;
629 	u_32_t j;
630 	u_int hv;
631 	u_int rhv;
632 	int k;
633 
634 	if (n->in_odstatype == FRI_NORMAL) {
635 		k = count4bits(n->in_odstmsk);
636 		ipf_inet_mask_add(k, &softn->ipf_nat_rdr_mask);
637 		j = (n->in_odstaddr & n->in_odstmsk);
638 		rhv = NAT_HASH_FN(j, 0, 0xffffffff);
639 	} else {
640 		ipf_inet_mask_add(0, &softn->ipf_nat_rdr_mask);
641 		j = 0;
642 		rhv = 0;
643 	}
644 	hv = rhv % softn->ipf_nat_rdrrules_sz;
645 	np = softn->ipf_nat_rdr_rules + hv;
646 	while (*np != NULL)
647 		np = &(*np)->in_rnext;
648 	n->in_rnext = NULL;
649 	n->in_prnext = np;
650 	n->in_hv[0] = hv;
651 	n->in_use++;
652 	*np = n;
653 }
654 
655 
656 /* ------------------------------------------------------------------------ */
657 /* Function:    ipf_nat_addmap                                              */
658 /* Returns:     Nil                                                         */
659 /* Parameters:  n(I) - pointer to NAT rule to add                           */
660 /*                                                                          */
661 /* Adds a NAT map rule to the hash table of rules and the list of  loaded   */
662 /* NAT rules.  Updates the bitmask indicating which netmasks are in use by  */
663 /* redirect rules.                                                          */
664 /* ------------------------------------------------------------------------ */
665 static void
ipf_nat_addmap(ipf_nat_softc_t * softn,ipnat_t * n)666 ipf_nat_addmap(ipf_nat_softc_t *softn, ipnat_t *n)
667 {
668 	ipnat_t **np;
669 	u_32_t j;
670 	u_int hv;
671 	u_int rhv;
672 	int k;
673 
674 	if (n->in_osrcatype == FRI_NORMAL) {
675 		k = count4bits(n->in_osrcmsk);
676 		ipf_inet_mask_add(k, &softn->ipf_nat_map_mask);
677 		j = (n->in_osrcaddr & n->in_osrcmsk);
678 		rhv = NAT_HASH_FN(j, 0, 0xffffffff);
679 	} else {
680 		ipf_inet_mask_add(0, &softn->ipf_nat_map_mask);
681 		j = 0;
682 		rhv = 0;
683 	}
684 	hv = rhv % softn->ipf_nat_maprules_sz;
685 	np = softn->ipf_nat_map_rules + hv;
686 	while (*np != NULL)
687 		np = &(*np)->in_mnext;
688 	n->in_mnext = NULL;
689 	n->in_pmnext = np;
690 	n->in_hv[1] = rhv;
691 	n->in_use++;
692 	*np = n;
693 }
694 
695 
696 /* ------------------------------------------------------------------------ */
697 /* Function:    ipf_nat_delrdr                                              */
698 /* Returns:     Nil                                                         */
699 /* Parameters:  n(I) - pointer to NAT rule to delete                        */
700 /*                                                                          */
701 /* Removes a redirect rule from the hash table of redirect rules.           */
702 /* ------------------------------------------------------------------------ */
703 void
ipf_nat_delrdr(ipf_nat_softc_t * softn,ipnat_t * n)704 ipf_nat_delrdr(ipf_nat_softc_t *softn, ipnat_t *n)
705 {
706 	if (n->in_odstatype == FRI_NORMAL) {
707 		int k = count4bits(n->in_odstmsk);
708 		ipf_inet_mask_del(k, &softn->ipf_nat_rdr_mask);
709 	} else {
710 		ipf_inet_mask_del(0, &softn->ipf_nat_rdr_mask);
711 	}
712 	if (n->in_rnext)
713 		n->in_rnext->in_prnext = n->in_prnext;
714 	*n->in_prnext = n->in_rnext;
715 	n->in_use--;
716 }
717 
718 
719 /* ------------------------------------------------------------------------ */
720 /* Function:    ipf_nat_delmap                                              */
721 /* Returns:     Nil                                                         */
722 /* Parameters:  n(I) - pointer to NAT rule to delete                        */
723 /*                                                                          */
724 /* Removes a NAT map rule from the hash table of NAT map rules.             */
725 /* ------------------------------------------------------------------------ */
726 void
ipf_nat_delmap(ipf_nat_softc_t * softn,ipnat_t * n)727 ipf_nat_delmap(ipf_nat_softc_t *softn, ipnat_t *n)
728 {
729 	if (n->in_osrcatype == FRI_NORMAL) {
730 		int k = count4bits(n->in_osrcmsk);
731 		ipf_inet_mask_del(k, &softn->ipf_nat_map_mask);
732 	} else {
733 		ipf_inet_mask_del(0, &softn->ipf_nat_map_mask);
734 	}
735 	if (n->in_mnext != NULL)
736 		n->in_mnext->in_pmnext = n->in_pmnext;
737 	*n->in_pmnext = n->in_mnext;
738 	n->in_use--;
739 }
740 
741 
742 /* ------------------------------------------------------------------------ */
743 /* Function:    ipf_nat_hostmap                                             */
744 /* Returns:     struct hostmap* - NULL if no hostmap could be created,      */
745 /*                                else a pointer to the hostmapping to use  */
746 /* Parameters:  np(I)   - pointer to NAT rule                               */
747 /*              real(I) - real IP address                                   */
748 /*              map(I)  - mapped IP address                                 */
749 /*              port(I) - destination port number                           */
750 /* Write Locks: ipf_nat                                                     */
751 /*                                                                          */
752 /* Check if an ip address has already been allocated for a given mapping    */
753 /* that is not doing port based translation.  If is not yet allocated, then */
754 /* create a new entry if a non-NULL NAT rule pointer has been supplied.     */
755 /* ------------------------------------------------------------------------ */
756 static struct hostmap *
ipf_nat_hostmap(ipf_nat_softc_t * softn,ipnat_t * np,struct in_addr src,struct in_addr dst,struct in_addr map,u_32_t port)757 ipf_nat_hostmap(ipf_nat_softc_t *softn, ipnat_t *np, struct in_addr src,
758 	struct in_addr dst, struct in_addr map, u_32_t port)
759 {
760 	hostmap_t *hm;
761 	u_int hv, rhv;
762 
763 	hv = (src.s_addr ^ dst.s_addr);
764 	hv += src.s_addr;
765 	hv += dst.s_addr;
766 	rhv = hv;
767 	hv %= softn->ipf_nat_hostmap_sz;
768 	for (hm = softn->ipf_hm_maptable[hv]; hm; hm = hm->hm_hnext)
769 		if ((hm->hm_osrcip.s_addr == src.s_addr) &&
770 		    (hm->hm_odstip.s_addr == dst.s_addr) &&
771 		    ((np == NULL) || (np == hm->hm_ipnat)) &&
772 		    ((port == 0) || (port == hm->hm_port))) {
773 			softn->ipf_nat_stats.ns_hm_addref++;
774 			hm->hm_ref++;
775 			return (hm);
776 		}
777 
778 	if (np == NULL) {
779 		softn->ipf_nat_stats.ns_hm_nullnp++;
780 		return (NULL);
781 	}
782 
783 	KMALLOC(hm, hostmap_t *);
784 	if (hm) {
785 		hm->hm_next = softn->ipf_hm_maplist;
786 		hm->hm_pnext = &softn->ipf_hm_maplist;
787 		if (softn->ipf_hm_maplist != NULL)
788 			softn->ipf_hm_maplist->hm_pnext = &hm->hm_next;
789 		softn->ipf_hm_maplist = hm;
790 		hm->hm_hnext = softn->ipf_hm_maptable[hv];
791 		hm->hm_phnext = softn->ipf_hm_maptable + hv;
792 		if (softn->ipf_hm_maptable[hv] != NULL)
793 			softn->ipf_hm_maptable[hv]->hm_phnext = &hm->hm_hnext;
794 		softn->ipf_hm_maptable[hv] = hm;
795 		hm->hm_ipnat = np;
796 		np->in_use++;
797 		hm->hm_osrcip = src;
798 		hm->hm_odstip = dst;
799 		hm->hm_nsrcip = map;
800 		hm->hm_ndstip.s_addr = 0;
801 		hm->hm_ref = 1;
802 		hm->hm_port = port;
803 		hm->hm_hv = rhv;
804 		hm->hm_v = 4;
805 		softn->ipf_nat_stats.ns_hm_new++;
806 	} else {
807 		softn->ipf_nat_stats.ns_hm_newfail++;
808 	}
809 	return (hm);
810 }
811 
812 
813 /* ------------------------------------------------------------------------ */
814 /* Function:    ipf_nat_hostmapdel                                          */
815 /* Returns:     Nil                                                         */
816 /* Parameters:  hmp(I) - pointer to hostmap structure pointer               */
817 /* Write Locks: ipf_nat                                                     */
818 /*                                                                          */
819 /* Decrement the references to this hostmap structure by one.  If this      */
820 /* reaches zero then remove it and free it.                                 */
821 /* ------------------------------------------------------------------------ */
822 void
ipf_nat_hostmapdel(ipf_main_softc_t * softc,struct hostmap ** hmp)823 ipf_nat_hostmapdel(ipf_main_softc_t *softc, struct hostmap **hmp)
824 {
825 	struct hostmap *hm;
826 
827 	hm = *hmp;
828 	*hmp = NULL;
829 
830 	hm->hm_ref--;
831 	if (hm->hm_ref == 0) {
832 		ipf_nat_rule_deref(softc, &hm->hm_ipnat);
833 		if (hm->hm_hnext)
834 			hm->hm_hnext->hm_phnext = hm->hm_phnext;
835 		*hm->hm_phnext = hm->hm_hnext;
836 		if (hm->hm_next)
837 			hm->hm_next->hm_pnext = hm->hm_pnext;
838 		*hm->hm_pnext = hm->hm_next;
839 		KFREE(hm);
840 	}
841 }
842 
843 
844 /* ------------------------------------------------------------------------ */
845 /* Function:    ipf_fix_outcksum                                            */
846 /* Returns:     Nil                                                         */
847 /* Parameters:  cksum(I) - ipf_cksum_t, value of fin_cksum                  */
848 /*              sp(I)  - location of 16bit checksum to update               */
849 /*              n(I)  - amount to adjust checksum by                        */
850 /*		partial(I) - partial checksum				    */
851 /*                                                                          */
852 /* Adjusts the 16bit checksum by "n" for packets going out.                 */
853 /* ------------------------------------------------------------------------ */
854 void
ipf_fix_outcksum(int cksum,u_short * sp,u_32_t n,u_32_t partial)855 ipf_fix_outcksum(int cksum, u_short *sp, u_32_t n, u_32_t partial)
856 {
857 	u_short sumshort;
858 	u_32_t sum1;
859 
860 	if (n == 0)
861 		return;
862 
863 	if (cksum == 4) {
864 		*sp = 0;
865 		return;
866 	}
867 	if (cksum == 2) {
868 		sum1 = partial;
869 		sum1 = (sum1 & 0xffff) + (sum1 >> 16);
870 		*sp = htons(sum1);
871 		return;
872 	}
873 	sum1 = (~ntohs(*sp)) & 0xffff;
874 	sum1 += (n);
875 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
876 	/* Again */
877 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
878 	sumshort = ~(u_short)sum1;
879 	*(sp) = htons(sumshort);
880 }
881 
882 
883 /* ------------------------------------------------------------------------ */
884 /* Function:    ipf_fix_incksum                                             */
885 /* Returns:     Nil                                                         */
886 /* Parameters:  cksum(I) - ipf_cksum_t, value of fin_cksum                  */
887 /*              sp(I)  - location of 16bit checksum to update               */
888 /*              n(I)  - amount to adjust checksum by                        */
889 /*		partial(I) - partial checksum				    */
890 /*                                                                          */
891 /* Adjusts the 16bit checksum by "n" for packets going in.                  */
892 /* ------------------------------------------------------------------------ */
893 void
ipf_fix_incksum(int cksum,u_short * sp,u_32_t n,u_32_t partial)894 ipf_fix_incksum(int cksum, u_short *sp, u_32_t n, u_32_t partial)
895 {
896 	u_short sumshort;
897 	u_32_t sum1;
898 
899 	if (n == 0)
900 		return;
901 
902 	if (cksum == 4) {
903 		*sp = 0;
904 		return;
905 	}
906 	if (cksum == 2) {
907 		sum1 = partial;
908 		sum1 = (sum1 & 0xffff) + (sum1 >> 16);
909 		*sp = htons(sum1);
910 		return;
911 	}
912 
913 	sum1 = (~ntohs(*sp)) & 0xffff;
914 	sum1 += ~(n) & 0xffff;
915 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
916 	/* Again */
917 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
918 	sumshort = ~(u_short)sum1;
919 	*(sp) = htons(sumshort);
920 }
921 
922 
923 /* ------------------------------------------------------------------------ */
924 /* Function:    ipf_fix_datacksum                                           */
925 /* Returns:     Nil                                                         */
926 /* Parameters:  sp(I)  - location of 16bit checksum to update               */
927 /*              n(I)  - amount to adjust checksum by                        */
928 /*                                                                          */
929 /* Fix_datacksum is used *only* for the adjustments of checksums in the     */
930 /* data section of an IP packet.                                            */
931 /*                                                                          */
932 /* The only situation in which you need to do this is when NAT'ing an       */
933 /* ICMP error message. Such a message, contains in its body the IP header   */
934 /* of the original IP packet, that causes the error.                        */
935 /*                                                                          */
936 /* You can't use fix_incksum or fix_outcksum in that case, because for the  */
937 /* kernel the data section of the ICMP error is just data, and no special   */
938 /* processing like hardware cksum or ntohs processing have been done by the */
939 /* kernel on the data section.                                              */
940 /* ------------------------------------------------------------------------ */
941 void
ipf_fix_datacksum(u_short * sp,u_32_t n)942 ipf_fix_datacksum(u_short *sp, u_32_t n)
943 {
944 	u_short sumshort;
945 	u_32_t sum1;
946 
947 	if (n == 0)
948 		return;
949 
950 	sum1 = (~ntohs(*sp)) & 0xffff;
951 	sum1 += (n);
952 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
953 	/* Again */
954 	sum1 = (sum1 >> 16) + (sum1 & 0xffff);
955 	sumshort = ~(u_short)sum1;
956 	*(sp) = htons(sumshort);
957 }
958 
959 
960 /* ------------------------------------------------------------------------ */
961 /* Function:    ipf_nat_ioctl                                               */
962 /* Returns:     int - 0 == success, != 0 == failure                         */
963 /* Parameters:  softc(I) - pointer to soft context main structure           */
964 /*              data(I)  - pointer to ioctl data                            */
965 /*              cmd(I)   - ioctl command integer                            */
966 /*              mode(I)  - file mode bits used with open                    */
967 /*              uid(I)   - uid of calling process                           */
968 /*              ctx(I)   - pointer used as key for finding context          */
969 /*                                                                          */
970 /* Processes an ioctl call made to operate on the IP Filter NAT device.     */
971 /* ------------------------------------------------------------------------ */
972 int
ipf_nat_ioctl(ipf_main_softc_t * softc,caddr_t data,ioctlcmd_t cmd,int mode,int uid,void * ctx)973 ipf_nat_ioctl(ipf_main_softc_t *softc, caddr_t data, ioctlcmd_t cmd,
974 	int mode, int uid, void *ctx)
975 {
976 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
977 	int error = 0, ret, arg, getlock;
978 	ipnat_t *nat, *nt, *n;
979 	ipnat_t natd;
980 	SPL_INT(s);
981 
982 #if !SOLARIS && defined(_KERNEL)
983 # if NETBSD_GE_REV(399002000)
984 	if ((mode & FWRITE) &&
985 	     kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_FIREWALL,
986 				     KAUTH_REQ_NETWORK_FIREWALL_FW,
987 				     NULL, NULL, NULL))
988 # else
989 #  if defined(__FreeBSD__)
990 	if (securelevel_ge(curthread->td_ucred, 3) && (mode & FWRITE))
991 #  else
992 	if ((securelevel >= 3) && (mode & FWRITE))
993 #  endif
994 # endif
995 	{
996 		IPFERROR(60001);
997 		return (EPERM);
998 	}
999 # if defined(__FreeBSD__)
1000 	if (jailed_without_vnet(curthread->td_ucred)) {
1001 		IPFERROR(60076);
1002 		return (EOPNOTSUPP);
1003 	}
1004 # endif
1005 #endif
1006 
1007 	getlock = (mode & NAT_LOCKHELD) ? 0 : 1;
1008 
1009 	n = NULL;
1010 	nt = NULL;
1011 	nat = NULL;
1012 
1013 	if ((cmd == (ioctlcmd_t)SIOCADNAT) || (cmd == (ioctlcmd_t)SIOCRMNAT) ||
1014 	    (cmd == (ioctlcmd_t)SIOCPURGENAT)) {
1015 		if (mode & NAT_SYSSPACE) {
1016 			bcopy(data, (char *)&natd, sizeof(natd));
1017 			nat = &natd;
1018 			error = 0;
1019 		} else {
1020 			bzero(&natd, sizeof(natd));
1021 			error = ipf_inobj(softc, data, NULL, &natd,
1022 					  IPFOBJ_IPNAT);
1023 			if (error != 0)
1024 				goto done;
1025 
1026 			if (natd.in_size < sizeof(ipnat_t)) {
1027 				error = EINVAL;
1028 				goto done;
1029 			}
1030 			KMALLOCS(nt, ipnat_t *, natd.in_size);
1031 			if (nt == NULL) {
1032 				IPFERROR(60070);
1033 				error = ENOMEM;
1034 				goto done;
1035 			}
1036 			bzero(nt, natd.in_size);
1037 			error = ipf_inobjsz(softc, data, nt, IPFOBJ_IPNAT,
1038 					    natd.in_size);
1039 			if (error)
1040 				goto done;
1041 			nat = nt;
1042 		}
1043 
1044 		/*
1045 		 * For add/delete, look to see if the NAT entry is
1046 		 * already present
1047 		 */
1048 		nat->in_flags &= IPN_USERFLAGS;
1049 		if ((nat->in_redir & NAT_MAPBLK) == 0) {
1050 			if (nat->in_osrcatype == FRI_NORMAL ||
1051 			    nat->in_osrcatype == FRI_NONE)
1052 				nat->in_osrcaddr &= nat->in_osrcmsk;
1053 			if (nat->in_odstatype == FRI_NORMAL ||
1054 			    nat->in_odstatype == FRI_NONE)
1055 				nat->in_odstaddr &= nat->in_odstmsk;
1056 			if ((nat->in_flags & (IPN_SPLIT|IPN_SIPRANGE)) == 0) {
1057 				if (nat->in_nsrcatype == FRI_NORMAL)
1058 					nat->in_nsrcaddr &= nat->in_nsrcmsk;
1059 				if (nat->in_ndstatype == FRI_NORMAL)
1060 					nat->in_ndstaddr &= nat->in_ndstmsk;
1061 			}
1062 		}
1063 
1064 		error = ipf_nat_rule_init(softc, softn, nat);
1065 		if (error != 0)
1066 			goto done;
1067 
1068 		MUTEX_ENTER(&softn->ipf_nat_io);
1069 		for (n = softn->ipf_nat_list; n != NULL; n = n->in_next)
1070 			if (ipf_nat_cmp_rules(nat, n) == 0)
1071 				break;
1072 	}
1073 
1074 	switch (cmd)
1075 	{
1076 #ifdef  IPFILTER_LOG
1077 	case SIOCIPFFB :
1078 	{
1079 		int tmp;
1080 
1081 		if (!(mode & FWRITE)) {
1082 			IPFERROR(60002);
1083 			error = EPERM;
1084 		} else {
1085 			tmp = ipf_log_clear(softc, IPL_LOGNAT);
1086 			error = BCOPYOUT(&tmp, data, sizeof(tmp));
1087 			if (error != 0) {
1088 				IPFERROR(60057);
1089 				error = EFAULT;
1090 			}
1091 		}
1092 		break;
1093 	}
1094 
1095 	case SIOCSETLG :
1096 		if (!(mode & FWRITE)) {
1097 			IPFERROR(60003);
1098 			error = EPERM;
1099 		} else {
1100 			error = BCOPYIN(data, &softn->ipf_nat_logging,
1101 					sizeof(softn->ipf_nat_logging));
1102 			if (error != 0)
1103 				error = EFAULT;
1104 		}
1105 		break;
1106 
1107 	case SIOCGETLG :
1108 		error = BCOPYOUT(&softn->ipf_nat_logging, data,
1109 				 sizeof(softn->ipf_nat_logging));
1110 		if (error != 0) {
1111 			IPFERROR(60004);
1112 			error = EFAULT;
1113 		}
1114 		break;
1115 
1116 	case FIONREAD :
1117 		arg = ipf_log_bytesused(softc, IPL_LOGNAT);
1118 		error = BCOPYOUT(&arg, data, sizeof(arg));
1119 		if (error != 0) {
1120 			IPFERROR(60005);
1121 			error = EFAULT;
1122 		}
1123 		break;
1124 #endif
1125 	case SIOCADNAT :
1126 		if (!(mode & FWRITE)) {
1127 			IPFERROR(60006);
1128 			error = EPERM;
1129 		} else if (n != NULL) {
1130 			natd.in_flineno = n->in_flineno;
1131 			(void) ipf_outobj(softc, data, &natd, IPFOBJ_IPNAT);
1132 			IPFERROR(60007);
1133 			error = EEXIST;
1134 		} else if (nt == NULL) {
1135 			IPFERROR(60008);
1136 			error = ENOMEM;
1137 		}
1138 		if (error != 0) {
1139 			MUTEX_EXIT(&softn->ipf_nat_io);
1140 			break;
1141 		}
1142 		if (nat != nt)
1143 			bcopy((char *)nat, (char *)nt, sizeof(*n));
1144 		error = ipf_nat_siocaddnat(softc, softn, nt, getlock);
1145 		MUTEX_EXIT(&softn->ipf_nat_io);
1146 		if (error == 0) {
1147 			nat = NULL;
1148 			nt = NULL;
1149 		}
1150 		break;
1151 
1152 	case SIOCRMNAT :
1153 	case SIOCPURGENAT :
1154 		if (!(mode & FWRITE)) {
1155 			IPFERROR(60009);
1156 			error = EPERM;
1157 			n = NULL;
1158 		} else if (n == NULL) {
1159 			IPFERROR(60010);
1160 			error = ESRCH;
1161 		}
1162 
1163 		if (error != 0) {
1164 			MUTEX_EXIT(&softn->ipf_nat_io);
1165 			break;
1166 		}
1167 		if (cmd == (ioctlcmd_t)SIOCPURGENAT) {
1168 			error = ipf_outobjsz(softc, data, n, IPFOBJ_IPNAT,
1169 					     n->in_size);
1170 			if (error) {
1171 				MUTEX_EXIT(&softn->ipf_nat_io);
1172 				goto done;
1173 			}
1174 			n->in_flags |= IPN_PURGE;
1175 		}
1176 		ipf_nat_siocdelnat(softc, softn, n, getlock);
1177 
1178 		MUTEX_EXIT(&softn->ipf_nat_io);
1179 		n = NULL;
1180 		break;
1181 
1182 	case SIOCGNATS :
1183 	    {
1184 		natstat_t *nsp = &softn->ipf_nat_stats;
1185 
1186 		nsp->ns_side[0].ns_table = softn->ipf_nat_table[0];
1187 		nsp->ns_side[1].ns_table = softn->ipf_nat_table[1];
1188 		nsp->ns_list = softn->ipf_nat_list;
1189 		nsp->ns_maptable = softn->ipf_hm_maptable;
1190 		nsp->ns_maplist = softn->ipf_hm_maplist;
1191 		nsp->ns_nattab_sz = softn->ipf_nat_table_sz;
1192 		nsp->ns_nattab_max = softn->ipf_nat_table_max;
1193 		nsp->ns_rultab_sz = softn->ipf_nat_maprules_sz;
1194 		nsp->ns_rdrtab_sz = softn->ipf_nat_rdrrules_sz;
1195 		nsp->ns_hostmap_sz = softn->ipf_nat_hostmap_sz;
1196 		nsp->ns_instances = softn->ipf_nat_instances;
1197 		nsp->ns_ticks = softc->ipf_ticks;
1198 #ifdef IPFILTER_LOGGING
1199 		nsp->ns_log_ok = ipf_log_logok(softc, IPF_LOGNAT);
1200 		nsp->ns_log_fail = ipf_log_failures(softc, IPF_LOGNAT);
1201 #else
1202 		nsp->ns_log_ok = 0;
1203 		nsp->ns_log_fail = 0;
1204 #endif
1205 		error = ipf_outobj(softc, data, nsp, IPFOBJ_NATSTAT);
1206 		break;
1207 	    }
1208 
1209 	case SIOCGNATL :
1210 	    {
1211 		natlookup_t nl;
1212 
1213 		error = ipf_inobj(softc, data, NULL, &nl, IPFOBJ_NATLOOKUP);
1214 		if (error == 0) {
1215 			void *ptr;
1216 
1217 			if (getlock) {
1218 				READ_ENTER(&softc->ipf_nat);
1219 			}
1220 
1221 			switch (nl.nl_v)
1222 			{
1223 			case 4 :
1224 				ptr = ipf_nat_lookupredir(&nl);
1225 				break;
1226 #ifdef USE_INET6
1227 			case 6 :
1228 				ptr = ipf_nat6_lookupredir(&nl);
1229 				break;
1230 #endif
1231 			default:
1232 				ptr = NULL;
1233 				break;
1234 			}
1235 
1236 			if (getlock) {
1237 				RWLOCK_EXIT(&softc->ipf_nat);
1238 			}
1239 			if (ptr != NULL) {
1240 				error = ipf_outobj(softc, data, &nl,
1241 						   IPFOBJ_NATLOOKUP);
1242 			} else {
1243 				IPFERROR(60011);
1244 				error = ESRCH;
1245 			}
1246 		}
1247 		break;
1248 	    }
1249 
1250 	case SIOCIPFFL :	/* old SIOCFLNAT & SIOCCNATL */
1251 		if (!(mode & FWRITE)) {
1252 			IPFERROR(60012);
1253 			error = EPERM;
1254 			break;
1255 		}
1256 		if (getlock) {
1257 			WRITE_ENTER(&softc->ipf_nat);
1258 		}
1259 
1260 		error = BCOPYIN(data, &arg, sizeof(arg));
1261 		if (error != 0) {
1262 			IPFERROR(60013);
1263 			error = EFAULT;
1264 		} else {
1265 			if (arg == 0)
1266 				ret = ipf_nat_flushtable(softc, softn);
1267 			else if (arg == 1)
1268 				ret = ipf_nat_clearlist(softc, softn);
1269 			else
1270 				ret = ipf_nat_extraflush(softc, softn, arg);
1271 			ipf_proxy_flush(softc->ipf_proxy_soft, arg);
1272 		}
1273 
1274 		if (getlock) {
1275 			RWLOCK_EXIT(&softc->ipf_nat);
1276 		}
1277 		if (error == 0) {
1278 			error = BCOPYOUT(&ret, data, sizeof(ret));
1279 		}
1280 		break;
1281 
1282 	case SIOCMATCHFLUSH :
1283 		if (!(mode & FWRITE)) {
1284 			IPFERROR(60014);
1285 			error = EPERM;
1286 			break;
1287 		}
1288 		if (getlock) {
1289 			WRITE_ENTER(&softc->ipf_nat);
1290 		}
1291 
1292 		error = ipf_nat_matchflush(softc, softn, data);
1293 
1294 		if (getlock) {
1295 			RWLOCK_EXIT(&softc->ipf_nat);
1296 		}
1297 		break;
1298 
1299 	case SIOCPROXY :
1300 		error = ipf_proxy_ioctl(softc, data, cmd, mode, ctx);
1301 		break;
1302 
1303 	case SIOCSTLCK :
1304 		if (!(mode & FWRITE)) {
1305 			IPFERROR(60015);
1306 			error = EPERM;
1307 		} else {
1308 			error = ipf_lock(data, &softn->ipf_nat_lock);
1309 		}
1310 		break;
1311 
1312 	case SIOCSTPUT :
1313 		if ((mode & FWRITE) != 0) {
1314 			error = ipf_nat_putent(softc, data, getlock);
1315 		} else {
1316 			IPFERROR(60016);
1317 			error = EACCES;
1318 		}
1319 		break;
1320 
1321 	case SIOCSTGSZ :
1322 		if (softn->ipf_nat_lock) {
1323 			error = ipf_nat_getsz(softc, data, getlock);
1324 		} else {
1325 			IPFERROR(60017);
1326 			error = EACCES;
1327 		}
1328 		break;
1329 
1330 	case SIOCSTGET :
1331 		if (softn->ipf_nat_lock) {
1332 			error = ipf_nat_getent(softc, data, getlock);
1333 		} else {
1334 			IPFERROR(60018);
1335 			error = EACCES;
1336 		}
1337 		break;
1338 
1339 	case SIOCGENITER :
1340 	    {
1341 		ipfgeniter_t iter;
1342 		ipftoken_t *token;
1343 		ipfobj_t obj;
1344 
1345 		error = ipf_inobj(softc, data, &obj, &iter, IPFOBJ_GENITER);
1346 		if (error != 0)
1347 			break;
1348 
1349 		SPL_SCHED(s);
1350 		token = ipf_token_find(softc, iter.igi_type, uid, ctx);
1351 		if (token != NULL) {
1352 			error  = ipf_nat_iterator(softc, token, &iter, &obj);
1353 			WRITE_ENTER(&softc->ipf_tokens);
1354 			ipf_token_deref(softc, token);
1355 			RWLOCK_EXIT(&softc->ipf_tokens);
1356 		}
1357 		SPL_X(s);
1358 		break;
1359 	    }
1360 
1361 	case SIOCIPFDELTOK :
1362 		error = BCOPYIN(data, &arg, sizeof(arg));
1363 		if (error == 0) {
1364 			SPL_SCHED(s);
1365 			error = ipf_token_del(softc, arg, uid, ctx);
1366 			SPL_X(s);
1367 		} else {
1368 			IPFERROR(60019);
1369 			error = EFAULT;
1370 		}
1371 		break;
1372 
1373 	case SIOCGTQTAB :
1374 		error = ipf_outobj(softc, data, softn->ipf_nat_tcptq,
1375 				   IPFOBJ_STATETQTAB);
1376 		break;
1377 
1378 	case SIOCGTABL :
1379 		error = ipf_nat_gettable(softc, softn, data);
1380 		break;
1381 
1382 	default :
1383 		IPFERROR(60020);
1384 		error = EINVAL;
1385 		break;
1386 	}
1387 done:
1388 	if (nat != NULL)
1389 		ipf_nat_rule_fini(softc, nat);
1390 	if (nt != NULL)
1391 		KFREES(nt, nt->in_size);
1392 	return (error);
1393 }
1394 
1395 
1396 /* ------------------------------------------------------------------------ */
1397 /* Function:    ipf_nat_siocaddnat                                          */
1398 /* Returns:     int - 0 == success, != 0 == failure                         */
1399 /* Parameters:  softc(I) - pointer to soft context main structure           */
1400 /*              softn(I) - pointer to NAT context structure                 */
1401 /*              n(I)       - pointer to new NAT rule                        */
1402 /*              np(I)      - pointer to where to insert new NAT rule        */
1403 /*              getlock(I) - flag indicating if lock on  is held            */
1404 /* Mutex Locks: ipf_nat_io                                                  */
1405 /*                                                                          */
1406 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
1407 /* from information passed to the kernel, then add it  to the appropriate   */
1408 /* NAT rule table(s).                                                       */
1409 /* ------------------------------------------------------------------------ */
1410 static int
ipf_nat_siocaddnat(ipf_main_softc_t * softc,ipf_nat_softc_t * softn,ipnat_t * n,int getlock)1411 ipf_nat_siocaddnat(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *n,
1412 	int getlock)
1413 {
1414 	int error = 0;
1415 
1416 	if (ipf_nat_resolverule(softc, n) != 0) {
1417 		IPFERROR(60022);
1418 		return (ENOENT);
1419 	}
1420 
1421 	if ((n->in_age[0] == 0) && (n->in_age[1] != 0)) {
1422 		IPFERROR(60023);
1423 		return (EINVAL);
1424 	}
1425 
1426 	if (n->in_redir == (NAT_DIVERTUDP|NAT_MAP)) {
1427 		/*
1428 		 * Prerecord whether or not the destination of the divert
1429 		 * is local or not to the interface the packet is going
1430 		 * to be sent out.
1431 		 */
1432 		n->in_dlocal = ipf_deliverlocal(softc, n->in_v[1],
1433 						n->in_ifps[1], &n->in_ndstip6);
1434 	}
1435 
1436 	if (getlock) {
1437 		WRITE_ENTER(&softc->ipf_nat);
1438 	}
1439 	n->in_next = NULL;
1440 	n->in_pnext = softn->ipf_nat_list_tail;
1441 	*n->in_pnext = n;
1442 	softn->ipf_nat_list_tail = &n->in_next;
1443 	n->in_use++;
1444 
1445 	if (n->in_redir & NAT_REDIRECT) {
1446 		n->in_flags &= ~IPN_NOTDST;
1447 		switch (n->in_v[0])
1448 		{
1449 		case 4 :
1450 			ipf_nat_addrdr(softn, n);
1451 			break;
1452 #ifdef USE_INET6
1453 		case 6 :
1454 			ipf_nat6_addrdr(softn, n);
1455 			break;
1456 #endif
1457 		default :
1458 			break;
1459 		}
1460 		ATOMIC_INC32(softn->ipf_nat_stats.ns_rules_rdr);
1461 	}
1462 
1463 	if (n->in_redir & (NAT_MAP|NAT_MAPBLK)) {
1464 		n->in_flags &= ~IPN_NOTSRC;
1465 		switch (n->in_v[0])
1466 		{
1467 		case 4 :
1468 			ipf_nat_addmap(softn, n);
1469 			break;
1470 #ifdef USE_INET6
1471 		case 6 :
1472 			ipf_nat6_addmap(softn, n);
1473 			break;
1474 #endif
1475 		default :
1476 			break;
1477 		}
1478 		ATOMIC_INC32(softn->ipf_nat_stats.ns_rules_map);
1479 	}
1480 
1481 	if (n->in_age[0] != 0)
1482 		n->in_tqehead[0] = ipf_addtimeoutqueue(softc,
1483 						       &softn->ipf_nat_utqe,
1484 						       n->in_age[0]);
1485 
1486 	if (n->in_age[1] != 0)
1487 		n->in_tqehead[1] = ipf_addtimeoutqueue(softc,
1488 						       &softn->ipf_nat_utqe,
1489 						       n->in_age[1]);
1490 
1491 	MUTEX_INIT(&n->in_lock, "ipnat rule lock");
1492 
1493 	n = NULL;
1494 	ATOMIC_INC32(softn->ipf_nat_stats.ns_rules);
1495 #if SOLARIS && !defined(INSTANCES)
1496 	pfil_delayed_copy = 0;
1497 #endif
1498 	if (getlock) {
1499 		RWLOCK_EXIT(&softc->ipf_nat);			/* WRITE */
1500 	}
1501 
1502 	return (error);
1503 }
1504 
1505 
1506 /* ------------------------------------------------------------------------ */
1507 /* Function:    ipf_nat_ruleaddrinit                                        */
1508 /* Parameters:  softc(I) - pointer to soft context main structure           */
1509 /*              softn(I) - pointer to NAT context structure                 */
1510 /*              n(I)     - pointer to NAT rule                              */
1511 /*                                                                          */
1512 /* Initialise all of the NAT address structures in a NAT rule.              */
1513 /* ------------------------------------------------------------------------ */
1514 static int
ipf_nat_ruleaddrinit(ipf_main_softc_t * softc,ipf_nat_softc_t * softn,ipnat_t * n)1515 ipf_nat_ruleaddrinit(ipf_main_softc_t *softc, ipf_nat_softc_t *softn,
1516 	ipnat_t *n)
1517 {
1518 	int idx, error;
1519 
1520 	if ((n->in_ndst.na_atype == FRI_LOOKUP) &&
1521 	    (n->in_ndst.na_type != IPLT_DSTLIST)) {
1522 		IPFERROR(60071);
1523 		return (EINVAL);
1524 	}
1525 	if ((n->in_nsrc.na_atype == FRI_LOOKUP) &&
1526 	    (n->in_nsrc.na_type != IPLT_DSTLIST)) {
1527 		IPFERROR(60069);
1528 		return (EINVAL);
1529 	}
1530 
1531 	if (n->in_redir == NAT_BIMAP) {
1532 		n->in_ndstaddr = n->in_osrcaddr;
1533 		n->in_ndstmsk = n->in_osrcmsk;
1534 		n->in_odstaddr = n->in_nsrcaddr;
1535 		n->in_odstmsk = n->in_nsrcmsk;
1536 
1537 	}
1538 
1539 	if (n->in_redir & NAT_REDIRECT)
1540 		idx = 1;
1541 	else
1542 		idx = 0;
1543 	/*
1544 	 * Initialise all of the address fields.
1545 	 */
1546 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_osrc, 1,
1547 				     n->in_ifps[idx]);
1548 	if (error != 0)
1549 		return (error);
1550 
1551 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_odst, 1,
1552 				     n->in_ifps[idx]);
1553 	if (error != 0)
1554 		return (error);
1555 
1556 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_nsrc, 1,
1557 				     n->in_ifps[idx]);
1558 	if (error != 0)
1559 		return (error);
1560 
1561 	error = ipf_nat_nextaddrinit(softc, n->in_names, &n->in_ndst, 1,
1562 				     n->in_ifps[idx]);
1563 	if (error != 0)
1564 		return (error);
1565 
1566 	if (n->in_redir & NAT_DIVERTUDP)
1567 		ipf_nat_builddivertmp(softn, n);
1568 
1569 	return (0);
1570 }
1571 
1572 
1573 /* ------------------------------------------------------------------------ */
1574 /* Function:    ipf_nat_resolvrule                                          */
1575 /* Returns:     Nil                                                         */
1576 /* Parameters:  softc(I) - pointer to soft context main structure           */
1577 /*              n(I)     - pointer to NAT rule                              */
1578 /*                                                                          */
1579 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
1580 /* from information passed to the kernel, then add it  to the appropriate   */
1581 /* NAT rule table(s).                                                       */
1582 /* ------------------------------------------------------------------------ */
1583 static int
ipf_nat_resolverule(ipf_main_softc_t * softc,ipnat_t * n)1584 ipf_nat_resolverule(ipf_main_softc_t *softc, ipnat_t *n)
1585 {
1586 	char *base;
1587 
1588 	base = n->in_names;
1589 
1590 	n->in_ifps[0] = ipf_resolvenic(softc, base + n->in_ifnames[0],
1591 				       n->in_v[0]);
1592 
1593 	if (n->in_ifnames[1] == -1) {
1594 		n->in_ifnames[1] = n->in_ifnames[0];
1595 		n->in_ifps[1] = n->in_ifps[0];
1596 	} else {
1597 		n->in_ifps[1] = ipf_resolvenic(softc, base + n->in_ifnames[1],
1598 					       n->in_v[1]);
1599 	}
1600 
1601 	if (n->in_plabel != -1) {
1602 		if (n->in_redir & NAT_REDIRECT)
1603 			n->in_apr = ipf_proxy_lookup(softc->ipf_proxy_soft,
1604 						     n->in_pr[0],
1605 						     base + n->in_plabel);
1606 		else
1607 			n->in_apr = ipf_proxy_lookup(softc->ipf_proxy_soft,
1608 						     n->in_pr[1],
1609 						     base + n->in_plabel);
1610 		if (n->in_apr == NULL)
1611 			return (-1);
1612 	}
1613 	return (0);
1614 }
1615 
1616 
1617 /* ------------------------------------------------------------------------ */
1618 /* Function:    ipf_nat_siocdelnat                                          */
1619 /* Returns:     int - 0 == success, != 0 == failure                         */
1620 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1621 /*              softn(I)   - pointer to NAT context structure               */
1622 /*              n(I)       - pointer to new NAT rule                        */
1623 /*              getlock(I) - flag indicating if lock on  is held            */
1624 /* Mutex Locks: ipf_nat_io                                                  */
1625 /*                                                                          */
1626 /* Handle SIOCADNAT.  Resolve and calculate details inside the NAT rule     */
1627 /* from information passed to the kernel, then add it  to the appropriate   */
1628 /* NAT rule table(s).                                                       */
1629 /* ------------------------------------------------------------------------ */
1630 static void
ipf_nat_siocdelnat(ipf_main_softc_t * softc,ipf_nat_softc_t * softn,ipnat_t * n,int getlock)1631 ipf_nat_siocdelnat(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *n,
1632 	int getlock)
1633 {
1634 	if (getlock) {
1635 		WRITE_ENTER(&softc->ipf_nat);
1636 	}
1637 
1638 	ipf_nat_delrule(softc, softn, n, 1);
1639 
1640 	if (getlock) {
1641 		RWLOCK_EXIT(&softc->ipf_nat);			/* READ/WRITE */
1642 	}
1643 }
1644 
1645 
1646 /* ------------------------------------------------------------------------ */
1647 /* Function:    ipf_nat_getsz                                               */
1648 /* Returns:     int - 0 == success, != 0 is the error value.                */
1649 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1650 /*              data(I)    - pointer to natget structure with kernel        */
1651 /*                           pointer get the size of.                       */
1652 /*              getlock(I) - flag indicating whether or not the caller      */
1653 /*                           holds a lock on ipf_nat                        */
1654 /*                                                                          */
1655 /* Handle SIOCSTGSZ.                                                        */
1656 /* Return the size of the nat list entry to be copied back to user space.   */
1657 /* The size of the entry is stored in the ng_sz field and the enture natget */
1658 /* structure is copied back to the user.                                    */
1659 /* ------------------------------------------------------------------------ */
1660 static int
ipf_nat_getsz(ipf_main_softc_t * softc,caddr_t data,int getlock)1661 ipf_nat_getsz(ipf_main_softc_t *softc, caddr_t data, int getlock)
1662 {
1663 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1664 	ap_session_t *aps;
1665 	nat_t *nat, *n;
1666 	natget_t ng;
1667 	int error;
1668 
1669 	error = BCOPYIN(data, &ng, sizeof(ng));
1670 	if (error != 0) {
1671 		IPFERROR(60024);
1672 		return (EFAULT);
1673 	}
1674 
1675 	if (getlock) {
1676 		READ_ENTER(&softc->ipf_nat);
1677 	}
1678 
1679 	nat = ng.ng_ptr;
1680 	if (!nat) {
1681 		nat = softn->ipf_nat_instances;
1682 		ng.ng_sz = 0;
1683 		/*
1684 		 * Empty list so the size returned is 0.  Simple.
1685 		 */
1686 		if (nat == NULL) {
1687 			if (getlock) {
1688 				RWLOCK_EXIT(&softc->ipf_nat);
1689 			}
1690 			error = BCOPYOUT(&ng, data, sizeof(ng));
1691 			if (error != 0) {
1692 				IPFERROR(60025);
1693 				return (EFAULT);
1694 			}
1695 			return (0);
1696 		}
1697 	} else {
1698 		/*
1699 		 * Make sure the pointer we're copying from exists in the
1700 		 * current list of entries.  Security precaution to prevent
1701 		 * copying of random kernel data.
1702 		 */
1703 		for (n = softn->ipf_nat_instances; n; n = n->nat_next)
1704 			if (n == nat)
1705 				break;
1706 		if (n == NULL) {
1707 			if (getlock) {
1708 				RWLOCK_EXIT(&softc->ipf_nat);
1709 			}
1710 			IPFERROR(60026);
1711 			return (ESRCH);
1712 		}
1713 	}
1714 
1715 	/*
1716 	 * Include any space required for proxy data structures.
1717 	 */
1718 	ng.ng_sz = sizeof(nat_save_t);
1719 	aps = nat->nat_aps;
1720 	if (aps != NULL) {
1721 		ng.ng_sz += sizeof(ap_session_t) - 4;
1722 		if (aps->aps_data != 0)
1723 			ng.ng_sz += aps->aps_psiz;
1724 	}
1725 	if (getlock) {
1726 		RWLOCK_EXIT(&softc->ipf_nat);
1727 	}
1728 
1729 	error = BCOPYOUT(&ng, data, sizeof(ng));
1730 	if (error != 0) {
1731 		IPFERROR(60027);
1732 		return (EFAULT);
1733 	}
1734 	return (0);
1735 }
1736 
1737 
1738 /* ------------------------------------------------------------------------ */
1739 /* Function:    ipf_nat_getent                                              */
1740 /* Returns:     int - 0 == success, != 0 is the error value.                */
1741 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1742 /*              data(I)    - pointer to natget structure with kernel pointer*/
1743 /*                           to NAT structure to copy out.                  */
1744 /*              getlock(I) - flag indicating whether or not the caller      */
1745 /*                           holds a lock on ipf_nat                        */
1746 /*                                                                          */
1747 /* Handle SIOCSTGET.                                                        */
1748 /* Copies out NAT entry to user space.  Any additional data held for a      */
1749 /* proxy is also copied, as to is the NAT rule which was responsible for it */
1750 /* ------------------------------------------------------------------------ */
1751 static int
ipf_nat_getent(ipf_main_softc_t * softc,caddr_t data,int getlock)1752 ipf_nat_getent(ipf_main_softc_t *softc, caddr_t data, int getlock)
1753 {
1754 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1755 	int error, outsize;
1756 	ap_session_t *aps;
1757 	nat_save_t *ipn, ipns;
1758 	nat_t *n, *nat;
1759 
1760 	error = ipf_inobj(softc, data, NULL, &ipns, IPFOBJ_NATSAVE);
1761 	if (error != 0)
1762 		return (error);
1763 
1764 	if ((ipns.ipn_dsize < sizeof(ipns)) || (ipns.ipn_dsize > 81920)) {
1765 		IPFERROR(60028);
1766 		return (EINVAL);
1767 	}
1768 
1769 	KMALLOCS(ipn, nat_save_t *, ipns.ipn_dsize);
1770 	if (ipn == NULL) {
1771 		IPFERROR(60029);
1772 		return (ENOMEM);
1773 	}
1774 	bzero(ipn, ipns.ipn_dsize);
1775 
1776 	if (getlock) {
1777 		READ_ENTER(&softc->ipf_nat);
1778 	}
1779 
1780 	ipn->ipn_dsize = ipns.ipn_dsize;
1781 	nat = ipns.ipn_next;
1782 	if (nat == NULL) {
1783 		nat = softn->ipf_nat_instances;
1784 		if (nat == NULL) {
1785 			if (softn->ipf_nat_instances == NULL) {
1786 				IPFERROR(60030);
1787 				error = ENOENT;
1788 			}
1789 			goto finished;
1790 		}
1791 	} else {
1792 		/*
1793 		 * Make sure the pointer we're copying from exists in the
1794 		 * current list of entries.  Security precaution to prevent
1795 		 * copying of random kernel data.
1796 		 */
1797 		for (n = softn->ipf_nat_instances; n; n = n->nat_next)
1798 			if (n == nat)
1799 				break;
1800 		if (n == NULL) {
1801 			IPFERROR(60031);
1802 			error = ESRCH;
1803 			goto finished;
1804 		}
1805 	}
1806 	ipn->ipn_next = nat->nat_next;
1807 
1808 	/*
1809 	 * Copy the NAT structure.
1810 	 */
1811 	bcopy((char *)nat, &ipn->ipn_nat, sizeof(*nat));
1812 
1813 	/*
1814 	 * If we have a pointer to the NAT rule it belongs to, save that too.
1815 	 */
1816 	if (nat->nat_ptr != NULL)
1817 		bcopy((char *)nat->nat_ptr, (char *)&ipn->ipn_ipnat,
1818 		      sizeof(nat->nat_ptr));
1819 
1820 	/*
1821 	 * If we also know the NAT entry has an associated filter rule,
1822 	 * save that too.
1823 	 */
1824 	if (nat->nat_fr != NULL)
1825 		bcopy((char *)nat->nat_fr, (char *)&ipn->ipn_fr,
1826 		      sizeof(ipn->ipn_fr));
1827 
1828 	/*
1829 	 * Last but not least, if there is an application proxy session set
1830 	 * up for this NAT entry, then copy that out too, including any
1831 	 * private data saved along side it by the proxy.
1832 	 */
1833 	aps = nat->nat_aps;
1834 	outsize = ipn->ipn_dsize - sizeof(*ipn) + sizeof(ipn->ipn_data);
1835 	if (aps != NULL) {
1836 		char *s;
1837 
1838 		if (outsize < sizeof(*aps)) {
1839 			IPFERROR(60032);
1840 			error = ENOBUFS;
1841 			goto finished;
1842 		}
1843 
1844 		s = ipn->ipn_data;
1845 		bcopy((char *)aps, s, sizeof(*aps));
1846 		s += sizeof(*aps);
1847 		outsize -= sizeof(*aps);
1848 		if ((aps->aps_data != NULL) && (outsize >= aps->aps_psiz))
1849 			bcopy(aps->aps_data, s, aps->aps_psiz);
1850 		else {
1851 			IPFERROR(60033);
1852 			error = ENOBUFS;
1853 		}
1854 	}
1855 	if (error == 0) {
1856 		error = ipf_outobjsz(softc, data, ipn, IPFOBJ_NATSAVE,
1857 				     ipns.ipn_dsize);
1858 	}
1859 
1860 finished:
1861 	if (ipn != NULL) {
1862 		KFREES(ipn, ipns.ipn_dsize);
1863 	}
1864 	if (getlock) {
1865 		RWLOCK_EXIT(&softc->ipf_nat);
1866 	}
1867 	return (error);
1868 }
1869 
1870 
1871 /* ------------------------------------------------------------------------ */
1872 /* Function:    ipf_nat_putent                                              */
1873 /* Returns:     int - 0 == success, != 0 is the error value.                */
1874 /* Parameters:  softc(I)   - pointer to soft context main structure         */
1875 /*              data(I)    - pointer to natget structure with NAT           */
1876 /*                           structure information to load into the kernel  */
1877 /*              getlock(I) - flag indicating whether or not a write lock    */
1878 /*                           on is already held.                            */
1879 /*                                                                          */
1880 /* Handle SIOCSTPUT.                                                        */
1881 /* Loads a NAT table entry from user space, including a NAT rule, proxy and */
1882 /* firewall rule data structures, if pointers to them indicate so.          */
1883 /* ------------------------------------------------------------------------ */
1884 static int
ipf_nat_putent(ipf_main_softc_t * softc,caddr_t data,int getlock)1885 ipf_nat_putent(ipf_main_softc_t *softc, caddr_t data, int getlock)
1886 {
1887 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
1888 	nat_save_t ipn, *ipnn;
1889 	ap_session_t *aps;
1890 	nat_t *n, *nat;
1891 	frentry_t *fr;
1892 	fr_info_t fin;
1893 	ipnat_t *in;
1894 	int error;
1895 
1896 	error = ipf_inobj(softc, data, NULL, &ipn, IPFOBJ_NATSAVE);
1897 	if (error != 0)
1898 		return (error);
1899 
1900 	/*
1901 	 * Initialise early because of code at junkput label.
1902 	 */
1903 	n = NULL;
1904 	in = NULL;
1905 	aps = NULL;
1906 	nat = NULL;
1907 	ipnn = NULL;
1908 	fr = NULL;
1909 
1910 	/*
1911 	 * New entry, copy in the rest of the NAT entry if it's size is more
1912 	 * than just the nat_t structure.
1913 	 */
1914 	if (ipn.ipn_dsize > sizeof(ipn)) {
1915 		if (ipn.ipn_dsize > 81920) {
1916 			IPFERROR(60034);
1917 			error = ENOMEM;
1918 			goto junkput;
1919 		}
1920 
1921 		KMALLOCS(ipnn, nat_save_t *, ipn.ipn_dsize);
1922 		if (ipnn == NULL) {
1923 			IPFERROR(60035);
1924 			return (ENOMEM);
1925 		}
1926 
1927 		bzero(ipnn, ipn.ipn_dsize);
1928 		error = ipf_inobjsz(softc, data, ipnn, IPFOBJ_NATSAVE,
1929 				    ipn.ipn_dsize);
1930 		if (error != 0) {
1931 			goto junkput;
1932 		}
1933 	} else
1934 		ipnn = &ipn;
1935 
1936 	KMALLOC(nat, nat_t *);
1937 	if (nat == NULL) {
1938 		IPFERROR(60037);
1939 		error = ENOMEM;
1940 		goto junkput;
1941 	}
1942 
1943 	bcopy((char *)&ipnn->ipn_nat, (char *)nat, sizeof(*nat));
1944 
1945 	switch (nat->nat_v[0])
1946 	{
1947 	case 4:
1948 #ifdef USE_INET6
1949 	case 6 :
1950 #endif
1951 		break;
1952 	default :
1953 		IPFERROR(60061);
1954 		error = EPROTONOSUPPORT;
1955 		goto junkput;
1956 		/*NOTREACHED*/
1957 	}
1958 
1959 	/*
1960 	 * Initialize all these so that ipf_nat_delete() doesn't cause a crash.
1961 	 */
1962 	bzero((char *)nat, offsetof(struct nat, nat_tqe));
1963 	nat->nat_tqe.tqe_pnext = NULL;
1964 	nat->nat_tqe.tqe_next = NULL;
1965 	nat->nat_tqe.tqe_ifq = NULL;
1966 	nat->nat_tqe.tqe_parent = nat;
1967 
1968 	/*
1969 	 * Restore the rule associated with this nat session
1970 	 */
1971 	in = ipnn->ipn_nat.nat_ptr;
1972 	if (in != NULL) {
1973 		KMALLOCS(in, ipnat_t *, ipnn->ipn_ipnat.in_size);
1974 		nat->nat_ptr = in;
1975 		if (in == NULL) {
1976 			IPFERROR(60038);
1977 			error = ENOMEM;
1978 			goto junkput;
1979 		}
1980 		bcopy((char *)&ipnn->ipn_ipnat, (char *)in,
1981 		      ipnn->ipn_ipnat.in_size);
1982 		in->in_use = 1;
1983 		in->in_flags |= IPN_DELETE;
1984 
1985 		ATOMIC_INC32(softn->ipf_nat_stats.ns_rules);
1986 
1987 		if (ipf_nat_resolverule(softc, in) != 0) {
1988 			IPFERROR(60039);
1989 			error = ESRCH;
1990 			goto junkput;
1991 		}
1992 	}
1993 
1994 	/*
1995 	 * Check that the NAT entry doesn't already exist in the kernel.
1996 	 *
1997 	 * For NAT_OUTBOUND, we're lookup for a duplicate MAP entry.  To do
1998 	 * this, we check to see if the inbound combination of addresses and
1999 	 * ports is already known.  Similar logic is applied for NAT_INBOUND.
2000 	 *
2001 	 */
2002 	bzero((char *)&fin, sizeof(fin));
2003 	fin.fin_v = nat->nat_v[0];
2004 	fin.fin_p = nat->nat_pr[0];
2005 	fin.fin_rev = nat->nat_rev;
2006 	fin.fin_ifp = nat->nat_ifps[0];
2007 	fin.fin_data[0] = ntohs(nat->nat_ndport);
2008 	fin.fin_data[1] = ntohs(nat->nat_nsport);
2009 
2010 	switch (nat->nat_dir)
2011 	{
2012 	case NAT_OUTBOUND :
2013 	case NAT_DIVERTOUT :
2014 		if (getlock) {
2015 			READ_ENTER(&softc->ipf_nat);
2016 		}
2017 
2018 		fin.fin_v = nat->nat_v[1];
2019 		if (nat->nat_v[1] == 4) {
2020 			n = ipf_nat_inlookup(&fin, nat->nat_flags, fin.fin_p,
2021 					     nat->nat_ndstip, nat->nat_nsrcip);
2022 #ifdef USE_INET6
2023 		} else if (nat->nat_v[1] == 6) {
2024 			n = ipf_nat6_inlookup(&fin, nat->nat_flags, fin.fin_p,
2025 					      &nat->nat_ndst6.in6,
2026 					      &nat->nat_nsrc6.in6);
2027 #endif
2028 		}
2029 
2030 		if (getlock) {
2031 			RWLOCK_EXIT(&softc->ipf_nat);
2032 		}
2033 		if (n != NULL) {
2034 			IPFERROR(60040);
2035 			error = EEXIST;
2036 			goto junkput;
2037 		}
2038 		break;
2039 
2040 	case NAT_INBOUND :
2041 	case NAT_DIVERTIN :
2042 		if (getlock) {
2043 			READ_ENTER(&softc->ipf_nat);
2044 		}
2045 
2046 		if (fin.fin_v == 4) {
2047 			n = ipf_nat_outlookup(&fin, nat->nat_flags, fin.fin_p,
2048 					      nat->nat_ndstip,
2049 					      nat->nat_nsrcip);
2050 #ifdef USE_INET6
2051 		} else if (fin.fin_v == 6) {
2052 			n = ipf_nat6_outlookup(&fin, nat->nat_flags, fin.fin_p,
2053 					       &nat->nat_ndst6.in6,
2054 					       &nat->nat_nsrc6.in6);
2055 #endif
2056 		}
2057 
2058 		if (getlock) {
2059 			RWLOCK_EXIT(&softc->ipf_nat);
2060 		}
2061 		if (n != NULL) {
2062 			IPFERROR(60041);
2063 			error = EEXIST;
2064 			goto junkput;
2065 		}
2066 		break;
2067 
2068 	default :
2069 		IPFERROR(60042);
2070 		error = EINVAL;
2071 		goto junkput;
2072 	}
2073 
2074 	/*
2075 	 * Restore ap_session_t structure.  Include the private data allocated
2076 	 * if it was there.
2077 	 */
2078 	aps = nat->nat_aps;
2079 	if (aps != NULL) {
2080 		KMALLOC(aps, ap_session_t *);
2081 		nat->nat_aps = aps;
2082 		if (aps == NULL) {
2083 			IPFERROR(60043);
2084 			error = ENOMEM;
2085 			goto junkput;
2086 		}
2087 		bcopy(ipnn->ipn_data, (char *)aps, sizeof(*aps));
2088 		if (in != NULL)
2089 			aps->aps_apr = in->in_apr;
2090 		else
2091 			aps->aps_apr = NULL;
2092 		if (aps->aps_psiz != 0) {
2093 			if (aps->aps_psiz > 81920) {
2094 				IPFERROR(60044);
2095 				error = ENOMEM;
2096 				goto junkput;
2097 			}
2098 			KMALLOCS(aps->aps_data, void *, aps->aps_psiz);
2099 			if (aps->aps_data == NULL) {
2100 				IPFERROR(60045);
2101 				error = ENOMEM;
2102 				goto junkput;
2103 			}
2104 			bcopy(ipnn->ipn_data + sizeof(*aps), aps->aps_data,
2105 			      aps->aps_psiz);
2106 		} else {
2107 			aps->aps_psiz = 0;
2108 			aps->aps_data = NULL;
2109 		}
2110 	}
2111 
2112 	/*
2113 	 * If there was a filtering rule associated with this entry then
2114 	 * build up a new one.
2115 	 */
2116 	fr = nat->nat_fr;
2117 	if (fr != NULL) {
2118 		if ((nat->nat_flags & SI_NEWFR) != 0) {
2119 			KMALLOC(fr, frentry_t *);
2120 			nat->nat_fr = fr;
2121 			if (fr == NULL) {
2122 				IPFERROR(60046);
2123 				error = ENOMEM;
2124 				goto junkput;
2125 			}
2126 			ipnn->ipn_nat.nat_fr = fr;
2127 			fr->fr_ref = 1;
2128 			(void) ipf_outobj(softc, data, ipnn, IPFOBJ_NATSAVE);
2129 			bcopy((char *)&ipnn->ipn_fr, (char *)fr, sizeof(*fr));
2130 
2131 			fr->fr_ref = 1;
2132 			fr->fr_dsize = 0;
2133 			fr->fr_data = NULL;
2134 			fr->fr_type = FR_T_NONE;
2135 
2136 			MUTEX_NUKE(&fr->fr_lock);
2137 			MUTEX_INIT(&fr->fr_lock, "nat-filter rule lock");
2138 		} else {
2139 			if (getlock) {
2140 				READ_ENTER(&softc->ipf_nat);
2141 			}
2142 			for (n = softn->ipf_nat_instances; n; n = n->nat_next)
2143 				if (n->nat_fr == fr)
2144 					break;
2145 
2146 			if (n != NULL) {
2147 				MUTEX_ENTER(&fr->fr_lock);
2148 				fr->fr_ref++;
2149 				MUTEX_EXIT(&fr->fr_lock);
2150 			}
2151 			if (getlock) {
2152 				RWLOCK_EXIT(&softc->ipf_nat);
2153 			}
2154 
2155 			if (n == NULL) {
2156 				IPFERROR(60047);
2157 				error = ESRCH;
2158 				goto junkput;
2159 			}
2160 		}
2161 	}
2162 
2163 	if (ipnn != &ipn) {
2164 		KFREES(ipnn, ipn.ipn_dsize);
2165 		ipnn = NULL;
2166 	}
2167 
2168 	if (getlock) {
2169 		WRITE_ENTER(&softc->ipf_nat);
2170 	}
2171 
2172 	if (fin.fin_v == 4)
2173 		error = ipf_nat_finalise(&fin, nat);
2174 #ifdef USE_INET6
2175 	else
2176 		error = ipf_nat6_finalise(&fin, nat);
2177 #endif
2178 
2179 	if (getlock) {
2180 		RWLOCK_EXIT(&softc->ipf_nat);
2181 	}
2182 
2183 	if (error == 0)
2184 		return (0);
2185 
2186 	IPFERROR(60048);
2187 	error = ENOMEM;
2188 
2189 junkput:
2190 	if (fr != NULL) {
2191 		(void) ipf_derefrule(softc, &fr);
2192 	}
2193 
2194 	if ((ipnn != NULL) && (ipnn != &ipn)) {
2195 		KFREES(ipnn, ipn.ipn_dsize);
2196 	}
2197 	if (nat != NULL) {
2198 		if (aps != NULL) {
2199 			if (aps->aps_data != NULL) {
2200 				KFREES(aps->aps_data, aps->aps_psiz);
2201 			}
2202 			KFREE(aps);
2203 		}
2204 		if (in != NULL) {
2205 			if (in->in_apr)
2206 				ipf_proxy_deref(in->in_apr);
2207 			KFREES(in, in->in_size);
2208 		}
2209 		KFREE(nat);
2210 	}
2211 	return (error);
2212 }
2213 
2214 
2215 /* ------------------------------------------------------------------------ */
2216 /* Function:    ipf_nat_delete                                              */
2217 /* Returns:     Nil                                                         */
2218 /* Parameters:  softc(I)   - pointer to soft context main structure         */
2219 /*              nat(I)     - pointer to NAT structure to delete             */
2220 /*              logtype(I) - type of LOG record to create before deleting   */
2221 /* Write Lock:  ipf_nat                                                     */
2222 /*                                                                          */
2223 /* Delete a nat entry from the various lists and table.  If NAT logging is  */
2224 /* enabled then generate a NAT log record for this event.                   */
2225 /* ------------------------------------------------------------------------ */
2226 void
ipf_nat_delete(ipf_main_softc_t * softc,struct nat * nat,int logtype)2227 ipf_nat_delete(ipf_main_softc_t *softc, struct nat *nat, int logtype)
2228 {
2229 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2230 	int madeorphan = 0, bkt, removed = 0;
2231 	nat_stat_side_t *nss;
2232 	struct ipnat *ipn;
2233 
2234 	if (logtype != 0 && softn->ipf_nat_logging != 0)
2235 		ipf_nat_log(softc, softn, nat, logtype);
2236 
2237 	/*
2238 	 * Take it as a general indication that all the pointers are set if
2239 	 * nat_pnext is set.
2240 	 */
2241 	if (nat->nat_pnext != NULL) {
2242 		removed = 1;
2243 
2244 		bkt = nat->nat_hv[0] % softn->ipf_nat_table_sz;
2245 		nss = &softn->ipf_nat_stats.ns_side[0];
2246 		if (nss->ns_bucketlen[bkt] > 0)
2247 			nss->ns_bucketlen[bkt]--;
2248 		if (nss->ns_bucketlen[bkt] == 0) {
2249 			nss->ns_inuse--;
2250 		}
2251 
2252 		bkt = nat->nat_hv[1] % softn->ipf_nat_table_sz;
2253 		nss = &softn->ipf_nat_stats.ns_side[1];
2254 		if (nss->ns_bucketlen[bkt] > 0)
2255 			nss->ns_bucketlen[bkt]--;
2256 		if (nss->ns_bucketlen[bkt] == 0) {
2257 			nss->ns_inuse--;
2258 		}
2259 
2260 		*nat->nat_pnext = nat->nat_next;
2261 		if (nat->nat_next != NULL) {
2262 			nat->nat_next->nat_pnext = nat->nat_pnext;
2263 			nat->nat_next = NULL;
2264 		}
2265 		nat->nat_pnext = NULL;
2266 
2267 		*nat->nat_phnext[0] = nat->nat_hnext[0];
2268 		if (nat->nat_hnext[0] != NULL) {
2269 			nat->nat_hnext[0]->nat_phnext[0] = nat->nat_phnext[0];
2270 			nat->nat_hnext[0] = NULL;
2271 		}
2272 		nat->nat_phnext[0] = NULL;
2273 
2274 		*nat->nat_phnext[1] = nat->nat_hnext[1];
2275 		if (nat->nat_hnext[1] != NULL) {
2276 			nat->nat_hnext[1]->nat_phnext[1] = nat->nat_phnext[1];
2277 			nat->nat_hnext[1] = NULL;
2278 		}
2279 		nat->nat_phnext[1] = NULL;
2280 
2281 		if ((nat->nat_flags & SI_WILDP) != 0) {
2282 			ATOMIC_DEC32(softn->ipf_nat_stats.ns_wilds);
2283 		}
2284 		madeorphan = 1;
2285 	}
2286 
2287 	if (nat->nat_me != NULL) {
2288 		*nat->nat_me = NULL;
2289 		nat->nat_me = NULL;
2290 		nat->nat_ref--;
2291 		ASSERT(nat->nat_ref >= 0);
2292 	}
2293 
2294 	if (nat->nat_tqe.tqe_ifq != NULL) {
2295 		/*
2296 		 * No call to ipf_freetimeoutqueue() is made here, they are
2297 		 * garbage collected in ipf_nat_expire().
2298 		 */
2299 		(void) ipf_deletequeueentry(&nat->nat_tqe);
2300 	}
2301 
2302 	if (nat->nat_sync) {
2303 		ipf_sync_del_nat(softc->ipf_sync_soft, nat->nat_sync);
2304 		nat->nat_sync = NULL;
2305 	}
2306 
2307 	if (logtype == NL_EXPIRE)
2308 		softn->ipf_nat_stats.ns_expire++;
2309 
2310 	MUTEX_ENTER(&nat->nat_lock);
2311 	/*
2312 	 * NL_DESTROY should only be passed in when we've got nat_ref >= 2.
2313 	 * This happens when a nat'd packet is blocked and we want to throw
2314 	 * away the NAT session.
2315 	 */
2316 	if (logtype == NL_DESTROY) {
2317 		if (nat->nat_ref > 2) {
2318 			nat->nat_ref -= 2;
2319 			MUTEX_EXIT(&nat->nat_lock);
2320 			if (removed)
2321 				softn->ipf_nat_stats.ns_orphans++;
2322 			return;
2323 		}
2324 	} else if (nat->nat_ref > 1) {
2325 		nat->nat_ref--;
2326 		MUTEX_EXIT(&nat->nat_lock);
2327 		if (madeorphan == 1)
2328 			softn->ipf_nat_stats.ns_orphans++;
2329 		return;
2330 	}
2331 	ASSERT(nat->nat_ref >= 0);
2332 	MUTEX_EXIT(&nat->nat_lock);
2333 
2334 	nat->nat_ref = 0;
2335 
2336 	if (madeorphan == 0)
2337 		softn->ipf_nat_stats.ns_orphans--;
2338 
2339 	/*
2340 	 * At this point, nat_ref can be either 0 or -1
2341 	 */
2342 	softn->ipf_nat_stats.ns_proto[nat->nat_pr[0]]--;
2343 
2344 	if (nat->nat_fr != NULL) {
2345 		(void) ipf_derefrule(softc, &nat->nat_fr);
2346 	}
2347 
2348 	if (nat->nat_hm != NULL) {
2349 		ipf_nat_hostmapdel(softc, &nat->nat_hm);
2350 	}
2351 
2352 	/*
2353 	 * If there is an active reference from the nat entry to its parent
2354 	 * rule, decrement the rule's reference count and free it too if no
2355 	 * longer being used.
2356 	 */
2357 	ipn = nat->nat_ptr;
2358 	nat->nat_ptr = NULL;
2359 
2360 	if (ipn != NULL) {
2361 		ipn->in_space++;
2362 		ipf_nat_rule_deref(softc, &ipn);
2363 	}
2364 
2365 	if (nat->nat_aps != NULL) {
2366 		ipf_proxy_free(softc, nat->nat_aps);
2367 		nat->nat_aps = NULL;
2368 	}
2369 
2370 	MUTEX_DESTROY(&nat->nat_lock);
2371 
2372 	softn->ipf_nat_stats.ns_active--;
2373 
2374 	/*
2375 	 * If there's a fragment table entry too for this nat entry, then
2376 	 * dereference that as well.  This is after nat_lock is released
2377 	 * because of Tru64.
2378 	 */
2379 	ipf_frag_natforget(softc, (void *)nat);
2380 
2381 	KFREE(nat);
2382 }
2383 
2384 
2385 /* ------------------------------------------------------------------------ */
2386 /* Function:    ipf_nat_flushtable                                          */
2387 /* Returns:     int - number of NAT rules deleted                           */
2388 /* Parameters:  softc(I) - pointer to soft context main structure           */
2389 /*              softn(I) - pointer to NAT context structure                 */
2390 /* Write Lock:  ipf_nat                                                     */
2391 /*                                                                          */
2392 /* Deletes all currently active NAT sessions.  In deleting each NAT entry a */
2393 /* log record should be emitted in ipf_nat_delete() if NAT logging is       */
2394 /* enabled.                                                                 */
2395 /* ------------------------------------------------------------------------ */
2396 /*
2397  * nat_flushtable - clear the NAT table of all mapping entries.
2398  */
2399 static int
ipf_nat_flushtable(ipf_main_softc_t * softc,ipf_nat_softc_t * softn)2400 ipf_nat_flushtable(ipf_main_softc_t *softc, ipf_nat_softc_t *softn)
2401 {
2402 	nat_t *nat;
2403 	int j = 0;
2404 
2405 	/*
2406 	 * ALL NAT mappings deleted, so lets just make the deletions
2407 	 * quicker.
2408 	 */
2409 	if (softn->ipf_nat_table[0] != NULL)
2410 		bzero((char *)softn->ipf_nat_table[0],
2411 		      sizeof(softn->ipf_nat_table[0]) *
2412 		      softn->ipf_nat_table_sz);
2413 	if (softn->ipf_nat_table[1] != NULL)
2414 		bzero((char *)softn->ipf_nat_table[1],
2415 		      sizeof(softn->ipf_nat_table[1]) *
2416 		      softn->ipf_nat_table_sz);
2417 
2418 	while ((nat = softn->ipf_nat_instances) != NULL) {
2419 		ipf_nat_delete(softc, nat, NL_FLUSH);
2420 		j++;
2421 	}
2422 
2423 	return (j);
2424 }
2425 
2426 
2427 /* ------------------------------------------------------------------------ */
2428 /* Function:    ipf_nat_clearlist                                           */
2429 /* Returns:     int - number of NAT/RDR rules deleted                       */
2430 /* Parameters:  softc(I) - pointer to soft context main structure           */
2431 /*              softn(I) - pointer to NAT context structure                 */
2432 /*                                                                          */
2433 /* Delete all rules in the current list of rules.  There is nothing elegant */
2434 /* about this cleanup: simply free all entries on the list of rules and     */
2435 /* clear out the tables used for hashed NAT rule lookups.                   */
2436 /* ------------------------------------------------------------------------ */
2437 static int
ipf_nat_clearlist(ipf_main_softc_t * softc,ipf_nat_softc_t * softn)2438 ipf_nat_clearlist(ipf_main_softc_t *softc, ipf_nat_softc_t *softn)
2439 {
2440 	ipnat_t *n;
2441 	int i = 0;
2442 
2443 	if (softn->ipf_nat_map_rules != NULL) {
2444 		bzero((char *)softn->ipf_nat_map_rules,
2445 		      sizeof(*softn->ipf_nat_map_rules) *
2446 		      softn->ipf_nat_maprules_sz);
2447 	}
2448 	if (softn->ipf_nat_rdr_rules != NULL) {
2449 		bzero((char *)softn->ipf_nat_rdr_rules,
2450 		      sizeof(*softn->ipf_nat_rdr_rules) *
2451 		      softn->ipf_nat_rdrrules_sz);
2452 	}
2453 
2454 	while ((n = softn->ipf_nat_list) != NULL) {
2455 		ipf_nat_delrule(softc, softn, n, 0);
2456 		i++;
2457 	}
2458 #if SOLARIS && !defined(INSTANCES)
2459 	pfil_delayed_copy = 1;
2460 #endif
2461 	return (i);
2462 }
2463 
2464 
2465 /* ------------------------------------------------------------------------ */
2466 /* Function:    ipf_nat_delrule                                             */
2467 /* Returns:     Nil                                                         */
2468 /* Parameters:  softc(I) - pointer to soft context main structure           */
2469 /*              softn(I) - pointer to NAT context structure                 */
2470 /*              np(I)    - pointer to NAT rule to delete                    */
2471 /*              purge(I) - 1 == allow purge, 0 == prevent purge             */
2472 /* Locks:       WRITE(ipf_nat)                                              */
2473 /*                                                                          */
2474 /* Preventing "purge" from occuring is allowed because when all of the NAT  */
2475 /* rules are being removed, allowing the "purge" to walk through the list   */
2476 /* of NAT sessions, possibly multiple times, would be a large performance   */
2477 /* hit, on the order of O(N^2).                                             */
2478 /* ------------------------------------------------------------------------ */
2479 static void
ipf_nat_delrule(ipf_main_softc_t * softc,ipf_nat_softc_t * softn,ipnat_t * np,int purge)2480 ipf_nat_delrule(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, ipnat_t *np,
2481 	int purge)
2482 {
2483 
2484 	if (np->in_pnext != NULL) {
2485 		*np->in_pnext = np->in_next;
2486 		if (np->in_next != NULL)
2487 			np->in_next->in_pnext = np->in_pnext;
2488 		if (softn->ipf_nat_list_tail == &np->in_next)
2489 			softn->ipf_nat_list_tail = np->in_pnext;
2490 	}
2491 
2492 	if ((purge == 1) && ((np->in_flags & IPN_PURGE) != 0)) {
2493 		nat_t *next;
2494 		nat_t *nat;
2495 
2496 		for (next = softn->ipf_nat_instances; (nat = next) != NULL;) {
2497 			next = nat->nat_next;
2498 			if (nat->nat_ptr == np)
2499 				ipf_nat_delete(softc, nat, NL_PURGE);
2500 		}
2501 	}
2502 
2503 	if ((np->in_flags & IPN_DELETE) == 0) {
2504 		if (np->in_redir & NAT_REDIRECT) {
2505 			switch (np->in_v[0])
2506 			{
2507 			case 4 :
2508 				ipf_nat_delrdr(softn, np);
2509 				break;
2510 #ifdef USE_INET6
2511 			case 6 :
2512 				ipf_nat6_delrdr(softn, np);
2513 				break;
2514 #endif
2515 			}
2516 		}
2517 		if (np->in_redir & (NAT_MAPBLK|NAT_MAP)) {
2518 			switch (np->in_v[0])
2519 			{
2520 			case 4 :
2521 				ipf_nat_delmap(softn, np);
2522 				break;
2523 #ifdef USE_INET6
2524 			case 6 :
2525 				ipf_nat6_delmap(softn, np);
2526 				break;
2527 #endif
2528 			}
2529 		}
2530 	}
2531 
2532 	np->in_flags |= IPN_DELETE;
2533 	ipf_nat_rule_deref(softc, &np);
2534 }
2535 
2536 
2537 /* ------------------------------------------------------------------------ */
2538 /* Function:    ipf_nat_newmap                                              */
2539 /* Returns:     int - -1 == error, 0 == success                             */
2540 /* Parameters:  fin(I) - pointer to packet information                      */
2541 /*              nat(I) - pointer to NAT entry                               */
2542 /*              ni(I)  - pointer to structure with misc. information needed */
2543 /*                       to create new NAT entry.                           */
2544 /*                                                                          */
2545 /* Given an empty NAT structure, populate it with new information about a   */
2546 /* new NAT session, as defined by the matching NAT rule.                    */
2547 /* ni.nai_ip is passed in uninitialised and must be set, in host byte order,*/
2548 /* to the new IP address for the translation.                               */
2549 /* ------------------------------------------------------------------------ */
2550 static int
ipf_nat_newmap(fr_info_t * fin,nat_t * nat,natinfo_t * ni)2551 ipf_nat_newmap(fr_info_t *fin, nat_t *nat, natinfo_t *ni)
2552 {
2553 	ipf_main_softc_t *softc = fin->fin_main_soft;
2554 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2555 	u_short st_port, dport, sport, port, sp, dp;
2556 	struct in_addr in, inb;
2557 	hostmap_t *hm;
2558 	u_32_t flags;
2559 	u_32_t st_ip;
2560 	ipnat_t *np;
2561 	nat_t *natl;
2562 	int l;
2563 
2564 	/*
2565 	 * If it's an outbound packet which doesn't match any existing
2566 	 * record, then create a new port
2567 	 */
2568 	l = 0;
2569 	hm = NULL;
2570 	np = ni->nai_np;
2571 	st_ip = np->in_snip;
2572 	st_port = np->in_spnext;
2573 	flags = nat->nat_flags;
2574 
2575 	if (flags & IPN_ICMPQUERY) {
2576 		sport = fin->fin_data[1];
2577 		dport = 0;
2578 	} else {
2579 		sport = htons(fin->fin_data[0]);
2580 		dport = htons(fin->fin_data[1]);
2581 	}
2582 
2583 	/*
2584 	 * Do a loop until we either run out of entries to try or we find
2585 	 * a NAT mapping that isn't currently being used.  This is done
2586 	 * because the change to the source is not (usually) being fixed.
2587 	 */
2588 	do {
2589 		port = 0;
2590 		in.s_addr = htonl(np->in_snip);
2591 		if (l == 0) {
2592 			/*
2593 			 * Check to see if there is an existing NAT
2594 			 * setup for this IP address pair.
2595 			 */
2596 			hm = ipf_nat_hostmap(softn, np, fin->fin_src,
2597 					     fin->fin_dst, in, 0);
2598 			if (hm != NULL)
2599 				in.s_addr = hm->hm_nsrcip.s_addr;
2600 		} else if ((l == 1) && (hm != NULL)) {
2601 			ipf_nat_hostmapdel(softc, &hm);
2602 		}
2603 		in.s_addr = ntohl(in.s_addr);
2604 
2605 		nat->nat_hm = hm;
2606 
2607 		if ((np->in_nsrcmsk == 0xffffffff) && (np->in_spnext == 0)) {
2608 			if (l > 0) {
2609 				NBUMPSIDEX(1, ns_exhausted, ns_exhausted_1);
2610 				DT4(ns_exhausted_1, fr_info_t *, fin, nat_t *, nat, natinfo_t *, ni, ipnat_t *, np);
2611 				return (-1);
2612 			}
2613 		}
2614 
2615 		if (np->in_redir == NAT_BIMAP &&
2616 		    np->in_osrcmsk == np->in_nsrcmsk) {
2617 			/*
2618 			 * map the address block in a 1:1 fashion
2619 			 */
2620 			in.s_addr = np->in_nsrcaddr;
2621 			in.s_addr |= fin->fin_saddr & ~np->in_osrcmsk;
2622 			in.s_addr = ntohl(in.s_addr);
2623 
2624 		} else if (np->in_redir & NAT_MAPBLK) {
2625 			if ((l >= np->in_ppip) || ((l > 0) &&
2626 			     !(flags & IPN_TCPUDP))) {
2627 				NBUMPSIDEX(1, ns_exhausted, ns_exhausted_2);
2628 				DT4(ns_exhausted_2, fr_info_t *, fin, nat_t *, nat, natinfo_t *, ni, ipnat_t *, np);
2629 				return (-1);
2630 			}
2631 			/*
2632 			 * map-block - Calculate destination address.
2633 			 */
2634 			in.s_addr = ntohl(fin->fin_saddr);
2635 			in.s_addr &= ntohl(~np->in_osrcmsk);
2636 			inb.s_addr = in.s_addr;
2637 			in.s_addr /= np->in_ippip;
2638 			in.s_addr &= ntohl(~np->in_nsrcmsk);
2639 			in.s_addr += ntohl(np->in_nsrcaddr);
2640 			/*
2641 			 * Calculate destination port.
2642 			 */
2643 			if ((flags & IPN_TCPUDP) &&
2644 			    (np->in_ppip != 0)) {
2645 				port = ntohs(sport) + l;
2646 				port %= np->in_ppip;
2647 				port += np->in_ppip *
2648 					(inb.s_addr % np->in_ippip);
2649 				port += MAPBLK_MINPORT;
2650 				port = htons(port);
2651 			}
2652 
2653 		} else if ((np->in_nsrcaddr == 0) &&
2654 			   (np->in_nsrcmsk == 0xffffffff)) {
2655 			i6addr_t in6;
2656 
2657 			/*
2658 			 * 0/32 - use the interface's IP address.
2659 			 */
2660 			if ((l > 0) ||
2661 			    ipf_ifpaddr(softc, 4, FRI_NORMAL, fin->fin_ifp,
2662 				       &in6, NULL) == -1) {
2663 				NBUMPSIDEX(1, ns_new_ifpaddr, ns_new_ifpaddr_1);
2664 				DT4(ns_new_ifpaddr_1, fr_info_t *, fin, nat_t *, nat, natinfo_t *, ni, ipnat_t *, np);
2665 				return (-1);
2666 			}
2667 			in.s_addr = ntohl(in6.in4.s_addr);
2668 
2669 		} else if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0)) {
2670 			/*
2671 			 * 0/0 - use the original source address/port.
2672 			 */
2673 			if (l > 0) {
2674 				NBUMPSIDEX(1, ns_exhausted, ns_exhausted_3);
2675 				DT4(ns_exhausted_3, fr_info_t *, fin, nat_t *, nat, natinfo_t *, ni, ipnat_t *, np);
2676 				return (-1);
2677 			}
2678 			in.s_addr = ntohl(fin->fin_saddr);
2679 
2680 		} else if ((np->in_nsrcmsk != 0xffffffff) &&
2681 			   (np->in_spnext == 0) && ((l > 0) || (hm == NULL)))
2682 			np->in_snip++;
2683 
2684 		natl = NULL;
2685 
2686 		if ((flags & IPN_TCPUDP) &&
2687 		    ((np->in_redir & NAT_MAPBLK) == 0) &&
2688 		    (np->in_flags & IPN_AUTOPORTMAP)) {
2689 			/*
2690 			 * "ports auto" (without map-block)
2691 			 */
2692 			if ((l > 0) && (l % np->in_ppip == 0)) {
2693 				if ((l > np->in_ppip) &&
2694 				    np->in_nsrcmsk != 0xffffffff)
2695 					np->in_snip++;
2696 			}
2697 			if (np->in_ppip != 0) {
2698 				port = ntohs(sport);
2699 				port += (l % np->in_ppip);
2700 				port %= np->in_ppip;
2701 				port += np->in_ppip *
2702 					(ntohl(fin->fin_saddr) %
2703 					 np->in_ippip);
2704 				port += MAPBLK_MINPORT;
2705 				port = htons(port);
2706 			}
2707 
2708 		} else if (((np->in_redir & NAT_MAPBLK) == 0) &&
2709 			   (flags & IPN_TCPUDPICMP) && (np->in_spnext != 0)) {
2710 			/*
2711 			 * Standard port translation.  Select next port.
2712 			 */
2713 			if (np->in_flags & IPN_SEQUENTIAL) {
2714 				port = np->in_spnext;
2715 			} else {
2716 				port = ipf_random() % (np->in_spmax -
2717 						       np->in_spmin + 1);
2718 				port += np->in_spmin;
2719 			}
2720 			port = htons(port);
2721 			np->in_spnext++;
2722 
2723 			if (np->in_spnext > np->in_spmax) {
2724 				np->in_spnext = np->in_spmin;
2725 				if (np->in_nsrcmsk != 0xffffffff)
2726 					np->in_snip++;
2727 			}
2728 		}
2729 
2730 		if (np->in_flags & IPN_SIPRANGE) {
2731 			if (np->in_snip > ntohl(np->in_nsrcmsk))
2732 				np->in_snip = ntohl(np->in_nsrcaddr);
2733 		} else {
2734 			if ((np->in_nsrcmsk != 0xffffffff) &&
2735 			    ((np->in_snip + 1) & ntohl(np->in_nsrcmsk)) >
2736 			    ntohl(np->in_nsrcaddr))
2737 				np->in_snip = ntohl(np->in_nsrcaddr) + 1;
2738 		}
2739 
2740 		if ((port == 0) && (flags & (IPN_TCPUDPICMP|IPN_ICMPQUERY)))
2741 			port = sport;
2742 
2743 		/*
2744 		 * Here we do a lookup of the connection as seen from
2745 		 * the outside.  If an IP# pair already exists, try
2746 		 * again.  So if you have A->B becomes C->B, you can
2747 		 * also have D->E become C->E but not D->B causing
2748 		 * another C->B.  Also take protocol and ports into
2749 		 * account when determining whether a pre-existing
2750 		 * NAT setup will cause an external conflict where
2751 		 * this is appropriate.
2752 		 */
2753 		inb.s_addr = htonl(in.s_addr);
2754 		sp = fin->fin_data[0];
2755 		dp = fin->fin_data[1];
2756 		fin->fin_data[0] = fin->fin_data[1];
2757 		fin->fin_data[1] = ntohs(port);
2758 		natl = ipf_nat_inlookup(fin, flags & ~(SI_WILDP|NAT_SEARCH),
2759 					(u_int)fin->fin_p, fin->fin_dst, inb);
2760 		fin->fin_data[0] = sp;
2761 		fin->fin_data[1] = dp;
2762 
2763 		/*
2764 		 * Has the search wrapped around and come back to the
2765 		 * start ?
2766 		 */
2767 		if ((natl != NULL) &&
2768 		    (np->in_spnext != 0) && (st_port == np->in_spnext) &&
2769 		    (np->in_snip != 0) && (st_ip == np->in_snip)) {
2770 			NBUMPSIDED(1, ns_wrap);
2771 			DT4(ns_wrap, fr_info_t *, fin, nat_t *, nat, natinfo_t *, ni, ipnat_t *, np);
2772 			return (-1);
2773 		}
2774 		l++;
2775 	} while (natl != NULL);
2776 
2777 	/* Setup the NAT table */
2778 	nat->nat_osrcip = fin->fin_src;
2779 	nat->nat_nsrcaddr = htonl(in.s_addr);
2780 	nat->nat_odstip = fin->fin_dst;
2781 	nat->nat_ndstip = fin->fin_dst;
2782 	if (nat->nat_hm == NULL)
2783 		nat->nat_hm = ipf_nat_hostmap(softn, np, fin->fin_src,
2784 					      fin->fin_dst, nat->nat_nsrcip,
2785 					      0);
2786 
2787 	if (flags & IPN_TCPUDP) {
2788 		nat->nat_osport = sport;
2789 		nat->nat_nsport = port;	/* sport */
2790 		nat->nat_odport = dport;
2791 		nat->nat_ndport = dport;
2792 		((tcphdr_t *)fin->fin_dp)->th_sport = port;
2793 	} else if (flags & IPN_ICMPQUERY) {
2794 		nat->nat_oicmpid = fin->fin_data[1];
2795 		((icmphdr_t *)fin->fin_dp)->icmp_id = port;
2796 		nat->nat_nicmpid = port;
2797 	}
2798 	return (0);
2799 }
2800 
2801 
2802 /* ------------------------------------------------------------------------ */
2803 /* Function:    ipf_nat_newrdr                                              */
2804 /* Returns:     int - -1 == error, 0 == success (no move), 1 == success and */
2805 /*                    allow rule to be moved if IPN_ROUNDR is set.          */
2806 /* Parameters:  fin(I) - pointer to packet information                      */
2807 /*              nat(I) - pointer to NAT entry                               */
2808 /*              ni(I)  - pointer to structure with misc. information needed */
2809 /*                       to create new NAT entry.                           */
2810 /*                                                                          */
2811 /* ni.nai_ip is passed in uninitialised and must be set, in host byte order,*/
2812 /* to the new IP address for the translation.                               */
2813 /* ------------------------------------------------------------------------ */
2814 static int
ipf_nat_newrdr(fr_info_t * fin,nat_t * nat,natinfo_t * ni)2815 ipf_nat_newrdr(fr_info_t *fin, nat_t *nat, natinfo_t *ni)
2816 {
2817 	ipf_main_softc_t *softc = fin->fin_main_soft;
2818 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
2819 	u_short nport, dport, sport;
2820 	struct in_addr in, inb;
2821 	u_short sp, dp;
2822 	hostmap_t *hm;
2823 	u_32_t flags;
2824 	ipnat_t *np;
2825 	nat_t *natl;
2826 	int move;
2827 
2828 	move = 1;
2829 	hm = NULL;
2830 	in.s_addr = 0;
2831 	np = ni->nai_np;
2832 	flags = nat->nat_flags;
2833 
2834 	if (flags & IPN_ICMPQUERY) {
2835 		dport = fin->fin_data[1];
2836 		sport = 0;
2837 	} else {
2838 		sport = htons(fin->fin_data[0]);
2839 		dport = htons(fin->fin_data[1]);
2840 	}
2841 
2842 	/* TRACE sport, dport */
2843 
2844 
2845 	/*
2846 	 * If the matching rule has IPN_STICKY set, then we want to have the
2847 	 * same rule kick in as before.  Why would this happen?  If you have
2848 	 * a collection of rdr rules with "round-robin sticky", the current
2849 	 * packet might match a different one to the previous connection but
2850 	 * we want the same destination to be used.
2851 	 */
2852 	if (((np->in_flags & (IPN_ROUNDR|IPN_SPLIT)) != 0) &&
2853 	    ((np->in_flags & IPN_STICKY) != 0)) {
2854 		hm = ipf_nat_hostmap(softn, NULL, fin->fin_src, fin->fin_dst,
2855 				     in, (u_32_t)dport);
2856 		if (hm != NULL) {
2857 			in.s_addr = ntohl(hm->hm_ndstip.s_addr);
2858 			np = hm->hm_ipnat;
2859 			ni->nai_np = np;
2860 			move = 0;
2861 			ipf_nat_hostmapdel(softc, &hm);
2862 		}
2863 	}
2864 
2865 	/*
2866 	 * Otherwise, it's an inbound packet. Most likely, we don't
2867 	 * want to rewrite source ports and source addresses. Instead,
2868 	 * we want to rewrite to a fixed internal address and fixed
2869 	 * internal port.
2870 	 */
2871 	if (np->in_flags & IPN_SPLIT) {
2872 		in.s_addr = np->in_dnip;
2873 		inb.s_addr = htonl(in.s_addr);
2874 
2875 		if ((np->in_flags & (IPN_ROUNDR|IPN_STICKY)) == IPN_STICKY) {
2876 			hm = ipf_nat_hostmap(softn, NULL, fin->fin_src,
2877 					     fin->fin_dst, inb, (u_32_t)dport);
2878 			if (hm != NULL) {
2879 				in.s_addr = hm->hm_ndstip.s_addr;
2880 				move = 0;
2881 			}
2882 		}
2883 
2884 		if (hm == NULL || hm->hm_ref == 1) {
2885 			if (np->in_ndstaddr == htonl(in.s_addr)) {
2886 				np->in_dnip = ntohl(np->in_ndstmsk);
2887 				move = 0;
2888 			} else {
2889 				np->in_dnip = ntohl(np->in_ndstaddr);
2890 			}
2891 		}
2892 		if (hm != NULL)
2893 			ipf_nat_hostmapdel(softc, &hm);
2894 
2895 	} else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0xffffffff)) {
2896 		i6addr_t in6;
2897 
2898 		/*
2899 		 * 0/32 - use the interface's IP address.
2900 		 */
2901 		if (ipf_ifpaddr(softc, 4, FRI_NORMAL, fin->fin_ifp,
2902 			       &in6, NULL) == -1) {
2903 			NBUMPSIDEX(0, ns_new_ifpaddr, ns_new_ifpaddr_2);
2904 			DT3(ns_new_ifpaddr_2, fr_info_t *, fin, nat_t *, nat, natinfo_t, ni);
2905 			return (-1);
2906 		}
2907 		in.s_addr = ntohl(in6.in4.s_addr);
2908 
2909 	} else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk== 0)) {
2910 		/*
2911 		 * 0/0 - use the original destination address/port.
2912 		 */
2913 		in.s_addr = ntohl(fin->fin_daddr);
2914 
2915 	} else if (np->in_redir == NAT_BIMAP &&
2916 		   np->in_ndstmsk == np->in_odstmsk) {
2917 		/*
2918 		 * map the address block in a 1:1 fashion
2919 		 */
2920 		in.s_addr = np->in_ndstaddr;
2921 		in.s_addr |= fin->fin_daddr & ~np->in_ndstmsk;
2922 		in.s_addr = ntohl(in.s_addr);
2923 	} else {
2924 		in.s_addr = ntohl(np->in_ndstaddr);
2925 	}
2926 
2927 	if ((np->in_dpnext == 0) || ((flags & NAT_NOTRULEPORT) != 0))
2928 		nport = dport;
2929 	else {
2930 		/*
2931 		 * Whilst not optimized for the case where
2932 		 * pmin == pmax, the gain is not significant.
2933 		 */
2934 		if (((np->in_flags & IPN_FIXEDDPORT) == 0) &&
2935 		    (np->in_odport != np->in_dtop)) {
2936 			nport = ntohs(dport) - np->in_odport + np->in_dpmax;
2937 			nport = htons(nport);
2938 		} else {
2939 			nport = htons(np->in_dpnext);
2940 			np->in_dpnext++;
2941 			if (np->in_dpnext > np->in_dpmax)
2942 				np->in_dpnext = np->in_dpmin;
2943 		}
2944 	}
2945 
2946 	/*
2947 	 * When the redirect-to address is set to 0.0.0.0, just
2948 	 * assume a blank `forwarding' of the packet.  We don't
2949 	 * setup any translation for this either.
2950 	 */
2951 	if (in.s_addr == 0) {
2952 		if (nport == dport) {
2953 			NBUMPSIDED(0, ns_xlate_null);
2954 			return (-1);
2955 		}
2956 		in.s_addr = ntohl(fin->fin_daddr);
2957 	}
2958 
2959 	/*
2960 	 * Check to see if this redirect mapping already exists and if
2961 	* it does, return "failure" (allowing it to be created will just
2962 	 * cause one or both of these "connections" to stop working.)
2963 	 */
2964 	inb.s_addr = htonl(in.s_addr);
2965 	sp = fin->fin_data[0];
2966 	dp = fin->fin_data[1];
2967 	fin->fin_data[1] = fin->fin_data[0];
2968 	fin->fin_data[0] = ntohs(nport);
2969 	natl = ipf_nat_outlookup(fin, flags & ~(SI_WILDP|NAT_SEARCH),
2970 			     (u_int)fin->fin_p, inb, fin->fin_src);
2971 	fin->fin_data[0] = sp;
2972 	fin->fin_data[1] = dp;
2973 	if (natl != NULL) {
2974 		DT2(ns_new_xlate_exists, fr_info_t *, fin, nat_t *, natl);
2975 		NBUMPSIDE(0, ns_xlate_exists);
2976 		return (-1);
2977 	}
2978 
2979 	inb.s_addr = htonl(in.s_addr);
2980 	nat->nat_ndstaddr = htonl(in.s_addr);
2981 	nat->nat_odstip = fin->fin_dst;
2982 	nat->nat_nsrcip = fin->fin_src;
2983 	nat->nat_osrcip = fin->fin_src;
2984 	if ((nat->nat_hm == NULL) && ((np->in_flags & IPN_STICKY) != 0))
2985 		nat->nat_hm = ipf_nat_hostmap(softn, np, fin->fin_src,
2986 					      fin->fin_dst, inb, (u_32_t)dport);
2987 
2988 	if (flags & IPN_TCPUDP) {
2989 		nat->nat_odport = dport;
2990 		nat->nat_ndport = nport;
2991 		nat->nat_osport = sport;
2992 		nat->nat_nsport = sport;
2993 		((tcphdr_t *)fin->fin_dp)->th_dport = nport;
2994 	} else if (flags & IPN_ICMPQUERY) {
2995 		nat->nat_oicmpid = fin->fin_data[1];
2996 		((icmphdr_t *)fin->fin_dp)->icmp_id = nport;
2997 		nat->nat_nicmpid = nport;
2998 	}
2999 
3000 	return (move);
3001 }
3002 
3003 /* ------------------------------------------------------------------------ */
3004 /* Function:    ipf_nat_add                                                 */
3005 /* Returns:     nat_t* - NULL == failure to create new NAT structure,       */
3006 /*                       else pointer to new NAT structure                  */
3007 /* Parameters:  fin(I)       - pointer to packet information                */
3008 /*              np(I)        - pointer to NAT rule                          */
3009 /*              natsave(I)   - pointer to where to store NAT struct pointer */
3010 /*              flags(I)     - flags describing the current packet          */
3011 /*              direction(I) - direction of packet (in/out)                 */
3012 /* Write Lock:  ipf_nat                                                     */
3013 /*                                                                          */
3014 /* Attempts to create a new NAT entry.  Does not actually change the packet */
3015 /* in any way.                                                              */
3016 /*                                                                          */
3017 /* This function is in three main parts: (1) deal with creating a new NAT   */
3018 /* structure for a "MAP" rule (outgoing NAT translation); (2) deal with     */
3019 /* creating a new NAT structure for a "RDR" rule (incoming NAT translation) */
3020 /* and (3) building that structure and putting it into the NAT table(s).    */
3021 /*                                                                          */
3022 /* NOTE: natsave should NOT be used to point back to an ipstate_t struct    */
3023 /*       as it can result in memory being corrupted.                        */
3024 /* ------------------------------------------------------------------------ */
3025 nat_t *
ipf_nat_add(fr_info_t * fin,ipnat_t * np,nat_t ** natsave,u_int flags,int direction)3026 ipf_nat_add(fr_info_t *fin, ipnat_t *np, nat_t **natsave, u_int flags,
3027 	int direction)
3028 {
3029 	ipf_main_softc_t *softc = fin->fin_main_soft;
3030 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3031 	hostmap_t *hm = NULL;
3032 	nat_t *nat, *natl;
3033 	natstat_t *nsp;
3034 	u_int nflags;
3035 	natinfo_t ni;
3036 	int move;
3037 
3038 	nsp = &softn->ipf_nat_stats;
3039 
3040 	if ((nsp->ns_active * 100 / softn->ipf_nat_table_max) >
3041 	    softn->ipf_nat_table_wm_high) {
3042 		softn->ipf_nat_doflush = 1;
3043 	}
3044 
3045 	if (nsp->ns_active >= softn->ipf_nat_table_max) {
3046 		NBUMPSIDED(fin->fin_out, ns_table_max);
3047 		DT2(ns_table_max, nat_stat_t *, nsp, ipf_nat_softc_t *, softn);
3048 		return (NULL);
3049 	}
3050 
3051 	move = 1;
3052 	nflags = np->in_flags & flags;
3053 	nflags &= NAT_FROMRULE;
3054 
3055 	ni.nai_np = np;
3056 	ni.nai_dport = 0;
3057 	ni.nai_sport = 0;
3058 
3059 	/* Give me a new nat */
3060 	KMALLOC(nat, nat_t *);
3061 	if (nat == NULL) {
3062 		DT(ns_memfail);
3063 		NBUMPSIDED(fin->fin_out, ns_memfail);
3064 		/*
3065 		 * Try to automatically tune the max # of entries in the
3066 		 * table allowed to be less than what will cause kmem_alloc()
3067 		 * to fail and try to eliminate panics due to out of memory
3068 		 * conditions arising.
3069 		 */
3070 		if ((softn->ipf_nat_table_max > softn->ipf_nat_table_sz) &&
3071 		    (nsp->ns_active > 100)) {
3072 			softn->ipf_nat_table_max = nsp->ns_active - 100;
3073 			printf("table_max reduced to %d\n",
3074 				softn->ipf_nat_table_max);
3075 		}
3076 		return (NULL);
3077 	}
3078 
3079 	if (flags & IPN_ICMPQUERY) {
3080 		/*
3081 		 * In the ICMP query NAT code, we translate the ICMP id fields
3082 		 * to make them unique. This is indepedent of the ICMP type
3083 		 * (e.g. in the unlikely event that a host sends an echo and
3084 		 * an tstamp request with the same id, both packets will have
3085 		 * their ip address/id field changed in the same way).
3086 		 */
3087 		/* The icmp_id field is used by the sender to identify the
3088 		 * process making the icmp request. (the receiver justs
3089 		 * copies it back in its response). So, it closely matches
3090 		 * the concept of source port. We overlay sport, so we can
3091 		 * maximally reuse the existing code.
3092 		 */
3093 		ni.nai_sport = fin->fin_data[1];
3094 		ni.nai_dport = 0;
3095 	}
3096 
3097 	bzero((char *)nat, sizeof(*nat));
3098 	nat->nat_flags = flags;
3099 	nat->nat_redir = np->in_redir;
3100 	nat->nat_dir = direction;
3101 	nat->nat_pr[0] = fin->fin_p;
3102 	nat->nat_pr[1] = fin->fin_p;
3103 
3104 	/*
3105 	 * Search the current table for a match and create a new mapping
3106 	 * if there is none found.
3107 	 */
3108 	if (np->in_redir & NAT_DIVERTUDP) {
3109 		move = ipf_nat_newdivert(fin, nat, &ni);
3110 
3111 	} else if (np->in_redir & NAT_REWRITE) {
3112 		move = ipf_nat_newrewrite(fin, nat, &ni);
3113 
3114 	} else if (direction == NAT_OUTBOUND) {
3115 		/*
3116 		 * We can now arrange to call this for the same connection
3117 		 * because ipf_nat_new doesn't protect the code path into
3118 		 * this function.
3119 		 */
3120 		natl = ipf_nat_outlookup(fin, nflags, (u_int)fin->fin_p,
3121 				     fin->fin_src, fin->fin_dst);
3122 		if (natl != NULL) {
3123 			KFREE(nat);
3124 			nat = natl;
3125 			goto done;
3126 		}
3127 
3128 		move = ipf_nat_newmap(fin, nat, &ni);
3129 	} else {
3130 		/*
3131 		 * NAT_INBOUND is used for redirects rules
3132 		 */
3133 		natl = ipf_nat_inlookup(fin, nflags, (u_int)fin->fin_p,
3134 					fin->fin_src, fin->fin_dst);
3135 		if (natl != NULL) {
3136 			KFREE(nat);
3137 			nat = natl;
3138 			goto done;
3139 		}
3140 
3141 		move = ipf_nat_newrdr(fin, nat, &ni);
3142 	}
3143 	if (move == -1)
3144 		goto badnat;
3145 
3146 	np = ni.nai_np;
3147 
3148 	nat->nat_mssclamp = np->in_mssclamp;
3149 	nat->nat_me = natsave;
3150 	nat->nat_fr = fin->fin_fr;
3151 	nat->nat_rev = fin->fin_rev;
3152 	nat->nat_ptr = np;
3153 	nat->nat_dlocal = np->in_dlocal;
3154 
3155 	if ((np->in_apr != NULL) && ((nat->nat_flags & NAT_SLAVE) == 0)) {
3156 		if (ipf_proxy_new(fin, nat) == -1) {
3157 			NBUMPSIDED(fin->fin_out, ns_appr_fail);
3158 			DT3(ns_appr_fail, fr_info_t *, fin, nat_t *, nat, ipnat_t *, np);
3159 			goto badnat;
3160 		}
3161 	}
3162 
3163 	nat->nat_ifps[0] = np->in_ifps[0];
3164 	if (np->in_ifps[0] != NULL) {
3165 		COPYIFNAME(np->in_v[0], np->in_ifps[0], nat->nat_ifnames[0]);
3166 	}
3167 
3168 	nat->nat_ifps[1] = np->in_ifps[1];
3169 	if (np->in_ifps[1] != NULL) {
3170 		COPYIFNAME(np->in_v[1], np->in_ifps[1], nat->nat_ifnames[1]);
3171 	}
3172 
3173 	if (ipf_nat_finalise(fin, nat) == -1) {
3174 		goto badnat;
3175 	}
3176 
3177 	np->in_use++;
3178 
3179 	if ((move == 1) && (np->in_flags & IPN_ROUNDR)) {
3180 		if ((np->in_redir & (NAT_REDIRECT|NAT_MAP)) == NAT_REDIRECT) {
3181 			ipf_nat_delrdr(softn, np);
3182 			ipf_nat_addrdr(softn, np);
3183 		} else if ((np->in_redir & (NAT_REDIRECT|NAT_MAP)) == NAT_MAP) {
3184 			ipf_nat_delmap(softn, np);
3185 			ipf_nat_addmap(softn, np);
3186 		}
3187 	}
3188 
3189 	if (flags & SI_WILDP)
3190 		nsp->ns_wilds++;
3191 	nsp->ns_proto[nat->nat_pr[0]]++;
3192 
3193 	goto done;
3194 badnat:
3195 	DT3(ns_badnatnew, fr_info_t *, fin, nat_t *, nat, ipnat_t *, np);
3196 	NBUMPSIDE(fin->fin_out, ns_badnatnew);
3197 	if ((hm = nat->nat_hm) != NULL)
3198 		ipf_nat_hostmapdel(softc, &hm);
3199 	KFREE(nat);
3200 	nat = NULL;
3201 done:
3202 	if (nat != NULL && np != NULL)
3203 		np->in_hits++;
3204 	if (natsave != NULL)
3205 		*natsave = nat;
3206 	return (nat);
3207 }
3208 
3209 
3210 /* ------------------------------------------------------------------------ */
3211 /* Function:    ipf_nat_finalise                                            */
3212 /* Returns:     int - 0 == sucess, -1 == failure                            */
3213 /* Parameters:  fin(I) - pointer to packet information                      */
3214 /*              nat(I) - pointer to NAT entry                               */
3215 /* Write Lock:  ipf_nat                                                     */
3216 /*                                                                          */
3217 /* This is the tail end of constructing a new NAT entry and is the same     */
3218 /* for both IPv4 and IPv6.                                                  */
3219 /* ------------------------------------------------------------------------ */
3220 /*ARGSUSED*/
3221 static int
ipf_nat_finalise(fr_info_t * fin,nat_t * nat)3222 ipf_nat_finalise(fr_info_t *fin, nat_t *nat)
3223 {
3224 	ipf_main_softc_t *softc = fin->fin_main_soft;
3225 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3226 	u_32_t sum1, sum2, sumd;
3227 	frentry_t *fr;
3228 #if SOLARIS && defined(_KERNEL) && defined(ICK_M_CTL_MAGIC)
3229 	qpktinfo_t *qpi = fin->fin_qpi;
3230 #endif
3231 
3232 	switch (nat->nat_pr[0])
3233 	{
3234 	case IPPROTO_ICMP :
3235 		sum1 = LONG_SUM(ntohs(nat->nat_oicmpid));
3236 		sum2 = LONG_SUM(ntohs(nat->nat_nicmpid));
3237 		CALC_SUMD(sum1, sum2, sumd);
3238 		nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
3239 
3240 		break;
3241 
3242 	default :
3243 		sum1 = LONG_SUM(ntohl(nat->nat_osrcaddr) + \
3244 				ntohs(nat->nat_osport));
3245 		sum2 = LONG_SUM(ntohl(nat->nat_nsrcaddr) + \
3246 				ntohs(nat->nat_nsport));
3247 		CALC_SUMD(sum1, sum2, sumd);
3248 		nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
3249 
3250 		sum1 = LONG_SUM(ntohl(nat->nat_odstaddr) + \
3251 				ntohs(nat->nat_odport));
3252 		sum2 = LONG_SUM(ntohl(nat->nat_ndstaddr) + \
3253 				ntohs(nat->nat_ndport));
3254 		CALC_SUMD(sum1, sum2, sumd);
3255 		nat->nat_sumd[0] += (sumd & 0xffff) + (sumd >> 16);
3256 		break;
3257 	}
3258 
3259 	/*
3260 	 * Compute the partial checksum, just in case.
3261 	 * This is only ever placed into outbound packets so care needs
3262 	 * to be taken over which pair of addresses are used.
3263 	 */
3264 	if (nat->nat_dir == NAT_OUTBOUND) {
3265 		sum1 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
3266 		sum1 += LONG_SUM(ntohl(nat->nat_ndstaddr));
3267 	} else {
3268 		sum1 = LONG_SUM(ntohl(nat->nat_osrcaddr));
3269 		sum1 += LONG_SUM(ntohl(nat->nat_odstaddr));
3270 	}
3271 	sum1 += nat->nat_pr[1];
3272 	nat->nat_sumd[1] = (sum1 & 0xffff) + (sum1 >> 16);
3273 
3274 	sum1 = LONG_SUM(ntohl(nat->nat_osrcaddr));
3275 	sum2 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
3276 	CALC_SUMD(sum1, sum2, sumd);
3277 	nat->nat_ipsumd = (sumd & 0xffff) + (sumd >> 16);
3278 
3279 	sum1 = LONG_SUM(ntohl(nat->nat_odstaddr));
3280 	sum2 = LONG_SUM(ntohl(nat->nat_ndstaddr));
3281 	CALC_SUMD(sum1, sum2, sumd);
3282 	nat->nat_ipsumd += (sumd & 0xffff) + (sumd >> 16);
3283 
3284 	nat->nat_v[0] = 4;
3285 	nat->nat_v[1] = 4;
3286 
3287 	if ((nat->nat_ifps[0] != NULL) && (nat->nat_ifps[0] != (void *)-1)) {
3288 		nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
3289 	}
3290 
3291 	if ((nat->nat_ifps[1] != NULL) && (nat->nat_ifps[1] != (void *)-1)) {
3292 		nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
3293 	}
3294 
3295 	if ((nat->nat_flags & SI_CLONE) == 0)
3296 		nat->nat_sync = ipf_sync_new(softc, SMC_NAT, fin, nat);
3297 
3298 	if (ipf_nat_insert(softc, softn, nat) == 0) {
3299 		if (softn->ipf_nat_logging)
3300 			ipf_nat_log(softc, softn, nat, NL_NEW);
3301 		fr = nat->nat_fr;
3302 		if (fr != NULL) {
3303 			MUTEX_ENTER(&fr->fr_lock);
3304 			fr->fr_ref++;
3305 			MUTEX_EXIT(&fr->fr_lock);
3306 		}
3307 		return (0);
3308 	}
3309 
3310 	NBUMPSIDED(fin->fin_out, ns_unfinalised);
3311 	DT2(ns_unfinalised, fr_info_t *, fin, nat_t *, nat);
3312 	/*
3313 	 * nat_insert failed, so cleanup time...
3314 	 */
3315 	if (nat->nat_sync != NULL)
3316 		ipf_sync_del_nat(softc->ipf_sync_soft, nat->nat_sync);
3317 	return (-1);
3318 }
3319 
3320 
3321 /* ------------------------------------------------------------------------ */
3322 /* Function:    ipf_nat_insert                                              */
3323 /* Returns:     int - 0 == sucess, -1 == failure                            */
3324 /* Parameters:  softc(I) - pointer to soft context main structure           */
3325 /*              softn(I) - pointer to NAT context structure                 */
3326 /*              nat(I) - pointer to NAT structure                           */
3327 /* Write Lock:  ipf_nat                                                     */
3328 /*                                                                          */
3329 /* Insert a NAT entry into the hash tables for searching and add it to the  */
3330 /* list of active NAT entries.  Adjust global counters when complete.       */
3331 /* ------------------------------------------------------------------------ */
3332 int
ipf_nat_insert(ipf_main_softc_t * softc,ipf_nat_softc_t * softn,nat_t * nat)3333 ipf_nat_insert(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, nat_t *nat)
3334 {
3335 	u_int hv0, hv1;
3336 	u_int sp, dp;
3337 	ipnat_t *in;
3338 	int ret;
3339 
3340 	/*
3341 	* Try and return an error as early as possible, so calculate the hash
3342 	 * entry numbers first and then proceed.
3343 	 */
3344 	if ((nat->nat_flags & (SI_W_SPORT|SI_W_DPORT)) == 0) {
3345 		if ((nat->nat_flags & IPN_TCPUDP) != 0) {
3346 			sp = nat->nat_osport;
3347 			dp = nat->nat_odport;
3348 		} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
3349 			sp = 0;
3350 			dp = nat->nat_oicmpid;
3351 		} else {
3352 			sp = 0;
3353 			dp = 0;
3354 		}
3355 		hv0 = NAT_HASH_FN(nat->nat_osrcaddr, sp, 0xffffffff);
3356 		hv0 = NAT_HASH_FN(nat->nat_odstaddr, hv0 + dp, 0xffffffff);
3357 		/*
3358 		 * TRACE nat_osrcaddr, nat_osport, nat_odstaddr,
3359 		 * nat_odport, hv0
3360 		 */
3361 
3362 		if ((nat->nat_flags & IPN_TCPUDP) != 0) {
3363 			sp = nat->nat_nsport;
3364 			dp = nat->nat_ndport;
3365 		} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
3366 			sp = 0;
3367 			dp = nat->nat_nicmpid;
3368 		} else {
3369 			sp = 0;
3370 			dp = 0;
3371 		}
3372 		hv1 = NAT_HASH_FN(nat->nat_nsrcaddr, sp, 0xffffffff);
3373 		hv1 = NAT_HASH_FN(nat->nat_ndstaddr, hv1 + dp, 0xffffffff);
3374 		/*
3375 		 * TRACE nat_nsrcaddr, nat_nsport, nat_ndstaddr,
3376 		 * nat_ndport, hv1
3377 		 */
3378 	} else {
3379 		hv0 = NAT_HASH_FN(nat->nat_osrcaddr, 0, 0xffffffff);
3380 		hv0 = NAT_HASH_FN(nat->nat_odstaddr, hv0, 0xffffffff);
3381 		/* TRACE nat_osrcaddr, nat_odstaddr, hv0 */
3382 
3383 		hv1 = NAT_HASH_FN(nat->nat_nsrcaddr, 0, 0xffffffff);
3384 		hv1 = NAT_HASH_FN(nat->nat_ndstaddr, hv1, 0xffffffff);
3385 		/* TRACE nat_nsrcaddr, nat_ndstaddr, hv1 */
3386 	}
3387 
3388 	nat->nat_hv[0] = hv0;
3389 	nat->nat_hv[1] = hv1;
3390 
3391 	MUTEX_INIT(&nat->nat_lock, "nat entry lock");
3392 
3393 	in = nat->nat_ptr;
3394 	nat->nat_ref = nat->nat_me ? 2 : 1;
3395 
3396 	nat->nat_ifnames[0][LIFNAMSIZ - 1] = '\0';
3397 	nat->nat_ifps[0] = ipf_resolvenic(softc, nat->nat_ifnames[0], 4);
3398 
3399 	if (nat->nat_ifnames[1][0] != '\0') {
3400 		nat->nat_ifnames[1][LIFNAMSIZ - 1] = '\0';
3401 		nat->nat_ifps[1] = ipf_resolvenic(softc,
3402 						  nat->nat_ifnames[1], 4);
3403 	} else if (in->in_ifnames[1] != -1) {
3404 		char *name;
3405 
3406 		name = in->in_names + in->in_ifnames[1];
3407 		if (name[1] != '\0' && name[0] != '-' && name[0] != '*') {
3408 			(void) strncpy(nat->nat_ifnames[1],
3409 				       nat->nat_ifnames[0], LIFNAMSIZ);
3410 			nat->nat_ifnames[1][LIFNAMSIZ - 1] = '\0';
3411 			nat->nat_ifps[1] = nat->nat_ifps[0];
3412 		}
3413 	}
3414 	if ((nat->nat_ifps[0] != NULL) && (nat->nat_ifps[0] != (void *)-1)) {
3415 		nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
3416 	}
3417 	if ((nat->nat_ifps[1] != NULL) && (nat->nat_ifps[1] != (void *)-1)) {
3418 		nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
3419 	}
3420 
3421 	ret = ipf_nat_hashtab_add(softc, softn, nat);
3422 	if (ret == -1)
3423 		MUTEX_DESTROY(&nat->nat_lock);
3424 	return (ret);
3425 }
3426 
3427 
3428 /* ------------------------------------------------------------------------ */
3429 /* Function:    ipf_nat_hashtab_add                                         */
3430 /* Returns:     int - 0 == sucess, -1 == failure                            */
3431 /* Parameters:  softc(I) - pointer to soft context main structure           */
3432 /*              softn(I) - pointer to NAT context structure                 */
3433 /*              nat(I) - pointer to NAT structure                           */
3434 /*                                                                          */
3435 /* Handle the insertion of a NAT entry into the table/list.                 */
3436 /* ------------------------------------------------------------------------ */
3437 int
ipf_nat_hashtab_add(ipf_main_softc_t * softc,ipf_nat_softc_t * softn,nat_t * nat)3438 ipf_nat_hashtab_add(ipf_main_softc_t *softc, ipf_nat_softc_t *softn,
3439 	nat_t *nat)
3440 {
3441 	nat_t **natp;
3442 	u_int hv0;
3443 	u_int hv1;
3444 
3445 	if (nat->nat_dir == NAT_INBOUND || nat->nat_dir == NAT_DIVERTIN) {
3446 		hv1 = nat->nat_hv[0] % softn->ipf_nat_table_sz;
3447 		hv0 = nat->nat_hv[1] % softn->ipf_nat_table_sz;
3448 	} else {
3449 		hv0 = nat->nat_hv[0] % softn->ipf_nat_table_sz;
3450 		hv1 = nat->nat_hv[1] % softn->ipf_nat_table_sz;
3451 	}
3452 
3453 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen[hv0] >=
3454 	    softn->ipf_nat_maxbucket) {
3455 		DT1(ns_bucket_max_0, int,
3456 		    softn->ipf_nat_stats.ns_side[0].ns_bucketlen[hv0]);
3457 		NBUMPSIDE(0, ns_bucket_max);
3458 		return (-1);
3459 	}
3460 
3461 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen[hv1] >=
3462 	    softn->ipf_nat_maxbucket) {
3463 		DT1(ns_bucket_max_1, int,
3464 		    softn->ipf_nat_stats.ns_side[1].ns_bucketlen[hv1]);
3465 		NBUMPSIDE(1, ns_bucket_max);
3466 		return (-1);
3467 	}
3468 
3469 	/*
3470 	 * The ordering of operations in the list and hash table insertion
3471 	 * is very important.  The last operation for each task should be
3472 	 * to update the top of the list, after all the "nexts" have been
3473 	 * done so that walking the list while it is being done does not
3474 	 * find strange pointers.
3475 	 *
3476 	 * Global list of NAT instances
3477 	 */
3478 	nat->nat_next = softn->ipf_nat_instances;
3479 	nat->nat_pnext = &softn->ipf_nat_instances;
3480 	if (softn->ipf_nat_instances)
3481 		softn->ipf_nat_instances->nat_pnext = &nat->nat_next;
3482 	softn->ipf_nat_instances = nat;
3483 
3484 	/*
3485 	 * Inbound hash table.
3486 	 */
3487 	natp = &softn->ipf_nat_table[0][hv0];
3488 	nat->nat_phnext[0] = natp;
3489 	nat->nat_hnext[0] = *natp;
3490 	if (*natp) {
3491 		(*natp)->nat_phnext[0] = &nat->nat_hnext[0];
3492 	} else {
3493 		NBUMPSIDE(0, ns_inuse);
3494 	}
3495 	*natp = nat;
3496 	NBUMPSIDE(0, ns_bucketlen[hv0]);
3497 
3498 	/*
3499 	 * Outbound hash table.
3500 	 */
3501 	natp = &softn->ipf_nat_table[1][hv1];
3502 	nat->nat_phnext[1] = natp;
3503 	nat->nat_hnext[1] = *natp;
3504 	if (*natp)
3505 		(*natp)->nat_phnext[1] = &nat->nat_hnext[1];
3506 	else {
3507 		NBUMPSIDE(1, ns_inuse);
3508 	}
3509 	*natp = nat;
3510 	NBUMPSIDE(1, ns_bucketlen[hv1]);
3511 
3512 	ipf_nat_setqueue(softc, softn, nat);
3513 
3514 	if (nat->nat_dir & NAT_OUTBOUND) {
3515 		NBUMPSIDE(1, ns_added);
3516 	} else {
3517 		NBUMPSIDE(0, ns_added);
3518 	}
3519 	softn->ipf_nat_stats.ns_active++;
3520 	return (0);
3521 }
3522 
3523 
3524 /* ------------------------------------------------------------------------ */
3525 /* Function:    ipf_nat_icmperrorlookup                                     */
3526 /* Returns:     nat_t* - point to matching NAT structure                    */
3527 /* Parameters:  fin(I) - pointer to packet information                      */
3528 /*              dir(I) - direction of packet (in/out)                       */
3529 /*                                                                          */
3530 /* Check if the ICMP error message is related to an existing TCP, UDP or    */
3531 /* ICMP query nat entry.  It is assumed that the packet is already of the   */
3532 /* the required length.                                                     */
3533 /* ------------------------------------------------------------------------ */
3534 nat_t *
ipf_nat_icmperrorlookup(fr_info_t * fin,int dir)3535 ipf_nat_icmperrorlookup(fr_info_t *fin, int dir)
3536 {
3537 	ipf_main_softc_t *softc = fin->fin_main_soft;
3538 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3539 	int flags = 0, minlen;
3540 	icmphdr_t *orgicmp;
3541 	nat_stat_side_t *nside;
3542 	tcphdr_t *tcp = NULL;
3543 	u_short data[2];
3544 	nat_t *nat;
3545 	ip_t *oip;
3546 	u_int p;
3547 
3548 	nside = &softn->ipf_nat_stats.ns_side[fin->fin_out];
3549 	/*
3550 	 * Does it at least have the return (basic) IP header ?
3551 	 * Only a basic IP header (no options) should be with an ICMP error
3552 	 * header.  Also, if it's not an error type, then return.
3553 	 */
3554 	if ((fin->fin_hlen != sizeof(ip_t)) || !(fin->fin_flx & FI_ICMPERR)) {
3555 		ATOMIC_INCL(nside->ns_icmp_basic);
3556 		return (NULL);
3557 	}
3558 
3559 	/*
3560 	 * Check packet size
3561 	 */
3562 	oip = (ip_t *)((char *)fin->fin_dp + 8);
3563 	minlen = IP_HL(oip) << 2;
3564 	if ((minlen < sizeof(ip_t)) ||
3565 	    (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen)) {
3566 		ATOMIC_INCL(nside->ns_icmp_size);
3567 		return (NULL);
3568 	}
3569 
3570 	/*
3571 	 * Is the buffer big enough for all of it ?  It's the size of the IP
3572 	 * header claimed in the encapsulated part which is of concern.  It
3573 	 * may be too big to be in this buffer but not so big that it's
3574 	 * outside the ICMP packet, leading to TCP deref's causing problems.
3575 	 * This is possible because we don't know how big oip_hl is when we
3576 	 * do the pullup early in ipf_check() and thus can't gaurantee it is
3577 	 * all here now.
3578 	 */
3579 #ifdef  ipf_nat_KERNEL
3580 	{
3581 	mb_t *m;
3582 
3583 	m = fin->fin_m;
3584 # if SOLARIS
3585 	if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN >
3586 	    (char *)m->b_wptr) {
3587 		ATOMIC_INCL(nside->ns_icmp_mbuf);
3588 		return (NULL);
3589 	}
3590 # else
3591 	if ((char *)oip + fin->fin_dlen - ICMPERR_ICMPHLEN >
3592 	    (char *)fin->fin_ip + M_LEN(m)) {
3593 		ATOMIC_INCL(nside->ns_icmp_mbuf);
3594 		return (NULL);
3595 	}
3596 # endif
3597 	}
3598 #endif
3599 
3600 	if (fin->fin_daddr != oip->ip_src.s_addr) {
3601 		ATOMIC_INCL(nside->ns_icmp_address);
3602 		return (NULL);
3603 	}
3604 
3605 	p = oip->ip_p;
3606 	if (p == IPPROTO_TCP)
3607 		flags = IPN_TCP;
3608 	else if (p == IPPROTO_UDP)
3609 		flags = IPN_UDP;
3610 	else if (p == IPPROTO_ICMP) {
3611 		orgicmp = (icmphdr_t *)((char *)oip + (IP_HL(oip) << 2));
3612 
3613 		/* see if this is related to an ICMP query */
3614 		if (ipf_nat_icmpquerytype(orgicmp->icmp_type)) {
3615 			data[0] = fin->fin_data[0];
3616 			data[1] = fin->fin_data[1];
3617 			fin->fin_data[0] = 0;
3618 			fin->fin_data[1] = orgicmp->icmp_id;
3619 
3620 			flags = IPN_ICMPERR|IPN_ICMPQUERY;
3621 			/*
3622 			 * NOTE : dir refers to the direction of the original
3623 			 *        ip packet. By definition the icmp error
3624 			 *        message flows in the opposite direction.
3625 			 */
3626 			if (dir == NAT_INBOUND)
3627 				nat = ipf_nat_inlookup(fin, flags, p,
3628 						       oip->ip_dst,
3629 						       oip->ip_src);
3630 			else
3631 				nat = ipf_nat_outlookup(fin, flags, p,
3632 							oip->ip_dst,
3633 							oip->ip_src);
3634 			fin->fin_data[0] = data[0];
3635 			fin->fin_data[1] = data[1];
3636 			return (nat);
3637 		}
3638 	}
3639 
3640 	if (flags & IPN_TCPUDP) {
3641 		minlen += 8;		/* + 64bits of data to get ports */
3642 		/* TRACE (fin,minlen) */
3643 		if (fin->fin_plen < ICMPERR_IPICMPHLEN + minlen) {
3644 			ATOMIC_INCL(nside->ns_icmp_short);
3645 			return (NULL);
3646 		}
3647 
3648 		data[0] = fin->fin_data[0];
3649 		data[1] = fin->fin_data[1];
3650 		tcp = (tcphdr_t *)((char *)oip + (IP_HL(oip) << 2));
3651 		fin->fin_data[0] = ntohs(tcp->th_dport);
3652 		fin->fin_data[1] = ntohs(tcp->th_sport);
3653 
3654 		if (dir == NAT_INBOUND) {
3655 			nat = ipf_nat_inlookup(fin, flags, p, oip->ip_dst,
3656 					       oip->ip_src);
3657 		} else {
3658 			nat = ipf_nat_outlookup(fin, flags, p, oip->ip_dst,
3659 					    oip->ip_src);
3660 		}
3661 		fin->fin_data[0] = data[0];
3662 		fin->fin_data[1] = data[1];
3663 		return (nat);
3664 	}
3665 	if (dir == NAT_INBOUND)
3666 		nat = ipf_nat_inlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
3667 	else
3668 		nat = ipf_nat_outlookup(fin, 0, p, oip->ip_dst, oip->ip_src);
3669 
3670 	return (nat);
3671 }
3672 
3673 
3674 /* ------------------------------------------------------------------------ */
3675 /* Function:    ipf_nat_icmperror                                           */
3676 /* Returns:     nat_t* - point to matching NAT structure                    */
3677 /* Parameters:  fin(I)    - pointer to packet information                   */
3678 /*              nflags(I) - NAT flags for this packet                       */
3679 /*              dir(I)    - direction of packet (in/out)                    */
3680 /*                                                                          */
3681 /* Fix up an ICMP packet which is an error message for an existing NAT      */
3682 /* session.  This will correct both packet header data and checksums.       */
3683 /*                                                                          */
3684 /* This should *ONLY* be used for incoming ICMP error packets to make sure  */
3685 /* a NAT'd ICMP packet gets correctly recognised.                           */
3686 /* ------------------------------------------------------------------------ */
3687 nat_t *
ipf_nat_icmperror(fr_info_t * fin,u_int * nflags,int dir)3688 ipf_nat_icmperror(fr_info_t *fin, u_int *nflags, int dir)
3689 {
3690 	ipf_main_softc_t *softc = fin->fin_main_soft;
3691 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3692 	u_32_t sum1, sum2, sumd, sumd2;
3693 	struct in_addr a1, a2, a3, a4;
3694 	int flags, dlen, odst;
3695 	icmphdr_t *icmp;
3696 	u_short *csump;
3697 	tcphdr_t *tcp;
3698 	nat_t *nat;
3699 	ip_t *oip;
3700 	void *dp;
3701 
3702 	if ((fin->fin_flx & (FI_SHORT|FI_FRAGBODY))) {
3703 		NBUMPSIDED(fin->fin_out, ns_icmp_short);
3704 		return (NULL);
3705 	}
3706 
3707 	/*
3708 	* ipf_nat_icmperrorlookup() will return NULL for `defective' packets.
3709 	 */
3710 	if ((fin->fin_v != 4) || !(nat = ipf_nat_icmperrorlookup(fin, dir))) {
3711 		NBUMPSIDED(fin->fin_out, ns_icmp_notfound);
3712 		return (NULL);
3713 	}
3714 
3715 	tcp = NULL;
3716 	csump = NULL;
3717 	flags = 0;
3718 	sumd2 = 0;
3719 	*nflags = IPN_ICMPERR;
3720 	icmp = fin->fin_dp;
3721 	oip = (ip_t *)&icmp->icmp_ip;
3722 	dp = (((char *)oip) + (IP_HL(oip) << 2));
3723 	if (oip->ip_p == IPPROTO_TCP) {
3724 		tcp = (tcphdr_t *)dp;
3725 		csump = (u_short *)&tcp->th_sum;
3726 		flags = IPN_TCP;
3727 	} else if (oip->ip_p == IPPROTO_UDP) {
3728 		udphdr_t *udp;
3729 
3730 		udp = (udphdr_t *)dp;
3731 		tcp = (tcphdr_t *)dp;
3732 		csump = (u_short *)&udp->uh_sum;
3733 		flags = IPN_UDP;
3734 	} else if (oip->ip_p == IPPROTO_ICMP)
3735 		flags = IPN_ICMPQUERY;
3736 	dlen = fin->fin_plen - ((char *)dp - (char *)fin->fin_ip);
3737 
3738 	/*
3739 	 * Need to adjust ICMP header to include the real IP#'s and
3740 	 * port #'s.  Only apply a checksum change relative to the
3741 	 * IP address change as it will be modified again in ipf_nat_checkout
3742 	 * for both address and port.  Two checksum changes are
3743 	 * necessary for the two header address changes.  Be careful
3744 	 * to only modify the checksum once for the port # and twice
3745 	 * for the IP#.
3746 	 */
3747 
3748 	/*
3749 	 * Step 1
3750 	 * Fix the IP addresses in the offending IP packet. You also need
3751 	 * to adjust the IP header checksum of that offending IP packet.
3752 	 *
3753 	 * Normally, you would expect that the ICMP checksum of the
3754 	 * ICMP error message needs to be adjusted as well for the
3755 	 * IP address change in oip.
3756 	 * However, this is a NOP, because the ICMP checksum is
3757 	 * calculated over the complete ICMP packet, which includes the
3758 	 * changed oip IP addresses and oip->ip_sum. However, these
3759 	 * two changes cancel each other out (if the delta for
3760 	 * the IP address is x, then the delta for ip_sum is minus x),
3761 	 * so no change in the icmp_cksum is necessary.
3762 	 *
3763 	 * Inbound ICMP
3764 	 * ------------
3765 	 * MAP rule, SRC=a,DST=b -> SRC=c,DST=b
3766 	 * - response to outgoing packet (a,b)=>(c,b) (OIP_SRC=c,OIP_DST=b)
3767 	 * - OIP_SRC(c)=nat_newsrcip,          OIP_DST(b)=nat_newdstip
3768 	 *=> OIP_SRC(c)=nat_oldsrcip,          OIP_DST(b)=nat_olddstip
3769 	 *
3770 	 * RDR rule, SRC=a,DST=b -> SRC=a,DST=c
3771 	 * - response to outgoing packet (c,a)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3772 	 * - OIP_SRC(b)=nat_olddstip,          OIP_DST(a)=nat_oldsrcip
3773 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3774 	 *
3775 	 * REWRITE out rule, SRC=a,DST=b -> SRC=c,DST=d
3776 	 * - response to outgoing packet (a,b)=>(c,d) (OIP_SRC=c,OIP_DST=d)
3777 	 * - OIP_SRC(c)=nat_newsrcip,          OIP_DST(d)=nat_newdstip
3778 	 *=> OIP_SRC(c)=nat_oldsrcip,          OIP_DST(d)=nat_olddstip
3779 	 *
3780 	 * REWRITE in rule, SRC=a,DST=b -> SRC=c,DST=d
3781 	 * - response to outgoing packet (d,c)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3782 	 * - OIP_SRC(b)=nat_olddstip,          OIP_DST(a)=nat_oldsrcip
3783 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3784 	 *
3785 	 * Outbound ICMP
3786 	 * -------------
3787 	 * MAP rule, SRC=a,DST=b -> SRC=c,DST=b
3788 	 * - response to incoming packet (b,c)=>(b,a) (OIP_SRC=b,OIP_DST=a)
3789 	 * - OIP_SRC(b)=nat_olddstip,          OIP_DST(a)=nat_oldsrcip
3790 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3791 	 *
3792 	 * RDR rule, SRC=a,DST=b -> SRC=a,DST=c
3793 	 * - response to incoming packet (a,b)=>(a,c) (OIP_SRC=a,OIP_DST=c)
3794 	 * - OIP_SRC(a)=nat_newsrcip,          OIP_DST(c)=nat_newdstip
3795 	 *=> OIP_SRC(a)=nat_oldsrcip,          OIP_DST(c)=nat_olddstip
3796 	 *
3797 	 * REWRITE out rule, SRC=a,DST=b -> SRC=c,DST=d
3798 	 * - response to incoming packet (d,c)=>(b,a) (OIP_SRC=c,OIP_DST=d)
3799 	 * - OIP_SRC(c)=nat_olddstip,          OIP_DST(d)=nat_oldsrcip
3800 	 *=> OIP_SRC(b)=nat_newdstip,          OIP_DST(a)=nat_newsrcip
3801 	 *
3802 	 * REWRITE in rule, SRC=a,DST=b -> SRC=c,DST=d
3803 	 * - response to incoming packet (a,b)=>(c,d) (OIP_SRC=b,OIP_DST=a)
3804 	 * - OIP_SRC(b)=nat_newsrcip,          OIP_DST(a)=nat_newdstip
3805 	 *=> OIP_SRC(a)=nat_oldsrcip,          OIP_DST(c)=nat_olddstip
3806 	 */
3807 
3808 	if (((fin->fin_out == 0) && ((nat->nat_redir & NAT_MAP) != 0)) ||
3809 	    ((fin->fin_out == 1) && ((nat->nat_redir & NAT_REDIRECT) != 0))) {
3810 		a1.s_addr = ntohl(nat->nat_osrcaddr);
3811 		a4.s_addr = ntohl(oip->ip_src.s_addr);
3812 		a3.s_addr = ntohl(nat->nat_odstaddr);
3813 		a2.s_addr = ntohl(oip->ip_dst.s_addr);
3814 		oip->ip_src.s_addr = htonl(a1.s_addr);
3815 		oip->ip_dst.s_addr = htonl(a3.s_addr);
3816 		odst = 1;
3817 	} else {
3818 		a1.s_addr = ntohl(nat->nat_ndstaddr);
3819 		a2.s_addr = ntohl(oip->ip_dst.s_addr);
3820 		a3.s_addr = ntohl(nat->nat_nsrcaddr);
3821 		a4.s_addr = ntohl(oip->ip_src.s_addr);
3822 		oip->ip_dst.s_addr = htonl(a3.s_addr);
3823 		oip->ip_src.s_addr = htonl(a1.s_addr);
3824 		odst = 0;
3825 	}
3826 	sum1 = 0;
3827 	sum2 = 0;
3828 	sumd = 0;
3829 	CALC_SUMD(a2.s_addr, a3.s_addr, sum1);
3830 	CALC_SUMD(a4.s_addr, a1.s_addr, sum2);
3831 	sumd = sum2 + sum1;
3832 	if (sumd != 0)
3833 		ipf_fix_datacksum(&oip->ip_sum, sumd);
3834 
3835 	sumd2 = sumd;
3836 	sum1 = 0;
3837 	sum2 = 0;
3838 
3839 	/*
3840 	 * Fix UDP pseudo header checksum to compensate for the
3841 	 * IP address change.
3842 	 */
3843 	if (((flags & IPN_TCPUDP) != 0) && (dlen >= 4)) {
3844 		u_32_t sum3, sum4, sumt;
3845 
3846 		/*
3847 		 * Step 2 :
3848 		 * For offending TCP/UDP IP packets, translate the ports as
3849 		 * well, based on the NAT specification. Of course such
3850 		 * a change may be reflected in the ICMP checksum as well.
3851 		 *
3852 		 * Since the port fields are part of the TCP/UDP checksum
3853 		 * of the offending IP packet, you need to adjust that checksum
3854 		 * as well... except that the change in the port numbers should
3855 		 * be offset by the checksum change.  However, the TCP/UDP
3856 		 * checksum will also need to change if there has been an
3857 		 * IP address change.
3858 		 */
3859 		if (odst == 1) {
3860 			sum1 = ntohs(nat->nat_osport);
3861 			sum4 = ntohs(tcp->th_sport);
3862 			sum3 = ntohs(nat->nat_odport);
3863 			sum2 = ntohs(tcp->th_dport);
3864 
3865 			tcp->th_sport = htons(sum1);
3866 			tcp->th_dport = htons(sum3);
3867 		} else {
3868 			sum1 = ntohs(nat->nat_ndport);
3869 			sum2 = ntohs(tcp->th_dport);
3870 			sum3 = ntohs(nat->nat_nsport);
3871 			sum4 = ntohs(tcp->th_sport);
3872 
3873 			tcp->th_dport = htons(sum3);
3874 			tcp->th_sport = htons(sum1);
3875 		}
3876 		CALC_SUMD(sum4, sum1, sumt);
3877 		sumd += sumt;
3878 		CALC_SUMD(sum2, sum3, sumt);
3879 		sumd += sumt;
3880 
3881 		if (sumd != 0 || sumd2 != 0) {
3882 			/*
3883 			 * At this point, sumd is the delta to apply to the
3884 			 * TCP/UDP header, given the changes in both the IP
3885 			 * address and the ports and sumd2 is the delta to
3886 			 * apply to the ICMP header, given the IP address
3887 			 * change delta that may need to be applied to the
3888 			 * TCP/UDP checksum instead.
3889 			 *
3890 			 * If we will both the IP and TCP/UDP checksums
3891 			 * then the ICMP checksum changes by the address
3892 			 * delta applied to the TCP/UDP checksum.  If we
3893 			 * do not change the TCP/UDP checksum them we
3894 			 * apply the delta in ports to the ICMP checksum.
3895 			 */
3896 			if (oip->ip_p == IPPROTO_UDP) {
3897 				if ((dlen >= 8) && (*csump != 0)) {
3898 					ipf_fix_datacksum(csump, sumd);
3899 				} else {
3900 					CALC_SUMD(sum1, sum4, sumd2);
3901 					CALC_SUMD(sum3, sum2, sumt);
3902 					sumd2 += sumt;
3903 				}
3904 			} else if (oip->ip_p == IPPROTO_TCP) {
3905 				if (dlen >= 18) {
3906 					ipf_fix_datacksum(csump, sumd);
3907 				} else {
3908 					CALC_SUMD(sum1, sum4, sumd2);
3909 					CALC_SUMD(sum3, sum2, sumt);
3910 					sumd2 += sumt;
3911 				}
3912 			}
3913 			if (sumd2 != 0) {
3914 				sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3915 				sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3916 				sumd2 = (sumd2 & 0xffff) + (sumd2 >> 16);
3917 				ipf_fix_incksum(0, &icmp->icmp_cksum, sumd2, 0);
3918 			}
3919 		}
3920 	} else if (((flags & IPN_ICMPQUERY) != 0) && (dlen >= 8)) {
3921 		icmphdr_t *orgicmp;
3922 
3923 		/*
3924 		 * XXX - what if this is bogus hl and we go off the end ?
3925 		 * In this case, ipf_nat_icmperrorlookup() will have
3926 		 * returned NULL.
3927 		 */
3928 		orgicmp = (icmphdr_t *)dp;
3929 
3930 		if (odst == 1) {
3931 			if (orgicmp->icmp_id != nat->nat_osport) {
3932 
3933 				/*
3934 				 * Fix ICMP checksum (of the offening ICMP
3935 				 * query packet) to compensate the change
3936 				 * in the ICMP id of the offending ICMP
3937 				 * packet.
3938 				 *
3939 				 * Since you modify orgicmp->icmp_id with
3940 				 * a delta (say x) and you compensate that
3941 				 * in origicmp->icmp_cksum with a delta
3942 				 * minus x, you don't have to adjust the
3943 				 * overall icmp->icmp_cksum
3944 				 */
3945 				sum1 = ntohs(orgicmp->icmp_id);
3946 				sum2 = ntohs(nat->nat_oicmpid);
3947 				CALC_SUMD(sum1, sum2, sumd);
3948 				orgicmp->icmp_id = nat->nat_oicmpid;
3949 				ipf_fix_datacksum(&orgicmp->icmp_cksum, sumd);
3950 			}
3951 		} /* nat_dir == NAT_INBOUND is impossible for icmp queries */
3952 	}
3953 	return (nat);
3954 }
3955 
3956 
3957 /*
3958  *       MAP-IN    MAP-OUT   RDR-IN   RDR-OUT
3959  * osrc    X       == src    == src      X
3960  * odst    X       == dst    == dst      X
3961  * nsrc  == dst      X         X      == dst
3962  * ndst  == src      X         X      == src
3963  * MAP = NAT_OUTBOUND, RDR = NAT_INBOUND
3964  */
3965 /*
3966  * NB: these lookups don't lock access to the list, it assumed that it has
3967  * already been done!
3968  */
3969 /* ------------------------------------------------------------------------ */
3970 /* Function:    ipf_nat_inlookup                                            */
3971 /* Returns:     nat_t* - NULL == no match,                                  */
3972 /*                       else pointer to matching NAT entry                 */
3973 /* Parameters:  fin(I)    - pointer to packet information                   */
3974 /*              flags(I)  - NAT flags for this packet                       */
3975 /*              p(I)      - protocol for this packet                        */
3976 /*              src(I)    - source IP address                               */
3977 /*              mapdst(I) - destination IP address                          */
3978 /*                                                                          */
3979 /* Lookup a nat entry based on the mapped destination ip address/port and   */
3980 /* real source address/port.  We use this lookup when receiving a packet,   */
3981 /* we're looking for a table entry, based on the destination address.       */
3982 /*                                                                          */
3983 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY.         */
3984 /*                                                                          */
3985 /* NOTE: IT IS ASSUMED THAT  IS ONLY HELD WITH A READ LOCK WHEN             */
3986 /*       THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags.             */
3987 /*                                                                          */
3988 /* flags   -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if   */
3989 /*            the packet is of said protocol                                */
3990 /* ------------------------------------------------------------------------ */
3991 nat_t *
ipf_nat_inlookup(fr_info_t * fin,u_int flags,u_int p,struct in_addr src,struct in_addr mapdst)3992 ipf_nat_inlookup(fr_info_t *fin, u_int flags, u_int p,
3993 	struct in_addr src , struct in_addr mapdst)
3994 {
3995 	ipf_main_softc_t *softc = fin->fin_main_soft;
3996 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
3997 	u_short sport, dport;
3998 	ipnat_t *ipn;
3999 	nat_t *nat;
4000 	int nflags;
4001 	u_32_t dst;
4002 	void *ifp;
4003 	u_int hv, rhv;
4004 
4005 	ifp = fin->fin_ifp;
4006 	dst = mapdst.s_addr;
4007 
4008 	switch (p)
4009 	{
4010 	case IPPROTO_TCP :
4011 	case IPPROTO_UDP :
4012 		sport = htons(fin->fin_data[0]);
4013 		dport = htons(fin->fin_data[1]);
4014 		break;
4015 	case IPPROTO_ICMP :
4016 		sport = 0;
4017 		dport = fin->fin_data[1];
4018 		break;
4019 	default :
4020 		sport = 0;
4021 		dport = 0;
4022 		break;
4023 	}
4024 
4025 
4026 	if ((flags & SI_WILDP) != 0)
4027 		goto find_in_wild_ports;
4028 
4029 	rhv = NAT_HASH_FN(dst, dport, 0xffffffff);
4030 	rhv = NAT_HASH_FN(src.s_addr, rhv + sport, 0xffffffff);
4031 	hv = rhv % softn->ipf_nat_table_sz;
4032 	nat = softn->ipf_nat_table[1][hv];
4033 	/* TRACE dst, dport, src, sport, hv, nat */
4034 
4035 	for (; nat; nat = nat->nat_hnext[1]) {
4036 		if (nat->nat_ifps[0] != NULL) {
4037 			if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
4038 				continue;
4039 		}
4040 
4041 		if (nat->nat_pr[0] != p)
4042 			continue;
4043 
4044 		switch (nat->nat_dir)
4045 		{
4046 		case NAT_INBOUND :
4047 		case NAT_DIVERTIN :
4048 			if (nat->nat_v[0] != 4)
4049 				continue;
4050 			if (nat->nat_osrcaddr != src.s_addr ||
4051 			    nat->nat_odstaddr != dst)
4052 				continue;
4053 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4054 				if (nat->nat_osport != sport)
4055 					continue;
4056 				if (nat->nat_odport != dport)
4057 					continue;
4058 
4059 			} else if (p == IPPROTO_ICMP) {
4060 				if (nat->nat_osport != dport) {
4061 					continue;
4062 				}
4063 			}
4064 			break;
4065 		case NAT_DIVERTOUT :
4066 			if (nat->nat_dlocal)
4067 				continue;
4068 		case NAT_OUTBOUND :
4069 			if (nat->nat_v[1] != 4)
4070 				continue;
4071 			if (nat->nat_dlocal)
4072 				continue;
4073 			if (nat->nat_dlocal)
4074 				continue;
4075 			if (nat->nat_ndstaddr != src.s_addr ||
4076 			    nat->nat_nsrcaddr != dst)
4077 				continue;
4078 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4079 				if (nat->nat_ndport != sport)
4080 					continue;
4081 				if (nat->nat_nsport != dport)
4082 					continue;
4083 
4084 			} else if (p == IPPROTO_ICMP) {
4085 				if (nat->nat_osport != dport) {
4086 					continue;
4087 				}
4088 			}
4089 			break;
4090 		}
4091 
4092 
4093 		if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4094 			ipn = nat->nat_ptr;
4095 			if ((ipn != NULL) && (nat->nat_aps != NULL))
4096 				if (ipf_proxy_match(fin, nat) != 0)
4097 					continue;
4098 		}
4099 		if ((nat->nat_ifps[0] == NULL) && (ifp != NULL)) {
4100 			nat->nat_ifps[0] = ifp;
4101 			nat->nat_mtu[0] = GETIFMTU_4(ifp);
4102 		}
4103 		return (nat);
4104 	}
4105 
4106 	/*
4107 	 * So if we didn't find it but there are wildcard members in the hash
4108 	 * table, go back and look for them.  We do this search and update here
4109 	 * because it is modifying the NAT table and we want to do this only
4110 	 * for the first packet that matches.  The exception, of course, is
4111 	 * for "dummy" (FI_IGNORE) lookups.
4112 	 */
4113 find_in_wild_ports:
4114 	if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH)) {
4115 		NBUMPSIDEX(0, ns_lookup_miss, ns_lookup_miss_0);
4116 		return (NULL);
4117 	}
4118 	if (softn->ipf_nat_stats.ns_wilds == 0 || (fin->fin_flx & FI_NOWILD)) {
4119 		NBUMPSIDEX(0, ns_lookup_nowild, ns_lookup_nowild_0);
4120 		return (NULL);
4121 	}
4122 
4123 	RWLOCK_EXIT(&softc->ipf_nat);
4124 
4125 	hv = NAT_HASH_FN(dst, 0, 0xffffffff);
4126 	hv = NAT_HASH_FN(src.s_addr, hv, softn->ipf_nat_table_sz);
4127 	WRITE_ENTER(&softc->ipf_nat);
4128 
4129 	nat = softn->ipf_nat_table[1][hv];
4130 	/* TRACE dst, src, hv, nat */
4131 	for (; nat; nat = nat->nat_hnext[1]) {
4132 		if (nat->nat_ifps[0] != NULL) {
4133 			if ((ifp != NULL) && (ifp != nat->nat_ifps[0]))
4134 				continue;
4135 		}
4136 
4137 		if (nat->nat_pr[0] != fin->fin_p)
4138 			continue;
4139 
4140 		switch (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND))
4141 		{
4142 		case NAT_INBOUND :
4143 			if (nat->nat_v[0] != 4)
4144 				continue;
4145 			if (nat->nat_osrcaddr != src.s_addr ||
4146 			    nat->nat_odstaddr != dst)
4147 				continue;
4148 			break;
4149 		case NAT_OUTBOUND :
4150 			if (nat->nat_v[1] != 4)
4151 				continue;
4152 			if (nat->nat_ndstaddr != src.s_addr ||
4153 			    nat->nat_nsrcaddr != dst)
4154 				continue;
4155 			break;
4156 		}
4157 
4158 		nflags = nat->nat_flags;
4159 		if (!(nflags & (NAT_TCPUDP|SI_WILDP)))
4160 			continue;
4161 
4162 		if (ipf_nat_wildok(nat, (int)sport, (int)dport, nflags,
4163 				   NAT_INBOUND) == 1) {
4164 			if ((fin->fin_flx & FI_IGNORE) != 0)
4165 				break;
4166 			if ((nflags & SI_CLONE) != 0) {
4167 				nat = ipf_nat_clone(fin, nat);
4168 				if (nat == NULL)
4169 					break;
4170 			} else {
4171 				MUTEX_ENTER(&softn->ipf_nat_new);
4172 				softn->ipf_nat_stats.ns_wilds--;
4173 				MUTEX_EXIT(&softn->ipf_nat_new);
4174 			}
4175 
4176 			if (nat->nat_dir == NAT_INBOUND) {
4177 				if (nat->nat_osport == 0) {
4178 					nat->nat_osport = sport;
4179 					nat->nat_nsport = sport;
4180 				}
4181 				if (nat->nat_odport == 0) {
4182 					nat->nat_odport = dport;
4183 					nat->nat_ndport = dport;
4184 				}
4185 			} else if (nat->nat_dir == NAT_OUTBOUND) {
4186 				if (nat->nat_osport == 0) {
4187 					nat->nat_osport = dport;
4188 					nat->nat_nsport = dport;
4189 				}
4190 				if (nat->nat_odport == 0) {
4191 					nat->nat_odport = sport;
4192 					nat->nat_ndport = sport;
4193 				}
4194 			}
4195 			if ((nat->nat_ifps[0] == NULL) && (ifp != NULL)) {
4196 				nat->nat_ifps[0] = ifp;
4197 				nat->nat_mtu[0] = GETIFMTU_4(ifp);
4198 			}
4199 			nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
4200 			ipf_nat_tabmove(softn, nat);
4201 			break;
4202 		}
4203 	}
4204 
4205 	MUTEX_DOWNGRADE(&softc->ipf_nat);
4206 
4207 	if (nat == NULL) {
4208 		NBUMPSIDE(0, ns_lookup_miss);
4209 	}
4210 	return (nat);
4211 }
4212 
4213 
4214 /* ------------------------------------------------------------------------ */
4215 /* Function:    ipf_nat_tabmove                                             */
4216 /* Returns:     Nil                                                         */
4217 /* Parameters:  softn(I) - pointer to NAT context structure                 */
4218 /*              nat(I)   - pointer to NAT structure                         */
4219 /* Write Lock:  ipf_nat                                                     */
4220 /*                                                                          */
4221 /* This function is only called for TCP/UDP NAT table entries where the     */
4222 /* original was placed in the table without hashing on the ports and we now */
4223 /* want to include hashing on port numbers.                                 */
4224 /* ------------------------------------------------------------------------ */
4225 static void
ipf_nat_tabmove(ipf_nat_softc_t * softn,nat_t * nat)4226 ipf_nat_tabmove(ipf_nat_softc_t *softn, nat_t *nat)
4227 {
4228 	u_int hv0, hv1, rhv0, rhv1;
4229 	natstat_t *nsp;
4230 	nat_t **natp;
4231 
4232 	if (nat->nat_flags & SI_CLONE)
4233 		return;
4234 
4235 	nsp = &softn->ipf_nat_stats;
4236 	/*
4237 	 * Remove the NAT entry from the old location
4238 	 */
4239 	if (nat->nat_hnext[0])
4240 		nat->nat_hnext[0]->nat_phnext[0] = nat->nat_phnext[0];
4241 	*nat->nat_phnext[0] = nat->nat_hnext[0];
4242 	nsp->ns_side[0].ns_bucketlen[nat->nat_hv[0] %
4243 				     softn->ipf_nat_table_sz]--;
4244 
4245 	if (nat->nat_hnext[1])
4246 		nat->nat_hnext[1]->nat_phnext[1] = nat->nat_phnext[1];
4247 	*nat->nat_phnext[1] = nat->nat_hnext[1];
4248 	nsp->ns_side[1].ns_bucketlen[nat->nat_hv[1] %
4249 				     softn->ipf_nat_table_sz]--;
4250 
4251 	/*
4252 	 * Add into the NAT table in the new position
4253 	 */
4254 	rhv0 = NAT_HASH_FN(nat->nat_osrcaddr, nat->nat_osport, 0xffffffff);
4255 	rhv0 = NAT_HASH_FN(nat->nat_odstaddr, rhv0 + nat->nat_odport,
4256 			   0xffffffff);
4257 	rhv1 = NAT_HASH_FN(nat->nat_nsrcaddr, nat->nat_nsport, 0xffffffff);
4258 	rhv1 = NAT_HASH_FN(nat->nat_ndstaddr, rhv1 + nat->nat_ndport,
4259 			   0xffffffff);
4260 
4261 	hv0 = rhv0 % softn->ipf_nat_table_sz;
4262 	hv1 = rhv1 % softn->ipf_nat_table_sz;
4263 
4264 	if (nat->nat_dir == NAT_INBOUND || nat->nat_dir == NAT_DIVERTIN) {
4265 		u_int swap;
4266 
4267 		swap = hv0;
4268 		hv0 = hv1;
4269 		hv1 = swap;
4270 	}
4271 
4272 	/* TRACE nat_osrcaddr, nat_osport, nat_odstaddr, nat_odport, hv0 */
4273 	/* TRACE nat_nsrcaddr, nat_nsport, nat_ndstaddr, nat_ndport, hv1 */
4274 
4275 	nat->nat_hv[0] = rhv0;
4276 	natp = &softn->ipf_nat_table[0][hv0];
4277 	if (*natp)
4278 		(*natp)->nat_phnext[0] = &nat->nat_hnext[0];
4279 	nat->nat_phnext[0] = natp;
4280 	nat->nat_hnext[0] = *natp;
4281 	*natp = nat;
4282 	nsp->ns_side[0].ns_bucketlen[hv0]++;
4283 
4284 	nat->nat_hv[1] = rhv1;
4285 	natp = &softn->ipf_nat_table[1][hv1];
4286 	if (*natp)
4287 		(*natp)->nat_phnext[1] = &nat->nat_hnext[1];
4288 	nat->nat_phnext[1] = natp;
4289 	nat->nat_hnext[1] = *natp;
4290 	*natp = nat;
4291 	nsp->ns_side[1].ns_bucketlen[hv1]++;
4292 }
4293 
4294 
4295 /* ------------------------------------------------------------------------ */
4296 /* Function:    ipf_nat_outlookup                                           */
4297 /* Returns:     nat_t* - NULL == no match,                                  */
4298 /*                       else pointer to matching NAT entry                 */
4299 /* Parameters:  fin(I)   - pointer to packet information                    */
4300 /*              flags(I) - NAT flags for this packet                        */
4301 /*              p(I)     - protocol for this packet                         */
4302 /*              src(I)   - source IP address                                */
4303 /*              dst(I)   - destination IP address                           */
4304 /*              rw(I)    - 1 == write lock on  held, 0 == read lock.        */
4305 /*                                                                          */
4306 /* Lookup a nat entry based on the source 'real' ip address/port and        */
4307 /* destination address/port.  We use this lookup when sending a packet out, */
4308 /* we're looking for a table entry, based on the source address.            */
4309 /*                                                                          */
4310 /* NOTE: THE PACKET BEING CHECKED (IF FOUND) HAS A MAPPING ALREADY.         */
4311 /*                                                                          */
4312 /* NOTE: IT IS ASSUMED THAT  IS ONLY HELD WITH A READ LOCK WHEN             */
4313 /*       THIS FUNCTION IS CALLED WITH NAT_SEARCH SET IN nflags.             */
4314 /*                                                                          */
4315 /* flags   -> relevant are IPN_UDP/IPN_TCP/IPN_ICMPQUERY that indicate if   */
4316 /*            the packet is of said protocol                                */
4317 /* ------------------------------------------------------------------------ */
4318 nat_t *
ipf_nat_outlookup(fr_info_t * fin,u_int flags,u_int p,struct in_addr src,struct in_addr dst)4319 ipf_nat_outlookup(fr_info_t *fin, u_int flags, u_int p,
4320 	struct in_addr src , struct in_addr dst)
4321 {
4322 	ipf_main_softc_t *softc = fin->fin_main_soft;
4323 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4324 	u_short sport, dport;
4325 	ipnat_t *ipn;
4326 	nat_t *nat;
4327 	void *ifp;
4328 	u_int hv;
4329 
4330 	ifp = fin->fin_ifp;
4331 
4332 	switch (p)
4333 	{
4334 	case IPPROTO_TCP :
4335 	case IPPROTO_UDP :
4336 		sport = htons(fin->fin_data[0]);
4337 		dport = htons(fin->fin_data[1]);
4338 		break;
4339 	case IPPROTO_ICMP :
4340 		sport = 0;
4341 		dport = fin->fin_data[1];
4342 		break;
4343 	default :
4344 		sport = 0;
4345 		dport = 0;
4346 		break;
4347 	}
4348 
4349 	if ((flags & SI_WILDP) != 0)
4350 		goto find_out_wild_ports;
4351 
4352 	hv = NAT_HASH_FN(src.s_addr, sport, 0xffffffff);
4353 	hv = NAT_HASH_FN(dst.s_addr, hv + dport, softn->ipf_nat_table_sz);
4354 	nat = softn->ipf_nat_table[0][hv];
4355 
4356 	/* TRACE src, sport, dst, dport, hv, nat */
4357 
4358 	for (; nat; nat = nat->nat_hnext[0]) {
4359 		if (nat->nat_ifps[1] != NULL) {
4360 			if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
4361 				continue;
4362 		}
4363 
4364 		if (nat->nat_pr[1] != p)
4365 			continue;
4366 
4367 		switch (nat->nat_dir)
4368 		{
4369 		case NAT_INBOUND :
4370 		case NAT_DIVERTIN :
4371 			if (nat->nat_v[1] != 4)
4372 				continue;
4373 			if (nat->nat_ndstaddr != src.s_addr ||
4374 			    nat->nat_nsrcaddr != dst.s_addr)
4375 				continue;
4376 
4377 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4378 				if (nat->nat_ndport != sport)
4379 					continue;
4380 				if (nat->nat_nsport != dport)
4381 					continue;
4382 
4383 			} else if (p == IPPROTO_ICMP) {
4384 				if (nat->nat_osport != dport) {
4385 					continue;
4386 				}
4387 			}
4388 			break;
4389 		case NAT_OUTBOUND :
4390 		case NAT_DIVERTOUT :
4391 			if (nat->nat_v[0] != 4)
4392 				continue;
4393 			if (nat->nat_osrcaddr != src.s_addr ||
4394 			    nat->nat_odstaddr != dst.s_addr)
4395 				continue;
4396 
4397 			if ((nat->nat_flags & IPN_TCPUDP) != 0) {
4398 				if (nat->nat_odport != dport)
4399 					continue;
4400 				if (nat->nat_osport != sport)
4401 					continue;
4402 
4403 			} else if (p == IPPROTO_ICMP) {
4404 				if (nat->nat_osport != dport) {
4405 					continue;
4406 				}
4407 			}
4408 			break;
4409 		}
4410 
4411 		ipn = nat->nat_ptr;
4412 		if ((ipn != NULL) && (nat->nat_aps != NULL))
4413 			if (ipf_proxy_match(fin, nat) != 0)
4414 				continue;
4415 
4416 		if ((nat->nat_ifps[1] == NULL) && (ifp != NULL)) {
4417 			nat->nat_ifps[1] = ifp;
4418 			nat->nat_mtu[1] = GETIFMTU_4(ifp);
4419 		}
4420 		return (nat);
4421 	}
4422 
4423 	/*
4424 	 * So if we didn't find it but there are wildcard members in the hash
4425 	 * table, go back and look for them.  We do this search and update here
4426 	 * because it is modifying the NAT table and we want to do this only
4427 	 * for the first packet that matches.  The exception, of course, is
4428 	 * for "dummy" (FI_IGNORE) lookups.
4429 	 */
4430 find_out_wild_ports:
4431 	if (!(flags & NAT_TCPUDP) || !(flags & NAT_SEARCH)) {
4432 		NBUMPSIDEX(1, ns_lookup_miss, ns_lookup_miss_1);
4433 		return (NULL);
4434 	}
4435 	if (softn->ipf_nat_stats.ns_wilds == 0 || (fin->fin_flx & FI_NOWILD)) {
4436 		NBUMPSIDEX(1, ns_lookup_nowild, ns_lookup_nowild_1);
4437 		return (NULL);
4438 	}
4439 
4440 	RWLOCK_EXIT(&softc->ipf_nat);
4441 
4442 	hv = NAT_HASH_FN(src.s_addr, 0, 0xffffffff);
4443 	hv = NAT_HASH_FN(dst.s_addr, hv, softn->ipf_nat_table_sz);
4444 
4445 	WRITE_ENTER(&softc->ipf_nat);
4446 
4447 	nat = softn->ipf_nat_table[0][hv];
4448 	for (; nat; nat = nat->nat_hnext[0]) {
4449 		if (nat->nat_ifps[1] != NULL) {
4450 			if ((ifp != NULL) && (ifp != nat->nat_ifps[1]))
4451 				continue;
4452 		}
4453 
4454 		if (nat->nat_pr[1] != fin->fin_p)
4455 			continue;
4456 
4457 		switch (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND))
4458 		{
4459 		case NAT_INBOUND :
4460 			if (nat->nat_v[1] != 4)
4461 				continue;
4462 			if (nat->nat_ndstaddr != src.s_addr ||
4463 			    nat->nat_nsrcaddr != dst.s_addr)
4464 				continue;
4465 			break;
4466 		case NAT_OUTBOUND :
4467 			if (nat->nat_v[0] != 4)
4468 				continue;
4469 			if (nat->nat_osrcaddr != src.s_addr ||
4470 			    nat->nat_odstaddr != dst.s_addr)
4471 				continue;
4472 			break;
4473 		}
4474 
4475 		if (!(nat->nat_flags & (NAT_TCPUDP|SI_WILDP)))
4476 			continue;
4477 
4478 		if (ipf_nat_wildok(nat, (int)sport, (int)dport, nat->nat_flags,
4479 				   NAT_OUTBOUND) == 1) {
4480 			if ((fin->fin_flx & FI_IGNORE) != 0)
4481 				break;
4482 			if ((nat->nat_flags & SI_CLONE) != 0) {
4483 				nat = ipf_nat_clone(fin, nat);
4484 				if (nat == NULL)
4485 					break;
4486 			} else {
4487 				MUTEX_ENTER(&softn->ipf_nat_new);
4488 				softn->ipf_nat_stats.ns_wilds--;
4489 				MUTEX_EXIT(&softn->ipf_nat_new);
4490 			}
4491 
4492 			if (nat->nat_dir == NAT_OUTBOUND) {
4493 				if (nat->nat_osport == 0) {
4494 					nat->nat_osport = sport;
4495 					nat->nat_nsport = sport;
4496 				}
4497 				if (nat->nat_odport == 0) {
4498 					nat->nat_odport = dport;
4499 					nat->nat_ndport = dport;
4500 				}
4501 			} else if (nat->nat_dir == NAT_INBOUND) {
4502 				if (nat->nat_osport == 0) {
4503 					nat->nat_osport = dport;
4504 					nat->nat_nsport = dport;
4505 				}
4506 				if (nat->nat_odport == 0) {
4507 					nat->nat_odport = sport;
4508 					nat->nat_ndport = sport;
4509 				}
4510 			}
4511 			if ((nat->nat_ifps[1] == NULL) && (ifp != NULL)) {
4512 				nat->nat_ifps[1] = ifp;
4513 				nat->nat_mtu[1] = GETIFMTU_4(ifp);
4514 			}
4515 			nat->nat_flags &= ~(SI_W_DPORT|SI_W_SPORT);
4516 			ipf_nat_tabmove(softn, nat);
4517 			break;
4518 		}
4519 	}
4520 
4521 	MUTEX_DOWNGRADE(&softc->ipf_nat);
4522 
4523 	if (nat == NULL) {
4524 		NBUMPSIDE(1, ns_lookup_miss);
4525 	}
4526 	return (nat);
4527 }
4528 
4529 
4530 /* ------------------------------------------------------------------------ */
4531 /* Function:    ipf_nat_lookupredir                                         */
4532 /* Returns:     nat_t* - NULL == no match,                                  */
4533 /*                       else pointer to matching NAT entry                 */
4534 /* Parameters:  np(I) - pointer to description of packet to find NAT table  */
4535 /*                      entry for.                                          */
4536 /*                                                                          */
4537 /* Lookup the NAT tables to search for a matching redirect                  */
4538 /* The contents of natlookup_t should imitate those found in a packet that  */
4539 /* would be translated - ie a packet coming in for RDR or going out for MAP.*/
4540 /* We can do the lookup in one of two ways, imitating an inbound or         */
4541 /* outbound  packet.  By default we assume outbound, unless IPN_IN is set.  */
4542 /* For IN, the fields are set as follows:                                   */
4543 /*     nl_real* = source information                                        */
4544 /*     nl_out* = destination information (translated)                       */
4545 /* For an out packet, the fields are set like this:                         */
4546 /*     nl_in* = source information (untranslated)                           */
4547 /*     nl_out* = destination information (translated)                       */
4548 /* ------------------------------------------------------------------------ */
4549 nat_t *
ipf_nat_lookupredir(natlookup_t * np)4550 ipf_nat_lookupredir(natlookup_t *np)
4551 {
4552 	fr_info_t fi;
4553 	nat_t *nat;
4554 
4555 	bzero((char *)&fi, sizeof(fi));
4556 	if (np->nl_flags & IPN_IN) {
4557 		fi.fin_data[0] = ntohs(np->nl_realport);
4558 		fi.fin_data[1] = ntohs(np->nl_outport);
4559 	} else {
4560 		fi.fin_data[0] = ntohs(np->nl_inport);
4561 		fi.fin_data[1] = ntohs(np->nl_outport);
4562 	}
4563 	if (np->nl_flags & IPN_TCP)
4564 		fi.fin_p = IPPROTO_TCP;
4565 	else if (np->nl_flags & IPN_UDP)
4566 		fi.fin_p = IPPROTO_UDP;
4567 	else if (np->nl_flags & (IPN_ICMPERR|IPN_ICMPQUERY))
4568 		fi.fin_p = IPPROTO_ICMP;
4569 
4570 	/*
4571 	 * We can do two sorts of lookups:
4572 	 * - IPN_IN: we have the `real' and `out' address, look for `in'.
4573 	 * - default: we have the `in' and `out' address, look for `real'.
4574 	 */
4575 	if (np->nl_flags & IPN_IN) {
4576 		if ((nat = ipf_nat_inlookup(&fi, np->nl_flags, fi.fin_p,
4577 					    np->nl_realip, np->nl_outip))) {
4578 			np->nl_inip = nat->nat_odstip;
4579 			np->nl_inport = nat->nat_odport;
4580 		}
4581 	} else {
4582 		/*
4583 		 * If nl_inip is non null, this is a lookup based on the real
4584 		 * ip address. Else, we use the fake.
4585 		 */
4586 		if ((nat = ipf_nat_outlookup(&fi, np->nl_flags, fi.fin_p,
4587 					 np->nl_inip, np->nl_outip))) {
4588 
4589 			if ((np->nl_flags & IPN_FINDFORWARD) != 0) {
4590 				fr_info_t fin;
4591 				bzero((char *)&fin, sizeof(fin));
4592 				fin.fin_p = nat->nat_pr[0];
4593 				fin.fin_data[0] = ntohs(nat->nat_ndport);
4594 				fin.fin_data[1] = ntohs(nat->nat_nsport);
4595 				if (ipf_nat_inlookup(&fin, np->nl_flags,
4596 						     fin.fin_p, nat->nat_ndstip,
4597 						     nat->nat_nsrcip) != NULL) {
4598 					np->nl_flags &= ~IPN_FINDFORWARD;
4599 				}
4600 			}
4601 
4602 			np->nl_realip = nat->nat_odstip;
4603 			np->nl_realport = nat->nat_odport;
4604 		}
4605  	}
4606 
4607 	return (nat);
4608 }
4609 
4610 
4611 /* ------------------------------------------------------------------------ */
4612 /* Function:    ipf_nat_match                                               */
4613 /* Returns:     int - 0 == no match, 1 == match                             */
4614 /* Parameters:  fin(I)   - pointer to packet information                    */
4615 /*              np(I)    - pointer to NAT rule                              */
4616 /*                                                                          */
4617 /* Pull the matching of a packet against a NAT rule out of that complex     */
4618 /* loop inside ipf_nat_checkin() and lay it out properly in its own function. */
4619 /* ------------------------------------------------------------------------ */
4620 static int
ipf_nat_match(fr_info_t * fin,ipnat_t * np)4621 ipf_nat_match(fr_info_t *fin, ipnat_t *np)
4622 {
4623 	ipf_main_softc_t *softc = fin->fin_main_soft;
4624 	frtuc_t *ft;
4625 	int match;
4626 
4627 	match = 0;
4628 	switch (np->in_osrcatype)
4629 	{
4630 	case FRI_NORMAL :
4631 		match = ((fin->fin_saddr & np->in_osrcmsk) != np->in_osrcaddr);
4632 		break;
4633 	case FRI_LOOKUP :
4634 		match = (*np->in_osrcfunc)(softc, np->in_osrcptr,
4635 					   4, &fin->fin_saddr, fin->fin_plen);
4636 		break;
4637 	}
4638 	match ^= ((np->in_flags & IPN_NOTSRC) != 0);
4639 	if (match)
4640 		return (0);
4641 
4642 	match = 0;
4643 	switch (np->in_odstatype)
4644 	{
4645 	case FRI_NORMAL :
4646 		match = ((fin->fin_daddr & np->in_odstmsk) != np->in_odstaddr);
4647 		break;
4648 	case FRI_LOOKUP :
4649 		match = (*np->in_odstfunc)(softc, np->in_odstptr,
4650 					   4, &fin->fin_daddr, fin->fin_plen);
4651 		break;
4652 	}
4653 
4654 	match ^= ((np->in_flags & IPN_NOTDST) != 0);
4655 	if (match)
4656 		return (0);
4657 
4658 	ft = &np->in_tuc;
4659 	if (!(fin->fin_flx & FI_TCPUDP) ||
4660 	    (fin->fin_flx & (FI_SHORT|FI_FRAGBODY))) {
4661 		if (ft->ftu_scmp || ft->ftu_dcmp)
4662 			return (0);
4663 		return (1);
4664 	}
4665 
4666 	return (ipf_tcpudpchk(&fin->fin_fi, ft));
4667 }
4668 
4669 
4670 /* ------------------------------------------------------------------------ */
4671 /* Function:    ipf_nat_update                                              */
4672 /* Returns:     Nil                                                         */
4673 /* Parameters:  fin(I) - pointer to packet information                      */
4674 /*              nat(I) - pointer to NAT structure                           */
4675 /*                                                                          */
4676 /* Updates the lifetime of a NAT table entry for non-TCP packets.  Must be  */
4677 /* called with fin_rev updated - i.e. after calling ipf_nat_proto().        */
4678 /*                                                                          */
4679 /* This *MUST* be called after ipf_nat_proto() as it expects fin_rev to     */
4680 /* already be set.                                                          */
4681 /* ------------------------------------------------------------------------ */
4682 void
ipf_nat_update(fr_info_t * fin,nat_t * nat)4683 ipf_nat_update(fr_info_t *fin, nat_t *nat)
4684 {
4685 	ipf_main_softc_t *softc = fin->fin_main_soft;
4686 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4687 	ipftq_t *ifq, *ifq2;
4688 	ipftqent_t *tqe;
4689 	ipnat_t *np = nat->nat_ptr;
4690 
4691 	tqe = &nat->nat_tqe;
4692 	ifq = tqe->tqe_ifq;
4693 
4694 	/*
4695 	 * We allow over-riding of NAT timeouts from NAT rules, even for
4696 	 * TCP, however, if it is TCP and there is no rule timeout set,
4697 	 * then do not update the timeout here.
4698 	 */
4699 	if (np != NULL) {
4700 		np->in_bytes[fin->fin_rev] += fin->fin_plen;
4701 		ifq2 = np->in_tqehead[fin->fin_rev];
4702 	} else {
4703 		ifq2 = NULL;
4704 	}
4705 
4706 	if (nat->nat_pr[0] == IPPROTO_TCP && ifq2 == NULL) {
4707 		(void) ipf_tcp_age(&nat->nat_tqe, fin, softn->ipf_nat_tcptq,
4708 				   0, 2);
4709 	} else {
4710 		if (ifq2 == NULL) {
4711 			if (nat->nat_pr[0] == IPPROTO_UDP)
4712 				ifq2 = fin->fin_rev ? &softn->ipf_nat_udpacktq :
4713 						      &softn->ipf_nat_udptq;
4714 			else if (nat->nat_pr[0] == IPPROTO_ICMP ||
4715 				 nat->nat_pr[0] == IPPROTO_ICMPV6)
4716 				ifq2 = fin->fin_rev ? &softn->ipf_nat_icmpacktq:
4717 						      &softn->ipf_nat_icmptq;
4718 			else
4719 				ifq2 = &softn->ipf_nat_iptq;
4720 		}
4721 
4722 		ipf_movequeue(softc->ipf_ticks, tqe, ifq, ifq2);
4723 	}
4724 }
4725 
4726 
4727 /* ------------------------------------------------------------------------ */
4728 /* Function:    ipf_nat_checkout                                            */
4729 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
4730 /*                     0 == no packet translation occurred,                 */
4731 /*                     1 == packet was successfully translated.             */
4732 /* Parameters:  fin(I)   - pointer to packet information                    */
4733 /*              passp(I) - pointer to filtering result flags                */
4734 /*                                                                          */
4735 /* Check to see if an outcoming packet should be changed.  ICMP packets are */
4736 /* first checked to see if they match an existing entry (if an error),      */
4737 /* otherwise a search of the current NAT table is made.  If neither results */
4738 /* in a match then a search for a matching NAT rule is made.  Create a new  */
4739 /* NAT entry if a we matched a NAT rule.  Lastly, actually change the       */
4740 /* packet header(s) as required.                                            */
4741 /* ------------------------------------------------------------------------ */
4742 int
ipf_nat_checkout(fr_info_t * fin,u_32_t * passp)4743 ipf_nat_checkout(fr_info_t *fin, u_32_t *passp)
4744 {
4745 	ipnat_t *np = NULL, *npnext;
4746 	struct ifnet *ifp, *sifp;
4747 	ipf_main_softc_t *softc;
4748 	ipf_nat_softc_t *softn;
4749 	tcphdr_t *tcp = NULL;
4750 	int rval, natfailed;
4751 	u_int nflags = 0;
4752 	u_32_t ipa, iph;
4753 	int natadd = 1;
4754 	frentry_t *fr;
4755 	nat_t *nat;
4756 
4757 	if (fin->fin_v == 6) {
4758 #ifdef USE_INET6
4759 		return (ipf_nat6_checkout(fin, passp));
4760 #else
4761 		return (0);
4762 #endif
4763 	}
4764 
4765 	softc = fin->fin_main_soft;
4766 	softn = softc->ipf_nat_soft;
4767 
4768 	if (softn->ipf_nat_lock != 0)
4769 		return (0);
4770 	if (softn->ipf_nat_stats.ns_rules == 0 &&
4771 	    softn->ipf_nat_instances == NULL)
4772 		return (0);
4773 
4774 	natfailed = 0;
4775 	fr = fin->fin_fr;
4776 	sifp = fin->fin_ifp;
4777 	if (fr != NULL) {
4778 		ifp = fr->fr_tifs[fin->fin_rev].fd_ptr;
4779 		if ((ifp != NULL) && (ifp != (void *)-1))
4780 			fin->fin_ifp = ifp;
4781 	}
4782 	ifp = fin->fin_ifp;
4783 
4784 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
4785 		switch (fin->fin_p)
4786 		{
4787 		case IPPROTO_TCP :
4788 			nflags = IPN_TCP;
4789 			break;
4790 		case IPPROTO_UDP :
4791 			nflags = IPN_UDP;
4792 			break;
4793 		case IPPROTO_ICMP :
4794 			/*
4795 			 * This is an incoming packet, so the destination is
4796 			 * the icmp_id and the source port equals 0
4797 			 */
4798 			if ((fin->fin_flx & FI_ICMPQUERY) != 0)
4799 				nflags = IPN_ICMPQUERY;
4800 			break;
4801 		default :
4802 			break;
4803 		}
4804 
4805 		if ((nflags & IPN_TCPUDP))
4806 			tcp = fin->fin_dp;
4807 	}
4808 
4809 	ipa = fin->fin_saddr;
4810 
4811 	READ_ENTER(&softc->ipf_nat);
4812 
4813 	if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
4814 	    (nat = ipf_nat_icmperror(fin, &nflags, NAT_OUTBOUND)))
4815 		/*EMPTY*/;
4816 	else if ((fin->fin_flx & FI_FRAG) && (nat = ipf_frag_natknown(fin)))
4817 		natadd = 0;
4818 	else if ((nat = ipf_nat_outlookup(fin, nflags|NAT_SEARCH,
4819 				      (u_int)fin->fin_p, fin->fin_src,
4820 				      fin->fin_dst))) {
4821 		nflags = nat->nat_flags;
4822 	} else if (fin->fin_off == 0) {
4823 		u_32_t hv, msk, nmsk = 0;
4824 
4825 		/*
4826 		 * If there is no current entry in the nat table for this IP#,
4827 		 * create one for it (if there is a matching rule).
4828 		 */
4829 maskloop:
4830 		msk = softn->ipf_nat_map_active_masks[nmsk];
4831 		iph = ipa & msk;
4832 		hv = NAT_HASH_FN(iph, 0, softn->ipf_nat_maprules_sz);
4833 retry_roundrobin:
4834 		for (np = softn->ipf_nat_map_rules[hv]; np; np = npnext) {
4835 			npnext = np->in_mnext;
4836 			if ((np->in_ifps[1] && (np->in_ifps[1] != ifp)))
4837 				continue;
4838 			if (np->in_v[0] != 4)
4839 				continue;
4840 			if (np->in_pr[1] && (np->in_pr[1] != fin->fin_p))
4841 				continue;
4842 			if ((np->in_flags & IPN_RF) &&
4843 			    !(np->in_flags & nflags))
4844 				continue;
4845 			if (np->in_flags & IPN_FILTER) {
4846 				switch (ipf_nat_match(fin, np))
4847 				{
4848 				case 0 :
4849 					continue;
4850 				case -1 :
4851 					rval = -3;
4852 					goto outmatchfail;
4853 				case 1 :
4854 				default :
4855 					break;
4856 				}
4857 			} else if ((ipa & np->in_osrcmsk) != np->in_osrcaddr)
4858 				continue;
4859 
4860 			if ((fr != NULL) &&
4861 			    !ipf_matchtag(&np->in_tag, &fr->fr_nattag))
4862 				continue;
4863 
4864 			if (np->in_plabel != -1) {
4865 				if (((np->in_flags & IPN_FILTER) == 0) &&
4866 				    (np->in_odport != fin->fin_data[1]))
4867 					continue;
4868 				if (ipf_proxy_ok(fin, tcp, np) == 0)
4869 					continue;
4870 			}
4871 
4872 			if (np->in_flags & IPN_NO) {
4873 				np->in_hits++;
4874 				break;
4875 			}
4876 			MUTEX_ENTER(&softn->ipf_nat_new);
4877 			/*
4878 			 * If we've matched a round-robin rule but it has
4879 			 * moved in the list since we got it, start over as
4880 			 * this is now no longer correct.
4881 			 */
4882 			if (npnext != np->in_mnext) {
4883 				if ((np->in_flags & IPN_ROUNDR) != 0) {
4884 					MUTEX_EXIT(&softn->ipf_nat_new);
4885 					goto retry_roundrobin;
4886 				}
4887 				npnext = np->in_mnext;
4888 			}
4889 
4890 			nat = ipf_nat_add(fin, np, NULL, nflags, NAT_OUTBOUND);
4891 			MUTEX_EXIT(&softn->ipf_nat_new);
4892 			if (nat != NULL) {
4893 				natfailed = 0;
4894 				break;
4895 			}
4896 			natfailed = -2;
4897 		}
4898 		if ((np == NULL) && (nmsk < softn->ipf_nat_map_max)) {
4899 			nmsk++;
4900 			goto maskloop;
4901 		}
4902 	}
4903 
4904 	if (nat != NULL) {
4905 		rval = ipf_nat_out(fin, nat, natadd, nflags);
4906 		if (rval == 1) {
4907 			MUTEX_ENTER(&nat->nat_lock);
4908 			ipf_nat_update(fin, nat);
4909 			nat->nat_bytes[1] += fin->fin_plen;
4910 			nat->nat_pkts[1]++;
4911 			fin->fin_pktnum = nat->nat_pkts[1];
4912 			MUTEX_EXIT(&nat->nat_lock);
4913 		}
4914 	} else
4915 		rval = natfailed;
4916 outmatchfail:
4917 	RWLOCK_EXIT(&softc->ipf_nat);
4918 
4919 	switch (rval)
4920 	{
4921 	case -3 :
4922 		/* ipf_nat_match() failure */
4923 		/* FALLTHROUGH */
4924 	case -2 :
4925 		/* retry_roundrobin loop failure */
4926 		/* FALLTHROUGH */
4927 	case -1 :
4928 		/* proxy failure detected by ipf_nat_out() */
4929 		if (passp != NULL) {
4930 			DT2(frb_natv4out, fr_info_t *, fin, int, rval);
4931 			NBUMPSIDED(1, ns_drop);
4932 			*passp = FR_BLOCK;
4933 			fin->fin_reason = FRB_NATV4;
4934 		}
4935 		fin->fin_flx |= FI_BADNAT;
4936 		NBUMPSIDED(1, ns_badnat);
4937 		rval = -1;	/* We only return -1 on error. */
4938 		break;
4939 	case 0 :
4940 		NBUMPSIDE(1, ns_ignored);
4941 		break;
4942 	case 1 :
4943 		NBUMPSIDE(1, ns_translated);
4944 		break;
4945 	}
4946 	fin->fin_ifp = sifp;
4947 	return (rval);
4948 }
4949 
4950 /* ------------------------------------------------------------------------ */
4951 /* Function:    ipf_nat_out                                                 */
4952 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
4953 /*                     1 == packet was successfully translated.             */
4954 /* Parameters:  fin(I)    - pointer to packet information                   */
4955 /*              nat(I)    - pointer to NAT structure                        */
4956 /*              natadd(I) - flag indicating if it is safe to add frag cache */
4957 /*              nflags(I) - NAT flags set for this packet                   */
4958 /*                                                                          */
4959 /* Translate a packet coming "out" on an interface.                         */
4960 /* ------------------------------------------------------------------------ */
4961 int
ipf_nat_out(fr_info_t * fin,nat_t * nat,int natadd,u_32_t nflags)4962 ipf_nat_out(fr_info_t *fin, nat_t *nat, int natadd, u_32_t nflags)
4963 {
4964 	ipf_main_softc_t *softc = fin->fin_main_soft;
4965 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
4966 	icmphdr_t *icmp;
4967 	tcphdr_t *tcp;
4968 	ipnat_t *np;
4969 	int skip;
4970 	int i;
4971 
4972 	tcp = NULL;
4973 	icmp = NULL;
4974 	np = nat->nat_ptr;
4975 
4976 	if ((natadd != 0) && (fin->fin_flx & FI_FRAG) && (np != NULL))
4977 		(void) ipf_frag_natnew(softc, fin, 0, nat);
4978 
4979 	/*
4980 	 * Fix up checksums, not by recalculating them, but
4981 	 * simply computing adjustments.
4982 	 * This is only done for STREAMS based IP implementations where the
4983 	 * checksum has already been calculated by IP.  In all other cases,
4984 	 * IPFilter is called before the checksum needs calculating so there
4985 	 * is no call to modify whatever is in the header now.
4986 	 */
4987 	if (nflags == IPN_ICMPERR) {
4988 		u_32_t s1, s2, sumd, msumd;
4989 
4990 		s1 = LONG_SUM(ntohl(fin->fin_saddr));
4991 		if (nat->nat_dir == NAT_OUTBOUND) {
4992 			s2 = LONG_SUM(ntohl(nat->nat_nsrcaddr));
4993 		} else {
4994 			s2 = LONG_SUM(ntohl(nat->nat_odstaddr));
4995 		}
4996 		CALC_SUMD(s1, s2, sumd);
4997 		msumd = sumd;
4998 
4999 		s1 = LONG_SUM(ntohl(fin->fin_daddr));
5000 		if (nat->nat_dir == NAT_OUTBOUND) {
5001 			s2 = LONG_SUM(ntohl(nat->nat_ndstaddr));
5002 		} else {
5003 			s2 = LONG_SUM(ntohl(nat->nat_osrcaddr));
5004 		}
5005 		CALC_SUMD(s1, s2, sumd);
5006 		msumd += sumd;
5007 
5008 		ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, msumd, 0);
5009 	}
5010 #if !defined(_KERNEL) || SOLARIS || \
5011     defined(BRIDGE_IPF) || defined(__FreeBSD__)
5012 	else {
5013 		/*
5014 		 * We always do this on FreeBSD because this code doesn't
5015 		 * exist in fastforward.
5016 		 */
5017 		switch (nat->nat_dir)
5018 		{
5019 		case NAT_OUTBOUND :
5020 			ipf_fix_outcksum(fin->fin_cksum & FI_CK_L4PART,
5021 					 &fin->fin_ip->ip_sum,
5022 					 nat->nat_ipsumd, 0);
5023 			break;
5024 
5025 		case NAT_INBOUND :
5026 			ipf_fix_incksum(fin->fin_cksum & FI_CK_L4PART,
5027 					&fin->fin_ip->ip_sum,
5028 					nat->nat_ipsumd, 0);
5029 			break;
5030 
5031 		default :
5032 			break;
5033 		}
5034 	}
5035 #endif
5036 
5037 	/*
5038 	 * Address assignment is after the checksum modification because
5039 	 * we are using the address in the packet for determining the
5040 	 * correct checksum offset (the ICMP error could be coming from
5041 	 * anyone...)
5042 	 */
5043 	switch (nat->nat_dir)
5044 	{
5045 	case NAT_OUTBOUND :
5046 		fin->fin_ip->ip_src = nat->nat_nsrcip;
5047 		fin->fin_saddr = nat->nat_nsrcaddr;
5048 		fin->fin_ip->ip_dst = nat->nat_ndstip;
5049 		fin->fin_daddr = nat->nat_ndstaddr;
5050 		break;
5051 
5052 	case NAT_INBOUND :
5053 		fin->fin_ip->ip_src = nat->nat_odstip;
5054 		fin->fin_saddr = nat->nat_ndstaddr;
5055 		fin->fin_ip->ip_dst = nat->nat_osrcip;
5056 		fin->fin_daddr = nat->nat_nsrcaddr;
5057 		break;
5058 
5059 	case NAT_DIVERTIN :
5060 	    {
5061 		mb_t *m;
5062 
5063 		skip = ipf_nat_decap(fin, nat);
5064 		if (skip <= 0) {
5065 			NBUMPSIDED(1, ns_decap_fail);
5066 			return (-1);
5067 		}
5068 
5069 		m = fin->fin_m;
5070 
5071 #if SOLARIS && defined(_KERNEL)
5072 		m->b_rptr += skip;
5073 #else
5074 		m->m_data += skip;
5075 		m->m_len -= skip;
5076 
5077 # ifdef M_PKTHDR
5078 		if (m->m_flags & M_PKTHDR)
5079 			m->m_pkthdr.len -= skip;
5080 # endif
5081 #endif
5082 
5083 		MUTEX_ENTER(&nat->nat_lock);
5084 		ipf_nat_update(fin, nat);
5085 		MUTEX_EXIT(&nat->nat_lock);
5086 		fin->fin_flx |= FI_NATED;
5087 		if (np != NULL && np->in_tag.ipt_num[0] != 0)
5088 			fin->fin_nattag = &np->in_tag;
5089 		return (1);
5090 		/* NOTREACHED */
5091 	    }
5092 
5093 	case NAT_DIVERTOUT :
5094 	    {
5095 		u_32_t s1, s2, sumd;
5096 		udphdr_t *uh;
5097 		ip_t *ip;
5098 		mb_t *m;
5099 
5100 		m = M_DUP(np->in_divmp);
5101 		if (m == NULL) {
5102 			NBUMPSIDED(1, ns_divert_dup);
5103 			return (-1);
5104 		}
5105 
5106 		ip = MTOD(m, ip_t *);
5107 		ip_fillid(ip, V_ip_random_id);
5108 		s2 = ntohs(ip->ip_id);
5109 
5110 		s1 = ip->ip_len;
5111 		ip->ip_len = ntohs(ip->ip_len);
5112 		ip->ip_len += fin->fin_plen;
5113 		ip->ip_len = htons(ip->ip_len);
5114 		s2 += ntohs(ip->ip_len);
5115 		CALC_SUMD(s1, s2, sumd);
5116 
5117 		uh = (udphdr_t *)(ip + 1);
5118 		uh->uh_ulen += fin->fin_plen;
5119 		uh->uh_ulen = htons(uh->uh_ulen);
5120 #if !defined(_KERNEL) || SOLARIS || \
5121     defined(BRIDGE_IPF) || defined(__FreeBSD__)
5122 		ipf_fix_outcksum(0, &ip->ip_sum, sumd, 0);
5123 #endif
5124 
5125 		PREP_MB_T(fin, m);
5126 
5127 		fin->fin_src = ip->ip_src;
5128 		fin->fin_dst = ip->ip_dst;
5129 		fin->fin_ip = ip;
5130 		fin->fin_plen += sizeof(ip_t) + 8;	/* UDP + IPv4 hdr */
5131 		fin->fin_dlen += sizeof(ip_t) + 8;	/* UDP + IPv4 hdr */
5132 
5133 		nflags &= ~IPN_TCPUDPICMP;
5134 
5135 		break;
5136 	    }
5137 
5138 	default :
5139 		break;
5140 	}
5141 
5142 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5143 		u_short *csump;
5144 
5145 		if ((nat->nat_nsport != 0) && (nflags & IPN_TCPUDP)) {
5146 			tcp = fin->fin_dp;
5147 
5148 			switch (nat->nat_dir)
5149 			{
5150 			case NAT_OUTBOUND :
5151 				tcp->th_sport = nat->nat_nsport;
5152 				fin->fin_data[0] = ntohs(nat->nat_nsport);
5153 				tcp->th_dport = nat->nat_ndport;
5154 				fin->fin_data[1] = ntohs(nat->nat_ndport);
5155 				break;
5156 
5157 			case NAT_INBOUND :
5158 				tcp->th_sport = nat->nat_odport;
5159 				fin->fin_data[0] = ntohs(nat->nat_odport);
5160 				tcp->th_dport = nat->nat_osport;
5161 				fin->fin_data[1] = ntohs(nat->nat_osport);
5162 				break;
5163 			}
5164 		}
5165 
5166 		if ((nat->nat_nsport != 0) && (nflags & IPN_ICMPQUERY)) {
5167 			icmp = fin->fin_dp;
5168 			icmp->icmp_id = nat->nat_nicmpid;
5169 		}
5170 
5171 		csump = ipf_nat_proto(fin, nat, nflags);
5172 
5173 		/*
5174 		 * The above comments do not hold for layer 4 (or higher)
5175 		 * checksums...
5176 		 */
5177 		if (csump != NULL) {
5178 			if (nat->nat_dir == NAT_OUTBOUND)
5179 				ipf_fix_outcksum(fin->fin_cksum, csump,
5180 						 nat->nat_sumd[0],
5181 						 nat->nat_sumd[1] +
5182 						 fin->fin_dlen);
5183 			else
5184 				ipf_fix_incksum(fin->fin_cksum, csump,
5185 						nat->nat_sumd[0],
5186 						nat->nat_sumd[1] +
5187 						fin->fin_dlen);
5188 		}
5189 	}
5190 
5191 	ipf_sync_update(softc, SMC_NAT, fin, nat->nat_sync);
5192 	/* ------------------------------------------------------------- */
5193 	/* A few quick notes:                                            */
5194 	/*      Following are test conditions prior to calling the       */
5195 	/*      ipf_proxy_check routine.                                 */
5196 	/*                                                               */
5197 	/*      A NULL tcp indicates a non TCP/UDP packet.  When dealing */
5198 	/*      with a redirect rule, we attempt to match the packet's   */
5199 	/*      source port against in_dport, otherwise we'd compare the */
5200 	/*      packet's destination.                                    */
5201 	/* ------------------------------------------------------------- */
5202 	if ((np != NULL) && (np->in_apr != NULL)) {
5203 		i = ipf_proxy_check(fin, nat);
5204 		if (i == -1) {
5205 			NBUMPSIDED(1, ns_ipf_proxy_fail);
5206 		}
5207 	} else {
5208 		i = 1;
5209 	}
5210 	fin->fin_flx |= FI_NATED;
5211 	return (i);
5212 }
5213 
5214 
5215 /* ------------------------------------------------------------------------ */
5216 /* Function:    ipf_nat_checkin                                             */
5217 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
5218 /*                     0 == no packet translation occurred,                 */
5219 /*                     1 == packet was successfully translated.             */
5220 /* Parameters:  fin(I)   - pointer to packet information                    */
5221 /*              passp(I) - pointer to filtering result flags                */
5222 /*                                                                          */
5223 /* Check to see if an incoming packet should be changed.  ICMP packets are  */
5224 /* first checked to see if they match an existing entry (if an error),      */
5225 /* otherwise a search of the current NAT table is made.  If neither results */
5226 /* in a match then a search for a matching NAT rule is made.  Create a new  */
5227 /* NAT entry if a we matched a NAT rule.  Lastly, actually change the       */
5228 /* packet header(s) as required.                                            */
5229 /* ------------------------------------------------------------------------ */
5230 int
ipf_nat_checkin(fr_info_t * fin,u_32_t * passp)5231 ipf_nat_checkin(fr_info_t *fin, u_32_t *passp)
5232 {
5233 	ipf_main_softc_t *softc;
5234 	ipf_nat_softc_t *softn;
5235 	u_int nflags, natadd;
5236 	ipnat_t *np, *npnext;
5237 	int rval, natfailed;
5238 	struct ifnet *ifp;
5239 	struct in_addr in;
5240 	icmphdr_t *icmp;
5241 	tcphdr_t *tcp;
5242 	u_short dport;
5243 	nat_t *nat;
5244 	u_32_t iph;
5245 
5246 	softc = fin->fin_main_soft;
5247 	softn = softc->ipf_nat_soft;
5248 
5249 	if (softn->ipf_nat_lock != 0)
5250 		return (0);
5251 	if (softn->ipf_nat_stats.ns_rules == 0 &&
5252 	    softn->ipf_nat_instances == NULL)
5253 		return (0);
5254 
5255 	tcp = NULL;
5256 	icmp = NULL;
5257 	dport = 0;
5258 	natadd = 1;
5259 	nflags = 0;
5260 	natfailed = 0;
5261 	ifp = fin->fin_ifp;
5262 
5263 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5264 		switch (fin->fin_p)
5265 		{
5266 		case IPPROTO_TCP :
5267 			nflags = IPN_TCP;
5268 			break;
5269 		case IPPROTO_UDP :
5270 			nflags = IPN_UDP;
5271 			break;
5272 		case IPPROTO_ICMP :
5273 			icmp = fin->fin_dp;
5274 
5275 			/*
5276 			 * This is an incoming packet, so the destination is
5277 			 * the icmp_id and the source port equals 0
5278 			 */
5279 			if ((fin->fin_flx & FI_ICMPQUERY) != 0) {
5280 				nflags = IPN_ICMPQUERY;
5281 				dport = icmp->icmp_id;
5282 			} break;
5283 		default :
5284 			break;
5285 		}
5286 
5287 		if ((nflags & IPN_TCPUDP)) {
5288 			tcp = fin->fin_dp;
5289 			dport = fin->fin_data[1];
5290 		}
5291 	}
5292 
5293 	in = fin->fin_dst;
5294 
5295 	READ_ENTER(&softc->ipf_nat);
5296 
5297 	if ((fin->fin_p == IPPROTO_ICMP) && !(nflags & IPN_ICMPQUERY) &&
5298 	    (nat = ipf_nat_icmperror(fin, &nflags, NAT_INBOUND)))
5299 		/*EMPTY*/;
5300 	else if ((fin->fin_flx & FI_FRAG) && (nat = ipf_frag_natknown(fin)))
5301 		natadd = 0;
5302 	else if ((nat = ipf_nat_inlookup(fin, nflags|NAT_SEARCH,
5303 					 (u_int)fin->fin_p,
5304 					 fin->fin_src, in))) {
5305 		nflags = nat->nat_flags;
5306 	} else if (fin->fin_off == 0) {
5307 		u_32_t hv, msk, rmsk = 0;
5308 
5309 		/*
5310 		 * If there is no current entry in the nat table for this IP#,
5311 		 * create one for it (if there is a matching rule).
5312 		 */
5313 maskloop:
5314 		msk = softn->ipf_nat_rdr_active_masks[rmsk];
5315 		iph = in.s_addr & msk;
5316 		hv = NAT_HASH_FN(iph, 0, softn->ipf_nat_rdrrules_sz);
5317 retry_roundrobin:
5318 		/* TRACE (iph,msk,rmsk,hv,softn->ipf_nat_rdrrules_sz) */
5319 		for (np = softn->ipf_nat_rdr_rules[hv]; np; np = npnext) {
5320 			npnext = np->in_rnext;
5321 			if (np->in_ifps[0] && (np->in_ifps[0] != ifp))
5322 				continue;
5323 			if (np->in_v[0] != 4)
5324 				continue;
5325 			if (np->in_pr[0] && (np->in_pr[0] != fin->fin_p))
5326 				continue;
5327 			if ((np->in_flags & IPN_RF) && !(np->in_flags & nflags))
5328 				continue;
5329 			if (np->in_flags & IPN_FILTER) {
5330 				switch (ipf_nat_match(fin, np))
5331 				{
5332 				case 0 :
5333 					continue;
5334 				case -1 :
5335 					rval = -3;
5336 					goto inmatchfail;
5337 				case 1 :
5338 				default :
5339 					break;
5340 				}
5341 			} else {
5342 				if ((in.s_addr & np->in_odstmsk) !=
5343 				    np->in_odstaddr)
5344 					continue;
5345 				if (np->in_odport &&
5346 				    ((np->in_dtop < dport) ||
5347 				     (dport < np->in_odport)))
5348 					continue;
5349 			}
5350 
5351 			if (np->in_plabel != -1) {
5352 				if (!ipf_proxy_ok(fin, tcp, np)) {
5353 					continue;
5354 				}
5355 			}
5356 
5357 			if (np->in_flags & IPN_NO) {
5358 				np->in_hits++;
5359 				break;
5360 			}
5361 
5362 			MUTEX_ENTER(&softn->ipf_nat_new);
5363 			/*
5364 			 * If we've matched a round-robin rule but it has
5365 			 * moved in the list since we got it, start over as
5366 			 * this is now no longer correct.
5367 			 */
5368 			if (npnext != np->in_rnext) {
5369 				if ((np->in_flags & IPN_ROUNDR) != 0) {
5370 					MUTEX_EXIT(&softn->ipf_nat_new);
5371 					goto retry_roundrobin;
5372 				}
5373 				npnext = np->in_rnext;
5374 			}
5375 
5376 			nat = ipf_nat_add(fin, np, NULL, nflags, NAT_INBOUND);
5377 			MUTEX_EXIT(&softn->ipf_nat_new);
5378 			if (nat != NULL) {
5379 				natfailed = 0;
5380 				break;
5381 			}
5382 			natfailed = -2;
5383 		}
5384 		if ((np == NULL) && (rmsk < softn->ipf_nat_rdr_max)) {
5385 			rmsk++;
5386 			goto maskloop;
5387 		}
5388 	}
5389 
5390 	if (nat != NULL) {
5391 		rval = ipf_nat_in(fin, nat, natadd, nflags);
5392 		if (rval == 1) {
5393 			MUTEX_ENTER(&nat->nat_lock);
5394 			ipf_nat_update(fin, nat);
5395 			nat->nat_bytes[0] += fin->fin_plen;
5396 			nat->nat_pkts[0]++;
5397 			fin->fin_pktnum = nat->nat_pkts[0];
5398 			MUTEX_EXIT(&nat->nat_lock);
5399 		}
5400 	} else
5401 		rval = natfailed;
5402 inmatchfail:
5403 	RWLOCK_EXIT(&softc->ipf_nat);
5404 
5405 	DT2(frb_natv4in, fr_info_t *, fin, int, rval);
5406 	switch (rval)
5407 	{
5408 	case -3 :
5409 		/* ipf_nat_match() failure */
5410 		/* FALLTHROUGH */
5411 	case -2 :
5412 		/* retry_roundrobin loop failure */
5413 		/* FALLTHROUGH */
5414 	case -1 :
5415 		/* proxy failure detected by ipf_nat_in() */
5416 		if (passp != NULL) {
5417 			NBUMPSIDED(0, ns_drop);
5418 			*passp = FR_BLOCK;
5419 			fin->fin_reason = FRB_NATV4;
5420 		}
5421 		fin->fin_flx |= FI_BADNAT;
5422 		NBUMPSIDED(0, ns_badnat);
5423 		rval = -1;	/* We only return -1 on error. */
5424 		break;
5425 	case 0 :
5426 		NBUMPSIDE(0, ns_ignored);
5427 		break;
5428 	case 1 :
5429 		NBUMPSIDE(0, ns_translated);
5430 		break;
5431 	}
5432 	return (rval);
5433 }
5434 
5435 
5436 /* ------------------------------------------------------------------------ */
5437 /* Function:    ipf_nat_in                                                  */
5438 /* Returns:     int - -1 == packet failed NAT checks so block it,           */
5439 /*                     1 == packet was successfully translated.             */
5440 /* Parameters:  fin(I)    - pointer to packet information                   */
5441 /*              nat(I)    - pointer to NAT structure                        */
5442 /*              natadd(I) - flag indicating if it is safe to add frag cache */
5443 /*              nflags(I) - NAT flags set for this packet                   */
5444 /* Locks Held:  ipf_nat(READ)                                               */
5445 /*                                                                          */
5446 /* Translate a packet coming "in" on an interface.                          */
5447 /* ------------------------------------------------------------------------ */
5448 int
ipf_nat_in(fr_info_t * fin,nat_t * nat,int natadd,u_32_t nflags)5449 ipf_nat_in(fr_info_t *fin, nat_t *nat, int natadd, u_32_t nflags)
5450 {
5451 	ipf_main_softc_t *softc = fin->fin_main_soft;
5452 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5453 	u_32_t sumd, sum1, sum2;
5454 #if !defined(_KERNEL) || SOLARIS
5455 	u_32_t ipsumd;
5456 #endif
5457 	icmphdr_t *icmp;
5458 	tcphdr_t *tcp;
5459 	ipnat_t *np;
5460 	int skip;
5461 	int i;
5462 
5463 	tcp = NULL;
5464 	np = nat->nat_ptr;
5465 	fin->fin_fr = nat->nat_fr;
5466 
5467 	if (np != NULL) {
5468 		if ((natadd != 0) && (fin->fin_flx & FI_FRAG))
5469 			(void) ipf_frag_natnew(softc, fin, 0, nat);
5470 
5471 	/* ------------------------------------------------------------- */
5472 	/* A few quick notes:                                            */
5473 	/*      Following are test conditions prior to calling the       */
5474 	/*      ipf_proxy_check routine.                                 */
5475 	/*                                                               */
5476 	/*      A NULL tcp indicates a non TCP/UDP packet.  When dealing */
5477 	/*      with a map rule, we attempt to match the packet's        */
5478 	/*      source port against in_dport, otherwise we'd compare the */
5479 	/*      packet's destination.                                    */
5480 	/* ------------------------------------------------------------- */
5481 		if (np->in_apr != NULL) {
5482 			i = ipf_proxy_check(fin, nat);
5483 			if (i == -1) {
5484 				NBUMPSIDED(0, ns_ipf_proxy_fail);
5485 				return (-1);
5486 			}
5487 		}
5488 	}
5489 
5490 	ipf_sync_update(softc, SMC_NAT, fin, nat->nat_sync);
5491 
5492 #if !defined(_KERNEL) || SOLARIS
5493 	ipsumd = nat->nat_ipsumd;
5494 #endif
5495 	/*
5496 	 * Fix up checksums, not by recalculating them, but
5497 	 * simply computing adjustments.
5498 	 * Why only do this for some platforms on inbound packets ?
5499 	 * Because for those that it is done, IP processing is yet to happen
5500 	 * and so the IPv4 header checksum has not yet been evaluated.
5501 	 * Perhaps it should always be done for the benefit of things like
5502 	 * fast forwarding (so that it doesn't need to be recomputed) but with
5503 	 * header checksum offloading, perhaps it is a moot point.
5504 	 */
5505 
5506 	switch (nat->nat_dir)
5507 	{
5508 	case NAT_INBOUND :
5509 		if ((fin->fin_flx & FI_ICMPERR) == 0) {
5510 			fin->fin_ip->ip_src = nat->nat_nsrcip;
5511 			fin->fin_saddr = nat->nat_nsrcaddr;
5512 		} else {
5513 			sum1 = nat->nat_osrcaddr;
5514 			sum2 = nat->nat_nsrcaddr;
5515 			CALC_SUMD(sum1, sum2, sumd);
5516 #if !defined(_KERNEL) || SOLARIS
5517 			ipsumd -= sumd;
5518 #endif
5519 		}
5520 		fin->fin_ip->ip_dst = nat->nat_ndstip;
5521 		fin->fin_daddr = nat->nat_ndstaddr;
5522 #if !defined(_KERNEL) || SOLARIS
5523 		ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, ipsumd, 0);
5524 #endif
5525 		break;
5526 
5527 	case NAT_OUTBOUND :
5528 		if ((fin->fin_flx & FI_ICMPERR) == 0) {
5529 			fin->fin_ip->ip_src = nat->nat_odstip;
5530 			fin->fin_saddr = nat->nat_odstaddr;
5531 		} else {
5532 			sum1 = nat->nat_odstaddr;
5533 			sum2 = nat->nat_ndstaddr;
5534 			CALC_SUMD(sum1, sum2, sumd);
5535 #if !defined(_KERNEL) || SOLARIS
5536 			ipsumd -= sumd;
5537 #endif
5538 		}
5539 		fin->fin_ip->ip_dst = nat->nat_osrcip;
5540 		fin->fin_daddr = nat->nat_osrcaddr;
5541 #if !defined(_KERNEL) || SOLARIS
5542 		ipf_fix_incksum(0, &fin->fin_ip->ip_sum, ipsumd, 0);
5543 #endif
5544 		break;
5545 
5546 	case NAT_DIVERTIN :
5547 	    {
5548 		udphdr_t *uh;
5549 		ip_t *ip;
5550 		mb_t *m;
5551 
5552 		m = M_DUP(np->in_divmp);
5553 		if (m == NULL) {
5554 			NBUMPSIDED(0, ns_divert_dup);
5555 			return (-1);
5556 		}
5557 
5558 		ip = MTOD(m, ip_t *);
5559 		ip_fillid(ip, V_ip_random_id);
5560 		sum1 = ntohs(ip->ip_len);
5561 		ip->ip_len = ntohs(ip->ip_len);
5562 		ip->ip_len += fin->fin_plen;
5563 		ip->ip_len = htons(ip->ip_len);
5564 
5565 		uh = (udphdr_t *)(ip + 1);
5566 		uh->uh_ulen += fin->fin_plen;
5567 		uh->uh_ulen = htons(uh->uh_ulen);
5568 
5569 		sum2 = ntohs(ip->ip_id) + ntohs(ip->ip_len);
5570 		sum2 += ntohs(ip->ip_off) & IP_DF;
5571 		CALC_SUMD(sum1, sum2, sumd);
5572 
5573 #if !defined(_KERNEL) || SOLARIS
5574 		ipf_fix_outcksum(0, &ip->ip_sum, sumd, 0);
5575 #endif
5576 		PREP_MB_T(fin, m);
5577 
5578 		fin->fin_ip = ip;
5579 		fin->fin_plen += sizeof(ip_t) + 8;	/* UDP + new IPv4 hdr */
5580 		fin->fin_dlen += sizeof(ip_t) + 8;	/* UDP + old IPv4 hdr */
5581 
5582 		nflags &= ~IPN_TCPUDPICMP;
5583 
5584 		break;
5585 	    }
5586 
5587 	case NAT_DIVERTOUT :
5588 	    {
5589 		mb_t *m;
5590 
5591 		skip = ipf_nat_decap(fin, nat);
5592 		if (skip <= 0) {
5593 			NBUMPSIDED(0, ns_decap_fail);
5594 			return (-1);
5595 		}
5596 
5597 		m = fin->fin_m;
5598 
5599 #if SOLARIS && defined(_KERNEL)
5600 		m->b_rptr += skip;
5601 #else
5602 		m->m_data += skip;
5603 		m->m_len -= skip;
5604 
5605 # ifdef M_PKTHDR
5606 		if (m->m_flags & M_PKTHDR)
5607 			m->m_pkthdr.len -= skip;
5608 # endif
5609 #endif
5610 
5611 		ipf_nat_update(fin, nat);
5612 		nflags &= ~IPN_TCPUDPICMP;
5613 		fin->fin_flx |= FI_NATED;
5614 		if (np != NULL && np->in_tag.ipt_num[0] != 0)
5615 			fin->fin_nattag = &np->in_tag;
5616 		return (1);
5617 		/* NOTREACHED */
5618 	    }
5619 	}
5620 	if (nflags & IPN_TCPUDP)
5621 		tcp = fin->fin_dp;
5622 
5623 	if (!(fin->fin_flx & FI_SHORT) && (fin->fin_off == 0)) {
5624 		u_short *csump;
5625 
5626 		if ((nat->nat_odport != 0) && (nflags & IPN_TCPUDP)) {
5627 			switch (nat->nat_dir)
5628 			{
5629 			case NAT_INBOUND :
5630 				tcp->th_sport = nat->nat_nsport;
5631 				fin->fin_data[0] = ntohs(nat->nat_nsport);
5632 				tcp->th_dport = nat->nat_ndport;
5633 				fin->fin_data[1] = ntohs(nat->nat_ndport);
5634 				break;
5635 
5636 			case NAT_OUTBOUND :
5637 				tcp->th_sport = nat->nat_odport;
5638 				fin->fin_data[0] = ntohs(nat->nat_odport);
5639 				tcp->th_dport = nat->nat_osport;
5640 				fin->fin_data[1] = ntohs(nat->nat_osport);
5641 				break;
5642 			}
5643 		}
5644 
5645 
5646 		if ((nat->nat_odport != 0) && (nflags & IPN_ICMPQUERY)) {
5647 			icmp = fin->fin_dp;
5648 
5649 			icmp->icmp_id = nat->nat_nicmpid;
5650 		}
5651 
5652 		csump = ipf_nat_proto(fin, nat, nflags);
5653 
5654 		/*
5655 		 * The above comments do not hold for layer 4 (or higher)
5656 		 * checksums...
5657 		 */
5658 		if (csump != NULL) {
5659 			if (nat->nat_dir == NAT_OUTBOUND)
5660 				ipf_fix_incksum(0, csump, nat->nat_sumd[0], 0);
5661 			else
5662 				ipf_fix_outcksum(0, csump, nat->nat_sumd[0], 0);
5663 		}
5664 	}
5665 
5666 	fin->fin_flx |= FI_NATED;
5667 	if (np != NULL && np->in_tag.ipt_num[0] != 0)
5668 		fin->fin_nattag = &np->in_tag;
5669 	return (1);
5670 }
5671 
5672 
5673 /* ------------------------------------------------------------------------ */
5674 /* Function:    ipf_nat_proto                                               */
5675 /* Returns:     u_short* - pointer to transport header checksum to update,  */
5676 /*                         NULL if the transport protocol is not recognised */
5677 /*                         as needing a checksum update.                    */
5678 /* Parameters:  fin(I)    - pointer to packet information                   */
5679 /*              nat(I)    - pointer to NAT structure                        */
5680 /*              nflags(I) - NAT flags set for this packet                   */
5681 /*                                                                          */
5682 /* Return the pointer to the checksum field for each protocol so understood.*/
5683 /* If support for making other changes to a protocol header is required,    */
5684 /* that is not strictly 'address' translation, such as clamping the MSS in  */
5685 /* TCP down to a specific value, then do it from here.                      */
5686 /* ------------------------------------------------------------------------ */
5687 u_short *
ipf_nat_proto(fr_info_t * fin,nat_t * nat,u_int nflags)5688 ipf_nat_proto(fr_info_t *fin, nat_t *nat, u_int nflags)
5689 {
5690 	icmphdr_t *icmp;
5691 	u_short *csump;
5692 	tcphdr_t *tcp;
5693 	udphdr_t *udp;
5694 
5695 	csump = NULL;
5696 	if (fin->fin_out == 0) {
5697 		fin->fin_rev = (nat->nat_dir & NAT_OUTBOUND);
5698 	} else {
5699 		fin->fin_rev = ((nat->nat_dir & NAT_OUTBOUND) == 0);
5700 	}
5701 
5702 	switch (fin->fin_p)
5703 	{
5704 	case IPPROTO_TCP :
5705 		tcp = fin->fin_dp;
5706 
5707 		if ((nflags & IPN_TCP) != 0)
5708 			csump = &tcp->th_sum;
5709 
5710 		/*
5711 		 * Do a MSS CLAMPING on a SYN packet,
5712 		 * only deal IPv4 for now.
5713 		 */
5714 		if ((nat->nat_mssclamp != 0) && (tcp_get_flags(tcp) & TH_SYN) != 0)
5715 			ipf_nat_mssclamp(tcp, nat->nat_mssclamp, fin, csump);
5716 
5717 		break;
5718 
5719 	case IPPROTO_UDP :
5720 		udp = fin->fin_dp;
5721 
5722 		if ((nflags & IPN_UDP) != 0) {
5723 			if (udp->uh_sum != 0)
5724 				csump = &udp->uh_sum;
5725 		}
5726 		break;
5727 
5728 	case IPPROTO_ICMP :
5729 		icmp = fin->fin_dp;
5730 
5731 		if ((nflags & IPN_ICMPQUERY) != 0) {
5732 			if (icmp->icmp_cksum != 0)
5733 				csump = &icmp->icmp_cksum;
5734 		}
5735 		break;
5736 
5737 #ifdef USE_INET6
5738 	case IPPROTO_ICMPV6 :
5739 	    {
5740 		struct icmp6_hdr *icmp6 = (struct icmp6_hdr *)fin->fin_dp;
5741 
5742 		icmp6 = fin->fin_dp;
5743 
5744 		if ((nflags & IPN_ICMPQUERY) != 0) {
5745 			if (icmp6->icmp6_cksum != 0)
5746 				csump = &icmp6->icmp6_cksum;
5747 		}
5748 		break;
5749 	    }
5750 #endif
5751 	}
5752 	return (csump);
5753 }
5754 
5755 
5756 /* ------------------------------------------------------------------------ */
5757 /* Function:    ipf_nat_expire                                              */
5758 /* Returns:     Nil                                                         */
5759 /* Parameters:  softc(I) - pointer to soft context main structure           */
5760 /*                                                                          */
5761 /* Check all of the timeout queues for entries at the top which need to be  */
5762 /* expired.                                                                 */
5763 /* ------------------------------------------------------------------------ */
5764 void
ipf_nat_expire(ipf_main_softc_t * softc)5765 ipf_nat_expire(ipf_main_softc_t *softc)
5766 {
5767 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5768 	ipftq_t *ifq, *ifqnext;
5769 	ipftqent_t *tqe, *tqn;
5770 	int i;
5771 	SPL_INT(s);
5772 
5773 	SPL_NET(s);
5774 	WRITE_ENTER(&softc->ipf_nat);
5775 	for (ifq = softn->ipf_nat_tcptq, i = 0; ifq != NULL;
5776 	     ifq = ifq->ifq_next) {
5777 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
5778 			if (tqe->tqe_die > softc->ipf_ticks)
5779 				break;
5780 			tqn = tqe->tqe_next;
5781 			ipf_nat_delete(softc, tqe->tqe_parent, NL_EXPIRE);
5782 		}
5783 	}
5784 
5785 	for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifq->ifq_next) {
5786 		for (tqn = ifq->ifq_head; ((tqe = tqn) != NULL); i++) {
5787 			if (tqe->tqe_die > softc->ipf_ticks)
5788 				break;
5789 			tqn = tqe->tqe_next;
5790 			ipf_nat_delete(softc, tqe->tqe_parent, NL_EXPIRE);
5791 		}
5792 	}
5793 
5794 	for (ifq = softn->ipf_nat_utqe; ifq != NULL; ifq = ifqnext) {
5795 		ifqnext = ifq->ifq_next;
5796 
5797 		if (((ifq->ifq_flags & IFQF_DELETE) != 0) &&
5798 		    (ifq->ifq_ref == 0)) {
5799 			ipf_freetimeoutqueue(softc, ifq);
5800 		}
5801 	}
5802 
5803 	if (softn->ipf_nat_doflush != 0) {
5804 		ipf_nat_extraflush(softc, softn, 2);
5805 		softn->ipf_nat_doflush = 0;
5806 	}
5807 
5808 	RWLOCK_EXIT(&softc->ipf_nat);
5809 	SPL_X(s);
5810 }
5811 
5812 
5813 /* ------------------------------------------------------------------------ */
5814 /* Function:    ipf_nat_sync                                                */
5815 /* Returns:     Nil                                                         */
5816 /* Parameters:  softc(I) - pointer to soft context main structure           */
5817 /*              ifp(I) - pointer to network interface                       */
5818 /*                                                                          */
5819 /* Walk through all of the currently active NAT sessions, looking for those */
5820 /* which need to have their translated address updated.                     */
5821 /* ------------------------------------------------------------------------ */
5822 void
ipf_nat_sync(ipf_main_softc_t * softc,void * ifp)5823 ipf_nat_sync(ipf_main_softc_t *softc, void *ifp)
5824 {
5825 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
5826 	u_32_t sum1, sum2, sumd;
5827 	i6addr_t in;
5828 	ipnat_t *n;
5829 	nat_t *nat;
5830 	void *ifp2;
5831 	int idx;
5832 	SPL_INT(s);
5833 
5834 	if (softc->ipf_running <= 0)
5835 		return;
5836 
5837 	/*
5838 	 * Change IP addresses for NAT sessions for any protocol except TCP
5839 	 * since it will break the TCP connection anyway.  The only rules
5840 	 * which will get changed are those which are "map ... -> 0/32",
5841 	 * where the rule specifies the address is taken from the interface.
5842 	 */
5843 	SPL_NET(s);
5844 	WRITE_ENTER(&softc->ipf_nat);
5845 
5846 	if (softc->ipf_running <= 0) {
5847 		RWLOCK_EXIT(&softc->ipf_nat);
5848 		return;
5849 	}
5850 
5851 	for (nat = softn->ipf_nat_instances; nat; nat = nat->nat_next) {
5852 		if ((nat->nat_flags & IPN_TCP) != 0)
5853 			continue;
5854 
5855 		n = nat->nat_ptr;
5856 		if (n != NULL) {
5857 			if (n->in_v[1] == 4) {
5858 				if (n->in_redir & NAT_MAP) {
5859 					if ((n->in_nsrcaddr != 0) ||
5860 					    (n->in_nsrcmsk != 0xffffffff))
5861 						continue;
5862 				} else if (n->in_redir & NAT_REDIRECT) {
5863 					if ((n->in_ndstaddr != 0) ||
5864 					    (n->in_ndstmsk != 0xffffffff))
5865 						continue;
5866 				}
5867 			}
5868 #ifdef USE_INET6
5869 			if (n->in_v[1] == 4) {
5870 				if (n->in_redir & NAT_MAP) {
5871 					if (!IP6_ISZERO(&n->in_nsrcaddr) ||
5872 					    !IP6_ISONES(&n->in_nsrcmsk))
5873 						continue;
5874 				} else if (n->in_redir & NAT_REDIRECT) {
5875 					if (!IP6_ISZERO(&n->in_ndstaddr) ||
5876 					    !IP6_ISONES(&n->in_ndstmsk))
5877 						continue;
5878 				}
5879 			}
5880 #endif
5881 		}
5882 
5883 		if (((ifp == NULL) || (ifp == nat->nat_ifps[0]) ||
5884 		     (ifp == nat->nat_ifps[1]))) {
5885 			nat->nat_ifps[0] = GETIFP(nat->nat_ifnames[0],
5886 						  nat->nat_v[0]);
5887 			if ((nat->nat_ifps[0] != NULL) &&
5888 			    (nat->nat_ifps[0] != (void *)-1)) {
5889 				nat->nat_mtu[0] = GETIFMTU_4(nat->nat_ifps[0]);
5890 			}
5891 			if (nat->nat_ifnames[1][0] != '\0') {
5892 				nat->nat_ifps[1] = GETIFP(nat->nat_ifnames[1],
5893 							  nat->nat_v[1]);
5894 			} else {
5895 				nat->nat_ifps[1] = nat->nat_ifps[0];
5896 			}
5897 			if ((nat->nat_ifps[1] != NULL) &&
5898 			    (nat->nat_ifps[1] != (void *)-1)) {
5899 				nat->nat_mtu[1] = GETIFMTU_4(nat->nat_ifps[1]);
5900 			}
5901 			ifp2 = nat->nat_ifps[0];
5902 			if (ifp2 == NULL)
5903 				continue;
5904 
5905 			/*
5906 			 * Change the map-to address to be the same as the
5907 			 * new one.
5908 			 */
5909 			sum1 = NATFSUM(nat, nat->nat_v[1], nat_nsrc6);
5910 			if (ipf_ifpaddr(softc, nat->nat_v[0], FRI_NORMAL, ifp2,
5911 				       &in, NULL) != -1) {
5912 				if (nat->nat_v[0] == 4)
5913 					nat->nat_nsrcip = in.in4;
5914 			}
5915 			sum2 = NATFSUM(nat, nat->nat_v[1], nat_nsrc6);
5916 
5917 			if (sum1 == sum2)
5918 				continue;
5919 			/*
5920 			 * Readjust the checksum adjustment to take into
5921 			 * account the new IP#.
5922 			 */
5923 			CALC_SUMD(sum1, sum2, sumd);
5924 			/* XXX - dont change for TCP when solaris does
5925 			 * hardware checksumming.
5926 			 */
5927 			sumd += nat->nat_sumd[0];
5928 			nat->nat_sumd[0] = (sumd & 0xffff) + (sumd >> 16);
5929 			nat->nat_sumd[1] = nat->nat_sumd[0];
5930 		}
5931 	}
5932 
5933 	for (n = softn->ipf_nat_list; (n != NULL); n = n->in_next) {
5934 		char *base = n->in_names;
5935 
5936 		if ((ifp == NULL) || (n->in_ifps[0] == ifp))
5937 			n->in_ifps[0] = ipf_resolvenic(softc,
5938 						       base + n->in_ifnames[0],
5939 						       n->in_v[0]);
5940 		if ((ifp == NULL) || (n->in_ifps[1] == ifp))
5941 			n->in_ifps[1] = ipf_resolvenic(softc,
5942 						       base + n->in_ifnames[1],
5943 						       n->in_v[1]);
5944 
5945 		if (n->in_redir & NAT_REDIRECT)
5946 			idx = 1;
5947 		else
5948 			idx = 0;
5949 
5950 		if (((ifp == NULL) || (n->in_ifps[idx] == ifp)) &&
5951 		    (n->in_ifps[idx] != NULL &&
5952 		     n->in_ifps[idx] != (void *)-1)) {
5953 
5954 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_osrc,
5955 					     0, n->in_ifps[idx]);
5956 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_odst,
5957 					     0, n->in_ifps[idx]);
5958 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_nsrc,
5959 					     0, n->in_ifps[idx]);
5960 			ipf_nat_nextaddrinit(softc, n->in_names, &n->in_ndst,
5961 					     0, n->in_ifps[idx]);
5962 		}
5963 	}
5964 	RWLOCK_EXIT(&softc->ipf_nat);
5965 	SPL_X(s);
5966 }
5967 
5968 
5969 /* ------------------------------------------------------------------------ */
5970 /* Function:    ipf_nat_icmpquerytype                                       */
5971 /* Returns:     int - 1 == success, 0 == failure                            */
5972 /* Parameters:  icmptype(I) - ICMP type number                              */
5973 /*                                                                          */
5974 /* Tests to see if the ICMP type number passed is a query/response type or  */
5975 /* not.                                                                     */
5976 /* ------------------------------------------------------------------------ */
5977 static int
ipf_nat_icmpquerytype(int icmptype)5978 ipf_nat_icmpquerytype(int icmptype)
5979 {
5980 
5981 	/*
5982 	 * For the ICMP query NAT code, it is essential that both the query
5983 	 * and the reply match on the NAT rule. Because the NAT structure
5984 	 * does not keep track of the icmptype, and a single NAT structure
5985 	 * is used for all icmp types with the same src, dest and id, we
5986 	 * simply define the replies as queries as well. The funny thing is,
5987 	 * altough it seems silly to call a reply a query, this is exactly
5988 	 * as it is defined in the IPv4 specification
5989 	 */
5990 	switch (icmptype)
5991 	{
5992 	case ICMP_ECHOREPLY:
5993 	case ICMP_ECHO:
5994 	/* route advertisement/solicitation is currently unsupported: */
5995 	/* it would require rewriting the ICMP data section          */
5996 	case ICMP_TSTAMP:
5997 	case ICMP_TSTAMPREPLY:
5998 	case ICMP_IREQ:
5999 	case ICMP_IREQREPLY:
6000 	case ICMP_MASKREQ:
6001 	case ICMP_MASKREPLY:
6002 		return (1);
6003 	default:
6004 		return (0);
6005 	}
6006 }
6007 
6008 
6009 /* ------------------------------------------------------------------------ */
6010 /* Function:    nat_log                                                     */
6011 /* Returns:     Nil                                                         */
6012 /* Parameters:  softc(I) - pointer to soft context main structure           */
6013 /*              softn(I) - pointer to NAT context structure                 */
6014 /*              nat(I)    - pointer to NAT structure                        */
6015 /*              action(I) - action related to NAT structure being performed */
6016 /*                                                                          */
6017 /* Creates a NAT log entry.                                                 */
6018 /* ------------------------------------------------------------------------ */
6019 void
ipf_nat_log(ipf_main_softc_t * softc,ipf_nat_softc_t * softn,struct nat * nat,u_int action)6020 ipf_nat_log(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, struct nat *nat,
6021 	u_int action)
6022 {
6023 #ifdef	IPFILTER_LOG
6024 	struct ipnat *np;
6025 	int rulen;
6026 	struct natlog natl;
6027 	void *items[1];
6028 	size_t sizes[1];
6029 	int types[1];
6030 
6031 	bcopy((char *)&nat->nat_osrc6, (char *)&natl.nl_osrcip,
6032 	      sizeof(natl.nl_osrcip));
6033 	bcopy((char *)&nat->nat_nsrc6, (char *)&natl.nl_nsrcip,
6034 	      sizeof(natl.nl_nsrcip));
6035 	bcopy((char *)&nat->nat_odst6, (char *)&natl.nl_odstip,
6036 	      sizeof(natl.nl_odstip));
6037 	bcopy((char *)&nat->nat_ndst6, (char *)&natl.nl_ndstip,
6038 	      sizeof(natl.nl_ndstip));
6039 
6040 	natl.nl_bytes[0] = nat->nat_bytes[0];
6041 	natl.nl_bytes[1] = nat->nat_bytes[1];
6042 	natl.nl_pkts[0] = nat->nat_pkts[0];
6043 	natl.nl_pkts[1] = nat->nat_pkts[1];
6044 	natl.nl_odstport = nat->nat_odport;
6045 	natl.nl_osrcport = nat->nat_osport;
6046 	natl.nl_nsrcport = nat->nat_nsport;
6047 	natl.nl_ndstport = nat->nat_ndport;
6048 	natl.nl_p[0] = nat->nat_pr[0];
6049 	natl.nl_p[1] = nat->nat_pr[1];
6050 	natl.nl_v[0] = nat->nat_v[0];
6051 	natl.nl_v[1] = nat->nat_v[1];
6052 	natl.nl_type = nat->nat_redir;
6053 	natl.nl_action = action;
6054 	natl.nl_rule = -1;
6055 
6056 	bcopy(nat->nat_ifnames[0], natl.nl_ifnames[0],
6057 	      sizeof(nat->nat_ifnames[0]));
6058 	bcopy(nat->nat_ifnames[1], natl.nl_ifnames[1],
6059 	      sizeof(nat->nat_ifnames[1]));
6060 
6061 	if (softc->ipf_large_nat && nat->nat_ptr != NULL) {
6062 		for (rulen = 0, np = softn->ipf_nat_list; np != NULL;
6063 		     np = np->in_next, rulen++)
6064 			if (np == nat->nat_ptr) {
6065 				natl.nl_rule = rulen;
6066 				break;
6067 			}
6068 	}
6069 	items[0] = &natl;
6070 	sizes[0] = sizeof(natl);
6071 	types[0] = 0;
6072 
6073 	(void) ipf_log_items(softc, IPL_LOGNAT, NULL, items, sizes, types, 1);
6074 #endif
6075 }
6076 
6077 
6078 
6079 
6080 /* ------------------------------------------------------------------------ */
6081 /* Function:    ipf_nat_rule_deref                                          */
6082 /* Returns:     Nil                                                         */
6083 /* Parameters:  softc(I) - pointer to soft context main structure           */
6084 /*              inp(I)   - pointer to pointer to NAT rule                   */
6085 /* Write Locks: ipf_nat                                                     */
6086 /*                                                                          */
6087 /* Dropping the refernce count for a rule means that whatever held the      */
6088 /* pointer to this rule (*inp) is no longer interested in it and when the   */
6089 /* reference count drops to zero, any resources allocated for the rule can  */
6090 /* be released and the rule itself free'd.                                  */
6091 /* ------------------------------------------------------------------------ */
6092 void
ipf_nat_rule_deref(ipf_main_softc_t * softc,ipnat_t ** inp)6093 ipf_nat_rule_deref(ipf_main_softc_t *softc, ipnat_t **inp)
6094 {
6095 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6096 	ipnat_t *n;
6097 
6098 	n = *inp;
6099 	*inp = NULL;
6100 	n->in_use--;
6101 	if (n->in_use > 0)
6102 		return;
6103 
6104 	if (n->in_apr != NULL)
6105 		ipf_proxy_deref(n->in_apr);
6106 
6107 	ipf_nat_rule_fini(softc, n);
6108 
6109 	if (n->in_redir & NAT_REDIRECT) {
6110 		if ((n->in_flags & IPN_PROXYRULE) == 0) {
6111 			ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules_rdr);
6112 		}
6113 	}
6114 	if (n->in_redir & (NAT_MAP|NAT_MAPBLK)) {
6115 		if ((n->in_flags & IPN_PROXYRULE) == 0) {
6116 			ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules_map);
6117 		}
6118 	}
6119 
6120 	if (n->in_tqehead[0] != NULL) {
6121 		if (ipf_deletetimeoutqueue(n->in_tqehead[0]) == 0) {
6122 			ipf_freetimeoutqueue(softc, n->in_tqehead[0]);
6123 		}
6124 	}
6125 
6126 	if (n->in_tqehead[1] != NULL) {
6127 		if (ipf_deletetimeoutqueue(n->in_tqehead[1]) == 0) {
6128 			ipf_freetimeoutqueue(softc, n->in_tqehead[1]);
6129 		}
6130 	}
6131 
6132 	if ((n->in_flags & IPN_PROXYRULE) == 0) {
6133 		ATOMIC_DEC32(softn->ipf_nat_stats.ns_rules);
6134 	}
6135 
6136 	MUTEX_DESTROY(&n->in_lock);
6137 
6138 	KFREES(n, n->in_size);
6139 
6140 #if SOLARIS && !defined(INSTANCES)
6141 	if (softn->ipf_nat_stats.ns_rules == 0)
6142 		pfil_delayed_copy = 1;
6143 #endif
6144 }
6145 
6146 
6147 /* ------------------------------------------------------------------------ */
6148 /* Function:    ipf_nat_deref                                               */
6149 /* Returns:     Nil                                                         */
6150 /* Parameters:  softc(I) - pointer to soft context main structure           */
6151 /*              natp(I)  - pointer to pointer to NAT table entry            */
6152 /*                                                                          */
6153 /* Decrement the reference counter for this NAT table entry and free it if  */
6154 /* there are no more things using it.                                       */
6155 /*                                                                          */
6156 /* IF nat_ref == 1 when this function is called, then we have an orphan nat */
6157 /* structure *because* it only gets called on paths _after_ nat_ref has been*/
6158 /* incremented.  If nat_ref == 1 then we shouldn't decrement it here        */
6159 /* because nat_delete() will do that and send nat_ref to -1.                */
6160 /*                                                                          */
6161 /* Holding the lock on nat_lock is required to serialise nat_delete() being */
6162 /* called from a NAT flush ioctl with a deref happening because of a packet.*/
6163 /* ------------------------------------------------------------------------ */
6164 void
ipf_nat_deref(ipf_main_softc_t * softc,nat_t ** natp)6165 ipf_nat_deref(ipf_main_softc_t *softc, nat_t **natp)
6166 {
6167 	nat_t *nat;
6168 
6169 	nat = *natp;
6170 	*natp = NULL;
6171 
6172 	MUTEX_ENTER(&nat->nat_lock);
6173 	if (nat->nat_ref > 1) {
6174 		nat->nat_ref--;
6175 		ASSERT(nat->nat_ref >= 0);
6176 		MUTEX_EXIT(&nat->nat_lock);
6177 		return;
6178 	}
6179 	MUTEX_EXIT(&nat->nat_lock);
6180 
6181 	WRITE_ENTER(&softc->ipf_nat);
6182 	ipf_nat_delete(softc, nat, NL_EXPIRE);
6183 	RWLOCK_EXIT(&softc->ipf_nat);
6184 }
6185 
6186 
6187 /* ------------------------------------------------------------------------ */
6188 /* Function:    ipf_nat_clone                                               */
6189 /* Returns:     ipstate_t* - NULL == cloning failed,                        */
6190 /*                           else pointer to new state structure            */
6191 /* Parameters:  fin(I) - pointer to packet information                      */
6192 /*              is(I)  - pointer to master state structure                  */
6193 /* Write Lock:  ipf_nat                                                     */
6194 /*                                                                          */
6195 /* Create a "duplcate" state table entry from the master.                   */
6196 /* ------------------------------------------------------------------------ */
6197 nat_t *
ipf_nat_clone(fr_info_t * fin,nat_t * nat)6198 ipf_nat_clone(fr_info_t *fin, nat_t *nat)
6199 {
6200 	ipf_main_softc_t *softc = fin->fin_main_soft;
6201 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6202 	frentry_t *fr;
6203 	nat_t *clone;
6204 	ipnat_t *np;
6205 
6206 	KMALLOC(clone, nat_t *);
6207 	if (clone == NULL) {
6208 		NBUMPSIDED(fin->fin_out, ns_clone_nomem);
6209 		return (NULL);
6210 	}
6211 	bcopy((char *)nat, (char *)clone, sizeof(*clone));
6212 
6213 	MUTEX_NUKE(&clone->nat_lock);
6214 
6215 	clone->nat_rev = fin->fin_rev;
6216 	clone->nat_aps = NULL;
6217 	/*
6218 	 * Initialize all these so that ipf_nat_delete() doesn't cause a crash.
6219 	 */
6220 	clone->nat_tqe.tqe_pnext = NULL;
6221 	clone->nat_tqe.tqe_next = NULL;
6222 	clone->nat_tqe.tqe_ifq = NULL;
6223 	clone->nat_tqe.tqe_parent = clone;
6224 
6225 	clone->nat_flags &= ~SI_CLONE;
6226 	clone->nat_flags |= SI_CLONED;
6227 
6228 	if (clone->nat_hm)
6229 		clone->nat_hm->hm_ref++;
6230 
6231 	if (ipf_nat_insert(softc, softn, clone) == -1) {
6232 		KFREE(clone);
6233 		NBUMPSIDED(fin->fin_out, ns_insert_fail);
6234 		return (NULL);
6235 	}
6236 
6237 	np = clone->nat_ptr;
6238 	if (np != NULL) {
6239 		if (softn->ipf_nat_logging)
6240 			ipf_nat_log(softc, softn, clone, NL_CLONE);
6241 		np->in_use++;
6242 	}
6243 	fr = clone->nat_fr;
6244 	if (fr != NULL) {
6245 		MUTEX_ENTER(&fr->fr_lock);
6246 		fr->fr_ref++;
6247 		MUTEX_EXIT(&fr->fr_lock);
6248 	}
6249 
6250 
6251 	/*
6252 	 * Because the clone is created outside the normal loop of things and
6253 	 * TCP has special needs in terms of state, initialise the timeout
6254 	 * state of the new NAT from here.
6255 	 */
6256 	if (clone->nat_pr[0] == IPPROTO_TCP) {
6257 		(void) ipf_tcp_age(&clone->nat_tqe, fin, softn->ipf_nat_tcptq,
6258 				   clone->nat_flags, 2);
6259 	}
6260 	clone->nat_sync = ipf_sync_new(softc, SMC_NAT, fin, clone);
6261 	if (softn->ipf_nat_logging)
6262 		ipf_nat_log(softc, softn, clone, NL_CLONE);
6263 	return (clone);
6264 }
6265 
6266 
6267 /* ------------------------------------------------------------------------ */
6268 /* Function:   ipf_nat_wildok                                               */
6269 /* Returns:    int - 1 == packet's ports match wildcards                    */
6270 /*                   0 == packet's ports don't match wildcards              */
6271 /* Parameters: nat(I)   - NAT entry                                         */
6272 /*             sport(I) - source port                                       */
6273 /*             dport(I) - destination port                                  */
6274 /*             flags(I) - wildcard flags                                    */
6275 /*             dir(I)   - packet direction                                  */
6276 /*                                                                          */
6277 /* Use NAT entry and packet direction to determine which combination of     */
6278 /* wildcard flags should be used.                                           */
6279 /* ------------------------------------------------------------------------ */
6280 int
ipf_nat_wildok(nat_t * nat,int sport,int dport,int flags,int dir)6281 ipf_nat_wildok(nat_t *nat, int sport, int dport, int flags, int dir)
6282 {
6283 	/*
6284 	 * When called by       dir is set to
6285 	 * nat_inlookup         NAT_INBOUND (0)
6286 	 * nat_outlookup        NAT_OUTBOUND (1)
6287 	 *
6288 	 * We simply combine the packet's direction in dir with the original
6289 	 * "intended" direction of that NAT entry in nat->nat_dir to decide
6290 	 * which combination of wildcard flags to allow.
6291 	 */
6292 	switch ((dir << 1) | (nat->nat_dir & (NAT_INBOUND|NAT_OUTBOUND)))
6293 	{
6294 	case 3: /* outbound packet / outbound entry */
6295 		if (((nat->nat_osport == sport) ||
6296 		    (flags & SI_W_SPORT)) &&
6297 		    ((nat->nat_odport == dport) ||
6298 		    (flags & SI_W_DPORT)))
6299 			return (1);
6300 		break;
6301 	case 2: /* outbound packet / inbound entry */
6302 		if (((nat->nat_osport == dport) ||
6303 		    (flags & SI_W_SPORT)) &&
6304 		    ((nat->nat_odport == sport) ||
6305 		    (flags & SI_W_DPORT)))
6306 			return (1);
6307 		break;
6308 	case 1: /* inbound packet / outbound entry */
6309 		if (((nat->nat_osport == dport) ||
6310 		    (flags & SI_W_SPORT)) &&
6311 		    ((nat->nat_odport == sport) ||
6312 		    (flags & SI_W_DPORT)))
6313 			return (1);
6314 		break;
6315 	case 0: /* inbound packet / inbound entry */
6316 		if (((nat->nat_osport == sport) ||
6317 		    (flags & SI_W_SPORT)) &&
6318 		    ((nat->nat_odport == dport) ||
6319 		    (flags & SI_W_DPORT)))
6320 			return (1);
6321 		break;
6322 	default:
6323 		break;
6324 	}
6325 
6326 	return (0);
6327 }
6328 
6329 
6330 /* ------------------------------------------------------------------------ */
6331 /* Function:    nat_mssclamp                                                */
6332 /* Returns:     Nil                                                         */
6333 /* Parameters:  tcp(I)    - pointer to TCP header                           */
6334 /*              maxmss(I) - value to clamp the TCP MSS to                   */
6335 /*              fin(I)    - pointer to packet information                   */
6336 /*              csump(I)  - pointer to TCP checksum                         */
6337 /*                                                                          */
6338 /* Check for MSS option and clamp it if necessary.  If found and changed,   */
6339 /* then the TCP header checksum will be updated to reflect the change in    */
6340 /* the MSS.                                                                 */
6341 /* ------------------------------------------------------------------------ */
6342 static void
ipf_nat_mssclamp(tcphdr_t * tcp,u_32_t maxmss,fr_info_t * fin,u_short * csump)6343 ipf_nat_mssclamp(tcphdr_t *tcp, u_32_t maxmss, fr_info_t *fin, u_short *csump)
6344 {
6345 	u_char *cp, *ep, opt;
6346 	int hlen, advance;
6347 	u_32_t mss, sumd;
6348 
6349 	hlen = TCP_OFF(tcp) << 2;
6350 	if (hlen > sizeof(*tcp)) {
6351 		cp = (u_char *)tcp + sizeof(*tcp);
6352 		ep = (u_char *)tcp + hlen;
6353 
6354 		while (cp < ep) {
6355 			opt = cp[0];
6356 			if (opt == TCPOPT_EOL)
6357 				break;
6358 			else if (opt == TCPOPT_NOP) {
6359 				cp++;
6360 				continue;
6361 			}
6362 
6363 			if (cp + 1 >= ep)
6364 				break;
6365 			advance = cp[1];
6366 			if ((cp + advance > ep) || (advance <= 0))
6367 				break;
6368 			switch (opt)
6369 			{
6370 			case TCPOPT_MAXSEG:
6371 				if (advance != 4)
6372 					break;
6373 				mss = cp[2] * 256 + cp[3];
6374 				if (mss > maxmss) {
6375 					cp[2] = maxmss / 256;
6376 					cp[3] = maxmss & 0xff;
6377 					CALC_SUMD(mss, maxmss, sumd);
6378 					ipf_fix_outcksum(0, csump, sumd, 0);
6379 				}
6380 				break;
6381 			default:
6382 				/* ignore unknown options */
6383 				break;
6384 			}
6385 
6386 			cp += advance;
6387 		}
6388 	}
6389 }
6390 
6391 
6392 /* ------------------------------------------------------------------------ */
6393 /* Function:    ipf_nat_setqueue                                            */
6394 /* Returns:     Nil                                                         */
6395 /* Parameters:  softc(I) - pointer to soft context main structure           */
6396 /*              softn(I) - pointer to NAT context structure                 */
6397 /*              nat(I)- pointer to NAT structure                            */
6398 /* Locks:       ipf_nat (read or write)                                     */
6399 /*                                                                          */
6400 /* Put the NAT entry on its default queue entry, using rev as a helped in   */
6401 /* determining which queue it should be placed on.                          */
6402 /* ------------------------------------------------------------------------ */
6403 void
ipf_nat_setqueue(ipf_main_softc_t * softc,ipf_nat_softc_t * softn,nat_t * nat)6404 ipf_nat_setqueue(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, nat_t *nat)
6405 {
6406 	ipftq_t *oifq, *nifq;
6407 	int rev = nat->nat_rev;
6408 
6409 	if (nat->nat_ptr != NULL)
6410 		nifq = nat->nat_ptr->in_tqehead[rev];
6411 	else
6412 		nifq = NULL;
6413 
6414 	if (nifq == NULL) {
6415 		switch (nat->nat_pr[0])
6416 		{
6417 		case IPPROTO_UDP :
6418 			nifq = &softn->ipf_nat_udptq;
6419 			break;
6420 		case IPPROTO_ICMP :
6421 			nifq = &softn->ipf_nat_icmptq;
6422 			break;
6423 		case IPPROTO_TCP :
6424 			nifq = softn->ipf_nat_tcptq +
6425 			       nat->nat_tqe.tqe_state[rev];
6426 			break;
6427 		default :
6428 			nifq = &softn->ipf_nat_iptq;
6429 			break;
6430 		}
6431 	}
6432 
6433 	oifq = nat->nat_tqe.tqe_ifq;
6434 	/*
6435 	 * If it's currently on a timeout queue, move it from one queue to
6436 	 * another, else put it on the end of the newly determined queue.
6437 	 */
6438 	if (oifq != NULL)
6439 		ipf_movequeue(softc->ipf_ticks, &nat->nat_tqe, oifq, nifq);
6440 	else
6441 		ipf_queueappend(softc->ipf_ticks, &nat->nat_tqe, nifq, nat);
6442 	return;
6443 }
6444 
6445 
6446 /* ------------------------------------------------------------------------ */
6447 /* Function:    nat_getnext                                                 */
6448 /* Returns:     int - 0 == ok, else error                                   */
6449 /* Parameters:  softc(I) - pointer to soft context main structure           */
6450 /*              t(I)   - pointer to ipftoken structure                      */
6451 /*              itp(I) - pointer to ipfgeniter_t structure                  */
6452 /*                                                                          */
6453 /* Fetch the next nat/ipnat structure pointer from the linked list and      */
6454 /* copy it out to the storage space pointed to by itp_data.  The next item  */
6455 /* in the list to look at is put back in the ipftoken struture.             */
6456 /* ------------------------------------------------------------------------ */
6457 static int
ipf_nat_getnext(ipf_main_softc_t * softc,ipftoken_t * t,ipfgeniter_t * itp,ipfobj_t * objp)6458 ipf_nat_getnext(ipf_main_softc_t *softc, ipftoken_t *t, ipfgeniter_t *itp,
6459 	ipfobj_t *objp)
6460 {
6461 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6462 	hostmap_t *hm, *nexthm = NULL, zerohm;
6463 	ipnat_t *ipn, *nextipnat = NULL, zeroipn;
6464 	nat_t *nat, *nextnat = NULL, zeronat;
6465 	int error = 0;
6466 	void *nnext;
6467 
6468 	if (itp->igi_nitems != 1) {
6469 		IPFERROR(60075);
6470 		return (ENOSPC);
6471 	}
6472 
6473 	READ_ENTER(&softc->ipf_nat);
6474 
6475 	switch (itp->igi_type)
6476 	{
6477 	case IPFGENITER_HOSTMAP :
6478 		hm = t->ipt_data;
6479 		if (hm == NULL) {
6480 			nexthm = softn->ipf_hm_maplist;
6481 		} else {
6482 			nexthm = hm->hm_next;
6483 		}
6484 		if (nexthm != NULL) {
6485 			ATOMIC_INC32(nexthm->hm_ref);
6486 			t->ipt_data = nexthm;
6487 		} else {
6488 			bzero(&zerohm, sizeof(zerohm));
6489 			nexthm = &zerohm;
6490 			t->ipt_data = NULL;
6491 		}
6492 		nnext = nexthm->hm_next;
6493 		break;
6494 
6495 	case IPFGENITER_IPNAT :
6496 		ipn = t->ipt_data;
6497 		if (ipn == NULL) {
6498 			nextipnat = softn->ipf_nat_list;
6499 		} else {
6500 			nextipnat = ipn->in_next;
6501 		}
6502 		if (nextipnat != NULL) {
6503 			ATOMIC_INC32(nextipnat->in_use);
6504 			t->ipt_data = nextipnat;
6505 		} else {
6506 			bzero(&zeroipn, sizeof(zeroipn));
6507 			nextipnat = &zeroipn;
6508 			t->ipt_data = NULL;
6509 		}
6510 		nnext = nextipnat->in_next;
6511 		break;
6512 
6513 	case IPFGENITER_NAT :
6514 		nat = t->ipt_data;
6515 		if (nat == NULL) {
6516 			nextnat = softn->ipf_nat_instances;
6517 		} else {
6518 			nextnat = nat->nat_next;
6519 		}
6520 		if (nextnat != NULL) {
6521 			MUTEX_ENTER(&nextnat->nat_lock);
6522 			nextnat->nat_ref++;
6523 			MUTEX_EXIT(&nextnat->nat_lock);
6524 			t->ipt_data = nextnat;
6525 		} else {
6526 			bzero(&zeronat, sizeof(zeronat));
6527 			nextnat = &zeronat;
6528 			t->ipt_data = NULL;
6529 		}
6530 		nnext = nextnat->nat_next;
6531 		break;
6532 
6533 	default :
6534 		RWLOCK_EXIT(&softc->ipf_nat);
6535 		IPFERROR(60055);
6536 		return (EINVAL);
6537 	}
6538 
6539 	RWLOCK_EXIT(&softc->ipf_nat);
6540 
6541 	objp->ipfo_ptr = itp->igi_data;
6542 
6543 	switch (itp->igi_type)
6544 	{
6545 	case IPFGENITER_HOSTMAP :
6546 		error = COPYOUT(nexthm, objp->ipfo_ptr, sizeof(*nexthm));
6547 		if (error != 0) {
6548 			IPFERROR(60049);
6549 			error = EFAULT;
6550 		}
6551 		if (hm != NULL) {
6552 			WRITE_ENTER(&softc->ipf_nat);
6553 			ipf_nat_hostmapdel(softc, &hm);
6554 			RWLOCK_EXIT(&softc->ipf_nat);
6555 		}
6556 		break;
6557 
6558 	case IPFGENITER_IPNAT :
6559 		objp->ipfo_size = nextipnat->in_size;
6560 		objp->ipfo_type = IPFOBJ_IPNAT;
6561 		error = ipf_outobjk(softc, objp, nextipnat);
6562 		if (ipn != NULL) {
6563 			WRITE_ENTER(&softc->ipf_nat);
6564 			ipf_nat_rule_deref(softc, &ipn);
6565 			RWLOCK_EXIT(&softc->ipf_nat);
6566 		}
6567 		break;
6568 
6569 	case IPFGENITER_NAT :
6570 		objp->ipfo_size = sizeof(nat_t);
6571 		objp->ipfo_type = IPFOBJ_NAT;
6572 		error = ipf_outobjk(softc, objp, nextnat);
6573 		if (nat != NULL)
6574 			ipf_nat_deref(softc, &nat);
6575 
6576 		break;
6577 	}
6578 
6579 	if (nnext == NULL)
6580 		ipf_token_mark_complete(t);
6581 
6582 	return (error);
6583 }
6584 
6585 
6586 /* ------------------------------------------------------------------------ */
6587 /* Function:    nat_extraflush                                              */
6588 /* Returns:     int - 0 == success, -1 == failure                           */
6589 /* Parameters:  softc(I) - pointer to soft context main structure           */
6590 /*              softn(I) - pointer to NAT context structure                 */
6591 /*              which(I) - how to flush the active NAT table                */
6592 /* Write Locks: ipf_nat                                                     */
6593 /*                                                                          */
6594 /* Flush nat tables.  Three actions currently defined:                      */
6595 /* which == 0 : flush all nat table entries                                 */
6596 /* which == 1 : flush TCP connections which have started to close but are   */
6597 /*	      stuck for some reason.                                        */
6598 /* which == 2 : flush TCP connections which have been idle for a long time, */
6599 /*	      starting at > 4 days idle and working back in successive half-*/
6600 /*	      days to at most 12 hours old.  If this fails to free enough   */
6601 /*            slots then work backwards in half hour slots to 30 minutes.   */
6602 /*            If that too fails, then work backwards in 30 second intervals */
6603 /*            for the last 30 minutes to at worst 30 seconds idle.          */
6604 /* ------------------------------------------------------------------------ */
6605 static int
ipf_nat_extraflush(ipf_main_softc_t * softc,ipf_nat_softc_t * softn,int which)6606 ipf_nat_extraflush(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, int which)
6607 {
6608 	nat_t *nat, **natp;
6609 	ipftqent_t *tqn;
6610 	ipftq_t *ifq;
6611 	int removed;
6612 	SPL_INT(s);
6613 
6614 	removed = 0;
6615 
6616 	SPL_NET(s);
6617 	switch (which)
6618 	{
6619 	case 0 :
6620 		softn->ipf_nat_stats.ns_flush_all++;
6621 		/*
6622 		 * Style 0 flush removes everything...
6623 		 */
6624 		for (natp = &softn->ipf_nat_instances;
6625 		     ((nat = *natp) != NULL); ) {
6626 			ipf_nat_delete(softc, nat, NL_FLUSH);
6627 			removed++;
6628 		}
6629 		break;
6630 
6631 	case 1 :
6632 		softn->ipf_nat_stats.ns_flush_closing++;
6633 		/*
6634 		 * Since we're only interested in things that are closing,
6635 		 * we can start with the appropriate timeout queue.
6636 		 */
6637 		for (ifq = softn->ipf_nat_tcptq + IPF_TCPS_CLOSE_WAIT;
6638 		     ifq != NULL; ifq = ifq->ifq_next) {
6639 
6640 			for (tqn = ifq->ifq_head; tqn != NULL; ) {
6641 				nat = tqn->tqe_parent;
6642 				tqn = tqn->tqe_next;
6643 				if (nat->nat_pr[0] != IPPROTO_TCP ||
6644 				    nat->nat_pr[1] != IPPROTO_TCP)
6645 					break;
6646 				ipf_nat_delete(softc, nat, NL_EXPIRE);
6647 				removed++;
6648 			}
6649 		}
6650 
6651 		/*
6652 		 * Also need to look through the user defined queues.
6653 		 */
6654 		for (ifq = softn->ipf_nat_utqe; ifq != NULL;
6655 		     ifq = ifq->ifq_next) {
6656 			for (tqn = ifq->ifq_head; tqn != NULL; ) {
6657 				nat = tqn->tqe_parent;
6658 				tqn = tqn->tqe_next;
6659 				if (nat->nat_pr[0] != IPPROTO_TCP ||
6660 				    nat->nat_pr[1] != IPPROTO_TCP)
6661 					continue;
6662 
6663 				if ((nat->nat_tcpstate[0] >
6664 				     IPF_TCPS_ESTABLISHED) &&
6665 				    (nat->nat_tcpstate[1] >
6666 				     IPF_TCPS_ESTABLISHED)) {
6667 					ipf_nat_delete(softc, nat, NL_EXPIRE);
6668 					removed++;
6669 				}
6670 			}
6671 		}
6672 		break;
6673 
6674 		/*
6675 		 * Args 5-11 correspond to flushing those particular states
6676 		 * for TCP connections.
6677 		 */
6678 	case IPF_TCPS_CLOSE_WAIT :
6679 	case IPF_TCPS_FIN_WAIT_1 :
6680 	case IPF_TCPS_CLOSING :
6681 	case IPF_TCPS_LAST_ACK :
6682 	case IPF_TCPS_FIN_WAIT_2 :
6683 	case IPF_TCPS_TIME_WAIT :
6684 	case IPF_TCPS_CLOSED :
6685 		softn->ipf_nat_stats.ns_flush_state++;
6686 		tqn = softn->ipf_nat_tcptq[which].ifq_head;
6687 		while (tqn != NULL) {
6688 			nat = tqn->tqe_parent;
6689 			tqn = tqn->tqe_next;
6690 			ipf_nat_delete(softc, nat, NL_FLUSH);
6691 			removed++;
6692 		}
6693 		break;
6694 
6695 	default :
6696 		if (which < 30)
6697 			break;
6698 
6699 		softn->ipf_nat_stats.ns_flush_timeout++;
6700 		/*
6701 		 * Take a large arbitrary number to mean the number of seconds
6702 		 * for which which consider to be the maximum value we'll allow
6703 		 * the expiration to be.
6704 		 */
6705 		which = IPF_TTLVAL(which);
6706 		for (natp = &softn->ipf_nat_instances;
6707 		     ((nat = *natp) != NULL); ) {
6708 			if (softc->ipf_ticks - nat->nat_touched > which) {
6709 				ipf_nat_delete(softc, nat, NL_FLUSH);
6710 				removed++;
6711 			} else
6712 				natp = &nat->nat_next;
6713 		}
6714 		break;
6715 	}
6716 
6717 	if (which != 2) {
6718 		SPL_X(s);
6719 		return (removed);
6720 	}
6721 
6722 	softn->ipf_nat_stats.ns_flush_queue++;
6723 
6724 	/*
6725 	 * Asked to remove inactive entries because the table is full, try
6726 	 * again, 3 times, if first attempt failed with a different criteria
6727 	 * each time.  The order tried in must be in decreasing age.
6728 	 * Another alternative is to implement random drop and drop N entries
6729 	 * at random until N have been freed up.
6730 	 */
6731 	if (softc->ipf_ticks - softn->ipf_nat_last_force_flush >
6732 	    IPF_TTLVAL(5)) {
6733 		softn->ipf_nat_last_force_flush = softc->ipf_ticks;
6734 
6735 		removed = ipf_queueflush(softc, ipf_nat_flush_entry,
6736 					 softn->ipf_nat_tcptq,
6737 					 softn->ipf_nat_utqe,
6738 					 &softn->ipf_nat_stats.ns_active,
6739 					 softn->ipf_nat_table_sz,
6740 					 softn->ipf_nat_table_wm_low);
6741 	}
6742 
6743 	SPL_X(s);
6744 	return (removed);
6745 }
6746 
6747 
6748 /* ------------------------------------------------------------------------ */
6749 /* Function:    ipf_nat_flush_entry                                         */
6750 /* Returns:     0 - always succeeds                                         */
6751 /* Parameters:  softc(I) - pointer to soft context main structure           */
6752 /*              entry(I) - pointer to NAT entry                             */
6753 /* Write Locks: ipf_nat                                                     */
6754 /*                                                                          */
6755 /* This function is a stepping stone between ipf_queueflush() and           */
6756 /* nat_dlete().  It is used so we can provide a uniform interface via the   */
6757 /* ipf_queueflush() function.  Since the nat_delete() function returns void */
6758 /* we translate that to mean it always succeeds in deleting something.      */
6759 /* ------------------------------------------------------------------------ */
6760 static int
ipf_nat_flush_entry(ipf_main_softc_t * softc,void * entry)6761 ipf_nat_flush_entry(ipf_main_softc_t *softc, void *entry)
6762 {
6763 	ipf_nat_delete(softc, entry, NL_FLUSH);
6764 	return (0);
6765 }
6766 
6767 
6768 /* ------------------------------------------------------------------------ */
6769 /* Function:    ipf_nat_iterator                                            */
6770 /* Returns:     int - 0 == ok, else error                                   */
6771 /* Parameters:  softc(I) - pointer to soft context main structure           */
6772 /*              token(I) - pointer to ipftoken structure                    */
6773 /*              itp(I)   - pointer to ipfgeniter_t structure                */
6774 /*              obj(I)   - pointer to data description structure            */
6775 /*                                                                          */
6776 /* This function acts as a handler for the SIOCGENITER ioctls that use a    */
6777 /* generic structure to iterate through a list.  There are three different  */
6778 /* linked lists of NAT related information to go through: NAT rules, active */
6779 /* NAT mappings and the NAT fragment cache.                                 */
6780 /* ------------------------------------------------------------------------ */
6781 static int
ipf_nat_iterator(ipf_main_softc_t * softc,ipftoken_t * token,ipfgeniter_t * itp,ipfobj_t * obj)6782 ipf_nat_iterator(ipf_main_softc_t *softc, ipftoken_t *token, ipfgeniter_t *itp,
6783 	ipfobj_t *obj)
6784 {
6785 	int error;
6786 
6787 	if (itp->igi_data == NULL) {
6788 		IPFERROR(60052);
6789 		return (EFAULT);
6790 	}
6791 
6792 	switch (itp->igi_type)
6793 	{
6794 	case IPFGENITER_HOSTMAP :
6795 	case IPFGENITER_IPNAT :
6796 	case IPFGENITER_NAT :
6797 		error = ipf_nat_getnext(softc, token, itp, obj);
6798 		break;
6799 
6800 	case IPFGENITER_NATFRAG :
6801 		error = ipf_frag_nat_next(softc, token, itp);
6802 		break;
6803 	default :
6804 		IPFERROR(60053);
6805 		error = EINVAL;
6806 		break;
6807 	}
6808 
6809 	return (error);
6810 }
6811 
6812 
6813 /* ------------------------------------------------------------------------ */
6814 /* Function:    ipf_nat_setpending                                          */
6815 /* Returns:     Nil                                                         */
6816 /* Parameters:  softc(I) - pointer to soft context main structure           */
6817 /*              nat(I)   - pointer to NAT structure                         */
6818 /* Locks:       ipf_nat (read or write)                                     */
6819 /*                                                                          */
6820 /* Put the NAT entry on to the pending queue - this queue has a very short  */
6821 /* lifetime where items are put that can't be deleted straight away because */
6822 /* of locking issues but we want to delete them ASAP, anyway.  In calling   */
6823 /* this function, it is assumed that the owner (if there is one, as shown   */
6824 /* by nat_me) is no longer interested in it.                                */
6825 /* ------------------------------------------------------------------------ */
6826 void
ipf_nat_setpending(ipf_main_softc_t * softc,nat_t * nat)6827 ipf_nat_setpending(ipf_main_softc_t *softc, nat_t *nat)
6828 {
6829 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
6830 	ipftq_t *oifq;
6831 
6832 	oifq = nat->nat_tqe.tqe_ifq;
6833 	if (oifq != NULL)
6834 		ipf_movequeue(softc->ipf_ticks, &nat->nat_tqe, oifq,
6835 			      &softn->ipf_nat_pending);
6836 	else
6837 		ipf_queueappend(softc->ipf_ticks, &nat->nat_tqe,
6838 				&softn->ipf_nat_pending, nat);
6839 
6840 	if (nat->nat_me != NULL) {
6841 		*nat->nat_me = NULL;
6842 		nat->nat_me = NULL;
6843 		nat->nat_ref--;
6844 		ASSERT(nat->nat_ref >= 0);
6845 	}
6846 }
6847 
6848 
6849 /* ------------------------------------------------------------------------ */
6850 /* Function:    nat_newrewrite                                              */
6851 /* Returns:     int - -1 == error, 0 == success (no move), 1 == success and */
6852 /*                    allow rule to be moved if IPN_ROUNDR is set.          */
6853 /* Parameters:  fin(I) - pointer to packet information                      */
6854 /*              nat(I) - pointer to NAT entry                               */
6855 /*              ni(I)  - pointer to structure with misc. information needed */
6856 /*                       to create new NAT entry.                           */
6857 /* Write Lock:  ipf_nat                                                     */
6858 /*                                                                          */
6859 /* This function is responsible for setting up an active NAT session where  */
6860 /* we are changing both the source and destination parameters at the same   */
6861 /* time.  The loop in here works differently to elsewhere - each iteration  */
6862 /* is responsible for changing a single parameter that can be incremented.  */
6863 /* So one pass may increase the source IP#, next source port, next dest. IP#*/
6864 /* and the last destination port for a total of 4 iterations to try each.   */
6865 /* This is done to try and exhaustively use the translation space available.*/
6866 /* ------------------------------------------------------------------------ */
6867 static int
ipf_nat_newrewrite(fr_info_t * fin,nat_t * nat,natinfo_t * nai)6868 ipf_nat_newrewrite(fr_info_t *fin, nat_t *nat, natinfo_t *nai)
6869 {
6870 	int src_search = 1;
6871 	int dst_search = 1;
6872 	fr_info_t frnat;
6873 	u_32_t flags;
6874 	u_short swap;
6875 	ipnat_t *np;
6876 	nat_t *natl;
6877 	int l = 0;
6878 	int changed;
6879 
6880 	natl = NULL;
6881 	changed = -1;
6882 	np = nai->nai_np;
6883 	flags = nat->nat_flags;
6884 	bcopy((char *)fin, (char *)&frnat, sizeof(*fin));
6885 
6886 	nat->nat_hm = NULL;
6887 
6888 	do {
6889 		changed = -1;
6890 		/* TRACE (l, src_search, dst_search, np) */
6891 		DT4(ipf_nat_rewrite_1, int, l, int, src_search, int, dst_search, ipnat_t *, np);
6892 
6893 		if ((src_search == 0) && (np->in_spnext == 0) &&
6894 		    (dst_search == 0) && (np->in_dpnext == 0)) {
6895 			if (l > 0)
6896 				return (-1);
6897 		}
6898 
6899 		/*
6900 		 * Find a new source address
6901 		 */
6902 		if (ipf_nat_nextaddr(fin, &np->in_nsrc, &frnat.fin_saddr,
6903 				     &frnat.fin_saddr) == -1) {
6904 			return (-1);
6905 		}
6906 
6907 		if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0xffffffff)) {
6908 			src_search = 0;
6909 			if (np->in_stepnext == 0)
6910 				np->in_stepnext = 1;
6911 
6912 		} else if ((np->in_nsrcaddr == 0) && (np->in_nsrcmsk == 0)) {
6913 			src_search = 0;
6914 			if (np->in_stepnext == 0)
6915 				np->in_stepnext = 1;
6916 
6917 		} else if (np->in_nsrcmsk == 0xffffffff) {
6918 			src_search = 0;
6919 			if (np->in_stepnext == 0)
6920 				np->in_stepnext = 1;
6921 
6922 		} else if (np->in_nsrcmsk != 0xffffffff) {
6923 			if (np->in_stepnext == 0 && changed == -1) {
6924 				np->in_snip++;
6925 				np->in_stepnext++;
6926 				changed = 0;
6927 			}
6928 		}
6929 
6930 		if ((flags & IPN_TCPUDPICMP) != 0) {
6931 			if (np->in_spnext != 0)
6932 				frnat.fin_data[0] = np->in_spnext;
6933 
6934 			/*
6935 			 * Standard port translation.  Select next port.
6936 			 */
6937 			if ((flags & IPN_FIXEDSPORT) != 0) {
6938 				np->in_stepnext = 2;
6939 			} else if ((np->in_stepnext == 1) &&
6940 				   (changed == -1) && (natl != NULL)) {
6941 				np->in_spnext++;
6942 				np->in_stepnext++;
6943 				changed = 1;
6944 				if (np->in_spnext > np->in_spmax)
6945 					np->in_spnext = np->in_spmin;
6946 			}
6947 		} else {
6948 			np->in_stepnext = 2;
6949 		}
6950 		np->in_stepnext &= 0x3;
6951 
6952 		/*
6953 		 * Find a new destination address
6954 		 */
6955 		/* TRACE (fin, np, l, frnat) */
6956 		DT4(ipf_nat_rewrite_2, frinfo_t *, fin, ipnat_t *, np, int, l, frinfo_t *, &frnat);
6957 
6958 		if (ipf_nat_nextaddr(fin, &np->in_ndst, &frnat.fin_daddr,
6959 				     &frnat.fin_daddr) == -1)
6960 			return (-1);
6961 		if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0xffffffff)) {
6962 			dst_search = 0;
6963 			if (np->in_stepnext == 2)
6964 				np->in_stepnext = 3;
6965 
6966 		} else if ((np->in_ndstaddr == 0) && (np->in_ndstmsk == 0)) {
6967 			dst_search = 0;
6968 			if (np->in_stepnext == 2)
6969 				np->in_stepnext = 3;
6970 
6971 		} else if (np->in_ndstmsk == 0xffffffff) {
6972 			dst_search = 0;
6973 			if (np->in_stepnext == 2)
6974 				np->in_stepnext = 3;
6975 
6976 		} else if (np->in_ndstmsk != 0xffffffff) {
6977 			if ((np->in_stepnext == 2) && (changed == -1) &&
6978 			    (natl != NULL)) {
6979 				changed = 2;
6980 				np->in_stepnext++;
6981 				np->in_dnip++;
6982 			}
6983 		}
6984 
6985 		if ((flags & IPN_TCPUDPICMP) != 0) {
6986 			if (np->in_dpnext != 0)
6987 				frnat.fin_data[1] = np->in_dpnext;
6988 
6989 			/*
6990 			 * Standard port translation.  Select next port.
6991 			 */
6992 			if ((flags & IPN_FIXEDDPORT) != 0) {
6993 				np->in_stepnext = 0;
6994 			} else if (np->in_stepnext == 3 && changed == -1) {
6995 				np->in_dpnext++;
6996 				np->in_stepnext++;
6997 				changed = 3;
6998 				if (np->in_dpnext > np->in_dpmax)
6999 					np->in_dpnext = np->in_dpmin;
7000 			}
7001 		} else {
7002 			if (np->in_stepnext == 3)
7003 				np->in_stepnext = 0;
7004 		}
7005 
7006 		/* TRACE (frnat) */
7007 		DT1(ipf_nat_rewrite_3, frinfo_t *, &frnat);
7008 
7009 		/*
7010 		 * Here we do a lookup of the connection as seen from
7011 		 * the outside.  If an IP# pair already exists, try
7012 		 * again.  So if you have A->B becomes C->B, you can
7013 		 * also have D->E become C->E but not D->B causing
7014 		 * another C->B.  Also take protocol and ports into
7015 		 * account when determining whether a pre-existing
7016 		 * NAT setup will cause an external conflict where
7017 		 * this is appropriate.
7018 		 *
7019 		 * fin_data[] is swapped around because we are doing a
7020 		 * lookup of the packet is if it were moving in the opposite
7021 		 * direction of the one we are working with now.
7022 		 */
7023 		if (flags & IPN_TCPUDP) {
7024 			swap = frnat.fin_data[0];
7025 			frnat.fin_data[0] = frnat.fin_data[1];
7026 			frnat.fin_data[1] = swap;
7027 		}
7028 		if (fin->fin_out == 1) {
7029 			natl = ipf_nat_inlookup(&frnat,
7030 						flags & ~(SI_WILDP|NAT_SEARCH),
7031 						(u_int)frnat.fin_p,
7032 						frnat.fin_dst, frnat.fin_src);
7033 
7034 		} else {
7035 			natl = ipf_nat_outlookup(&frnat,
7036 						 flags & ~(SI_WILDP|NAT_SEARCH),
7037 						 (u_int)frnat.fin_p,
7038 						 frnat.fin_dst, frnat.fin_src);
7039 		}
7040 		if (flags & IPN_TCPUDP) {
7041 			swap = frnat.fin_data[0];
7042 			frnat.fin_data[0] = frnat.fin_data[1];
7043 			frnat.fin_data[1] = swap;
7044 		}
7045 
7046 		/* TRACE natl, in_stepnext, l */
7047 		DT3(ipf_nat_rewrite_2, nat_t *, natl, ipnat_t *, np , int, l);
7048 
7049 		if ((natl != NULL) && (l > 8))	/* XXX 8 is arbitrary */
7050 			return (-1);
7051 
7052 		np->in_stepnext &= 0x3;
7053 
7054 		l++;
7055 		changed = -1;
7056 	} while (natl != NULL);
7057 
7058 	nat->nat_osrcip = fin->fin_src;
7059 	nat->nat_odstip = fin->fin_dst;
7060 	nat->nat_nsrcip = frnat.fin_src;
7061 	nat->nat_ndstip = frnat.fin_dst;
7062 
7063 	if ((flags & IPN_TCPUDP) != 0) {
7064 		nat->nat_osport = htons(fin->fin_data[0]);
7065 		nat->nat_odport = htons(fin->fin_data[1]);
7066 		nat->nat_nsport = htons(frnat.fin_data[0]);
7067 		nat->nat_ndport = htons(frnat.fin_data[1]);
7068 	} else if ((flags & IPN_ICMPQUERY) != 0) {
7069 		nat->nat_oicmpid = fin->fin_data[1];
7070 		nat->nat_nicmpid = frnat.fin_data[1];
7071 	}
7072 
7073 	return (0);
7074 }
7075 
7076 
7077 /* ------------------------------------------------------------------------ */
7078 /* Function:    nat_newdivert                                               */
7079 /* Returns:     int - -1 == error, 0 == success                             */
7080 /* Parameters:  fin(I) - pointer to packet information                      */
7081 /*              nat(I) - pointer to NAT entry                               */
7082 /*              ni(I)  - pointer to structure with misc. information needed */
7083 /*                       to create new NAT entry.                           */
7084 /* Write Lock:  ipf_nat                                                     */
7085 /*                                                                          */
7086 /* Create a new NAT  divert session as defined by the NAT rule.  This is    */
7087 /* somewhat different to other NAT session creation routines because we     */
7088 /* do not iterate through either port numbers or IP addresses, searching    */
7089 /* for a unique mapping, however, a complimentary duplicate check is made.  */
7090 /* ------------------------------------------------------------------------ */
7091 static int
ipf_nat_newdivert(fr_info_t * fin,nat_t * nat,natinfo_t * nai)7092 ipf_nat_newdivert(fr_info_t *fin, nat_t *nat, natinfo_t *nai)
7093 {
7094 	ipf_main_softc_t *softc = fin->fin_main_soft;
7095 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7096 	fr_info_t frnat;
7097 	ipnat_t *np;
7098 	nat_t *natl;
7099 	int p;
7100 
7101 	np = nai->nai_np;
7102 	bcopy((char *)fin, (char *)&frnat, sizeof(*fin));
7103 
7104 	nat->nat_pr[0] = 0;
7105 	nat->nat_osrcaddr = fin->fin_saddr;
7106 	nat->nat_odstaddr = fin->fin_daddr;
7107 	frnat.fin_saddr = htonl(np->in_snip);
7108 	frnat.fin_daddr = htonl(np->in_dnip);
7109 	if ((nat->nat_flags & IPN_TCPUDP) != 0) {
7110 		nat->nat_osport = htons(fin->fin_data[0]);
7111 		nat->nat_odport = htons(fin->fin_data[1]);
7112 	} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
7113 		nat->nat_oicmpid = fin->fin_data[1];
7114 	}
7115 
7116 	if (np->in_redir & NAT_DIVERTUDP) {
7117 		frnat.fin_data[0] = np->in_spnext;
7118 		frnat.fin_data[1] = np->in_dpnext;
7119 		frnat.fin_flx |= FI_TCPUDP;
7120 		p = IPPROTO_UDP;
7121 	} else {
7122 		frnat.fin_flx &= ~FI_TCPUDP;
7123 		p = IPPROTO_IPIP;
7124 	}
7125 
7126 	if (fin->fin_out == 1) {
7127 		natl = ipf_nat_inlookup(&frnat, 0, p,
7128 					frnat.fin_dst, frnat.fin_src);
7129 
7130 	} else {
7131 		natl = ipf_nat_outlookup(&frnat, 0, p,
7132 					 frnat.fin_dst, frnat.fin_src);
7133 	}
7134 
7135 	if (natl != NULL) {
7136 		NBUMPSIDED(fin->fin_out, ns_divert_exist);
7137 		DT3(ns_divert_exist, fr_info_t *, fin, nat_t *, nat, natinfo_t, nai);
7138 		return (-1);
7139 	}
7140 
7141 	nat->nat_nsrcaddr = frnat.fin_saddr;
7142 	nat->nat_ndstaddr = frnat.fin_daddr;
7143 	if ((nat->nat_flags & IPN_TCPUDP) != 0) {
7144 		nat->nat_nsport = htons(frnat.fin_data[0]);
7145 		nat->nat_ndport = htons(frnat.fin_data[1]);
7146 	} else if ((nat->nat_flags & IPN_ICMPQUERY) != 0) {
7147 		nat->nat_nicmpid = frnat.fin_data[1];
7148 	}
7149 
7150 	nat->nat_pr[fin->fin_out] = fin->fin_p;
7151 	nat->nat_pr[1 - fin->fin_out] = p;
7152 
7153 	if (np->in_redir & NAT_REDIRECT)
7154 		nat->nat_dir = NAT_DIVERTIN;
7155 	else
7156 		nat->nat_dir = NAT_DIVERTOUT;
7157 
7158 	return (0);
7159 }
7160 
7161 
7162 /* ------------------------------------------------------------------------ */
7163 /* Function:    nat_builddivertmp                                           */
7164 /* Returns:     int - -1 == error, 0 == success                             */
7165 /* Parameters:  softn(I) - pointer to NAT context structure                 */
7166 /*              np(I)    - pointer to a NAT rule                            */
7167 /*                                                                          */
7168 /* For divert rules, a skeleton packet representing what will be prepended  */
7169 /* to the real packet is created.  Even though we don't have the full       */
7170 /* packet here, a checksum is calculated that we update later when we       */
7171 /* fill in the final details.  At present a 0 checksum for UDP is being set */
7172 /* here because it is expected that divert will be used for localhost.      */
7173 /* ------------------------------------------------------------------------ */
7174 static int
ipf_nat_builddivertmp(ipf_nat_softc_t * softn,ipnat_t * np)7175 ipf_nat_builddivertmp(ipf_nat_softc_t *softn, ipnat_t *np)
7176 {
7177 	udphdr_t *uh;
7178 	size_t len;
7179 	ip_t *ip;
7180 
7181 	if ((np->in_redir & NAT_DIVERTUDP) != 0)
7182 		len = sizeof(ip_t) + sizeof(udphdr_t);
7183 	else
7184 		len = sizeof(ip_t);
7185 
7186 	ALLOC_MB_T(np->in_divmp, len);
7187 	if (np->in_divmp == NULL) {
7188 		NBUMPD(ipf_nat_stats, ns_divert_build);
7189 		return (-1);
7190 	}
7191 
7192 	/*
7193 	 * First, the header to get the packet diverted to the new destination
7194 	 */
7195 	ip = MTOD(np->in_divmp, ip_t *);
7196 	IP_V_A(ip, 4);
7197 	IP_HL_A(ip, 5);
7198 	ip->ip_tos = 0;
7199 	if ((np->in_redir & NAT_DIVERTUDP) != 0)
7200 		ip->ip_p = IPPROTO_UDP;
7201 	else
7202 		ip->ip_p = IPPROTO_IPIP;
7203 	ip->ip_ttl = 255;
7204 	ip->ip_off = 0;
7205 	ip->ip_sum = 0;
7206 	ip->ip_len = htons(len);
7207 	ip->ip_id = 0;
7208 	ip->ip_src.s_addr = htonl(np->in_snip);
7209 	ip->ip_dst.s_addr = htonl(np->in_dnip);
7210 	ip->ip_sum = ipf_cksum((u_short *)ip, sizeof(*ip));
7211 
7212 	if (np->in_redir & NAT_DIVERTUDP) {
7213 		uh = (udphdr_t *)(ip + 1);
7214 		uh->uh_sum = 0;
7215 		uh->uh_ulen = 8;
7216 		uh->uh_sport = htons(np->in_spnext);
7217 		uh->uh_dport = htons(np->in_dpnext);
7218 	}
7219 
7220 	return (0);
7221 }
7222 
7223 
7224 #define	MINDECAP	(sizeof(ip_t) + sizeof(udphdr_t) + sizeof(ip_t))
7225 
7226 /* ------------------------------------------------------------------------ */
7227 /* Function:    nat_decap                                                   */
7228 /* Returns:     int - -1 == error, 0 == success                             */
7229 /* Parameters:  fin(I) - pointer to packet information                      */
7230 /*              nat(I) - pointer to current NAT session                     */
7231 /*                                                                          */
7232 /* This function is responsible for undoing a packet's encapsulation in the */
7233 /* reverse of an encap/divert rule.  After removing the outer encapsulation */
7234 /* it is necessary to call ipf_makefrip() again so that the contents of 'fin'*/
7235 /* match the "new" packet as it may still be used by IPFilter elsewhere.    */
7236 /* We use "dir" here as the basis for some of the expectations about the    */
7237 /* outer header.  If we return an error, the goal is to leave the original  */
7238 /* packet information undisturbed - this falls short at the end where we'd  */
7239 /* need to back a backup copy of "fin" - expensive.                         */
7240 /* ------------------------------------------------------------------------ */
7241 static int
ipf_nat_decap(fr_info_t * fin,nat_t * nat)7242 ipf_nat_decap(fr_info_t *fin, nat_t *nat)
7243 {
7244 	ipf_main_softc_t *softc = fin->fin_main_soft;
7245 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7246 	char *hdr;
7247 	int hlen;
7248 	int skip;
7249 	mb_t *m;
7250 
7251 	if ((fin->fin_flx & FI_ICMPERR) != 0) {
7252 		/*
7253 		 * ICMP packets don't get decapsulated, instead what we need
7254 		 * to do is change the ICMP reply from including (in the data
7255 		 * portion for errors) the encapsulated packet that we sent
7256 		 * out to something that resembles the original packet prior
7257 		 * to encapsulation.  This isn't done here - all we're doing
7258 		 * here is changing the outer address to ensure that it gets
7259 		 * targetted back to the correct system.
7260 		 */
7261 
7262 		if (nat->nat_dir & NAT_OUTBOUND) {
7263 			u_32_t sum1, sum2, sumd;
7264 
7265 			sum1 = ntohl(fin->fin_daddr);
7266 			sum2 = ntohl(nat->nat_osrcaddr);
7267 			CALC_SUMD(sum1, sum2, sumd);
7268 			fin->fin_ip->ip_dst = nat->nat_osrcip;
7269 			fin->fin_daddr = nat->nat_osrcaddr;
7270 #if !defined(_KERNEL) || SOLARIS
7271 			ipf_fix_outcksum(0, &fin->fin_ip->ip_sum, sumd, 0);
7272 #endif
7273 		}
7274 		return (0);
7275 	}
7276 
7277 	m = fin->fin_m;
7278 	skip = fin->fin_hlen;
7279 
7280 	switch (nat->nat_dir)
7281 	{
7282 	case NAT_DIVERTIN :
7283 	case NAT_DIVERTOUT :
7284 		if (fin->fin_plen < MINDECAP)
7285 			return (-1);
7286 		skip += sizeof(udphdr_t);
7287 		break;
7288 
7289 	case NAT_ENCAPIN :
7290 	case NAT_ENCAPOUT :
7291 		if (fin->fin_plen < (skip + sizeof(ip_t)))
7292 			return (-1);
7293 		break;
7294 	default :
7295 		return (-1);
7296 		/* NOTREACHED */
7297 	}
7298 
7299 	/*
7300 	 * The aim here is to keep the original packet details in "fin" for
7301 	 * as long as possible so that returning with an error is for the
7302 	 * original packet and there is little undoing work to do.
7303 	 */
7304 	if (M_LEN(m) < skip + sizeof(ip_t)) {
7305 		if (ipf_pr_pullup(fin, skip + sizeof(ip_t)) == -1)
7306 			return (-1);
7307 	}
7308 
7309 	hdr = MTOD(fin->fin_m, char *);
7310 	fin->fin_ip = (ip_t *)(hdr + skip);
7311 	hlen = IP_HL(fin->fin_ip) << 2;
7312 
7313 	if (ipf_pr_pullup(fin, skip + hlen) == -1) {
7314 		NBUMPSIDED(fin->fin_out, ns_decap_pullup);
7315 		return (-1);
7316 	}
7317 
7318 	fin->fin_hlen = hlen;
7319 	fin->fin_dlen -= skip;
7320 	fin->fin_plen -= skip;
7321 	fin->fin_ipoff += skip;
7322 
7323 	if (ipf_makefrip(hlen, (ip_t *)hdr, fin) == -1) {
7324 		NBUMPSIDED(fin->fin_out, ns_decap_bad);
7325 		return (-1);
7326 	}
7327 
7328 	return (skip);
7329 }
7330 
7331 
7332 /* ------------------------------------------------------------------------ */
7333 /* Function:    nat_nextaddr                                                */
7334 /* Returns:     int - -1 == bad input (no new address),                     */
7335 /*                     0 == success and dst has new address                 */
7336 /* Parameters:  fin(I) - pointer to packet information                      */
7337 /*              na(I)  - how to generate new address                        */
7338 /*              old(I) - original address being replaced                    */
7339 /*              dst(O) - where to put the new address                       */
7340 /* Write Lock:  ipf_nat                                                     */
7341 /*                                                                          */
7342 /* This function uses the contents of the "na" structure, in combination    */
7343 /* with "old" to produce a new address to store in "dst".  Not all of the   */
7344 /* possible uses of "na" will result in a new address.                      */
7345 /* ------------------------------------------------------------------------ */
7346 static int
ipf_nat_nextaddr(fr_info_t * fin,nat_addr_t * na,u_32_t * old,u_32_t * dst)7347 ipf_nat_nextaddr(fr_info_t *fin, nat_addr_t *na, u_32_t *old, u_32_t *dst)
7348 {
7349 	ipf_main_softc_t *softc = fin->fin_main_soft;
7350 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7351 	u_32_t new;
7352 	i6addr_t newip;
7353 	int error;
7354 
7355 	new = 0;
7356 
7357 	switch (na->na_atype)
7358 	{
7359 	case FRI_RANGE :
7360 	case FRI_NETMASKED :
7361 	case FRI_DYNAMIC :
7362 	case FRI_NORMAL :
7363 	case FRI_LOOKUP :
7364 		break;
7365 
7366 	case FRI_BROADCAST :
7367 	case FRI_PEERADDR :
7368 	case FRI_NETWORK :
7369 	default :
7370 		DT4(ns_na_atype, fr_info_t *, fin, nat_addr_t *, na, u_32_t *, old, u_32_t *, new);
7371 		return (-1);
7372 	}
7373 
7374 	error = -1;
7375 
7376 	if (na->na_atype == FRI_LOOKUP) {
7377 		if (na->na_type == IPLT_DSTLIST) {
7378 			error = ipf_dstlist_select_node(fin, na->na_ptr, dst,
7379 							NULL);
7380 		} else {
7381 			NBUMPSIDE(fin->fin_out, ns_badnextaddr);
7382 			DT4(ns_badnextaddr_1, fr_info_t *, fin, nat_addr_t *, na, u_32_t *, old, u_32_t *, new);
7383 		}
7384 
7385 	} else if (na->na_atype == IPLT_NONE) {
7386 		/*
7387 		 * 0/0 as the new address means leave it alone.
7388 		 */
7389 		if (na->na_addr[0].in4.s_addr == 0 &&
7390 		    na->na_addr[1].in4.s_addr == 0) {
7391 			new = *old;
7392 
7393 		/*
7394 		 * 0/32 means get the interface's address
7395 		 */
7396 		} else if (na->na_addr[0].in4.s_addr == 0 &&
7397 			   na->na_addr[1].in4.s_addr == 0xffffffff) {
7398 			if (ipf_ifpaddr(softc, 4, na->na_atype,
7399 					fin->fin_ifp, &newip, NULL) == -1) {
7400 				NBUMPSIDED(fin->fin_out, ns_ifpaddrfail);
7401 				DT4(ns_ifpaddrfail, fr_info_t *, fin, nat_addr_t *, na, u_32_t *, old, u_32_t *, new);
7402 				return (-1);
7403 			}
7404 			new = newip.in4.s_addr;
7405 		} else {
7406 			new = htonl(na->na_nextip);
7407 		}
7408 		*dst = new;
7409 		error = 0;
7410 
7411 	} else {
7412 		NBUMPSIDE(fin->fin_out, ns_badnextaddr);
7413 		DT4(ns_badnextaddr_2, fr_info_t *, fin, nat_addr_t *, na, u_32_t *, old, u_32_t *, new);
7414 	}
7415 
7416 	return (error);
7417 }
7418 
7419 
7420 /* ------------------------------------------------------------------------ */
7421 /* Function:    nat_nextaddrinit                                            */
7422 /* Returns:     int - 0 == success, else error number                       */
7423 /* Parameters:  softc(I) - pointer to soft context main structure           */
7424 /*              na(I)      - NAT address information for generating new addr*/
7425 /*              initial(I) - flag indicating if it is the first call for    */
7426 /*                           this "na" structure.                           */
7427 /*              ifp(I)     - network interface to derive address            */
7428 /*                           information from.                              */
7429 /*                                                                          */
7430 /* This function is expected to be called in two scenarious: when a new NAT */
7431 /* rule is loaded into the kernel and when the list of NAT rules is sync'd  */
7432 /* up with the valid network interfaces (possibly due to them changing.)    */
7433 /* To distinguish between these, the "initial" parameter is used.  If it is */
7434 /* 1 then this indicates the rule has just been reloaded and 0 for when we  */
7435 /* are updating information.  This difference is important because in       */
7436 /* instances where we are not updating address information associated with  */
7437 /* a network interface, we don't want to disturb what the "next" address to */
7438 /* come out of ipf_nat_nextaddr() will be.                                  */
7439 /* ------------------------------------------------------------------------ */
7440 static int
ipf_nat_nextaddrinit(ipf_main_softc_t * softc,char * base,nat_addr_t * na,int initial,void * ifp)7441 ipf_nat_nextaddrinit(ipf_main_softc_t *softc, char *base, nat_addr_t *na,
7442 	int initial, void *ifp)
7443 {
7444 
7445 	switch (na->na_atype)
7446 	{
7447 	case FRI_LOOKUP :
7448 		if (na->na_subtype == 0) {
7449 			na->na_ptr = ipf_lookup_res_num(softc, IPL_LOGNAT,
7450 							na->na_type,
7451 							na->na_num,
7452 							&na->na_func);
7453 		} else if (na->na_subtype == 1) {
7454 			na->na_ptr = ipf_lookup_res_name(softc, IPL_LOGNAT,
7455 							 na->na_type,
7456 							 base + na->na_num,
7457 							 &na->na_func);
7458 		}
7459 		if (na->na_func == NULL) {
7460 			IPFERROR(60060);
7461 			return (ESRCH);
7462 		}
7463 		if (na->na_ptr == NULL) {
7464 			IPFERROR(60056);
7465 			return (ESRCH);
7466 		}
7467 		break;
7468 
7469 	case FRI_DYNAMIC :
7470 	case FRI_BROADCAST :
7471 	case FRI_NETWORK :
7472 	case FRI_NETMASKED :
7473 	case FRI_PEERADDR :
7474 		if (ifp != NULL)
7475 			(void )ipf_ifpaddr(softc, 4, na->na_atype, ifp,
7476 					   &na->na_addr[0], &na->na_addr[1]);
7477 		break;
7478 
7479 	case FRI_SPLIT :
7480 	case FRI_RANGE :
7481 		if (initial)
7482 			na->na_nextip = ntohl(na->na_addr[0].in4.s_addr);
7483 		break;
7484 
7485 	case FRI_NONE :
7486 		na->na_addr[0].in4.s_addr &= na->na_addr[1].in4.s_addr;
7487 		return (0);
7488 
7489 	case FRI_NORMAL :
7490 		na->na_addr[0].in4.s_addr &= na->na_addr[1].in4.s_addr;
7491 		break;
7492 
7493 	default :
7494 		IPFERROR(60054);
7495 		return (EINVAL);
7496 	}
7497 
7498 	if (initial && (na->na_atype == FRI_NORMAL)) {
7499 		if (na->na_addr[0].in4.s_addr == 0) {
7500 			if ((na->na_addr[1].in4.s_addr == 0xffffffff) ||
7501 			    (na->na_addr[1].in4.s_addr == 0)) {
7502 				return (0);
7503 			}
7504 		}
7505 
7506 		if (na->na_addr[1].in4.s_addr == 0xffffffff) {
7507 			na->na_nextip = ntohl(na->na_addr[0].in4.s_addr);
7508 		} else {
7509 			na->na_nextip = ntohl(na->na_addr[0].in4.s_addr) + 1;
7510 		}
7511 	}
7512 
7513 	return (0);
7514 }
7515 
7516 
7517 /* ------------------------------------------------------------------------ */
7518 /* Function:    ipf_nat_matchflush                                          */
7519 /* Returns:     int - -1 == error, 0 == success                             */
7520 /* Parameters:  softc(I) - pointer to soft context main structure           */
7521 /*              softn(I) - pointer to NAT context structure                 */
7522 /*              nat(I)   - pointer to current NAT session                   */
7523 /*                                                                          */
7524 /* ------------------------------------------------------------------------ */
7525 static int
ipf_nat_matchflush(ipf_main_softc_t * softc,ipf_nat_softc_t * softn,caddr_t data)7526 ipf_nat_matchflush(ipf_main_softc_t *softc, ipf_nat_softc_t *softn,
7527 	caddr_t data)
7528 {
7529 	int *array, flushed, error;
7530 	nat_t *nat, *natnext;
7531 	ipfobj_t obj;
7532 
7533 	error = ipf_matcharray_load(softc, data, &obj, &array);
7534 	if (error != 0)
7535 		return (error);
7536 
7537 	flushed = 0;
7538 
7539 	for (nat = softn->ipf_nat_instances; nat != NULL; nat = natnext) {
7540 		natnext = nat->nat_next;
7541 		if (ipf_nat_matcharray(nat, array, softc->ipf_ticks) == 0) {
7542 			ipf_nat_delete(softc, nat, NL_FLUSH);
7543 			flushed++;
7544 		}
7545 	}
7546 
7547 	obj.ipfo_retval = flushed;
7548 	error = BCOPYOUT(&obj, data, sizeof(obj));
7549 
7550 	KFREES(array, array[0] * sizeof(*array));
7551 
7552 	return (error);
7553 }
7554 
7555 
7556 /* ------------------------------------------------------------------------ */
7557 /* Function:    ipf_nat_matcharray                                          */
7558 /* Returns:     int - -1 == error, 0 == success                             */
7559 /* Parameters:  fin(I) - pointer to packet information                      */
7560 /*              nat(I) - pointer to current NAT session                     */
7561 /*                                                                          */
7562 /* ------------------------------------------------------------------------ */
7563 static int
ipf_nat_matcharray(nat_t * nat,int * array,u_long ticks)7564 ipf_nat_matcharray(nat_t *nat, int *array, u_long ticks)
7565 {
7566 	int i, n, *x, e, p;
7567 
7568 	e = 0;
7569 	n = array[0];
7570 	x = array + 1;
7571 
7572 	for (; n > 0; x += 3 + x[2]) {
7573 		if (x[0] == IPF_EXP_END)
7574 			break;
7575 		e = 0;
7576 
7577 		n -= x[2] + 3;
7578 		if (n < 0)
7579 			break;
7580 
7581 		p = x[0] >> 16;
7582 		if (p != 0 && p != nat->nat_pr[1])
7583 			break;
7584 
7585 		switch (x[0])
7586 		{
7587 		case IPF_EXP_IP_PR :
7588 			for (i = 0; !e && i < x[2]; i++) {
7589 				e |= (nat->nat_pr[1] == x[i + 3]);
7590 			}
7591 			break;
7592 
7593 		case IPF_EXP_IP_SRCADDR :
7594 			if (nat->nat_v[0] == 4) {
7595 				for (i = 0; !e && i < x[2]; i++) {
7596 					e |= ((nat->nat_osrcaddr & x[i + 4]) ==
7597 					      x[i + 3]);
7598 				}
7599 			}
7600 			if (nat->nat_v[1] == 4) {
7601 				for (i = 0; !e && i < x[2]; i++) {
7602 					e |= ((nat->nat_nsrcaddr & x[i + 4]) ==
7603 					      x[i + 3]);
7604 				}
7605 			}
7606 			break;
7607 
7608 		case IPF_EXP_IP_DSTADDR :
7609 			if (nat->nat_v[0] == 4) {
7610 				for (i = 0; !e && i < x[2]; i++) {
7611 					e |= ((nat->nat_odstaddr & x[i + 4]) ==
7612 					      x[i + 3]);
7613 				}
7614 			}
7615 			if (nat->nat_v[1] == 4) {
7616 				for (i = 0; !e && i < x[2]; i++) {
7617 					e |= ((nat->nat_ndstaddr & x[i + 4]) ==
7618 					      x[i + 3]);
7619 				}
7620 			}
7621 			break;
7622 
7623 		case IPF_EXP_IP_ADDR :
7624 			for (i = 0; !e && i < x[2]; i++) {
7625 				if (nat->nat_v[0] == 4) {
7626 					e |= ((nat->nat_osrcaddr & x[i + 4]) ==
7627 					      x[i + 3]);
7628 				}
7629 				if (nat->nat_v[1] == 4) {
7630 					e |= ((nat->nat_nsrcaddr & x[i + 4]) ==
7631 					      x[i + 3]);
7632 				}
7633 				if (nat->nat_v[0] == 4) {
7634 					e |= ((nat->nat_odstaddr & x[i + 4]) ==
7635 					      x[i + 3]);
7636 				}
7637 				if (nat->nat_v[1] == 4) {
7638 					e |= ((nat->nat_ndstaddr & x[i + 4]) ==
7639 					      x[i + 3]);
7640 				}
7641 			}
7642 			break;
7643 
7644 #ifdef USE_INET6
7645 		case IPF_EXP_IP6_SRCADDR :
7646 			if (nat->nat_v[0] == 6) {
7647 				for (i = 0; !e && i < x[3]; i++) {
7648 					e |= IP6_MASKEQ(&nat->nat_osrc6,
7649 							x + i + 7, x + i + 3);
7650 				}
7651 			}
7652 			if (nat->nat_v[1] == 6) {
7653 				for (i = 0; !e && i < x[3]; i++) {
7654 					e |= IP6_MASKEQ(&nat->nat_nsrc6,
7655 							x + i + 7, x + i + 3);
7656 				}
7657 			}
7658 			break;
7659 
7660 		case IPF_EXP_IP6_DSTADDR :
7661 			if (nat->nat_v[0] == 6) {
7662 				for (i = 0; !e && i < x[3]; i++) {
7663 					e |= IP6_MASKEQ(&nat->nat_odst6,
7664 							x + i + 7,
7665 							x + i + 3);
7666 				}
7667 			}
7668 			if (nat->nat_v[1] == 6) {
7669 				for (i = 0; !e && i < x[3]; i++) {
7670 					e |= IP6_MASKEQ(&nat->nat_ndst6,
7671 							x + i + 7,
7672 							x + i + 3);
7673 				}
7674 			}
7675 			break;
7676 
7677 		case IPF_EXP_IP6_ADDR :
7678 			for (i = 0; !e && i < x[3]; i++) {
7679 				if (nat->nat_v[0] == 6) {
7680 					e |= IP6_MASKEQ(&nat->nat_osrc6,
7681 							x + i + 7,
7682 							x + i + 3);
7683 				}
7684 				if (nat->nat_v[0] == 6) {
7685 					e |= IP6_MASKEQ(&nat->nat_odst6,
7686 							x + i + 7,
7687 							x + i + 3);
7688 				}
7689 				if (nat->nat_v[1] == 6) {
7690 					e |= IP6_MASKEQ(&nat->nat_nsrc6,
7691 							x + i + 7,
7692 							x + i + 3);
7693 				}
7694 				if (nat->nat_v[1] == 6) {
7695 					e |= IP6_MASKEQ(&nat->nat_ndst6,
7696 							x + i + 7,
7697 							x + i + 3);
7698 				}
7699 			}
7700 			break;
7701 #endif
7702 
7703 		case IPF_EXP_UDP_PORT :
7704 		case IPF_EXP_TCP_PORT :
7705 			for (i = 0; !e && i < x[2]; i++) {
7706 				e |= (nat->nat_nsport == x[i + 3]) ||
7707 				     (nat->nat_ndport == x[i + 3]);
7708 			}
7709 			break;
7710 
7711 		case IPF_EXP_UDP_SPORT :
7712 		case IPF_EXP_TCP_SPORT :
7713 			for (i = 0; !e && i < x[2]; i++) {
7714 				e |= (nat->nat_nsport == x[i + 3]);
7715 			}
7716 			break;
7717 
7718 		case IPF_EXP_UDP_DPORT :
7719 		case IPF_EXP_TCP_DPORT :
7720 			for (i = 0; !e && i < x[2]; i++) {
7721 				e |= (nat->nat_ndport == x[i + 3]);
7722 			}
7723 			break;
7724 
7725 		case IPF_EXP_TCP_STATE :
7726 			for (i = 0; !e && i < x[2]; i++) {
7727 				e |= (nat->nat_tcpstate[0] == x[i + 3]) ||
7728 				     (nat->nat_tcpstate[1] == x[i + 3]);
7729 			}
7730 			break;
7731 
7732 		case IPF_EXP_IDLE_GT :
7733 			e |= (ticks - nat->nat_touched > x[3]);
7734 			break;
7735 		}
7736 		e ^= x[1];
7737 
7738 		if (!e)
7739 			break;
7740 	}
7741 
7742 	return (e);
7743 }
7744 
7745 
7746 /* ------------------------------------------------------------------------ */
7747 /* Function:    ipf_nat_gettable                                            */
7748 /* Returns:     int     - 0 = success, else error                           */
7749 /* Parameters:  softc(I) - pointer to soft context main structure           */
7750 /*              softn(I) - pointer to NAT context structure                 */
7751 /*              data(I)  - pointer to ioctl data                            */
7752 /*                                                                          */
7753 /* This function handles ioctl requests for tables of nat information.      */
7754 /* At present the only table it deals with is the hash bucket statistics.   */
7755 /* ------------------------------------------------------------------------ */
7756 static int
ipf_nat_gettable(ipf_main_softc_t * softc,ipf_nat_softc_t * softn,char * data)7757 ipf_nat_gettable(ipf_main_softc_t *softc, ipf_nat_softc_t *softn, char *data)
7758 {
7759 	ipftable_t table;
7760 	int error;
7761 
7762 	error = ipf_inobj(softc, data, NULL, &table, IPFOBJ_GTABLE);
7763 	if (error != 0)
7764 		return (error);
7765 
7766 	switch (table.ita_type)
7767 	{
7768 	case IPFTABLE_BUCKETS_NATIN :
7769 		error = COPYOUT(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
7770 				table.ita_table,
7771 				softn->ipf_nat_table_sz * sizeof(u_int));
7772 		break;
7773 
7774 	case IPFTABLE_BUCKETS_NATOUT :
7775 		error = COPYOUT(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
7776 				table.ita_table,
7777 				softn->ipf_nat_table_sz * sizeof(u_int));
7778 		break;
7779 
7780 	default :
7781 		IPFERROR(60058);
7782 		return (EINVAL);
7783 	}
7784 
7785 	if (error != 0) {
7786 		IPFERROR(60059);
7787 		error = EFAULT;
7788 	}
7789 	return (error);
7790 }
7791 
7792 
7793 /* ------------------------------------------------------------------------ */
7794 /* Function:    ipf_nat_settimeout                                          */
7795 /* Returns:     int  - 0 = success, else failure			    */
7796 /* Parameters:  softc(I) - pointer to soft context main structure           */
7797 /*              t(I) - pointer to tunable                                   */
7798 /*              p(I) - pointer to new tuning data                           */
7799 /*                                                                          */
7800 /* Apply the timeout change to the NAT timeout queues.                      */
7801 /* ------------------------------------------------------------------------ */
7802 int
ipf_nat_settimeout(struct ipf_main_softc_s * softc,ipftuneable_t * t,ipftuneval_t * p)7803 ipf_nat_settimeout(struct ipf_main_softc_s *softc, ipftuneable_t *t,
7804 	ipftuneval_t *p)
7805 {
7806 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7807 
7808 	if (!strncmp(t->ipft_name, "tcp_", 4))
7809 		return (ipf_settimeout_tcp(t, p, softn->ipf_nat_tcptq));
7810 
7811 	if (!strcmp(t->ipft_name, "udp_timeout")) {
7812 		ipf_apply_timeout(&softn->ipf_nat_udptq, p->ipftu_int);
7813 	} else if (!strcmp(t->ipft_name, "udp_ack_timeout")) {
7814 		ipf_apply_timeout(&softn->ipf_nat_udpacktq, p->ipftu_int);
7815 	} else if (!strcmp(t->ipft_name, "icmp_timeout")) {
7816 		ipf_apply_timeout(&softn->ipf_nat_icmptq, p->ipftu_int);
7817 	} else if (!strcmp(t->ipft_name, "icmp_ack_timeout")) {
7818 		ipf_apply_timeout(&softn->ipf_nat_icmpacktq, p->ipftu_int);
7819 	} else if (!strcmp(t->ipft_name, "ip_timeout")) {
7820 		ipf_apply_timeout(&softn->ipf_nat_iptq, p->ipftu_int);
7821 	} else {
7822 		IPFERROR(60062);
7823 		return (ESRCH);
7824 	}
7825 	return (0);
7826 }
7827 
7828 
7829 /* ------------------------------------------------------------------------ */
7830 /* Function:    ipf_nat_rehash                                              */
7831 /* Returns:     int  - 0 = success, else failure			    */
7832 /* Parameters:  softc(I) - pointer to soft context main structure           */
7833 /*              t(I) - pointer to tunable                                   */
7834 /*              p(I) - pointer to new tuning data                           */
7835 /*                                                                          */
7836 /* To change the size of the basic NAT table, we need to first allocate the */
7837 /* new tables (lest it fails and we've got nowhere to store all of the NAT  */
7838 /* sessions currently active) and then walk through the entire list and     */
7839 /* insert them into the table.  There are two tables here: an inbound one   */
7840 /* and an outbound one.  Each NAT entry goes into each table once.          */
7841 /* ------------------------------------------------------------------------ */
7842 int
ipf_nat_rehash(ipf_main_softc_t * softc,ipftuneable_t * t,ipftuneval_t * p)7843 ipf_nat_rehash(ipf_main_softc_t *softc, ipftuneable_t *t, ipftuneval_t *p)
7844 {
7845 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
7846 	nat_t **newtab[2], *nat, **natp;
7847 	u_int *bucketlens[2];
7848 	u_int maxbucket;
7849 	u_int newsize;
7850 	int error;
7851 	u_int hv;
7852 	int i;
7853 
7854 	newsize = p->ipftu_int;
7855 	/*
7856 	 * In case there is nothing to do...
7857 	 */
7858 	if (newsize == softn->ipf_nat_table_sz)
7859 		return (0);
7860 
7861 	newtab[0] = NULL;
7862 	newtab[1] = NULL;
7863 	bucketlens[0] = NULL;
7864 	bucketlens[1] = NULL;
7865 	/*
7866 	 * 4 tables depend on the NAT table size: the inbound looking table,
7867 	 * the outbound lookup table and the hash chain length for each.
7868 	 */
7869 	KMALLOCS(newtab[0], nat_t **, newsize * sizeof(nat_t *));
7870 	if (newtab[0] == NULL) {
7871 		error = 60063;
7872 		goto badrehash;
7873 	}
7874 
7875 	KMALLOCS(newtab[1], nat_t **, newsize * sizeof(nat_t *));
7876 	if (newtab[1] == NULL) {
7877 		error = 60064;
7878 		goto badrehash;
7879 	}
7880 
7881 	KMALLOCS(bucketlens[0], u_int *, newsize * sizeof(u_int));
7882 	if (bucketlens[0] == NULL) {
7883 		error = 60065;
7884 		goto badrehash;
7885 	}
7886 
7887 	KMALLOCS(bucketlens[1], u_int *, newsize * sizeof(u_int));
7888 	if (bucketlens[1] == NULL) {
7889 		error = 60066;
7890 		goto badrehash;
7891 	}
7892 
7893 	/*
7894 	 * Recalculate the maximum length based on the new size.
7895 	 */
7896 	for (maxbucket = 0, i = newsize; i > 0; i >>= 1)
7897 		maxbucket++;
7898 	maxbucket *= 2;
7899 
7900 	bzero((char *)newtab[0], newsize * sizeof(nat_t *));
7901 	bzero((char *)newtab[1], newsize * sizeof(nat_t *));
7902 	bzero((char *)bucketlens[0], newsize * sizeof(u_int));
7903 	bzero((char *)bucketlens[1], newsize * sizeof(u_int));
7904 
7905 	WRITE_ENTER(&softc->ipf_nat);
7906 
7907 	if (softn->ipf_nat_table[0] != NULL) {
7908 		KFREES(softn->ipf_nat_table[0],
7909 		       softn->ipf_nat_table_sz *
7910 		       sizeof(*softn->ipf_nat_table[0]));
7911 	}
7912 	softn->ipf_nat_table[0] = newtab[0];
7913 
7914 	if (softn->ipf_nat_table[1] != NULL) {
7915 		KFREES(softn->ipf_nat_table[1],
7916 		       softn->ipf_nat_table_sz *
7917 		       sizeof(*softn->ipf_nat_table[1]));
7918 	}
7919 	softn->ipf_nat_table[1] = newtab[1];
7920 
7921 	if (softn->ipf_nat_stats.ns_side[0].ns_bucketlen != NULL) {
7922 		KFREES(softn->ipf_nat_stats.ns_side[0].ns_bucketlen,
7923 		       softn->ipf_nat_table_sz * sizeof(u_int));
7924 	}
7925 	softn->ipf_nat_stats.ns_side[0].ns_bucketlen = bucketlens[0];
7926 
7927 	if (softn->ipf_nat_stats.ns_side[1].ns_bucketlen != NULL) {
7928 		KFREES(softn->ipf_nat_stats.ns_side[1].ns_bucketlen,
7929 		       softn->ipf_nat_table_sz * sizeof(u_int));
7930 	}
7931 	softn->ipf_nat_stats.ns_side[1].ns_bucketlen = bucketlens[1];
7932 
7933 #ifdef USE_INET6
7934 	if (softn->ipf_nat_stats.ns_side6[0].ns_bucketlen != NULL) {
7935 		KFREES(softn->ipf_nat_stats.ns_side6[0].ns_bucketlen,
7936 		       softn->ipf_nat_table_sz * sizeof(u_int));
7937 	}
7938 	softn->ipf_nat_stats.ns_side6[0].ns_bucketlen = bucketlens[0];
7939 
7940 	if (softn->ipf_nat_stats.ns_side6[1].ns_bucketlen != NULL) {
7941 		KFREES(softn->ipf_nat_stats.ns_side6[1].ns_bucketlen,
7942 		       softn->ipf_nat_table_sz * sizeof(u_int));
7943 	}
7944 	softn->ipf_nat_stats.ns_side6[1].ns_bucketlen = bucketlens[1];
7945 #endif
7946 
7947 	softn->ipf_nat_maxbucket = maxbucket;
7948 	softn->ipf_nat_table_sz = newsize;
7949 	/*
7950 	 * Walk through the entire list of NAT table entries and put them
7951 	 * in the new NAT table, somewhere.  Because we have a new table,
7952 	 * we need to restart the counter of how many chains are in use.
7953 	 */
7954 	softn->ipf_nat_stats.ns_side[0].ns_inuse = 0;
7955 	softn->ipf_nat_stats.ns_side[1].ns_inuse = 0;
7956 #ifdef USE_INET6
7957 	softn->ipf_nat_stats.ns_side6[0].ns_inuse = 0;
7958 	softn->ipf_nat_stats.ns_side6[1].ns_inuse = 0;
7959 #endif
7960 
7961 	for (nat = softn->ipf_nat_instances; nat != NULL; nat = nat->nat_next) {
7962 		nat->nat_hnext[0] = NULL;
7963 		nat->nat_phnext[0] = NULL;
7964 		hv = nat->nat_hv[0] % softn->ipf_nat_table_sz;
7965 
7966 		natp = &softn->ipf_nat_table[0][hv];
7967 		if (*natp) {
7968 			(*natp)->nat_phnext[0] = &nat->nat_hnext[0];
7969 		} else {
7970 			NBUMPSIDE(0, ns_inuse);
7971 		}
7972 		nat->nat_phnext[0] = natp;
7973 		nat->nat_hnext[0] = *natp;
7974 		*natp = nat;
7975 		NBUMPSIDE(0, ns_bucketlen[hv]);
7976 
7977 		nat->nat_hnext[1] = NULL;
7978 		nat->nat_phnext[1] = NULL;
7979 		hv = nat->nat_hv[1] % softn->ipf_nat_table_sz;
7980 
7981 		natp = &softn->ipf_nat_table[1][hv];
7982 		if (*natp) {
7983 			(*natp)->nat_phnext[1] = &nat->nat_hnext[1];
7984 		} else {
7985 			NBUMPSIDE(1, ns_inuse);
7986 		}
7987 		nat->nat_phnext[1] = natp;
7988 		nat->nat_hnext[1] = *natp;
7989 		*natp = nat;
7990 		NBUMPSIDE(1, ns_bucketlen[hv]);
7991 	}
7992 	RWLOCK_EXIT(&softc->ipf_nat);
7993 
7994 	return (0);
7995 
7996 badrehash:
7997 	if (bucketlens[1] != NULL) {
7998 		KFREES(bucketlens[0], newsize * sizeof(u_int));
7999 	}
8000 	if (bucketlens[0] != NULL) {
8001 		KFREES(bucketlens[0], newsize * sizeof(u_int));
8002 	}
8003 	if (newtab[0] != NULL) {
8004 		KFREES(newtab[0], newsize * sizeof(nat_t *));
8005 	}
8006 	if (newtab[1] != NULL) {
8007 		KFREES(newtab[1], newsize * sizeof(nat_t *));
8008 	}
8009 	IPFERROR(error);
8010 	return (ENOMEM);
8011 }
8012 
8013 
8014 /* ------------------------------------------------------------------------ */
8015 /* Function:    ipf_nat_rehash_rules                                        */
8016 /* Returns:     int  - 0 = success, else failure			    */
8017 /* Parameters:  softc(I) - pointer to soft context main structure           */
8018 /*              t(I) - pointer to tunable                                   */
8019 /*              p(I) - pointer to new tuning data                           */
8020 /*                                                                          */
8021 /* All of the NAT rules hang off of a hash table that is searched with a    */
8022 /* hash on address after the netmask is applied.  There is a different table*/
8023 /* for both inbound rules (rdr) and outbound (map.)  The resizing will only */
8024 /* affect one of these two tables.                                          */
8025 /* ------------------------------------------------------------------------ */
8026 int
ipf_nat_rehash_rules(ipf_main_softc_t * softc,ipftuneable_t * t,ipftuneval_t * p)8027 ipf_nat_rehash_rules(ipf_main_softc_t *softc, ipftuneable_t *t,
8028 	ipftuneval_t *p)
8029 {
8030 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8031 	ipnat_t **newtab, *np, ***old, **npp;
8032 	u_int newsize;
8033 	u_int mask;
8034 	u_int hv;
8035 
8036 	newsize = p->ipftu_int;
8037 	/*
8038 	 * In case there is nothing to do...
8039 	 */
8040 	if (newsize == *t->ipft_pint)
8041 		return (0);
8042 
8043 	/*
8044 	 * All inbound rules have the NAT_REDIRECT bit set in in_redir and
8045 	 * all outbound rules have either NAT_MAP or MAT_MAPBLK set.
8046 	 * This if statement allows for some more generic code to be below,
8047 	 * rather than two huge gobs of code that almost do the same thing.
8048 	 */
8049 	if (t->ipft_pint == &softn->ipf_nat_rdrrules_sz) {
8050 		old = &softn->ipf_nat_rdr_rules;
8051 		mask = NAT_REDIRECT;
8052 	} else {
8053 		old = &softn->ipf_nat_map_rules;
8054 		mask = NAT_MAP|NAT_MAPBLK;
8055 	}
8056 
8057 	KMALLOCS(newtab, ipnat_t **, newsize * sizeof(ipnat_t *));
8058 	if (newtab == NULL) {
8059 		IPFERROR(60067);
8060 		return (ENOMEM);
8061 	}
8062 
8063 	bzero((char *)newtab, newsize * sizeof(ipnat_t *));
8064 
8065 	WRITE_ENTER(&softc->ipf_nat);
8066 
8067 	if (*old != NULL) {
8068 		KFREES(*old, *t->ipft_pint * sizeof(ipnat_t **));
8069 	}
8070 	*old = newtab;
8071 	*t->ipft_pint = newsize;
8072 
8073 	for (np = softn->ipf_nat_list; np != NULL; np = np->in_next) {
8074 		if ((np->in_redir & mask) == 0)
8075 			continue;
8076 
8077 		if (np->in_redir & NAT_REDIRECT) {
8078 			np->in_rnext = NULL;
8079 			hv = np->in_hv[0] % newsize;
8080 			for (npp = newtab + hv; *npp != NULL; )
8081 				npp = &(*npp)->in_rnext;
8082 			np->in_prnext = npp;
8083 			*npp = np;
8084 		}
8085 		if (np->in_redir & NAT_MAP) {
8086 			np->in_mnext = NULL;
8087 			hv = np->in_hv[1] % newsize;
8088 			for (npp = newtab + hv; *npp != NULL; )
8089 				npp = &(*npp)->in_mnext;
8090 			np->in_pmnext = npp;
8091 			*npp = np;
8092 		}
8093 
8094 	}
8095 	RWLOCK_EXIT(&softc->ipf_nat);
8096 
8097 	return (0);
8098 }
8099 
8100 
8101 /* ------------------------------------------------------------------------ */
8102 /* Function:    ipf_nat_hostmap_rehash                                      */
8103 /* Returns:     int  - 0 = success, else failure			    */
8104 /* Parameters:  softc(I) - pointer to soft context main structure           */
8105 /*              t(I) - pointer to tunable                                   */
8106 /*              p(I) - pointer to new tuning data                           */
8107 /*                                                                          */
8108 /* Allocate and populate a new hash table that will contain a reference to  */
8109 /* all of the active IP# translations currently in place.                   */
8110 /* ------------------------------------------------------------------------ */
8111 int
ipf_nat_hostmap_rehash(ipf_main_softc_t * softc,ipftuneable_t * t,ipftuneval_t * p)8112 ipf_nat_hostmap_rehash(ipf_main_softc_t *softc, ipftuneable_t *t,
8113 	ipftuneval_t *p)
8114 {
8115 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8116 	hostmap_t *hm, **newtab;
8117 	u_int newsize;
8118 	u_int hv;
8119 
8120 	newsize = p->ipftu_int;
8121 	/*
8122 	 * In case there is nothing to do...
8123 	 */
8124 	if (newsize == *t->ipft_pint)
8125 		return (0);
8126 
8127 	KMALLOCS(newtab, hostmap_t **, newsize * sizeof(hostmap_t *));
8128 	if (newtab == NULL) {
8129 		IPFERROR(60068);
8130 		return (ENOMEM);
8131 	}
8132 
8133 	bzero((char *)newtab, newsize * sizeof(hostmap_t *));
8134 
8135 	WRITE_ENTER(&softc->ipf_nat);
8136 	if (softn->ipf_hm_maptable != NULL) {
8137 		KFREES(softn->ipf_hm_maptable,
8138 		       softn->ipf_nat_hostmap_sz * sizeof(hostmap_t *));
8139 	}
8140 	softn->ipf_hm_maptable = newtab;
8141 	softn->ipf_nat_hostmap_sz = newsize;
8142 
8143 	for (hm = softn->ipf_hm_maplist; hm != NULL; hm = hm->hm_next) {
8144 		hv = hm->hm_hv % softn->ipf_nat_hostmap_sz;
8145 		hm->hm_hnext = softn->ipf_hm_maptable[hv];
8146 		hm->hm_phnext = softn->ipf_hm_maptable + hv;
8147 		if (softn->ipf_hm_maptable[hv] != NULL)
8148 			softn->ipf_hm_maptable[hv]->hm_phnext = &hm->hm_hnext;
8149 		softn->ipf_hm_maptable[hv] = hm;
8150 	}
8151 	RWLOCK_EXIT(&softc->ipf_nat);
8152 
8153 	return (0);
8154 }
8155 
8156 
8157 /* ------------------------------------------------------------------------ */
8158 /* Function:    ipf_nat_add_tq                                              */
8159 /* Parameters:  softc(I) - pointer to soft context main structure           */
8160 /*                                                                          */
8161 /* ------------------------------------------------------------------------ */
8162 ipftq_t *
ipf_nat_add_tq(ipf_main_softc_t * softc,int ttl)8163 ipf_nat_add_tq(ipf_main_softc_t *softc, int ttl)
8164 {
8165 	ipf_nat_softc_t *softs = softc->ipf_nat_soft;
8166 
8167 	return (ipf_addtimeoutqueue(softc, &softs->ipf_nat_utqe, ttl));
8168 }
8169 
8170 /* ------------------------------------------------------------------------ */
8171 /* Function:    ipf_nat_uncreate                                            */
8172 /* Returns:     Nil                                                         */
8173 /* Parameters:  fin(I) - pointer to packet information                      */
8174 /*                                                                          */
8175 /* This function is used to remove a NAT entry from the NAT table when we   */
8176 /* decide that the create was actually in error. It is thus assumed that    */
8177 /* fin_flx will have both FI_NATED and FI_NATNEW set. Because we're dealing */
8178 /* with the translated packet (not the original), we have to reverse the    */
8179 /* lookup. Although doing the lookup is expensive (relatively speaking), it */
8180 /* is not anticipated that this will be a frequent occurance for normal     */
8181 /* traffic patterns.                                                        */
8182 /* ------------------------------------------------------------------------ */
8183 void
ipf_nat_uncreate(fr_info_t * fin)8184 ipf_nat_uncreate(fr_info_t *fin)
8185 {
8186 	ipf_main_softc_t *softc = fin->fin_main_soft;
8187 	ipf_nat_softc_t *softn = softc->ipf_nat_soft;
8188 	int nflags;
8189 	nat_t *nat;
8190 
8191 	switch (fin->fin_p)
8192 	{
8193 	case IPPROTO_TCP :
8194 		nflags = IPN_TCP;
8195 		break;
8196 	case IPPROTO_UDP :
8197 		nflags = IPN_UDP;
8198 		break;
8199 	default :
8200 		nflags = 0;
8201 		break;
8202 	}
8203 
8204 	WRITE_ENTER(&softc->ipf_nat);
8205 
8206 	if (fin->fin_out == 0) {
8207 		nat = ipf_nat_outlookup(fin, nflags, (u_int)fin->fin_p,
8208 					fin->fin_dst, fin->fin_src);
8209 	} else {
8210 		nat = ipf_nat_inlookup(fin, nflags, (u_int)fin->fin_p,
8211 				       fin->fin_src, fin->fin_dst);
8212 	}
8213 
8214 	if (nat != NULL) {
8215 		NBUMPSIDE(fin->fin_out, ns_uncreate[0]);
8216 		ipf_nat_delete(softc, nat, NL_DESTROY);
8217 	} else {
8218 		NBUMPSIDE(fin->fin_out, ns_uncreate[1]);
8219 	}
8220 
8221 	RWLOCK_EXIT(&softc->ipf_nat);
8222 }
8223 
8224 
8225 /* ------------------------------------------------------------------------ */
8226 /* Function:    ipf_nat_cmp_rules                                           */
8227 /* Returns:     int   - 0 == success, else rules do not match.              */
8228 /* Parameters:  n1(I) - first rule to compare                               */
8229 /*              n2(I) - first rule to compare                               */
8230 /*                                                                          */
8231 /* Compare two rules using pointers to each rule. A straight bcmp will not  */
8232 /* work as some fields (such as in_dst, in_pkts) actually do change once    */
8233 /* the rule has been loaded into the kernel. Whilst this function returns   */
8234 /* various non-zero returns, they're strictly to aid in debugging. Use of   */
8235 /* this function should simply care if the result is zero or not.           */
8236 /* ------------------------------------------------------------------------ */
8237 static int
ipf_nat_cmp_rules(ipnat_t * n1,ipnat_t * n2)8238 ipf_nat_cmp_rules(ipnat_t *n1, ipnat_t *n2)
8239 {
8240 	if (n1->in_size != n2->in_size)
8241 		return (1);
8242 
8243 	if (bcmp((char *)&n1->in_v, (char *)&n2->in_v,
8244 		 offsetof(ipnat_t, in_ndst) - offsetof(ipnat_t, in_v)) != 0)
8245 		return (2);
8246 
8247 	if (bcmp((char *)&n1->in_tuc, (char *)&n2->in_tuc,
8248 		 n1->in_size - offsetof(ipnat_t, in_tuc)) != 0)
8249 		return (3);
8250 	if (n1->in_ndst.na_atype != n2->in_ndst.na_atype)
8251 		return (5);
8252 	if (n1->in_ndst.na_function != n2->in_ndst.na_function)
8253 		return (6);
8254 	if (bcmp((char *)&n1->in_ndst.na_addr, (char *)&n2->in_ndst.na_addr,
8255 		 sizeof(n1->in_ndst.na_addr)))
8256 		return (7);
8257 	if (n1->in_nsrc.na_atype != n2->in_nsrc.na_atype)
8258 		return (8);
8259 	if (n1->in_nsrc.na_function != n2->in_nsrc.na_function)
8260 		return (9);
8261 	if (bcmp((char *)&n1->in_nsrc.na_addr, (char *)&n2->in_nsrc.na_addr,
8262 		 sizeof(n1->in_nsrc.na_addr)))
8263 		return (10);
8264 	if (n1->in_odst.na_atype != n2->in_odst.na_atype)
8265 		return (11);
8266 	if (n1->in_odst.na_function != n2->in_odst.na_function)
8267 		return (12);
8268 	if (bcmp((char *)&n1->in_odst.na_addr, (char *)&n2->in_odst.na_addr,
8269 		 sizeof(n1->in_odst.na_addr)))
8270 		return (13);
8271 	if (n1->in_osrc.na_atype != n2->in_osrc.na_atype)
8272 		return (14);
8273 	if (n1->in_osrc.na_function != n2->in_osrc.na_function)
8274 		return (15);
8275 	if (bcmp((char *)&n1->in_osrc.na_addr, (char *)&n2->in_osrc.na_addr,
8276 		 sizeof(n1->in_osrc.na_addr)))
8277 		return (16);
8278 	return (0);
8279 }
8280 
8281 
8282 /* ------------------------------------------------------------------------ */
8283 /* Function:    ipf_nat_rule_init                                           */
8284 /* Returns:     int   - 0 == success, else rules do not match.              */
8285 /* Parameters:  softc(I) - pointer to soft context main structure           */
8286 /*              softn(I) - pointer to NAT context structure                 */
8287 /*              n(I)     - first rule to compare                            */
8288 /*                                                                          */
8289 /* ------------------------------------------------------------------------ */
8290 static int
ipf_nat_rule_init(ipf_main_softc_t * softc,ipf_nat_softc_t * softn,ipnat_t * n)8291 ipf_nat_rule_init(ipf_main_softc_t *softc, ipf_nat_softc_t *softn,
8292 	ipnat_t *n)
8293 {
8294 	int error = 0;
8295 
8296 	if ((n->in_flags & IPN_SIPRANGE) != 0)
8297 		n->in_nsrcatype = FRI_RANGE;
8298 
8299 	if ((n->in_flags & IPN_DIPRANGE) != 0)
8300 		n->in_ndstatype = FRI_RANGE;
8301 
8302 	if ((n->in_flags & IPN_SPLIT) != 0)
8303 		n->in_ndstatype = FRI_SPLIT;
8304 
8305 	if ((n->in_redir & (NAT_MAP|NAT_REWRITE|NAT_DIVERTUDP)) != 0)
8306 		n->in_spnext = n->in_spmin;
8307 
8308 	if ((n->in_redir & (NAT_REWRITE|NAT_DIVERTUDP)) != 0) {
8309 		n->in_dpnext = n->in_dpmin;
8310 	} else if (n->in_redir == NAT_REDIRECT) {
8311 		n->in_dpnext = n->in_dpmin;
8312 	}
8313 
8314 	n->in_stepnext = 0;
8315 
8316 	switch (n->in_v[0])
8317 	{
8318 	case 4 :
8319 		error = ipf_nat_ruleaddrinit(softc, softn, n);
8320 		if (error != 0)
8321 			return (error);
8322 		break;
8323 #ifdef USE_INET6
8324 	case 6 :
8325 		error = ipf_nat6_ruleaddrinit(softc, softn, n);
8326 		if (error != 0)
8327 			return (error);
8328 		break;
8329 #endif
8330 	default :
8331 		break;
8332 	}
8333 
8334 	if (n->in_redir == (NAT_DIVERTUDP|NAT_MAP)) {
8335 		/*
8336 		 * Prerecord whether or not the destination of the divert
8337 		 * is local or not to the interface the packet is going
8338 		 * to be sent out.
8339 		 */
8340 		n->in_dlocal = ipf_deliverlocal(softc, n->in_v[1],
8341 						n->in_ifps[1], &n->in_ndstip6);
8342 	}
8343 
8344 	return (error);
8345 }
8346 
8347 
8348 /* ------------------------------------------------------------------------ */
8349 /* Function:    ipf_nat_rule_fini                                           */
8350 /* Returns:     int   - 0 == success, else rules do not match.              */
8351 /* Parameters:  softc(I) - pointer to soft context main structure           */
8352 /*              n(I)     - rule to work on                                  */
8353 /*                                                                          */
8354 /* This function is used to release any objects that were referenced during */
8355 /* the rule initialisation. This is useful both when free'ing the rule and  */
8356 /* when handling ioctls that need to initialise these fields but not        */
8357 /* actually use them after the ioctl processing has finished.               */
8358 /* ------------------------------------------------------------------------ */
8359 static void
ipf_nat_rule_fini(ipf_main_softc_t * softc,ipnat_t * n)8360 ipf_nat_rule_fini(ipf_main_softc_t *softc, ipnat_t *n)
8361 {
8362 	if (n->in_odst.na_atype == FRI_LOOKUP && n->in_odst.na_ptr != NULL)
8363 		ipf_lookup_deref(softc, n->in_odst.na_type, n->in_odst.na_ptr);
8364 
8365 	if (n->in_osrc.na_atype == FRI_LOOKUP && n->in_osrc.na_ptr != NULL)
8366 		ipf_lookup_deref(softc, n->in_osrc.na_type, n->in_osrc.na_ptr);
8367 
8368 	if (n->in_ndst.na_atype == FRI_LOOKUP && n->in_ndst.na_ptr != NULL)
8369 		ipf_lookup_deref(softc, n->in_ndst.na_type, n->in_ndst.na_ptr);
8370 
8371 	if (n->in_nsrc.na_atype == FRI_LOOKUP && n->in_nsrc.na_ptr != NULL)
8372 		ipf_lookup_deref(softc, n->in_nsrc.na_type, n->in_nsrc.na_ptr);
8373 
8374 	if (n->in_divmp != NULL)
8375 		FREE_MB_T(n->in_divmp);
8376 }
8377